Report on renewable energy supply in Europe addressing technological and political preconditions

Size: px
Start display at page:

Download "Report on renewable energy supply in Europe addressing technological and political preconditions"

Transcription

1 GA No Report on renewable energy supply in Europe addressing technological and political preconditions Rolf Golombek, Finn Roar Aune, Hilde Hallre The Frisch Centre (FCO) Brigitte Knopf, Paul Nahmmacher, Eva Schmid Potsdam Institute for Climate Impact Research (PIK)

2 FP7-ENV-2012 Executive Summary The prospect of renewable energy in Europe depends on climate policy as well as other technological and political preconditions. This report examines the role of renewable energy if either nuclear power is phased out in Europe or a strict renewable target is imposed in addition to greenhouse gas emissions targets. Using a numerical simulation model of the European energy industry (LIBEMOD) we explore the impact of a nuclear phase out provided the proposal of the EU Commission to reduce greenhouse gas emissions by 40 percent in 2030 relative to 1990 is implemented. We find that a complete nuclear phase out in Europe by 2030 has a moderate impact on total production of electricity and only a tiny impact on total consumption of energy. Lower nuclear production is to a large extent replaced by more renewable electricity production, in particular wind power and bio power. Second, using the European electricity market model LIMES-EU we analyze what would be a cost-effective share of renewable energy sources that could serve as a target for the year We perform a sensitivity analysis by considering a variety of scenarios. In particular, we show that a high renewable energy source share triggers more wind power and solar power, whereas the impact on bio power is negligible. We conclude that the additional cost of a renewable target in the electricity sector is moderate, but it increases non-linearly with the level of the renewable energy source target. 2

3 FP7-ENV-2012 Table of Contents Executive Summary... 2 PART 1: Phasing out nuclear power in Europe Introduction Summary of Work Performed Results and Conclusion... 5 PART 2: A European renewable electricity target for Introduction Summary of Work Performed Results and Conclusion References List of Abbreviations...22 Appendixes Phasing out nuclear power in Europe by Finn Roar Aune, Rolf Golombek and Hilde Hallre Sensitivity analysis and costs of a European renewable target for 2030 by Brigitte Knopf, Paul Nahmmacher and Eva Schmid Carpe diem: A novel approach to select representative days for long-term power system models with high shares of renewable energy sources by Paul Nahmmacher, Eva Schmid, Lion Hirth and Brigitte Knopf 3

4 FP7-ENV-2012 PART 1: Phasing out nuclear power in Europe 1.1 Introduction Until the Fukushima accident in Japan in February 2011, nuclear power was by many seen as an important part of a low-carbon future. The accident sparked security concerns and antinuclear sentiments in many European countries causing three EU member states to phase out nuclear power. In Belgium, three reactors are to be phased out by 2015 and the remaining four reactors will be shut down by In Germany, the seven oldest reactors where shut down very fast and a plan for a complete phase out of nuclear by 2022 was agreed upon. In Switzerland, the parliament agreed not to replace any of the country s nuclear reactors, which will result in a complete phase-out by For other EU countries, the response to the Fukushima accident was more mixed. For example, in France a European Pressurized Reactor (EPR) is under construction but the President has pledged to reduce the share of nuclear electricity production from 75 percent (2011) to 50 percent by In some East-European countries, there are plans to either extend the lifetime of current reactors (for example Bulgaria) or build new reactors (for example Romania), but currently plans are on hold because of lack of financing. Hence, the future of nuclear power in Europe is uncertain. We examine the outcome if all EU member states copy the long-run strategy of Belgium, Germany and Switzerland to phase out nuclear power. We focus on two questions. First, to what extent will a phase out of nuclear power be replaced by supply from other electricity technologies? Second, how will a phase out change the composition of electricity technologies? 1.2 Summary of Work Performed The short-run partial effect of a nuclear phase out is lower supply of electricity, which, cet. par., should increase the price of electricity, thereby providing incentives to invest in fossilfuel based and renewable electricity production capacity. A higher price of electricity may also lead to substitution effects between consumption of electricity and consumption of primary energy. Hence, the effect of a nuclear phase out may be smaller on total consumption of energy than on consumption of electricity. This suggests that in analyzing the impact of a nuclear phase out a model that captures the whole energy industry, not only the electricity sector, should be used. In this study we use the numerical multi-good, multi-period model LIBEMOD to analyze impacts of a nuclear phase out. This model meets the requirements specified above: it covers the entire energy industry in 30 European countries (EU-27 plus Iceland, Norway and Switzerland, henceforth referred to as EU-30). In the model, eight energy goods, that is, three types of coal, oil, natural gas, two types of bioenergy and electricity, are extracted, produced, traded and consumed in each of the 30 European countries. In each country, electricity can be produced by a number of technologies; nuclear, fuel based technologies (using either steam coal, lignite, oil, natural gas or biomass as an input), fossil-fuel based CCS (using either steam coal or natural gas), hydro (reservoir hydro, run-of-river hydro and pumped storage hydro), wind power and solar. We make a distinction between plants with pre-existing capacities in the data year of the model (2009) and new plants; the latters are built if such investments are profitable. All markets for energy goods are assumed to be competitive in While steam coal, coking coal and biofuel are traded in global markets in LIBEMOD, natural gas, 4

5 FP7-ENV-2012 electricity and biomass are traded in European markets, although there is import of these goods from non-european countries. For the latter group of energy goods, trade takes place between pair of countries, and such trade requires electricity transmission lines/gas pipelines. These networks have pre-existing capacities in the data year of the model, but through profitable investments capacities can be expanded. LIBEMOD determines all prices and quantities in the European energy industry as well as prices and quantities of energy goods traded globally. In addition, the model determines emissions of CO2 by country and sectors (households; services and the public sector; manufacturing; transport; electricity generation). As part of this project we have developed a strategy to model profitable investments in solar power and wind power in numerical energy-market models (like LIBEMOD) that takes into account that i) the production sites of these technologies differ, that is, the number of solar and wind hours differ between sites, and ii) access to sites is regulated. Both wind power and solar power will in general use surface area that has an opportunity cost; we therefore make explicit assumptions on how much land that may be available for this type of electricity production in each country. The endogenous determination of investment in solar power and wind power is based on a combination of technical factors the degree to which production sites differ political factors the degree to which agents get access to production sites and economic factors the profitability of investment given access to a set of production sites. We have also gathered and processed information to provide an overview of costs of producing electricity by different technologies both total cost of electricity and specific cost elements. These cost elements have consistent assumptions about factors like duration of a new plant, rate of interest, operational hours throughout the year, and fossil fuel prices. We also compare our cost assumptions to other studies. Finally, we use the numerical model LIBEMOD to quantify the effects of a nuclear phase out in EU-30 and test the sensitivity of our main results by varying factors like i) the GHG emissions target, ii) the carbon policy instruments imposed by the EU, and iii) cost of electricity production, for example, cost of investment in CCS power stations. 1.3 Results and Conclusion To examine the effects of a nuclear phase out we consider a number of scenarios for 2030, see Table 1. In our reference scenario we assume that the nuclear capacities in 2030 reflect decisions taken in 2014 or earlier at the country level with respect to whether nuclear plants will be phased out or new nuclear capacity will come online before 2030, see Table 2. Based on information from IEA Electricity Information Statistics database and Eurelectric (2011) there may be a net decrease in nuclear capacity in EU-30 between 2009 and 2030 of about 23.2 GW, see Table 8, which amounts to roughly 20 percent of the 2009 nuclear capacity in EU-30. Hence, in the reference scenario total nuclear capacity in EU-30 is 23.2 GW lower than in the data year

6 FP7-ENV-2012 Table 1 Scenarios for 2030 Reference scenario 50 percent phase out 100 percent phase out No policy Efficient High emissions Cheap CCS EU renewable target National renewable policy Balancing power Nuclear capacities reflect decisions after percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. Nuclear capacities reduced by 50 percent in 2030 relative to percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. Complete nuclear phase out by percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. Complete nuclear phase out by No environmental target. Complete nuclear phase out by percent GHG reduction in 2030 relative to One common emission target for ETS and non-ets sectors. Complete nuclear phase out by percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. Complete nuclear phase out by percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. Costs of CCS investment reduced by 50 percent relative to reference scenario Complete nuclear phase out by percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. One common EU target for share of renewable energy of 40 percent. Complete nuclear phase out by percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. Subsidies to renewable energy in selected countries. Complete nuclear phase out by percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. Increased requirement of balancing power. 6

7 FP7-ENV-2012 Table 2 Nuclear policy in EU member states COUNTRY POLICY PLANNED CAPACITY CHANGE Belgium Complete phase-out by MWe phase-out by MWe phase-out by 2025 Bulgaria Plans to extend lifetime of current reactors. Plans for a new reactor on hold due to lack of financing. Czech Rep National energy plan to 2060 assumes 50% nuclear capacity, however plans for two reactors are put on hold after the government refused to provide state support. Finland One EPR reactor under construction, expected to be in commercial operation by Another two reactors planned. France One EPR reactor under construction. The current President has pledged to reduce the share of electricity from nuclear to 50% by Germany Closed down 8 reactors in March Plans for complete phase-out by Hungary Plans for two new reactors under government ownership. Italy Plans to revive the national nuclear industry rejected by referendum in Lithuania Closed down two reactors in 2009 due to EU safety concerns. Plans for one new reactor, expected to start operating in Netherlands Previous decision on phase-out was reversed in However, plans for new reactors are on hold due to economic uncertainties. Poland Cabinet decision to move to nuclear power in Currently two planned reactors. Romania Two new reactors planned, but currently lacking financing. Slovakia Plans for new reactors outlined in the 2008 Energy Security Strategy, aiming to keep the share of electricity from nuclear at 50%. Slovenia Spain Sweden Switzerland United Kingdom Considering capacity expansion, but no plans confirmed. Political uncertainty surrounding nuclear future. No plans for new reactors, but in 2011 the legal limitation to plant operating lives was removed (previously 40 years). Phase-out plan from 1980 repealed in June Currently plans to uprate/replace old units when decommissioned. Parliament decision in June 2011 to not replace any reactors. Complete phase-out by Plans for several new reactors between 2023 and Government goal is 16 GWe new capacity by MWe in MWe in MWe in MWe around MWe in MWE in MWe shut down in MWe phase-out by MWe in MWe in after MWe in MWe in MWe in MWe in MWe in MWe in by MWe in by MWe phase-out by 2022 (net) 985 MWe phase-out by 2030 (net) 1165 MWe phase-out by 2034 (net) MWe by

8 FP7-ENV-2012 In the reference scenario we assume that the climate policy of EU in 2030 will follow the proposal of the EU Commission from January 2014 to reduced GHG emissions by 40 percent relative to This proposal distinguishes between the ETS sector (electricity generation and large carbon-intensive manufacturing firms) and the remaining sectors. Whereas the ETS sector has to reduce its GHG emissions by 43 percent relative to 2005, the corresponding number for the non-ets sector is 30 percent. In addition, the renewable share of consumption should be 27 percent; the EU Commission indicates that the latter target may be reached if the emission targets are reached. All targets are at the EU level and hence not broken down to national targets. In the reference scenario we follow the proposal of the EU Commission and hence have one common EU-30 target for emissions in the ETS sector implemented by a common quota system and one common EU-30 target for emissions in the non-ets sector implemented by a common uniform tax. Because LIBEMOD only covers CO2, the most important GHG gas, we transform the GHG emissions targets to CO2 targets. The transformation to CO2 targets also reflect that LIBEMOD cannot distinguish between manufacturing firms that belong to the ETS sector (large carbon-intensive units) and those firms not covered by the ETS sector. In the next two scenarios we reduce the capacities of nuclear power in all model countries that did not completely phase out nuclear power in the reference scenario by either 50 percent relative to 2009 ( 50 % Phase out ) or by 100 percent ( 100 % Phase out ). The climate goals and climate instruments are, however, the same as those in the reference scenario. For the remaining scenarios we stick to the assumption that there has been a complete nuclear phase out. Figure 1 shows the capacity of electricity technologies in EU-30 in 2009, in the reference scenario, under a 50 percent nuclear phase out and under a complete phase out by As indicated by the figure, total capacity increases sharply from the observed 2009 value (917 GW) to the reference scenario for 2030 (1494 GW). The increase, which is roughly 60 percent, is mainly due to economic growth between 2009 and 2030, but it also reflects the climate policy in the reference scenario (40 percent emissions reduction): there is a sharp increase in wind power capacity from 74 GW in 2009 to 482 GW in the reference scenario. In 2009 the capacity share of wind power was 8 percent, whereas the share in the reference scenario is as high as 32 percent. In 2009 the share of nuclear capacity was 14 percent, whereas in the reference scenario this share is down to 7 percent. The change reflects partly the assumption that there will be 23.2 GW less nuclear capacity in the reference scenario than in 2009, and partly the sharp increase in total electricity capacity. 8

9 FP7-ENV-2012 Figure 1 Net capacity by technology in EU 30 in 2009 and 2030 (GW) As seen from Figure 1, total capacity increases slightly (by 60 GW) from the reference scenario to the case of a 50 percent phase out, and it increases further by 59 GW if nuclear is completely phased out. The corresponding increases in wind power capacity are in fact slightly higher. If nuclear is phased out, the capacity share of wind power is 40 percent whereas the capacity share of renewable electricity is 72 percent (34 percent in 2009). The small increase in total capacity from the reference scenario to a complete nuclear phase out suggests that total production of electricity increases, but the higher share of renewables, in particular for wind power, suggests the opposite because the capacity rate of wind power is by far lower than for nuclear, and also lower than for most fossil fuel plants. Figure 2 shows production of electricity by technology. As seen from the figure, total production of electricity is roughly at the same level in the reference scenario, under a 50 percent nuclear phase out and also under a complete phase out (This common production level is almost 30 percent higher than in 2009). In these three scenarios, the demand functions for energy, which are derived from nested CES utility functions, are identical. Because the endogenous prices of energy differ slightly among the scenarios the producer price of electricity is, for example, two percent higher in the 100 percent scenario than in the 50 percent scenario, which again is two percent higher than in the reference scenario - the equilibrium level of consumption of electricity differs only moderately among the three scenarios. This implies that the short-run marginal cost curve of electricity does not differ much among the scenarios, which again reflects that as nuclear capacity is phased out, it is profitable to invest in more renewable capacity, in particular wind power. In fact, total capacity increases as more nuclear is phased out, see Figure 1. The low short-run marginal cost of wind power tends to decrease the price of electricity (in periods with much wind), but on the other hand the low operating rate of wind power relative to nuclear tends to increase the equilibrium electricity prices (in periods with 9

10 FP7-ENV-2012 moderate wind). Our results suggest that as an annual average, the equilibrium price of electricity does not shift much. We stress that the high investment in renewable energy is per construction consistent with profit-maximizing behavior. Whereas total production of electricity does not differ much among the three scenarios in Figurer 2, the composition of technologies changes, however, radically. Bio power production, which has a high rate of capacity utilization, is almost twice as high under a complete phase out than in the reference scenario, whereas the corresponding increase for wind power is almost 50 percent. The share of renewables in electricity production is 73 percent after a complete nuclear phase out, that is, almost identical to the capacity share of renewables. 10

11 FP7-ENV-2012 Figure 2 Electricity production in EU 30 in 2009 and 2030 (TWh) Figures 3-5 show electricity capacity, electricity production and consumption of energy by scenario. In each figure we first repeat the case of a full nuclear phase when GHG emissions are imposed to be 40 percent lower than in 1990 and there are separate ETS and non-ets targets (henceforth referred to as 100 percent ). Then the results of the remaining scenarios from Table 1 are shown. Without any climate policy ( no policy ) total capacity is somewhat more than 10 percent lower than in the 100 percent scenario. However, total production of electricity is almost identical in the two cases, reflecting that under no policy there is massive production of coal power its market share is roughly two third and this technology has a high rate of capacity utilization. The high level of coal power production tends to decrease the price of electricity and therefore production of gas power is substantially lower than in the 100 percent scenario. The market share of renewable is almost 30 percent, that is, somewhat higher than in 2009 (24 percent). This exercise suggests that the impact of a nuclear phase out in the absence of a climate policy is mainly that nuclear production is replaced by fossil fuel based production. In the case of one climate target and therefore one common price of emissions of CO2 ( efficient ), total electricity capacity is almost 10 percent above the capacity in the 100 percent scenario, but total production of electricity does not differ much between the two cases. In order to reach the climate target a uniform CO2 tax at 69 euro/tco2 has to be imposed. This is almost twice as high as the price of emissions in the ETS sector in the 100 percent scenario (38 euro/tco2), and therefore fossil fuel based technologies are punished hard in the efficient scenario. In fact, there is almost no conventional production of gas power and coal power in the efficient scenario, but CCS gas obtains a market share of roughly 5 percent. Thus, the market share of renewable electricity exceeds 90 percent. If total emissions are to be 20 percent lower than in 1990 (not 40 percent as in the 100 percent nuclear phase out scenario), total production of electricity is marginally lower than in the 100 percent scenario. As expected, the technology mix differs significantly 11

12 FP7-ENV-2012 between the two cases. In the case of high emissions, there is substantially coal power production, which crowds out some of the gas fired power production. The resulting market share of renewable is 37 percent, which is significantly lower than in the 100 percent scenario (73 percent). In the 100 percent phase-out scenario each national system operator is to make sure that (at least) 5 percent of the maintained capacity is available for reserve power production in case demand for electricity suddenly increases or supply suddenly drops. If this requirement is increased to 20 percent due to the increased market share of intermittent renewable electricity maintained capacity is increased by almost 200 GW whereas installed capacity is only slightly affected. Hence, a higher share of the installed capacity of preexisting plants is being maintained; it is much cheaper to increase maintained capacity by maintaining plants with low efficiency than to buy new power plants. Moreover, because the increase in maintained capacity is of the same magnitude as the increase in capacity acquired by the national operators, the available capacity for production is almost similar in the two cases. Hence, also production of electricity is similar in these two cases, see Figure 4. In the 100 percent phase-out scenario there is only marginal CCS production. If the government subsidizes 50 percent of all CCS investment costs ( cheap CCS ), the effect on total production of electricity is minor, but the market share of CCS becomes 11 percent. The resulting market share of renewable is around two third. Figure 6 shows the share of renewables in energy consumption across scenarios. In the reference scenario this share is 25.9 percent, that is, slightly below the EU target of 27 percent. Hence, our study suggests that if the climate targets are reached, the renewable target is almost reached. With a complete phase out, the share is 32.9 percent and hence the renewable target is by far reached. To explore the partial effect of tightening the renewable share in energy consumption we have imposed a renewable target of 35 percent when nuclear is fully phased out and GHG emissions are to be 40 percent below the 1990 level. The renewable target is implemented by offering a subsidy to producers of renewable electricity and to end-users of bioenergy (same subsidy per unit of energy). With a renewable subsidy the price of emissions of CO2 in the ETS sector decreases: the CO2 price in the ETS sector is now 29 euro/tco2, which is 9 euro/tco2 lower than in the 100 percent scenario. As seen from Figure 11, a higher renewable share in energy consumption increases total production of electricity slightly (by 2 percent). The main effects are that gas power decreases (by around 160 TWh) whereas wind power increases by almost 200 TWh. To sum up, if nuclear power is fully phased out total production of electricity, and to a large extent also total consumption of energy, do not differ much between the scenarios. Moreover, these levels do not differ much from the reference scenario. However, the mix of electricity technologies differs significantly between scenarios. The equilibrium composition of electricity technologies reflects the stringency of the climate target and whether some technologies are being promoted, either directly through subsidies or indirectly through a tailor-made policy goal. 12

13 FP7-ENV-2012 Figure 3 Robustness: Net capacity by technology in EU 30 in 2030 by scenario (GW) 13

14 FP7-ENV-2012 Figure 4 Robustness: Electricity production in EU 30 in 2030 by scenario (TWh) Figure 5 Robustness: Energy consumption in EU 30 in 2030 (Mtoe) 14

15 FP7-ENV-2012 Figure 6 Renewable share in final energy demand in EU 30 by scenario 15

16 FP7-ENV-2012 PART 2: A European renewable electricity target for Introduction In Europe, the share of renewable energy sources (RES) has been increasing steadily over the past decade: It rose from just above 8 % in 2004 to 14 % in Especially in the electricity sector the share of RES increased significantly, reaching 24 % in This development is mainly due to the dedicated support of RES deployment by means of feed-in tariffs or other policy instruments designed to fulfil the EU-wide endeavor of increasing the share of renewables. In 2009, the European Union (EU) adopted the EU climate and energy package (the so-called package) that includes (i) a 20% reduction in EU greenhouse-gas (GHG) emissions from 1990 levels, (ii) raising the share of renewables in the EU s final energy consumption to 20% (including a renewable share of 10% in the transport sector), and (iii) a 20% improvement in the EU's energy efficiency. With separate targets for renewables and for energy efficiency, there is a clear overlap between different policies to address GHG reduction. These additional targets, especially for renewables, are often justified in the political debate by referring to potential co-benefits that renewables create, such as employment effects, local value added, additional environmental benefits and industrial policy. In January 2014 the European Commission suggested an EU-wide renewable target of 27% in final energy consumption. This number is derived by the EU Commission s own modeling analysis after imposing a 40% GHG reduction target (with a specific distribution between the ETS sector and the non-ets sector). In the following we refer to this benchmark as the cost-effective RES target it is the share of RES following from achieving a given climate target at least costs over time. If the RES share is set above this cost-effective share, the underlying rationale is not only climate change mitigation but also other considerations. So far, a systematic comparative quantification of the potential co-benefits of renewable energy deployment and the costs of an additional renewable target is lacking. This study aims to contribute to the quantification of the latter aspect in a systematic manner. 2.2 Summary of Work Performed The key questions addressed in this study are: 1. What is the cost-effective RES share in the European electricity sector for the year 2030 that is consistent with a 40% GHG reduction target in 2030, given varying key input assumptions concerning technology availability and institutional settings? What are the decisive drivers of the cost-effective RES share in the context of climate mitigation? 2. What are the additional economic costs and system effects if the RES target is set higher than this cost-effective share? To answer these questions we have performed a sensitivity analysis with a refined version of the European electricity sector model LIMES-EU. This is an optimization model that determines investment in electricity production capacity, investment in international transmission capacity and production of electricity by technology, whereas demand for electricity is exogenous and mainly builds on European Commission (2014). In addition to an 16

17 FP7-ENV-2012 update of data, techno-economic assumptions and the calibration strategy, we have improved LIMES-EU in several respects, especially regarding the representation of variable RES (vres) to better capture the key effects of an increasingly high share of intermittent power. To this end we have developed a novel approach for deriving time-slices for longterm power system models. Our approach is mainly based on Ward s hierarchical clustering algorithm. We apply this algorithm on historic electricity demand and weather data to group days with similar demand and vres load patterns. After convergence of the developed algorithm, each group of days is reflected by a representative day in the power system model. The method is easily reproducible and applicable to all sorts of input data with multiple fluctuating time-series. 2.3 Results and Conclusion In order to be comparable to the European Commission s Impact Assessment, in our default scenario, which is referred to COM policy, there is a GHG reduction of 40% in 2030 and a GHG reduction of 80% by 2050 (relative to 1990). Because we implement these boundary conditions under the assumption that annual consumption of electricity is exogenous in each country, a change in the RES share will not have impact on total production of electricity. To some extent this can be justified by the LIBEMOD study of a nuclear phase out as such a policy has minor impact of total production of electricity in the long run. The GHG targets for the entire EU economy have been transformed to a CO2 target for the electricity sector. For the default scenario this leads to a CO 2 reduction target of electricity of 95% by 2050 in comparison to Intermediate targets between 2010 and 2050 are increasing linearly and result in a CO 2 reduction in the electricity sector of 52% by In addition to the CO2 constraint there are region specific constraints for the deployment of certain generation technologies. In fact, hard coal and natural gas are the only primary energy sources for which there is no such regional consumption constraint. We assume their global supply is large enough to satisfy the demand of the European electricity sector at all points in time. In order to reflect the political situation of nuclear power in Europe, we implement a nuclear phase-out in Germany, Belgium and Switzerland. New installations in other countries are limited to those currently under construction or planned and those expected to replace depreciated capacities. In order to address the key research questions we analyze a number of scenarios that differ from COM policy with respect to technology development or institutional settings, see Table 3. 17

18 RES target Institutional setting Technology availability FP7-ENV-2012 Table 3 Overview of scenarios Technology / Setting Issue Scenario Name As in GHG40 of the Impact Assessment for the 2030 COM policy framework by the European Commission (2014) Nuclear Power Phase-out until 2030 nucout 2030 Phase-out until 2050 nucout 2050 (-50% in 2030) CCS Not available no ccs vres (wind, solar) Higher investment costs high cost vres (+10% in 2020; +20% in 2030 and thereafter) Lower investment costs low cost vres (-10% in 2020; -20% in 2030 and thereafter) Only sites with average low pot vres capacity factors available Biomass Higher fuel costs (+100%) high cost bio Storage Higher investment costs high cost stor (+100%) Transmission Capacity expansion process is gridlocked Transmission Capacity expansion process gains momentum / Completion of the Internal Market Energy security strictly domestic Energy efficiency programs are successful Reference scenario of the European Commission (2013) No Capacity expansion beyond today, completion of internal market impeded Faster capacity expansion (and market integration) possible (expansion rate +100%) More than 95% of electricity supply from domestic power plants Electricity demand lower by 5% in 2030 and thereafter Electricity demand lower by 10% in 2030 and thereafter 44% GHG reduction target in the year 2050 no trans exp high trans exp 95% national demand -5% demand -10% COM reference minimum 50% RES share in 2030 and thereafter RES target 50% minimum 55% RES share in 2030 and thereafter RES target 55% minimum 60% RES share in 2030 and thereafter RES target 60% minimum 65% RES share in 2030 and thereafter RES target 65% minimum 70% RES share in 2030 and thereafter RES target 70% 18

19 FP7-ENV-2012 With LIMES-EU we find that the COM policy scenario results in a RES-share of electricity of 50% in This is very close to 49 %, which was attained by the Impact Assessment of the European Commission. However, Figure 7 shows that the different scenarios considered in our sensitivity analysis lead to a range of cost-effective RES shares between 43% and 56%. Interestingly, the 49%-target proposed by the Impact Assessment is in the middle of this range. Figure 7 Cost-effective share of renewables in the European electricity sector in per cent of total electricity provision and the respective technology mix of renewables in the year Red shading indicates the range across scenarios. Source: Model results of LIMES-EU. Black line: Share of RES in the electricity sector (49%) in the scenario GHG40 of the Impact Assessment for the 2030 framework by the European Commission (2014). The model results obtained with LIMES-EU indicate that the cost-effective share of RES in 2030 is higher than 49% if a nuclear phase-out is pursued all over Europe, investment costs for wind and solar technologies decrease faster than expected in the default scenario or the CCS technology is not available. It is lower than 49%, however, if electricity demand decreases due to progress in energy efficiency programs, investment costs for wind and solar technologies decrease slower than expected in the default scenario, only RES sites with average instead of high potential are available or biomass costs are higher than in the default scenario. The technology mix that emerges for the year 2030 is rather similar across all scenarios. The share of biomass is lower if fuel costs are high (high cost bio) or the investment costs for wind and solar technologies develop very favorably (low cost vres). Wind onshore plays a very important role in all scenarios. However, if high quality wind sites cannot be used (low pot vres), or investment costs develop on the pessimistic side of the projected range (high cost vres), the role of onshore wind is somewhat smaller. In those scenarios where the share of biomass is lower, the difference is made up for by a higher share of wind. Solar PV is used in all scenarios to a certain extent. However, in the scenarios with the highest total RES share (nucout 2030, low cost vres, nucout 2050), the share of solar PV is also highest. Concentrated solar power (CSP) does not play an important role in any of the scenarios. Comparatively, this technology has higher costs of electricity production 19

20 FP7-ENV-2012 in the year 2030 and is hence not deployed. Only in the scenario with optimistic investment cost developments (low cost vres), CSP is deployed to an extent that is at least visible in the graph. Only wind offshore plays an even less important role than CSP. Figure 8 RES shares in the electricity sector over time for the scenarios with additional RES target for 2030 (left) and technology mix of renewables in these scenarios in the year 2030 (right). Source: Model results of LIMES-EU. In order to study the economic costs of a RES-share that is set higher than the cost-effective one, we consider default scenarios with imposed RES-shares up to 70% in Figure 8 (left) shows the resulting trajectory of RES-shares over time. Until 2025 they follow the same path. Between 2025 and 2030 as much investments in RES capacities as needed to fulfill the target in 2050 are undertaken. In all scenarios the RES-share stagnates until it again meets the cost-effective trajectory for achieving the 95% mitigation target in The right panel of figure 8 illustrates that these additional investments into RES capacities between 2025 and 2030 consist primarily of solar photovoltaic and onshore wind. Also CSP comes into play. Wind offshore is again not deployed in these scenarios, due to its prohibitively high investment costs. The amount of hydro and biomass is unaffected by the imposed RESshare in What are the economic costs of setting a RES target above the cost-effective one, that is, in excess of 50%? Figure 9 provides answers to this question by displaying changes in cost (y-axes) depending on the RES target that is set in 2030 (x-axes). The graph shows the additional discounted cost over the period (in bn ) relative to the default scenario. These costs increase non-linearly with the RES-target and reach 59 bn for the 70% target. 20

21 FP7-ENV-2012 Figure 9 Present value of additional system costs over the period for scenarios with different exogenously set RES targets from 2030 and onwards (in bn ). 3. References IEA (2013). OECD - Net electrical capacity. IEA Electricity Information Statistics database. doi: /data en EURELECTRIC. (2011). Power Statistics and Trends Brussels: Union of the Electricity Industry - EURELECTRIC. European Commission (2014). Impact Assessment accompanying the document Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions A policy framework for climate and energy in the period from 2020 up to 2030 (No. SWD/2014/015 final). Brussels. 21

22 FP7-ENV List of Abbreviations CCS: carbon capture and storage CSP: concentrated solar power ETS: emission trading system EU-30: The EU-27 member states plus Iceland, Norway and Switzerland. GHG: greenhouse gases RES: renewable energy sources vres: variable renewable energy sources 22

23 DRAFT 26 August 2014 Phasing out nuclear power in Europe 1 Finn Roar Aune, Rolf Golombek and Hilde Hallre Abstract Following the Fukushima accident in 2011, some EU member states decided to phase out nuclear power. We explore the impact of an EU-wide nuclear phase out provided the proposal of the EU Commission to reduce GHG emissions by 40 percent in 2030 relative to 1990 is implemented. Using a numerical simulation model of the European energy industry (LIBEMOD), we find that a complete nuclear phase out in Europe by 2030 has a moderate impact on total production of electricity and only a tiny impact on total consumption of energy. Lower nuclear production is to a large extent replaced by more renewable electricity production, in particular wind power and bio power. More generally, the equilibrium composition of electricity technologies reflects the stringency of the climate target, which climate instruments that are imposed and whether some technologies are being promoted, either directly through subsidies or indirectly through a tailor-made policy goal. JEL classification: Q28; Q41; Q42 ; Q48 ; Q54 Key words: nuclear power, renewable electricity, CCS, carbon policy, energy modeling 1 All authors are associated with CREE - the Oslo Centre for Research on Environmentally friendly Energy, which is supported by the Research Council of Norway. Earlier versions of this paper have been presented at the CREE work-shop for policy makers and at DIW Berlin we thank the participants for their comments. This research was made possible through financial support from the Research Council of Norway to the CREE centre as well as through financial support from the European Commission under the 7 th Framework Programme of the European Union to the project Economic iinstruments to Achieve Climate Treaties in Europe (ENRACTE), project number

24 1 Introduction Until the Fukushima accident in Japan in February 2011, nuclear power was by many seen as an important part of a low-carbon future. The accident sparked security concerns and antinuclear sentiments in many European countries causing three EU member states to phase out nuclear power. In Belgium, three reactors are to be phased out by 2015 and the remaining four reactors will be shut down by In Germany, the seven oldest reactors where shut down very fast and a plan for a complete phase out of nuclear by 2022 was agreed upon. In Switzerland, the parliament agreed not to replace any of the country s nuclear reactors, which will result in a complete phase-out by For other EU countries, the response to the Fukushima accident was more mixed. For example, in France a European Pressurized Reactor (EPR) is under construction but the President has pledged to reduce the share of nuclear electricity production from 75 percent (2011) to 50 percent by In some East-European countries, there are plans to either extend the lifetime of current reactors (for example Bulgaria) or build new reactors (for example Romania), but currently plans are on hold because of lack of financing. Hence, the future of nuclear power in Europe is uncertain. In this paper we examine the outcome if all EU member states follow the long-run strategy of Belgium, Germany and Switzerland to phase out nuclear power. We focus on two questions. First, to what extent will a phase out of nuclear power be replaced by supply from other electricity technologies? Second, how will a phase out change the composition of electricity technologies? The short-run partial effect of a nuclear phase out is lower supply of electricity, which, cet. par., should increase the price of electricity, thereby providing incentives to invest in fossil-fuel based and renewable electricity production capacity. A higher price of electricity may also lead to substitution effects between consumption of electricity and consumption of primary energy. Hence, the effect of a nuclear phase out may be smaller on total consumption of energy than on consumption of electricity. This suggests that in analyzing the impact of a nuclear phase out a model that captures the whole energy industry, not only the electricity sector, should be used. Of course, the impact of a nuclear phase out depends on a number of factors. First, what is the guiding principle of investment in the electricity industry? One corner case is a centralized economy where the government solely decides investment in order to achieve some political goals, for example, a warranted level of electricity production. This is hardly a 2

25 suitable description of the current energy industry in Europe. Rather, EU bodies and European governments impose energy and environmental goals and policy instruments and leave most investment and production decisions to the private sector this is the approach taken in the present study. In particular, we will assume that profit maximization is the guiding principle of investment in the European energy industry. Second, the time horizon of a nuclear phase out is important. While a nuclear plant may be shut down immediately, it takes time to build up new generation capacity: To set up and run a new electricity plant requires detailed planning, concessions, construction, and adjustment of facilities and technologies, which may easily take 10 years. Thus in this study we examine a nuclear phase out for 2030 and hence short-run bottlenecks are not an issue. Third, the impact of a nuclear phase out will depend on costs of electricity, in particular costs of new power plants. In general, costs can be decomposed into three elements: cost of investment; cost of daily operation, which, for thermal power, reflects the cost of purchasing the amount of a fuel necessary to produces 1 kwh with the efficiency of the installed technology; and other costs, for example, ramp-up costs, costs of maintenance and fixed costs. These costs components differ between technologies and will change over time. For most renewable electricity technologies, for example, solar power and wind power, there are negligible costs of daily operation. For fuel based electricity technologies, including bio power, this cost component is, however, substantial, in particular for gas power. Over time, costs of investment of renewable electricity technologies like solar and wind power may continue to fall, see, for example, European Commission (2013) and Schröder(2013), and thus in the future these technologies may increase their market shares radically. Nuclear power has low cost of operation but excessive start-up and ramp-up cost, and is therefore used as the base load technology. If nuclear is phased out, the short-run marginal cost curve of electricity shifts upwards. Similarly, the short-run marginal cost curve of electricity shifts downwards if solar and wind power is phased in to replace nuclear. If, hypothetically, the nuclear phase out is replaced by solar and wind power with an annual production capacity equal to that of nuclear, the new annual short-run marginal cost curve will be below the annual short-run marginal cost curve prior to the nuclear phase out; this is because of negligible marginal costs of solar and wind power. This suggests that the average annual price of electricity will fall. However, thinking in terms of an annual marginal cost curve may easily lead to false conclusions: due to the intermittency of solar and wind power, in periods with no sun and wind the price of electricity will be high, whereas in periods with 3

26 lots of sun or wind and moderate demand the price will be low. Hence, price volatility will increase and the impact on the average annual price is not obvious. Finally, we will assume that EU bodies will be successful in establishing efficient internal energy markets and also that there is a common EU climate policy. Hence, we will examine the case of a nuclear phase out under the assumption of competitive markets, profitmaximizing energy producers and common EU climate targets and instruments. Needless to say, these key assumptions should be kept in mind when interpreting our results, and hence in Section 5 we will discuss how our main results are sensitive to these assumptions. Our assumption of EU-wide climate instruments reflects that the era of national tailormade subsidies to new renewable generators may have come to an end: in some European countries with non-negligible solar and wind capacity, for example, Spain, policy instruments to spur investment in renewables are now being removed, partly because the competitive position of solar and wind power has improved radically over the last 10 years, and partly because the large transfers to the private sector are regarded as a financial problem. In addition, the EU Commission has recently adopted new rules on public support for projects aiming at environmental protection. The guidelines promote a gradual move to an EU-wide market-based support for renewable energy by replacing feed-in tariffs by feed-in premiums, which are supposed to expose renewable energy to market signals through bidding processes for allocation of public support, see European Commission (2014). The discussion above suggests that an adequate analysis of a nuclear phase out should incorporate a detailed modelling of different electricity technologies to determine how the market price of electricity will vary over time periods. Because equilibrium prices determine the profitability of investment, and hence future supply of electricity, also the determination of investment should be an integral part of the model to ensure consistency. While a theoretical study for sure will determine the sign of several effects, for example, the impact on total production of renewable electricity, of course the magnitude of the effects cannot be determined. Moreover, even the sign of some effects are truly ambiguous. For example, a higher price of CO2 emissions, which may reflect a stricter emissions target, will weaken the position of fossil-fuel based electricity relative to renewables, but it will also strengthen the position of gas-fired plants relative to coal- and oilfired plants. Hence, the net effect on natural gas-fired plants is ambiguous. In addition, while the short-run partial effect of a nuclear phase out is a higher price of electricity, which should improve the position of gas power, the introduction of new technologies, for example, gasfired plants with integrated Carbon Capture and Storage (CCS) facilities, may weaken the 4

27 position of conventional gas-fired plants. Again, the sign of the gross effects are clear, but the net effect on conventional gas-fired plants is ambiguous - to identify the net effect a numerical model is required. In this study we use the numerical multi-good multi-period model LIBEMOD to analyze impacts of a nuclear phase out. This model meets the requirements specified above: it covers the entire energy industry in 30 European countries (EU-27 plus Iceland, Norway and Switzerland, henceforth referred to as EU-30). In the model, eight energy goods, that is, three types of coal, oil, natural gas, two types of bioenergy and electricity, are extracted, produced, traded and consumed in each of the 30 European countries. In each country, electricity can be produced by a number of technologies; nuclear, fuel based technologies (using either steam coal, lignite, oil, natural gas or biomass as an input), fossil-fuel based CCS (using either steam coal or natural gas), hydro (reservoir hydro, run-of-river hydro and pumped storage hydro), wind power and solar. We make a distinction between plants with pre-existing capacities in the data year of the model (2009) and new plants; the latter are built if such investments are profitable. All markets for energy goods are assumed to be competitive in While steam coal, coking coal and biofuel are traded in global markets in LIBEMOD, natural gas, electricity and biomass are traded in European markets, although there is import of these goods from non- European countries. For the latter group of energy goods, trade takes place between pairs of countries, and such trade requires electricity transmission lines/gas pipelines. These networks have pre-existing capacities in the data year of the model, but through profitable investments capacities can be expanded. LIBEMOD determines all prices and quantities in the European energy industry as well as prices and quantities of energy goods traded globally. In addition, the model determines emissions of CO2 by country and sectors (households; services and the public sector; manufacturing; transport; electricity generation). In Section 2 we provide a description of LIBEMOD, focusing mainly on supply of electricity. This section builds on an earlier version of the model, see Aune et al. (2008). In the new version of the model more countries have been added (13 East-European countries); the end-user sectors have been refined (the service and public sector has been separated from the household segment); the modeling of wind power has been changed and more renewable technologies have been included (run-of-river hydro and solar power); the modeling of natural gas has been refined; bioenergy has been split into biomass and biofuel; all data have been 5

28 updated (the data year has been changed from 2000 to 2009) and the complete model has been recalibrated. In LIBEMOD all electricity producers maximize profits with respect to, for example, investment in production capacity and production of electricity in each time period, subject to a number of technology-specific constraints. In particular, LIBEMOD offers a strategy to model profitable investments in solar power and wind power taking into account that i) the production sites of these technologies differ, that is, the number of solar and wind hours differ between sites, and ii) access to sites is regulated. Both wind power and solar power will in general use surface area that has an opportunity cost; we therefore make explicit assumptions on how much land that may be available for this type of electricity production in each country. The endogenous determination of investment in solar power and wind power is based on a combination of technical factors the degree to which production sites differ political factors the degree to which agents get access to production sites and economic factors the profitability of investment given access to a set of production sites. To the best of our knowledge, LIBEMOD is the first energy market model with endogenous prices and truly endogenous investment in renewable electricity. 2 In addition, we make two other contributions to the literature. First, we present an overview of costs of producing electricity by comparing total cost of electricity, as well as different cost elements, between different electricity technologies. These cost elements have consistent assumptions about factors like duration of a new plant, rate of interest, operational hours throughout the year, and fossil fuel prices. We also compare our cost assumptions to other studies, see Section 3. Second, in Section 4 we use the numerical model LIBEMOD to quantify the effects of a nuclear phase out in EU-30 and test (in Section 5) the sensitivity of our main results by varying factors like i) the GHG emissions target, ii) the carbon policy instruments imposed by the EU, iii) cost of electricity production, for example, cost of investment in CCS power stations, and iv) the market structure. To the best of our knowledge, the impact of an EU-wide 2 There is a number of energy models covering different parts of Europe. Most of these models are pure electricity models, see, for example, the ATLANTIS model (Gutschi et al., 2009) and the LIMES model (Haller et al., 2012), whereas LIBEMOD covers fossil fuels and as well as bio energy. Typically, pure electricity models have exogenous demand for electricity, whereas LIBEMOD endogenizes consumption of energy. Some of the pure electricity models offer very detailed description of production of electricity as well as the electricity infrastructure, see, for example, ATLANTIS, but less attention on investment. Others have a higher level of aggregation and minimize total costs (optimizing models), see, for example, LIMES. In most of the models, supply of renewable electricity contains a substantial fraction of exogenous elements, but improvements are expected. 6

29 nuclear phase out has not been examined earlier. 3 We find that a complete nuclear phase out in EU-30 by 2030 has a moderate impact on total production of electricity and only a tiny impact on total consumption of energy. If the proposal of the EU Commission to reduce GHG emissions by 40 percent in 2030 relative to 1990 is implemented, a nuclear phase out is to a large extent replaced by more renewable electricity production, in particular wind power and bio power. More generally, the equilibrium composition of electricity technologies reflects the stringency of the climate target and whether some technologies are being promoted, either directly through subsidies or indirectly through a tailor-made policy goal. 2 Libemod In this section we describe the numerical multi-market multi-good equilibrium model LIBEMOD. This model allows for a detailed study of the energy markets in Europe, taking into account factors like fossil fuel extraction, inter-fuel competition, technological differences in electricity supply, key characteristics of renewable electricity technologies, transport of energy through gas pipes/electricity lines and investment in the energy industry. The model determines all energy prices and all energy quantities invested, extracted, produced, traded and consumed in each sector in each of 30 European countries; EU-27 plus Iceland, Norway and Switzerland henceforth referred to as EU-30. The model also determines all energy prices and quantities traded in world markets, as well as emissions of CO 2 by country and sector, see Figure 1. Figure 1 The LIBEMOD model 2.1 General description The core of LIBEMOD is a set of competitive markets for eight energy goods: natural gas, oil, steam coal, coking coal, lignite, biomass, biofuel and electricity. All energy goods are extracted, produced and consumed in each country in EU-30. Natural gas, biomass and electricity are traded in competitive European markets. Trade in natural gas requires gas pipes 3 There are, however, some studies on the impact of a nuclear phase out in Germany. For example, Fürsch et al. (2012) find that nuclear will be replaced by more coal-fire power and new gas fire capacity in Germany, as well as with increased imports of electricity. Knopf et al. (2014) examine the impact on German electricity prices and CO2 emissions under a number of scenarios, stressing that the effects critically depend on which scenario that is examined. 7

30 that connect pairs of countries. Similarly, trade in electricity requires electricity transmission lines that connect pairs of countries. There are competitive world markets for coking coal, steam coal, oil and bio fuel, but competitive domestic markets for lignite. While fuels are traded in annual markets, there are seasonal (summer vs. winter) and time-of-day markets for electricity. In each country in EU-30 (henceforth referred to as a model country) there is demand for all types of energy from four groups of end users; the household sector, the service and the public sector, the industry sector and the transport sector. Demand from each end-user group (in each model country) is derived from a nested multi-good multi-period constant elasticity of substitution (CES) utility function; this is a truly non-linear function, making LIBEMOD a non-linear model. 4 In addition, there is intermediate demand for fuels from fuel-based electricity producers; gas-fired power stations demand natural gas, bio power stations demand biomass, etc. Extraction of all fossil fuels, as well as production of biomass, is modelled by standard supply functions. Energy is traded between countries. In addition, there are domestic transport and distribution costs for energy; these differ across countries, energy carriers and user groups. 5 For all energy goods, there is a competitive equilibrium; this is the case i) for all goods traded in a model country, ii) for oil, steam coal, coking coal and bio fuel traded in world markets, and iii) for transport services of natural gas and electricity between model countries. The price of each transport service consists of a unit cost and a non-negative (endogenous) capacity term; the latter ensures that demand for transport does not exceed the capacity of the gas pipe/electricity line. The capacities for international transport consist of two terms: pre-determined capacities (according to observed capacities in the data year of the model) and investment in capacities; the latter is undertaken if it is profitable. We now turn to electricity supply, which is the most detailed model block in LIBEMOD. In each model country electricity can (with several exceptions) be produced by a number of technologies: steam coal power, lignite power, gas power, oil power, reservoir hydropower, run-of-river hydropower, pumped storage hydropower, nuclear power, biomass power, wind power and solar. 4 There are also other non-linear functions in LIBEMOD, for example, in extraction of fossil fuels. 5 End-users also face different types of taxes, in particular value added taxes. The end-user price of an energy good is the sum of i) the producer price of this good, ii) costs of domestic transport and distribution of this energy good (which differ between countries, end-user groups and energy goods), iii) end-user taxes (which also differ between countries, end-user groups and energy goods), and finally iv) losses in domestic transport and distribution. 8

31 In each model country there are six pre-existing ( old ) electricity technologies: gas power, steam coal power, lignite power, oil power, bio power and a composite technology referred to as renewable. Moreover, there are four new fuel-based technologies using the same fuels (except lignite) as the pre-existing technologies and two new renewable technologies; wind power and solar power. In general, for each old fuel-based technology and each model country, efficiency varies across electricity plants. However, instead of specifying heterogeneous plants for each old technology (in each model countries), we model the supply of electricity from each old fuel-based technology (in each model countries) as if there were one single plant with decreasing efficiency; this implies increasing marginal costs. For each type of new fuel-based technology, we assume, however, that all plants have the same efficiency (in all model countries). Whereas for pre-existing technologies the capacity is exogenous (in each model country), for new plants the capacity is in general determined by the model. 6 There are six types of costs involved in electricity supplied from combustion of fuels. First, there are non-fuel monetary costs directly related to production of electricity, formulated as a constant unit operating cost c O. When in period t, the monetary cost in each period is E y t (TWh) is the production of power O E c y t, which must be summed over all periods to get the total annual operating costs. Second, there are fuel costs. Third, production of electricity requires that capacity is maintained: in addition to choosing an electricity output level, the producer is assumed to choose the level of power capacity (GW) that is maintained, PM K, thus incurring a unit maintenance cost M c per power unit (GW). Fourth, if the producer chooses to produce more electricity in one period than in the previous period in the same season, he will incur start-up or ramping up costs. In LIBEMOD these costs are partly expressed as an extra fuel requirement, but also as a monetary cost per unit of started power capacity in each period. For investments in new power capacity, K inv, there are annualised capital costs c inv related to the investment. Finally, for new plants there are costs related to connecting to the grid; these reflect either that the site of the plant is not located at the grid and/or that connecting a new plant to the grid requires upgrading of the grid and these costs may partly be borne by the plant. Under the assumption that the distance to the grid is increasing in the 6 For the pre-existing electricity technologies we use information from ENTSO-E (2011) on capacities for 2020, that we depreciate to 2030, in the model simulations for 2030 in sections 4 and 5. Thus, capacities that are expected to come online by 2020 are included in our study (as pre-existing technologies). 9

32 number of new plants, that is, increasing in new capacity, and/or costs of upgrading the grid is gc inv inv increasing and convex, the cost of grid connection, c ( K ) K, is also increasing and convex. Each plant maximizes profits subject to a number of technology constraints; for example, i) maintained power capacity should be less than or equal to total installed power capacity, ii) production of electricity in a time period should not exceed the net power capacity multiplied by the number of hours available for electricity production in that time period, and iii) because power plants need some down-time for technical maintenance total annual production cannot exceed a share of the maintained annual production capacity. For a more detailed discussion of electricity supply from fuel-based technologies, see Aune et al. (2008). We now turn to the modelling of renewables. In LIBEMOD there are now three types of hydroelectricity technologies; reservoir hydro, run-of-river hydro and pumped-storage hydro. Relative to the modelling of electricity supply from fuel-based technologies, reservoir hydro, which has a reservoir to store water, has two additional technology constraints. First, the reservoir filling at the end of season s cannot exceed the reservoir capacity. Second, total use of water, that is, total production of reservoir hydro power in season s plus the reservoir filling at the end of season s should not exceed total supply of water, that is, the sum of the reservoir filling at the end of the previous season and the seasonal inflow capacity (expressed in energy units, TWh). For the run-of-river hydro power technology, which is an extension of the LIBEMOD model presented in Aune et al. (2008), there is per definition no reservoir. Like for reservoir hydro there is, however, a restriction on use of water relative to availability of water; production in each time period cannot exceed the inflow of water. Finally, the pumped storage hydro power technology is characterized by buying electricity in one period (e.g. during the night) and using that energy to pump water up to the reservoir in order to produce electricity in a different (higher-price) period (e.g. during the day) by letting the water flow down through the generator again. As demonstrated in Aune et al. (2008), the optimization problem of this technology is similar to the one for fuel-based technologies, except that the pumped storage producer uses electricity (and not fuels) as an input. Bio power is modelled in exactly the same way as electricity supply from fuel-based technologies. The only difference is that bio power uses (carbon free) biomass as an input. Similarly to fossil fuels, biomass is supplied competitively. Whereas for solar power and 10

33 wind power we assume that production sites differ (with respect to magnitude of solar hours and wind hours), and we also take into consideration the amount of land available for solar and wind power, see sections 2.2 and 2.3, these factors are not relevant for bio power: the thermal efficiency of a bio power plant is of course independent of its location. Moreover, for the equilibrium biomass prices in our simulations bio mass mainly consists of waste and byproducts from agriculture and industry, that is, products not requiring separate land to be manufactured Wind power - modeling We assume that wind sites differ with respect to annual wind hours and that the best site for wind power (in terms of annual wind hours) is developed for wind power production before the second best site is developed, and so on. This is formalized by f ( K ), which shows average number of wind hours per year (measured in kh) as a decreasing function of aggregate capacity of wind power plants. Because production of wind power depends on the PM amount of the capacity that is maintained, K, we define the annual energy (electricity PM PM production) capacity of wind power (TWh) by f ( K ) K. Also for wind power we have some technical constraints. First, maintained power capacity should be less or equal to installed power capacity, which for a new power plant is equal to investment in electricity production capacity: PM inv E K K 0 (1) E where is the shadow price of installed power capacity. Second, let W t be the share in period t of the annual number of wind hours. This means that maximum production of wind power in period t is t f ( K ) K there is an upper limit on production of electricity in this period: W PM PM, and hence E W PM PM y f( K ) K 0 (2) t t t 7 For hypothetically higher biomass prices, other types of biomass products would be supplied, and some of these require agricultural land. Note that for biofuels, that is, energy carriers used in the transport sector, the alternative value of land may be substantial in several countries, see, for example, Searchinger et al. (2008). In 2012, 2 percent of the agricultural land was used for biofuel production in the EU. Because the growth in equilibrium consumption of biofuel is moderate in LIBEMOD, there is no need to introduce restrictions on land use for biofuel production. 11

34 where t is the shadow price of the periodic energy capacity. Finally, also for wind power there is need for technical maintenance. Therefore, total annual production cannot exceed a share ( ) of the maintained annual production capacity: E PM yt tk 0 (3) t t where t is the number of hours available for electricity production in period t (kh) and is the shadow price of the annual energy capacity. PM PM Note that when we defined f ( K ) K as the annual energy capacity of wind power we (implicitly) assumed that technical maintenance of wind power plants takes place in periods without any wind; this assumption reflects that the number of wind hours even at the best site is by far below total number of hours in a year. We also (implicitly) assumed that if the installed capacity of some (new) wind power plants is not maintained, then these plants are located at sites with the lowest number of annual wind hours. This assumption will be fulfilled if producers maximize profits, as we assume. In fact, with profit-maximizing wind power producers (and no uncertainty) the entire invested capacity will be maintained in the model. Like for fuel-based technologies, wind power has a constant operating unit cost, c O, as well as a constant unit maintenance cost, M c. However, there is of course no fuel cost and there are no start-up costs for a wind power plant. Therefore, the Lagrangian of the optimizing problem of new wind power is: L P y c y c K c K c ( K ) K E YE E o E M PM inv inv gc inv inv t t t t T t T E PM inv E W PM PM E PM K K t yt t f( K ) K yt tk. t T t T t T (4) The first-order condition for produced electricity in each period is: YE O E P c y 0. (5) t t t 12

35 E This is a standard first-order condition, simply stating that an interior solution, that is, y 0, requires that the difference between the price of electricity P YE t and the marginal operating cost of production c O should be equal to the sum of two shadow prices. The first is the shadow price of the periodic energy capacity where t 0 reflects that increased production in period t is not possible for a given maintained capacity PM K. The second is the shadow price of the annual energy capacity. Because the maximum number of operating hours during the year ( t ) will, for reasonable values of, always exceed the number of wind t T hours at the best site, we have 0. The first-order condition for maintained capacity is: t (6) ( W )( f( K PM ) f ( K PM ) K PM ) c M E K PM 0. t T t t t t T This first-order condition states that the cost of increasing maintained capacity marginally M E the sum of the maintenance cost ( c ) and the shadow price of installed capacity ( ) should (in an interior solution) be equal to the value of increased annual production following from this policy. Increased maintained capacity raises potential periodic and annual electricity production. Therefore, the value of increased production is i) the shadow price of periodic W energy capacity ( t ) weighted by the wind share in this period ( t ) and summed over the year when the effect on annual production of wind power due to increased maintained PM PM PM capacity ( f( K ) f ( K ) K ) is taken into account, plus ii) the value of increased potential annual production, which is the shadow price of the annual energy capacity ( ) times the maximum number of operating hours during the year ( ). Finally, the first-order condition for investment is given by t T t gc inv E inv gc inv dc ( K ) inv inv c c ( K ) K K 0. (7) inv dk This condition implies that if investment is positive, then the total annualised investment cost, which includes the marginal cost of connecting to the grid, must equal the shadow price of 13

36 E installed capacity ( ), i.e. the increase in operating surplus resulting from one extra unit of capacity. As always, in addition to the FOCs with respect to the decision variables the FOCs with respect to the multipliers recover the original optimisation restrictions. 2.3 Wind power - calibration PM We impose a linear function on f( K ): PM W W PM f( K ) a b K. (8) PM Because f( K ) shows average number of wind hours (per year) as a decreasing function of aggregate maintained capacity, a W should be interpreted as the number of wind hours (per year) at the best site (in a country). We have determined this parameter by using information from Storm Weather Centre (2004), EEA (2009) and Hoefnagels et al. (2011). From these sources we found the best location for wind power in each model country, with annual load hours ranging from 1500 to 3700, see Table 1. The load hours are defined as the ratio between annual electricity output of a wind turbine and its rated capacity (for details on how this is estimated see Hoefnagels et al. (2011)). Table 1 Wind hours at best site and wind power potential in model countries In order to determine the value of b W we have to solve the optimization problem of a profit- maximizing agent investing in new wind power. To simplify, we assume that maintained capacity is equal to invested capacity (which is the case for a profit-maximizing agent). We YE also assume that the price of electricity is constant over the year ( P ), and hence we focus only on annual production ( E y ). This implies that we have only one restriction on wind power production; this restriction is related to total annual production of wind power. The Lagrangian of the optimizing problem of new wind power is then: L E P YE y E c o y E c M K PM c inv K PM y E f( K PM ) K PM. (9) t T 14

37 Note that relative to the real decision problem of a wind power producer, see (4), we have gc PM PM removed costs of grid connection ( c ( K ) K ) because the price of electricity in (9) is measured at the production node. The first-order condition for annual produced electricity is: YE O E P c y 0. (10) Further, the first-order condition for investment is Using (8), this condition can be rewritten as: PM PM df ( K ) PM M inv ( f ( K ) K ) c c. PM dk ( a w 2 b w K PM ) c M c inv K PM 0. (11) E PM PM Finally, the first-order condition wrt. the multiplier is y f( K ) K. Using (8) and the fact that a profit-maximizing producer always will use the entire maintained capacity, this first-order condition can be rewritten as E W W PM PM y ( a b K ) K. (12) Based on available data we solve the system (10), (11) and (12) by treating E y, YE P and W a as exogenous variables. Then this system determines (from (10)), PM W K and. b The parameters are calibrated for each country for the year We assume cost of investment in inv wind power, c, decreases by an annual rate of 1 percent. This implies that this annualized cost falls from 1000 /kw in 2005 to 576 /kw in We now explain how we set values for E y and YE P. Our calibration of b W draws on Eerens and Visser (2008), which has data for wind power potential (TWh) in Europe for This report provides a technical potential for each country, which is then reduced by excluding all sites with wind speeds below 4 m/s and land where biodiversity issues could prevent development (all land registered in the Natura

38 database or as nationally designated areas). For each country the remaining generation potential, referred to as the market potential, has been categorised into three cost classes. These are labelled Competitive, Most likely competitive and Not competitive, and the potential within the two first classes are sites with production costs below /kwh. Thus, the Eerens and Viser study provides information about profitable potential wind power production in 2030 (in a country) if the price of electricity is constant over the year and equal to 0.07 /kwh in Because wind power requires use of land, which typically has an opportunity cost, actual wind power production will only be a small share of potential wind power production. It is hard to estimate this share, but in this study we assume that if the price of electricity is 0.07 /kwh in 2030, total production of wind power in 2030 will be of the same magnitude as total production of electricity in EU-30 in our data year 2009 (3399 TWh). To be more specific, we assume that 10 percent of the wind power potential in 2030, which amounts to 3816 TWh, see Table 1, will be available for electricity generation in By fixing annual E wind power production in 2030, y, to 10 percent of the potential wind power production if the annual price of electricity is 0.07 /kwh in 2030 ( P (wind hours at best site in a country) from Table 1, we can determine YE ), and using the values for a W W b for the year Finally, we have made some rough estimates of land use by wind power under the assumption that actual production of wind power amounts to 10 percent of potential wind power production in In the literature two approaches are common: either to include areas directly related to wind power production (the mills, access roads to the mills, and other facilities) or the entire area of the wind park (which may encompass areas used for, say, agricultural production between the mills). According to Denholm et al. (2009), direct land use amounts to between 3 km 2 /GW and 17 km 2 /GW, whereas total area affected varies between 100 and 500 km 2 /GW. Hence, if 10 percent of the wind power potential in 2030 is developed, and the average land use rates are applied (10 km 2 /GW vs. 300 km 2 /GW), direct land use would be almost 1 percent of the EU-30 land mass, whereas the total area affected would equal 24 percent of the EU-30 land mass. 2.4 Solar power - modeling The main solar power technologies are Centralized Solar Power (CSP) and Photovoltaics (PV). The latter is a method of generating electrical power by converting solar radiation into 16

39 direct current electricity by using solar panels containing photovoltaic material. We have chosen to model PV, which, based on available cost estimates, seems to be the most promising technology. The PV technology requires land to produce electricity. Let be actual use of land (measured in 2 Gm ) to produce solar power in a country in a year. Under ideal conditions, the PV technology requires 1 2 m to produce 1 kw momentarily, and therefore is the momentarily production of electricity (KW per m 2 ) under ideal conditions. The actual momentarily production capacity of solar under ideal conditions (measured in GW) is therefore K. (13) Further, let ˆ be the amount of land available to solar power (in a country in a year) where ˆ. Then the maximum momentarily production capacity is must have K Kˆ. K ˆ ˆ, and obviously we We now derive measures for annual energy capacity of solar power. First, let be annual solar irradiance (kwh per ) in a country. Then measures received energy by the solar panels throughout a year. Second, let be the share of energy received by the solar panels that is actually transformed to solar power. Annual energy capacity of solar power (TWh) is then. Alternatively, annual energy capacity can be expressed by zk where z measures annual solar hours (measured in kh), defined from the identity zk. Using (13) this identity can be rewritten as 2 m z. (14) So far we have implicitly assumed that each solar panel receives the same amount of energy. However, sites differ wrt. solar irradiance. We now assume that there is a continuum of sites (in a country) and these can be ranked according to their solar irradiance. Further, we assume that when solar production capacity is developed the best solar site is used before the second best site, etc. Hence, the more solar power that is developed, the lower is the average amount of energy received by the solar panels. This mechanism is captured by letting the measure of 17

40 K solar irradiance,, be a downward sloping function of the capacity utilization: ( ). Kˆ K Note that ( ) should be interpreted as the average solar irradiance. Kˆ Using the identity (14), we now define our measure of annual solar hours: K ( ) K ˆ z( ) K. (15) K ˆ By letting S t be the share of annual solar hours in period t, we have a measure of energy PM S K PM capacity of solar power in this time period: t z( ) K. Here we have substituted actual Kˆ PM production capacity ( K ) by maintained production capacity ( K ) because production requires that panels are maintained and we assume that producers always maintain the panels at the best sites (A profit-maximizing actor investing in solar will in fact maintain the entire installed capacity). A producer investing in solar power faces the same type of technical constraints as an agent investing in wind power: First, maintained power capacity should be less or equal to PM inv installed power capacity, that is, K K. Second, there is a restriction in periodic PM E S K PM production of electricity: yt t z( ) K. Finally, due to technical maintenance there is Kˆ E PM a restriction on total annual production of electricity: y K 0. t t In addition, t t because of limited availability of land for solar power, there is also a restriction on investment: inv ˆ E K K 0 (16) E where is the shadow price of land. Thus for solar power, which has the same type of costs as wind power, the Lagrangian of the optimisation problem is: 18

41 L P y c y c K c K c ( K ) K E YE E O E M PM inv inv gc inv inv t t t t T t T PM E PM inv E inv ( ) ( ˆ M E S K PM K K K K) t { yt t z ( ) K } Kˆ E yt tk t T t T PM. t T (17) The first-order condition with respect to electricity produced in each period is the same as the one for wind power, see (5). The first-order condition for maintained capacity is t T PM PM PM M S K K K M E PM t t ( z( ) z ( ) ) t c K 0. Kˆ Kˆ Kˆ (18) t T Finally, the first-order condition for investment is given by gc inv E E inv gc inv dc ( K ) inv inv c c ( K ) K K 0. (19) inv dk These conditions have similar interpretations as those for wind power. 2.5 Solar power - calibration In the model it is assumed that all solar power is based on photovoltaic (PV) technology and organised as centralised power plants. The PV cells are assembled as modules that are used for electricity generation (IEA ETSAP 2011). There are several PV technologies on the market and under development. These are often divided into three categories; (i) firstgeneration PV systems based on crystalline silicon technology, (ii) second-generation thin film PV (based on several different materials) and (iii) third-generation PV which includes new technologies like concentrated PV, organic solar cells and dye sensitized solar cells. The first-generation PV systems are fully commercial, whereas the second-generation are in the stages of early market deployment (IRENA 2012a). In the model we use technical data and costs of first-generation PV systems. To estimate the potential of the solar resource in each model country data for solar insolation around the world from the NASA Surface Meteorology and Solar Energy database 19

42 has been used. 8 This gives information about the monthly average insolation incident, measured in kwh/m 2 /day, based on a 22-year average. We use the data for tilted collectors, choosing the tilt angle that gives the highest annual average for each location. 9 We have created a dataset with a best and worst location for solar insolation (kwh/m 2 /year) for each model country, see Table 2. These locations have been chosen based on an assessment of each model country using a map of PV potential in the EU regions 10 and sampling from the NASA database. The data have been aggregated to our two seasons (summer/winter). Table 2 Solar insolation kwh/m 2 /year (Average radiation incident on an equatorpointed tilted surface) K We assume that the function ( ) is linear: K K a. Kˆ S b S Because ( ) should be 2Kˆ Kˆ interpreted as the average solar irradiance, the marginal solar irradiance is given by S S a b. This means that a should be interpreted as the irradiance at the best solar site of K ˆ a country. To determine the value of b S note that if the entire amount of land for solar power S K is used, then the marginal site receives a solar irradiance of for each country, the values of a S (best site) and a S b the value of S b for each country. S a S b S. From Table 2 we know, (worst site), and hence we can find In the model we assume that over time more land will be available for solar. In particular, we rely on Hoefnagels et al. (2011) which assumes that 0.5 percent of the agricultural land 11 will be made available for solar power plants in each model country by The increase of land available for solar power is captured by the function l( T 2009) where the parameters k and l are calibrated so that h(2050) 1 ht ( ) ke ( k 2.5, l ). This means that around 0.3 percent of the agricultural land will be made There are various ways to measure solar irradiance. Global horizontal irradiance (GHI) is a measure of the density of the available solar resource per surface area. However, GHI can also be measured with tilted collectors that have a fixed optimal angle for the location or even with devices that track the sun. We use data for tilted collectors that have a fixed optimal angle Potential.pdf 11 Data on agricultural land are gathered from: According to this source, for EU-30 agricultural land amounts to 41 percent of total land mass. 20

43 available for solar power plants in each model country in For EU-30 the share of total land mass used to solar power production would then be 0.2 percent in IEA ETSAP (2011) has data for land use (m 2 /kw) for PV technologies. According to this study, the typical current international range for crystalline Si PV cells is between 6 and 9 m 2 /kw. In the model 7 m 2 /kw has been used, which means that 7 m 2 is required to generate 1 1 kw instantly under optimal conditions. Hence,. Based on the assumptions in IEA 7 ETSAP (2011) and IPCC (2011), the maximum module efficiency of PV panels is assumed to be 18 %, that is, Finally, also for solar we assume that cost of investment is decreasing over time; the annual rate is set to three percent. Above we derived that annual production of solar power can be calculated from K K ( ). Using i) 0.18, ii) average solar insolation ( ( )) by country from Table 2, Kˆ Kˆ and iii) the assumption that 0.3 % of the agricultural land will be made available for solar power plants in each model country in 2030 ( ), we can calculate maximum solar power by country in 2030, see Table 3. According to this table, maximum solar power in 2030 amounts to 1620 TWh, which is close to 50 percent of total electricity production in EU-30 in Table 3 Potential solar power production in 2030 by country (TWh) 3 Costs of electricity A key factor in determining the impacts of a nuclear phase out is costs of electricity, in particular cost of electricity from new power plants. Costs of electricity will affect to what extent a phase out will be replaced by new capacity and also the mix of the electricity technologies, that is, the two main research questions in this paper. Figure 2 shows average cost of new electricity in 2030 measured in 2009 euro per MWh by technology, that is, new gas power, new coal power, new bio power, new wind power, new solar, new CCS based on natural gas (termed gas CCS greenfield) and new CCS based on coal. Figure 2 Average costs of electricity in 2030 ( 2009/MWh) 21

44 In the figure costs have been split into three factors; costs of investment, costs of operation and maintenance (O&M), and fuel costs. Because Figure 2 provides information about costs in 2030, we have taken into account that costs of investment (per GW) will fall over time due to learning (see below). For fuel costs, we have used observed fuel prices in 2009 (including taxes) for electricity producers, averaged over EU-30, and specific assumptions about the efficiency of new fuel based plants (see below). For wind power and solar we show cost of electricity for very good locations in Europe (3500 wind hours and 2500 solar hours annually). As seen from Figure 2, average cost per TWh varies from 40.3 (bio power) to 79.4 (CCS gas greenfield). We now comment on the different cost factors in more detail. 3.1 Cost of investment and efficiency of new plants The LIBEMOD model distinguishes between steam coal and lignite power plants, however it is only possible to invest in new steam coal plants. According to Burnard and Bhattacharya (2011), the super-critical (SC) technology is currently the standard for new plants in industrialised countries: despite emerging types of coal power plants like integrated gasification combined cycle (IGCC) and circulating flue gas desulphurisation (CFGD), the super critical and ultra-super critical pulverised coal plants continue to dominate the new orders. For coal power plants coming online in 2030 we have therefore used cost data for an ultra-super critical (USC) pulverised coal plant; the OECD (2010) estimate for this technology is /kw (data from the Netherlands). For natural gas the majority of the estimates from OECD (2010) are for combined cycle gas turbine (CCGT) plants. The estimates differ between the reporting countries. In the model the cost estimate from Belgium ( /kw) has been used, see Table 4, which is very close to the average of all the CCGT-estimates in the publication. Table 4 Investment costs in 2010 ( 2009/kW) Tyma (2010) and Schröder et al. (2013) are among the few studies that provide cost estimates for new oil-fired power plants. After assessing the available sources an investment cost of /kw was assumed. 22

45 The investment cost for new wind power plants was based on an assessment of various sources (Mott MacDonald 2010; OECD 2010; IPCC 2011; NVE 2011; Black & Veatch (2012) and IRENA 2012b). Offshore wind power is not included in the LIBEMOD model. The cost estimates for onshore wind in OECD (2010) range from 1419 /kw to 2742 /kw. In LIBEMOD it is assumed that the investment cost falls over time at a rate of 1 % per anno. Based on these considerations, in the LIBEMOD model the investment cost of a new wind power plant is /kw for 2009 ( /kw for 2030). Numerous sources were reviewed for the cost of solar PV (OECD 2010; IEA 2011a; IEA ETSAP 2011; IPCC 2011; IRENA 2012a; Bazilian et al and Schröder et al. 2013). An estimate of /kw is used for 2009, which is towards the lower end of the estimates of these sources. The reason is partly that some of the publications are several years old, and that the cost of solar PV installations has been dropping dramatically in recent years. Schröder et al. (2013) goes even lower, using 1560 /kw after reviewing numerous sources. They base their decision on the dynamics of the solar power market in recent years and argue that this leaves even the lower estimates in the literature outdated. However, because the base year in the LIBEMOD model is 2009 a higher estimate than 1560 /kw seems reasonable for However, we assume that investment cost per GW falls with 3 percent per anno from 2009 this gives us /kw for IPCC (2011) defines biomass as Material of biological origin (plants or animal matter), excluding material embedded in geological formations and transformed to fossil fuels or peat. This wide definition of biomass and also the variety of technologies that come under the term bio power means that landing on a cost estimate for a generic biomass-based power plant is problematic. The cost of bio power depends on type of feedstock used, boiler technology, plant capacity and type of plant. The estimates from OECD (2010) vary considerably from country to country, mainly due to differences in the reported technologies. IEA ETSAP (2010a) has a range for typical values for a biomass CHP plant in 2010 and an estimate for expected costs in For new plants it seems reasonable to go with the lower end of the IEA ETSAP (2010a) estimates; we assume that the cost of a new biomass power plant is /kw for For the hydro technologies apart from pumped storage, cost data for Norway from the Norwegian Water Resources and Energy Directorate, for example, NVE (2011), has been used. The costs for other model countries are then based on this, but adjusted with an investment cost coefficient creating country specific costs for run-of-river and reservoir hydro plants. This coefficient is based on the load hours for each technology compared to Norway. 23

46 We can then construct country specific investment cost estimates for reservoir hydro and runof-river plants. The cost of new pumped storage is taken from IEA ETSAP (2010b). In this technology brief they use 2900 /kw for a typical large hydro power plant (with costs ranging from 1300 to 4500 /kw). According to the IEA, for pumped storage costs can be up to twice as high as for equivalent plants without pumps. Based on this the investment cost for a new pumped storage plant in the model is set to 1.5 times the cost of a typical plant given by IEA ETSAP (2012b), that is, 4363 /kw. Efficiencies for new power plants have generally been taken from OECD (2010), which has efficiency estimates for plants coming online in Because of the assumption that the cost of a new plant (of a given technology) is the same for all model countries, the same applies for the efficiencies. However, for new pumped storage there is constant efficiency within each country, but these efficiencies differ across countries because of, e.g. topological differences. For each model country, the efficiency for new pumped storage is set equal to the efficiency of pumped storage in the base year. Table 5 shows the efficiencies used in the model for new power plants. For gas-fired power plants an efficiency of 60 percent is assumed, and for coal-fired power plants an efficiency of 46 percent is assumed. Table 5 Efficiencies for new power plants 3.2 CCS technologies We now turn to carbon capture and storage (CCS) technologies, which is a process to prevent CO 2 from being released into the atmosphere. A power plant with CCS is able to capture (most of) the CO 2 and transport it to a suitable location where it can be permanently stored. 12 CCS is still an immature technology, and there are various capture technologies under development. There are four different carbon capture and storage technologies in the LIBEMOD model; retrofit CCS for existing coal power plants, retrofit CCS for existing gas power plants, greenfield CCS coal power plants and greenfield CCS gas power plants. The greenfield plants are new gas and coal power plants complete with CCS. The costs of the two retrofit options are based on the CCS technology being retrofitted to an already existing power plant. A CO 2 capture level of 90 % is assumed for all CCS technologies

47 The costs of greenfield gas and greenfield coal plants are taken from ZEP (2011). 13 The report distinguishes between several different types of power plants with CCS. After consultation with industry experts, a combined cycle gas turbine (CCGT) plant and an integrated gasification combined cycle (IGCC) coal power plant were chosen. 14 The investment costs for these were /kw and /kw respectively for 2030, see Table 6. For retrofit CCS costs there were fewer sources. When an already existing power plant is being retrofitted with CCS equipment, the investment costs involved will be power plant and site specific. These costs are therefore more difficult to predict. However, for the LIBEMOD model we assume that there is one retrofit technology for natural gas and one for coal. IEA GHG (2011) has investment costs for several different retrofit solutions for natural gas and coal power plants. After consultation with industry experts, we decided to use the costs for the integrated retrofit solution. For a natural gas plant the investment cost for this type of retrofit is /kw, whereas for a coal plant it is /kw (for 2030). These estimates assume that the investment costs for all CCS technologies fall with 0.5 percent per anno. Table 6 Investment costs of power plants with CCS for 2030 ( 2009/kW) The cost of the integrated retrofit option only includes retrofitting the plant; the initial costs of the power plant are considered sunk. The costs in Table 6 do not cover the cost of transportation and storage of the CO 2. ZEP (2011) has cost data for these activities. According to this report, existing studies on transportation costs were inadequate for a review, so the costs in the report are based on input from EU-member states and in-house ZEP analysis. 13 The ZEP report compares several studies on the costs of CCS greenfield power plants. Compared to other studies, see, for example, IEA (2013a), ZEPs costs are at the lower end of the scale. This is partly due to some of the estimates being older, and probably also to the difference in type of power plants. Because the technology is still new and untested in full-scale plants, it is to be expected that the estimates differ. 14 The IEA report Power Generation from Coal (Burnard and Bhattacharya, 2011) supports our coal plant choice by describing IGCC as well placed to embrace CO 2 -capture and that the cost of CCS with this type of power plant is expected to be lower than for pulverised coal systems. 25

48 The two main transport options for CO 2 from a power plant are through a pipeline network or with ship. We have chosen to base our estimates on the pipeline option. 15 ZEP (2011) provides two sets of cost estimates for pipelines. One is for a typical capacity of 2.5 million tonnes per annum (Mtpa), which is considered to be appropriate for CCS demonstration projects and commercial natural gas plants with CCS, and the other is for a pipeline with typical capacity of 20 Mtpa, 16 which is thought more realistic for commercial large-scale networks. The unit transportation costs for CO 2 /tonne vary with distance and whether it is an onshore or offshore pipeline. We have assumed a cost of 6 /tco 2 for transportation. This is based on an offshore pipeline of 500 km with a capacity of 20 Mtpa. Storage costs depend on factors like field capacity, well injection rate and type of reservoir, and are thought to vary considerably between sites. ZEP (2011) provides low, medium and high cost scenarios for storage depending on type of well (depleted oil and gas field or saline aquifer) and whether it is located onshore or offshore. In Europe there is more offshore than onshore capacity, and more capacity in saline aquifers than in depleted oil and gas fields (ZEP 2011). This means that the majority of the potential European storage sites are of the most expensive kind. There has also been public resistance to storage onshore near where people live due to the risk of leakages. 17 Taking this into consideration we assume a storage cost of 10 /tco 2, 18 which is based on depleted offshore oil and gas fields in ZEP s medium cost scenario. Due to the carbon capture, CCS plants will incur an efficiency penalty compared to power plants without CCS. Most of the literature assumes that there is little difference in the actual efficiency penalty between greenfield plants and plants that are retrofitted. The difference in actual efficiency can mainly be attributed to how older existing plants that are candidates for being retrofitted have a lower efficiency than a newly built plant made specifically for CCS. The reduction in efficiency for retrofits is plant specific, and the plants efficiency will fall and costs increase depending on to what degree it is suitable for CCS (IEA GHG 2011). Many existing plants may not be good candidates for CO 2 capture due to being 15 The alternative to pipeline transportation of the CO 2 is ship. Transportation costs with ship are less dependent on distance and on the scale of the transport. However, to transport CO 2 by ship one has to factor in the costs of liquefaction. 16 It is assumed that the 20 Mtpa pipeline can serve a cluster of CO 2 sources and that it has double feeders from the source to the pipeline and double distribution pipelines. 17 According to the Special Eurobarometer (European Commission, 2011b), six out of ten people in Europe expressed concerned when asked how they would feel about a deep underground CO 2 storage site within 5 km of their home. For an overview of studies looking at public perception and acceptance of CO 2 storage, see IPCC (2005). 18 None of the above estimates include costs for monitoring the storage sites. IPCC (2007) estimates it to lie between 0.05 and 0.09 /tco 2. 26

49 too small and/or too inefficient. Burnard and Bhattacharya (2011) assume that the higher the efficiency of the existing plant, the more favourable it will be to retrofit. Due to the limited experience with retrofit projects, there is considerable uncertainty regarding how low a power plants initial efficiency can be before the plant is unsuitable for retrofit. According to industry experts at Gassnova, for coal plants there may also be a higher penalty for retrofitted plants if the damp from the turbine is not completely compatible with what the capture process requires. The IEA GHG study uses a 9 percentage point reduction for both greenfield and retrofitted plants. ZEP and NETL also use the same reduction across types of plants; 8 and 10 percentage points respectively. Based on this literature and advice from industry experts, we assume that the penalty for natural gas plants (greenfield and retrofit) is an 8 percentage point reduction in efficiency compared to a new power plant without CCS, and likewise a 9 percentage point penalty for both types of coal power plants. Figure 3 shows average costs of electricity from CCS plants. For CCS retrofit, cost of investment is solely CCS investment cost and fuel costs reflect efficiencies for good existing power plants. For all technologies we have used average EU-30 fuel prices for electricity generation in As seen from the figure, CCS coal is cheaper than CCS gas, and for both CCS coal and CCS gas retrofitting the most efficient plants is cheaper than building new CCS stations. Figure 3 Average costs of CCS electricity ( 2009/MWh) 3.3 Operation and maintenance cost In the model we differentiate between fixed and variable operation and maintenance costs (O&M). Fixed O&M costs are costs that incur irrespective of use of the plant and therefore can be viewed as long-run maintenance costs, whereas variable O&M costs are linked to the maintenance of the capacity that has been used during a year. The OECD-publication Projected Costs of Generating Electricity 2010 (OECD 2010) provides estimates for total O&M costs, so other sources have been used for the split between fixed and variable costs. Tidball et al. (2010), Black & Veatch (2012) and Mott MacDonald (2010) provide more detailed information about O&M costs. Schröder et al. (2013) provides a compilation of 27

50 different studies and their assumptions for fixed and variable O&M costs for different technologies. Based on an assessment of these sources a dataset has been created. 19 O&M costs from OECD (2010) have been used for natural gas, steam coal, lignite and nuclear power plants. For steam coal we assume that of the total O&M costs 54 percent are variable and 46 percent are fixed, whilst for lignite the allocation is 35 percent variable and 65 percent fixed. For natural gas (combined cycle) we assume that variable costs make up 55 percent and fixed 45 percent, and for nuclear 4 percent variable and 96 percent fixed is assumed. For oil power Tyma (2010) provides an overview of personnel costs, fuel costs and chemical costs, which have been allocated to fixed and variable costs in keeping with the above definition. For biopower we have used IRENA (2012c), and assumed that 42 percent of the O&M costs are variable, and 58 percent are fixed. In the overview made by Schröder et al. (2013) the majority of the studies on hydro power categorise all O&M costs as fixed. In their own dataset they report only fixed O&M. The O&M costs for pumped storage, reservoir and run-of-river hydropower in LIBEMOD are based on this. The O&M costs for solar power are based on data from the technology briefs from IEA ETSAP (2011). The costs for wind power are based on OECD (2010) and IRENA (2012b). For wind power the four studies evaluated by Tidball et al. (2010) differ considerably with respect to the allocation between fixed and variable costs. Two of the studies assume 100 percent fixed costs, and two assume 25 percent fixed costs and 75 percent variable costs. Schröder et al. (2013) compares O&M costs for onshore and offshore wind power from various sources and they vary between only fixed costs and a split between the two. In their cost proposal Schröder et al. (2013) assume all O&M costs are fixed. In LIBEMOD it is assumed a 25/75 split between fixed and variable costs. For mature technologies the same O&M costs have been used for existing plants and new plants. For bio, solar and wind power the costs for new plants are based on the same sources, but they are lower than for existing plants reflecting cost reductions as these technologies mature over time, see Table For the CCS technologies the O&M costs for greenfield plants are taken from ZEP (2011) and for retrofitted plants from IEA GHG (2011). However, the O&M costs for 19 For the thermal technologies a 70 % load factor has been assumed. 20 The IEA ETSAP technology briefs and IRENA reports provide intervals for costs, so for the existing technologies the higher end of the interval has been used, whereas for new plants the costs are assumed to be towards the lower end. 28

51 retrofitted coal plants have been adjusted somewhat as they were lower than for greenfield plants. Table 7 Operation and maintenance (O&M) costs for new power plants 29

52 4 Results 4.1 Scenarios To examine the effects of a nuclear phase out we consider a number of scenarios for 2030, see Table 8. In our reference scenario we assume that the nuclear capacities in 2030 reflect decisions taken in 2014 or earlier at the country level with respect to whether nuclear plants will be phased out or new nuclear capacity will come online before 2030, see Table 9. As indicated in Section 1, whereas some countries, for example, Belgium and Germany, have decided to completely phase out nuclear power, other countries, for example, Finland and the UK, are planning to build new nuclear stations. In addition, in several countries old nuclear stations will be taken out of production without being replaced. Based on information from IEA (2013b) and Eurelectric (2011) there may be a net decrease in nuclear capacity in EU-30 between 2009 and 2030 of about 23.2 GW, see Table 9, which amounts to roughly 20 percent of the 2009 nuclear capacity in EU-30. Hence, in the reference scenario total nuclear capacity in EU-30 is 23.2 GW lower than in the data year Table 8 Scenarios for 2030 Table 9 Nuclear policy in EU-30 In the reference scenario we assume that the climate policy of the EU in 2030 will follow the proposal of the EU Commission from January 2014 to reduced GHG emissions by 40 percent relative to 1990, see European Commission (2011a). This proposal distinguishes between the ETS sector (electricity generation and large carbon-intensive manufacturing firms) and the remaining sectors. Whereas the ETS sector has to reduce its GHG emissions by 43 percent relative to 2005, the corresponding number for the non-ets sector is 30 percent. In addition, the renewable share of energy consumption should be 27 percent; the EU Commission indicates that the latter target may be reached if the emission targets are reached. All targets are at the EU level and hence not broken down to national targets. 21 For other electricity technologies we use data from ENTSO-E System Adequacy Forecast (scenario B) on (predicted) capacities in 2020 by country. These reflect current capacities adjusted by planed investments and disinvestments. 30

53 In the reference scenario we follow the proposal of the EU Commission and hence have one common EU-30 target for emissions in the ETS sector implemented by a common quota system and one common EU-30 target for emissions in the non-ets sector implemented by a common uniform tax. Because LIBEMOD only covers CO2, the most important GHG gas, we transform the GHG emissions targets to CO2 targets. 22 In the next two scenarios we reduce the capacities of nuclear power in all model countries that did not completely phase out nuclear power in the reference scenario by either 50 percent relative to 2009 ( 50 % Phase out ) or by 100 percent ( 100 % Phase out ). The climate goals and climate instruments are, however, the same as those in the reference scenario. For the remaining scenarios we stick to the assumption that there has been a complete nuclear phase out. We first explore the impact of other assumptions with respect to emissions targets - either no climate policy (referred to as no policy ), or that GHG emissions are to be reduced by only 20 percent ( high emissions ) relative to For the latter scenario we assume, like in the reference scenario, that there are ETS and non-ets sector specific emissions targets, and the estimation of these targets follow the same procedure as in the reference case. Next, we examine the case of a 40 percent GHG reduction under the assumption of no specific targets for ETS and non-ets, that is, there is one common emissions target for EU-30. In this scenario ( effective ) we use a common uniform CO2 tax to reach the climate goal. In the scenarios above there was no target for the share of renewables in energy consumption. As stated above, the EU Commission suggested a common EU wide renewable target of a least 27 percent, which will be met in our reference case here the renewable share is 32 percent. Still, in order to explore the effect of a higher renewable share we impose - in the scenario referred to as EU renewable target - a renewable share of 35 percent. 23 This 22 Our strategy to find CO2 emission targets for EU-30 is mainly as follows. We use EEA (2013) to find GHG emissions for EU-27 in 1990, which is 40 percent above the 2030 emission target. Because Iceland, Norway and Switzerland each has committed to a conditional emissions reduction of at least 30 percent, we assume that also these countries will commit to a 40 percent GHG reduction by Based on Höglund-Isaksson (2011), which has projections for non-co2 emissions for ETS and non-ets, we find CO2 targets for ETS/non-ETS. Further, we take into account that LIBEMOD cannot distinguish between manufacturing firms that belong to the ETS sector (large carbon-intensive units) and those firms not covered by the ETS sector. We also take into consideration that the emission target for the transport sector is part of the 40 percent climate target, which is not compatible with the LIBEMOD model because the CES demand structure gives little room for substitution in transport due to the initial share of oil being very close to 100 percent. A more detailed description of the calculations of the LIBEMOD climate targets is available upon request. 23 We define the share of renewables in final energy demand as i) the sum of renewable electricity production (except from bio power) and total use of bioenergy relative to ii) total consumption of electricity (less of electricity used in pumped storage hydro) and total consumption of primary energy among end users. 31

54 target is reached by offering a production subsidy to all producers of renewable electricity (bio power, hydro power, solar power and wind power) and also a subsidy to all end-users of bioenergy (biomass and biofuel). The subsidies are identical measured per energy unit. One new electricity technology that may replace nuclear power is CCS. Both the EU and the IEA have published reports estimating that this technology may have a great future potential; according to the Energy Roadmap, see European Commission (2011c), the share of CCS in EU power generation in 2050 may become as high as one third. Likewise, the IEA Technology Roadmap from 2013, see IEA (2013a), predicts that in 2050 the annual amount of CO2 captured and stored globally (in electricity generation and in manufacturing processes) may be around 8000 MtCO2, which is roughly 25 percent of current global emissions of CO2. Costs of CCS are, however, high because of additional costs of investment (relative to conventional fossil fuel plants) and also due to additional energy use, see Section 3.2. In the scenario termed Cheap CO2 we explore the market outcome if a substantial share of CCS investment costs (50 percent) is covered by the government. The climate policy goals and instruments are the same as in the reference scenario. A nuclear phase-out will decrease total supply of electricity in the short run, and thereby push up investment in other electricity technologies because, cet. par., the price of electricity will increase. It seems reasonable to expect that also production of renewable electricity will increase, including supply from solar and wind power. The intermittency of these technologies will easily cause more price volatility in the electricity market, and the probability of a black out - triggered if consumers of electricity at a point in time try to use more electricity than the amount of electricity fed into the system - will also increase. In order to cope with these challenges national regulators design and implement arrangements that seek to ensure an effective electricity market. In LIBEMOD there are national capacity markets, and each national regulator buys maintained capacity (from non-intermittent technologies except nuclear power) according to a rule-of thumb; at least five percent of the maintained capacity should always be available for additional production. This potential production capacity is frequently referred to as balancing power. In the scenario termed balancing power we examine the impact of tightening the rule-of thumb by replacing 5 percent with 20 percent. 32

55 4.2 The effects of a nuclear phase out Figure 4 shows the capacity of electricity technologies in EU-30 in 2009, in the reference scenario, under a 50 percent nuclear phase out and under a complete phase out by As indicated by the figure, total capacity increases sharply from the observed 2009 value (917 GW) to the reference scenario for 2030 (1494 GW). The increase, which is roughly 60 percent, is mainly due to economic growth between 2009 and 2030, but it also reflects the climate policy in the reference scenario (40 percent emissions reduction): there is a sharp increase in wind power capacity from 74 GW in 2009 to 482 GW in the reference scenario. In 2009 the capacity share of wind power was 8 percent, whereas the share in the reference scenario is as high as 32 percent, see Table 10 which show capacity share by technology. As seen from Table 10, the total share of renewable power (bio power, hydro power, wind power, solar power and renewable electricity from other technologies) was 34 percent in 2009 and increases to 62 percent in the reference scenario. In 2009 the share of nuclear capacity was 14 percent, whereas in the reference scenario this share is down to 7 percent. The change reflects party the assumption that there will be 23.2 GW less nuclear capacity in the reference scenario than in 2009, and partly the sharp increase in total electricity capacity. Figure 4 Net capacity by technology in EU-30 in 2009 and 2030 (GW) Figure 5 Capacity share by technology in EU-30 in 2009 and 2030 (GW) Table 10 Capacity shares in EU-30 in 2030 As seen from Figure 4, total capacity increases slightly (by 60 GW) from the reference scenario to the case of a 50 percent phase out, and it increases further by 59 GW if nuclear is completely phased out. The corresponding increases in wind power capacity are in fact slightly higher. If nuclear is phased out, the capacity share of wind power is 40 percent whereas the capacity share of renewable electricity is 72 percent (34 percent in 2009). The small increase in total capacity from the reference scenario to a complete nuclear phase out suggests that total production of electricity increases, but the higher share of renewables, in particular for wind power, suggests the opposite because the capacity rate of wind power is by far lower than for nuclear, and also lower than for most fossil fuel plants. Figure 6 shows production of electricity by technology. As seen from the figure, total 33

56 production of electricity is roughly at the same level in the reference scenario, under a 50 percent nuclear phase out and also under a complete phase out (This common production level is almost 30 percent higher than in 2009). The composition of technologies changes, however, radically. Bio power production, which has a high rate of capacity utilization, is almost twice as high under a complete phase out as in the reference scenario, whereas the corresponding increase for wind power is almost 50 percent. The share of renewables in electricity production is 73 percent after a complete nuclear phase out, that is, almost identical to the capacity share of renewables. Figure 6 Electricity production in EU-30 in 2030 (TWh) In order to reach the climate targets there are separate uniform prices for CO2 emissions in the ETS sector and in the non-ets sector. As seen from Figure 7, the ETS price is 36 euro/tco2, whereas the non-ets price is much higher; 611 euro/tco2. The difference reflects much more flexibility in the power sector than among the end users. In the electricity generation sector, LIBEMOD specifies a number of alternative technologies. The composition of these may change radically if prices are altered. In contrast, end-user demand is derived from nested CES utility functions, that is, there is no direct substitution between technologies. With a CES utility function even a moderate change in consumption requires significant price changes. However, in the real world large changes in end-user prices may trigger installation and use of alternative technologies, for example, solar panels for domestic heating and electric cars in the transport sector. Because LIBEMOD neglects end-user technology substitution, the model overestimates the non-ets CO2 price. Figure 7 CO2 prices in EU-30 in 2030 ( 2009/tCO2) Figure 8 shows how total consumption of energy varies across scenarios. Here we have merged consumption of primary energy and consumption of electricity. There is no obvious way how these should be compared; in the figure we have transformed consumption of electricity from nuclear, hydro, solar and wind power to consumption of primary energy by using a transformation rate of MWh/toe, which is a standard conversion factor. Using 34

57 this transformation rate we see that consumption of energy is roughly at the same level in the three scenarios shown in Figure Figure 8 Energy consumption in EU-30 in 2030 (Mtoe) Finally, Table 11 shows (annual) producer and consumer prices by energy good in 2009 and in the three different scenarios. Because total production of electricity is almost the same in the three 2030 scenarios, and net import of electricity to the model countries is (per assumption) independent of scenario, consumption of electricity does not change much between the three 2030 scenarios. This is the reason why the consumer price of electricity is almost equal in the three 2030 scenarios. As seen from Table 11, the (annual) producer price of electricity does not change much between the three 2030 scenarios either. This is also the case for most other energy prices; the price that changes the most is the price of biomass, a fuel that is used primarily for bio power production. The increase reflects higher bio power production when nuclear is phased out (see discussion above) along with a biomass market with limited international trade because of high costs of transportation. Table 11 Producer and consumer prices in EU-30 in 2030 ( 2009/MWh or 2009/toe) 5 Robustness The main results from Section 4 are that consumption of electricity, as well as total consumption of energy, is not affected much by a nuclear phase out in In contrast, the mix of electricity technologies depends on the extent to which nuclear is phased out: the more nuclear capacity that is phased out, the higher is renewable electricity production. A nuclear phase out is almost entirely replaced by renewable electricity, that is, mainly wind power and bio power. We now examine to what extent these conclusions are robust relative to the specification of scenarios. Figures 9-12 show electricity capacity, electricity production, prices for emissions of CO2 and consumption of energy by scenario. In each figure we first repeat the case of a full 24 Consumption of fuel-based electricity, for example, coal power, is measured by the use of coal (toe) to produce electricity. 35

58 nuclear phase-out when GHG emissions are imposed to be 40 percent lower than in 1990 and there are separate ETS and non-ets targets (henceforth referred to as 100 percent ). Then the results of the remaining scenarios from Table 8 are shown. Without any climate policy ( no policy ) total capacity is somewhat more than 10 percent lower than in the 100 percent scenario. However, total production of electricity is almost identical in the two cases, reflecting that under no policy there is massive generation from coal power its market share is roughly two third and this technology has a high rate of capacity utilization. The high level of coal power production tends to reduce the price of electricity and therefore production of gas power is substantially lower than in the 100 percent scenario. The market share of renewable is almost 30 percent, that is, somewhat higher than in 2009 (24 percent). This exercise suggests that the impact of a nuclear phase out in the absence of a climate policy is mainly that nuclear production is replaced by fossil fuel based production. 25 In the case of one climate target and therefore one common price of emissions of CO2 ( efficient ), total electricity capacity is almost 10 percent above the capacity in the 100 percent scenario, but total production of electricity does not differ much between the two cases. In order to reach the climate target a uniform CO2 tax at 69 euro/tco2 has to be imposed. This is almost twice as high as the price of emissions in the ETS sector in the 100 percent scenario (38 euro/tco2), and therefore fossil fuel based technologies are punished hard in the efficient scenario. In fact, there is almost no conventional production of gas power and coal power in the efficient scenario, but CCS gas obtains a market share of roughly 5 percent. Thus, the market share of renewable electricity exceeds 90 percent. If total emissions are to be 20 percent lower than in 1990 (not 40 percent as in the 100 percent nuclear phase out scenario), total production of electricity is marginally lower than in the 100 percent scenario. As expected, the technology mix differs significantly between the two cases. In the case of high emissions, there is substantial coal power production, which crowds out some of the gas-fired power production. The resulting market share of renewable is 37 percent, which is significantly lower than in the 100 percent scenario (73 percent). In the 100 percent phase-out scenario each national system operator has to make sure that (at least) 5 percent of the maintained capacity is available for reserve power production in 25 In order to determine the effect of a nuclear phase out in the case of no climate policy, the equilibrium with nuclear capacities as in the reference scenario should be compared with the equilibrium after a complete nuclear phase out (when there is no climate policy in both cases). Such an exercise confirms the conjecture above. 36

59 case demand for electricity suddenly increases or supply suddenly drops. 26 If this requirement is increased to 20 percent due to the increased market share of intermittent renewable electricity maintained capacity is increased by almost 200 GW whereas installed capacity is only slightly affected, see Figure 9. Hence, a higher share of the installed capacity of preexisting plants is being maintained; it is much cheaper to increase maintained capacity by maintaining plants with low efficiency than to buy new power plants. Moreover, because the increase in maintained capacity is of the same magnitude as the increase in capacity acquired by the national operators, the available capacity for production is almost similar in the two cases. Hence, also production of electricity is similar in these two cases, see Figure 10. In the scenarios examined in Section 4 there is only marginal CCS production (3 TWh in Malta). If the government subsidizes 50 percent of all CCS investment costs ( cheap CCS ), the effect on total production of electricity is minor, but the market share of CCS becomes 11 percent. The resulting market share of renewable is around two third. Figure 13 shows the share of renewable in energy consumption across scenarios. In the reference scenario this share is 25.9 percent, that is, slightly below the EU target of 27 percent. Hence, our study suggests that if the climate targets are reached, the renewable target is almost reached. With a complete phase out, the share is 32.9 percent and hence the renewable target is by far accomplished. To explore the partial effect of tightening the renewable share in energy consumption we have imposed a renewable target of 35 percent when nuclear is fully phased out and GHG emissions are to be 40 percent below the 1990 level. The renewable target is implemented by offering a subsidy to producers of renewable electricity and to end-users of bioenergy (same subsidy per unit of energy). With a renewable subsidy the price of emissions of CO2 in the ETS sector decreases: the CO2 price in the ETS sector is now 29 euro/tco2, which is 9 euro/tco2 lower than in the 100 percent scneario. As seen from Figure 9, a higher renewable share in energy consumption increases total production of electricity slightly (by 2 percent). The main effects are that gas power decreases (by around 160 TWh) whereas wind power increases by almost 200 TWh. Finally, Table 12 shows producer and consumer prices by scenarios. The discussion above, along with an examination of Figures 11-13, shows that total production of electricity, and to a large extent also total consumption of energy, do not differ much between the scenarios where nuclear power is fully phased out. Moreover, from the discussion in Section 4 we know that these levels do not differ much from the reference 26 Because LIBEMOD is a deterministic model, the maintained capacity that is available for reserve power production is never actually used for electricity production. 37

60 scenario. However, the mix of electricity technologies differs significantly between scenarios. The equilibrium composition of electricity technologies reflects the stringency of the climate target and whether some technologies are being promoted, either directly through subsidies or indirectly through a tailor-made policy goal. Figure 9 Net capacity by technology in EU-30 in 2030 by scenario (GW) Figure 10 Electricity production in EU-30 in 2030 by scenario (TWh) Figure 11 CO2 prices in EU-30 in 2030 by scenario ( 2009/tCO2) Figure 12 Energy consumption in EU-30 in 2030 by scenario (Mtoe) Figure 13 Renewable share in EU-30 in 2030 by scenario Table 12 Producer and consumer prices in EU-30 in 2030 by scenario ( 2009/MWh or 2009/toe) 38

61 References Aune, F., R. Golombek, S. A. C. Kittelsen and K. E. Rosendahl (2008). Liberalizing European Energy Markets: An Economic Analysis. Cheltenham, UK and Northampton, US. : Edward Elgar Publishing. Bazilian, M., I. Onyeji, M. Liebreich, I. MacGill, J. Chase, J. Shah, D. Gielen, D. Arent, D. Landfear and S. Zengrong (2013). Re-considering the economics of photovoltaic power. Renewable Energy, Black & Veatch. (2012). Cost Report: Cost and Performance Data for Power Generation Technologies. National Renwable Energy Laboratory. Burnard, K. and S. Bhattacharya (2011). Power Generation from Coal - Ongoing Developments and Outlook (Information paper). Paris: International Energy Agency, OECD Publishing. Denholm, P., M. Hand, M. Jackson and S. Ong (2009). Land-use requirements of modern wind power plants in the United States. Technical Report NREL/TP-6A , August National Renewable Energy Laboratory (NREL). EEA (2009). Europe's onshore and offshore wind energy potential, An assessment of environmental and economic constraints, EEA Technical report 6/2009. Copenhagen: European Environment Agency (EEA). EEA (2013). Annual European Union greenhouse gas inventory and inventory report Submission to the UNFCCC Secretariat, Technical report no 8/2013, European Environment Agency, Copenhagen. Eerens, H. and E. de Visser (2008). Wind-energy potential in Europe , Technical Paper 2008/6. Blithoven: European Topic centre on Air and Climate Change (ETC/ACC). EURELECTRIC. (2011). Power Statistics and Trends Brussels: Union of the Electricity Industry - EURELECTRIC. European Commission (2011a). A roadmap for moving to a competitive low carbon economy in 2050, Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions, COM 112, Brussels. lex.europa.eu/resource.html?uri=cellar:5db26ecc-ba4e-4de2-ae08- dba649109d /doc_1&format=pdf European Commission (2011b). Special Eurobarometer 365, Public awareness and Acceptance of CO2 capture and storage. Brussels: European Commission, TNS Opinion & Social. European Commission (2011c). Energy Roadmap Impact Assessment. SEC(2011) 1565 final

62 European Commission (2013). EU Energy, transport and GHG emissions. Trends to 2050: Reference scenario Available at: pdf European Commission (2014). Guidelines on State aid for environmental protection and energy /C 200/ Fürsch, M., D. Lindenberger, R.Malischek, S. Nagl, T. Panke and J. Trüby (2012). German Nuclear Policy Reconsidered. Implications for the Electrcity Market. Economics of Energy & Environmental Policy, Vol. 1(3), Golombek, R., K. A. Brekke and S.A.C. Kittelsen (2013): Is electricity more important than natural gas? Partial liberalizations of the Western European energy markets. Economic Modelling, 35, Gutschi, C., U. Bachhiesl, C. Huber, G. Nischler, A. Jagl, W. Süßsenbacher and H. Stigler (2009). ATLANTIS Simulationsmodell der europäischen Elektrizitätswirtschaft bis Elektrotechnik & Informationstechnik, 126/12, Haller, M., S. Ludig and N. Bauer (2012). Decarbonization scenarios for the EU and MENA power system: Considering spatial distribution and short term dynamics of renewable generation. Energy Policy, 47, Hoefnagels, R., M. Junginger and A. Held (2011). Long Term Potentials and Costs of RES, Part I: Potentials, Diffusion and Technological Learning. RE-Shaping, Intelligent Energy - Europe. Höglund-Isaksson, L., W. Winiwarter, F. Wagner, Z. Klimont and M. Amann (2010). Potentials and costs for mitigation of non-co2 greenhouse gas emissions in the European Union until Results. Report to the European Commission, DG Climate Action Contract No /2009/545854/SER/C5. International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria. IEA (2011a). Renewable Energy Technologies: Solar Energy Perspectives. Paris: OECD/IEA. IEA (2011b). World Energy Outlook Paris: OECD Publishing. IEA (2013a). Technology roadmap: Carbon capture and storage, 2013 edition. Paris:OECD/IEA. IEA (2013b). OECD - Net electrical capacity. IEA Electricity Information Statistics database. doi: /data en IEA ETSAP (2010a). Biomass for Heat and Power Technology Brief E05. IEA Energy Technology Network - Energy Technology Systems Analysis Programme. IEA ETSAP (2010b). Hydropower Technology Brief E12. IEA Energy Technology Network - Energy Technology Systems Analysis Programme. IEA ETSAP (2011). Photovoltaic Solar Power Technology Brief E11. IEA Energy Technology Network - Energy Technology Systems Analysis Programme. 40

63 IEA GHG (2011). Retrofitting CO2 Capture to Existing Power Plants. Paris: IEA. IPCC. (2005). IPCC Special Report on Carbon Dioxide Capture and Storage, Prepared by Working Group III of the Intergovernmental Panel on Climate Change [Metz, B., O. Davidson, H. C. de Coninck, M. Loos, and L. A. Meyer (eds.)]. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press. IPCC (2007). "Summary for policymakers". In: Climate Change 2007: Mitigation, Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [B. Metz, O.R. Davidson, P.R. Bosch, R. Dave, L.A. Meyer (eds)]. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press. IPCC (2011). IPCC Special Report on Renewable Energy Sources and Climate Change Mitigation. Prepared by Working Group III of the Intergovernmental Panel on Climate Change [Edenhofer, O., R. Pichs-Madruga, Y. Sokona, K. Seyboth, P. Matschoss, S. Kadner, T. Zwickel and P. Ei.] Cambridge, United Kingdom and New York, NY, USA: Cambridge Univerwity Press. IRENA (2012a). Solar Photovoltaics, Renewable Energy Technologies: Cost Analysis Series, Volume 1 Power Sector, Issue 4/5. International Renewable Energy Agency. IRENA (2012b). Wind Power, Renewable Energy Technologies: Cost Analysis Series, Volume 1, Power Sector Issue 5/5. International Renewable Energy Agency. IRENA (2012c). Biomass for Power Generation, Renewable Energy Technologies: Cost Analysis Series, Volume 1 Power Sector, Issue 1/5. Bonn: International Renwable Energy Agency. Knopf, B., M. Pahle, H. Kondziella, F. Joas, O. Edenhofer and T. Bruckner (2014). Germany s nuclear phase-out: Sensitivities and impacts on electricity prices and CO2 emissions. Economics of Energy & Environmental Policy, Vol. 3(1), Mott MacDonald. (2010). UK Electricity Generation Costs, Update London: Department of Energy and Climate Change. Natura (2005). A data overview of the network of special protection areas in the EU25, A working paper from the European Topic Centre on Biological Diversity. Paris. NETL (2013). Carbon dioxide transport and storage costs in NETL studies (DOE/NETL- 2013/1614). National Energy Technology Laboratory. NVE (2011). Kostnader ved produksjon av kraft og varme, Håndbok 1/2011 (Costs of producing power and heat). Oslo: Norges vassdrags- og energidirektorat. OECD/Nuclear Energy Agency (2010). Projected Costs of Generating Electricity Paris: OECD Publishing. Schröder, A., F. Kunz, J. Meiss, R. Mendelevitch and C. von Hirschhausen (2013). Data documentation, Current Prospective Costs of Electricity Generation until Berlin: Deutsches Institut für Wirtschaftsforschung (DIW). 41

64 Searchinger, T., R. Heimlich, R.A. Houghton, F. Dong, A. Elobeid, J. Fabiosa, S. Tokgoz, D. Hayes and T-H. Yu (2008). Use of U.S. Croplands for Biofuels Increases Greenhouse Gases Through Emissions from Land-Use Change. Science 319: Storm Weather Center (2004). Et røft estimat av vindkraftpotensialet i Europa (A rough estimate of the potential for wind power in Europe). Bergen: Storm Weather Center. Tidball, R., J. Bluestein, N. Rodriguez and S. Knoke (2010). Cost and performance assumptions for modelling electricity generation technologies. Farifax, Virginia: ICF International for National Renewable Energy Laboratory (NREL). Tyma, F. (2010). Fixenkostendeckung über den stromgrosshandelsmarkt und wohlfartsökonomische optimale preise. Technische Universität Graz, Institut für Electrizitätswirtschaft und Energieinnovation. Zero Emissions Platform. (2011). The cost of CO2 Capture, Transport and Storage - Postdemonstration CCS in the EU. European Technology Platform for Zero Emission Fossil Fuel Power Plants. Retrieved from 42

65 Table 1 Wind hours at best site and wind power potential in model countries Country Best (load hours) Potential* 2030 (TWh) Country Best (load hours) Potential* 2030 (TWh) AT IE BE IS BG IT CH LT CY LU CZ LV DE MT DK NL EE NO ES PL FI PT FR RO GB SE GR SI HU SK Sources: Eerens and Visser (2008), EEA (2009), Hoefnagels et al. (2011a) and Storm Weather Centre. *In the model only 10 % of the potential from Hoefnagels et al. (2011a) has been used. Table 2 Solar insolation in kwh/m2/year (Average radiation incident on an equator-pointed tilted surface) Country Best site kwh/m2/yr Worst site kwh/m2/yr Country Best site kwh/m2/yr Worst site kwh/m2/yr AT IE BE IS BG IT CH LT CY LU CZ LV DE MT DK NL EE NO ES PL FI PT FR RO GB SE GR SI HU SK Sources: All data from the NASA Surface meteorology and solar energy database 43

66 Table 3 Potential solar production in model countries in 2030 (TWh)* Country Potential production (TWh) Country Potential production (TWh) AT 24.4 IE 28.4 BE 9.8 IS 13.2 BG 46.1 IT CH 12.6 LT 19.6 CY 1.5 LU 0.9 CZ 15.4 LV 13.6 DE MT 0.1 DK 18.3 NL 16.1 EE 6.9 NO 6.2 ES PL FI 15.5 PT 42.1 FR RO GB SE 21.5 GR 86.5 SI 4.0 HU 45.7 SK 13.9 *Based on solar panel efficiency of 18%, maximum available land for solar power in 2030 (0.33 % of agricultural land in each country) and average insolation for each country. Table 4 Investment costs in 2010 ( 2009/kW) Technology LIBEMOD IEA ETSAP (2010) Schröder et al. (2013) IEA (2010) Mott MacDonald (2010) 1 EU (2013) 2 Natural gas (CCGT) Coal (PC SC) Oil Nuclear (EPR) Biomass Solar (PV) Wind (onshore) The data from Mott MacDonald are for nth of a kind plant in their medium scenario. 2 EU data is for EU coal plant is IGCC, not PC SC. 4 The data from Schröder et al. includes decommissioning and waste disposal. 44

67 Table 5 Efficiencies for new power plants for 2030 Technology Efficiency Coal 46 % Coal CCS greenfield 37 % Gas 60 % Gas CCS greenfield 52 % Table 6 Investment costs of power plants with CCS for 2030 ( 2009/kW) Type of CCS plant Technology Investment costs Natural gas - greenfield Combined Cycle Gas Turbine (CCGT) 1829 /kw Coal greenfield Integrated gasification combined cycle (IGCC) 3080 /kw Natural gas retrofit Integrated retrofit (CCGT) 665 /kw Coal retrofit Integrated retrofit (PC) 1035 /kw Table 7 Operation and maintenance (O&M) costs for new power plants in 2030 ( 2009) Variable O&M costs /MWh Fixed O&M costs /kw/year Natural gas Coal Lignite Oil Nuclear Bio Pumped storage Reservoir hydro Run-of-river Solar PV Wind CCS coal greenfield CCS coal retrofit CCS gas greenfield CCS gas retrofit

68 Table 8 Scenarios for 2030 Reference scenario 50 percent phase out 100 percent phase out No policy Efficient High emissions Cheap CCS EU renewable target Balancing power Nuclear capacities reflect decisions after percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. Nuclear capacities reduced by 50 percent in 2030 relative to percent GHG reduction in 2030 relative to Separate targets for ETS and non- ETS sectors. Complete nuclear phase out by percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. Complete nuclear phase out by No environmental target. Complete nuclear phase out by percent GHG reduction in 2030 relative to One common emission target for ETS and non-ets sectors. Complete nuclear phase out by percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. Complete nuclear phase out by percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. Costs of CCS investment reduced by 50 percent relative to reference scenario Complete nuclear phase out by percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. One common EU target for share of renewable energy of 35 percent. Complete nuclear phase out by percent GHG reduction in 2030 relative to Separate targets for ETS and non-ets sectors. Increased requirement of balancing power. 46

69 Table 9 Nuclear policy in EU member states COUNTRY POLICY PLANNED CAPACITY CHANGE Belgium Complete phase-out by MWe phase-out by MWe phase-out by 2025 Bulgaria Plans to extend lifetime of current reactors. Plans for a new reactor on hold due to lack of financing. Czech Rep National energy plan to 2060 assumes 50% nuclear capacity, however plans for two reactors are put on hold after the government refused to provide state support. Finland France Germany Hungary Italy Lithuania Netherlands One EPR reactor under construction, expected to be in commercial operation by Another two reactors planned. One EPR reactor under construction. The current President has pledged to reduce the share of electricity from nuclear to 50% by Closed down 8 reactors in March Plans for complete phase-out by Plans for two new reactors under government ownership. Plans to revive the national nuclear industry rejected by referendum in Closed down two reactors in 2009 due to EU safety concerns. Plans for one new reactor, expected to start operating in Previous decision on phase-out was reversed in However, plans for new reactors are on hold due to economic uncertainties. Poland Cabinet decision to move to nuclear power in Currently two planned reactors. Romania Two new reactors planned, but currently lacking financing. Slovakia Plans for new reactors outlined in the 2008 Energy Security Strategy, aiming to keep the share of electricity from nuclear at 50%. Slovenia Spain Considering capacity expansion, but no plans confirmed. Political uncertainty surrounding nuclear future. No plans for new reactors, but in 2011 the legal limitation to plant operating lives was removed (previously 40 years). Sweden Phase-out plan from 1980 repealed in June Currently plans to uprate/replace old units when decommissioned. Switzerland United Kingdom Parliament decision in June 2011 to not replace any reactors. Complete phase-out by Plans for several new reactors between 2023 and Government goal is 16 GWe new capacity by MWe in MWe in MWe in MWe around MWe in MWE in MWe shut down in MWe phase-out by MWe in MWe in after MWe in MWe in MWe in MWe in MWe in MWe in by MWe in by MWe phase-out by 2022 (net) 985 MWe phase-out by 2030 (net) 1165 MWe phase-out by 2034 (net) MWe by

70 Table 10 Capacity shares in EU 30 in 2030 (%) 2009 Reference 100% phase-out Nuclear power Oil power Coal power Coal power CCS Gas power Gas power CCS Bio power Hydropower Wind power Solar power Other renewables Table 11 Producer and consumer prices ( 2009/MWh or 2009/toe) Reference 50% Phase-out 100% Phase-out Producer Consumer Producer Consumer Producer Consumer Electricity price Natural gas price Steam coal price Coking Coal price Lignite price Oil price Biofuel price Biomass price Table 12 Robustness: Producer and consumer prices ( 2009/MWh or 2009/toe) No Policy Efficient High emissions Cheap CCS EU Renewable target Balancing Power Producer Consumer Producer Consumer Producer Consumer Producer Consumer Producer Consumer Producer Consumer Electricity price Natural gas price Steam coal price Coking coal price Lignite price Oil price Biofuel price Biomass price

71 Figure 1 The LIBEMOD model 49

72 Figure 2 Average costs of electricity in 2030 ( /2009/MWh) Figure 3 Average costs of CCS electricity in 2030 ( /2009/MWh) 50

73 Figure 4 Net capacity by technology in EU 30 in 2009 and 2030 (GW) Figure 5 Capacity shares in EU 30 in 2009 and

74 Figure 6 Electricity production in EU 30 in 2009 and 2030 (TWh) 52

75 Figure 7 CO 2 prices in EU 30 in 2030 ( 2009/tCO 2 ) 53

76 Figure 8 Energy consumption in EU 30 in 2009 and 2030 (Mtoe) Figure 9 Robustness: Net capacity by technology in EU 30 in 2030 (GW) 54

77 Figure 10 Robustness: Electricity production in EU 30 in 2030 (TWh) Figure 11 Robustness: CO 2 prices in EU 30 ( 2009/tCO 2 ) 55

78 Figure 12 Robustness: Energy consumption in EU 30 in 2030 (Mtoe) Figure 13 Renewable share in final energy demand in EU 30 by scenario 56

79 Sensitivity analysiss and costs of a European renewable target for Abstractt Brigitte Knopf 2, Paul Nahmmacher, Eva Schmid (PIK) In Europe, the share of renewable energy sources (RES) has been increasing i steadily over the past decades. The European Union (EU) has set both a target for the reduction of greenhouse gases (GHG), and one for the share of renewable energies in final energy demand for The currentt political debate is not only on how to continue with the GHG target for 2030, but also whether an additional RES target should be set again. Such an additional target for renewables is often justified in the political debate by referring to their potential co benefits, e.g. employment effects, local value added, additional environmental benefits or representing industrial policy aspects. In this paper we intend i to quantify the additional costs that a renewable target in the electricity sector would have, depending on its ambition. Moreover, by means of a sensitivity analysis we show that, from a GHG mitigation perspective, the cost effective share of renewables varies considerably with different plausible input assumptions. We conclude that the additional costs for RES targets set at higher than cost effectivambition off the RES levels are moderate; however, they t increase non linearly with the level of the target. 1. Introduction In Europe, the share of renewable energy sources (RES) has been increasing i steadily over the past decade: It rose from just above 8% in thee year 2004 to 14% in the yearr 2012 (EUROSTAT, 2014a).. Especially in the electricity sector the share of RES increased significantly, reaching 23.5% in 2013 (EUROSTAT, 2014b). Thiss development was mainly due to the dedicated supportt of RES deployment by means of feed in tariffs or other policy instruments designed to fulfil the EU wide endeavor of increasing the share of renewables. In 2009, the European Union (EU) has adopted the EU climate andd energy package (the so called package) that includes (i) a 20% reduction in EU greenhouse gas (GHG) emissions from 1990 levels, (ii) raising the share of renewables in the EU s final energy consumption to 20% 1 This rese earch was made possible through financial support from the European Commission under the 7th Framework Programme of the European Union to the project Economic iinstruments to Achieve Climate Treaties in Europe (ENRACTE), project number Corresponding author. Potsdam Institute for Climate Impact Research (PIK), Research Domain III: Sustainablee Solutions, P.O. Box , Potsdam, Germany. potsdam.de 1

80 (including a renewable share of 10% in the transport sector), and (iii) a 20% improvement in the EU's energy efficiency. With separate targets for renewables and for energy efficiency, there is a clear overlap between different policies that have an impact on GHG reduction. These additional targets, especially for renewables, are often justified in the political debate by referring to potential cobenefits that renewables create, such as employment effects, local value added, additional environmental benefits and industrial policy (Edenhofer et al., 2013; Lehmann and Gawel, 2013). The European Commission refers to co benefits of renewable energy deployment, when stating that "increased shares in renewables [...] contribute to more indigenous energy sources, reduced energy import dependence and jobs and growth (European Commission, 2013a). These co benefits play an increasingly dominant role in the current debate on environmental, climate and energy policy, especially against the background of the present economic down turn (Creutzig et al., 2014). This suggests there are multiple objectives implicitly or explicitly considered by national policymakers in the choice of renewable policies apart from GHG mitigation (Edenhofer et al., 2013). Hence, policy makers should carefully consider whether renewable deployment is the most cost efficient way of achieving other policy objectives that are associated with renewables (Knopf et al., 2013). So far, a systematic comparative quantification of the potential co benefits, on the one hand, and the costs of an additional renewable target, on the other hand, is lacking. This paper aims to contribute to the quantification of the latter aspect in a systematic manner. In the context of the proposed 2030 framework for climate and energy policies, in January 2014 the European Commission suggested an EU wide renewable target of 27% in final energy consumption. Whether a RES target will be adopted after all is subject to the current political debate, contesting whether an explicit RES target (next to an emission reduction target) is useful; and if so, what is an appropriate number? The suggested number of 27% is derived from the EU Commission s own modeling analysis (European Commission, 2014) and is designed to be consistent with a 40% GHG reduction target for the year This means that a 27% share of renewable energy is costeffective when a target of 40% GHG emission reduction is set. In that sense, the renewable target of 27% does not constitute any additional ambition for the deployment of renewables beyond its contribution to reduce GHG emissions. In the following, we refer this benchmark as the costeffective RES target, meaning the share of RES that is necessary for achieving a given climate target at least costs over time 3. If the share of RES is meant to be higher than that cost effective share we postulate that the underlying rationale is not climate change mitigation, but other considerations. However, it is important to realize that both in our and the European Commission s quantitative analysis, these other considerations are neglected only the climate mitigation target is taken into account explicitly. In that sense, the term cost effective here means: cost effective with respect to the long term climate mitigation target and no other RES related targets. The main reason for not taking into account other targets and potential co benefits is that existing energy system models are not capable of representing the underlying processes in sufficient detail. The quantitative result of 27% as a cost effective RES share in 2030 is derived from an analysis based on one single energy system model, namely PRIMES (Capros et al., 2014; E3MGLab, 2011), and is thus subject to a variety of specific set of model and data related input assumptions. Given the high political relevance of the RES target, it is hence crucial to better understand the robustness of the 3 It also refers to the economic welfare perspective on these costs in contrast to costs seen by individual market participants. Our specific analysis looks only at the electricity sector, not the overall economy. 2

81 optimal RES target in 2030 with respect to such underlying assumptions. Therefore, the key questions addressed in this paper are: 1. What is the cost effective RES share in the European electricity sector for the year 2030 that is consistent with a 40% GHG reduction target in 2030, given varying key input assumptions concerned with both technology availability and institutional settings? What are the decisive drivers of the cost effective RES share in the context of climate mitigation? 2. What are the additional economic costs and system effects if the RES target is set higher than this cost effective share? For answering these questions, we perform a sensitivity analysis with a refined version of the European electricity sector model LIMES EU (Haller et al., 2012). Next to an update of the calibration data and techno economic assumptions (see Nahmmacher et al. (2014b)), we have improved LIMES EU in several respects, but especially regarding the representation of variable RES (vres), in order to better capture the key effects of an increasingly high share of intermittent feed in. To this end we have developed a novel approach for deriving time slices for long term power system models. Our approach is mainly based on Ward s (1963) hierarchical clustering algorithm. We apply this algorithm on historic electricity demand and weather data to group days with similar demand and vres load patterns. Eventually, each group of days is reflected by a representative day in the power system model. The method is easily reproducible and applicable to all sorts of input data with multiple fluctuating time series. It is described in more detail in Nahmmacher et al. (2014a) 4. The remainder of this paper is structured as follows. Section 2 briefly presents the model LIMES EU, the novel time slice approach developed in order to adequately represent the key features of intermittent renewables in the electricity system as well as a selection of input data for the model. While Section 3.1 outlines the scenario setup, Section 3.2 discusses the optimal RES target level that would be set by the EU in 2030 according to the results of the model LIMES EU and analyses the economic costs and system effects that are incurred when setting a higher than optimal RES target. Section 4 summarizes and concludes. 2. Method and data 2.1. The electricity sector model LIMES EU The partial electricity system model LIMES EU (for details see Haller et al., (2012) and Nahmmacher et al. (2014b)) is designed to generate quantitative scenarios that represent a consistent, systemcost optimal transition towards a decarbonized European electricity system in In its current version the model comprises 26 of the 28 EU Member States 5 plus Norway, Switzerland and the Balkan region. The capacities of generation and storage technologies are aggregated to national levels, with each country constituting one model region 6. The transmission grid in LIMES EU is represented via Net Transfer Capacities (NTCs) between the model regions. The model is calibrated to the base year 2010, for which installed capacities are fixed. Table 2 and Table 3 in the Appendix 4 potsdam.de/members/paulnah/nahmmacher carpediem wp.pdf 5 The insular states Malta and Cyprus are not included in the current model version 6 with the exception of the non EU countries in the Balkan region that are grouped to one model region 3

82 report the installed capacities of generation and storage as well as renewable technologies in 2010 for every model region based on Platts (2011) and EUROSTAT (2013a). The installed transmission network is reflected by the NTC summer values of 2010 as reported by ENTSO E (2013a). Endowed with perfect foresight, LIMES EU yields a social planner solution that specifies in time steps of 5 years for each model region the optimal (i) dispatch and curtailment of installed electricity generation technologies, (ii) electricity import balance from neighboring model regions, (iii) investments into installed capacities of electricity generation technologies and (iv) investments into NTCs between model regions. Specified as a linear optimization model, the objective function of LIMES EU is to minimize the total sum of discounted 7 electricity system costs (comprised of fuel, investment, fixed and variable operation and maintenance costs) jointly for all model regions between 2010 and 2050, given an exogenous demand of electricity and a number of technological and political boundary conditions. Climate policy is simulated by constraining annual CO 2 emissions as suggested by the Roadmap for moving to a competitive low carbon economy in 2050 (European Commission, 2011b), leading to a near decarbonization of the electricity system in 2050, see Section 3.1. In order to consider fluctuating feed in of vres and differences in electricity demand occurring on time scales that require higher than annual resolution for an appropriate representation, a new time slice approach was developed and applied, which is briefly described in the next Section and in more detail in Nahmmacher et al. (2014a) A novel time slice approach In order to explore scenarios on the future of power systems, a variety of numerical models have been developed. Due to computational restrictions many long term power system models refrain from an hourly resolution but use time slices that aggregate similar load and renewable electricity generation levels instead. As the share of vres is projected to increase significantly in future electricity systems, suitably accounting for their temporal and spatial variability becomes ever more important for developing sound long term scenarios. Recent model applications use a variety of approaches; however, to date there is no reproducible and validated method to derive and select time slices for power system models with multiple fluctuating time series. In Nahmmacher et al. (2014a), we have developed a novel and computationally efficient method that is easily applicable to input data for all kinds of power system models. We have designed the time slice approach to optimally fulfil the following three essential requirements. The derived time slices should sufficiently reflect (1) the annual electricity demand and average vres capacity factors for each region, (2) the region specific load duration curves of electricity demand and vres technologies, and (3) the spatial correlation of electricity demand and vres electricity feed in among regions. The first requirement ensures that the quality of a region with respect to solar and wind power is correctly reflected. By replicating both common and rare load situations as well as their respective frequency of occurrence (second requirement), the timeslices neither overestimate nor underestimate single events. This serves to correctly value both base and peak load plants. The third requirement ensures that the characteristics of an interconnected multi regional electricity system are correctly assessed. 7 We apply a social discount rate of 5%. 4

83 We explicitly developed our approach in order to be applied on a discretionary number of timeseries simultaneously. Multiple time series exist when there is more than one model region and/or more than one type of exogenously varying load, i.e. electricity demand and the feed in of vres technologies such as wind and solar. Our approach is based on the hierarchical clustering algorithm described by Ward (1963). It groups similar historic days to clusters in a way that the inner cluster variance respecting all dimensions (i.e. values of electricity demand and vres infeed) is kept to a minimum. Each cluster of historic days is represented by one representative day in the power system model. We choose that day as a representative day which leads to the smallest sum of squared errors with respect to the other days in its cluster. In this way, the approximation of each load duration curve is optimized while at the same time accounting for the other time series. For more details of this general method and its validation see Nahmmacher et al. (2014a) Input data for LIMES EU Time slice dependent input data LIMES EU contains several load series with time( slice) dependent values for each model region: Electricity demand and fluctuating infeed from the RES technologies wind onshore, wind offshore, photovoltaic and concentrated solar power. The input data for the time slice approach is based on ENTSO E (2013b) data for the historic electricity demand levels and on historic weather data from ECMWF (2012) for the vres infeed. Using weather data rather than historic infeed data allows for taking a longer time span into consideration and prevents the over estimation of unusual years. The ECMWF data set comprises 33 years of solar radiation, ground temperature and wind speed levels at 120m height for Europe. For every third hour between 1979 and 2011 the respective information is given for local data points in a spatial resolution of 0.75 x0.75. After converting this weather information to time series of infeed levels of representative wind and solar power plants, the data points are aggregated to replicate the vres infeed in the model s sub regions. Compared to the vres infeed, the intra year demand fluctuations are less stochastic and follow distinct diurnal, intra week and seasonal patterns. Though the absolute demand levels changed over the years for demographic and economic reasons, the relative intra year fluctuations are assumed to remain the same. The hourly demand data of 2010 and 2011 that is available from ENTSO E (2013b) for all model regions is therefore deemed to be representative for the intra year demand side fluctuations between 1979 and Future inter year growth of annual demand is subject to scenario assumptions. However, as we use normalized values for the clustering, these scenario assumptions do not affect the time slice approach. Nahmmacher et al. (2014a) shows that when applying the time slice approach on the input data required for LIMES EU, a total of 48 time slices are appropriate to model the crucial features of vres generation in Europe. It turns out that spring and fall days can be represented in similar time slices and less time slices are required for summer days as compared to winter days because their variation is lower. With this computationally efficient method we are able to give an appropriate representation of vres within Europe, which is crucial when analyzing potential RES targets for

84 Regional RES potential The potential for RES deployment varies across model regions. Table 4 in the Appendix shows the bioenergy potential used in LIMES EU. It is based on EEA (2006) 8, in a way that one third of the potential stated in this publication is considered eligible for electricity production in LIMES EU. This accounts for the fact that not all of the biomass potential stated in EEA (2006) can be deployed at competitive prices and that the transport and heat sector also demand a considerable amount of the available biomass stock. In case the potential calculated for a specific country is smaller than its biomass deployment target stated in the National Renewable Energy Action Plans (NREAPS) (European Commission, 2013b), the potential is adjusted to cover this target 9. The limited availability of sites suitable for deploying hydro power is reflected through an upper limit on the installable capacity of hydro power plants. As the potential for further hydro power capacities is low in most European countries, capacity additions are only allowed up to the NREAPS targets (European Commission, 2013b). In addition to the installable capacity, the capacity factors of hydro power plants vary across model regions. Average capacity factors are derived from EUROSTAT (2013a) and EUROSTAT (2013b). Both the maximum installable capacities and the associated capacity factors are given in Table 4 (see Appendix). A country s wind and solar power potential is defined through two aspects: (1) the installable capacity of wind and solar power plants and (2) the achievable capacity factors at the respective sites. The installable capacity is again determined by three factors: The area that is suitable for installing a specific technology is derived from land cover (FAO, 2013) and elevation (NGDC, 2013) data. Due to public acceptance and competing usage possibilities only a certain share of this area is actually available for power production. Third, the amount of capacity that can be installed in this area is subject to technology specific restrictions. Table 5 in the Appendix summarizes the parameters used for each technology to calculate their respective capacity potential. Onshore wind turbines can be installed in forests and agricultural areas. Additional usage, such as food production on agricultural land, is still possible as the wake effect 10 considerably limits the maximum density of wind turbines per square kilometer. Sites eligible for offshore wind power plants lie within a distance of less than 55km to the mainland and belong to the exclusive economic zone (VLIZ, 2012) of the respective model region. Sites with a water depth of more than 50m are excluded. Additionally, only a share of the resulting area may be used for offshore wind power to prevent wind turbines from being installed too close to the mainland shore or smaller islands as well as to account for shipping corridors. In order to determine the potential for solar PV one needs to differentiate between large PV systems that are installed on former agricultural land and small PV systems mounted on rooftops and facades. In contrast to onshore wind power, no other use of the land dedicated to solar power is possible, as 8 Biomass potentials of countries for which no data is available in EEA (2006) are calculated based on the extent of arable land and forests in these countries (FAO, 2013) as well as the land structure and biomass potential of the sourrounding countries with available data. 9 This is the case for Belgium, Denmark, Luxemburg and the Netherlands. 10 The wake effect describes the turbulence of the wind stream behind a turbine. This turbulence prohibits the installation of wind turbines in too close proximity. 6

85 the PV shade most of the ground. For that reason, only a small share of a model region s agricultural area is eligible for large PV systems. Following IEA (2002) the available rooftop and facade area for small PV systems is approximated based on a model region s population. However, to account for the deployment of solar heating panels only half of the area potential stated in (IEA, 2002) is available for solar PV (Held, 2010). Similar to large solar PV systems, CSP plants may only be installed on former agricultural land. However, as we assume a SM4 configuration 11 in LIMES EU the maximum installable capacity per square kilometer is much smaller compared to PV systems. To account for the varying quality of wind and solar sites within a country, we define three resource grades per intermittent renewable technology for every model region. Each resource grade comprises a certain share of the region s area and is assigned the average technology specific capacity factor of this area (based on the weather data in ECMWF (2012)). The assignment is made in a way that the first resource grade comprises the best resource sites of a region that together add up to 10% of the region s area. The second resource grade comprises the next best sites that add up to 30% of the region s area. Consequently the third resource grade contains 60% of a region s area subsuming the sites with the lowest capacity factors. The assignment of resource grades is done separately for every technology that is based on wind or solar power. Table 6 (Appendix) shows the technologies overall capacity potentials per model region, valid for all the entire time period until The capacity factors per model region and resource grade are given in Table 7 (Appendix) Default techno economic assumptions The future development of a host of techno economic parameters in the energy system is subject to large uncertainty. In the case of LIMES EU it is inter alia necessary to make assumptions on the investment cost and technical characteristics of generation, transmission and storage technologies, fossil fuel prices and electricity demand. In order to enable a comparison we use techno economic assumptions similar to those used in the Impact Assessment of the European Commission (2014) as a default. Table 8 and Table 9 in the Appendix report the respective assumptions on the future development of investment costs and other techno economic characteristics for vres, thermal and storage technologies. If not mentioned otherwise, all prices and costs stated in this paper are measured in Table 10 (Appendix) displays fuel costs, which are also based on the assumptions of the European Commission (2014). Final electricity demand in the model s calibration year 2010 is retrieved from EUROSTAT (2013c) and IEA (2012). Demand projections until 2050 are likewise taken from the projections of the European Commission (2014). Demand growth rates for model regions not mentioned there are estimated based on the growth rates of their neighboring countries for which data is available. Our default assumptions for future electricity demand are reported in Table 11 in the Appendix. To account for intra regional transmission and distribution losses, we impose that the required production of electricity has to exceed the reported final electricity consumption by 15% (cf. EUROSTAT (2014c)). The annual expansion of cross border transmission capacities is restricted to 0.2GW of net transfer capacity (NTC) per cross country connection in the default case. This constraint serves to proxy for the notion that there is a certain limit on the speed of capacity expansion that can be achieved, due to social or bureaucratic considerations. 11 A solar multiple 4 (SM4) configuration indicates that maximum input is four times higher than maximum output of the CSP plant. Such configurations include large storage to balance power input and output during the day (Trieb et al., 2009). 7

86 3. Sensitivity analysis of the cost effective RES share in 2030 This Section deals with the question of what would be a cost effective share of RES in the European electricity sector for the year 2030 according to the results of the model LIMES EU. As already mentioned in the Introduction, the notion of cost effectiveness in this context refers to costeffective with respect to the long term climate mitigation target (and no other RES related targets). As it depends on the choice of input assumptions, we vary decisive parameters. The analysis shows to what extent the cost effective RES share is dependent on the chosen input assumptions and thereby reveals which assumptions are particularly important to consider Scenario setup In order to be comparable to the European Commission s Impact Assessment (European Commission, 2014), we configure our default techno economic and policy assumptions similar to its scenario GHG40. This scenario assumes a GHG reduction of 40% in the year 2030 and 80% by 2050, relative to 1990 levels. In our context, we implement these boundary conditions to an electricity sector model and therefore use the respective numbers for the electricity sector and for CO 2 (instead for all GHG) from the Commission s analysis. This corresponds to a CO 2 reduction target of minus 95% by 2050 in comparison to 1990 (European Commission, 2014), which is an exogenously set constraint in our model. Intermediate targets between 2010 and 2050 are increasing linearly and result in a CO 2 reduction in the electricity sector of 52% 12 by Other policy constraints in LIMES EU are kept to a minimum, e.g. there are no additional policies implemented to foster the electricity production from RES. However, all primary resources except uranium and hard coal are subject to environmental and/or societal constraints. In order to reflect the political situation of nuclear power in Europe, we implement a nuclear phase out in Germany, Belgium and Switzerland. New installations in other countries are limited to those currently under construction or planned, or for replacing depreciated capacities, see Table 12. In order to address the key research questions given in Section 1, we analyse next to the default scenario, referred to as COM policy, a number of scenarios with different assumptions regarding technology development and institutional settings, see the upper part of Table 1 for an overview. Per scenario we vary one assumption. Concerning technology settings, we analyze the impacts resulting from different conceivable developments on the availability of nuclear power and carbon capture and storage (CCS), as well as investment costs for renewables and electricity storage technologies. We also look at the impact of assuming that only sites with average RES capacity factors would be available for wind and solar power plants as well as the impact of higher fuel prices for biomass. In addition, we analyse scenarios that represent different institutional settings, e.g. with respect to grid transmission capacity and electricity trade, the latter reflecting the range between a fully integrated European electricity market versus a more nationally focused setting. Further institutional issues are energy security concerns and energy efficiency improvements that affect the development of the total electricity demand by Finally, we consider the possibility that a less 12 European electricity sector emissions in 2010 were at around 90% of 1990 levels. To reach a 95% emission reduction until 2050 with a linearly decreasing emission pathway, 2030 emission levels have to be at 47.5% of 1990 levels (i.e. a reduction of about 52%). 8

87 ambitious mitigation policy is pursued as reflected in the European Commission s reference scenario. This considers GHG emission reductions of 32% by 2030 and 44% by 2050 (European Commission, 2013c), translating into 73% CO 2 emission reduction in the electricity sector by Table 1: Overview of scenarios Technology availability Institutional setting RES target Technology / Institution Issue Scenario Name As in GHG40 of the Impact Assessment for the 2030 framework by COM policy the European Commission (2014) Nuclear Power Phase out until 2030 nucout 2030 Phase out until 2050 nucout 2050 ( 50% in 2030) CCS Not available no ccs vres (wind, solar) Higher investment costs high cost vres (+10% in 2020; +20% in 2030 and thereafter) Lower investment costs low cost vres ( 10% in 2020; 20% in 2030 and thereafter) Only sites with average low pot vres capacity factors available Biomass Higher fuel costs (+100%) high cost bio Storage Higher investment costs high cost stor (+100%) Transmission capacity expansion process is gridlocked Transmission capacity expansion process gains momentum / Completion of the internal market Energy security concerns strictly domestic Energy efficiency programs are successful Reference scenario of the European Commission (2013) No capacity expansion beyond today, completion of internal market impeded Faster capacity expansion (and market integration) possible (expansion rate +100%) More than 95% of electricity supply from domestic power plants Electricity demand lower by 5% in 2030 and thereafter Electricity demand lower by 10% in 2030 and thereafter 44% GHG reduction target in the year 2050 no trans exp high trans exp 95% national demand 5% demand 10% COM reference minimum 50% RES share in 2030 and thereafter RES target 50% minimum 55% RES share in 2030 and thereafter RES target 55% minimum 60% RES share in 2030 and thereafter RES target 60% minimum 65% RES share in 2030 and thereafter RES target 65% minimum 70% RES share in 2030 and thereafter RES target 70% In the COM policy as well as in all sensitivity scenarios concerning technology availability and institutional settings only a constraint on CO 2 is set. As outlined, for this default scenario we calculate the cost effective share of renewables. This implies that the RES share is an output of the model, given a certain constraint for CO 2 emission reduction until It certainly varies across the 9

88 different sensitivity scenarios. In addition we investigate variants of the COMM policy scenario in which an additional RES target is set for 2030 with a share of 50, 55, 5 60, 65 and 75%, respectively (see the bottom of Table 1). In these model runs, a constraint enforces the RES share to remain at least at this level from the year 2030 onwards. It is clear that a RES target in addition to the CO 2 target will always lead to additional costss in the setting of LIMES EU, as it is an intertemporal optimization model endowed with perfect foresight. The aim of this analysis is to quantify exactly these costs. It should also be clearly stated here that in this modell setting, wee can neitherr explore potential co benefits of renewable energyy sources nor can we address a other externalities that appear in the context of RES deployment, for example learning and spillover externalities or externalities at the diffusion site Results We have posed the research question of what is the cost effective RES share in the European electricity sector for the year 2030 that is consistent with a 40% GHG reduction target. The Impact Assessment of the European Commission (2014) indicates this figure to be 49% in the scenario GHG40. With LIMES EU we find that thee COM policy scenario, mapping the GHG40 scenario, results in a RES share of 50% in the year 2030 a very similar outcome. However, Figure 1 shows that the different scenarios considered in our sensitivity analysis lead in factt to a range of cost by the effectivee RES sharess that extends from 43% % to 56%. Interestingly, the 49% target proposed Impact Assessment is in the middle of thiss range. Even the cost effectivshows that even e if only 32% GHG mitigation are to be achieved until 2030 a RES share of well above 40% in the electricity sector is cost effective. share occurring in the reference scenario is within thiss range withh 45%. This Figure 1: Cost effective share of renewables in the European electricity sector in % of total electricity provision and the respectivee technology mix of renewables in the year Red shading indicates the range across scenarios. Source: Model results of LIMES EU. Black line: Share of RES in the electricity sector inn the scenario GHG40 of the t Impact Assessment for the framework byy the Europeann Commission (2014), 49%. 10

89 The model results obtained with LIMES EU indicate that the cost effective share of RES in 2030 is higher than 49% if a nuclear phase out is pursued all over Europe, investmentt costs for wind and solar technologies decrease faster than expected in the default scenario or the CCS technology is not available. It is lower than 49%, however, if electricity demand decreases due too progress in energy efficiency programs, investment costs forr wind and solar technologies decrease slower than expected in the default scenario, only RES sites with average instead of high potential are available or biomass costs are higher than in the default scenario. This shows that setting a RES target that is cost effective as a is claimed by the European Commission is clearly not independent off choices regarding assumptions on the development of low carbon technologies and institutional framework conditions. Particularly decisive in thiss context are the assumptions on a nuclear phase out technologies and fuel costs for biomass. in Europe as well as the future developments of investment costs for wind and solar b The technology mix that emerges for the year 2030 is rather similar across all scenarios. The share of biomasss is lower if fuel costs are high (high cost bio) or the investment costss for wind and solar technologies develop very favorably (low cost vres). Wind onshore plays a veryy important role r in all scenarios, only if high quality wind sites cannot be used (low pot vres) or investment costss develop rather pessimistically (high cost vres) its share is somewhat smaller. In those scenarios in which the share of biomass is lower, the difference is made up for by a higher share of wind. Solar PV is used in all scenarios to a certain extent. However, in the scenarios with thee highest total RES share (nucout 2030, low cost vres, nucout 2050), the share of solar PV is also highest. Concentrated solar power (CSP) does not play an important role in any of the scenarios. Comparatively,, this technology has higher costs of electricity production in the year 2030 and is hence not deployed. Only in the scenario with optimistic investment cost developments (low cost vres), CSP is deployed to an extent that is at least visible in the graph. Only wind offshore plays an even less important role than CSP. This is due to their comparatively high costs of electricity generation, as opposed to the more mature technologies wind onshore and solarr PV. Figure 2: RES shares in the electricity sector over time for the scenarios with additional RES target for (left) and technology mix of renewables in thesee scenarios in the year 2030 (right). Source: Model M results of LIMES EU. In orderr to study the economic costs and system effects of a RES share that is set higher than costto t 70% effective, we consider variants of the COM policy scenario with imposed RES shares of up in 11

90 2030. Figure 2 (left) shows the resulting trajectory of RES shares over o time. Until 2025 they follow the same path. Between 2025 and 2030 as much investments in RES capacitiess as needed to fulfill the target in 2030 are pursued. Thereafter, the RES share stagnates over time inn all scenarios until it meets the cost effective trajectory for achieving the 95% mitigationn target in The right panel of Figure 2 illustrates that thesee additional investments into RES capacities between 2025 and 2030 primarily consist of solar photovoltaic and onshore wind. Also, increasingly, CSP comes into play. Wind offshore is again not deployed in thesee scenarios, due to its prohibitively high investment costs assumed scenario GHG40 of the t Impact Assessment by the European Commission (2014), which we proxy with our default scenario. The amount of hydro and biomass iss unaffected by the prescribed RES share in What are the economic costs if the RES target is set higher than would be cost effectivee from a mitigation point of view (i.e. >50% in 2030)?? Figure 3 provides answers to this question by displaying changes in cost indicators (y axes) depending on the magnitude of the t RES target that is sett in 2030 (x axes). The left panel shows the present value of cumulative discounted total system costs over the period (the objective function of LIMES EU) in bn. The right panel illustrates the difference with respect to the COM policy scenario, indicating the additional system costs over the period induced by a RES target in the year 2030 that is set at higher levels than can be justified by long term mitigation considerations. Thesee costs increase non linearly with the RES the target and reach up to 59 bn for f the 70% target. In order to set these t numbers into context, right axis of the figure also shows the additional costs as a percentage change in total system costs. For a RES target of 70%, this amounts to 1..5%, which seems rather small. However, the number is hard to grasp both due to long time horizon and its system character. Figure 3: Present value of total system costs over thee period for scenarios with different exogenously set RES targets from 2030 onwards, in bn (left). Absolute and relative difference of this number n with respect to the default case without an explicit RES target in 2030 (right).( Source: : Model resultss of LIMES EU. Figure 4: Shadow price of the RES targett in 2030 in /MWh. Source: Model M results of LIMES EU. 12

91 In orderr to better grasp the order of magnitude of the costs of an additional a RES target in 2030, we calculate the shadow price of the RES target. This shadow price indicates how much the total system costs would rise in case the RES target is increased by an additional MWh. M It shows that an additional unit of renewables would cost nearly 100 /MWh in case of the scenario with a RES target of 70% (Figure 4). In contrast to the changes c in total system costs that are increasing non linearly with higher RES targets, the shadow price rises linearly. Note that in case of the 60, 655 and 70% scenarios, the respective RES constraint leads to such high deployment of RES capacities c that it renderss the CO 2 emission pathway as a non binding constraint. Hence, the shadow price p of CO 2 emissions decreases to zero in this case ( cf. Figure 7 in the Appendix). Figure 5: Discounted costs ( 2050) in relation to the default scenario on the left (right) panel. Source: Model results of LIMES EU. Having established that a RES target higherr than the cost efficient t level derived from a long term mitigation target of 95% by the year 2050 leads to increases in total system costs raises the question of what kind of costs such a RES target entails. In LIMES EU, total system costs consist of investment costs, operation and maintenance costs and fuel costs. The left panel of Figure 5 displays the relation between each of these individual components for the scenarios with additionall RES targetss and the COM policy scenario for the period Here it becomes evident that the largest proportion of costs is investment costs which are up to 70% higher in the res scenario. The right panel of Figure 5 reveals that these additional investments amount to justt 15% above e the default scenario when considering the long term perspective until Reconsidering thee trajectory of RES deployment over time (left panel of Figure 2) it becomes evident that the additional investments for the fast deployment of RES capacities between 2025 and 2030 leadd to a stagnation of investments afterwards. It is important to acknowledge that investment costs are of a different nature than t fuel costs, as they can possibly stimulate economic development (Creutzig et al., 2014), e.g. due to a multiplier effect on the economy. In contrast fuel costs constitute a cash outflow of the economy if purchased from abroad, which iss largely the case for fossil fuels in Europe. Relatively higher shares of RES lead to a substitution of electricityy generationn based on fuel cost f leads to a lower share of gas power plants; cf. Figure 7 in the Appendix. Hence, the fast deployment of RES capacities by 2030 leads to fuel cost savings in the years between 2030 and 2050 that lower intensivee technologies and hence to a reduction in fuel costs. In case of LIMES EU, this particularly total system costs for these years. This development can be seen clearly in the right panel off Figure 5: Cumulative discounted fuel costs for the period is 13% lower in the res scenario and 6% lower in the res scenario. The effect of the additional RES targets on operation and maintenance costs is comparably moderate.. 13

92 Overall, the relativee cost increase stemming from substantially higher h investment costs for RES capacities until 2030 and moderately low increasess in operation and maintenance costs is counterbalanced by lower fuel costs c thereafter. Hence, the longer the time perspective over which one evaluates the additional system costs stemming from an additional RES target in 2030, the lower their relative magnitude. COM policy RES target 70% Figure 6: Additional transmission capacities installedd between and 2030 for the COM policy scenario (left) and the scenario RES target 70% (right). Source: Model results of LIMES EU. As LIMES EU considers both electricity generation and transmission technologies endogenously it shows that the additional RES target t also has an influence on the pan European transmission grid expansion. The left panel of Figure 6 shows the additional transmission capacities installed between 2011 and 2030 in the COM policy scenario. Particularly in central Western Europe (between France, Germany, Belgium, Luxemburg) the net transfer capacities are strengthened. Also, the connections to southern Europe and between Great Britain and Ireland as well as a between the Baltic and Nordic states are reinforced. When comparing the transmission grid expansion of the COM policy scenario with that of the most ambitiouss RES scenario res (right panel of Figure e 6), it becomes clear that for reaching this target until 2030 even more transfer capacities have too be built. For some country connections the capacities have to be larger. Also, additional country connectionss overall. This is especially the need to be increased, contributed to a more connected pan European system case for south Eastern Europe and its connection to the Northern states. However, the precise regional allocation of RES capacities and transmissionn capacities crucially hinges on the relative development of specific investment costs off wind and solar technologies (Schmid and Knopf, 2014). The analysis of transmission capacities reveals that the possibility too attain additional RES targets by 2030 requires a dedicated expansion of transmissionn capacities. Therefore it would be highly sensiblee to combine the RES target formulation in 2030 with an explicit infrastructure package. 4. Conclusions While it is non controversial to set further GHG reduction targets within w Europe for 2030, it is less clear whether a binding renewable target will finally be part of the overall o framework of climate and energy policy for The European Commission has proposed ann EU wide target with a share of renewables of 27%, what they call a cost effective target in that sense that it is in line with the ambition of 40% GHG reduction by 2030 that is also part of the proposal. This implies thatt the RES 14

93 target of 27% does not constitute an additional endeavor. In the political debate, however, some stakeholders argue for higher RES targets by referring to potential co benefits that renewables create, such as employment effects, local value added, additional environmental benefits and aspects of industrial policy. But the costs for these additional targets are often not quantified. Economists often claim that additional targets and additional policy instruments come at large additional costs. Our added value in the debate is that we quantify the additional costs for an additional RES target in the electricity sector. Over the time horizon they amount to 0 59 bn for the 50% respectively the 70% target. This figure can be compared to the overall additional costs of 72 bn for the cost effective target in the COM policy scenario relative to the reference scenario with only limited ambition of mitigation. It should be noted, however, that the largest part of the additional costs are investments costs for RES capacities in the period before the year These are de facto not pure costs, as they can have a multiplier effect on the overall economy. In principle, these numbers can now be compared to the additional benefits in terms of avoided air pollution, jobs etc. that such a target could have over the period Our analysis provides one part of the equation that is required to answer the question of whether an additional RES target might be justified in order to reap some co benefits associated with renewables beyond the climate externality. The costs for an additional RES targets deem relatively low; however, it is important to note that the additional system costs increase non linearly with the ambition of the RES target. In this sense one can conclude that an additional RES target that is only slightly above the costeffective share does not come at large additional costs, while a target with a strong ambition for RES deployment, as for example the 70% scenario, has notable implications on the overall system costs. This is particularly the case for the period , when substantial funding needs to be provided for deploying the required RES generation capacities. In addition, we show that it is indeed nearly impossible for the European Commission to set a costeffective RES target, as the cost effective RES share comes at considerable uncertainties. A particularly driving assumption is the assumption about the future development of nuclear power in Europe. The RES share in the electricity sector differs by 6 percentage points when either nuclear phase out by 2030 or a nearly constant level of nuclear power is assumed. Also factors such as the future development of bioenergy, here represented by different assumptions on the costs for bioenergy, have a strong impact and pessimistic assumptions reduce the cost effective share by 7 percentage points. Also, the required grid expansion across Europe is affected by an eventual RES target: more pan European capacities are required, the higher the RES target in This analysis cannot answer the question of whether it is reasonable to set an additional renewables target or not. However, our results emphasize that due to large uncertainties about future price developments, institutional settings and the availability of technologies such as nuclear or CCS, it is nearly impossible to set such an additional target cost effectively. On the other hand, a target that is only slightly higher than the cost effective target would not harm to the economy, as additional costs are moderate if the target is not too ambitious. The whole analysis comes with the caveat that it is only for the aggregated EU level. However, a number of analysis have shown that especially the regional distribution is important and also the aspect of a fair effort sharing within the EU (Boratyńsk et al.(2014), Enerdata (2014)). This aspect is beyond the scope of this paper but is of great importance for the question of a European framework for

94 References Boratyńsk, J., Zachłod Jelec, M., Antoszewsk, M., Wójtowic, K., Economic effects of the proposed 2030 climate and energy policy framework on Poland and other EU regions Results based on the PLACE global CGE model. Center for Climate Policy Analysis, Warsaw. Capros, P., Paroussos, L., Fragkos, P., Tsani, S., Boitier, B., Wagner, F., Busch, S., Resch, G., Blesl, M., Bollen, J., Description of models and scenarios used to assess European decarbonisation pathways. Energy Strategy Rev. 2, doi: /j.esr Creutzig, F., Goldschmidt, J.G., Lehmann, P., Schmid, E., von Blücher, F., Breyer, C., Fernandez, B., Jakob, M., Knopf, B., Lohrey, S., Susca, T., Wiegandt, K., Catching two European birds with one renewable stone: Mitigating climate change and Eurozone crisis by an energy transition. Renew. Sustain. Energy Rev. forthcoming. Denholm, P., Hand, M., Jackson, M., Ong, S., Land Use Requirements of Modern Wind Power Plants in the United States (No. NREL/TP 6A ). NREL National Renewable Energy Laboratory. E3MGLab, PRIMES Model Presentation for Peer Review. ECMWF, ERA Interim Reanalysis Data European Centre for Medium Range Weather Forecasts. Edenhofer, O., Hirth, L., Knopf, B., Pahle, M., Schlömer, S., Schmid, E., Ueckerdt, F., On the economics of renewable energy sources. Energy Econ. 40, S12 S23. doi: /j.eneco EEA, How much bioenergy can Europe produce without harming the environment? (No. 7). European Environment Agency, Copenhagen. Enerdata, Costs and Benefits to EU Member States of 2030 Climate and Energy Targets. ENTSO E, 2013a. NTC Values Summer 2010, final version (6 July 2010). European Network of Transmission System Operators for Electricity. ENTSO E, 2013b. Consumption Data Hourly Load Values. European Network of Transmission System Operators for Electricity. European Commission, 2013a. Green paper: A 2030 framework for climate and energy policies (No. COM(2013) 169 final). Brussels. European Commission, 2013b. National Renewable Energy Action Plans [WWW Document]. URL (accessed ). European Commission, 2013c. EU Energy, transport and GHG emissions. Trends to 2050: Reference scenario European Commission, Impact Assessment accompanying the document Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions A policy framework for climate and energy in the period from 2020 up to 2030 (No. SWD/2014/015 final). Brussels. EUROSTAT, 2013a. Infrastructure electricity annual data (nrg_113a). EUROSTAT, 2013b. Supply, transformation, consumption renewables (hydro, wind, photovoltaic) annual data (nrg_1072a). EUROSTAT, 2013c. Final energy consumption of electricity (ten00097). EUROSTAT, 2014a. Anteil erneuerbarer Energien am Bruttoendenergieverbrauch [WWW Document]. URL 20_31&plugin=0 (accessed ). EUROSTAT, 2014b. Electricity generated from renewable sources [WWW Document]. URL cc330&plugin=1 (accessed ). 16

95 EUROSTAT, 2014c. Gross inland energy consumption, by fuel [WWW Document]. URL sdcc320&language=en (accessed ). FAO, Land Resource Statistics Food and Agriculture Organization of the United Nations. Haller, M., Ludig, S., Bauer, N., Decarbonization scenarios for the EU and MENA power system: Considering spatial distribution and short term dynamics of renewable generation. Energy Policy 47, doi: /j.enpol Held, A., Modelling the future development of renewable energy technologies in the European electricity sector using agent based simulation. Karlsruhe Insitut für Technologie. IEA, Potential for Building Integrated Photovoltaics. International Energy Agency. IEA, Energy balances of non OECD countries International Energy Agency, Paris. Knopf, B., Bakken, B., Carrara, S., Kanudia, A., Keppo, I., Koljonen, T., Mima, S., Schmid, E., Van Vuuren, D.P., TRANSFORMING THE EUROPEAN ENERGY SYSTEM: MEMBER STATES PROSPECTS WITHIN THE EU FRAMEWORK. Clim. Change Econ. 04, doi: /s Lehmann, P., Gawel, E., Why should support schemes for renewable electricity complement the EU emissions trading scheme? Energy Policy 52, doi: /j.enpol Nahmmacher, P., Schmid, E., Hirth, L., Knopf, B., 2014a. Carpe diem: A novel approach to select representative days for long term power system models with high shares of renewable energy sources. Prep. Nahmmacher, P., Schmid, E., Knopf, B., 2014b. LIMES EU: A long term electricity system model for Europe Technical Documentation. Prep. NGDC, ETOPO1 Global Relief Model. United States National Geophysical Data Center. Ong, S., Campbell, C., Denholm, P., Margolis, R., Heath, G., Land Use Requirements for Solar Power Plants in the United States (No. NREL/TP 6A ). NREL National Renewable Energy Laboratory. Platts, UDI World Electric Power Plants Data Base (September 2011). Schmid, E., Knopf, B., Quantifying the Long Term Economic Benefits of European Electricity System Integration. FEEM Work. Pap. Ser Note di lavoro. Schmid, E., Knopf, B., Bauer, N., REMIND D: A Hybrid Energy Economy Model of Germany. FEEM Work. Pap. Ser Schröder, A., Kunz, F., Meiss, J., Mendelevitch, R., von Hirschhausen, C., Current and Prospective Costs of Electrictiy Generation until Deutsches Institut für Wirtschaftsforschung, Berlin. Trieb, F., Schillings, C., O Sullivan, M., Pregger, T., Hoyer Klick, C., Global Potential of Concentrating Solar Power. German Aerospace Center. VLIZ, Exclusive Economic Zone Boundaries (World EEZ v7). Ward, J.H. jr., Hierarchical Grouping to Optimize an Objective Function. J. Am. Stat. Assoc. 58. World Nuclear Association, World Nuclear Power Reactors & Uranium Requirements [WWW Document]. URL nuclear.org/info/reactors.html (accessed ). 17

96 Appendix Figure 7: CO2 price (top) and overall technology mix in the different scenarios (bottom) for the year Source: Model results of LIMES EU. 18

97 Table 2. Installed thermal power generation and storage capacities in 2010 (in GW). Source: Platts (2011) Model region Nuclear Hard Coal Lignite Natural Gas CC Natural Gas GT Storage AT BE BG CZ DE DK EE ES FI FR GB GR HR HU IE IT LT LU LV NL PL PT RO SE SI SK Balkan CH NO

98 Table 3. Installed RES power generation and storage capacities in 2010 (in GW). Source: Platts (2011) and EUROSTAT (2013a). Model region Hydro Biomass Wind Onshore Wind Offshore Solar PV CSP AT BE BG CZ DE DK EE ES FI FR GB GR HR HU IE IT LT LU LV NL PL PT RO SE SI SK Balkan CH NO

99 Table 4. Potential for biomass and hydro power deployment per region. Source: EEA (2006), FAO (2013), European Commission (2013b), EUROSTAT (2013a), EUROSTAT (2013b) and own assumptions. Biomass Annual primary energy potential in PJ Installable capacity in GW Hydro Annual capacity factor in % AT BE BG CZ DE DK EE ES FI FR GB GR HR HU IE IT LT LU LV NL PL PT RO SE SI SK Balkan CH NO

100 Table 5. Parameters used for each technology to calculate their capacity potential. Source: FAO (2013), Held (2010), IEA (2002) VLIZ (2012), NGDC (2013), Trieb et al. (2009), Denholm et al. (2009), Ong et al. (2013) and own assumptions. Wind Onshore Suitable areas Agricultural areas Forest areas Share of suitable areas available for RES 30% 5% Maximum capacity density (MW/km 2 ) on available area Wind Offshore Marine areas max. depth: 50m max. distance to shore: 55km 50% 6 within exclusive economic zone Solar PV Agricultural areas Roof tops & facades 2% 50% (12m 2 /capita) CSP Agricultural areas 2% 10 4 Table 6. Capacity potential for variable renewables (vres) per region in GW. Source: FAO (2013), Held (2010), IEA (2002) VLIZ (2012), NGDC (2013), Trieb et al. (2009), Denholm et al. (2009), Ong et al. (2013) and own assumptions. Model region Wind Onshore Wind Offshore Solar PV CSP AT BE BG CZ DE DK EE ES FI FR GB GR HR HU IE IT LT LU LV NL PL PT RO SE SI SK Balkan CH NO

101 Table 7. Average annual capacity factors for variable renewable (vres) technologies per region and resource grade in percent. Source: ECMWF (2012) and own assumptions. Model region Wind Onshore Wind Offshore Solar PV CSP 1st 2nd 3rd 1st 2nd 3rd 1st 2nd 3rd 1st 2nd 3rd AT BE BG CZ DE DK EE ES FI FR GB GR HR HU IE IT LT LU LV NL PL PT RO SE SI SK Balkan CH NO

102 Table 8. Investment costs for vres technologies in /kw. Source: European Commission (2014) Solar Wind PV CSP onshore offshore Table 9. Techno economic characteristics of thermal technologies. Source: Schröder et al. (2013), European Commission (2014), Haller et al. (2012), Schmid et al. (2012) and own assumptions. Technology Investment cost Efficiency new (old) Annual availability Fixed O&M cost Variable O&M cost Lifetime /kw % % % of inv. cost /MWh years Nuclear Hard Coal (37.4) Hard Coal CCS Lignite (36.6) Lignite CCS Gas CC Gas CC CCS Gas GT Hydro see Table Biomass Intraday storage Interday storage Table 10. Fuel costs in /GJ. Source: European Commission (2014) and own assumptions. Year Hard Coal Lignite Natural Gas Uranium Biomass

103 Table 11. Annual electricity demand in TWh. Source: European Commission (2014), IEA (2012) and own assumptions. Model region AT BE BG CZ DE DK EE ES FI FR GB GR HR HU IE IT LT LU LV NL PL PT RO SE SI SK Balkan CH NO Table 12. New nuclear capacities under construction or planned (in GW) per model region. Source: World Nuclear Association (2013). Model region [GW] BG 1.90 CZ 2.40 FI 1.70 FR 3.44 GB 3.80 LT 1.35 PL 6.00 RO 1.31 SK

104 Carpe diem: A novel approach to select representative days for long-term power system models with high shares of renewable energy sources 1 Paul Nahmmacher # *, Eva Schmid #, Lion Hirth #~, Brigitte Knopf # # Potsdam Institute for Climate Impact Research ~ Vattenfall Germany GmbH * Corresponding author: paulnah@pik-potsdam.de; P.O. Box , D Potsdam WORKING PAPER (June, 2014) Decarbonizing today's power systems is essential for reaching the reduction of greenhouse gas emissions that is needed for climate change mitigation. The use of renewable energy sources (RES), and particularly the variable ones wind and solar (vres), is likely to play a key role in this context. In order to explore scenarios on the future of power systems, a variety of numerical models have been developed. Due to computational restrictions many long-term power system models refrain from an hourly resolution but use time-slices that aggregate similar load and renewable electricity generation levels instead. As the share of vres is projected to increase significantly in future electricity systems, suitably accounting for their temporal and spatial variability becomes ever more important for developing sound long-term scenarios. Recent model applications use a variety of approaches; however, to date there is no reproducible and validated method to derive and select time-slices for power system models with multiple fluctuating time-series. In this paper, we present a novel and computationally efficient method that is easily applicable to input data for all kinds of power system models. We further apply the procedure on input data for the long-term power system model LIMES-EU and show that already a small number of representative days developed in this way are sufficient to reflect the characteristic fluctuations of the input data. Next to a validation of our approach, we analyze under what conditions a seasonal differentiation and the use of representative weeks instead of days is necessary in addition. 1 The esea h leadi g to these esults has pa tl e ei ed fu di g f o the Eu opea U io s e e th F a e o k Programme under grant agreement n

105 Contents 1. Introduction Overview of current time-slice approaches Requirements for a new time-slice approach A novel time-slice approach The clustering algorithm method Error metrics for measuring accuracy Application of the approach on electricity demand and vres time-series Time-series of electricity demand and vres infeed Deriving time-slices from the time-series Accuracy of the approximation Application to the electricity sector model LIMES-EU Model overview Appropriate number of representative days Analysis of resulting clusters with regard to seasons Using representative days instead of representative weeks Conclusion References

106 1. Introduction Decarbonizing today's power systems is essential for reaching the reduction of greenhouse gas emissions that is needed for climate change mitigation and the use of renewable energy sources (RES) is likely to play a key role in this context (EC, 2011; IPCC, 2011). In order to explore scenarios on the future of power systems, a variety of numerical models has been developed. These models study the technical and economic particularities of future electricity systems on a regional, national or international scale. Long-term models with endogenous investments are computationally demanding, especially when optimizing inter-temporally. Therefore, they usually require a reduction of complexity with regard to their temporal, geographical and technical resolution. A common way to reduce temporal resolution is to aggregate time-series data to a few characteristic time-slices, instead of modeling every hour of the day, and to include only a limited number of representative days per year in the model. This is usually done for aggregate model regions and representative technology classes. While for decisions on the latter political borders and engineering logic provide clear guidelines, the situation is much less obvious for the reduction of temporal complexity. How many and which days should be selected from the past data and how to decide whether the selection is appropriate? To date, the selection of days has mainly been based on heuristics and is rarely documented in detail in the description of power system models. For historical reasons, most established approaches are solely focused on reducing temporal complexity along the lines of demand side fluctuations, e.g. day vs. night or weekday vs. weekend. However, with variable renewable energy sources (vres) gaining importance in many electricity systems already and projected to increase significantly in the future, approaches that are able to cover multiple time-series simultaneously become necessary in order to describe the fluctuations of vres sufficiently. That is, the variability of the time-series describing solar and wind infeed needs to be considered equally next to the time-series of electricity demand. A structural underrepresentation of demand or vres variability would result in distorted model results and biased scenarios. Therefore, for developing sound long-term scenarios it is important to reduce complexity without losing the relevant information inherent to the time-series data across model regions. In this paper we present and apply a novel approach for deriving and selecting representative days from multiple fluctuating time-series. Our approach is easily reproducible in order to be applied on input data for other power system models. It is especially useful for those with multiple fluctuating time-series, i.e. models with multiple vres technologies and/or multiple regions. Section 1.1 presents a brief literature overview of approaches that have been used to date and identifies their shortcomings. Section 1.2 gives a first introduction to the approach and outlines the structure of this paper. 3

107 1.1. Overview of current time-slice approaches Before vres were introduced to power systems, the sole driver of variable load has been the structural variability of demand. Hence the traditional way for developing time-slices is to discern different seasonal and diurnal load levels, and additionally accounting for different load levels between working days and week-ends. Tab. 1 gives an overview of models following this tradition and applied in recent studies. The models vary considerably in the number of time-slices. Kannan and Turton (2012) discuss the variations in their model results when applying three different representative days per season with hourly resolution versus only having two time-slices per day and one representative day per season. They find that the diurnal time-slice aggregation and the removal of the intra-week differentiations substantially reduce the variation between the highest and lowest demand and thus lead to an overestimation of the value of base load plants. The comparison by Nicolosi (2011) between 8760, 288 and 16 time-slices lead to similar findings. Nicolosi (2011) adds that in a scenario with low levels of wind power capacities the adding of a super-peak time-slice like in Short et al. (2011) might avoid the shortcomings of a highly aggregated model resolution. Considering scenarios with a high share of wind or other vres technologies both Kannan and Turton (2012) and Nicolosi (2011) emphasize the importance of representing the fluctuations of these technologies i the odel s ti e-slices. Accordingly, a number of alternative approaches were developed in recent years that go beyond the demand-based approach of the studies mentioned in Tab. 1 to better account for the fluctuations of vres. Tab. 1: Models with time-slices that are mainly based on demand-side patterns and use averaged vres-infeed with seasonal and diurnal fluctuations No. of No. of No. of No. of days Model Region Applied in seasons time-slices time-slices per season per year per day Fürsch et al. DIMENSION Europe (2011) TIMES TIMES TIMES ReEDS Azores (Portugal) Switzerland Europe USA Pina et al. (2011) Kannan and Turton (2012) Blesl et al. (2012) Short et al. (2011) THEA Texas (USA) Nicolosi (2011) * +1 denotes the addition of a peak time-slice *

108 GEMS +CEEM Germany DENA (2005) seasons, 3 demand days and 3 wind infeed days per season, 12 time-slices per day Tab. 2: Models with time-slice approaches that specifically deal with more complex vres fluctuation patterns No. Model Region Applied in of Data Time-slice specifications timeslices basis DIMENSION +INTRES Europe Golling (2012) seasons, 8 combinations of low/high wind days over all regions, 12 time-slices per day DIMENSION Europe Nagl et al. (2012) simulated weather years with 30 days (2 seasons) each, hourly resolution US-REGEN USA Blanford and Niemeyer (2011) randomly selected weighted combinations of load and wind infeed 2007 LIMES-EU + Europe +MENA Haller et al. (2012) 49 4 seasons, 3 vres situations, 4 time-slices per day (plus one peak time-slice) 2009 URBS-EU Europe Schaber et al. (2012) years with 6 representative weeks each, hourly resolution unit commitment model Texas (USA) Sisternes and Webster (2013) weeks (each 7 days) with hourly resolution (plus one peak day) 2009 Tab. 2 presents some studies that specifically aim at representing the fluctuation patterns of vres in thei odels ti e-slices. As each of the mentioned studies applies a different method to generate their respective time-slices, the respective approaches are briefly described in the following. DENA (2005) combines the usual differentiations between seasons, between working days and weekends and between twelve diurnal time-slices with three different day patterns of wind load ep ese ti g da s ith lo, high a d ediu i d po e i feed. As DENA s odel fo uses o Germany, this differentiation is solely applied on wind power infeed in Germany. Blanford and Niemeyer (2011) do not make use of the traditional time-slice differentiation between seasons and diurnal load levels at all. Instead, they randomly pick 50 out of the 8760 combinations of demand and wind load levels, i.e. residual load levels, observed in 2007 and weight them to best fit the residual load demand curve of Golling (2012) develops a time-slice approach to capture the correlation between wind levels of different regions for a multi-regional model. After having clustered the model regions to a smaller number of wind regions Golling (2012) explores the frequency of occurrence of simultaneous combinations of different wind speeds among the regions. The day patterns with the highest frequencies of occurrence serve as input for the model. 5

109 Nagl et al. (2012) do not only account for different days of wind speed, but also for days with different levels of solar radiation within a season. They randomly pick a set of 30 days 2 with hourly resolution from a multi-regional data set of wind speeds and solar radiation in Europe for the years and epeat this 2000 ti es. Based o the load le els i p e iousl defi ed i po ta t i d a d sola regions they chose ten sets out of the 2000 that together represent five different levels of wind and five different levels of solar radiation in Europe. The high number of resulting time-slices is acceptable as the model in Nagl et al. (2012) is only applied on a single step greenfield investment decision for Schaber et al. (2012) make use of an even higher number of time-slices but apply their multi-regional model only to optimise the investments in transmission capacities while the generation capacities are given exogenously. The time-slices are composed of 48 representative weeks from an eight-year data set 3 with hourly resolution, thus amounting to a total number of 8064 time-slices. For their single region unit commitment model, Sisternes and Webster (2013) also use representative weeks with hourly resolution. The weeks are selected out of a one-year dataset in order to best fit the residual load 4 duration curve. Using the residual load allows for having only one time-series to be approximated but is not appropriate for investment models with endogenous vres capacities. Haller et al. (2012) apply a much stronger aggregation scheme for their inter-temporal investment optimisation model. They allow for only four time-slices per day and three different wind days per season. Like Nagl et al. (2012) they do not account for the possibility that the same overall wind load level can be realized by different combinations of regional wind levels in their multi-regional model. The variety of different time-slice approaches in recent model applications shows that the research community has approached the problem of how to adequately represent the variability of RES infeed and electricity demand in their models from different angles. However, all of the presented approaches are subject to certain shortcomings. They are either based on only one RES time-series (Blanford and Niemeyer, 2011; DENA, 2005; Golling, 2012; Haller et al., 2012), are focused on only one region or disregard different spatial compositions of load levels (Blanford and Niemeyer, 2011; DENA, 2005; Haller et al., 2012; Sisternes and Webster, 2013) or they lead to an amount of time-slices that is not suitable for long-term inter-temporal investment models (Nagl et al., 2012; Schaber et al., 2012; Sisternes and Webster, 2013). Most importantly, only Sisternes and Webster (2013) provide a distinct validation of their approach winter days and 12 summer days 3 six weeks per year 4 residual load is the actual load minus the infeed from vres 6

110 1.2. Requirements for a new time-slice approach In this paper we present a novel approach for deriving time-slices for long-term power system models. It is meant to optimally fulfil the following three essential requirements: The derived time-slices should sufficiently reflect (1) the annual electricity demand and average vres capacity factors for each region, (2) the region specific load duration curves of electricity demand and vres technologies, (3) the spatial correlation of electricity demand and vres electricity infeed among regions. The first requirement ensures that the quality of a region with respect to solar and wind power is correctly reflected. By replicating both common and rare load situations as well as their respective frequency of occurrence (second requirement), the time-slices neither overestimate nor underestimate single events. This serves to correctly value both base and peak load plants. The third requirement ensures that the characteristics of an interconnected multi-regional electricity system are correctly assessed and features such as large-area pooling and anti-correlated infeed are taken into account. Ou app oa h is ai l ased o Wa d s (1963) hierarchical clustering algorithm. We apply this algorithm on historic electricity demand and weather data to group days with similar demand and vres load patterns. Eventually each group of days is reflected by a representative day in the power system model. In this paper the approach is tailored in order to be applied on input data for LIMES-EU, a longterm electricity model for the European electricity system with several model regions and multiple vres and demand time-series per region. We use the approach to identify representative days; however it is also applicable to selecting hours or groups of days. Our method is easily reproducible and applicable to all sorts of input data with multiple fluctuating time-series. The remainder of the paper is structured as follows: Section 2 describes our novel time-slice approach and introduces error metrics in order to assess its accuracy of the approach. In Section 3 we apply the approach on historic European weather and electricity demand data and run a validation exercise based on the error metrics presented in the previous Section. In Section 4 we evaluate how many time-slices are needed for the LIMES-EU model when applying our approach. In addition we discuss selected results both from the time-slice approach and from LIMES-EU to address two central questions when aggregating time-series to time-slices, namely the merits of seasonal differentiation and the use of representative weeks instead of days for electricity system models. Section 5 summarizes and concludes. 7

111 2. A novel time-slice approach We explicitly developed our approach in order to be applied on a discretionary number of time-series simultaneously. Multiple time-series exist when there is more than one model region and/or more than o e t pe of e oge ousl a i g load. I this pape load does ot o l efe to ele t i it de a d ut also to the infeed of vres technologies such as wind and solar. The historic data used as input for our approach thus consist of multiple dimensions, i.e. load values for multiple regions and load types. Based on the hierarchical clustering algorithm described by Ward (1963), the days are grouped in a way that the inner cluster variance respecting all dimensions is kept to a minimum. For each cluster of days, one representative day is chosen that minimizes the sum of squared errors between that day and the other days of its cluster. In this way, the approximation of each load duration curve is optimized while at the same time accounting for the simultaneous load situations of the other time-series. After presenting the approach in Section 2.1, we propose a metric for assessing its accuracy in Section The clustering algorithm method The clustering of historic days is the core of our time-slice approach. The overall approach consists of six consecutive steps (see Fig. 1) that are described in detail in the following. Note that the final number of time-slices to be applied to the model results from the predefined number of time-slices per day multiplied by the number of representative days selected with our approach Normalization of all time-series Applying the hierarchical clustering algorithm Defining a candidate number of clusters (i.e. representative days) Choosing one representative day per cluster Weighting the representative day according to cluster size Scaling single time-series in order to reach annual average Fig. 1 Visualization of the clustering algorithm method for identifying representative time-slices. 8

112 Step 1: Normalization of all time-series The algorithm clusters the historic days based on their respective distance to each other. In order to value the fluctuations of different load series equally, all series are normalized to their maximum value. Step 2: Appling the hierarchical clustering algorithm A clustering algo ith g oups si ila o se atio s i to the sa e luste. i ila it is defi ed ia the cluster method and a distance measure. Hierarchical cluster algorithms start with single observations, which are clusters with just one member. They iteratively group those two clusters that are most similar until only one cluster containing all observation remains. In our case, the observations are multi-dimensional. They contain information about the different coincident loads of every region. In order to cluster similar days, the scope of each observation is one day, i.e. each observation does not only contain one value per load type and region but multiple ones. The number of values per load type and region depends on the diurnal temporal resolution. Accordingly, the resulting number of time-slices for the model is the number of representative days times the number of load situations per day. For a given target number of representative days, the clustering algorithm aims to minimize the deviation between observations and their representative. denotes a vector of load values, with being the index for days (observations) and the index for clusters; is the index of all days within cluster. To minimize the deviation between observations and their representative we apply the hierarchical clustering algorithm described by Ward (1963). The distance between two observations (d 1 and d 2 ) is defined as the Euclidean distance. A central figure i Wa d s algo ith is the e t oid which is the mean vector of the observations grouped into the same cluster ( denotes the number of days grouped into cluster ). 9

113 Wa d s algo ith ite ati el joi s those t o luste s hose o i atio esults i the s allest i ease in the overall sum of squared errors between the observations and their clusters centroids. In other words, it attempts to keep the inner cluster variance to a minimum level. With each iteration the sum of squared errors increases. The increase two observations d 1 and d 2 is caused by the merge of the ( ) ( ) We consequently define the distance overall sum of squared errors when those two clusters are merged. between two clusters based on the resulting increase in the After each merge of two clusters c 1 and c 2, we calculate the distance between the new cluster c 1 +c 2 to an existing cluster c 3 via the Lance-Williams formula (Lance and Williams, 1967; Wishart, 1969), where denotes the number of observations in a cluster. In each iteration the two clusters with the smallest distance are merged and the number of clusters reduces by one. The algorithm is repeated until only one cluster that contains all observations is left. Alternatively one can set a target number of remaining clusters in which case the iteration continues until the required number is reached. Fig. 2 gives a visualization of the clustering algorithm. 10

114 Increase in SSE observations (days) Fig. 2 Dendrogram of the clustering algorithm showing the consecutive grouping of two clusters to a joint cluster and the resulting increase in the overall sum of squared errors (SSE, y-axis). All days (x-axis) are consecutively grouped together until only one cluster is left. Step 3: Defining a candidate number of clusters The algorithm continues to merge observations and clusters until only one cluster is left or until a predefined number of clusters is reached. One may also define a threshold, e.g. for the sum of squared errors, that leads to a stop of the algorithm once it is surpassed. However, in our case the appropriate amount of clusters should not be defined within the algorithm but by testing the model with several numbers of representative days. It is the responsibility of the user to determine the optimal trade-off between model computation time and the accuracy of the model output. Consequently, the following steps have to be made for each candidate set of representative days. Step 4: Choosing one representative day per cluster All historic days that are grouped into the same cluster will be represented by the same representative day in the power system model. In order to minimize the SSE we chose this day as representative day that is losest to the luste s e t oid. The distance between the historic days and the centroid is again defined by the Euclidean distance. Step 5: Weighting the representative days according to cluster size The approach allows to account for typical days that reflect common load patterns for the particular electricity system as well as for extreme days that are only rarely occurring but are nevertheless important for the optimal composition of the electricity system. Extreme days are part of smaller clusters that long remain separated in the clustering algorithm (Step 2) due to their high distance to other clusters. We weight the representative days used for the power system model with a factor 11

115 according to the size of their respective cluster. The size of a cluster is defined by the number of historic days grouped into it ( ). Thereby we account for the fact that large clusters containing many days represent rather common load situations while small clusters represent only rarely occurring load pattern. Approaches that do not assign different weights to their time-slices 5 risk cutting out extreme events and tend to use an unnecessarily high number of time-slices for similar common days. Step 6: Scaling single time-series in order to reach annual average The eighted a e age of the ep ese tati e da s alues 6 may deviate from the average of the underlying historic time-series 7. To ensure correct average demand levels and capacity factors per technology and region the values of each representative day are scaled if necessary. The scaling is done for each dimension (i.e. each time-series) separately. denotes the load situations (e.g. hours) per day. Upscaling is pursued in a way that ensures that the normalized values of a representative day do not surpass 1. In a case that a value is scaled above 1, it is set to 1 and the other values of the time-series are scaled again in order to reach the correct average Error metrics for measuring accuracy To measure the accuracy of the approximation with different amounts of clusters, we use a set of five different error metrics as indicators. Tab. 3 gives an overview about the indicators that are presented in more detail in the following. 5 e.g. Sisternes and Webster (2013) and the established demand based approaches in Tab. 1 6 The eighted a e age of the ep ese tati e da s alues equals. 7 The average of historic time-series values equals. 12

116 Tab. 3 Overview of the indicators for measuring the accuracy of the approximation Indicator Description MSE of the historic values to the value of thei luste s e t oid ( ) ( ) RM E of the histo i alues to the alue of thei luste s e t oid (averaged over all regions and load types) RMSE of the load duration curves consisting of historic values to the load duration curves consisting of the clusters centroid values (averaged over all regions and load types) RMSE of the histo i alues to the alue of thei luste s representative day (averaged over all regions and load types) ( ) RMSE of the load duration curves consisting of historic values to the load duration curves o sisti g of the luste s ep ese tati e da alues (averaged over all regions and load types) Note: MSE Means Square Error; RMSE Root Mean Square Error; LDC Load Duration Curve The first indicator is the mean square error M E ith ega d to the luste s e t oids. It is di e tl de i ed f o the luste algo ith s o je ti e fu tio, the o e all su of s ua ed e o s E, divided by the number of time-series, days and load situations per day. The other four indicators are average root-mean-square errors (RMSE) derived from the deviation between historic days and their corresponding representative values. These corresponding representative values differ between the four indicators. denotes the average root-mean-square error per time-series between historic values and the values of their corresponding cluster s e t oid. ( ) ( ) The second indicator, ( ), also gives an average root-mean-square error between the observed values and their respective centroid values. However, this time the indicator refers to the approximation accuracy of the time-series specific load duration curves. Hence, the RMSE is not calculated for observed values and their centroid s alues but to the centroid value that has the same position in the load duration curve. To construct the load duration curve of the centroid 13

117 values, the occurrence of each centroid value is set according to its cluster s eight. This way, the load duration curve of the centroid values has eventually the same number of values as the load duration curve of the observed values. ( ) ( ) { } { } ( ) By analogy with, denotes the average root-mean-square error per time-series between historic values and the values of their corresponding representative day. ( ) ( ) Parallel to ( ), ( ) denotes the average root-mean-square error per time-series between the di e sio s histo i load du atio u es a d the espe ti e load du atio u es consisting of the representative days. ( ) ( ) 3. Application of the approach on electricity demand and vres time-series We apply our novel time-slice approach on input data for European electricity system model LIMES-EU. However, the application of the time-slice approach is rather independent from the actual characteristics of the power system model. Therefore, in this Section we concentrate on the characteristics of the input data, whereas the description of the model is postponed to the application in Section 4.1. The necessary input data consists of multiple time-series, including electricity demand as well as wind and solar infeed for 29 European regions. After presenting the time-slice specific data in 14

118 Section 3.1, Section 3.2 applies the six steps as detailed in Section 2.1 and Section 3.3 validates the results based on the error metrics given in Section Time-series of electricity demand and vres infeed Our power system model LIMES-EU contains several load series with time(-slice)-dependent values for each model region: Electricity demand and fluctuating infeed from the RES technologies wind onshore, wind offshore, photovoltaic and concentrated solar power. In order to account for the fact that wind speeds and solar radiation do not only vary between but also within the model regions, we consider three load series for each model region and vres technology in LIMES-EU. 8 The input data for the time-slice approach is based on ENTSO-E (2013) data for the historic electricity demand levels and on historic weather data from ECMWF (2012) for the vres infeed. Using weather data rather than historic infeed data allows for taking a longer time span into consideration and prevents the over-estimation of unusual years. The ECMWF data set comprises 33 years of solar radiation, ground temperature and wind speed levels at 120m height for Europe. For every third hour between 1979 and 2011 the respective information is given for local data points in a spatial resolution of 0.75 x0.75. After converting this weather information to time-series of infeed levels of representative i d a d sola po e pla ts, the data poi ts a e agg egated to epli ate the RE i feed i the odel s sub-regions. Compared to the vres infeed, the intra-year demand fluctuations are less stochastic and follow distinct diurnal, intra-week and seasonal patterns. Though the absolute demand levels change among different years due to demographic and economic reasons, the relative intra-year fluctuations are assumed to remain the same. The hourly demand data of 2010 and 2011 that is available from ENTSO-E (2013) for all model regions is therefore deemed to be representative for the intra-year demand side fluctuations between 1979 and Future inter-year growth of annual demand is subject to scenario assumptions. However, as we use normalized values for the clustering, these scenario assumptions do not affect the time-slice approach. 8 The first sub-region covers the 10% best wind sites (or solar sites respectively), the second sub-region covers the following 30% wind (solar) sites and the third sub- egio o p ises the 60% of a odel egio s i d sola potential with the lowest wind speeds (solar radiation). 15

119 3.2. Deriving time-slices from the time-series Based on the data sources presented in the previous Section, we assemble the input data for our timeslice approach. In order to find representative days, the observations used in the clustering algorithm comprise data of one whole day each. Each observation contains data for 29 model regions and 13 load types 9 per model region, resulting in 377 time-series. The time-series are given in three-hourly resolution and the scope of each observation is one day. Hence, each observation comprises eight load situations per time-series, resulting in a total of 3016 values per observation. The number of observations for the clustering algorithm amounts to 12053, comprising the years from 1979 to Before starting the clustering algorithm, each time-series is normalized to its maximum value (cf. Step 1). While solar and wind infeed virtually expand over the whole range between 0 and 1, none of the normalized electricity demand time-series has a situation with less than 30% of maximum demand. Consequently, the range of normalized demand side fluctuations is much smaller compared to the timeseries of wind and solar infeed, which possibly results in its underestimation in the clustering algorithm. However, it turns out that demand side fluctuations are properly covered by our approach. In fact, the ep ese tati e da s load du atio u es of the de a d a d sola ti e-series quickly converge to the original load duration curve. Even one single representative day covers the characteristic fluctuations of these time-series fairly well. This is due to the fact that most fluctuations in demand and solar irradiation are based on differences between day and night and the day-night shift is already covered when clustering whole days instead of single load situations. Fig. 3a-c show the original load duration curves of wind onshore, solar PV and electricity demand in Germany as well as approximations based on different numbers of representative days. The load duration curves of the representative days are constructed in a way that the relative weight of a representative day is accounted for (cf. Step 5). It can be seen that a higher number of representative days leads to a more accurate replication of the original load duration curve. The load duration curve of wind power is hardest to replicate as wind fluctuation occur rather between than within days. In addition to the visual validation of the time-slice approach given by Fig. 3a-c, the following Section provides a numerical discussion of the approximation accuracy. 9 For each region, the observations contain data about the electricity demand and the infeed of four vrestechnologies in three different resource grades per region, resulting in 13 load types per region. 16

120 Fig. 3a-c App o i atio of Ge a s a wind onshore, (b) solar PV and (c) electricity demand load duration curves, displaying original data (blue) and results for one (red), five (green) and ten (purple) clusters Accuracy of the approximation For an exemplary set of different amounts of representative days (clusters), Tab. 4 shows the values for the indicators introduced in Section 2.2. Naturally the mean square error between the o se atio s a d thei espe ti e luste s e t oids de eases ith a i easi g a ou t of luste s. If each observation had its own cluster, the MSE would be zero. For ease of reading Fig. 4 visualizes the values presented in Tab. 4. Tab. 4 Accuracy of the approximation depending on the number of time-slices; for a definition of the error measures see Section 2.2. ( ) ( ) ( ) ( ) 1 cluster clusters clusters clusters clusters clusters clusters

121 Fig. 4 Increasing approximation accuracy with an increasing number of time-slices based on the RMSEs. The use of representative days (solid blue line) yields a better approximation of the load duration curves than the use of the luste s centroids (dashed blue line). and represent the de iatio et ee the o se atio s a d thei luste s centroids. By design of the clustering algorithm, their values are decreasing with an increasing number of clusters. It is obvious that that reflects the deviation between the observation values and the values of their respective representative days is higher than : The sum of distances between all observations of o e luste a d thei luste s centroid is necessary smaller than the sum of distances to another specific observation of the cluster. Ho e e, it is i te esti g to see that the di e sio s load du atio u es a e ette epli ated the load duration curves of the representative da s tha the load du atio u es of the luste s centroids: ( ) is lower than ( ). One explanation could be that the representative days better manage to replicate the extreme values at the upper and lower end of the load duration curve (see Fig. 5). We consequently use real observations as representative days for our model i stead of the luste s e t oids. Fig. 5 Appro i atio of Ger a s i d o shore load duratio ur e, showing original data of wind onshore in Germany (blue) and the results for centroids (red) and representative days (green) based on 10 clusters. 18

122 4. Application to the electricity sector model LIMES-EU We use the data of the representative days generated with our time-slice approach in the previous Section as input for LIMES-EU. After giving a short model overview (Section 4.1), we run a simplified version of the model with different amounts of representative days in order to decide on how many days are needed for appropriately representing the vres fluctuations (Section 4.2). Afterwards, we discuss two interesting characteristics of other time-slice approaches that are not applied in our approach: In Section 4.3 we discuss the usefulness of seasonal differentiation in time-slice approaches, i.e. that representative days are chosen for each season separately; in Section 4.4 we discuss the use of representative groups of consecutive days (e.g. weeks) instead of representative days Model overview LIMES-EU (the Long-term Investment Model for the Electricity System of the European Union) was originally published in the earlier version LIMES-EU + by Haller et al. (2012). It is a linear model developed for determining cost-optimal investment pathways for the European power system. Starting with the installed capacities of 2010 it simultaneously optimizes dispatch and investment decisions for generation, storage and international transmission technologies for every fifth year until This simultaneous, inter-temporal optimization of the investment pathway from 2010 to 2050 is computational demanding. However, it is necessary to explore such scenarios as long-term targets like the CO 2 emission reductions envisioned by the EU Commission until 2050 affect the optimal investments pathways and technology choice in previous years. The optimal intra-annual dispatch is calculated for a limited amount of time-slices. Emissions and costs for producing electricity in these time-slices are scaled up according to a predefined weight of a timeslice in order to add up to the total annual values. The weights may differ among time-slices. In its current version the model comprises 26 of the 28 EU Member States 10 plus Norway, Switzerland and the Balkan region. The capacities of generation and storage technologies are aggregated to national levels, with each country constituting one model region 11. The international transmission grid in LIMES- EU is represented via Net Transfer Capacities (NTCs) between the model regions. 10 The insular states Malta and Cyprus are not included in the current model version 11 with the exception of the non-eu countries in the Balkan region that are grouped to one model region 19

123 For the applications in this paper we set an exogenous CO 2 reduction target of minus 95% until 2050 in comparison to The reduction pathway between 2010 and 2050 is increasing linearly. Other policy constraints are kept to a minimum, e.g. there are no additional policies implemented to foster the electricity production from RES. However, all primary resources except uranium and hard coal are subject to environmental and/or societal constraints. In order to reflect the political situation of nuclear power in Europe, we implement a nuclear phase-out in Germany, Belgium and Switzerland. New installations in other countries are limited to those currently under construction or planned and for replacing depreciated capacities Appropriate number of representative days In order to evaluate the approach and decide on the number of representative days 12 to use in the intertemporal model LIMES-EU, we test different numbers of time-slices in a non-inter-temporal version of the model with the same geographical scope and the same objective function. Similar to the intertemporal version, this model version also minimizes overall system costs given the electricity demand in the model regions. However, the model is solved for one single year only, annualizing the investment costs. To give the model as much freedom as possible, there are no historical assets that can be built on, i.e. the model is solved in a so-called 'greenfield' version optimizing the total power system. Compared to the inter-temporal model, this greenfield equilibrium version of LIMES-EU is less computationally demanding and thus capable to solve with a considerably higher temporal resolution. We compare the outcomes of model runs with different numbers of time-slices ranging from 1 to 100 representative days (i.e. 8 to 800 time-slices). As Fig. 6 shows, the share of vres is up to 16% higher than in model runs with only few time-slices. This is due to the fact that the fluctuations of vres are represented to a much lesser extent in these model runs. This overestimation of the system value of vres eventually leads to an underestimation of the total system cost: The model run with only one representative day results in 13% lower overall cost than the model run with 100 representative days. The more time-slices we use in the model, the closer the results are to the outcome of the run with 800 time-slices. However, this convergence is not monotonic; e.g. the total system cost of the run with 160 time-slices are higher than the cost of the neighboring runs with 120 and 200 time-slices respectively (cf. Fig. 6). This is due to the fact that we do not use average values (i.e. centroids) but values that are based on specific historic days to represent the days grouped into the same cluster. 13 Extreme values such as 12 The number of time-slices within each representative day is set to 8 (each of three hours length) from the outset. 13 cf. Step 4 of the time-slice approach (Section 2.1) as well as Section

124 peak demand or peak wind power infeed are therefore not necessarily more pronounced in a set of more representative days. However, the resulting distortions are only minor and the notion that a higher number of time-slices leads to a more accurate representation of variability and costs undoubtedly holds. Fig. 7 provides more detailed information about the cost-efficient electricity mix found by the different model runs. Nuclear power is virtually non-existent in the model run with one representative day, but increases to a share of 19% in the electricity mix of the model run with 100 representative days (800 time-slices). In contrast, the optimal electricity generation of wind power is 30% higher with one as compared to 100 representative days. This is because the levelized cost of electricity (LCOE) of wind power is lower than the LCOE of nuclear power; it is therefore favored in a model with low temporal resolution and a eak ep ese tatio of the i d po e s a ia ilit. With higher temporal resolution the value of vres decreases, which leads to a higher deployment of the more costly but to a certain extent dispatchable nuclear power. In addition, the share of coal-based CCS power plants decreases and is partially replaced by natural gas power plants, both combined cycle with CCS and ordinary gas turbines. Gas turbines are the most flexible thermal power plants but more CO 2 intensive than power plants with CCS. Therefore, the overall electricity production from fossil fuel power plants decreases and natural gas combined cycle plants with CCS the least CO 2 intensive type of fossil fuel power plants become more important. In short, one representative day is obviously not enough to sufficiently cover the fluctuations of electricity demand and vres electricity production. The most accurate results are obtained when including as many representative days as possible in the model. However, in practice the appropriate number of time-slices has to be found by a trade-off between computation time and accuracy. Based on Fig. 6 and Fig. 7 we find that in our case already 48 time-slices (six representative days) are sufficient to obtain reliable results. The share of vres in the electricity mix differs by just 2.5% between the model run with 100 representative days and the model run with six representative days. The difference in total system costs is 4%. To reach more accuracy the temporal resolution would have to be strongly increased. However, with exponentially increasing computation time we deem the accuracy obtainable with 48 time-slices to be sufficient. 21

125 Fig. 6 Total system cost (i ) and percentage share of variable renewable energy sources (vres) in the electricity generation mix depending on the number of time-slices used in the greenfield version of LIMES-EU. Fig. 7 Electricity generation mix depending on the number of time-slices used in the greenfield version of LIMES-EU Analysis of resulting clusters with regard to seasons Most conventional time-slice approaches derive their time-specific model input data by (criteria-based or randomly) selecting a predefined number of hours, days or weeks from each season of the year. The number of hours, days or weeks to select is equal for each season. In this Section we analyze the clusters generated by our approach with regard to their seasonal composition. In its default version, the approach does not specifically consider seasons when clustering historic days. Based on the input data used in this paper the compulsory differentiation of seasons would lead to a higher number of necessary clusters. When requiring an equal amount of clusters from each season the mean-square error (cf. Section 2.2 and 3.3) is 8% higher for the selection of 2 clusters per season and 7% higher when selecting 3 clusters per season (compared to 8 and 12 chosen clusters without such restrictions). Naturally, the approximation accuracy would even further decrease if the representative days had to be chosen in a way that each day is weighted the same in the model. 22

126 winter dominated clusters spring & autumn dominated clusters summer dominated clusters Fig. 8 Distribution of historic days (sorted by months) to 10 clusters, e.g. 57% of the 990 July days from 1979 to 2011 are grouped to cluster #1 (bright orange). For the case of ten clusters, Fig. 8 shows how the historic days (sorted by months) that are used in this paper are distributed among the clusters obtained with the time-slice approach. It suggests two explanations for why it is not useful to use the same number of representative days from each season when aiming to minimize the number of necessary representative days: (1) Spring and autumn days are fairly similar with regard to our input data, they are grouped into the same clusters; (2) the variance among winter days is higher than the variance among summer days, so less clusters are needed for the latter. Five of the ten clusters are dominated by winter days (blue in Fig. 8); three are dominated by summer days (orange). More than 90% of the days from May to August are covered by these three clusters. Four clusters even cover 99% of all summer days. Spring and autumn together just dominate two clusters (greenish in Fig. 8). Their days are widely spread over all clusters. This suggests that load patterns are more stable in summer and more variable in spring and autumn; however, they do not differ much between those two transition seasons. Time-slice approaches that select the same number of representative days from each season do not account for these patterns and result ceteris paribus in a less accurate representation of the overall load structure (cf. also Sisternes and Webster (2013)). Note that the seasonal approach might be less questionable with input data for other models. For example, the seasonally changing availability of water from glaciers may be important for models with high shares of hydro power. In such a case, time-series for water levels and precipitation could be easily added to the input data for our time-slice approach without any further need for a distinct seasonal differentiation in the presented algorithm. 23

127 4.4. Using representative days instead of representative weeks The presented approach derives representative days to be used in a power system model. Other recent time-slice approaches derive representative groups of consecutive days to be used in their models. Nagl et al. (2012) derives groups of three days; Sisternes and Webster (2013) and Schaber et al. (2012) even use weeks. The rationale behind this idea is that power systems sometimes have to cope with consecutive days of low wind and solar infeed. Such situations are best covered when groups of consecutive days are modelled. Models without this feature risk to shift high amounts of excess electricity from days with high vres-infeed to days with low vres-infeed, whereas in reality huge storage facilities would be needed to guarantee the electricity supply not only for single days but for complete weeks of low vres supply. In turn, the use of representative groups of consecutive days results in more time-slices that are rather similar to each other. In our time-slice approach such days would be grouped into one cluster. As the overall number of time-slices is limited by computational restrictions, more similar days caused by the consecutive-day-approach translates to less space for other characteristic days that are important to account for when determining the optimal structure of the power system. Ultimately, the decision whether to select single days or representative groups of days depends on the research question and the model in use. Despite the higher computational cost, the advantages of consecutive days should especially be considered for long-term models with a focus on different storage options. Our approach is easily modifiable in order to select groups of representative consecutive hours spanning over more than one day. For the case of LIMES-EU we favor computation time and an optimal representation of vres variability over the accurate representation of inter-day storage. This is justified by the fact that inter-day storage does not play a significant role in most model scenarios. 14 Fig. 9 shows a typical generation pattern retrieved from the inter-temporal version of LIMES-EU aggregated over all model regions for Of the two implemented storage options intra-day and inter-day only intra-day storage is being installed. The intra-day storage is used to shift electricity supply between time-slices of the same day. It is more efficient and less expensive than inter-day storage which is able to level out excess and deficit of power supply among different model days. Instead of inter-day storage, conventional generation capacities are used to secure sufficient power supply during days of low vres infeed. Hence, each day is self-sufficient and does not rely on a major shift of electricity from high vres days. 14 This is also found in a detailed storage study by Fraunhofer IWES et al. (2014) 24

128 Fig. 9 Electricity generation mix and demand for 48 time-slices in the year 2050, aggregated over Europe. 5. Conclusion We presented a novel approach to derive and select representative days as input data for power system models. As is the case for other time-slice approaches, the intended aim of our method is to keep computational requirements for the respective model to a minimum by decreasing temporal resolution while still ensuring reliable results. At the core of our approach is the hierarchical clustering method described by Ward (1963). Whereas most other approaches are bound to account for one single timeseries only (e.g. electricity demand in one region), we are able to apply our method on multidimensional input data. Thereby we can simultaneously optimize the replication of variability of multiple load series, e.g. electricity demand data and multiple vres electricity production data for multiple model regions. This is especially important considering that electricity production from vres is expected to play a dominant role in future power systems designed to be in line with ambitious climate mitigation efforts. By its design our approach is easily transferable to be applied on input data for a wide range of different kinds of power system models. It is neither restricted to certain technologies or geographical regions nor to finding representative days instead of representative hours or representative groups of days. 25

Phasing out nuclear power in Europe Rolf Golombek May 2015

Phasing out nuclear power in Europe Rolf Golombek May 2015 Oslo Centre of Research on Environmentally friendly Energy Phasing out nuclear power in Europe Rolf Golombek May 2015 Nuclear power in Europe Mixed picture after Fukushima 2011 Phasing out nuclear vs.

More information

Phasing out nuclear power in Europe Rolf Golombek, Finn Roar Aune and Hilde Hallre Le Tissier 39th IAEE International Conference Bergen, June 2016

Phasing out nuclear power in Europe Rolf Golombek, Finn Roar Aune and Hilde Hallre Le Tissier 39th IAEE International Conference Bergen, June 2016 Oslo Centre of Research on Environmentally friendly Energy Phasing out nuclear power in Europe Rolf Golombek, Finn Roar Aune and Hilde Hallre Le Tissier 39th IAEE International Conference Bergen, June

More information

Long-term Market Analysis Nordics and Europe Executive summary

Long-term Market Analysis Nordics and Europe Executive summary Long-term Market Analysis Nordics and Europe 2018 2040 Executive summary 1 December 2018 Europe is set to develop a low carbon power system with significant contribution from renewable energy sources.

More information

Originally published as:

Originally published as: Originally published as: Knopf, B., Nahmmacher, P., Schmid, E. (2015): The European renewable energy target for 2030 - An impact assessment of the electricity sector. - Energy Policy, 85, 50-60 DOI: 10.1016/j.enpol.2015.05.010

More information

Background paper. Electricity production from wind and solar photovoltaic power in the EU

Background paper. Electricity production from wind and solar photovoltaic power in the EU Background paper Electricity production from wind and solar photovoltaic power in the EU February 2018 1 The 2009 Lisbon Treaty gave the European Union (EU) the authority to develop an energy policy containing

More information

Energy, Electricity and Nuclear Power Estimates for the Period up to 2050

Energy, Electricity and Nuclear Power Estimates for the Period up to 2050 REFERENCE DATA SERIES No. 1 2018 Edition Energy, Electricity and Nuclear Power Estimates for the Period up to 2050 @ ENERGY, ELECTRICITY AND NUCLEAR POWER ESTIMATES FOR THE PERIOD UP TO 2050 REFERENCE

More information

Competitive energy landscape in Europe

Competitive energy landscape in Europe President of Energy Sector, South West Europe, Siemens Competitive energy landscape in Europe Brussels, siemens.com/answers Agenda Europe s competitiveness depends on an affordable and reliable energy

More information

WIND POWER TARGETS FOR EUROPE: 75,000 MW by 2010

WIND POWER TARGETS FOR EUROPE: 75,000 MW by 2010 About EWEA EWEA is the voice of the wind industry actively promoting the utilisation of wind power in Europe and worldwide. EWEA members from over 4 countries include 2 companies, organisations, and research

More information

EU wide energy scenarios until 2050 generated with the TIMES model

EU wide energy scenarios until 2050 generated with the TIMES model EU wide energy scenarios until 2050 generated with the TIMES model Rainer Friedrich, Markus Blesl Institut für Energiewirtschaft und Rationelle Energieanwendung, Universität Stuttgart EMEP CLRTAP TFIAM

More information

Methodology for calculating subsidies to renewables

Methodology for calculating subsidies to renewables 1 Introduction Each of the World Energy Outlook scenarios envisages growth in the use of renewable energy sources over the Outlook period. World Energy Outlook 2012 includes estimates of the subsidies

More information

Wind energy in Europe markets

Wind energy in Europe markets Wind energy in Europe markets Turkish Wind Energy Congress (TWEC 2012), 7 November 2012, Istanbul Christian Kjaer CEO European Wind Energy Association (EWEA) More than 600 members from almost 60 countries

More information

Photo: Thinkstock. Wind in power 2010 European statistics. February The European Wind energy association

Photo: Thinkstock. Wind in power 2010 European statistics. February The European Wind energy association Photo: Thinkstock Wind in power 21 European statistics February 211 1 WIND IN POWER: 21 EUROPEAN STATISTICS Contents Executive summary 21 annual installations Wind map 21 Wind power capacity installations

More information

Core projects and scientific studies as background for the NREAPs. 9th Inter-Parliamentary Meeting on Renewable Energy and Energy Efficiency

Core projects and scientific studies as background for the NREAPs. 9th Inter-Parliamentary Meeting on Renewable Energy and Energy Efficiency Core projects and scientific studies as background for the NREAPs 9th Inter-Parliamentary Meeting on Renewable Energy and Energy Efficiency Brussels, 18.11.2009 Mario Ragwitz Fraunhofer Institute Systems

More information

COMMISSION STAFF WORKING PAPER. Impact Assessment. Accompanying the document

COMMISSION STAFF WORKING PAPER. Impact Assessment. Accompanying the document EUROPEAN COMMISSION Brussels, XXX SEC(2011) 1565 Part 2/2 COMMISSION STAFF WORKING PAPER Impact Assessment Accompanying the document COMMUNICATION FROM THE COMMISSION TO THE EUROPEAN PARLIAMENT, THE COUNCIL,

More information

Merit Order Effect of Wind Power Impact on EU 2020 Electricity Prices

Merit Order Effect of Wind Power Impact on EU 2020 Electricity Prices Merit Order Effect of Wind Power Impact on EU 2020 Electricity Prices AEE Wind Energy Convention, Madrid June 15-16 Jesper Munksgaard Ph.D., Senior Consultant 2 Agenda Background Literature review Modeling

More information

Electricity and heat statistics

Electricity and heat statistics Electricity and heat statistics Statistics Explained Data extracted in June 2018. Planned article update: June 2019. Gross electricity production by fuel, GWh, EU-28, 2000-2016Source: Eurostat (nrg105a)

More information

Highlights. Figure 1. World Marketed Energy Consumption by Region,

Highlights. Figure 1. World Marketed Energy Consumption by Region, Highlights World energy consumption is projected to increase by 71 percent from 3 to 23. Fossil fuels continue to supply much of the energy used worldwide, and oil remains the dominant energy source. In

More information

Vattenfall Capital Markets Day 2009

Vattenfall Capital Markets Day 2009 Vattenfall Capital Markets Day 2009 Presentation by: Lars G. Josefsson CEO Amsterdam, 23 September 2009 Contents Vattenfall overview Industry trends Strategic direction 2 Vattenfall overview 3 Vattenfall

More information

Medium and long-term perspectives for PV. Dr. Paolo Frankl Division Head Renewable Energy Division International Energy Agency

Medium and long-term perspectives for PV. Dr. Paolo Frankl Division Head Renewable Energy Division International Energy Agency Medium and long-term perspectives for PV Dr. Paolo Frankl Division Head Renewable Energy Division International Energy Agency Solar Power Summit, Brussels, 7-8 March 2017 Annual additions (GW) Cumulative

More information

WIND ENERGY - THE FACTS PART VI SCENARIOS AND TARGETS

WIND ENERGY - THE FACTS PART VI SCENARIOS AND TARGETS WIND ENERGY - THE FACTS PART VI SCENARIOS AND TARGETS Acknowledgements Part VI was compiled by Arthouros Zervos of the National Technical University of Athens, Greece (www. ntua.gr), and Christian Kjaer

More information

REPORT FROM THE COMMISSION TO THE EUROPEAN PARLIAMENT AND THE COUNCIL

REPORT FROM THE COMMISSION TO THE EUROPEAN PARLIAMENT AND THE COUNCIL EUROPEAN COMMISSION Brussels, 29.7.2016 COM(2016) 464 final REPORT FROM THE COMMISSION TO THE EUROPEAN PARLIAMENT AND THE COUNCIL Progress by Member States in reaching cost-optimal levels of minimum energy

More information

Decarbonizing Europe s power sector by Analyzing the implications of alternative decarbonization pathways

Decarbonizing Europe s power sector by Analyzing the implications of alternative decarbonization pathways Decarbonizing Europe s power sector by 2050 - Analyzing the implications of alternative decarbonization pathways AUTHORS Cosima Jägemann Michaela Fürsch Simeon Hagspiel Stephan Nagl EWI Working Paper,

More information

Medium Term Renewable Energy Market Report Michael Waldron Senior Energy Market Analyst Renewable Energy Division International Energy Agency

Medium Term Renewable Energy Market Report Michael Waldron Senior Energy Market Analyst Renewable Energy Division International Energy Agency Medium Term Renewable Energy Market Report 13 Michael Waldron Senior Energy Market Analyst Renewable Energy Division International Energy Agency OECD/IEA 13 Methodology and Scope OECD/IEA 13 Analysis of

More information

Wind energy and Climate policy Fixing the Emission Trading System

Wind energy and Climate policy Fixing the Emission Trading System Wind energy and Climate policy Fixing the Emission Trading System Rémi Gruet Senior Advisor - Climate Change & Environment European Wind Energy Association 1st February 2012 EWEA Members Across entire

More information

Our strategy in challenging markets

Our strategy in challenging markets Our strategy in challenging markets Øystein Løseth President and CEO Solna/Stockholm, 3 December 2012 Today s focus Vattenfall at a glance Market trends & outlook Political and regulatory outlook Alignment

More information

Photo: Karpov. Wind in power 2009 European statistics. February 2010 THE EUROPEAN WIND ENERGY ASSOCIATION

Photo: Karpov. Wind in power 2009 European statistics. February 2010 THE EUROPEAN WIND ENERGY ASSOCIATION Photo: Karpov Wind in power 29 European statistics February 21 1 WIND IN POWER: 29 EUROPEAN STATISTICS Contents Executive summary 29 annual installations Wind map 29 Wind installations 29 Power capacity

More information

30/10/2013. The Belgian TIMES model

30/10/2013. The Belgian TIMES model 30/10/2013 The Belgian TIMES model 1.1 General overview: history» INTERNATIONAL DIMENSION» Developed by ETSAP implementing agreement of IEA as successor of MARKAL» History of 30 years development» Software

More information

Contribution of Renewables to Energy Security Cédric PHILIBERT Renewable Energy Division

Contribution of Renewables to Energy Security Cédric PHILIBERT Renewable Energy Division Contribution of Renewables to Energy Security Cédric PHILIBERT Renewable Energy Division EUFORES Parliamentary Dinner Debate, Brussels, 9 September, 2014 What Energy Security is about IEA defines energy

More information

Recent trends and projections in EU greenhouse gas emissions

Recent trends and projections in EU greenhouse gas emissions Approximated greenhouse gas emissions Recent trends and projections in EU greenhouse gas emissions Greenhouse gas (GHG) emissions in the European Union (EU) increased by 0.6 % in 2017, according to preliminary

More information

Danish and European plans for wind energy deployment

Danish and European plans for wind energy deployment Chapter 5 Danish and European plans for wind energy deployment By Peter Hjuler Jensen and Søren Knudsen, DTU Wind Energy; Poul Erik Morthorst, DTU Management Engineering Danish and European plans for wind

More information

Nuclear Energy s State of Play European Perspective. Andrei Goicea FORATOM Executive Manager 29 October 2018

Nuclear Energy s State of Play European Perspective. Andrei Goicea FORATOM Executive Manager 29 October 2018 Nuclear Energy s State of Play European Perspective Andrei Goicea FORATOM Executive Manager 29 October 2018 Nuclear Energy s State of Play 2 1. About FORATOM 2. EU Energy Policy 3. Nuclear Energy in EU

More information

Seminar Organised by INFORSE & EUFORES & EREF Brussels, November 9-10, 2004 Presentation by Giulio Volpi, WWF

Seminar Organised by INFORSE & EUFORES & EREF Brussels, November 9-10, 2004 Presentation by Giulio Volpi, WWF Brussels, November 9-1, 24 Bioenergy conversion chains Bioenergyfor Europe status, trends, gaps and future actions Seminar on New and Upcoming EU Policies for Sustainable Energy and Climate Protection

More information

Impact of CCS on the economics of coal-fired power plants. Why investment costs do and efficiency doesn't matter

Impact of CCS on the economics of coal-fired power plants. Why investment costs do and efficiency doesn't matter Impact of CCS on the economics of coal-fired power plants Why investment costs do and efficiency doesn't matter Richard Lohwasser*, Reinhard Madlener Institute for Future Energy Needs and Behavior (FCN)

More information

(How) can the European power market be decarbonized to 2050, without CCS? Lasse Torgersen CenSES årskonferanse Oslo 7/

(How) can the European power market be decarbonized to 2050, without CCS? Lasse Torgersen CenSES årskonferanse Oslo 7/ (How) can the European power market be decarbonized to 25, without CCS? Lasse Torgersen CenSES årskonferanse Oslo 7/12 217 25 low-carbon economy Cost-efficient pathway to reach the 8% target by 25 Reduction

More information

EU Climate and Energy Policy Framework: EU Renewable Energy Policies

EU Climate and Energy Policy Framework: EU Renewable Energy Policies EU Climate and Energy Policy Framework: EU Renewable Energy Policies Buenos Aires 26-27 May 2015 Dr Stefan Agne European Commission DG Climate Action Energy 1 EU Climate and Energy Policy Framework 2 Agreed

More information

Eeekonomics: BEG/CEE UT Annual Meeting. Commentary: European Live Issues, Post recession Demand. European gas demand post Fukushima Outline

Eeekonomics: BEG/CEE UT Annual Meeting. Commentary: European Live Issues, Post recession Demand. European gas demand post Fukushima Outline Gas Research Programme ENERGY STUDIES Natural G Eeekonomics: BEG/CEE UT Annual Meeting Houston, 7 December 2011 Commentary: European Live Issues, Post recession Demand Dr Anouk Honore Senior Research Fellow

More information

EUROPEAN GAS MARKET :

EUROPEAN GAS MARKET : EUROPEAN GAS MARKET : The challenges to face WEC-Europe regional meeting MONACO November 7, 2012 Jean-Marie DAUGER Executive Vice-President GDF SUEZ The challenges of natural gas in Europe Acceptability

More information

China Renewable Energy Outlook 2018

China Renewable Energy Outlook 2018 China Renewable Energy Outlook 218 The energy system for a Beautiful China 25 Time for a new era in the Chinese energy transition The China Renewable Energy Outlook 218 (CREO 218) is guided by the strategic

More information

The cost of renewable energy:

The cost of renewable energy: The cost of renewable energy: A critical assessment of the Impact Assessments underlying the Clean Energy for All Europeans-Package Andreas Graf BERLIN, 2 JUNE 2017 State of play: outlook Malta presidency

More information

European Commission. Communication on Support Schemes for electricity from renewable energy sources

European Commission. Communication on Support Schemes for electricity from renewable energy sources European Commission Communication on Support Schemes for electricity from renewable energy sources External Costs of energy and their internalisation in Europe Beatriz Yordi DG Energy and Transport External

More information

Finding an Optimal Path to 2050 Decarbonization Goals

Finding an Optimal Path to 2050 Decarbonization Goals Finding an Optimal Path to 2050 Decarbonization Goals John Bistline, Ph.D. Technical Leader 3 rd IEA-EPRI Workshop Paris October 17, 2016 Substantial Effort Beyond NDCs Will Be Required Billion tonnes

More information

138 ENVIRONMENTAL PROFILE OF SPAIN 2011

138 ENVIRONMENTAL PROFILE OF SPAIN 2011 2.9 ENERGY At the end of 2011, the EU Commission launched the Energy Road Map 2050, a communication analysing the challenges of decarbonising the EU, while at the same time ensuring security of energy

More information

Energy demand dynamics and infrastructure development plans in the EU. October 10 th, 2012 Jonas Akelis, Managing Partner - Baltics

Energy demand dynamics and infrastructure development plans in the EU. October 10 th, 2012 Jonas Akelis, Managing Partner - Baltics Energy demand dynamics and infrastructure development plans in the EU October 10 th, 2012 Jonas Akelis, Managing Partner - Baltics Forecasted energy demand dynamics of EU-11 will be significantly higher

More information

Risk managing cost-effective decarbonisation of the power sector in Germany

Risk managing cost-effective decarbonisation of the power sector in Germany Risk managing costeffective decarbonisation of the power sector in Germany FINAL RESULTS April 2013 This project is funded by the European Climate Foundation 1 Contents Objectives and the methodology Baseline

More information

Capacity reserves until 2025: declining, but sufficient

Capacity reserves until 2025: declining, but sufficient 1 Capacity reserves until 2025: declining, but sufficient Trends from ENTSO-E s Scenario Outlook & Adequacy Forecast 2015 Once a year, ENTSO-E 1 publishes a Scenario Outlook & Adequacy Forecast (SO&AF).

More information

SAMPLE. Reference Code: GDAE6214IDB. Publication Date: September GDAE6214IDB / Published SEP 2012

SAMPLE. Reference Code: GDAE6214IDB. Publication Date: September GDAE6214IDB / Published SEP 2012 Solar PV in Spain, Market Outlook to 2025 - Capacity, Generation, Levelized Cost of Energy (LCOE), Investment Trends, Regulations and Reference Code: GDAE6214IDB Publication Date: September 2012 GlobalData.

More information

Reference Code: GDAE6521IDB. Publication Date: March 2015

Reference Code: GDAE6521IDB. Publication Date: March 2015 Hydro Power in France, Market Outlook to 2025, Update 2015 Capacity, Generation, Levelized Cost of Energy (LCOE), Investment Trends, Regulations and Company Profiles Reference Code: GDAE6521IDB Publication

More information

Nuclear Energy and Greenhouse Gas Emissions Avoidance in the EU

Nuclear Energy and Greenhouse Gas Emissions Avoidance in the EU Position Paper Nuclear Energy and Greenhouse Gas Emissions Avoidance in the EU The European Atomic Forum (FORATOM) is the Brussels-based trade association for the nuclear energy industry in Europe. The

More information

Profound changes underway in energy markets Signs of decoupling of energy-related CO 2 emissions and global economic growth Oil prices have fallen pre

Profound changes underway in energy markets Signs of decoupling of energy-related CO 2 emissions and global economic growth Oil prices have fallen pre Keisuke Sadamori Director of Energy Markets and Security, IEA The 88th IEEJ Energy Seminar, 5th October 215 Profound changes underway in energy markets Signs of decoupling of energy-related CO 2 emissions

More information

Nuclear-renewable hybrid energy systems for non-electric applications

Nuclear-renewable hybrid energy systems for non-electric applications Nuclear-renewable hybrid energy systems for non-electric applications Cost issues Saied Dardour 3E Analysis Unit Planning and Economic Studies Section Division of Planning, Information and Knowledge Management

More information

Transformation towards a sustainable European Energy System Which Roadmaps?

Transformation towards a sustainable European Energy System Which Roadmaps? Transformation towards a sustainable European Energy System Which Roadmaps? Institut für Energiewirtschaft und Rationelle Energieanwendung Universität Stuttgart EUROPEAN ENERGY CONFERENCE 212 Maastricht,

More information

Part 1: Sustainability in the Electricity Sector

Part 1: Sustainability in the Electricity Sector Part 1: Sustainability in the Electricity Sector Franz Trieb MBA Energy Management, Vienna, September 9-10, 2010 Folie 1 A Scenario for Sustainability www.ren21.net Folie 2 Studies: MED-CSP TRANS-CSP AQUA-CSP

More information

Economic analysis of reaching a 20% share of renewable energy sources in 2020

Economic analysis of reaching a 20% share of renewable energy sources in 2020 Executive Summary Economic analysis of reaching a 20% share of renewable energy sources in 2020 Mario Ragwitz, Felipe Toro Fraunhofer - ISI Gustav Resch, Thomas Faber, Reinhard Haas - EEG Monique Hoogwijk,

More information

Options for structural measures in the EU ETS

Options for structural measures in the EU ETS CEPS Carbon Market Forum 23 April 2013, Brussels Options for structural measures in the EU ETS Stefan P. Schleicher Wegener Center for Climate and Global Change University of Graz A B B C C D E F F G G

More information

Future EU energy and climate regulation. Implications for Nordic energy development and Nordic stakeholders

Future EU energy and climate regulation. Implications for Nordic energy development and Nordic stakeholders Future EU energy and climate regulation Implications for Nordic energy development and Nordic stakeholders Future EU energy and climate regulation Implications for Nordic energy development and Nordic

More information

Supporting the deployment of selected low-carbon technologies in Europe

Supporting the deployment of selected low-carbon technologies in Europe Supporting the deployment of selected low-carbon technologies in Europe Implications of technoeconomic assumptions. An energy system perspective with the JRC-EU-TIMES model Wouter Nijs, Savvas Politis,

More information

Belgium Energy efficiency report

Belgium Energy efficiency report Belgium Energy efficiency report Objectives: 27.5 TWh of end-user energy savings in 216 12% share of renewables for electricity suppliers in 212 Overview - (% / year) Primary intensity (EU=) 1 132 - -1.8%

More information

Energy Efficiency Indicators: The Electric Power Sector

Energy Efficiency Indicators: The Electric Power Sector Energy Efficiency Indicators: 5 Sectors, 5 Challenges Mexico City, Mexico 14-15 March 2011 Energy Efficiency Indicators: The Electric Power Sector Robert Schnapp Head, Coal, Renewables, Electric and Heat

More information

Slovenia Energy efficiency report

Slovenia Energy efficiency report Slovenia Energy efficiency report Objectives: 4.3 TWh of end-use energy savings by 2016 Overview - (% / year) Primary intensity (EU=100) 1 115 - -1.9% + CO 2 intensity (EU=100) 120 - -1.6% - CO 2 emissions

More information

The PRIMES Energy Model

The PRIMES Energy Model EC4MACS Uncertainty Treatment The PRIMES Energy Model European Consortium for Modelling of Air Pollution and Climate Strategies - EC4MACS Editors: E3MLab, National Technical University of Athens (NTUA)

More information

Reference scenario with PRIMES

Reference scenario with PRIMES EUROPEAN ENERGY AND TRANSPORT TRENDS TO 2030 UPDATE 2009 Reference scenario with PRIMES Dr. Leonidas MANTZOS E3MLab National Technical University of Athens April 2010 PRIMES ENERGY SYSTEM MODEL Main Features

More information

Nuclear power is part of the solution for fighting climate change

Nuclear power is part of the solution for fighting climate change Nuclear power is part of the solution for fighting climate change "Nuclear for Climate" is an initiative undertaken by the members of the French Nuclear Energy Society (SFEN), the American Nuclear Society

More information

Bio-energy and the European Pulp and Paper Industry An Impact Assessment

Bio-energy and the European Pulp and Paper Industry An Impact Assessment Bio-energy and the European Pulp and Paper Industry An Impact Assessment MCKINSEY & COMPANY, INC. AND PÖYRY FOREST INDUSTRY CONSULTING FOR CEPI Project Summary July 16, 2007 This document is a summary

More information

IEA WORK ON FUTURE ELECTRICITY SYSTEMS

IEA WORK ON FUTURE ELECTRICITY SYSTEMS IEA WORK ON FUTURE ELECTRICITY SYSTEMS Power grids, demand response and the low carbon transition Dr. Luis Munuera Smart Grids Technology Lead IEA Symposium on Demand Flexibility and RES Integration SMART

More information

Medium Term Renewable Energy Market Report 2016

Medium Term Renewable Energy Market Report 2016 Medium Term Renewable Energy Market Report 2016 Clean Energy Investment and Trends IETA Pavilion COP22, Marrakech November 10, 2016 Liwayway Adkins Environment and Climate Change Unit International Energy

More information

Emissions Trading System (ETS): The UK needs to deliver its share of the total EU ETS emissions reduction of 21% by 2020, compared to 2005;

Emissions Trading System (ETS): The UK needs to deliver its share of the total EU ETS emissions reduction of 21% by 2020, compared to 2005; Emissions Trading System (ETS): The UK needs to deliver its share of the total EU ETS emissions reduction of 21% by 2020, compared to 2005; Non-ETS emissions: The Effort Sharing Decision sets a target

More information

Roadmap for Solar PV. Michael Waldron Renewable Energy Division International Energy Agency

Roadmap for Solar PV. Michael Waldron Renewable Energy Division International Energy Agency Roadmap for Solar PV Michael Waldron Renewable Energy Division International Energy Agency OECD/IEA 2014 IEA work on renewables IEA renewables website: http://www.iea.org/topics/renewables/ Renewable Policies

More information

Renewable energy technologies/sources path within EU 2020 strategy

Renewable energy technologies/sources path within EU 2020 strategy Renewable energy technologies/sources path within EU 2020 strategy Analysis according to national renewable energy action plans and 2013 progress reports Manjola Banja, Fabio Monforti-Ferrario, Katalin

More information

How to secure Europe s competitiveness in terms of energy and raw materials? The answer, my friend, is blowing in the wind

How to secure Europe s competitiveness in terms of energy and raw materials? The answer, my friend, is blowing in the wind How to secure Europe s competitiveness in terms of energy and raw materials? The answer, my friend, is blowing in the wind Iván Pineda Head of Policy Analysis, EWEA PolyTalk 2014, Brussels Around 600 members

More information

THE BALTIC SEA REGION STORAGE, GRID EXCHANGE AND FLEXIBLE ELECTRICITY GENERATION FOR THE TRANSITION TO A 100% RENEWABLE ENERGY SYSTEM

THE BALTIC SEA REGION STORAGE, GRID EXCHANGE AND FLEXIBLE ELECTRICITY GENERATION FOR THE TRANSITION TO A 100% RENEWABLE ENERGY SYSTEM THE BALTIC SEA REGION STORAGE, GRID EXCHANGE AND FLEXIBLE ELECTRICITY GENERATION FOR THE TRANSITION TO A 100% RENEWABLE ENERGY SYSTEM Michael Child, Dmitrii Bogdanov and Christian Breyer Lappeenranta University

More information

ENERGY PRIORITIES FOR EUROPE

ENERGY PRIORITIES FOR EUROPE ENERGY PRIORITIES FOR EUROPE Presentation of J.M. Barroso, President of the European Commission, to the European Council of 4 February 2011 Contents 1 I. Why energy policy matters II. Why we need to act

More information

Resource efficiency and waste

Resource efficiency and waste Municipal Municipal management across European See also: Country profiles on municipal management 1. Introduction Over the last two decades, European have increasingly shifted their focus with regard to

More information

Energy Technology Perspectives 2014 Harnessing Electricity s Potential

Energy Technology Perspectives 2014 Harnessing Electricity s Potential The Global Outlook An active transformation of the energy system is essential to meet long-term goals. (ETP 2014) charts a course by which policy and technology together become driving forces in transforming

More information

COMPETITIVENESS OF NUCLEAR ENERGY AN INTERNATIONAL VIEWPOINT. Nuclear Power Plants for Poland, Warsaw 1-2 June 2006

COMPETITIVENESS OF NUCLEAR ENERGY AN INTERNATIONAL VIEWPOINT. Nuclear Power Plants for Poland, Warsaw 1-2 June 2006 COMPETITIVENESS OF NUCLEAR ENERGY AN INTERNATIONAL VIEWPOINT Power Plants for Poland, Warsaw 1-2 June 2006 Evelyne BERTEL OECD Energy Agency Bertel@nea.fr Introduction Economic competitiveness, which always

More information

Wetter and Wilder: Impacts on the electricity industry in Western Europe of climate change

Wetter and Wilder: Impacts on the electricity industry in Western Europe of climate change Working Paper 3/2011 Wetter and Wilder: Impacts on the electricity industry in Western Europe of climate change Rolf Golombek, Sverre A.C. Kittelsen and Ingjerd Haddeland The CREE Centre acknowledges financial

More information

Electricity sector transformation in Europe Taking local idiosyncrasies into account

Electricity sector transformation in Europe Taking local idiosyncrasies into account Electricity generation in TWh IAEE Vienna September 217 4.5 4. 3.5 3. 2.5 2. 1.5 1. 5 215 22 225 23 235 24 245 25 Electricity sector transformation in Europe Taking local idiosyncrasies into account Clemens

More information

The role of energy efficient buildings in the EUs future power system

The role of energy efficient buildings in the EUs future power system The role of energy efficient buildings in the EUs future power system The role of energy efficient buildings in the EUs future power system Thomas Boermans, Georgios Papaefthymiou, Markus Offermann, Ashok

More information

Innovation benefits from nuclear phase-out: Can they compensate the costs?

Innovation benefits from nuclear phase-out: Can they compensate the costs? 2013 International Energy Workshop Paris, 19 th -21 st June 2013 Innovation benefits from nuclear phase-out: Can they compensate the costs? Enrica De Cian, Samuel Carrara, Massimo Tavoni FEEM (Fondazione

More information

Germany s Energiewende

Germany s Energiewende Germany s Energiewende at a glance Vienna, 4 MARCH 2014 Basic Facts 1 What is it all about? A Definition of the Energiewende In order to realise high shares of Renewables in the entire energy system, a

More information

ASSESSING GOOD PRACTICES IN POLICIES AND MEASURES TO MITIGATE CLIMATE CHANGE IN CENTRAL AND EASTERN EUROPE. Elena Petkova

ASSESSING GOOD PRACTICES IN POLICIES AND MEASURES TO MITIGATE CLIMATE CHANGE IN CENTRAL AND EASTERN EUROPE. Elena Petkova Workshop on Best Practices in Policies and Measures, 8-10 October 2001, Copenhagen ASSESSING GOOD PRACTICES IN POLICIES AND MEASURES TO MITIGATE CLIMATE CHANGE IN CENTRAL AND EASTERN EUROPE Elena Petkova

More information

RTE Réseau de transport d électricité shall not be liable for damages of any nature, direct or indirect, arising from the use, exploitation or

RTE Réseau de transport d électricité shall not be liable for damages of any nature, direct or indirect, arising from the use, exploitation or GENERATION ADEQUACY ACY REPORT on the electricity city supply-demand balance an in France 216 EDITION IO EXECUTIVE SUMMARY RTE Réseau de transport d électricité shall not be liable for damages of any nature,

More information

Medium Term Renewable Energy Market Report 2013

Medium Term Renewable Energy Market Report 2013 Renewable Energy Market Report 213 Michael Waldron Renewable Energy Division International Energy Agency OECD/IEA 213 OECD/IEA 213 MTRMR methodology and scope Analysis of drivers and challenges for RE

More information

Working Paper No. 4. November 2018

Working Paper No. 4. November 2018 Working Paper No. 4 November 2018 Carbon Price Floor in Ireland Author: Paul Deane 1, John FitzGerald 2 and Gemma O Reilly 3 1 MaREI Centre/Environmental Research Institute, University College Cork 2 Chair

More information

Renewables after COP-21 A global perspective. Dr. Fatih Birol Executive Director International Energy Agency

Renewables after COP-21 A global perspective. Dr. Fatih Birol Executive Director International Energy Agency Renewables after COP-21 A global perspective Dr. Fatih Birol Executive Director International Energy Agency 17 th Symposium, Syndicat des Énergies Renouvelables, Unesco, Paris, 4 February 2016 The start

More information

Opportunities in Renewable Energies. World Renewable Energy technology Congress Delhi, 27 th September 2013

Opportunities in Renewable Energies. World Renewable Energy technology Congress Delhi, 27 th September 2013 Opportunities in Renewable Energies Franzjosef Schafhausen Deputy Director General Energy Transition Federal Ministry of the Environment, Nature Conservation and Nuclear Safety, Berlin World Renewable

More information

The hydropower sector directly and indirectly contributes to the European economy in several ways:

The hydropower sector directly and indirectly contributes to the European economy in several ways: THE HYDROPOWER SECTOR S CONTRIBUTION TO A SUSTAINABLE AND PROSPEROUS EUROPE Executive Summary Scope and Purpose of this Report The climate and energy policy of the European Union (EU-28) and many European

More information

SAMPLE. Reference Code: GDAE6535IDB. Publication Date: May 2015

SAMPLE. Reference Code: GDAE6535IDB. Publication Date: May 2015 Hydropower in Austria, Market Outlook to 2025, Update 2015 Capacity, Generation, Levelized Cost of Energy (LCOE), Investment Trends, Regulations and Company Profiles Reference Code: GDAE6535IDB Publication

More information

Summary Report UK s Long Term Power Balance and Implications for Major Power Producers Gas Consumption

Summary Report UK s Long Term Power Balance and Implications for Major Power Producers Gas Consumption Updated 17th of June 015 Summary Report UK s Long Term Power Balance and Implications for Major Power Producers Gas Consumption Ambitious cuts in greenhouse gases and strong promotion of renewable energy

More information

Renewables 2O18 Analysis and Forecasts to 2O23

Renewables 2O18 Analysis and Forecasts to 2O23 Market Report Series Secure Sustainable Together Renewables 2O18 Analysis and Forecasts to 2O23 executive summary Market Report Series Renewables 2O18 Analysis and Forecasts to 2O23 executive summary INTERNATIONAL

More information

Energy consumption, GDP and energy intensity '85 '90 '95 '00 '05

Energy consumption, GDP and energy intensity '85 '90 '95 '00 '05 Index 198=1 The Danish example the way to an energy efficient and energy friendly economy February 29 Danish experience shows that through persistent and active energy policy focus on enhanced energy efficiency,

More information

Finland Energy efficiency report

Finland Energy efficiency report Finland Energy efficiency report Objectives: 37 TWh of energy savings by 22 Overview - (% / year) Primary intensity (EU=1) 1 167 -- -1.2% - CO 2 intensity (EU=1) 12 - -1.5% -- CO 2 emissions per capita

More information

What role subsidies? A CGE analysis of announcement effects of future policies on the development of emissions and energy consumption in Finland

What role subsidies? A CGE analysis of announcement effects of future policies on the development of emissions and energy consumption in Finland What role subsidies? A CGE analysis of announcement effects of future policies on the development of emissions and energy consumption in Finland Juha Honkatukia 1 Introduction The European Union has committed

More information

Nuclear Power in France

Nuclear Power in France 1 Nuclear Power in France Dr Sunil Félix, Nuclear Counsellor, French Embassy in Tokyo Seminar on WNE and Overseas Nuclear Industry February 9, 2016 2 AGENDA Energy Context and Key Trends Nuclear Power

More information

When the Wind Blows Over Europe European Experiences with RES

When the Wind Blows Over Europe European Experiences with RES When the Wind Blows Over Europe European Experiences with RES Harvard electricity Policy Group Fifty-Second Plenary Session Chicago, 02-03 October 2008 Hannes Weigt Chair of Energy Economics and Public

More information

German energy policy a blueprint for the world? Full survey results. World Energy Council Germany Berlin, January 2017

German energy policy a blueprint for the world? Full survey results. World Energy Council Germany Berlin, January 2017 German energy policy a blueprint for the world? Full survey results World Energy Council Germany Berlin, January 2017 WEC Survey German Energy Policy Survey Overview Global View 42 countries Europe vs.

More information

Potentials, costs, and environmental effects of electricity generation technologies.

Potentials, costs, and environmental effects of electricity generation technologies. Department of the Environment, Transport, Energy and Communication DETEC Swiss Federal Office of Energy SFOE Energy Supply and Monitoring Potentials, costs, and environmental effects of electricity generation

More information

CONTENTS TABLE OF PART A GLOBAL ENERGY TRENDS PART B SPECIAL FOCUS ON RENEWABLE ENERGY OECD/IEA, 2016 ANNEXES

CONTENTS TABLE OF PART A GLOBAL ENERGY TRENDS PART B SPECIAL FOCUS ON RENEWABLE ENERGY OECD/IEA, 2016 ANNEXES TABLE OF CONTENTS PART A GLOBAL ENERGY TRENDS PART B SPECIAL FOCUS ON RENEWABLE ENERGY ANNEXES INTRODUCTION AND SCOPE 1 OVERVIEW 2 OIL MARKET OUTLOOK 3 NATURAL GAS MARKET OUTLOOK 4 COAL MARKET OUTLOOK

More information

Statkraft-LSE Program: Credible, effective and publicly acceptable policies to decarbonise the European Union

Statkraft-LSE Program: Credible, effective and publicly acceptable policies to decarbonise the European Union Statkraft-LSE Program: Credible, effective and publicly acceptable policies to decarbonise the European Union Samuela Bassi, Maria Carvalho, Baran Doda and Sam Fankhauser in collaboration with: Alina Averchenkova

More information

GE OIL & GAS ANNUAL MEETING 2016 Florence, Italy, 1-2 February

GE OIL & GAS ANNUAL MEETING 2016 Florence, Italy, 1-2 February Navigating energy transition Keisuke Sadamori Director for Energy Markets and Security IEA GE OIL & GAS ANNUAL MEETING 2016 Florence, Italy, 1-2 February 2016 General Electric Company - All rights reserved

More information

AREVA s Vision of Global Nuclear Market

AREVA s Vision of Global Nuclear Market AREVA s Vision of Global Nuclear Market Didier Beutier ICO Marketing 16 March 2012 Table of Contents Global Post Fukushima Outlook Safety imperative AREVA position on the market Our prospects D. BEUTIER

More information