Long-term Transition Paths towards a Sustainable Energy Supply

Long-term Transition Paths towards a Sustainable Energy Supply Gerrit Jan Schaeffer PhD. Manager Transition and Innovation Group Policy Studies Unit of the Energy Research Center of the Netherlands ECN Harvard 22 May 2003

Four main parts of this presentation A short introduction to myself, ECN and the Transition and Innovation group Transition, a core concept in Dutch energy policy Modelling long-term transtions: example of the Western Europe electricity sector Innovation and learning: example of solar PV technology

Personal introduction Background in Physics and Sociology of Science and Technology ( Science and Technology Studies ) PhD. in 1998 on history of Fuel Cells and what can be learned from that for technology development mechanisms in general Renewable Energy policy specialist at ECN 1998-2001 Manager of Transition and Innovation group since 2002

ECN, the Energy research Center of the Netherlands

ECN, some characteristics Working force: 900 people Departments (units): 9 - Solar, Wind, Biomass, Fuel Cells, Clean Fossil Fuels, Energy Efficiency in the Industry, Nuclear, Renewable Buildings and.. Policy Studies Mission: Contribute to transition to sustainable energy system by research Clients: governments (national and European), (large) enterprises

Policy Studies Unit Supports Dutch Government (Ministry of Economics) for contents recurring energy policy documents Takes part in many EU-funded policy projects Provides services to multi-lateral organisations + private sector

Policy Studies: 4 themes Climate Change Studies - emission trading, CDM, JI etc. Gas and Electricity Market studies - market modelling, studying different market designs (etc.) New and Renewable Energy Policy - RPS, Feed-In, potential and cost of renewable energy technologies Transition and Innovation Group - Long-term energy modelling, societal aspects of technological change

The concept of transition

Characteristics of transitions Multi-level - novelties (in niches), regime en landschap - transition policy should be focussing on each of the three levels Multi-dimension - apart from technology also social networds, knowledge, user preferences, cultural meaning, sectoral policy and infrastructure Multi-actor - means that not everybody of today will join in forming the world of tomorrow - new connections and networks between actors will have to be constructed Multi-phase - RDD&D instead of R&D - one technology can pave the road for another (see bikes and cars)

Transition management? What does managing mean when you talk about such complex, long-term processes? Characteristics transition management - Long-term assessment part of shorter term decisions - Action/policy should be integral, I.e. focused on several domains, levels, actors - The concept of learning is very important. How to improve learning is crucial Steps - Achieve some convergence on the transition target - Assess different pictures of the future related to the target - Formulate intermediate targets and transition paths - Start and plan evaluation and learning cycles - Create societal support

Transition in Dutch policy Core concept in 4th National Environmental Report (2001) Focused on (Sustainable) Agriculture, Biodiversity, Mobility and Energy Ministry of Energy leads Energy Transition Policy Four transition trajectories have been chosen: New gas (hydrogen?), biomass, regional energy systems and the energy-intensive industry - Lots of workshops - Start of drafting visions - thinking about transition paths - dealing with uncertainties by defining transition experiments

The role of the T&I-group in this discussion Bring knowledge on the concept of transition to our energy acquaintances Bring knowledge on energy technologies, economics and sector to our transition acquaintances Competences (10 people): - long-term energy modelling (MARKAL) - sociology of technology - energy economics

Transition in a broader European perspective Long-term forecasts and studies in several countries of the EU - (e.g. Germany, Finland, Denmark, UK) Strategic discussions on energy going on in several countries - France, Germany, Belgium, UK EU Energy policy directives - Renewable electricity targets - Alternative Fuels communication (23 % in 2020) - High level hydrogen working group

Germany s global energy transformation

MARKAL Western Europe (start of part 3) Single region region model covering OECD 1990 Western Europe Time period: 1990-2100 in 10 year steps Price elastic demand version: impact on end use demand under severe constraints Energy system with biomass-food industry sector included Endogenous technology learning (ETL) using SFLC (learning-by doing) and cluster approach (This is a unique feature for large-scale models)

MARKAL CO 2 capture and storage IMPORT MINING STOCKS EXPORT REFINERIES FUEL PROCESSING and CONVERSION ELECTRICITY HEAT END USE DEVICES, PROCESSES & D E M A N D S TECHNO- LOGIES BIOMASS ACTIVITIES and PROCESSING, FORESTRY AGRICULTU RE and FOOD INDUSTRY Land use

Technological change Before 1999: technological progress induced through exogenous cost decrease 1999: learning-by-doing investment costs decrease as function of cumulative capacity learning curve: SC(C) = a C -b progress ratio (pr = 2 -b ): cost reduction factor for each doubling in cumulative capacity 2001: learning-by-searching investment costs also decrease as function of (public) R&D component and cluster approach (learning)

Learning curves in MARKAL Example: Solar PV modules (pr = 0.81) /kw] 18000 16000 14000 12000 1984 historical data trend 10000 8000 1994 6000 4000 2000 0 0 100 200 300 400 500 cumulative installed capacity (MW) 100000 /kw 10000 1000 historical data trend 100 10 1 1 10 100 1000 Cumulative installed capacity (MW)

Clusters of technologies (1) Definitions Cluster = a group of technologies sharing a common essential (i.e. learning) component; therefore the learning behaviour of these technologies is linked Component = the selected learning key technology shared by all technologies in a cluster Technologies are build by a number of components and a balance of system (infrastructure, non learning parts)

Clusters of technologies (2) Parts of investment costs over time Two examples of key technologies Gas turbine many applications, simple cycle (SC) or combined cycle (CC) CC on gas, coal or biomass (see note) learning gas CC dominated by learning gas turbine Investment cost 1200 1000 800 600 400 200 0 1 2 3 4 5 6 model period Note: coal and biomass CC (i.e. after gasification) = another cluster Solar PV module many markets, regional differences PV system = PV module (shared) + BOS (customized) ETL Comp 2 ETL Comp 1 N-ETL

Clusters of technologies (3) 19 clusters currently implemented Cluster no. Description # technologies in cluster 1 Solar PV modules 6 2 Wind turbine 8 3 Fuel cell 13 4 Hydro turbine 4 5 Gas turbine 32 6 Gasifier 20 7 Steam turbine 45 8 Boiler 14 9 Combined cycle boiler 25 10 Nuclear reactor 1 11-15 CO 2 sequestration 10 16-18 Up stream oil and gas 15 19 Fusion reactor 2

Multi Sectoral Learning - Clusters Learning components-clusters in: - up stream oil and gas exploitation (3) - electricity production (13) - CO 2 capture (3) more than 90 technologies with 1 or more learning components: - up stream oil and gas exploitation (9) - electricity production (69) & transport (fuel cell; 7) - CO 2 capture (9)

Long-term transitions runs for electricity sector Purpose: to test radical future images in a long term technology rich model Renewable electricity (green) target: - 100% by 2050, 2070 and 2100 - optimal path and linear fixed path from 26% in 2010 CO 2 less electricity production: - 0 Mton CO 2 in 2050, 2070 and 2100 - optimal path and fixed linear path from 850 Mton in 2010 No new investments in fission (LWR) reactors after 2000

Scenarios Determined path 100% renewable Free path Slow transition pace 2100 2070 Determined path 2050 Fast transition pace 100% CO2 emission free Free path

Base case: electricity production TWh 5000 4500 4000 3500 3000 2500 2000 1500 1000 500 0 1990 2000 Share renewable electricity 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 nuclear solid with sequestration solid oil with sequestration oil gas with sequstration gas others hydro wind solar biomass 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 19.9% 19.4% 18.8% 21.7% 23.5% 22.7% 21.7% 24.3% 23.4% 21.8% 20.8% 25.8%

Scenario results TWh TWh 100% renewable by 2050, fixed linear path 4500 4000 3500 3000 2500 2000 1500 1000 500 4500 4000 3500 3000 2500 2000 1500 1000 500 0 1990 2000 2010 Id for 2100 2020 2030 2040 2050 2060 2070 2080 2090 2100 0 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 nuclear solid with sequestration solid oil with sequestration oil gas with sequstration gas others hydro wind solar biomass nuclear solid with sequestration solid oil with sequestration oil gas with sequstration gas others hydro wind solar biomass TWh TWh 4500 4000 3500 3000 2500 2000 1500 1000 500 0 4500 4000 3500 3000 2500 2000 1500 1000 500 0 optimal path 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 nuclear solid with sequestration solid oil with sequestration oil gas with sequstration gas others hydro wind solar biomass nuclear solid with sequestration solid oil with sequestration oil gas with sequstration gas others hydro wind solar biomass

Scenario results 0 CO 2 and 100% renewable, with (left) and without nuclear (right) TWh TWh 4500 4000 3500 3000 2500 2000 1500 1000 500 0 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 4500 4000 3500 3000 2500 2000 1500 1000 500 0 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 nuclear solid with sequestration solid oil with sequestration oil gas with sequstration gas others hydro wind solar biomass nuclear solid with sequestration solid oil with sequestration oil gas with sequstration gas others hydro wind solar biomass TWh TWh 4500 4000 3500 3000 2500 2000 1500 1000 500 4500 4000 3500 3000 2500 2000 1500 1000 500 0 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 0 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 nuclear solid with sequestration solid oil with sequestration oil gas with sequstration gas others hydro wind solar biomass nuclear solid with sequestration solid oil with sequestration oil gas with sequstration gas others hydro wind solar biomass 2100 2070

Results: Learning Effects - Clusters E.g. IGCC with flue gas CO 2 capture 2500 700 Investment cost [ /kwe] 2000 1500 1000 500 0 600 500 400 300 200 100 0 Cumulative capacity [GW] CO2 capture CC boiler steamturbine gasturbine gasifier N-ETL INVESTMENT COST 10000 1000 100 10 100 1000 CUMULATIVE CAPACITY => PR = 0.907 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 Clusters lead to technology spill-over and shared learning experience

Results: Learning Effects - Clusters (2) 160 Capacity Investments Steam Turbines [GWel] 140 120 100 80 60 40 20 0 ----> time Mature technology example: steam turbine Capacity investments per technolog in a cluster indicates relative contribution to learning experience Cumulative capaity Steam Turbines [GWel] 900 800 700 600 500 400 300 200 100 0 ----> time Cumulative capacity of component is equally built up by contributions from all technologies in the cluster

Results: Learning Effects - Clusters (3) Capacity Investments Gasifier [GWel] 250 200 150 100 50 0 Leader = coal ----> time Promising technology example: gasifier Capacity investments per technolog in a cluster indicate which technolog is leader and which is follower preferential areas/technologies for (policy) intervention Intermediate follower = coal with CO 2 capture and SOFC Last follower = biomass

Results: comments All scenarios show a dip in electricity production in the target year: - system (model) postpones efforts to the end - effect most moderate for 2100 System (model) uses fossil or non renewables as much as possible in the free path scenarios => even with perfect foresight still some inertia to reach severe targets; all changes occur in 10-20 years before target year. Emission free scenarios not only achieved by renewables, also nuclear and CO 2 sequestration play a role. Emission reductions in electricity sector partially (20% on average) annulled by increased CO 2 emissions in other sectors (price electricity so more competition with other energy carriers (fossils) in end use).

Results comments (2) Costs analysis is complicated by use of price elastic model version: system cost for some reduction scenarios becomes cheaper than base case, but loss of welfare is significant. Electricity costs see peak in target year, for free and determined scenarios, afterwards more relaxed, so cheaper. Welfare loss over whole period is about 0.05 to 0.30 % of total GDP (28500 trillion ). Nuclear phase out also investigated => targets can also be reached without new nuclear but electricity costs increase with 6% on average.

Results: comments (3) /kw 8000 7000 6000 5000 4000 3000 2000 1000 0 Technology learning is very much scenario and scenario assumptions dependent (e.g. discount rate or fuel prices) as well as from the learning curve parameters (e.g progress ratio) Solar PV 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 Specific Investment Costs [ 90/kWel] Wind turbines /kw 1 1600 2 3 14004 5 6 1200 7 8 10009 13 14 800 15 16 600 17 18 19 40020 21 25 200 26 27 028 29 1990 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 1 2 3 4 5 6 7 8 9 13 14 15 16 17 18 19 20 21 25 26 27 28 29

Conclusions modelling part MARKAL as technology-rich model is extremely suitable to include learning (by doing). Not only new, promising technologies but also mature technologies balance the scenario results. The cluster approach offers wider areas for analysis and allows technology transfer and spill-over (between sectors and regions). The first model runs looking at transition scenarios are promising. Single year targets give rather unlikely results and path or intermediate targets lead to more acceptable and realisable changes in the energy system. Further elaboration of transition items (regimes, niche markets,..) and future images (policy targets) in the model will increase its relevance and applicability

In search of the right balance between market deployment stimulation and R&D The case of Solar PV

Cost development for solar and wind $70 $60 Solar-PV $50 $40 $30 $20 $10 $0 1970 1975 1980 1985 1990 1995 2000 2005 3500 3000 2500 2000 Wind 1500 1000 500 0 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998

The learning curve $100 APS ($2001) 1976-2001 ASP for Power Modules Strategies Unlimited, March 2003 Progress ratio = 80% $10 Lower Learning learning investment investment y = 40.408x -0.3216 R 2 = 0.988 PR = 80.0% (was 80.1 for 1976-1999) After each cumulative doubling of installed capacity costs are just 80% of costs before Assume, this is break-even $1 Improved experience curve (for instance because of additional R&D) Power Module Cumulative Shipment (MW p ) 0 1 10 100 1 000 10 000 1999

What do we know about learning processes?

Learning processes Occur because of combination of R&D and market deployment R&D influences progress ratio and thus diminishes the market deployment support needed The search is for an optimal mix - just market deployment support, or just R&D can be sub-optimal

Progress ratio a.o dependent on growth rate 85% 83% 81% 79% 1977-1986 (R^2 = 0.98) 1976-1985 (R^2 = 0.98) 1978-1987 (R^2 = 0.97) 1980-1989 (R^2 = 0.95) 1979-1988 (R^2 = 0.97) 1981-1990 (R^2 = 0.93) 1985-1994 (R^2 = 0.91) 1984-1993 (R^2 = 0.92) 1983-1992 (R^2 = 0.91) 1982-1991 (R^2 = 0.94) 1986-1995 (R^2 = 0.89) 1987-1996 (R^2 = 0.83) 77% 75% [MWp] 2000 1800 1600 1400 SU Total Modules Cumulative world-wide PV shipment according to several literature sources Maycock world total shipments EPIA, 1996, 2001 Maycock, 1988-2002 100% 90% 80% 70% 1988-1997 (R^2 = 0.90) 73% 71% 1200 1000 800 600 400 Annual growth Maycock [in %, right axis] PR (R^2 acceptable) PR (R^2 not acceptable) 60% 50% 40% 30% 20% 1989-1998 (R^2 = 0.95) 1990-1999 (R^2 = 0.98) 69% 200 10% 1976-1985 1977 (R^2-1986 = 0.98) 1978 (R^2-1987 = 0.98) 1979 (R^2-1988 = 0.97) 1980 (R^2-1989 = 0.97) 1981 (R^2-1990 = 0.95) 1982 (R^2-1991 = 0.93) 1983 (R^2-1992 = 0.94) 1984 (R^2-1993 = 0.91) 1985 (R^2-1994 = 0.92) 1986 (R^2-1995 = 0.91) 1987 (R^2-1996 = 1990 1989 1988 0.89) (R^2-1999 1998 1997 = 0.83) (R^2 = 0.98) 0.95) 0.90) 0 0% 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 1976-1985 (R^2 = 0.98) 1977-1986 (R^2 = 0.98) 1978-1987 (R^2 = 0.97) 1979-1988 (R^2 = 0.97) 1980-1989 (R^2 = 0.95) 1981-1990 (R^2 = 0.93) 1982-1991 (R^2 = 0.94) 1983-1992 (R^2 = 0.91) 1984-1993 (R^2 = 0.92) 1985-1994 (R^2 = 0.91) 1986-1995 (R^2 = 0.89) 1987-1996 (R^2 = 0.83) 1988-1997 (R^2 = 0.90) 1989-1998 (R^2 = 0.95) 1990-1999 (R^2 = 0.98)

The right balane for PV: 3 scenarios 1 2 3 Progress ratio 0.8 0.75 0.65 Growth rate 0.2 0.15 0.1 Results break-even year 2044 2045 2044 total learning investment (billions Euro) 1239 310 69 Room for extra R&D-spending (learning improvement) (billions Euro) 0 929 1169 Annual room for additional R&D-spending (learning improvement) 0 21.6 27.8 1974-1998 $ 5 miljard 1= matches history 2= relax market growth, more R&D 3= Strongly R&D-focused

Lessons Question of balance between R&D and deployment support can be addressed R&D is needed to at least sustain historical trend There is good reason to support also short-term R&D by public money, as long as the technology is supported by market deployment measures PV-case: grow a little slower, put some more R&D-money in

Thank you for having me here