THE CONTEXT FOR ENERGY STORAGE TO FACILITATE RENEWABLE ELECTRICITY IN MINNESOTA

Size: px
Start display at page:

Download "THE CONTEXT FOR ENERGY STORAGE TO FACILITATE RENEWABLE ELECTRICITY IN MINNESOTA"

Transcription

1 THE CONTEXT FOR ENERGY STORAGE TO FACILITATE RENEWABLE ELECTRICITY IN MINNESOTA JULY 2016 MARTHA HEWETT CONSULTANT JOSH QUINNELL, PH.D. CENTER FOR ENERGY AND ENVIRONMENT

2

3 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota Table of Contents Executive Summary... 1 Objective... 1 Findings from Past Renewable Integration Studies... 1 The 2014 Minnesota Renewable Integration and Transmission Study (MRITS)... 2 Other Renewable Integration Studies... 9 Current Penetrations of Renewables Strategies to Facilitate Integration of Renewables into the Electric System Transmission Balancing Area Geographic Extent and Generation Capacity Markets Forecasting of VG Output Advanced Control of VG Different Types of VG Low Load Flexibility Energy Storage Background and Basics Potential Roles of Energy Storage in the Electric System Characteristics of Storage Technologies Market and Policy Barriers to Increased Use of Storage Existing and Planned Storage on the MISO System MISO Studies and Activities Related to Storage MISO Energy Storage Study Phase 1, Energy Storage Study (MISO, Policy Studies), Manitoba Hydro Wind Synergy Study Addition of Short-Term Stored Energy Resource Pay-for-Performance Regulation Ramp Capability Product Current Work on Non-Transmission Alternatives Current Market Subcommittee Work on Storage References The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. i

4 List of Tables Table 1. MRITS Study Scenarios for 2028 (GE Energy Consulting 2014, Table 1-1, p. 1-3)... 2 Table 2. Incremental wind and solar MW to be added for each MRITS scenario (GE Energy Consulting 2014a Table 2-1)... 5 Table 3. Incremental wind and solar additions for Minnesota-centric and non-minnesota-centric areas (GE Energy Consulting 2014a, Table 3-1)... 5 Table 4. ISO/RTO Renewables 2014 (PJM 2015) Table 5. Multi-Value Project Status as of Q (MISO 2015b) Table 6. Wind and Solar Curtailment for MRITS Study Scenarios (GE Energy Consulting 2014) Table 7. MISO Capacity and Energy Output by Fuel (Potomac Economics 2015, p. 5) Table 8. MISO North+Central Wind and Solar Resources for MRITS Study (GE Energy Consulting 2014 p. 2-6) Table 9. MISO Fuel Diversity (PJM 2015) Table 10. DOE/EPRI Handbook Electric Grid Energy Storage Services (Akhil et al. 2013) Table 11. Storage Characteristics Needed for Various Electric System Functions (Akhil et al. 2013) Table 12. Primary Applications and Status of Electricity Storage Technologies (DOE 2013) Table 13. Cost and Performance Projection for 262 MW, 15 Hour Compressed Air Energy Storage Plant (Black & Veatch 2012) 49 Table 14. Cost and Performance Projection for 500 MW, 10 Hour Pumped Hydro Storage (PHS) Plant (Black & Veatch 2012) Table 15. Cost and Performance Projection for 7.2 MW, 8.1 Hour Sodium Sulfide Battery Storage Plant (Black & Veatch 2012) 50 Table 16. Cost and Performance Projection for 211 MW Gas Turbine Plant (Black & Veatch 2012) Table 17. Cost and Performance Projection for 580 MW Combined-Cycle Plant (Black & Veatch 2012) Table 18. Cost and Performance Projection for Onshore Wind (Black & Veatch 2012) Table 19. Overnight Capital Costs for Selected Storage Types in Various Applications (Akhil et al. 2013) Table 20. ISO Wholesale Power Costs, $/MWh 2014 (source: PJM 2015) Table 21. Projects in the DOE International Energy Storage Database Located in MISO Table 22. Storage Projects in MISO by Technology Type and Capacity (DOE) Table 23. Storage Projects in MISO by Technology Type and Capacity, Excluding Thermal Storage (DOE) Table 24. Pumped Hydro Storage Projects in MISO (DOE) Table 25. Lithium-Ion Battery Storage Projects in MISO (DOE) Table 26. Lithium-Ion Titanate Battery Storage Projects in MISO (DOE) Table 27. Sodium-Sulfur Battery Storage Projects in MISO (DOE) Table 28. Zinc Iron Flow Battery Storage Projects in MISO (DOE) Table 29. Flywheel Storage Projects in MISO (DOE) Table 30. Advanced Lead-Acid Battery Storage Projects in MISO (DOE) Table 31. Lead-Acid Battery Storage Projects in MISO (DOE) Table 32. Lead-Acid Battery Project Service/Use Case Information (DOE) Table 33. Lead-Acid Battery Storage Project Descriptions (DOE) Table 34. Electric and Plug-In Hybrid Vehicles in MISO's Footprint (MISO 2014g) Table 35. Average bidirectional ramp rate offered into MISO by resource type (MISO 2014h) Table 36. Average of monthly ratio between penalty and gross revenue before penalty for 2013 (MISO 2014h) p. ii The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

5 List of Figures Figure 1. MISO North/Central (green market footprint from GE Energy Consulting 2014a Figure 6-2; outlined area from information provided by MISO staff) Figure 2. Minnesota-Centric region for extraction of production simulation results (GE Energy Consulting 2014a Figure 7-1)... 8 Figure 3. Total curtailment as a function of usable wind energy penetration for different system flexibility factors (system with 60 GW peak demand) (Denholm & Hand 2011) Figure 4. Total curtailment as a function of VG energy penetration for different solar/wind energy mixes, assuming a 100% flexible system (Denholm & Hand 2011) Figure 5. Total curtailment as a function of VG energy penetration for different amounts of energy storage (or load shifting) with 80% round trip efficiency, for a 30/70 solar/wind energy mix and 100% flexible system. Each hour of storage represents one hour of average system demand (34.4 GWh). Denholm & Hand Figure 6. CAPX-2020 Project Status (CapX Quarterly, Winter 2015) Figure 7. MVP Portfolio (MISO 2012) Figure 8. North American ISO/RTOs (FERC 2016) Figure 9. MISO Market Area and Reliability Coordination Area Figure 10. U.S. Wind Speed at 100 m (NREL 2013) Figure 11. U.S. Solar Photovoltaic Resource (NREL 2012) Figure 12. Load Following Impact of Wind (DeCesaro & Porter, 2009) Figure 13. CAISO Graph of the Impact of Solar Generation on Load Shape and Ramping (so-called duck curve, cited in Lazar & Linvill 2016) Figure 14. MISO Average Hourly Wind Forecasting Accuracy (MISO 2015e) Figure 15. MISO Monthly Wind Generation 2015 (MISO 2016c, p. 34) Figure 16. MISO Monthly DIR Wind Energy Generation (MISO 2016c) Figure 17. Wind Curtailment by Hour of Day for Minnesota-Centric Region (GE Energy Consulting 2014) Figure 18. Approximate Ranges of Capacity and DTRP for Storage Technologies (Akhil et al. 2013) Figure 19. Rated Power of Current and Announced U.S. Grid Storage Projects Figure 20. Maturity of Electricity Storage Technologies (DOE 2013) Figure 21. ISO Wholesale Power Costs, $/MWh, 2014 (source: PJM 2015) Figure 22. Comparison of Estimated Net Revenue to Estimated Annualized Cost of New Entry, Midwest Region (Potomac Economics 2015) Figure 23. MISO Planned and Actual Reserves (PJM 2015, p. 175) Figure 24. Annual Load Duration Curve in 2033 with Medium Retirements, Low Construction Costs, Gas at $10/MMBtu Carbon at $0/ton (MISO 2014g) Figure 25. Major Salt Deposits in the U.S. (DeVries et al. 2005) Figure 26. Synergy between MISO wind production and MH-MISO interface flow (Bakke et al. 2014, p. 35) Figure 27. Gross regulation revenue, performance penalties, and net revenue (MISO 2014f, p. 24) Figure 28. Day-Ahead and Real Time Market Clearing Prices for Ancillary Services in 2015 (MISO 2016c) The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. iii

6

7 Executive Summary The objective of this paper is to assess the need for energy storage to facilitate increased penetration of variable renewable generation (VG) in the regional electric system that serves Minnesota. Thus it is as much about that system and electric systems in general as it is about electricity storage. Many renewable integration studies have found that the electric systems that were analyzed can accommodate targeted amounts of VG without operational problems and without the addition of storage. Most directly relevant is the Minnesota Renewable Energy Integration and Transmission Study (MRITS), ordered by the Minnesota Legislature to assess the reliability impacts and costs of increasing Minnesota s Renewable Energy Standard to 40% by 2030 and to higher levels thereafter. Completed in 2014, the study looked at Minnesota in the context of the Midcontinent Independent System Operator (MISO) North/Central region, which includes Minnesota. It found that the MISO system can be successfully operated for all hours of the year (no unserved load, no reserve violations, and minimal curtailment of renewable energy) with wind and solar resources increased to achieve 40% renewable energy in Minnesota and with current renewable energy standards fully implemented in neighboring MISO North/Central states. This scenario corresponds to an overall VG penetration in MISO North/Central of 15%. The study likewise found that [w]ith wind and solar resources increased to achieve 50% renewable energy in Minnesota and 25% renewable energy in MISO North/Central (10% above current renewable energy standards in neighboring states), with significant transmission upgrades and expansions in the five state area, the power system can be successfully operated for all hours of the year. 1 Energy storage was on the list of strategies to be considered if there were difficulties integrating the targeted amounts of variable renewables, but integration was not found to be a problem, so these strategies were not analyzed. Many other renewable integration studies have found that the electric systems that they analyzed can accommodate up to 35% VG. Although some of these studies discussed storage and even conducted some analysis of storage, none found that adding storage is necessary to achieve these penetrations. Some more general analysis conducted by National Renewable Energy Laboratory (NREL) scientists for the Electric Reliability Council of Texas (ERCOT) region suggests that at system-wide VG penetrations in the neighborhood of 50% to 80%, either large scale storage or load shifting or both may be necessary to keep curtailments of VG below 10% (low curtailment rates help to hold down total system costs and maintain adequate returns on investments in VG). The current VG penetration on the MISO North/Central system is 7.3%. Meeting all of the state renewable portfolio standards (RPS s) currently in place in MISO North/Central states would bring the penetration to 14%. An increase in Minnesota s RPS to 40% would bring the penetration in MISO North/Central to 15%. More aggressive standards in other states would be necessary to achieve substantially higher penetrations. Thus it will likely be some time before increased storage is required in MISO to facilitate increases in VG. Many attributes of the electric system can facilitate high levels of VG. These include sufficient transmission to enable delivery of VG from high resource areas to high load areas; large balancing areas 1 The study did not complete dynamic analysis for the second scenario, and stated that such analysis is necessary to ensure system reliability. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. ES-1

8 that have a diversity of load, VG and conventional generation; and markets that include both day-ahead (hourly) and real-time (sub-hourly) timeframes as well as both energy and ancillary services. Other valuable features include accurate forecasting of VG output, advanced control of VG itself, a balance of VG types (e.g., wind and solar), and conventional generation that is sufficiently flexible under low load conditions. The MISO system has many of these features. MISO member utilities, together with MISO, have planned and are in the process of building major expansions of the transmission system, for which a key objective is transmitting wind energy from high resource areas to high load areas. MISO is the largest ISO in geographic extent and the second largest in generation capacity in the U.S. It operates as a single, functionally consolidated balancing area and has both day-ahead and real-time (5 minute) markets for both energy and ancillary services. The real-time energy market provides a low-cost way to accommodate the higher rates of ramping of net load (load minus VG output) that can be caused by high penetrations of VG, without requiring a large amount of more expensive frequency regulation. In May 2016, MISO added a ramp capability product to its ancillary services market, which provides additional ability to deal with both predictable and unpredictable ramping of net load. Accurate forecasting of VG output reduces its uncertainty and thereby reduces the costs the system incurs due to committing and scheduling sufficient resources to compensate for uncertainty in net load. MISO uses both multiday and six-hour wind forecasts for this purpose. MISO requires wind and solar systems to meet minimum performance requirements that limit VG impacts on the system. It has also implemented a requirement that most wind plants operate under a Dispatchable Intermittent Resource (DIR) tariff that enables them to be dispatched down like other resources in response to price signals. This has greatly reduced manual curtailments and over-curtailments at times of transmission congestion or low load. In 2015, 6.4% of potential DIR wind generation was lost due to dispatch down. Based on the MRITS modeling, total curtailments of wind and solar output should be much lower by 2028 (wind 0.4% to 2.1% and solar 0.1% to 0.4%), even with substantial increases in VG, due to the transmission expansion in progress and planned retirements of baseload coal plants. MISO does have some characteristics that could potentially make it more difficult to integrate high levels of VG. MISO currently does not have very much solar generation, which would follow a different daily output pattern than wind. This is presumably due to the higher capacity factor of wind in the MISO footprint as well as the lower capital cost of wind. In addition, the MISO system s flexibility at low load is somewhat limited. MISO has a large amount of baseload coal capacity that is designated by its owners to operate on a must-run basis. While this does not result in much wind curtailment at present, it could be a more important factor at higher penetrations of VG. This was shown in the MRITS study, where most of the curtailment in Scenario 2 (50% VG Minnesota, 25% VG MISO) was due to system lowload operating limits at night. Energy storage can provide services at various levels within the electric system. The key applications that could potentially facilitate large increases in VG are bulk energy services (capacity, energy arbitrage), ancillary services (frequency regulation, ramp capability, contingency reserves), and transmission infrastructure services (upgrade deferral, congestion relief). Other resources can also provide these services, so the potential for storage in these applications will be determined by the relative economics. Energy storage can also provide services at the distribution level (upgrade deferral, voltage support) or at the customer level (reliability, power quality, etc.), but these applications either p. ES-2 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

9 are not related to renewable generation or are unlikely to influence VG penetration at scale in Minnesota any time soon. Storage technologies can be divided into long-term storage systems such as pumped hydro storage (PHS) or compressed air energy storage (CAES); medium-term systems, primarily various types of batteries; and short-term storage systems such as flywheels and some battery systems. These three types differ in their suitability for various electric system services. Long-term storage technologies can provide capacity, energy arbitrage, frequency regulation, ramp capability, contingency reserves, and transmission congestion relief or upgrade deferral. Short-term systems (storage less than an hour) cannot provide a significant amount of energy arbitrage and do not meet the North American Electric Reliability Corporation (NERC) minimum requirements for contingency reserves (90 minutes of operation following a contingency event), nor do they meet the minimum requirements to qualify as a capacity resource (four hours). They can provide frequency regulation and in some cases ramp capability (depending on their discharge time), and in general they have much faster ramp rates than conventional generation. Any medium-term storage would likely qualify for contingency reserves in addition to frequency regulation and ramp capability. Like short-term storage technologies, it would offer fast ramp rates. Some medium-term storage may have a discharge time long enough to qualify as a capacity resource or to provide a useful amount of energy arbitrage or congestion relief. The U.S. Department of Energy (DOE) identifies the key barriers to wider use of energy storage as costcompetitiveness, validated reliability and safety, an equitable regulatory environment, and industry acceptance. Energy storage is currently not cost-competitive with other alternatives in most applications, and this is particularly the case in low-cost markets such as MISO. However, it can be difficult to unambiguously quantify the value of storage, particularly when it provides multiple services. The processes and tools used by the electric industry to evaluate alternatives in system planning and to commit and dispatch resources in system operation are designed for more conventional system components (generation, transmission, and demand response) and not for storage. Thus they may not have the capability to optimize allocation of a storage system s capabilities among various services. This is particularly the case for medium-term storage systems, which can provide a broader set of services than short-term systems provide, but have more limited discharge times than long-term systems do. A further issue in Minnesota is that the planning and acquisition of generation and transmission assets are divided between the vertically integrated utilities and MISO, which affects both approval processes and valuation. It is worth noting that dedicated storage tied to a particular VG plant is likely to result in suboptimal economics and may therefore be attractive primarily as a way to obtain regulatory approval for storage as a novel resource. In the near term, the ability of other resources to manage VG and the high cost of storage mean that deployments of storage are unlikely to be a sound investment as a means to achieve large increases in the penetration of VG in Minnesota, beyond demonstration projects that allow utilities, MISO, and regulators to gain experience. However, it will be worthwhile to improve methods of valuation, modeling, commitment/dispatch, and regulatory treatment of storage in the interests of its longer term application when VG increases to much higher penetrations and storage prices decline. MISO has undertaken several projects and activities related to storage. A study was conducted in 2011, and repeated in 2014 (due to changes in the MISO footprint and membership and coal plant retirements), to assess the circumstances under which adding storage to MISO would be economically The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. ES-3

10 justified. The studies used Electric Power Research Institute s (EPRI s) EGEAS software, an integrated resource planning tool that allows analysis of many scenarios in a relatively short time. EGEAS is able to capture the capacity and energy arbitrage value of storage, but not the potential transmission or ancillary service benefits. The studies concluded that while large scale investment in storage is not economic under current conditions, investments in low-cost storage (CAES at $833/kW) could be economic under some of the future scenarios modeled. Under the more plausible of these, low-cost offpeak coal would be used to charge the storage unit, and higher-cost on-peak gas would be displaced by discharge from the storage unit. Since coal releases more carbon per unit of energy produced than gas and since the round-trip efficiency of any storage system is less than 100%, such a use of storage would increase carbon emissions. MISO undertook a much more in-depth study the Manitoba Hydro Wind Synergy study to assess the potential for hydro power from Manitoba to serve a storage-like function in mitigating the effects of variable renewables in MISO. The study analyzed the costs and benefits of expanding transmission capacity from the Canadian border into the MISO system. It was conducted using PLEXOS, a detailed production cost simulation model. MISO and the PLEXOS vendor developed a new Interleave functionality for this project, which alternates day-ahead and real-time simulations for a full year. This allowed modeled day-ahead bids to take into account the dispatch in the previous day s real-time market, which is critical for storage resources since they are energy-limited. Using this model, MISO was able to capture the value of Manitoba Hydro storage for ancillary services, transmission congestion relief, and energy arbitrage under realistic market conditions. Two system changes were found to be economic and were subsequently approved by the relevant regulatory bodies for implementation. One was a change in the use of the existing high voltage direct current (HVDC) transmission line between Minnesota and Manitoba from unidirectional to bidirectional, which allowed Manitoba Hydro to buy power from MISO where it previously could only supply power to MISO. The other was the addition of a new 500 kv transmission line from the Canada border into Minnesota. The modeling showed that, as intended, these modifications would cause Manitoba Hydro to reduce hydro generation when wind energy production was high and increase hydro generation when wind energy production was low. Minnesota Power is now building the 500 kv transmission line and has contracts with Manitoba Hydro to store up to a million MWh of wind energy per year by reducing hydro generation and return that energy to Minnesota Power when requested. The agreements also provide 250 MW of firm capacity to Minnesota Power. This arrangement is similar to a system that is used in Europe to coordinate use of Danish wind power with Swedish and Norwegian hydro power. In response to a Federal Energy Regulatory Commission (FERC) requirement that ISOs ensure participation of non-generation resources in markets on a fair and equitable basis, MISO developed a new tariff for short term stored energy resources (SERs). This tariff is focused on units with discharge times of five minutes to one hour that only provide frequency regulation. MISO developed an automated method to optimize SER dispatch such that the maximum amount of frequency regulation can be procured from SERs. They also developed a method to incorporate SERs into day-ahead unit commitment. In spite of this tariff, MISO s market for frequency regulation has thus far attracted almost no SER participation, presumably because of the relatively low market prices. Also in response to a FERC requirement, MISO developed a pay-for-performance tariff for frequency regulation. This tariff provides more compensation for fast-responding units (storage or others) that respond with greater speed and accuracy to automatic generation control (AGC) signals. This has p. ES-4 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

11 increased participation in the market by fast-ramping resources and decreased participation by slowramping resources. In May 2016, MISO added a ramp capability product to its ancillary service market. This was recommended by MISO s independent market monitor to help MISO manage fast ramps in net load at lower cost. Many types of resources, including storage, will be able to provide this service. MISO has a process to consider non-transmission alternatives (NTAs) in transmission planning, and MISO s Planning Subcommittee is currently discussing potential modifications to this process with stakeholders. An improved process could potentially increase opportunities for storage to serve as an NTA. MISO s Market Subcommittee is currently looking at storage because market participants are considering more alternatives to generation and because both storage technology and relevant aspects of market design have improved. Several market participants have or are planning battery storage projects that appear to be efforts to gain experience with the technology. A number of key issues in the treatment of storage are under discussion among stakeholders. In summary, additional electricity storage is not needed in the near term to substantially increase the penetration of variable renewable generation in Minnesota or the MISO North/Central region. Many other system attributes can achieve this objective at lower cost. While electricity storage could possibly help to achieve other policy objectives, these are beyond the scope of this study. For example, storage could be useful on the distribution grid to defer distribution upgrades otherwise necessitated by load growth or to provide increased reliability for customers. Since these applications do not relate to increased renewable generation they are outside the scope of this study. Likewise, storage could be helpful in integrating concentrated installations of renewable generation connected at the distribution level. If there is a sound policy rationale to increase the penetration of such distributed VG as opposed to utility-scale VG then storage could be one option to assist in integrating it. Such an application would be of value to achieve higher levels of VG in a particular way rather than to achieve high levels in the most cost-effective manner possible. Various distribution level applications of storage are currently being evaluated in Minnesota as part of activities such as Xcel Energy s Distribution Grid Modernization work. Recommendations based on this study are that: Minnesota should not provide incentives for energy storage or otherwise encourage energy storage with the intent of enabling increasing penetrations of renewable generation unless it is shown that (a) the electric system will have difficulty integrating the target levels of renewables without storage and (b) storage is economically competitive with other alternatives to manage these levels of renewables. Minnesota should support the concept of limited storage demonstration projects to gain experience with storage in transmission-level applications. Minnesota should support limited storage demonstration projects in distribution-level applications only if there is a policy rationale to encourage more expensive distributed VG as opposed to utility-scale VG or to provide distribution system benefits not related to VG. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. ES-5

12 Minnesota should support efforts to develop improved processes and tools that will help to more fully quantify the value of storage in electric system planning and optimize the use of storage in electric system operation. p. ES-6 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

13 Objective The objective of this paper is to assess the need for energy storage to facilitate increased penetrations of variable renewable generation (VG) in the regional electric system that serves Minnesota. From a policy perspective, CEE s primary interest in storage is its potential to facilitate substantially increased penetrations of cost-effective renewable electricity, not to promote storage per se nor to encourage particular types of renewable energy (e.g., solar PV vs. wind) or particular scales of renewable energy (e.g., distributed solar vs. utility-scale solar). Therefore, this report focuses on whether and how storage could be a cost-effective strategy to facilitate large increases in renewable generation. Many good sources summarize the types of storage available, their characteristics, technology trends, costs, and other related information. Given project time constraints, little time was devoted to resummarizing this information. The project did not examine other possible uses of storage. For example, storage located on the distribution grid could be useful to defer distribution upgrades otherwise necessitated by load growth, or to provide reliability or demand charge management for individual customers. These applications do not relate to increased renewable generation and are therefore outside the scope of this study. Likewise, storage could be helpful in integrating concentrated installations of renewable generation that is connected at the distribution level. If there is a sound policy rationale to increase the penetration of distributed VG as opposed to utility-scale VG then storage could be one option to assist in integrating that VG. Such an application would be of value not to facilitate high levels of VG per se, but rather to achieve high levels of VG in a particular way. Various distribution level applications of storage are currently being evaluated in Minnesota as part of activities such as Xcel Energy s Distribution Grid Modernization work. Findings from Past Renewable Integration Studies Renewable integration studies are the best source of insights into the need for storage to increase renewable generation. These studies assess the operational impacts and cost of adding targeted amounts of variable renewable generation (VG) to specific electric systems by some future date. They analyze operational impacts using production simulation software, with simulation results providing such performance metrics as annual energy production by type of generating resource, renewable energy resource utilization and curtailment, cycling duty of thermal plants, adequacy of ramping capability of the generation fleet, and risk of reserve violations and unserved load (GE Energy Consulting 2014a). 2 Integration studies may also assess options to resolve any operational problems identified through the simulations. Costs analyzed typically include the capital costs of new VG required to meet targets, capital costs of additional conventional generation required to meet future load, capital costs of new transmission, operating costs attributable specifically to VG integration, 3 and all other production costs. The operational performance and costs of targeted high-vg scenarios are compared with those for a reference, business-as-usual scenario. Renewable integration studies typically have a 2 The MRITS study also included transient stability analysis (dynamic analysis) under challenging operating conditions identified through the production simulation, but this is not typically done as part of the same study. 3 Costs due to increased needs for unit commitment, load following and regulation owing to the variability, and uncertainty in VG output. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 1

14 study team that includes a number of organizations and a technical review committee that includes a larger and broader group. Renewable integration studies have found that the systems studied can accommodate targeted amounts of VG up to about 35% of retail sales without insurmountable operational problems, which is considerably higher than the current 7.3% penetration in MISO North/Central. 4 Higher penetrations have not been analyzed. Although some of these studies have discussed or even conducted some analysis of storage, none has found that additional storage 5 is necessary to achieve these penetrations of VG. The 2014 Minnesota Renewable Integration and Transmission Study (MRITS) The Minnesota Renewable Energy Integration and Transmission Study (MRITS) (GE Energy Consulting 2014a) was ordered by the Minnesota Legislature to assess the reliability impacts and costs of increasing Minnesota s Renewable Energy Standard to 40% by 2030 and to higher levels after that. The research team studied the expected baseline configuration and two increased-vg scenarios for 2028, as shown in Table 1. Table 1. MRITS Study Scenarios for 2028 (GE Energy Consulting 2014, Table 1-1, p. 1-3) The Baseline scenario was defined to have enough renewable generation to meet current renewable and solar energy standards for all states in MISO. (Note that MISO here is defined as MISO North plus MISO Central (see Figure 1) and does not include MISO South 6 (GE Energy Consulting 2014a, p. 2-6), which only joined MISO near the end of 2013.) Scenario 1 increased wind and solar generation to 40% of retail sales in Minnesota, but left the other states at the same levels as in the Baseline. It thus increased the MISO VG penetration in 2028 by only 1%, from 14% to 15%. Scenario 2 increased Minnesota s renewable energy to 50% and increased other states wind and solar energy enough to bring the total within MISO to 25%. The study found that: [W]ith upgrades to existing transmission, the power system can be successfully operated for all hours of the year (no unserved load, no reserve violations, and minimal curtailment of renewable energy) with wind and solar resources increased to achieve 40% renewable energy in Minnesota and with current renewable energy standards fully implemented in neighboring MISO North/Central states. 4 See section titled Current Penetrations of Renewables for further information. 5 Some of the systems studied including MISO already have some storage, mostly pumped hydro. 6 With the exception of the portion in Texas, most of MISO South is in states that do not have renewable portfolio standards (namely MO, LA, & MS). p. 2 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

15 With wind and solar resources increased to achieve 50% renewable energy in Minnesota and 25% renewable energy in MISO North/Central (10% above current renewable energy standards in neighboring states), MRITS production simulation results show that, with significant transmission upgrades and expansions in the five state area, the power system can be successfully operated for all hours of the year (no unserved load, no reserve violations, and minimal curtailment of renewable energy). Due to study schedule limitations, no dynamic analysis was performed for 50% renewable energy in Minnesota (Scenarios 2 and 2a) and this analysis is necessary to ensure system reliability. One of the MRITS study objectives was to identify and develop options to manage the impacts of renewable energy resources. According to Bill Grant, Deputy Commissioner of Division of Energy Resources at the Minnesota Department of Commerce (pers. comm.), energy storage was on the list of strategies to be considered if there were difficulties integrating the targeted amounts of variable renewables, but integration was not found to be a problem so these alternatives were not analyzed. MISO North/Central Figure 1. MISO North/Central (green market footprint from GE Energy Consulting 2014a Figure 6-2; outlined area from information provided by MISO staff 7 ). Given the central relevance of the MRITS findings to the issue of energy storage in Minnesota, it is worthwhile to understand MRITS in more depth. The MRITS study steps were to: Develop study scenarios and site the required wind and solar generation; Perform power flow analysis and develop conceptual transmission plans; Perform production simulation analysis; 7 According to MISO staff, MISO North and Central is the same as what is described in many other MISO documents as MISO West, East and Central. See for example the map in MTEP 15 Chapter 2.1. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 3

16 Review production simulation results in order to evaluate operational performance, including violations of required generating reserves, unserved load, wind or solar curtailments, and cycling and ramping of conventional generating units; Screen for challenging periods; Evaluate system stability issues including transient stability, voltage regulation, and adequacy of dynamic reactive support, as well as weak system strength issues (completed only for Scenario 1); and Identify and develop mitigations and solutions, if necessary. Minnesota utilities and the Minnesota Department of Commerce took the lead in developing study scenarios and siting the wind and solar generation. To determine the amount of wind and solar capacity required to meet the scenarios in Table 1, it was first necessary to project retail electricity sales (GWh) for 2028, then to compute the wind plus solar energy (GWh) required to meet the specified percentages of sales, and finally to determine the plant capacity (MW) required to produce that much energy. Retail sales estimates for 2028 were computed assuming load growth of 0.5%/year for Minnesota and 0.75% per year for MISO (GE Energy Consulting 2014a, p. 2-5). The total renewable energy (GWh) required was then computed based on the target percentages of renewables for each scenario. In order to determine the total amount of wind and solar plant capacity required to provide this energy, the following assumptions were made about the wind/solar split, the utility-scale solar vs. distributed solar split, and the annual capacity factors 8 for wind and solar (GE Energy Consulting 2014a, p. 3-1 & 3-2): For Minnesota, wind capacity as a percentage of total wind + solar capacity is assumed to be about 92% in the Baseline scenario, 85% in Scenario 1, and 64% in Scenario 2. 9 For MISO, wind capacity as a percentage of total wind + solar capacity is assumed to be about 94% in the Baseline scenario, 91% in Scenario 1, and 81% in Scenario 2. According to Bill Grant (pers. comm.) the split between wind and solar in each scenario was established in an open stakeholder process that developed the overall scope for the study prior to assembling the study team. For the Minnesota Baseline scenario and Scenario 1, it was assumed that 80% of solar capacity would be utility-scale PV and 20% would be residential and commercial distributed PV. For Minnesota Scenario 2, it was assumed that 85% would be utility-scale and 15% residential or commercial. These values exceed the 10% distributed PV requirement included in the 1.5% solar mandate enacted by the Minnesota legislature in For MISO, it was assumed that 90% of solar would be utility-scale and 10% would be residential or commercial for all three scenarios. According to Bill Grant (pers. comm.) the utility-scale PV vs. distributed PV split was set by the study team and reviewed by the Technical Review Committee. It was guided by statutory requirements in the Minnesota solar energy standard. 8 The annual capacity factor is the ratio of actual output over the year to the output that would be produced if the unit could operate at its full nameplate capacity for all hours of the year. 9 Since wind has a much higher capacity factor than solar in this region, wind accounts for even higher percentages of the combined wind + solar energy output. Based on the total MW and capacity factors given, wind accounts for about 96% of combined wind + solar output in the Baseline scenario, 93% in Scenario 1, and 81% in Scenario 2. p. 4 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

17 For Minnesota, wind was estimated to have an average capacity factor of 38% in the Baseline scenario and 42% in Scenarios 1 and 2. Utility-scale PV was estimated to have an average capacity factor of 18% and distributed PV an average capacity factor of 17% in all three scenarios. For MISO, wind was estimated to have a 37% capacity factor for all three scenarios. Utility-scale PV was estimated to have a capacity factor of 17% and distributed PV a capacity factor of 16% for all three scenarios. The resulting amounts of wind and solar capacity (MW) added to the Minnesota-centric area and to MISO North/Central in the various scenarios are shown in Table 2. Table 3 separates the incremental wind and solar capacity additions from the second half of Table 2 into Minnesota-centric and non- Minnesota-centric portions. Table 2. Incremental wind and solar MW to be added for each MRITS scenario (GE Energy Consulting 2014a Table 2-1) Table 3. Incremental wind and solar additions for Minnesota-centric and non-minnesota-centric areas (GE Energy Consulting 2014a, Table 3-1) The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 5

18 The MRITS report provides considerable detail on the assumptions that were made regarding wind and solar siting (see GE Energy Consulting 2014a, chapter 3). In general, Minnesota wind and solar resources were sited in the Minnesota-centric area (MN, ND, SD, northern Iowa) based on existing wind and solar, planned wind and solar (including those with signed Interconnection Agreements, wind sites in MVP portfolio planning), and MN utility announced projects. Wind and solar resources in the interconnection queues also helped inform the siting selection process. MISO future wind and solar was sited per MTEP guidelines (e.g. at expanded RGOS 10 zones on a pro rata basis). All of the baseline Minnesota-centric wind and 69% of the baseline MISO North/Central wind (including Minnesota) was sited based on existing wind plants plus signed Generation Interconnection Agreements (GIAs). The remaining baseline 6,900 MW of non-minnesota-centric wind was distributed across 29 of the wind zones defined in the Regional Generation Outlet Study 11 (RGOS) in 10 states (GE Energy Consulting 2014a, Table 3-4, Figure 3-1). An additional 1,931 MW of Minnesota-centric wind capacity was required for Scenario 1 (beyond that in the Baseline). Of this, 750 MW (39%) was expected to come from four wind plants in four separate RGOS zones for which Xcel Energy had recently signed GIAs. The remainder was divided across 12 other RGOS zones in Minnesota, North Dakota, South Dakota, and Iowa (GE Energy Consulting 2014a, Table 3-5, Figure 3-2). For Scenario 1, no wind was added beyond the Baseline to serve non-minnesota portions of MISO. An additional 610 MW of Minnesota-centric wind capacity was required for Scenario 2 (beyond that in Scenario 1). This was divided across eight RGOS zones in Minnesota, North Dakota, South Dakota, and Iowa. A very substantial 13,026 MW of additional non-minnesota-centric MISO wind was required for Scenario 2. This was divided over 32 ROGS zones in 10 states (GE Energy Consulting 2014a, Table 3-7, Figure 3-3). Production simulation analysis with the initial configuration showed significant transmission congestion for some RGOS sites in the western portion of MISO. To mitigate this, some of the wind capacity serving non-minnesota loads was moved from the four most congested wind zones to the 10 least congested zones (GE Energy Consulting 2014a, Table 3-9, Figure 4). The less-congested zones generally had lower capacity factors so the 851 MW of wind moved from the four congested zones had to be replaced by 947 MW of wind in the 10 least congested zones in order to provide the same annual energy output. Based on the solar resource, the utility-scale solar PV in Minnesota was distributed largely over the southern half of the state. However, since some utilities in the northern part of the state wanted PV sited closer to their own service territories, some PV was located in northern Minnesota, even though the solar resource is about 10% lower. General locations and the transmission lines to which each plant would be connected were identified (GE Energy Consulting 2014a, Table 3-10, Figure 3-6 and 3-7). The distributed PV in Minnesota was assumed to be sited at various load centers since this is where most residential and commercial customers are located (GE Energy Consulting 2014a, Table 3-11, Figure 3-8). 10 Regional Generation Outlet Study (MISO 2010). See section titled Transmission for further information on the RGOS. 11 Ibid. p. 6 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

19 The non-minnesota MISO solar was located at the same locations used in NREL s Eastern Renewable Generation Integration Study (ERGIS an NREL study in progress). Utility-scale solar was distributed over seven states (GE Energy Consulting 2014a, Table 3-12). Distributed solar was located in 25 different cities in six states (GE Energy Consulting 2014a, Table 3-13). Conceptual transmission plans were developed for each study scenario using traditional transmission planning processes. More details on the role of transmission in increasing penetration of VG, transmission expansion in MISO over the past 10+ years, and additional transmission required to accommodate MRITS scenarios is given in the Transmission section. Production simulation analysis was used to analyze the operational performance of the power system with the levels of renewable generation in each of the three scenarios. MISO conducted this analysis using the commercial PLEXOS simulation tool, which simulates hourly system operation. The particular implementation they employed interleaves day-ahead security constrained unit commitment and realtime security constrained economic dispatch on a daily basis throughout a full year, in order to capture the impact of forecast uncertainties on differences between day-ahead and real-time markets. This is particularly important in modeling performance and cost for systems with high penetrations of variable renewable generation. The wind and solar resource data used in production simulation were taken from data developed by NREL for ERGIS. The solar data for Minnesota were refined to 10 km resolution by NREL. The simulation model footprint includes all areas in the Eastern Interconnect, with the exception of Florida, ISO New England and Eastern Canada (GE Energy Consulting 2014a p. 1-5). However, since the focus of the study was on Minnesota, modeling results were extracted and summarized in the report for the Minnesota-Centric Region (see Figure 2) consisting of "all generating resources operated by and system loads served by the Minnesota utilities" (GE Energy Consulting 2014a p. 7-1). The production simulation provides many of the most important study results. In particular, The production simulation results were analyzed to assess system operational performance with respect to the following parameters; annual energy production by type of generating resource, renewable energy resource utilization and curtailment, cycling duty of thermal plants, adequacy of ramping capability of the MISO generation fleet, and risk of reserve violations and unserved load. For Scenario 1, the results were also screened to select challenging operating conditions for dynamic performance, and these operating points were subsequently analyzed with fault simulations in the dynamics task. ((GE Energy Consulting 2014a, p. 1-5). According to study team members, the operational findings are not particularly sensitive to the details of wind and solar plant siting. Rather, the findings are a realistic estimate of operational performance as long as good assumptions were made about the spread of the siting, since smoothing of renewable energy output occurs when plants are distributed over a large geographical area. According to Bill Grant (pers. comm.), the MRITS study had unprecedented engagement by the utilities, including both resource planning staff and transmission planning staff, so there is good reason to believe that the distribution of renewable plants in the study reasonably approximates the distribution that will actually occur. 12 Thus it is likely that the operational conclusions are robust to the level of divergence between the modeled and actual distributions of wind and solar plants that are likely to occur. 12 The vast majority of generation in Minnesota is added through the utilities integrated resource planning (IRP) processes (rather than as merchant plants that do not have a contract to sell their power to a particular utility). The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 7

20 Figure 2. Minnesota-Centric region for extraction of production simulation results (GE Energy Consulting 2014a Figure 7-1) Likewise, the MRITS transmission costs for Scenario 1 are likely reasonably good estimates. According to study team members, relatively little additional expenditure was required for transmission upgrades in Scenario 1 because plants were sited to populate the CapX and MVP transmission that has already been built forward in anticipation of future needs and future generation locations (see the Transmission section for more on CapX and MVP). Future wind and solar generation is highly likely to be located near this existing transmission in order to be cost-competitive. Even when utilities go through a competitive bidding process to acquire generation and so do not have direct control of power plant location, they require interconnection costs or transmission delivery costs to be included in the bid. Bids will only be competitive if plants are sited where they have low cost access to transmission. 13 Likewise, developers have an economic incentive to avoid sites with congested transmission. Anyone building a power plant must get that plant into the generation interconnection queue, and as part of that process they learn whether they will be able to get network transmission service (all energy generated can be delivered) or energy-only service (likely unable to deliver some of the energy that could be generated at times of 13 As an example, several solar plants with capacities in the hundreds of megawatts were just approved for Xcel Energy. All were acquired through competitive bidding and all be sited near existing transmission lines. p. 8 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

21 congestion) at the planned location. If output is likely to be curtailed due to congestion, the project will only proceed if the savings from using the site exceeded expected curtailment losses. For Scenario 2, a substantial amount of new transmission would be needed. Production simulation showed no problems with operational performance, assuming this transmission is added. However, given the substantial transmission expenditure required and the fact that stability analysis was not conducted, the MRITS team concluded that more study would be needed before proceeding to this level of renewables. To summarize, MRITS found that Minnesota could obtain 40 or 50% of its electricity from renewables without operational problems and without additional electricity storage, and it is likely that this finding would hold even if the actual siting of renewables ends up being somewhat different than that in the study. Thus promotion of storage or provision of incentives for storage is not necessary to achieve these levels of renewables. Note that the MRITS study, like other renewable integration studies, focused on the ability of the regional power system to operate at the targeted levels of renewables without significant curtailment of renewable plant output, reserve violations, unserved loads, inadequate ramping to follow net loads, or excessive cycling of thermal power plants. If energy policy strongly favors distributed generation, or if many individual customers decide to install their own renewable generation, this will lead to a large amount of variable renewable generation connected to the distribution system (rather than the transmission system) and possibly to large concentrations of VG on specific distribution lines. If this occurs, there could be local problems integrating this distributed VG without storage or other accommodations. Such problems are not impediments to achieving high levels of renewables, but rather to achieving them in a particular way, namely through distributed generation (DG). In general, distributed VG is more expensive than utility-scale VG (Black & Veatch 2012), so a policy preference for distributed VG must be driven by considerations other than achieving a target level of renewables. Other Renewable Integration Studies The Eastern Wind Integration and Transmission Study (EWITS) was completed by EnerNex Corporation 14 (2011) for the National Renewable Energy Laboratory (NREL), with funding from DOE. The study assessed the feasibility of incorporating 20 to 30% wind energy into the entire Eastern Interconnection 15 of the U.S. by It concluded that these wind penetrations are technically feasible with substantial expansion of transmission and that the costs for this transmission make up a relatively small portion of the total annualized costs in any of the scenarios studied. It further concluded that wind integration costs are manageable with large regional operating pools and significant market, tariff and operational changes. 16 With those changes, integration costs are measurable but very small relative to other factors. (Transmission and system/market attributes that facilitate integration of VG are discussed in the context of MISO in the section on Strategies to Facilitate Integration of Renewables into the Electric System.) The study noted that with the transmission and market assumptions, large 14 The study team included both Ventyx and MISO as well as EnerNex. 15 The Eastern Interconnection includes the U.S. from the Dakotas, Nebraska, Kansas, and Oklahoma eastward. 16 Integration costs were determined by running comparative production simulations in which the variability and short-term uncertainty of wind was eliminated. The integration cost is the difference in total production costs for simulations with and without this variability and uncertainty. It does not include transmission costs. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 9

22 amounts of wind generation can be accommodated without deploying additional energy storage resources. They note that while bulk storage might obviate the need for some of the transmission and reduce wind integration impacts, determining this would require analysis of the relative costs and benefits. The Western Wind and Solar Integration Study (WWSIS), Phase 1, was conducted by GE Energy (2010a) and others for NREL with funding from DOE. It investigate[d] the operational impact of up to 35% energy penetration of wind, photovoltaics (PVs), and concentrating solar power (CSP) on the power system operated by the WestConnect group of utilities in Arizona, Colorado, Nevada, New Mexico, and Wyoming. The study modeled the entire Western Interconnection and assumed that penetrations of wind and solar in the area outside of WestConnect were up to 23%. It concluded that, it is operationally feasible for WestConnect to accommodate 30% wind and 5% solar energy penetration, assuming the following changes to current practice could be made over time: Substantially increase balancing area cooperation or consolidation, real or virtual; Increase the use of sub-hourly scheduling for generation and interchanges; Increase utilization of transmission; Enable coordinated commitment and economic dispatch of generation over wider regions; Incorporate state-of-the-art wind and solar forecasts in unit commitment and grid operations; Increase the flexibility of dispatchable generation where appropriate (e.g., reduce minimum generation levels, increase ramp rates, reduce start/stop costs or minimum down time); Commit additional operating reserves as appropriate; Build transmission as appropriate to accommodate renewable energy expansion; Target new or existing demand response programs (load participation) to accommodate increased variability and uncertainty; and Require wind plants to provide down reserves. These strategies are discussed in the context of MISO in the Strategies to Facilitate Integration of Renewables into the Electric System section. The WWIS examined several issues related to storage. First, it found that the production simulations that increased renewables also increased use of existing (pumped hydro) storage slightly, but not to the point where more storage was needed on the system. WWSIS evaluated only the price arbitrage part of the value proposition for PSH and found it much less than sufficient to economically justify additional storage facilities. They explain why this is the case: They note: In the 10% and 20% wind penetration scenarios, gas generation is always on the margin (meaning that there are only small spot price variations during most days).as a result, there is no apparent opportunity to economically justify energy storage based on price arbitrage. Spot price variations increase in the 30% wind penetration scenarios, primarily due to errors in day-ahead wind energy forecasts. Occasionally, the price swings are very large. However, because this is driven by forecast uncertainty, it is not possible to strategically schedule the use of storage resources to take advantage of the price variations (and subsequently help eliminate the operational problems due to wind forecast errors). The best way to integrate wind and solar generation into the system is to make full use of the capabilities of all of the generating units in the system. Although the flexible operation of storage is attractive, when gas fired generation in [sic] on the margin both on peak and off peak there is little economic room for using the storage in an energy arbitrage manner due to the 25% losses associated with pumped storage hydro or batteries. As long as balancing area issues don t interfere, utilizing the ramping capabilities of system s [sic] dispatchable generation is generally more efficient. p. 10 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

23 The concentrating solar power plants accounted for 3 or 4% of renewable energy generation (out of 5% total for solar), and those plants did include storage. The collector field was oversized relative to the steam turbine and generator, and the additional heat was stored to allow the unit to continue producing electricity after the sun went down. They determined that, assuming perfect foreknowledge of prices, the annual revenue of the CSP plants could be increased by 10 to 12% if thermal energy were stored and electricity generated at times of day with higher prices, rather than simply used as it became available except for the amount that needed to be stored due to the intentionally undersized turbine and generator capacity limits. With regard to the use of storage for ancillary services, they point out that for this study area there were: substantial reserve requirements during periods when relatively limited amounts of flexible generation are on line. Unlike energy storage for arbitrage, energy storage technologies that can provide reserve services, both regulation and sub hourly spinning reserve, may be economically attractive. The economics of energy storage for these services must compete with the costs, both operational and opportunity, of providing these services with conventional generation. This is grounds for further investigation. Finally, they consider the use of plug-in hybrid electric vehicles (PHEV) not as storage, but as an off-peak controllable load that could have synergies with wind generation: Adding PHEVs to a system makes it more attractive to wind generation and adding high penetrations of wind reduces the cost of charging PHEV. In sum, the WWIS did not find that additional storage was necessary to integrate large amounts of renewables. Under the study conditions, storage could possibly compete with other system resources to provide ancillary services, but was unlikely to be competitive based on its energy arbitrage value alone. The New England Wind Integration Study (NEWIS) (GE Energy 2010b) was conducted to determine the operational, planning and market impacts of integrating substantial wind generation resources for the New England Balancing Authority Area, with due consideration to the neighboring areas, as well as, the measures that may be available to ISO-NE for mitigating any negative impacts while enabling the integration of wind. It looked at scenarios in which wind energy would meet 2.5 to 24% of annual energy needs in about The study results show that New England could potentially integrate wind resources to meet up to 24% of the region s total annual electric energy needs in 2020 if the system includes transmission upgrades comparable to the configurations identified in the Governors Study. It is important to note that this study assumes (1) the continued availability of existing supply-side and demand-side resources as cleared through the second FCA (in other words, no significant retirements relative to the capacity cleared through the second FCA), (2) the retention of the additional resources cleared in the second Forward Capacity Auction, and (3) increases in regulation and operating reserves as recommended in this study. The NEWIS found relatively small increases in the use of existing pumped-hydro storage (PHS) for large wind penetrations in the ISO-NE area, because balancing of net load was largely provided by the flexibility of the natural-gas-fired generation fleet. The production cost modeling did not find that the economics supported substantially increased use of storage, even with existing storage facilities: The lack of a price signal to increase use of energy storage is the primary reason the study showed small increases in the use of pumped-storage hydro in the higher wind penetrations. For energy arbitrage applications, like pumped storage hydro, a persistent spread in peak and off-peak prices is the most critical economic driver. The differences between on-peak and off-peak prices were small because natural-gas-fired The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 11

24 generation remained on the margin most hours of the year. Over the past six years, GE has completed wind integration studies in Texas, California, Ontario, the western region of the United States, and Hawaii. In many of these studies, as the wind power penetration increases, spot prices tend to decrease, particularly during high priced peak hours. The off-peak hours remain relatively the same. Therefore, the peak and offpeak price spread shrinks and no longer has sufficient range for economic storage operation (p. 25). This finding of decreasing peak-off-peak price spreads is consistent with some of MISO s early research on storage, discussed in the section titled MISO Energy Storage Study Phase 1, The PJM Renewable Integration Study (PRIS) (GE Energy Consulting 2014b) was conducted to [d]etermine, for the PJM balancing area, the operational, planning, and energy market effects of largescale integration of wind and solar power as well as mitigation/facilitation measures available to PJM. It looked at scenarios in which wind and solar provide 14 to 30% of PJM electricity by The study findings indicate that the PJM system, with adequate transmission expansion and additional regulating reserves, will not have any significant issues operating with up to 30% of its energy provided by wind and solar generation. With respect to storage, they say (p. 47): There is a growing industry trend to use energy storage and demand response resources as an alternative to generation resources for spinning reserves. This study considered a case where 1000 MW of storage or demand response resources were used in place of generator resources for spinning reserves in the 30% LOBO [low offshore and best onshore] scenario. Total system production costs were reduced by $17.41M/year, which corresponds to $1.99/MWh or $17.41/kW-year. Energy storage resources are emerging as viable contributors to regulation reserves in some operating areas where the market prices of regulation services are adequate to make the capital investment worthwhile. This [is] especially true in markets where the inherent fast-ramping capability of some storage technologies is financially rewarded (e.g., a mileage charge). In fact, some storage resources are already participating in PJM s regulation market. However, this study did not include economic assessment of the regulation market in PJM, so no specific conclusions can be drawn with respect to the economic competitiveness of energy storage devices as regulation resources in PJM as renewable penetration increases. The market price of regulation and the capital costs of energy storage devices will ultimately dictate viability. Thus, again, the PRIS found that storage was not necessary to integrate the target amount of renewables. For spinning reserves the study found that using storage rather than generation would reduce production costs but did not assess whether this would generate enough revenue to make the storage cost-effective. Nor did it evaluate the economics of storage for regulation in PJM. 17 DeCesaro & Porter (2009) reviewed 13 earlier wind integration studies that contemplated capacity penetrations of 3.5 to 30% including several that focused on Minnesota and concluded that there are no fundamental and insurmountable obstacles when incorporating greater levels of wind power into the U.S. grid assuming that appropriate strategies are adopted. Regarding storage, they concluded, It will be many years before the levels of wind generation will be significant enough in the United States [that] storage may be needed, and in that time, changes in resource mix, market rules, and the other factors discussed earlier may ease large-scale wind integration without the need for storage technologies. Consistent with this time frame, several storage technologies are at an advanced R&D or demonstration stage and are not available at a large scale currently. That said, storage systems have recently been installed and/or tested in PJM, NYISO, CAISO, and ISO New England, as well as by American Electric Power and Golden Valley Cooperative in Alaska. These systems were not installed to back up wind, but to provide 17 Note that prices in the PJM ancillary services market are considerably higher than those in MISO. p. 12 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

25 essential grid services such as regulation and voltage support. Indeed, storage should be viewed as providing grid services, not for supporting any individual plant or technology. This last point is echoed by Sioshansi et al. (2012), who note that the idea of firming variable renewable generator with storage is: something of a canard, in our opinion, since this service would likely be procured in some combination of balancing energy and AS [ancillary services] markets and would not be a separate storage application. Indeed, directly coupling storage to a particular renewable generator can yield inefficiencies, since other resources (including other renewables) may be able to counterbalance this variability to some extent. Likewise, a report from a DOE (2011) workshop of experts and stakeholders concluded that storage facilities dedicated to storing energy from a particular VG plant tend not to be economically optimal: It is worth noting, that the economics of combining variable renewable plants with dedicated energy storage tend to be sub-optimal, since the value of the curtailed energy plus the avoided cost of incremental transmission capacity is typically not sufficient to cover storage costs and the lost opportunity of providing other grid services. The renewable integration studies discussed above are prospective modeling studies. However, actual experience with substantial wind penetrations outside the U.S. has not shown insurmountable operational problems, either (Ackermann et al. 2009). In a white paper, the American Wind Energy Association (AWEA 2015) discussed at length their view that storage is not necessary and is perhaps the most expensive option to provide the flexibility needed to integrate wind: Some of the most common questions about wind power involve the role of energy storage in integrating wind power with the electric grid. It is important to understand that very large amounts of wind energy can be reliably integrated at low cost without a need for energy storage, and that energy storage provides a variety of services and is therefore best viewed as a power system resource and not a resource for wind energy or any other individual resource. Moreover, energy storage is typically a more expensive source of flexibility than grid operating reforms that allow greater use of the flexibility that already exists on the power system today. The reality is that, while several small-scale energy storage demonstration projects have been conducted, the U.S. has been able to add more than 65,000 MW of wind power to the grid without adding any largescale energy storage. Similarly, European countries like Denmark, Spain, Ireland, and Germany have successfully integrated very large amounts of wind energy without having to install new energy storage resources. In the U.S., numerous peer-reviewed studies have concluded that wind energy can provide 30% or more of our electricity without any need for energy storage. As noted earlier, none of the renewable integration studies to date have looked at very high penetrations of renewables. It may be that storage will be needed to integrate large percentages of VG economically. Denholm and Hand (2011) conducted a simplified study of extreme VG penetrations up to 80% of electricity on the ERCOT system in Texas. They note that at penetrations above those in most renewable integration studies to date, a primary constraint becomes the simple coincidence of renewable energy supply and demand for electricity, combined with the operational limits on generators providing baseload power and operating reserves. This may present an economic upper limit on variable renewable penetration without the use of enabling technologies. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 13

26 They analyzed various scenarios using a reduced form dispatch model that compares VG supply with demand and calculates the fraction of load potentially met by VG considering flexibility constraints [of the conventional generation] and curtailment [of the VG]. [This model] does not commit individual thermal units based on generator operating constraints. Instead it evaluates the ability of an entire system to accommodate VG based on its aggregated system minimum generation level. This allows for a general understanding of the system flexibility needs of many different combinations of VG, as opposed to a detailed technical and economic evaluation of any particular scenario. The system minimum is an input to the model based on a fraction of system peak, representing the limits of both baseload generators and generators that must remain online to reliably meet the variability and uncertainty of the net load. This minimum load constraint can also be expressed more generally as the system s flexibility factor, which is defined as the fraction below the annual peak to which conventional generators can cycle. A 0% flexible system would be unable to cycle below annual peak load at all, while a 100% flexible system could cycle down to zero load. In these simulations, the amount of must-run generation was based on fixed levels to examine sensitivity to different levels of system flexibility. The study found that: a highly flexible system with must-run baseload generators virtually eliminated allows for penetrations of up to about 50% variable generation with curtailment rates of less than 10% [although relative system costs increase due to curtailments]. For penetration levels up to 80% of the system s electricity demand, keeping curtailments to less than 10% requires a combination of load shifting and storage equal to about one day of average demand [Figure 3 - Figure 5]. According to Denholm (pers. comm. April 19, 2016), the 10% figure was simply chosen as one example to illustrate the concept. The full range of results can be read from the figures. Figure 3. Total curtailment as a function of usable wind energy penetration for different system flexibility factors (system with 60 GW peak demand) (Denholm & Hand 2011) p. 14 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

27 Figure 4. Total curtailment as a function of VG energy penetration for different solar/wind energy mixes, assuming a 100% flexible system (Denholm & Hand 2011) Figure 5. Total curtailment as a function of VG energy penetration for different amounts of energy storage (or load shifting) with 80% round trip efficiency, for a 30/70 solar/wind energy mix and 100% flexible system. Each hour of storage represents one hour of average system demand (34.4 GWh). Denholm & Hand The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 15

28 In summary, studies to date have shown that VG penetrations up to about 30 to 35% of electricity generation can be integrated into electrical systems without additional storage, assuming sufficient build-out of transmission and operational and market changes. Additional storage for energy arbitrage appears likely to have poor economics at these penetrations. 18 While storage for regulation or spinning reserves could turn out to be competitive in high cost markets, it is not the only option to meet these needs. At very high VG penetrations, storage and/or load-shifting may be necessary to minimize curtailments of VG output, thereby enhancing profitability of VG plants and holding down total system costs, particularly if the remainder of the generation on the system does not have a very high flexibility factor. 18 See the MISO Studies and Activities Related to Storage section for more on this. p. 16 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

29 Current Penetrations of Renewables The current penetration of non-hydro renewables is under 12% in the regions served by most independent system operators (ISOs) in the U.S. (Table 1). It is 6% in the area served by MISO. Even the higher scenario in MRITS for 2028 (25%) is within the range that past studies have found can be integrated without operational problems and without additional storage. Table 4. ISO/RTO Renewables 2014 (PJM 2015) Renewable Percent of Total Energy Renewable Percent of Total Capacity* Hydro Percent of Total Energy Hydro Percent of Total Capacity Non-Hydro Renewable Percent of Total Energy Non-Hydro Renewable Percent of Total Capacity ISO Basis Notes MISO ~8% ~12% ~1% ~3.8% ~6% ~7.8% Installed Percent non-hydro renewables declined to 6% in 2014 from 7.3% in 2013 due to addition of the South Region on 12/19/2013 CAISO ~23.75% ~24% ~9.75% ~6.25% ** ** Nameplate Includes units within balancing area and connected to CAISO grid & telemetry. Renewable excludes large hydro ISO-NE 16.7% 14.1% 8.0% 10.4% 8.7% 3.7% Summer Includes units under ISO dispatch control. Capacity excludes firm imports and exports of capacity. Energy excludes imports but includes exports. NYISO 25.5% 20.7% 20.5% ~14.8% 5% ~5.9% Unspecified Per NYPSC definition of renewables, hydro includes large hydro that existed prior to the order (2004) but only small run-of-river hydro going forward PJM ~4.4% ~6.3% ~1.9% ~4.8% ~2.5% ~1.5% Installed SPP 13% 13% 0.5% ~1% ~11.9% ~11.6% Installed *Generally includes small and large hydro (including pumped storage) wind, solar, wood, methane, refuse and other types. Capacity may be nameplate capacity, or may be a seasonal rating based on capability audits mandated by the ISO/RTO. **CAISO percent non-hydro cannot be determined since the renewable category excludes large hydro but the hydro category includes large hydro. Note: In 2014, CAISO implemented an Energy Imbalance Market to allow it to integrate renewable resources over a larger geographic footprint. Various neighboring utilities are planning to join the EIM. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 17

30 Strategies to Facilitate Integration of Renewables into the Electric System Many of the sources discussed above (GE Energy Consulting 2014a and 2014b, EnerNex Corporation 2011, GE Energy 2010a and 2010b, DeCesaro & Porter 2009, Denholm & Hand 2011) and others (Porter et al. 2012, Milligan et al. 2009, NERC 2009, 2015) conclude that successfully integrating large amounts of VG is greatly facilitated by substantial expansion of transmission, large balancing areas, and efficient regional markets: A substantial amount of new transmission and transmission upgrades is necessary to deliver wind and solar energy from high resource areas to high load areas, minimizing curtailments of VG output due to transmission congestion (and therefore total system costs to achieve a target percentage of renewables). Adequate transmission also enables the operation of large balancing areas. Balancing areas that are large both geographically and in terms of total generation capacity facilitate management of VG and lower the cost of VG integration because they provide more diversity of VG output, more diversity of load, and a larger number of dispatchable generating units that can provide increased regulation and ramping (net load following) required due to VG. Large and efficient markets that include both day ahead and sub-hourly timeframes and include both energy and ancillary services ensure that the lowest-priced resources are used to provide energy, regulation, and ramping. Open trading across balancing area boundaries makes available even more diversity of VG and load as well as more dispatchable generation, and can potentially further reduce wind curtailment, production costs, and integration costs. However, such interchange can also contribute to the variability of net load. In addition, these sources state the need for: Better measurement and forecasting of VG output; More advanced control of VG (e.g., wind turbines that ride through voltage and frequency disturbances, can be dispatched downward, can limit their own up- and down-ramp rates, and can provide inertial and frequency response and voltage support); Incorporation of different types of VG that have complementary output patterns (e.g., wind and solar); and More flexible resources on the system, including more flexible thermal generators (fast-ramping and capable of frequent starts and stops and modulating over a wide range), demand response, and possibly storage. The MISO situation with regard to each of these issues is described below. Transmission Major transmission expansion is inevitably required to deliver renewable energy from high resource areas to high load areas. In MISO, the best wind resources are in the Dakotas, Iowa, and southwestern Minnesota, while the major load centers are well to the east. Transmission lines typically take significantly longer to plan, approve, and build than renewable generation facilities. In the case of wind, this difference in timelines has been exacerbated by uncertainty about the ongoing availability of federal production tax credits (PTCs), which has caused wind developers to complete projects as quickly as possible without waiting for transmission upgrades in order to lock in those PTCs (MN PUC 2008). In past periods the transmission system has been unable to deliver a significant portion of available wind p. 18 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

31 energy from resource areas to the Twin Cities due to transmission congestion, resulting in wind power output being manually curtailed to avoid overloading transmission lines and utilities paying for renewable energy that they could not actually receive. Curtailment reached a peak of 14% of Minnesota s wind energy production in 2004 (MN PUC 2008). The EWITS study (EnerNex Corporation 2011) provides valuable background on transmission planning for VG integration. In EWITS, the authors noted that the existing U.S. bulk power grid is the result of a bottom-up process that makes decisions about transmission based on annual incremental expansions. Each transmission line is decided one at a time to meet near-term resource adequacy or delivery requirements, that is, the objectives have historically been to connect specific new power plants to load or to maintain or improve reliability. However in EWITS, the project team used top-down economic methods to develop the conceptual transmission capacity needed to deliver energy to load. These top down methods tend to create designs with more transmission than bottom-up methods. The primary reason is that the total economic potential of increasing the economic efficiency of the generation fleet including wind generation in the Eastern Interconnection is used to justify transmission expansion (p. 32). After conceptually siting the targeted amount of new wind generation in areas with optimal wind potential, as well as the additional non-wind generation required for reliability and [a]fter simulating system operation over an entire year of hourly data, study analysts then compared the results of this modeling simulation to those from a similar simulation in which constraints on the transmission system were removed. The comparison indicates how regional or interconnection-wide production costs increase because of transmission congestion, or put another way, what value could be achieved by eliminating or reducing transmission constraints. Differences between the constrained case and the unconstrained case yield the following information: The areas of economic energy sources and sinks The interface flow changes to determine the incremental transfer capacity needs The total benefit savings, which in turn gives a rough estimate of a potential budget for building transmission to relieve constraints and reduce congestion costs. Transmission flows between regions in EWITS are determined in part by the differences between production simulations using a copper sheet (i.e., no transmission constraints, no congestion) versus the existing transmission system. Transmission capacity is designed to deliver 80% of the desired energy flow 19 [T]he annual generation differences between the unconstrained and constrained cases [help] to define the energy source and sink areas and [give] insight into the optimal locations for potential transmission lines and substations. (p ). They note that, in spite of the massive transmission build-out contemplated in the study, transmission costs are a relatively small fraction of total annualized costs for all scenarios analyzed (p. 224) The 80% figure was based on an initial estimate of the economic level of transmission expansion. 20 Reading from a graph on p. 224 of the report, it appears that the annualized cost of the transmission expansion for the three 20% wind scenarios ranges from about 7 to 10% of total annualized costs (including wind capital costs, other new generation capital costs, transmission costs, wind integration costs, wind operation costs and production costs), or 18 to 29% of the annualized capital cost of the wind plants. For the 30% wind scenario the annualized cost of the transmission expansion appears to be about 9% of total annualized costs or about 17% of the annualized capital cost of the wind plants. The baseline scenario, though, also requires non-trivial transmission expansions. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 19

32 Several efforts have resulted in plans for regional transmission development in Minnesota and MISO with a view to connecting renewable resource areas to load centers. In 2004 several Minnesota utilities (Great River Energy, Minnesota Power, Otter Tail Power Company, and Xcel Energy) began planning for major high voltage transmission upgrades and expansion to accommodate projected long term load growth (Great River Energy et al. 2004). Eventually 11 transmission-owning utilities in Minnesota, North and South Dakota, and Wisconsin joined the effort. At the time planning began in 2004, it was the largest development of new transmission in the Upper Midwest in nearly 40 years ( In planning this transmission expansion, the member utilities used information from independent power producers, wind developers, utility resource planning staff, and MISO to develop three generation scenarios, each of which included 2,400 MW of renewable energy, consistent with the Minnesota Renewable Energy Objective at the time. The expansion also considered access to potential new coal generation in the Dakotas. Transmission that was common to two of the three scenarios was considered to be the core of the CapX 2020 Vision Plan (CapX 2020, 2005). Figure 6. CAPX-2020 Project Status (CapX Quarterly, Winter 2015) Cap-X 2020 included four 345 kv lines and one 230 kv line expected to cost more than $2 billion and cover about 800 miles ( The Group 1 portion (referenced p. 20 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

33 later) included three 345 kv lines, one 230 kv line, and associated substations, all of which are either in service currently (spring 2016) or projected to be completed by (Figure 6). MISO began a formal transmission expansion planning process in Currently, MISO s annual transmission expansion planning process (MTEP) involves analyzing the myriad regulatory policy and reliability issues impacting our energy sector and developing a portfolio of transmission projects designed to maintain a reliable electric grid and deliver the lowest-cost energy to customers in the MISO region (MISO 2015a p. 4). Their planning currently includes three categories of transmission projects (MISO 2015a p ): Bottom-up projects, including baseline reliability and other projects; Top-down projects, including market efficiency and multi-value projects; and Externally driven projects, including generation interconnection, transmission delivery service, and market participant funded projects. The top-down projects are of most interest in the present context (MISO 2015a p ): Top-down projects include transmission projects classified as Market Efficiency Projects and Multi-Value Projects. Regional or sub-regional top-down projects are developed by MISO working in conjunction with stakeholders to address regional economic and/or public policy transmission issues. Interregional top-down projects are developed by MISO and one or more additional planning regions in conjunction with stakeholders to address interregional transmission issues. Interregional projects are cost shared per provisions in the Joint Operating Agreement and/or MISO tariff, first between MISO and the other planning regions, then within MISO based on provisions in Attachment FF of the MISO tariff. Multi-Value Projects (MVP) meet Attachment FF requirements to provide regional public policy, economic and/or reliability benefits. Costs are shared with loads and export transactions in proportion to metered MWh consumption or export schedules. Market Efficiency Projects (MEP), formerly referred to as regionally beneficial projects, meet Attachment FF requirements for reduction in market congestion. MEPs are shared based on benefit-to-cost ratio, cost and voltage thresholds. MISO s Multi-Value Project (MVP) Portfolio is a group of 17 transmission projects that grew out of the Regional Generation Outlet Study (RGOS) (MISO 2010). The goal of the RGOS was to allow MISO states Renewable Portfolio Standards to be met at the lowest delivered wholesale energy cost. MISO and other stakeholders determined that the more traditional bottom-up process of planning transmission based on the 64.5 GW of wind requests in the generation interconnection queue at that time would not be an efficient means for building a cost-effective transmission system The RGOS included an extensive analysis of the trade-offs between wind generation located in lower wind resource areas closer to load centers requiring less transmission and more wind turbines and wind generation located in the highest wind resource areas with substantially more transmission to get it to the load centers. The least-cost approach included a combination of local and regional generation and was affirmed by the Midwest Governors Association as the preferred approach to wind zone selection. The MVP portfolio was deemed to 21 These are (1) the Brookings SD to Hampton, MN 345 kv line and the related Marshall, MN to Granite Falls, MN 345 kv line, (2) the Fargo, ND-St. Cloud, MN-Monticello, MN 345 kv line, (3) the Hampton, MN/Rochester, MN La Crosse, WI 345 kv line, and (4) the Bemidji, MN Grand Rapids, MN 230 kv line. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 21

34 provide substantial reliability, public policy, and economic benefits. The reliability benefits were based on conventional transmission analysis. The public policy benefits included facilitating the delivery of 41 million MWh of wind energy/year to meet renewable energy mandates and goals in various states. The economic benefits were determined based on a benefit/cost analysis that found B/C ratios of 1.8 to 3.0. The energy zones for which the RGOS and MVP transmission plan was designed were sited with strong consideration of wind capacity factors along with other factors such as the location of natural gas pipelines and existing transmission. The intent was that the energy zones, though created to serve renewable mandates, be usable for other types of generation as well as wind, and that the portfolio therefore consist of no regrets projects that would provide multiple kinds of reliability and economic benefits under all alternate futures studied (MISO 2012a). Figure 7. MVP Portfolio (MISO 2012). The MVP projects were expected to cost $5.2 billion and had targeted in-service years ranging from 2014 to 2020 (MISO 2012a). FERC approved and a federal appeals court upheld MISO s plan to allocate the costs of the MVP projects to utilities in proportion to their share of wholesale consumption of electricity (so-called postage stamp allocation) (FERC 2010, Varela p. 22 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

35 2013). 22 The location of the lines is shown in Figure 7 (dashed pink and green lines). Their current status and cost estimates 23 are shown in Table 5. Fourteen of the 17 have completed state regulatory approvals and, of these, three are complete and six have construction underway. Table 5. Multi-Value Project Status as of Q (MISO 2015b) The Great Northern Transmission Line (GNTL) is another major project of interest from the perspective of VG integration. The GNTL includes a 500 kv line planned to extend from the border with Manitoba, Canada, to the Blackberry Substation in Grand Rapids, Minnesota, and a 345 kv double circuit line from there to the Arrowhead Substation near Duluth (Minnesota Power 2012). It will connect with a new transmission line running from a converter station east of Winnipeg, Ontario, to the border with Minnesota. The GNTL is intended both to provide hydro power to Minnesota Power and to enable storage of wind energy (MN PUC 2015, Minnesota Power 2013b), as described further in the Manitoba Hydro Wind Synergy Study 22 In a parallel development, FERC Order 1000, effective October 11, 2011, required among other things that Local and regional transmission planning processes must consider transmission needs driven by public policy requirements established by state or federal laws or regulations. Each public utility transmission provider must establish procedures to identify transmission needs driven by public policy requirements and evaluate proposed solutions to those transmission needs. The order also established requirements for regional transmission cost allocation The MTEP Approved column shows the estimated cost at the time of these projects approval by MISO s independent Board of Directors (December 2011). The Q column shows estimates as of the fourth quarter of The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 23

36 section. The estimated cost of the 500 kv portion is $0.5B (Minnesota Power 2013). A certificate of need was granted June 30, 2015 (MN PUC 2015), and a route permit was granted on February 26, An additional federal permit required because the line crosses an international border is still pending. The MRITS Baseline transmission model incorporates all transmission lines already planned or built, including the CapX 2020 Group I lines, the MISO Multi-Value Project (MVP) portfolio, and the Great Northern Transmission line, which have total costs of nearly $7B, 25 in addition to numerous other lines and upgrades 26 (GE Energy Consulting 2014 p. 1-2). With this transmission build-out, the MRITS Baseline system will be able to deliver renewables sufficient to meet all current renewable and solar energy standards (see Table 1) with curtailment of less than 0.5% of total wind generation and less than 0.1% of total solar generation (Table 6). 27 Table 6. Wind and Solar Curtailment for MRITS Study Scenarios (GE Energy Consulting 2014) 28 Meeting the renewable requirements of MRITS Scenario 1 (see Table 1) did not necessitate any new transmission lines beyond those in the Baseline scenario, although it did require 54 transmission mitigations totaling $373 million (GE Energy Consulting 2014 p. 1-4). The study found that with the large expansion of transmission already included in the Baseline and the comparatively small amount of transmission mitigations added for Scenario 1, the system can operate successfully for all hours of the year with no unserved load, no reserve violations, and curtailment of only 1% of available wind energy and 0% of available solar energy (Table 6). Curtailment in this scenario is primarily due to local transmission congestion at a few wind plants. This congestion could be mitigated by transmission modifications, if economically justifiable. Meeting the renewable requirements of Scenario 2 was estimated to require nine new transmission lines and 30 transmission upgrades over and above those in Scenario 1, with an estimated cost of $ There is some overlap between CapX-2020 projects and MVP portfolio projects. Specifically, the Big Stone to Brookings 345 kv line and the Brookings to SE Twin Cities 345 kv line from CapX-2020 were subsequently approved for inclusion in the MVP portfolio. They had estimated combined costs of $0.886B. 26 The other lines and upgrades are the MISO Transmission Expansion Plan Appendix A projects with in-service dates prior to 2023, the Manitoba Hydro Bipole III (500kV HVDC line with a cost of C$3.28B, per Manitoba Bureau of Statistics 2011), Antelope Valley Station-Charlie Creek-Williston-Tioga 345 kv, Hazleton-Salem 345 kv, Square Butte HVDC increase to 550 MW, Center-Prairie 345 kv, and Winger-Thief River Falls 230 kv line. 27 Curtailment was defined in MRITS as the difference between the total available wind or solar energy output and the delivered wind or solar energy (GE Energy Consulting 2014 p. 7-3). Thus it includes output lost due to dispatch down of dispatchable intermittent resources (see section on Markets) and manual curtailment of nondispatchable intermittent resources. 28 Note Scenarios 1a and 2a are discussed in the section on Low Load Flexibility. p. 24 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

37 billion (GE Energy Consulting 2014 p. 1-4). Bill Grant (pers. comm.) indicated that the project team felt that that level of additional investment in transmission would have to be a regional discussion. In Scenario 2 curtailment was somewhat higher: over 2% of available wind energy and over 0.4% of available solar energy (Table 6). To get the curtailment down to this level with this amount of additional transmission build-out, the project team also had to relocate some of the wind plants not used to meet Minnesota s RPS to the south and east, generally from the Dakotas, Minnesota, and western Iowa to eastern Iowa, Illinois, Indiana and Michigan, where the wind resource is not as good but transmission congestion would be lower (p. 3-9 and 3-10). Because of the lower resource this also entailed increasing the number of wind turbines. Curtailment in Scenario 2 is discussed further in the Low Load Flexibility section. Balancing Area Geographic Extent and Generation Capacity Large balancing areas lower the cost of VG integration because they provide more diversity of VG output, more diversity of load, and a larger number of dispatchable generating units. MISO is geographically the largest independent system operator in North America (MISO 2014) (see Figure 8). It has the second largest generating capacity (after PJM), totaling 174,808 MW, with 189,390 MW in its reliability coordination area (shown in Figure 9)(MISO 2016). It is substantially larger both geographically and in installed capacity than CAISO, NYISO, SPP, ISO-NE, and ERCOT. The numerous local balancing areas within MISO are functionally consolidated (GE Energy Consulting 2014, p. 2-1). This can be contrasted, for example, to the situation in the Western Electricity Coordinating Council (WECC) area at the time of the WWSIS, which, aside from California and Alberta, operated as 37 separate balancing areas. It is much more difficult to integrate VG in such small balancing areas. As one example, Milligan et al. (2009) demonstrated the 14% reduction in ramping requirements that could be achieved by combining four balancing areas in Minnesota that were separate at the time of Milligan s study. He reports similar results for an analysis of the 11 zones in the New York system at that time. Aggregating balancing over an area as large as MISO, as is currently the case, reduces variability even further. The MISO footprint includes fairly widely distributed wind and solar resources, as can be seen by comparing Figure 9 with Figure 10 and Figure 11. In addition, MISO has seams agreements and processes with eight adjacent RTO and non-rto entities (MISO 2015c, d). MISO and PJM (at least) schedule transfers across their boundaries on a sub-hourly basis (Porter et al. 2012). The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 25

38 Figure 8. North American ISO/RTOs (FERC 2016). Figure 9. MISO Market Area and Reliability Coordination Area p. 26 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

39 Figure 10. U.S. Wind Speed at 100 m (NREL 2013) Figure 11. U.S. Solar Photovoltaic Resource (NREL 2012) The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 27

40 Markets Key market attributes to facilitate integration of VG include large market size, inclusion of both dayahead and sub-hourly timeframes, and inclusion of both energy and ancillary services (Milligan et al. 2009, DeCesaro & Porter 2009). Larger markets allow efficient access to the diversity of load, VG output, and conventional generation available in a large balancing area. In an energy market, each generating unit or other resource submits offers for energy at a price typically close to its short run marginal cost of production. Likewise, load-serving entities submit bids for needed energy. In an ancillary service market (ASM), generators and other resources submit offers for the capacity needed to provide moment to moment frequency regulation and the contingency reserves (spinning and non-spinning) needed to ensure reliability in case of an unexpected generator or transmission line outage. These ancillary services are purchased by the balancing authority with costs passed on at a later time. The price of ancillary services includes the opportunity cost (foregone profit) to the generator of withholding part or all of their capacity from the energy market in order to provide ancillary services. For all of these markets, the market clearing price is the crossing point between the supply and demand curves. 29 All resources (e.g. generators) whose offers are accepted for a particular product (energy, regulation, spinning reserves, or non-spinning reserves) are paid the market clearing price for that product. Since many generating units (and other resources) can provide energy as well as one or more of the ancillary services, and since they offer each at separate prices, units are committed and dispatched using a simultaneous co-optimization method that [m]inimizes the cost of committing sufficient Resources to meet forecasted demand, confirmed Interchange Schedules (imports and exports) and Operating Reserve requirements, all subject to system security constraints (MISO 2014d), and maximizes the total margin (payment above offers) for individual resources (MISO 2014c). The market process ensures (barring abuses of market power or other factors) that the lowest cost resources are operated to meet load and reserve requirements (subject to system security constraints). 30 In the day-ahead market, bids and offers are made and markets are cleared for each hour of the following day. Based on the bids and offers, the ISO performs a security constrained unit commitment analysis to determine which units will operate the next day and what their schedules will be. Generators and other resources are contractually obligated to follow their day-ahead commitments or pay for any energy or ancillary services they did not produce. In the real time market, bids and offers are made every time increment (5 minutes in MISO s case) to true-up the balance of supply and demand based on real conditions, as opposed to day-ahead expectations (MISO 2014b). 29 MISO uses a locational marginal price (LMP) approach for energy. The LMP is the market price required to supply the last incremental amount of energy at a specific node on the transmission grid, including a marginal energy component, a marginal transmission congestion component, and a marginal transmission loss component (MISO 2014c). 30 See the report by MISO s independent market monitor (Potomac Economics 2015 p ) for a discussion of MISO market competitiveness. They conclude that the system marginal price is only 1% above estimated marginal cost and that instances of apparent withholding account for less than 0.6% of load. p. 28 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

41 Among the challenges of VG for the bulk electric system are that (a) VG has high ramp rates (rates of increase or decrease in output per unit time), (b) VG ramps are often in the opposite direction from load ramps, creating even higher net ramping rates, and (c) there is significant variability and uncertainty in VG ramping: U.S. wind integration studies have typically found a larger increase in the need for load following [than frequency regulation] with higher levels of wind generation. This is due in large part to wind s diurnal output, which in many cases may be opposite of the peak demand period for electricity. For instance, wind output may fall off in the early morning hours when load is increasing, increasing the need for generating resources to ramp up to meet the increasing electric demand. Conversely, wind power production may be higher during off-peak hours when load is decreasing or at minimum levels, increasing the need for generating resources that can ramp down. Therefore, adding wind generation will typically require more load following to counteract the combined net variability of load and wind (DeCesaro & Porter 2009). the output of variable resources is characterized by steep ramps as opposed to the controlled, gradual ramp up or down generally experienced with electricity demand and the output of traditional generation. Managing these ramps can be challenging for system operators, particularly if down ramps occur as demand increases and vice versa. Insufficient ramping and dispatchable capability on the remainder of the bulk power system can exacerbate these challenges (NERC 2009). DeCesaro & Porter (2009) provide an example based on the 2004 wind integration study conducted for Xcel Energy and the Minnesota Department of Commerce. Figure 12 shows the distributions of hourly changes in load with and without 25% wind for specific summer and winter hours. Solar systems can likewise cause increased ramp requirements (Figure 13). EPRI (2010) notes that PV systems have large voltage sags and rapid demand shifts due to cloud effects and that [t]hese effects can be even more severe than wind ramps because they are much faster. These factors increase the ramp rate of net load (load minus VG output), which must be met by other resources (DeCesaro & Porter 2009, Denholm & Hand 2011). This can be handled in various ways. One approach is to use day-ahead hourly commitments or operate hourly markets and then meet any intrahour ramping requirements with regulation resources. Another approach is to operate sub-hourly energy markets, which provide load-following as a byproduct. A third option is to meet ramping requirements with spinning or non-spinning contingency reserves. Finally, sub-hourly markets can be complemented by other processes or products used to ensure that sufficient ramping capability is available on the system at all times. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 29

42 Figure 12. Load Following Impact of Wind (DeCesaro & Porter, 2009) p. 30 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

43 Figure 13. CAISO Graph of the Impact of Solar Generation on Load Shape and Ramping (so-called duck curve, cited in Lazar & Linvill 2016) Comparing sub-hourly energy markets to hourly markets that use regulation to meet intra-hour ramping requirements, DeCesaro & Porter (2009) state that: wind integration costs for independent system operators (ISOs) and regional transmission operators (RTOs) are typically lower than for non-isos and RTOs. One reason for these results is that ISOs and RTOs typically operate sub-hourly markets (i.e., they dispatch generation on a five- to fifteen-minute time frame), while many of the non-isos or RTOs require generators to follow hourly schedules and obtain all sub-hourly balancing from regulating units. Milligan et al. (2009) explain why sub-hourly markets are preferable to using regulation: [F]ast energy markets [which perform an economic re-dispatch every 5 to 15 minutes] make it possible to hold the regulating units closer to their preferred operating point because they can be brought back to the mid-point of their operating range much faster than if the re-dispatch did not occur for an hour. Therefore, there is less need for regulation in faster energy markets. This results in a significant reduction in costs because regulation is typically the most expensive ancillary service. Meeting intra-hour load-balancing requirements using spinning reserves has also been discussed and sometimes recommended (Milligan et al. 2009, GE Energy 2008) because these reserves clear at a substantially lower price than regulating reserves. However, this approach is complicated by the fact that spinning reserves are intended to meet specific contingencies (e.g., an unexpected power plant or transmission line outage), not to be load-following resources, and are subject to specific reliability rules (DeCesaro & Porter 2009). A number of studies note that the addition of VG does not increase true regulation requirements by very much (unless regulation is used to meet intra-hour balancing requirements). DeCesaro & Porter (2009) The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 31

44 report that, [b]ecause the variations of load and wind tend to be uncorrelated in short time scales, most U.S. wind integration studies have found that only modest amounts of additional regulation are necessary with more wind MISO s markets generally meet the criteria recommended for efficient integration of VG: MISO s market is large both geographically and in terms of total system capacity. MISO launched day ahead and real-time (5 minute) energy markets and a market for Financial Transmission Rights in 2005 (Potomac Energy 2015). 31,32 MISO began operating as a balancing authority and launched its ancillary services market (ASM) in 2009 (Potomac Energy 2015). It initially included regulation and contingency reserves. 33 MISO has added other market processes and products that help to manage integration of wind. In April 2012 it implemented a Look-Ahead Commitment tool. The tool is used on the operating day and covers the period three hours ahead. It allows system operators to optimize near-term unit commitment decisions and address intra-hour requirements given uncertainties in the load forecast, wind forecast, net scheduled interchange (NSI) across MISO system boundaries, and equipment outages (MISO 2013). On May 1, 2016, MISO added a ramp capability product to its ASM. This is another market feature that will facilitate integration of renewables. The Independent Market Monitor (IMM) for MISO (Potomac Economics) recommended this in 2011, and MISO filed a proposal for such a product with FERC in 2014 (Potomac Economics 2015). The rationale for this product is as follows: currently [r]amp shortages are the most common cause of short-term scarcities and price spikes in MISO. When the system has capacity online but it is not rampable, prices rise to scarcity level and then quickly drop 31 According to the independent market monitor (Potomac Energy 2015) one of the shortcomings of MISO s realtime (5 minute) market is that, while it: produces new dispatch instructions and prices every five minutes, settlements are based on hourly average prices. This inconsistency can create incentives for suppliers to be inflexible rather than being responsive to five-minute signals. For this reason, MISO instituted Price Volatility Make-Whole Payments (PVMWP) to ensure that suppliers are not harmed when they respond to MISO s five-minute dispatch instructions These payments would be substantially reduced if MISO settled with participants on a fiveminute basis. Additionally, flexible resources would have received more than $35 million in higher net revenues and inflexible resources would have received lower net revenues under a five-minute settlement. These changes in settlements would provide much better incentives to follow dispatch instructions and, in doing so, would generate production cost savings for the system and improve reliability. Hence, we continue to recommend that MISO implement five-minute settlements for generators and external transactions In addition, we have proposed improved uninstructed deviation thresholds to provide better incentives for generators to be flexible and perform well in following MISO s dispatch signals. 32 In the time available we were unable to find information on the mix of electricity on the MISO system that is purchased from the MISO market, self-generated, or obtained through bilateral agreements. 33 Per MISO 2014a: Regulating Reserves are used for the real-time balancing of supply and demand. Regulating Reserves must be online and have a response time of 5 minutes. Resources providing Regulating Reserves must be equipped with Automatic Generation Control (AGC) in order to respond to MISO dispatch signals. They are the most flexible and thus the highest-priced reserve product Contingency reserves are activated in case of a contingency (a sudden loss of generation). [They] consist of Spinning Reserves and Supplemental Reserves. Spinning Reserves must be online [and] have a response time of 10 minutes. They are not as flexible as Regulating Reserves and thus have a lower price. Supplemental Reserves can be online or offline and must have a response time of 10 minutes. As the least flexible reserve product, they have the lowest price. p. 32 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

45 back Transparent price signals are needed to provide economic incentives for resource flexibility (MISO 2016b). As explained in MISO s white paper on the product s development (Navid et al. 2013): Operating the power system to maintain power balance requires controllable resources to alter their power output to match varying net load. Maintaining this balance can be a challenge when there is a high rate of change from the non-controllable assets (e.g., load, net schedule interchange, renewable production). With increases in the proportion of generation from intermittent renewable resources and increases in the flexibility of interchange scheduling (e.g., 15 minute scheduling intervals), it is likely that the variability of the net load will increase in the future which will further tax the ramp response of controllable resources and could increase the frequency of short-term scarcity events due to shortages of rampable capacity. In real-time operations on the second or minute timeframe, regulation service provides the only response option for variations in the net load. On the dispatch horizon, controllable resources are poised for expected [emphasis added] changes in the net load, or for so called "load following." The current practice is to schedule units to provide the most economical solution to the expected [emphasis added] level of variation. The ability to cope with additional [emphasis added] variation is provided by residual capability of controllable resources. Deviations from expected net load or high rates of change beyond the visibility of the dispatch horizon can leave the dispatchable resources with sufficient capacity but without ramp capability to respond which can lead to short-term scarcity events. The overall objective of the investigation summarized in this paper is to identify an approach that can be applied from the Day-Ahead Market through Real-Time Dispatch to provide a determined quantity of ramp capability provided by controllable resources to respond to the variability of the net load served by these resources. The purpose of this approach is to increase the robustness of system operations for a wider range of potential operating conditions and to reduce the frequency and/or severity of short-term scarcity conditions which occur when resource ramp constraints cause a short-term inability to meet variations in the net load. With an expected increase in the variability of net load, it follows that scarcity conditions may increase in frequency without additional efforts to achieve higher net responsiveness from the existing controllable resources. Dispatching resources in a way that increases their response capability may result in additional costs, but it has the benefit of increased responsiveness and reduction in the frequency of scarcity events. Attempting to eliminate all scarcity events with this approach is not practical as there could be some very rare events with high variations beyond the level of ramp capability that can be achieved at reasonable expense. A balance must be struck between the additional operating costs that are required to provide additional ramp capability and the avoided costs of prevented scarcity events. One major factor in this investigation is to identify a balance where the avoided costs exceed the additional operating costs and any increase in the system total production cost is avoided. Based on the cost-benefit studies, the tangible annual cost savings are estimated to be in the range of $ M after consideration of the impact of additional costs of $ M in operational costs to provide the ramp capability products. Alternate solutions such as increasing the regulation requirement and using it for URC / DRC and increasing the spinning reserve requirement and using it as URC were studied and both have shown a significant increase in total production cost compare to the proposed URC / DRC products introduced in this paper. It will be up to the resource owners to decide whether to include each resource in ramp participation. Ramp capability will be simultaneously co-optimized with energy and the other ancillary services in order to choose the most economical resources to meet each requirement. While only about one fourth of resources are qualified to provide regulation, nearly all dispatchable resources 34 are qualified to 34 Except Demand Response Resources Type 1 and short term Stored Energy Resources (SER). The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 33

46 provide ramping. It is expected that ramp product prices will be much lower than regulation prices and generally lower than contingency reserve prices. 35 (MISO 2016b). Another feature of the MISO market that is of interest from the point of view of energy storage is payfor-performance regulation. This is discussed in the section titled Pay-for-Performance Regulation. Note that some services required by the system are obtained by MISO through means external to the market. For example, conventional generating plants provide inertial response as a direct physical consequence of their heavy rotating masses, but are not compensated for this. Voltage regulation and black start capability are also obtained through extra-market contracts. Forecasting of VG Output Accurate forecasting of output from VG is important to manage the costs of VG integration (DeCesaro & Porter 2009, EnerNex Corporation 2011). Past studies have found that unit commitment costs typically account for the largest share of wind integration costs and that wind forecast accuracy directly affects unit commitment. For the unit commitment time frame, wind generation introduces uncertainty in the day-ahead time frame of scheduling and committing generating units. This is the time scale that has the largest wind integration cost impacts, up to almost $9.00/MWh at wind capacity penetrations of up to 20% or 30%. Unit commitment cost impacts are contingent on the amount of and characteristics of dispatchable generation resources, the amount of the wind forecast error (and interactions with the load forecast error), the market and regulatory environment, and the characteristics of the wind generation resource as compared to load. The uncertainty of wind power production in the unit commitment time frame may result in higher variable costs through increased fuel consumption and increased operating costs. This may occur if too much generation is committed due to underestimating wind production, or if not enough generation is committed because of overestimating expected wind generation, necessitating the use of quick start units, or shortterm market purchases. Short term (same day) forecasting can also improve the accuracy and economic efficiency of look-ahead commitment and real-time ramp capability requirements. MISO forecasts wind output for each wind plant within MISO. An hourly forecast is developed for seven days into the future and is used in unit commitment, transmission security planning, and outage coordination. A 5-minute interval wind generation forecast is developed for six hours into the future and is used to establish wind resource high limits for Real-Time Economic Dispatch and Look-Ahead Unit Commitment For Economic Dispatch, a participant-supplied forecast is used when available [32% of wind farms}; MISO s forecast is used when not available (MISO 2015e). According to their report to FERC for the period from (PJM 2015 p. 163): Wind forecasting accuracy is calculated using an industry-wide methodology called Mean Absolute Error (MAE). The MAE is the average of the absolute value of the difference between forecasted and actual wind power output and is expressed as a percent of installed wind nameplate capacity. The wind forecasting accuracy is represented as one minus MAE.[This] methodology used is a common practice within the industry. 35 MISO currently employs Headroom Constraints that model the need for capacity to serve intra hour ramping in the day-ahead and reliability assessment unit commitments. This is a non-market strategy and may be phased out after the ramp capability product is added if experience shows it is no longer needed. p. 34 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

47 MISO is continuing to explore methods for improving the accuracy of its wind forecasting, but our current accuracy appears to be consistent with the accuracy obtained in other regions throughout the world. Figure 14. MISO Average Hourly Wind Forecasting Accuracy (MISO 2015e) MISO s day-ahead forecast accuracy has improved 2.5% since 2009 (Figure 14) (MISO 2015e). Note that because mean absolute error and forecast accuracy are expressed as a percent of nameplate capacity and because wind output is considerably lower than nameplate capacity on average, the actual error in wind output forecasting is substantially greater than it may appear. This can be particularly consequential at times when wind accounts for a substantial fraction of total generation. Advanced Control of VG The types of control available for VG itself also affect integration impacts. VG that can ride through voltage and frequency disturbances, can be dispatched downward, can limit its own up- and down-ramp rates, and can provide inertial and frequency response and voltage support requires less compensatory flexibility on the part of other system resources. MISO s Generator Interconnection Agreement (GIA) (MISO 2016g) requires new and replacement wind generating plants to have specific low-voltage ride-through capabilities (Appendix G). It also requires that they maintain a power factor between 0.95 leading and 0.95 lagging if the Transmission Provider s System Impact Study shows that such a requirement is necessary to ensure safety or reliability (Appendix G). In some cases dynamic voltage support is also required. Wind plants must also provide supervisory control and data acquisition (SCADA) capability to enable them to transmit data to MISO and receive instructions from MISO (Appendix G). Both wind and solar systems must provide meteorological data and forced outage data to MISO sufficient for MISO s development and deployment of power production forecasts for that class of Variable Energy Resources (Section 8.4). Wind plants must provide site-specific data on temperature, wind speed, wind direction, and atmospheric pressure. Solar plants must provide site-specific data on temperature, atmospheric pressure, and irradiance. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 35

48 The requirements from Appendix G of MISO s GIA are identical to those in Appendix G of the FERC pro forma Large Generator Interconnection Agreement. 36 FERC is currently considering removing the language related to power factor from Appendix G of its LGIA and the provision in section that exempts wind generators from meeting the power factor criteria applied to other generators. That section would, however, provide that non-synchronous generators be required to maintain a 0.95 leading to 0.95 lagging power factor only when their output is above 10% of capacity (FERC Docket No. RM ). 37 FERC is also considering requiring small generators (<=20 MW) to ride through system 38, 39 disturbances (FERC NOPR March 17, 2016, Docket No. RM ). One important market feature that MISO has instituted is the designation of a new resource type, Dispatchable Intermittent Resource (DIR) as part of their Tariff. MISO proposed this to FERC on November 1, 2010 (Kessler 2010), and FERC conditionally approved it on February 28, Prior to this Tariff change, intermittent resources were not dispatched through the real-time market and were not allowed to set market prices. They were simply price takers. If there was excessive transmission congestion or if wind was surplus to requirements during periods of low load due to the amount of must run coal generation on the system, MISO Reliability Coordinators contacted wind generators to direct them to curtail output. The amount of curtailment needed to balance the system had to be estimated by the Coordinators (MISO 2012b). The DIR tariff proposal was developed with stakeholder input as part of MISO s ongoing Wind Integration Initiative (Kessler 2010). The tariff requires that intermittent resources be capable of being dispatched down to lower output, as other resources can be. This allows them to participate in the market and potentially set market prices. It was expected to reduce manual curtailments, minimize over-curtailments, and maximize the use of low-cost wind, resulting in a more efficient market outcome. From the real-time energy market s perspective, the only difference between a DIR and a conventional resource is that the DIR s economic max output for dispatch purposes is determined by a 10-minute forecast (MISO 2012b). Most wind generators were required to switch to DIR status by March 1, As originally proposed to FERC, the DIR requirement would have covered solar generating plants as well as wind. FERC s approval limited mandatory DIR status to wind, but solar projects can choose to be a DIR. 36 Note that large generators are those > 20 MW. For low voltage ride-through, the FERC pro forma GIA also included a less rigorous standard for a transition period, but that transition period is long past The MRITS study (GE Energy Consulting 2014 p. 1-5) notes the following with regard to the VG plants included in their integration study: New utility-scale wind and solar photovoltaic (PV) plant models were consistent with current NERC and FERC minimum requirements (e.g. voltage regulation, power factor, voltage ride-through). Full commercial technical capability (e.g. synthetic inertia, frequency response) was not modeled. Distributed PV was modeled as lumped generation at locations (per the siting task) with no reactive power or voltage regulation capability. 40 The exceptions were intermittent resources that commenced commercial operations before April 1, 2005, and intermittent resources that had 100% of the capacity covered by certain types of transmission service (MISO 2012c). FERC rejected without prejudice the requirement for non-wind intermittent resources to register as DIRs (MISO 2011). p. 36 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

49 MISO s independent market monitor reports that the expansion of DIR has almost entirely eliminated manual curtailments as a means to manage congestion caused by wind output or to manage overgeneration conditions (Potomac Economics 2015). They cite averages of 201 MW/interval economic (DIR) curtailments in 2014 and only 3 MW/interval manual curtailments. Although an early report stated that the use of DIRs had resulted in more use of wind output (MISO 2012b), Potomac Economics (2015) reported more recently that, [w]ind output is being curtailed at approximately twice the rate compared to curtailments prior to DIR adoption in In 2015, DIR wind energy generation ranged from 82.5 to 85.2% of total wind generation in various months (Figure 15). A total of 6.4% of DIR wind generation was lost due to DIR being dispatched down (MISO 2016c) (Figure 16). The current split between dispatch down due to congestion and dispatch down due to low energy demand combined with must-run coal capacity is not readily available. An early report (MISO 2012b) for July through December 2011 showed DIRs at their economic maximum output 95.2% of the time, dispatched down due to congestion 3.7% of the time, and dispatched down due to low energy demand 0.2% of the time, which would suggest that congestion was, at least at that time, the more important driver of dispatch down. Figure 15. MISO Monthly Wind Generation 2015 (MISO 2016c, p. 34) The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 37

50 Figure 16. MISO Monthly DIR Wind Energy Generation (MISO 2016c). dispatched lost January February March April May June July August September October November December Total Total pct 93.6% 6.4% The analysis conducted in the MRITS study (GE Energy Consulting 2014) suggests that the gradual completion of the MVP portfolio and other transmission projects, together with the 12.6 GW of coal plant retirements that were included in the MRITS model based on the MISO-estimated effects of prior EPA regulations, will eventually reduce the dispatch down of DIRs considerably. One thing to be aware of in this regard is that MISO wind output has been running ahead of their renewable energy targets based on aggregate state RPS s (MISO 2016c p. 37), while transmission takes more time to plan, approve, and build than wind plants, and many of the coal plant retirements have not yet occurred. DIR s are now on the margin (and therefore setting prices) in at least part of the system close to half the time (Table 7). This is a bit misleading, though, since low prices set by wind units typical prevail in relatively small congested areas (Potomac Economics 2015). These prices tend to be negative and average $-7 per MWh. 41 FERC s order on MISO s DIR tariff required MISO to consider whether DIRs should be allowed to provide operating reserves. MISO investigated this but determined that there would be almost no time when DIRs would be able to provide operating reserves (McMullen 2012). 41 Elsewhere in the same report they say that the average price set by wind was $-11/MWh. p. 38 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

51 Table 7. MISO Capacity and Energy Output by Fuel (Potomac Economics 2015, p. 5) The changes between 2013 and 2014 reflect in part the incorporation of the South Region into MISO on 12/19/2013. The Price Setting percent for LMPs is confusing. According to MISO (2014f), Binding transmission constraints can produce instances where more than one unit is marginal in the system. Consequently, more than one fuel may be on the margin; and, since each marginal unit is included in the analysis, the percentage may sum to more than 100%. Different Types of VG Wind accounted for 6% of total energy output in MISO in 2014 (see Table 7), while solar accounted for a negligible percentage. The MRITS study Baseline for 2028 (GE Energy Consulting 2014 p. 2-6), which met all of the current renewable and solar energy standards for all of the states in the study region, included 22,229 MW of wind capacity and 1,509 MW of solar photovoltaic capacity (this is a 94%/6% wind/solar capacity split). Scenario 1 had modest increases over the Baseline, due to the potential increases in Minnesota renewable requirements under study, and a wind/solar capacity split of 91%/9%. Scenario 2 increased both wind and solar and decreased the wind/solar capacity split to 81%/19% (Table 8). The greater proportion of wind in all of these scenarios presumably reflects the substantial head start that wind generation has (Table 7, Table 8) as well as differences in capital costs and in the wind and solar resource quality in the MISO footprint (Figure 10 & Figure 11). Of the solar PV capacity, the MRITS Baseline scenario assumed that 94% would be utility-scale, Scenario 1 assumed that 87% would be utility-scale, and Scenario 2 assumed that 90% would be utility-scale (Table 8). Thus distributed small PV (non-utility-scale) would be 96 MWac in the Baseline, =306 MWac in Scenario 1, and =871 MWac in Scenario 2. This illustrates the fact that the study team and review committee for MRITS did not expect distributed PV to account for a significant share of total capacity or generation in The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 39

52 Table 8. MISO North+Central Wind and Solar Resources for MRITS Study (GE Energy Consulting 2014 p. 2-6) Low Load Flexibility As discussed earlier (Findings from Past Renewable Integration Studies section), Denholm and Hand (2011) found that at penetrations above those in most renewable integration studies to date, a primary constraint becomes the simple coincidence of renewable energy supply and demand for electricity, combined with the operational limits on generators providing baseload power and operating reserves. 42 Insufficient flexibility at low load can cause some curtailment of VG even at low penetrations. In their first study of energy storage (EPRI 2011) (see section on MISO Energy Storage Study Phase 1, 2011) MISO reported that: A significant proportion of the coal plant fleet is considered must run and therefore runs during off-peak hours. The need to keep coal plants running off-peak reduces the impact that free wind generation has in bringing down the off-peak power price since the system operator curtails the wind if the capacity is not needed. Although MISO s newer rules requiring most wind plants to be Dispatchable Intermittent Resources (DIR) have greatly reduced the amount of manual curtailment (see the Markets section), wind output is still occasionally reduced through dispatch during periods of low load, even with the current VG penetration of 6%. An early report (MISO 2012b) showed that for the period from July through December 2011, DIRs were at their economic maximum output 95.2% of the time, dispatched down due to congestion 3.7% of the time, and dispatched down due to low energy demand only 0.2% of the time. As MISO s market penetration of VG increases, reductions of wind output below maximum during periods of low load would also be likely to increase. The MRITS study (GE Energy Consulting 2014) investigated this issue by adding two Scenarios, 1a and 2a. The primary Scenarios 1 and 2 assumed that baseload coal plants would continue to operate as they do now, as must run units. Alternative Scenarios 1a and 2a assumed that coal units would be subject to economic commitment/dispatch. They found that the wind curtailment 43 in the baseline configuration and in Scenario 1 is primarily due to local transmission congestion at a few wind plants, so Scenario 1a provided no decrease in curtailment relative to Scenario 1. On the other hand, 42 See also Palchak & Denholm Defined in MRITS as the difference between available energy and delivered energy (p. 7-3), so in effect including both manual curtailments and dispatch down of DIRs. p. 40 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

53 Wind curtailment in Scenario 2 is due to system-wide operational limits during nighttime hours, when many baseload generators are dispatched to their minimum output levels. This type of curtailment could be reduced by decommitting some baseload generation via economic market signals. The effectiveness of this mitigation option is illustrated by comparing Scenario 2 (coal units must-run) with Scenario 2a (economic coal commitment). Wind curtailment decreases from 2.14% to 1.60% (reduction of 332 GWh of wind curtailment). Solar curtailment decreases from 0.42% to 0.24% (reduction of 12 GWh of solar curtailment) (GE Energy Consulting 2014 p. 1-9 and 1-10) [See Table 6.] Figure 17 shows that most of the curtailment in Scenarios 2 and 2a occurs during nighttime low load hours. Scenario 2a also resulted in an additional 100 to 200 starts/year for small coal units (under 300 MW) and an additional 20 to 100 starts/year for large coal units. 44 MRITS did not assess the costs of the increased cycling of the coal plans, though several other studies have done so (e.g., Connolly et al. 2011, Kumar et al. 2012, Lew et al. 2013, EPRI 2013). The MRITS study noted that there could be an intermediate option that would reduce VG curtailment while limiting start/stop wear-and-tear on coal plants; namely, to implement a multi-day unit commitment process, rather than day-ahead only (GE Energy Consulting p. 1-9). Figure 17. Wind Curtailment by Hour of Day for Minnesota-Centric Region (GE Energy Consulting 2014) 44 The additional start/stop and ramping of coal plants under economic dispatch may or may not be predominately due to VG: it is impossible to know for sure because the study did not run a model using economic commitment/dispatch of the coal plants without increased VG. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 41

54 The share of MISO s installed capacity and energy generation that is coal has decreased in recent years (Table 9) (PJM 2015), partly due to the addition of the South Region, which has much more natural gas generation, in December 2013 (Potomac Economics 2015). As discussed earlier (see Markets section), the MRITS study assumed 12.6 GW of coal plant retirements by 2028, based on MISO s estimate of the effects of prior EPA regulations (GE Energy Consulting 2014), so that factor is already reflected in the VG curtailment percentages projected by MRITS for Table 9. MISO Fuel Diversity (PJM 2015) Energy Storage Background and Basics A large amount of work has been done on energy storage in recent years. Some of the key activities are as follows: The U.S. Department of Energy (DOE) has coordinated energy storage research and development across the Office of Electricity Delivery and Energy Reliability, the Office of Energy Efficiency and Renewable Energy, the Office of Science Basic Energy Sciences, and the Advanced Research Projects Agency-Energy (DOE 2011). Numerous presentations from many years peer review meetings are available at Through Title XIII of the Energy Independence and Security Act of 2007 as modified by the American Recovery and Reinvestment Act of 2009, DOE co-funded 16 energy storage demonstration projects. These are being managed by the National Energy Technology Laboratory (Byrne et al. 2012). DOE created and is maintaining an internet database of energy storage projects worldwide ( In 2012 DOE started an Energy Storage Technology Advancement Partnership to work with interested states to accelerate commercialization and deployment of energy storage technologies via joint funding and coordination p. 42 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

55 ( It is facilitated by the Clean Energy States Alliance, of which the Xcel Energy Renewable Development Fund is a core member and the Minnesota Department of Commerce is an affiliate member ( FERC has issued a number of orders that directly or indirectly facilitated increased use of storage (Orders 755, 784, 719, 745, 792, 819, 890, 1000) (Bhatnagar et al. 2013, Behr, 2013, Hellrich- Dawson 2015, Stanton 2015, Akhil et al. 2013). In addition to basic technology R&D, DOE has funded research on the integration of energy storage into the electric system, including technical, economic, policy, and market barriers. (e.g., Batnagar et al. 2013, Byrne et al. 2012, Milligan et al. 2009). The Electric Power Research Institute (EPRI) has conducted extensive research and developed tools and guidance on energy storage from an industry perspective (search energy storage at epri.com) DOE and EPRI, along with the National Rural Electric Cooperative Association, published a comprehensive electricity storage handbook addressing everything from characteristics and uses of storage to procurement and installation (Akhil et al. 2013). Given the high volume and quality of materials available elsewhere and the limited budget of this project, it was deemed more beneficial to analyze and understand the context for storage in Minnesota than to review and summarize in detail the output of the activities listed above. This section provides a brief overview of how storage can be used in electric systems, key characteristics of various types of storage, and some of the market and policy barriers that storage faces. Potential Roles of Energy Storage in the Electric System Energy storage can provide many services at various levels within the electric system (Akhil et al. 2013, DOE 2013) (Table 10). From the point of view of increasing penetrations of VG on the electric system, the key potential applications are in bulk energy services, ancillary services, and transmission infrastructure services. 45 Other resources can also provide these services, so the potential for storage will be determined by its relative economics. 45 Although storage may provide value at the distribution or customer level, either to facilitate integration of distributed renewables or for other reasons, constraints on VG at these levels are not likely to have a significant impact on total VG penetration within MISO in the next decade or more. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 43

56 Table 10. DOE/EPRI Handbook Electric Grid Energy Storage Services (Akhil et al. 2013) Characteristics of Storage Technologies As described by Byrne et al. (2012), Akhil et al (2013), Black & Veatch (2012), and Lazard (2015), some of the key parameters of interest for storage technologies include power rating (capacity) (MW), energy storage capacity (MWh) (or equivalently, discharge time at rated power (DTRP) (in hours, minutes or seconds), round trip efficiency (energy output/energy input, %), ramp rate (MW/min), minimum load, technological maturity, capital cost, variable O&M cost, fixed O&M cost, reliability, safety, time required to plan and build an operating system, and plant life. The sources mentioned provide system descriptions and data on these parameters for a wide variety of storage technologies. 46 As shown in Figure 18 (Akhil et al. 2013), pumped hydroelectric storage and compressed air energy storage have the largest capacities and longest DTRPs. On the other end of the spectrum, flywheels, high power supercapacitors, and superconducting magnetic energy storage devices (SMES) have DTRPs of seconds to minutes. A wide range of battery technologies have intermediate DTRPs ranging from less than an hour to a number of hours. Table 11 (Akhil et al. 2013) shows the required power rating and discharge duration (as well as cycles/year and target lifetimes) for various electric system applications. 46 DOE has also initiated a process, conducted by NREL, to produce an Annual Technology Baseline with updated technology cost and performance data ( first year product contains very detailed information for renewable and conventional generation technologies, but does not address storage except in the context of concentrating solar power. It is expected that storage will be addressed in future years (see p. 67). p. 44 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

57 Figure 18. Approximate Ranges of Capacity and DTRP for Storage Technologies (Akhil et al. 2013) The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 45

58 Table 11. Storage Characteristics Needed for Various Electric System Functions (Akhil et al. 2013) Drawing from their database ( DOE reported 202 installed or announced storage projects in the U.S. as of August 2013 (Figure 19) (DOE 2013 p. 11). The great majority of the capacity is existing pumped hydro storage that was installed prior to p. 46 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

59 Figure 19. Rated Power of Current and Announced U.S. Grid Storage Projects In the same report (DOE 2013 p ), DOE also summarized the 2013 maturity and the status of many types of storage technologies (Figure 20, Table 12). Figure 20. Maturity of Electricity Storage Technologies (DOE 2013) The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 47

60 Table 12. Primary Applications and Status of Electricity Storage Technologies (DOE 2013) p. 48 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

61 In 2012, Black & Veatch 47 prepared a report on Cost and Performance Data for Power Generation Technologies for NREL. [E]stimated costs for fully demonstrated technologies were based on experience obtained in EPC [engineering, procurement and construction] projects, engineering studies, owner s engineer and due diligence work, and evaluation of power purchase agreement (PPA) pricing. Costs for other technologies or advanced versions of demonstrated technologies were based on engineering studies and other published sources. [p.5] The report included three storage technologies. Projected overnight capital costs, in 2009 dollars, are shown in Table 13, Table 14, and Table 15 for compressed air energy storage (CAES), pumped hydro storage (PHS), and sodium sulfide (NaS) battery storage respectively. While these tables give a sense of costs, the report should be consulted prior to use in order to understand general and technologyspecific assumptions as well as technology-specific uncertainties. As a point of comparison, Table 16, Table 17, and Table 18 provide cost projections for gas turbines, combined cycle plants, and wind plants. 48 Table 13. Cost and Performance Projection for 262 MW, 15 Hour Compressed Air Energy Storage Plant (Black & Veatch 2012) FOR = forced outage rate, POR = planned outage rate 47 Black & Veatch is a leading global engineering, consulting and construction company specializing in energy, among other market areas The report also provides projections for eight different types of solar PV installations and for numerous other conventional and renewable technologies. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 49

62 Table 14. Cost and Performance Projection for 500 MW, 10 Hour Pumped Hydro Storage (PHS) Plant (Black & Veatch 2012) Table 15. Cost and Performance Projection for 7.2 MW, 8.1 Hour Sodium Sulfide Battery Storage Plant (Black & Veatch 2012) p. 50 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

63 Table 16. Cost and Performance Projection for 211 MW Gas Turbine Plant (Black & Veatch 2012) Table 17. Cost and Performance Projection for 580 MW Combined-Cycle Plant (Black & Veatch 2012) The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 51

64 Table 18. Cost and Performance Projection for Onshore Wind (Black & Veatch 2012) The DOE/EPRI Electricity Storage Handbook (Akhil et al. 2013) also provides cost and performance data based on a 2011 study in which EPRI conducted detailed surveys of suppliers and system integrators (EPRI 2012). The handbook covers a much broader array of storage technologies than the Black & Veatch report, as well as a broader range of DTRPs for each technology. However, it estimates only then-current costs, not costs over time. Some technologies (e.g., lithium ion batteries) are evolving rapidly, and in those cases the actual storage component of plant costs may be lower today than it was in The project cost components included and excluded in the DOE/EPRI handbook are also different from those in the Black & Veatch estimates. Finally, in the main body of the report, costs were presented as present value of installed costs, life-cycle cost of energy, and life-cycle cost of capacity, which are not comparable to the overnight capital costs provided by Black & Veatch. Appendix B of the handbook does show overnight capital costs for many storage system types, applied for various anchor services (primary uses). Some of these capital cost estimates are summarized in Table 19. See the handbook for assumptions and caveats. Lazard 49 (2015) partnered with Enovation 50 to produce levelized cost estimates ($/MWh) and capital cost estimates for energy storage in a number of use cases, including five in front of the meter applications (transmission system, peaker replacement, frequency regulation, distribution services, and PV integration). The costs included and excluded are different from those in the Black & Veatch and DOE/EPRI analyses. Their summary graphics cannot be reproduced without permission. See the report for assumptions, caveats, and estimates. 49 Lazard is a preeminent financial advisory and asset management firm Enovation Partners focuses exclusively on energy and infrastructure We focus on strategy and innovation in distributed energy, renewables, and natural gas. p. 52 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

65 Table 19. Overnight Capital Costs for Selected Storage Types in Various Applications (Akhil et al. 2013) Storage Type (Vendor) Anchor Service (Application) Capacity, Net MW DTRP, hrs Total plant cost, $/kw Pumped Hydro Bulk Storage ,500 Pumped Hydro Bulk Storage ,200 Pumped Hydro Bulk Storage ,700 Pumped Hydro Bulk Storage ,850 CT-CAES (Above Ground) (S12-2) Bulk Storage ,762 Brayton-CAES (Below Ground) (S9-1) Bulk Storage ,050 Brayton-CAES (Below Ground) (S9) Bulk Storage NaS (S36) Bulk Storage ,168 NaS (S36) Bulk Storage ,071 NaS (S36) Utility T&D ,152 NaS (S36) Utility T&D ,434 Advanced Lead Acid (S11) Bulk Storage ,326 Advanced Lead Acid (S15) Bulk Storage ,897 Advanced Lead Acid (S13) Bulk Storage ,743 Advanced Lead Acid (S15) Bulk Storage ,876 Advanced Lead Acid (S15) Freq. Reg. & Renewables Integration ,176 Advanced Lead Acid (S11) Freq. Reg. & Renewables Integration ,692 Advanced Lead Acid (S15) Utility T&D 1 4 4,855 Advanced Lead Acid (S11) Utility T&D ,360 Flywheel (S5) Freq. Reg. & Renewables Integration ,159 Li-ion (S19-1) Freq. Reg. & Renewables Integration ,017 Li-ion (S22) Freq. Reg. & Renewables Integration ,144 Li-ion (S7) Wind Integration 1 1 1,634 Li-ion (S6) Utility T&D ,183 Li-ion (S1) Utility T&D 3 1 1,388 Li-ion (S7) Utility T&D 1 4 4,420 Li-ion (S25) Distributed Energy Storage ,570 Li-ion (S22) Distributed Energy Storage ,904 Combustion Turbine Comparison 100 NA 720 Combined Cycle Gas Turbine Comparison 500 NA 1,100 Market and Policy Barriers to Increased Use of Storage DOE (2013, p. 4) summarizes the key barriers to wider use of energy storage as follows: Cost competitive energy storage technology Achievement of this goal requires attention to factors such as life-cycle cost and performance (round-trip efficiency, energy density, cycle life, capacity fade, etc.) for energy storage technology as deployed. It is expected that early deployments will be in high value applications, but that long term success requires both cost reduction and the capacity to realize revenue for all grid services storage provides. Validated reliability and safety Validation of the safety, reliability, and performance of energy storage is essential for user confidence. Equitable regulatory environment Value propositions for grid storage depend on reducing institutional and regulatory hurdles to levels comparable with those of other grid resources. Industry acceptance Industry adoption requires that they have confidence storage will deploy as expected, and deliver as predicted and promised. In more detail, they explain (p 30-31): Cost competitive energy storage systems: The total cost of storage systems, including all the subsystem components, installation, and integration costs need to be cost competitive with other non-storage options available to electric utilities. While there is a strong focus on reducing the cost of the storage The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 53

66 components, such as batteries or the flywheel, the storage component still constitutes only 30% to 40% of the total system cost, thus the focus needs to be on the entire system. Additionally, there is a concurrent need to quantify the value of storage in the various services it provides to the grid, individually and in multiple or stacked services, where a single storage system has the potential to capture several revenue streams to achieve economic viability. This is important now and as the cost of storage systems decline to economically attractive levels. Validated performance and safety: The process for evaluating and reporting the performance of existing storage systems on a unified basis needs to be created. This combined with industry accepted codes and standards to specify desired performance parameters for each storage service, will lead to a wider acceptance of energy storage systems. For example, there is significant uncertainty over the usable life of batteries and the length of time that a storage installation can generate revenue; both of these issues directly impact investment calculations. According to stakeholder input, a fuller understanding of the true life of batteries through demonstrations and accelerated testing could help remove this barrier, since predicting reliability through improved testing is important in supporting commercialization. The operational safety of large storage systems is a concern and will be a barrier in their deployment in urban areas or in proximity of other grid resources such as substations. Design practices that incorporate safety standards and safety testing procedures for the different storage technologies need to be developed and codified. Equitable Regulatory Environment: Currently, a consistent pricing or market plan for providing grid storage does not exist and the uncertainty surrounding use-case economics inhibits investment. Without an established revenue generation model for storage operators, the case for investment will remain muted. While there have been demonstrations in areas such as frequency regulation, there are still enough revenue uncertainties in other applications to dissuade investment. Industry Acceptance: There is also significant uncertainty about how storage technology will be used in practice and how new storage technologies will perform over time in applications. Currently, systems operators have limited experience using deployed storage resources; stakeholder input suggests that development of algorithms to employ storage technology effectively and profitably could encourage investments. Similarly, today s utility planning, transmission and distribution design tools do not have the capability to analyze energy storage as an option on a consistent basis. Integrating storage into the planning tools that are currently used by industry (rather than developing stand-alone tools) could boost storage technologies. The same report also summarizes the strategies DOE is using to overcome these barriers and the specific activities they are funding in support of each of these strategies. Additional sources discussing barriers to energy storage include Bhatnagar et al (2013) and Sioshansi et al. (2012). In terms of cost, storage must compete with other technologies that can provide the same grid services. Currently, it is not cost-competitive in most applications (Lazard 2015). MISO s total wholesale power costs are currently much lower than those of the other ISOs reporting comparable data to FERC, with the exception of the Southwest Power Pool (Table 20). 51 This is likely to make storage less economically attractive in MISO than it may be in other markets. 51 SPP serves most of ND, SD, NE, all of KS and OK and parts of MT, MN, IA, MO, AK, TX and NM. See Figure 8. p. 54 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

67 Table 20. ISO Wholesale Power Costs, $/MWh 2014 (source: PJM 2015) Energy Capacity Transmission Operating Reserves Ancillary Services RTO Cost & Regulatory Fees NCPC Total ISO-NE $69.21 $8.05 $13.87 $0.00 $1.21 $1.28 $1.28 $94.90 MISO $37.34 $0.00 $3.00 $0.00 $0.08 $0.29 $40.71 NYISO $67.13 $21.05 $0.63 $0.40 $0.51 $0.69 $90.41 PJM $53.13 $8.91 $5.75 $1.23 $1.06 $0.32 $70.40 SPP $34.32 $0.00 $4.30 $0.50 $0.00 $0.43 $39.55 $100 $90 $80 $70 $60 $50 $40 $30 $20 $10 $0 ISO-NE MISO NYISO PJM SPP Energy Transmission Ancillary Services NCPC Capacity Operating Reserves RTO Cost & Regulatory Fees Figure 21. ISO Wholesale Power Costs, $/MWh, 2014 (source: PJM 2015) At present, it would be impossible for a combustion turbine or combined cycle plant, let alone a storage facility, to enter MISO on a merchant plant basis: In 2014, MISO markets would not have supported investment in either gas CT or CC generation units based on their annualized costs of new investment. The MISO footprint has a capacity surplus that prevents significant periods of shortage, particularly at reduced load levels (PJM 2015 p. 187). Figure 22, from MISO s independent market monitor s 2015 report, compares the net revenue (revenue above variable cost of production) a new generator would earn if it ran only when economic (bars) to the estimated annualized capital cost of a new combustion turbine (CT) or combined cycle plant (CC) The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 55

68 (Potomac Economics 2015). Note that the net revenues include stacked benefits of capacity, energy, and ancillary services. Figure 22. Comparison of Estimated Net Revenue to Estimated Annualized Cost of New Entry, Midwest Region (Potomac Economics 2015) Moreover, MISO does not operate a true capacity market. They do have a resource adequacy mechanism, through which they establish the minimum amount of capacity a load serving entity (LSE) must have, including reserve margins. LSEs can meet this requirement through so-called Fixed Resource Adequacy Plans [FRAPs], bilateral transactions, self-scheduling, capacity deficiency payments, and through MISO s voluntary Planning Resource Auction (PRA) (PJM 2015, p. 175). Prices in the PRA most recently cleared at about 7% of the cost of new entry (Potomac Economics 2015 p. v.), partly because of available excess capacity and partly because the annual PRA is really too short term to form the basis for new plant construction. Through the FRAP mechanism, utilities that have to submit an integrated resource plan for approval by their utility commission can submit the approved plan to satisfy their capacity requirements. 52 Plans approved by utility commissions of course allow utilities to recover capital costs through their rate base. Practically speaking, the current situation is such that a storage 52 In MISO Zone 1 (which includes MN, ND, and WI) in the planning year, about two-thirds of the required capacity was obtained through FRAPs, with the remainder cleared in the annual auction at a very low price of $3.29 per MW-day (vs. an average price for MISO of $16.75 per MW-day), due to transmission constraints on exports from Zone 1 (Potomac Economics 2015 p. 20). p. 56 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

69 system supplier would have to work with a utility to include a storage system in their integrated resource plan in order to secure cost recovery. 53 The other element of DOE s cost-competitiveness barrier is the question of how to quantify and then capture the value of the services that energy storage can provide to the grid. This is an area of considerable research and discussion. One complexity is that the value is influenced by the system in which the storage is installed and the specific location where it is installed. For example, the value of energy storage for arbitrage depends on which fuels are on the margin (i.e., setting prices) in on-peak and off-peak periods, and this varies for different electric systems (and over time for any one system). The value of energy storage for frequency regulation likewise depends on what other mechanisms, such as sub-hourly energy markets, are used to balance system output and load in particular systems (among other factors that may influence the price of regulation). The value of energy storage to relieve transmission congestion or defer transmission upgrades depends on the specific location of the storage system with respect to transmission constraints. The capacity value of storage depends on whether a particular system currently needs capacity or has a surplus. Moreover, the installation of a substantial amount of storage on a system can change the prices that may have been key to its cost-effectiveness. For example, storage used for arbitrage increases demand at off-peak times and may thereby increase off-peak prices; likewise, it increases supply at on-peak times and may thereby decrease on-peak prices (Denholm et al. 2013, MISO 2011). MISO has shown (Chen et al section II.B & C) that when short-term energy storage used for frequency regulation grows to the point of providing all regulation on a system, it can cause regulation prices to drop far below the price for contingency reserves, because short-term storage does not have a DTRP long enough to meet standards for contingency reserves. Similarly Denholm et al. (2013) have noted that the highest value service (regulation reserves) [that storage can provide] represents a relatively small market opportunity, and the introduction of large amounts of storage might quickly collapse the market for this service. Many have concluded (e.g., Eyer & Corey 2010, EPRI 2010, Akhil et al. 2013) that at least for the near term, storage is most likely to be cost-effective when it can serve more than one function. Yet there are unresolved questions as to how storage systems should be modeled in system planning and used in system operation in order to reliably provide multiple services at the same time: [E]conomics depend on the compatibility of multiple applications for shared storage capacity, both in terms of revenue recovery through markets and regulatory structures (less transaction costs), and in terms of technical and operational feasibility [DOE 2011]. Electricity storage can be used for any of the services listed above, but it is rare for a single service to generate sufficient revenue to justify its investment. However, the flexibility of storage can be leveraged to provide multiple or stacked services, or use cases, with a single storage system that captures several revenue streams and becomes economically viable. How these services are stacked depends on the location of the system within the grid and the storage technology used. However, due to regulatory and operating 53 The most recent information found (2009) indicated that stakeholders have divided opinions about MISO s resource adequacy construct. Some like the current construct, some want a mandatory forward capacity market, and some favor an energy-only approach (in which capacity is funded through energy shortage pricing). Many states with traditionally regulated utilities were opposed to a mandatory forward capacity market (Brattle Group 2009, Newell et al. 2010). The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 57

70 constraints, stacking services is a process that requires careful planning and should be considered on a caseby-case basis. [Akhil et al. 2013]. In addition, if storage is most cost-effective in situations where it can provide a value stack of multiple benefits, this inevitably reduces the market potential to those situations where all of those benefits (services) are needed. For example, a study by EPRI (2010) notes: The analysis summarized in this paper indicates that capturing multiple benefits, including transmission and distribution (T&D) deferral, local or system capacity, and frequency regulation, is key for high-value applications. Applications that achieve the highest revenues do so by aggregating several benefits across multiple categories. The number of locations at which all of these benefits can be realized together, however, is limited. One conclusion that can be drawn from these complexities is that the value of storage cannot be accurately assessed without the use of detailed production cost simulations (Denholm et al. 2013). These analyses are time-consuming and expensive, and they require that the analyst has a full model of the system in question. Only system planners and operators or researchers are likely to have such a model. In addition, the models have not historically included algorithms designed to simulate storage and capture all of the trade-offs of operating to provide different services at different times. Although detailed production cost simulations are the best tool with which to obtain full and accurate evaluation of storage, Akhil et al. (2013, p ) do provide a multi-step framework that utilities can use to evaluate storage systems. This framework delays costly production cost simulations until after several preliminary screening steps. Batnagar & Loose (2012) also provide guidance on how regulators can assess the value of storage projects proposed by utilities. Assuming that the value of storage projects can be quantified, a second hurdle is actually capturing those value streams. For example, Denholm et al. (2013) showed that the payment a merchant storage facility is likely to receive in an ISO environment is lower than the value the facility might have in a vertically integrated utility environment. Even within a given electrical system, there are institutional and modeling issues. While different entities may have the lead responsibility for planning and approving transmission and generation, and may use different processes and modeling tools to do so, a storage facility might have value for both purposes. A further issue is that the day ahead and real-time commercial models that commit and dispatch different resources do not include the algorithms they would need to optimize dispatch of storage for different functions. For example, as described in the Addition of Short-Term Stored Energy Resource section, MISO had to develop an algorithm that would optimize the operation of short-term storage simply to provide the single service of frequency regulation. Battery storage manufacturers have pointed out that MISO s model would not optimize dispatch of battery systems with longer DTRPs that might be able to provide contingency reserves, arbitrage, congestion relief, or other services in addition to regulation (IPL 2016). At the other extreme of storage system types, MISO s current algorithms for pumped-hydro storage allow it to participate in the market as a load, a demand side resource for demand reduction, or generation, but it is up to the owner to decide when to bid or offer it for each of these uses. MISO notes two challenges with their current approach: (1) charging and discharging are treated separately, so it is possible for the storage unit s generation to clear in the day-ahead market but for the load not to clear, leaving the storage owner committed to provide power from elsewhere at an unknown cost; and (2) if p. 58 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

71 the facility has multiple turbines, each is treated as a separate unit, making the arbitrage more complicated for the plant operator to work out. Also, at least as of 2011, the unit dispatch model did not include a look-ahead functionality to forecast when arbitrage was optimal. These problems would be even more significant for mid-term storage systems, which have substantially shorter DTRPs than pumped-hydro plants. Existing and Planned Storage on the MISO System The DOE energy storage database (DOE, no date) currently shows 25 storage projects within MISO. Their status is summarized in Table 21. Table 22 shows the projects by technology type and total capacity. The list includes some ice thermal storage projects. Thermal storage is a load-shifting (peak demand reduction) technology and, while it is typically cheaper than storage that can produce electricity (Lazar & Linvill 2016), it is not relevant to this project, which focuses on storage that can be used to produce electricity. In addition, other common thermal storage projects are not listed in the database, notably large electric water heaters that heat large volumes of water overnight at low rates for use during the day. For that reason, the summary is repeated in Table 23 without the ice thermal storage projects. Table 21. Projects in the DOE International Energy Storage Database Located in MISO Status Freq. Percent Operational 21 84% Under Construction 1 4% Announced 1 4% De-Commissioned 1 4% Offline/Under Repair 1 4% Total % Table 22. Storage Projects in MISO by Technology Type and Capacity (DOE) Technology Type Freq. Total Rated Power in kw pct of kw Open-loop Pumped Hydro Storage 3 2,370, % Lithium-ion Battery 4 21, % Ice Thermal Storage 4 4, % Sodium-sulfur Battery 1 1, % Zinc Iron Flow Battery 1 1, % Flywheel % Advanced Lead-acid Battery % Lead-acid Battery % Lithium Ion Titanate Battery % 25 2,399, % The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 59

72 Table 23. Storage Projects in MISO by Technology Type and Capacity, Excluding Thermal Storage (DOE) Technology Type Freq. Total Rated Power in kw pct of kw Open-loop Pumped Hydro Storage 3 2,370, % Lithium-ion Battery 4 21, % Sodium-sulfur Battery 1 1, % Zinc Iron Flow Battery 1 1, % Flywheel % Advanced Lead-acid Battery % Lead-acid Battery % Lithium Ion Titanate Battery % Total 21 2,394, % Of the 21 storage projects that produce electricity when discharging, 99% of the total capacity is in three large pumped hydro projects (Table 23). Further information about these is given in Table 24. The two that provide data on discharge time have discharge times of 8 hours. Table 24. Pumped Hydro Storage Projects in MISO (DOE) Project Name Rated Power, kw Duration HH:MM State Utility Grid Interconnection Ownership Model Clarence Cannon Dam Pumped Storage 58,000 00:00.0 Missouri Transmission Third-Party- Owned* Ludington Pumped 1,872,000 08:00.0 Michigan Consumers Transmission Utility-Owned Storage Energy Taum Sauk Hydroelectric Power Station 440,000 08:00.0 Missouri Ameren Transmission Utility-Owned *Owned by Army Corps of Engineers The next largest group, accounting for 0.9% of capacity, consists of four lithium-ion battery storage projects ranging from 15 kw to 20 MW. The three for which data is given have discharge times ranging from 30 minutes to 5 hours and 20 minutes (Table 25). Note that the IPL 20 MW project is listed in the DOE database as under construction. 54 There is also one lithium-ion titanate battery storage project (Table 26). 54 Time did not permit a detailed search of all 617 U.S. storage projects in the DOE database to verify whether any had been misclassified as to ISO. However, it was noted that two Entergy projects were not classified as in MISO (and so are not included in the tabulated data above), although Entergy is now a member of MISO. One of these projects is a 10 kw, two hour discharge lithium-ion battery system paired with a 25 kw rooftop PV system that is located in Louisiana. This is an operational, customer-owned project in Louisiana connected to the secondary distribution system. The other is a 1 MW, 30 minute discharge lithium ion battery system paired with a 1 MW solar array. This project has been announced for construction in 2016 and will be a utility-owned system located in Louisiana. It appears to be a pilot aimed at gaining more experience with and evaluating the performance of a paired solar/battery system. More information on that project is available at p. 60 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

73 Table 25. Lithium-Ion Battery Storage Projects in MISO (DOE) Project Name EnerDel Mobile Hybrid Power System DTE Community Energy Storage for Grid Support - Monroe Community College CES Unit DTE Community Energy Storage for Grid Support - Residential CES Units Indianapolis Power and Light - 20 MW Rated Power, kw Duration HH:MM State Utility Grid Interconnection Ownership Model 15 05:20.0 Illinois Customer- Owned* :30.0 Michigan Detroit Primary Utility-Owned Edison Distribution Energy :00.0 Michigan Detroit Edison Energy 20,000 00:00.0 Indiana Indianapolis Power & Light Secondary Distribution Utility-Owned Utility-Owned *U.S. Army Engineer Research & Development Center (ERDC), Construction Engineering Research Laboratory (CERL) project Table 26. Lithium-Ion Titanate Battery Storage Projects in MISO (DOE) Project Name Clay Terrace Plug-In Ecosystem Rated Duration Grid Ownership Power, kw HH:MM State Utility Interconnection Model 75 00:40.0 Indiana Duke Energy Utility-Owned Sodium-sulfur batteries, zinc-iron flow batteries, flywheels, and advanced lead-acid batteries are each represented by a single medium-sized project in MISO. The sodium-sulfur battery project (Table 27) is located in Minnesota and owned by Xcel Energy. It has a capacity of 1 MW and a discharge time of 7 hours 12 minutes. It was commissioned October 1, Its service/use cases are listed as: 1. Renewables energy time shift 2. Ramping 3. Voltage support 4. Frequency regulation Table 27. Sodium-Sulfur Battery Storage Projects in MISO (DOE) Project Name XCEL MinnWind Windto-Battery Project Rated Power, kw Duration HH:MM State Utility Grid Interconnection Ownership Model :12.0 Minnesota Xcel Energy Transmission Utility-Owned The project is described in the database as follows: In October 2008, Xcel Energy began testing a one-megawatt battery-storage technology to demonstrate its ability to store wind energy and move it to the electricity grid when needed. Xcel Energy purchased the battery from NGK Insulators Ltd. The sodium-sulfur battery is commercially available and versions of this technology are in use elsewhere in the U.S. and other parts of the world, but this is the first U.S. application of the battery as a direct wind energy storage device. The project is being conducted in Luverne, Minn., about 30 miles east of Sioux Falls, S.D. The battery installation is connected to a nearby 11-megawatt wind The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 61

74 farm owned by Minwind Energy, LLC. The project received a $1 million grant from Xcel Energy s Renewable Development Fund. The zinc-iron flow battery project (Table 28) is listed in the DOE database as announced. It is expected to have 1 MW capacity and have a discharge time of 2 hours. The flywheel project (Table 29) is 860 kw and has a discharge time of 0.4 minutes. The advanced lead-acid battery project (Table 30) is listed in the DOE database as offline/under repair. It has a capacity of 750 kw and a discharge time of 2 hours and 40 minutes. Table 28. Zinc Iron Flow Battery Storage Projects in MISO (DOE) Project Name Comed-Bronzeville- Demo Rated Duration Grid Ownership Power, kw HH:MM State Utility Interconnection Model :00.0 Illinois ComEd Utility-Owned Table 29. Flywheel Storage Projects in MISO (DOE) Project Name Delta Dental Data Center VYCON Flywheels Rated Power, kw Duration HH:MM State Utility Grid Interconnection Ownership Model :00.4 Michigan Customer- Owned Table 30. Advanced Lead-Acid Battery Storage Projects in MISO (DOE) Project Name Ford Manufacturing Assembly Plant Xtreme Power 750 kw (Dearhorn, Michigan) Rated Power, kw Duration HH:MM State Utility :40.0 Michigan Detroit Edison Energy Grid Interconnection Ownership Model Third-Party- Owned The nine remaining storage projects listed in MISO are lead-acid battery projects (Table 31). 55 All of these are in Minnesota, and all belong to either rural electric coops or municipal utilities. They range in size from 5 to 115 kw 56 and in discharge time from 2 to 3 hours. All are listed as operational except the Brainerd Public Utilities project, which is listed as decommissioned 3/1/2014. All are reported to be on the customer side of the meter and remotely controlled by the utility. The storage technology provider is listed as Silent Power Inc. for all except the Shakopee Public Utilities project, which does not list a provider. 55 Time did not permit resolution of the apparent overlap between the first project listed, which encompasses four utilities, and the individual projects listed for the same utilities. 56 See preceding note regarding the 115 kw case. p. 62 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

75 Table 31. Lead-Acid Battery Storage Projects in MISO (DOE) Project Name National Rural Electric Cooperative Association/Cooperativ e Research Network Distributed Energy Storage Research Project Wright-Hennepin Solar Community Austin Utilities Energy Storage Pilot Shakopee Public Utilities Environmental Learning Center NRECA/CRN Distributed Energy Storage Research Project (Federated) NRECA/CRN Distributed Energy Storage Research Project (Meeker) NRECA/CRN Distributed Energy Storage Research Project (Minnesota Valley Electric) NRECA/CRN Distributed Energy Storage Research Project (Wright- Hennepin) Brainerd Public Utilities Battery Pilot Rated Power, kw Duration HH:MM State Utility :00.0 Minnesota REC's: Minnesota Valley, Wright- Hennepin, Meeker, Federated Grid Interconnection Secondary Distribution 37 02:00.0 Minnesota Wright-Hennepin Secondary Cooperative Electric Distribution Association 37 02:00.0 Minnesota Austin Utilities Secondary Distribution 9 02:00.0 Minnesota Shakopee Public Utilities Commission 5 02:00.0 Minnesota Rural Electric Cooperative 5 02:00.0 Minnesota Rural Electric Cooperative 33 02:30.0 Minnesota Rural Electric Cooperative 51 02:00.0 Minnesota Rural Electric Cooperative 5 02:00.0 Minnesota Brainerd Public Utilities Secondary Distribution Secondary Distribution Secondary Distribution Secondary Distribution Secondary Distribution Secondary Distribution Commissioning Ownership Model Date Utility-Owned 8/1/2013 Utility-Owned 9/9/2013 Utility-Owned Utility-Owned 10/2/2013 Utility-Owned Utility-Owned Utility-Owned 1/1/2013 Utility-Owned Utility-Owned Most of the lead-acid battery projects appear to be geared primarily toward load-shifting, demand charge reduction, and reliability (Table 32 and Table 33), although the Shakopee Public Utilities 9 kw project has renewables capacity firming and renewables energy time shifting as its first and second purposes and the Wright-Hennepin Solar Community project has these as its second and third purposes. These two storage projects are co-located with solar arrays. Some of the other projects list renewables capacity firming or renewables energy time shifting as a third or fourth purpose. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 63

76 Table 32. Lead-Acid Battery Project Service/Use Case Information (DOE) Project Name ServiceUseCase1 ServiceUseCase2 ServiceUseCase3 ServiceUseCase4 National Rural Electric Cooperative Association/Cooperative Research Network Electric Energy Time Shift Grid-Connected Residential Renewables Energy Time Shift Distributed Energy Storage Research Project (Reliability) Wright-Hennepin Solar Community Electric Bill Management Renewables Capacity Firming Renewables Energy Time Shift Austin Utilities Energy Storage Pilot Electric Bill Management Electric Energy Time Shift Shakopee Public Utilities Environmental Learning Center NRECA/CRN Distributed Energy Storage Research Project (Federated) NRECA/CRN Distributed Energy Storage Research Project (Meeker) NRECA/CRN Distributed Energy Storage Research Project (Minnesota Valley Electric) NRECA/CRN Distributed Energy Storage Research Project (Wright-Hennepin) Brainerd Public Utilities Battery Pilot Renewables Capacity Firming Electric Energy Time Shift Electric Energy Time Shift Electric Energy Time Shift Electric Energy Time Shift Electric Energy Time Shift Renewables Energy Time Shift Grid-Connected Residential (Reliability) Grid-Connected Residential (Reliability) Grid-Connected Residential (Reliability) Grid-Connected Residential (Reliability) Electric Bill Management Grid-Connected Residential (Reliability) Renewables Capacity Firming Renewables Capacity Firming Renewables Energy Time Shift Renewables Energy Time Shift Table 33. Lead-Acid Battery Storage Project Descriptions (DOE) Project Name National Rural Electric Cooperative Association/Cooperative Research Network Distributed Energy Storage Research Project Wright-Hennepin Solar Community Austin Utilities Energy Storage Pilot Shakopee Public Utilities Environmental Learning Center NRECA/CRN Distributed Energy Storage Research Project (Federated) NRECA/CRN - Distributed Energy Storage Research Project (Meeker) NRECA/CRN Distributed Energy Storage Research Project (Minnesota Valley Electric) NRECA/CRN Distributed Energy Storage Research Project (Wright-Hennepin) Brainerd Public Utilities Battery Pilot Description Instillation includes eleven 5 kw/10 kwh and six 10 kw/20 kwh Silent Power storage units installed at cooperative member locations as part of DOE funded CRN administered study on residential storage kw/94.4 kwh hour Silent Power storage DC coupled with 31 kw community solar array. Four 9.2 kw/23.6 kwh Silent Power storage units installed in municipal buildings for peak demand management. 9.2 kw/23.6 kwh Silent Power storage unit co-located with solar array at high school environmental learning center. 4.6 kw/11.8 kwh Silent Power storage unit installed at utility office building for demand charge reduction and backup power. This Silent Power system uses one 4.6 kw/11.8 kwh Silent Power storage unit, which has been installed at the utility office building for demand charge reduction and backup power. This installation incorporates three 4.6 kw/11.8 kwh and two 9.6 kw/23.6 kwh Silent Power storage units for demand charge reduction and backup power. This system incorporated seven 4.6 kw/11.8 kwh and two 9.6 kw/23.6 kwh Silent Power storage units installed for demand charge reduction and backup power. This installation incorporates one 4.6 kw/11.8 kwh Silent Power storage unit installed at utility office building for demand charge reduction and backup power. p. 64 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

77 A project that is not yet listed in the DOE database is the solar plus battery storage demonstration project that Xcel Energy proposed to the Minnesota PUC in October As they describe the project (Xcel Energy 2015, p. 16): The City of Belle Plaine is scheduled for a new substation installation within the next five years due to the fact that the existing substation is nearing capacity. We are pursuing a new parcel of land for the new substation because the existing site is too small for full distribution build out and for the eventual conversion of the transmission source from 69 kv to 115 kv. We are also exploring plans to add a large battery (we have investigated batteries sized at 6 MWH, 2 MW output to meet the forecasted 2017 feeder peak) to reduce the load on a Belle Plaine feeder and transformer. We are also pursuing a 1 MW solar array to service in combination with the substation property to explore the benefits and complexities of storage working in conjunction with a variable generation resource. Our expectation is that if we go forward with the solar portion of the project, the solar contribution may shift the load curve so that the feeder has a smaller overload for a shorter length of time. They describe the expected benefits of the project as follows (Xcel Energy 2015, p ): The primary benefit to be explored is the ability to defer distribution capital investments associated with overloads. However, we also expect there will be additional benefits and learnings that we can apply to our overall service territory. These benefits include the following: Volt/Var Control The battery system would have the ability to assist in optimizing feeder voltage and reactive power flow either as a standalone system or in conjunction with the IVVO module of the ADMS system. Loss Impact Analysis Transmission and Distribution system loss impacts would be studied for the various individual and stacked operational modes of the battery system (i.e. peak demand reduction, Volt/Var, and DER smoothing, etc.). Regulation Data collected from the site could be used to explore the battery energy storage system s potential capability to participate in MISO regulation markets. Power Quality The battery system could have the ability to mitigate short duration events, such as voltage sags, to demonstrate capabilities to protect sensitive customer loads. DER Smoothing Using the battery system to smooth the output of DER, specifically PV in this case, would demonstrate capabilities to reduce voltage fluctuations in order to increase feeder DER hosting capacity and minimize maintenance costs for voltage regulation equipment. Recent studies on energy storage have supported the importance of capturing more than one value stream for projects to be economically feasible. We intend to learn about complex interactions and limitations involved with stacking multiple battery systems. For example, while we view the primary objectives as renewable integration and a battery energy storage system deferring a distribution substation. Other objectives include answering questions such as the following (1) what battery capability remains to participate in MISO regulation markets, (2) what is the total value and (3) how can the overall benefit be optimized? Reaching conclusions on these types of research questions, with field data specific to conditions in Minnesota, will aid in our ability to integrate energy storage into the grid in a way that is most costeffective and beneficial to our customers. Xcel Energy s filing also describes several other options they considered for a battery demonstration project, all of which would seek to have batteries defer a distribution capacity project. There are currently three battery storage projects in the MISO generator interconnection queue, totaling 70 MW. 57 One is the 20 MW Indianapolis Power & Light (IPL) project listed in the DOE database 57 Note that only projects that connect at the transmission level are included in the queue. The queue can be searched at The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 65

78 as under construction. In the MISO queue, it has a projected in-service date of March 31, 2016, and is also listed with a done status as of March 19, The other two are not listed in the DOE database. One of these is a 30 MW International Transmission Company (ITC) project located in Huron County, Michigan. In has a projected in-service date of October 1, The third is a 20 MW ITC Midwest project in Mower County, Minnesota. Its projected in-service date is September 15, No further information is available about these two ITC projects. MISO Studies and Activities Related to Storage MISO has conducted a number of past studies related to energy storage, and is currently involved in several activities that have a bearing on energy storage. These are described below. MISO Energy Storage Study Phase 1, 2011 MISO conducted their first study of energy storage in The work was published by EPRI (2011). Key drivers of the study were (1) state renewable portfolio standards, which necessitated that MISO continually integrate higher levels of variable renewable generation; (2) ongoing discussions and rulings between FERC and MISO regarding how storage should be treated in the ancillary services market; and (3) the need to enhance MISO s ability to model storage for use in transmission and capacity planning. They noted that MISO already had 2500 MW of pumped storage capacity at that time, had done some feasibility analysis for the proposed Iowa Stored Energy Project (a CAES project), and had worked with Xcel on a solid-state dry cell battery project. Phase 1, the only phase completed, sought to identify circumstances when adding energy storage resources to the MISO footprint is justified economically, over a 20 year capacity expansion planning period starting in The analytical work was done primarily with EPRI s Electric Generation Expansion Analysis System (EGEAS), 58 a tool designed to find the least cost integrated resource plan (IRP) (i.e., minimize the present value of revenue requirements) for a given amount of demand. EGEAS selects the resources that have the highest benefit to cost ratio, with the costs including long term capital costs. In EGEAS, storage resources take energy from the system during off-peak hours and provide energy during peak hours (EPRI 2015). Thus EGEAS considers as benefits only the energy arbitrage value of storage and its capacity. EGEAS is not an hourly production cost simulation tool; rather, it performs optimization using load duration curves. For that reason, it is not able to consider ancillary services benefits, does not model and consider the transmission congestion market, and does not have the granularity that might be required to capture optimal arbitrage economics. 59 However, because it is a simplified model that runs in a few hours, it permits simulation of many scenarios in a relatively short time. MISO s plan was to use EGEAS to identify scenarios that offered the best potential for storage, and then model them in more detail using their hourly production cost simulation model, PLEXOS (EPRI 2011). The storage facilities considered were a pumped hydro storage plants with capacity of 2400 MW, CAES plants with capacity of 2160 MW, and battery storage plants with capacity of 200 MW. Assumed discharge times were not reported. For each type of facility, a range of three different capital costs was assumed. 58 EGEAS is normally used by MISO for their long term resource adequacy planning. 59 MISO points out that EGEAS may also overstate the arbitrage opportunity because EGEAS has perfect hindsight as to when prices are highest and lowest, but an energy storage plant operator does not have perfect foresight of these prices. p. 66 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

79 They found that EGEAS only selected storage in 18 of the 405 sensitivity cases 60 and, in all 18 cases, only CAES was chosen due to it having the lowest modeled capacity cost of the three storage types. Furthermore, CAES was only selected when the lowest capital cost was used ($833 kw). Storage was only chosen when the cost of CO 2 emissions was zero, because the off-peak energy used to charge the storage is primarily coal, and it is not low-cost when a carbon emissions cost is added. As gas prices increased, storage tended to get selected for installation earlier in the study period, because a CAES plant uses less gas than conventional combined cycle or gas turbine plants. As renewable requirements increased, storage tended to get selected later and in smaller quantities because high wind penetrations reduced on-peak prices thereby reducing the energy arbitrage value of storage. They explained that, [w]here the EGEAS model did identify economic benefit from energy arbitrage, it was restricted by two significant market factors. The first is that MISO has more than enough existing generation capacity, including abundant coal generation. A significant proportion of the coal plant fleet is considered must run and therefore runs during off-peak hours. The need to keep coal plants running off-peak reduces the impact that free wind generation has in bringing down the off-peak power price since the system operator curtails the wind if the capacity is not needed. The consequence is that higher off-peak prices reduce energy arbitrage (and that the reduction is magnified when carbon emission tax costs are added to coal prices). The second market factor that reduces economic benefit from energy arbitrage is that wind energy is treated as nondispatchable and, due to the wind profile, some wind penetration is experienced even during peak hours. The EGEAS model uses wind first whenever it is available before considering alternative resources such as stored energy. During peak hours the result is that wind generation is effectively netted out of the load duration curve, which, in situations with higher wind penetrations, results in coal being the marginal unit. Lower peak prices squeeze energy arbitrage benefits from stored resources. Coal plants are run as baseload (i.e. 24 X 7) because their costs are low and a significant proportion (15,000 MW) of coal units are must run and are used for off-peak generation regardless of alternatives. In the EGEAS model, baseload off-peak is therefore using coal quite frequently (not wind) because the coal has to run. Baseload coal is quite inexpensive to run and, in fact, wind helps to lower the price even further. The arbitrage is lost when large amounts of wind force coal to be on the margin during peak demand. When that happens, the only energy arbitrage available is the difference between cheap, must-run, efficient coal and more expensive, less efficient, on peak coal. As discussed in the Advanced Control of VG section, most of the wind generation on the MISO system is no longer considered to be non-dispatchable and so is no longer manually curtailed. In addition, the amount of excess capacity has declined somewhat, especially since integration of the South Region into MISO (Figure 23), and coal plant retirements have accelerated. As they noted, if more coal plants are retired for environmental or end of life reasons, the arbitrage value of storage will increase. For these reasons, the results of the Phase 1 study are less applicable to current circumstances. 60 The 405 sensitivities included 5 natural gas prices * 3 levels of state and federal renewable portfolio standards * 3 levels of coal plant retirements due to EPA rules * 3 sets of construction capital costs * 3 CO 2 costs. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 67

80 Figure 23. MISO Planned and Actual Reserves (PJM 2015, p. 175) In the 2011 study they planned to use PLEXOS, an hourly production cost simulation model, to better understand how storage could best be used within the MISO market, but only ended up doing some preliminary testing with PLEXOS. This did, however, provide insights that enabled them to improve their modeling of energy storage in the later Manitoba Hydro Wind Synergy Study. Ultimately, they concluded that EGEAS isn t the right tool to understand storage potential, but that it is useful in identifying situations where storage is beneficial so that those cases can be analyzed further by PLEXOS. Energy Storage Study (MISO, Policy Studies), 2014 In July 2014 MISO published another report on energy storage (MISO 2014g). This appears to have been brought about by the fact that MISO was leading an initiative by Working Group 7 of GO to assess how increased amounts of variable renewables, demand response, and electric vehicles would affect the viability of large scale energy storage. They note several changes in the MISO system and modeling assumptions relative to the 2011 study, including changes in the MISO footprint and membership, power plant retirements and retrofits, and installation of new generation. Like its predecessor, this study used EGEAS for analysis, primarily because it was able to analyze a large number of scenarios with a limited amount of modeling time. They note again that this focuses on energy arbitrage and capacity benefits (contributions to the planning reserve margin) and is unable to 61 GO-15 is a voluntary initiative of the world s 18 largest Power Grid Operators (PGOs) representing together more than 70% of the world s electricity demand. p. 68 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

81 capture benefits from the ancillary services market. It also does not have the granularity of price data needed to capture optimal arbitrage economics, nor does it model congestion from transmission constraints. The study examined sensitivities to gas prices, renewable portfolio standards, carbon taxes, coal plant retirements, and storage capital costs. Gas prices again ranged from $4 to $12/million Btu. RPS mandates again ranged from about 10% to about 30% by The range of carbon taxes was reduced to $0 to $50/ton (vs. $0 to $100/ton in the 2011 study). Coal plant retirements ranged from 12.6 to 23 GW, 62 whereas in the 2011 study 12.6 GW had been the upper end of the coal retirement range. Capacities for all three storage types were assumed to be the same at 1200 MW. As with the 2011 study, discharge times were not stated. Overnight capital costs for storage systems were updated and changed rather substantially. The low end CAES plant cost increased 15%, while the high end cost decreased 23%, resulting in a range from $833 to $1667/kW. Battery systems underwent the same percentage changes, resulting in a cost range of $1667 to $3333/kW. Pumped hydro storage systems were assumed to be about two to three times as expensive in the 2014 study as in the 2011 study, resulting in costs ranging from $4050 to $5400/kW. 63 A total of 270 scenarios were run. The modeling assumed an amount of must-run coal equal to a percentage of MW from the first loading blocks of all must-run units on the system, recognizing that not all such units are online at the same time due to outages. The study does not explicitly state whether the effect of treating wind as a dispatchable intermittent resource was included in the modeling. The study also considered the potential impact of electric and plug-in hybrid cars and trucks. From 2006 to 2011, sales of these vehicles within the MISO footprint were reported to have accounted for 4.24% of total U.S. sales (Table 34). This percentage was assumed to remain the same over the study period, and growth was estimated by applying national growth projections from the EIA Annual Energy Outlook. The report also discussed the charging rates and load profiles assumed. Table 34. Electric and Plug-In Hybrid Vehicles in MISO's Footprint (MISO 2014g) The study concluded that: [C]urrent conditions in the MISO footprint do not find large-scale investment in storage to be economical. However, in certain scenarios, the energy arbitrage potential exists with coal units as the marginal unit during off-peak demand, and gas units as the marginal unit for peak demand. Furthermore, renewable penetration is found to have a positive impact on the energy arbitrage potential for storage because it helps bolster the amount of lower priced off-peak energy available for storage to utilize 62 This compares with 73.6 GW of installed coal capacity in 2014 (MISO 2014g). 63 The overnight construction costs assumed for gas turbines and combined cycle plants were $690/kW and $1045/kW respectively. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 69

82 Where the EGEAS model did identify economic benefit from energy arbitrage, the benefit was restricted by several factors... [T]he addition of the MISO southern region changes the resource mix of the system. The South footprint brings in a significant amount of gas units which increases the occurrence of gas units being the marginal units in off peak periods, thus reducing the potential for energy arbitrage With the current MISO footprint, the energy arbitrage potential exists with having coal units as the marginal unit during off-peak demand, and gas units as the marginal unit for peak demand. The divergence of production costs between coal and gas units creates the potential for storage to charge at the low marginal price of coal and discharge at the high marginal price of gas. [T]he potential for energy arbitrage was reduced [in scenarios where] the production costs of coal and gas units were similar. Renewable penetration was found to have a positive impact on the energy arbitrage potential for storage [R]enewable energy helps bolster the amount of lower priced off-peak energy available for storage to utilize. [By contrast, i]f there is low renewable penetration, the amount of baseload generation and available renewable energy would be at or slightly higher than the minimum demand of the system. When this occurs, there is little room for storage to benefit from energy arbitrage because baseload generation is not able to set the marginal price for enough periods. Figure 24. Annual Load Duration Curve in 2033 with Medium Retirements, Low Construction Costs, Gas at $10/MMBtu Carbon at $0/ton (MISO 2014g) The top 34 scenarios for storage (out of 270 total scenarios) selected 6 to 12 GW of storage over the time period from 2014 to The largest amounts of storage were generally selected when coal and age-related plant retirements were at the medium sensitivity level, 64 storage construction costs were low, and renewable portfolio standards were high. In those scenarios there was still enough coal generation that coal was on the margin off-peak (and accounted for 64 The high and low levels assumed 12.6 GW and 23 GW of coal-only retirements. The mid level assumed 12.6 GW of coal retirements and 11 GW of age-related retirements. p. 70 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

83 the great majority of energy used to charge the storage), while gas was on the margin on-peak (and accounted for the great majority of energy displaced when the storage discharged). The difference in cost between coal and gas generation provided the spread required for arbitrage. Figure 24 shows the annual load duration curve for one such scenario, illustrating the fact that there are hours when the sum of must-run coal plant output and wind output exceeds required generation, providing low-priced energy that can be used to charge storage for price arbitrage. Substantial amounts of storage were also selected in some scenarios where carbon costs were high ($50/ton) and gas prices were low ($4/MMBtu). In those scenarios, gas provided much of the baseload generation. The storage was charged with low-cost gas generation and discharged to displace high-carbon-cost coal. Whether gas prices would be likely to remain low with carbon costs of $50/ton burdening coal generation was not discussed. Compressed air energy storage was always selected over batteries and pumped hydro storage because of its much lower construction costs and higher efficiency. 65 This brings up the question of whether CAES is a feasible technology in the MISO footprint. The Iowa Stored Energy Park had attempted to develop a compressed air energy storage facility in a deep sandstone formation in Iowa (the Mt. Simon formation). After considerable development work, field tests showed that the permeability of the formation was not sufficient to allow the rates of air injection and withdrawal required for CAES (Schulte 2011). The only two existing CAES plants in the world use caverns in salt formations to store compressed air. Caverns like this either already exist from previous uses or are relatively lowcost to create via solution mining. Caverns offer very little resistance to air flow, unlike a porous rock formation. With their expansion into the Southern Region, MISO should have access to salt domes near the Gulf Coast that would provide suitable formations for CAES (Figure 25). However, storage that is not located close to the wind resource might not be able to take advantage of low energy prices from off-peak wind, depending on transmission constraints. There is currently a constraining 1,000-MW transfer limit between MISO Midwest (Zones 1-7) and MISO South (Zones 8 and 9) (Potomac Economics 2015, p. v), though whether this would hinder transfer at times of high wind generation is unknown. 65 They note again that the study did not consider provision of ancillary services, which might improve the economics for batteries. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 71

84 Figure 25. Major Salt Deposits in the U.S. (DeVries et al. 2005) Manitoba Hydro Wind Synergy Study The Manitoba Hydro Wind Synergy study (Bakke et al. 2013) analyzed the potential for hydro power from Manitoba to serve a storage-like function in mitigating the effects of variable renewables in MISO. Specifically, the study was intended to evaluate whether the cost of expanding the transmission capacity between Manitoba and MISO would enable greater wind participation in the MISO market. The overall concept was that at times with excess low-cost wind energy on the MISO system, this wind energy could be used by Manitoba Hydro to meet its load, and that the hydro power displaced by this wind power could be retained as water behind Manitoba Hydro dams and used later in the MISO system when needed. Whereas the two earlier storage studies looked at the overall potential in MISO, and therefore tested many scenarios at a high level, this study looked at a specific project and was intended to produce benefit/cost information accurate enough to enable MISO to decide whether to include the proposed transmission line in its transmission expansion plan. If the B/C analysis was favorable, it would also provide information needed to support Minnesota Power s application to the Minnesota PUC for a certificate of need for the transmission line. As a result, this project both required more detailed analysis and made that possible by choosing a specific location for the storage and limiting the scenarios to be examined. The study considered three transmission routing alternatives and two alternatives for sequencing the construction of hydro generation in Manitoba. It considered three energy futures drawn from MTEP12 studies, which differed in their assumptions about demand growth, energy growth, coal plant retirements, gas prices and price escalation, and wind penetration (see Bakke et al. 2013, p ). It p. 72 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

85 also considered low, median, and high water futures. In addition, it considered different wind offer prices including and excluding the production tax credit. Finally, in addition to analysis of the B/C ratios of new transmission lines, the study also analyzed the potential impacts of changing the use of the existing HVDC line ( External Asynchronous Resource ) between Minnesota and Manitoba from unidirectional (selling into the MISO market only) to bidirectional (both selling to and buying from the MISO market). MISO used PLEXOS to analyze the potential cost savings and wind curtailment reduction that could be realized. PLEXOS is typically set up to contain a detailed representation of the modeled system (in this case, much of the Eastern Interconnection) and used to simulate one year of system operation and production costs. As such, it is suitable for modeling specific projects in specific locations as opposed to analysis of a large number of scenarios. PLEXOS is capable of modeling with sub-hourly granularity and of employing the same simultaneous co-optimization strategy MISO uses to commit and dispatch resources for energy, regulation, spinning reserves, and non-spinning reserves. It therefore can estimate not only the energy arbitrage value of a particular storage project, but also the regulation and reserve value and transmission congestion impacts. MISO determined that to properly assess the impact of Manitoba Hydro storage on the system it would be necessary to simulate both the day-ahead and real-time markets. This would allow them to capture the impact of uncertainties in wind and load that are not captured in day-ahead only models. To enable this, MISO developed a new process with the assistance of Manitoba Hydro and Energy Exemplar, named Interleave, to capture the effect of the real-time response to changing forecasts. The Interleave simulation best represents the sequential nature of the day-ahead and real-time markets. After completion of a single day-ahead simulation, the unit commitment and other outputs are passed to the real-time simulation. After this simulation is completed, the ending conditions are then passed into the next day-ahead simulation. This continues for every day of the planning year, interleaving the days to create a realistic market simulation (p. 4). Most long-term production cost simulation tools replicate the hourly operations of the system (DA market) for the whole simulation timeframe without considering the interactions between the hourly simulation and sub-hourly simulation (if that function is available). In the real system, however, it is important for generation companies to evaluate the RT settlement results from the previous day in order to prepare the next DA bids. This need is especially critical for energy limited resources as the DA/RT dispatch difference will change the available energy for those units. The price and volume of bids and offers to the DA market is a function of the storage available, which is impacted by the prior day s RT dispatch. A DA-only or RT-only simulation model cannot adequately represent energy limited resources (p. 18). They also developed a new approach to model the natural storage of the hydro system and Manitoba Hydro s expected market participation: This was necessary to reflect the reality that traders adjust their bids and offers depending on what storage operations occurred as a result of previous day s day-ahead and real-time market activities. A value of water in storage (VWS) curve was introduced to take into account the opportunity cost of water of the entire planning period to allow for daily bids. Real-time bidding offers were calculated from the VWS curve along with offer bands representing the uncertainty presented between the day ahead and real time markets. New offers were determined after each simulated day (p. 4). While the value of water in storage is a conventional concept in hydroelectric generation, its implementation by MISO in PLEXOS was new. It seems possible that this concept could be used to provide improved modeling of other storage resources (such as batteries) in MISO s system. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 73

86 The modeling effort allowed a very realistic assessment of the impact of Manitoba Hydro storage on MISO production costs, at the expense of a model that took several days to run. In order to compare scenarios, MISO had to modify the production cost estimates to take into account differences in Manitoba Hydro storage levels at the end of different runs and differences between runs in imports and exports to neighboring areas (other than Manitoba). MISO also had to extrapolate from a single year of modeling to estimate total lifetime benefits. One outcome of the study was the determination that expanding use of the existing HVDC line between Minnesota and Manitoba (the so-called External Asynchronous Resource or EAR) from unidirectional to bidirectional, thus allowing Manitoba Hydro to submit price sensitive bids and offers into the MISO market, would result in $8.74 million in annual production cost savings for MISO, including $100,000 in reserve savings. That change was expected to take effect in Another outcome was the determination that adding a 500 kv line between Dorsey, Manitoba, and either Duluth, Minnesota or Fargo, North Dakota/Moorhead, Minnesota would provide a B/C ratio of 1.7 to 3.84 across all future scenarios analyzed, using the modified production cost metric. The benefit metrics included the benefits of incremental hydro generation. 66 The costs, however, included only the cost of the U.S. portion of the transmission line (not the Canadian line to the border nor the generating plant). It is standard MISO practice to include only MISO costs and MISO benefits in B/C analysis. Because the production cost model considers the impacts of the project on all aspects of system operation, it is not possible to isolate the effect on wind energy production per se. MISO assessed the impacts of the project on wind energy production in three ways: (1) by looking at wind curtailment reduction: (2) by looking at the correlation between imports from Manitoba Hydro and MISO wind generation; and (3) by looking at the improvements in utilization of both wind and hydro. Most of the wind energy production in MISO occurs in the North Region (MISO 2016c p. 58), and its use is sometimes dispatched down or curtailed due to transmission congestion. MISO s analyses of the impact on curtailment differed in Phases 3 and 4 of project modeling. In Phase 3, they assumed that wind would bid into the market at its variable cost. In Phase 4, they assumed that all wind plants would get the production tax credit and that their variable cost would be negative, specifically $-20/MWh. Thus in Phase 3, the curtailment reduction was estimated to be 30 to 100 GWh (22 to 55% decrease in curtailment) (p. 34). In Phase 4, the estimated Northern MISO wind curtailment reduction was only 1.3 to 4.5 GWh. 67 Although the executive summary reports reductions (50 to 100 GWh) more similar to those from Phase 3, the body of the report (p. 55) states that the Phase 4 results are considered more representative of likely reductions, 68 and the Phase 4 modeling results are used in the B/C analysis. Thus 66 Their Phase 3 analysis did not produce B/C ratios >1. In Phase 4 they changed their assumption about which of two proposed MH plants to include in the baseline and which to include as part of the change case, and they also changed their assumption about wind offer prices. The Phase 4 analysis found higher B/C ratios which were >1. 67 Most of the curtailment was due to congestion rather than low load. This curtailment is an outcome of the production cost minimization of the model, rather than a physical system limitation. When the offer price for wind is lowered, the model finds it more economic to manage congestion by cycling other resources rather than curtailing wind (J. Bakke, pers. comm. March 23, 2016). Thus in the Phase 4 modeling curtailment is already low, due to low wind offer prices, before the transmission line is added. 68 In fact the scenario with the highest curtailment reduction of ~100 GWh had a Central location for the transmission line, which was dropped in Phase 4 because its higher cost resulted in lower benefit-to-cost ratios. p. 74 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

87 it appears that the impact of the Manitoba Hydro Wind Synergy project on wind curtailments may be minimal. MISO s analysis showed that the electricity flow over the Manitoba Hydro to MISO interface would be inversely correlated with the wind generation in MISO: when the wind picks up and adds downward pressure on prices, Manitoba Hydro reduces its generation. Conversely, when the wind dies down, Manitoba Hydro increases its generation. Depending on the scenario, the correlation coefficient between hydro power flow over the interface and wind generation ranged from -0.1 to (p. 53) (This corresponds to an R 2 of 0.01 to 0.25). A visual example (Figure 26) shows both the large scale inverse correlation between interface flow and wind production and the short term response of interface flow to changes in wind production. Improvements in utilization of wind and hydro were shown by the reduction in production costs for MISO and the increase in revenue for Manitoba Hydro. In parallel with this study, Minnesota Power submitted three related filings to the Minnesota PUC. The first (Minnesota Power 2011, MN PUC 2012) was for approval of a 250 MW power purchase agreement (PPA) and associated Energy Exchange Agreement (EEA) with Manitoba Hydro. The second (Minnesota Power 2013, MN PUC 2015a) was for the transmission line from the Minnesota-Manitoba border to the Blackberry Substation in Itasca County. The third (Minnesota Power 2014, MN PUC 2015b) was for a 133 MW PPA and EEA. As Minnesota Power described it, the Energy Exchange Agreements offer the benefits of a very unique wind storage provision [that] facilitates timely shifts of energy resources between Minnesota Power and Manitoba Hydro, optimizing the generation of electricity from either wind or water resources to meet load The 250 MW EEA entitles Minnesota Power to effectively store 250,000 MWh per year of wind energy it produces with Manitoba Hydro by shifting delivery of those resources to Manitoba Hydro at appropriate times. The 133 MW EEA provides for an additional 750,000 MWh of wind storage. The Pumped Storage Provision or wind storage provision shifts delivery of Minnesota Power s wind energy to Manitoba Hydro when wind energy production is high relative to Minnesota Power s loads. When wind production is low or loads are high, Minnesota Power will schedule the reserved energy to be returned to its system. Each of the PPAs and EEAs is for a 20 year period beginning when the Great Northern Transmission line becomes operational. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 75

88 Figure 26. Synergy between MISO wind production and MH-MISO interface flow (Bakke et al. 2014, p. 35) This project is a different approach to the use of storage to facilitate integration of VG. A similar approach is used in Europe, where use of wind power from Denmark is coordinated with use of hydro power from Sweden and Norway (Ackermann et. Al. 2009). The cost of this project is low for MISO compared to other bulk storage systems that could be envisioned since the cost of the hydro plant and of the transmission line within Canada is not included. Although it serves a storage function, the project does not require construction of separate upper and lower reservoirs as would a pumped hydro system; instead, and more simply, it manages the flow from a dammed reservoir through hydroelectric generators. It is located comparatively close to the region within MISO that has the best wind resources and the most transmission congestion related to wind generation. 69 The project illustrates the level of effort required to quantify all benefits from a storage project, as well as the wealth of information available from such a modeling effort. Addition of Short-Term Stored Energy Resource MISO launched its ancillary service market in Also in 2009 MISO started working with stakeholders to incorporate short-term stored energy resources (SER) into the MISO market (Chen et al. 2011). This was required by the provisions of FERC Order No. 890 (FERC 2007) that ISOs ensure participation of nongeneration resources in markets on a fair and equitable basis, and it was consistent with the inclusion of Demand Response Resources (DRR) in MISO s ancillary service markets from the beginning (Chen et al. 2011). MISO submitted a proposed tariff revision to FERC to allow SERs to participate in MISO s 69 It potentially raises the same kinds of environmental and social issues as other Manitoba Hydro projects, but that is outside the scope of this discussion. p. 76 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

89 regulation reserve market with a starting date of January 1, This proposal was conditionally accepted by FERC in December 2009 (Chen et al. 2011). In their paper explaining how SERs were incorporated into the tariff (Chen et al. 2011), the authors also explain some of the benefits of SERs relative to conventional resources for regulation: The short-term SER has several unique characteristics that can benefit the energy and ancillary service markets. First of all, it is usually very fast-responsive and can provide significant value for regulation response in AGC. Pacific Northwest National Laboratory (PNNL) compared the performance of fastresponsive storage resources with conventional regulation resources like hydro, combustion turbines, steam turbines and combined-cycle units in the California ISO market... The conclusion is that the faster responsive resource can help to reduce California ISO s regulation procurement by up to 40% on average. The second important benefit of short-term SER is that it can help reduce CO2 emissions. KEMA reported on a study of Beacon Power s flywheel technology in PJM, California ISO and ISO New England. The conclusion is that the flywheel-based frequency regulation can be expected to produce significantly less CO2 emissions for all three regions. The references they mention are Makarov et al and KEMA The KEMA study found that frequency regulation increased fuel consumption of coal and gas-fired generation by about 0.5 to 1.5%, so switching to flywheel frequency regulation reduced CO 2. Impacts on SO 2 and NO x emissions varied in the three ISOs KEMA analyzed (which did not include MISO), depending on the source of the electricity used to charge the flywheel system. The devices for which MISO s Stored Energy Resource tariff category was created have discharge time at rated power (DTRP) of at least 5 minutes and typically less than one hour. The five minute minimum was necessitated by MISO s 5 minute dispatch interval in the real-time market. Products with short DTRPs cannot provide a meaningful amount of energy and hence are not allowed to bid into the energy market. Because of NERC s requirement that contingency reserves be able to operate for 90 minutes following a contingency event, SERs, as defined, cannot be used for spinning and non-spinning reserves either. Thus SERs are only allowed to make offers into MISO s market for regulating reserves. Unlike other types of resources, SER s cleared for regulation are not allowed to substitute for lower quality spinning and non-spinning reserves because of their shorter discharge capability. They are also not allowed to meet zonal regulation requirements, since zonal regulating reserves can be substituted for contingency reserves in the event of a contingency (Chen et al. 2011). As of 2011, the minimum SER offer was 1 MW per hour (EPRI 2011). The specific operating parameters that must be submitted with the resource offer are spelled out in MISO s Business Practice Manual and include things like ramp rates, minimum and maximum regulation limits, maximum charge and discharge rates, maximum storage, and storage loss rate (see EPRI 2011 for a summary current as of that time). Chen et al. (2011) describe the method developed to optimize real-time dispatch of SERs for regulation in MISO. The method allows the maximum amount of frequency regulation to be procured from SER. They also describe the method for incorporating SERs into day-ahead unit commitment. The addition of this resource type to the MISO tariff makes it possible for energy storage devices with short DTRPs, such as flywheels and certain battery systems, to participate in the market. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 77

90 Pay-for-Performance Regulation In February 2011, FERC issued notice of a proposed rulemaking (NOPR) that would require grid operators to implement pay-for-performance tariffs that would pay faster-ramping resources more for their frequency regulation service. According to EPRI (2011 p. 4-7): Commercial manufacturers of fast storage devices such as Beacon Power Corporation have requested this rulemaking. Beacon's flywheel systems react in seconds to a grid operator's control signal -- a response that is exponentially faster than conventional fossil fuel-based regulation resources, pay-for-performance tariffs would enable the company to earn increased revenue from any regulation services it provides in those markets. Such markets include the New York ISO, where Beacon is already operating a regulation facility that is expected to reach its full 20 megawatts of capacity in the second quarter of FERC issued Order 755, Frequency Regulation Compensation in the Organized Wholesale Power Markets, in October 2011, effective December 2011: Specifically, this Final Rule requires RTOs and ISOs to compensate frequency regulation resources based on the actual service provided, including a capacity payment that includes the marginal unit's opportunity costs and a payment for performance that reflects the quantity of frequency regulation service provided by a resource when the resource is accurately following the dispatch signal (FERC 2011). MISO filed proposed revisions to its tariff for regulation to FERC on April 30, 2012, and they were provisionally accepted December 17, 2012 (Murphy 2014), with a requirement that MISO provide a report 14 months after implementation. In that informational report to FERC (MISO 2014h), MISO describes their two part regulation offer process, including a Regulating Capacity Offer ($/MWh) and a Regulating Mileage Offer ($/MW). Regulation mileage is the absolute value of the up and down movement, in MW, of a resource in response to Automatic Generation Control ( AGC ) regulation deployment. The two offers are added together to produce one regulating total cost ($/MWh) used in the market clearing engines. The regulating mileage offer is adjusted by a deployment ratio that is updated once a month: 70 Regulating Total Cost ($/MWh) = (Regulating Capacity Offer ($/MWh) + Deployment Ratio*(Regulating Mileage Offer ($/MW))*12 MISO also implemented regulation performance accuracy measurement to compensate regulating resources based on the actual performance. Performance is measured at every 5-minute Dispatch Interval and every hour. A resource is paid or charged accordingly, based on the results of performance tests They note that [t]he new regulation compensation should encourage fast-response regulating resources to participate in the regulation market, eventually improving the overall regulating performance. The average ramp rates offered into the real-time market as of the Year One report are shown in Table 35. Clearly, SERs offered much higher ramp rates than the other resources. 71 Pumped storage also 70 The 12 in the formula probably refers to the 12 five-minute dispatch intervals in an hour. 71 MISO noted in the report that in year one, they only had one small Stored Energy Resource in the market and it [was] only offered and cleared occasionally for testing purposes. p. 78 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

91 offered a higher average ramp rate than any of the other resources. Coal (steam turbines) offered the lowest ramp rate. Table 35. Average bidirectional ramp rate offered into MISO by resource type (MISO 2014h) The informational report to FERC divided regulation resources into fast (ramp rate >= 10 MW/min), slow (ramp rate < 3 MW/min), and middle ramp rates. They found that the switch to pay-for-performance regulation did somewhat increase the amount of fast- and middle-ramping resources cleared in the market, relative to the year before implementation. 72 Their monitoring also found that, as would be expected, fast-ramping resources were better at following total hourly and five-minute setpoint changes than middle-ramp resources, and middle-ramp resources were better than slow-ramping resources. The new pay-for-performance regulation penalized resources when they did not pass five-minute or one-hour performance tests. The penalties cut into a resource s revenue from regulation. Steam turbines lost nearly half of their gross regulation revenue due to penalties, while demand response performed the best and lost only 13% of its regulation revenues (Table 36). MISO reported that the net regulation payment in 2013 was much less than the payment in 2012, even though the average regulation clear prices in 2013 were higher than in 2012 (see green line in Figure 27 from MISO 2014f it appears to have dropped from about $26M to $20M). System control performance also improved slightly, as measured by NERC standards. 72 From Table 35 it would appear that only SERs fell into their fast category, but since the one SER (see note above) was only offered and cleared occasionally for testing, yet statistics are reported for fast resources throughout, there must have been other individual units whose offered ramp rates were above 10, even though the average for each of the non-ser categories was below 10. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 79

92 Table 36. Average of monthly ratio between penalty and gross revenue before penalty for 2013 (MISO 2014h) Figure 27. Gross regulation revenue, performance penalties, and net revenue (MISO 2014f, p. 24) The addition of a SER resource category and pay-for-performance regulation undoubtedly improves the potential for short-term energy storage in MISO s market. However, this has to be viewed in the overall context of the size of MISO s regulation market and the MISO market s regulation prices. When MISO became the balancing authority and started its ancillary services market, the amount of regulation required within the MISO footprint dropped significantly. This is the outcome of the region moving to a centralized common footprint regulation target rather than several non-coordinated regulation targets (PJM 2015, p. 211). Moreover, ancillary services requirements were unchanged after the MISO South integration despite the far larger footprint (Potomac Economics 2015 p. 3). In the period before consolidation of balancing responsibility within MISO, total regulation requirements of the p. 80 The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota

93 multiple balancing areas within MISO averaged 1,559 MW. 73 After consolidation, they averaged 396 MW (MISO 2016f). In 2015, MISO monthly average regulation market clearing prices (MCPs) ranged from $5.71/MWh to $8.18/MWh in the day-ahead market and $5.63/MWh to $7.86/MWh in the real-time market (Figure 28) (MISO 2016c p. 15). Time did not permit comprehensive investigation of how these prices compare to other markets. The data that are readily available show that the weighted average clearing price for regulation in PJM in 2015 was $31.92 per effective MW (Monitoring Analytics 2016 p. 363), much higher than in MISO. The average day-ahead price in NYISO in 2014 was $12.87 (Potomac Economics 2014 p. 12), somewhat higher than in MISO. Time did not permit comparison of these prices to revenue requirements for short-term energy storage systems. Moreover, it is unclear to what extent zonal regulation requirements (for which SERs do not qualify) affect the average MCPs shown in Figure 28. Figure 28. Day-Ahead and Real Time Market Clearing Prices for Ancillary Services in 2015 (MISO 2016c) Ramp Capability Product MISO s ramp capability product was discussed in the Markets section. Any dispatchable resource, including storage, is eligible to provide ramp capability (MISO 2016b). 74 Current Work on Non-Transmission Alternatives Certainly much of the transmission built to facilitate renewable portfolio standards is unavoidable, simply because the best VG resources are not located close to load centers. However, given the large 73 Average from 4/1/2005 to 12/31/2008, adjusted for membership changes. 74 The Navid & Rosenwald (2013) white paper stated that SERs would not be qualified to provide ramp capability product because their discharge time is too short. The Context for Energy Storage to Facilitate Renewable Electricity in Minnesota p. 81