Readers of this publication are undoubtedly aware of a renewed interest in energy storage. Congress is considering the Storage Technology for Renewable and Green Energy Act of 20111, which would provide tax incentives. In the fall of 2010, California enacted AB 2514 2, which originally included a portfolio standard and was later changed to direct state and local governments to assess the need for a portfolio standard. AEP3, Xcel Energy4, Beacon Power5 and AES6 all have operating pilot or commercial-scale storage devices in operation. Storage has been prominently mentioned in connection with distributed generation, integrating renewables, and as an alternative for transmission7 and distribution upgrades. A report from EPRI8 cites numerous storage-related benefits, including capacity deferral, shifting energy from off-peak to on-peak periods, ancillary services and lower T&D losses.
However the high capital cost of storage technologies and a number of practical problems associated with storage implementation and operation are not quite as well documented. My purpose in writing this article is to point out some of the barriers and propose solutions. Rather than trying to address all possible storage applications, this article focuses on storage used to time-shift energy, provide ancillary services, provide peaking capacity, and act as long-term (multi-day) storage.
Perhaps the biggest obstacle to wider adoption of storage is its high capital cost. Pumped storage is probably the least expensive commercially available technology as well as one of the most efficient, but it has higher capital costs than gas-fired peaking plants and suitable locations are limited by geography and geology. Compressed Air Energy Storage is less site-specific and several conceptual designs have been proposed that could be even less costly than pumped storage, but these have not been proven at any scale. The diabatic9 process used at two existing sites employs gas-fired combustion turbines, costs about as much as pumped storage, and is relatively inefficient. Batteries and flywheels have no siting restrictions, but they currently cost significantly more to build than pumped storage, CAES or a gas-fired peaking plant. Lithium ion and sodium sulfur batteries are entering limited use in commercial applications, but while they are relatively efficient, they have very high capital costs. Flow batteries are less expensive to build, but they are also less efficient.
Many proponents are advocating for tax credits and other financial incentives to spur volume production in an effort to drive down costs. This is likely to waste enormous sums of money for several reasons. First, pumped storage costs won't be affected in any material way because most of their costs involve pouring concrete and moving dirt. Second, batteries and other, similar, technologies are still immature and unproven. Reductions in battery costs are likely to require advances in chemistry and materials science that increase the capacity and reduce the cost of existing chemical processes10. It makes more sense to invest money in R&D informed by technology roadmaps that identify obstacles to improving performance and reducing costs, with subsequent rounds of funding tied to cost reduction and performance improvement milestones. Particularly for the time-shifting and long-term storage applications, cost targets of $100/kWh or less have been mentioned in the literature11. For ancillary services applications, where one hour of storage capacity is sufficient, an overnight cost target of $300/kW12 would allow storage to compete with alternatives.
Markets and Prices
Even if storage intended for time-shifting, long-term storage or providing ancillary services could be built for free, it can't be operated cost-effectively in any of the organized wholesale (RTO and ISO) markets in the US. Organized markets employ incremental cost dispatch, which requires the operator of a generating plant to provide the grid or market operator with a price curve that typically reflects the generator's incremental running costs13. However a storage device's incremental costs depend on market prices that are unknown at the time bids (to buy energy) must be submitted to the market operator. Similarly, the prices a storage device would receive for energy it sells to the grid are unknown at the time offers (to sell) must be submitted. Consequently, a storage operator lacks the information it needs to formulate and submit price/quantity offers into the single clearing price auctions that form the basis for all organized electricity markets in the US. Instead, it must estimate next day prices and then blindly submit price and quantity offers without knowing whether their results from the auction are operationally and economically feasible14. As Xcel Energy discovered in the course of its wind-to-battery pilot, estimates are problematic because even small errors have a disproportionate impact on operating profits15.
The grid operator could accommodate storage devices by including them in its market optimization16 and producing an "optimal" charging and discharging schedule. PJM apparently does this for the pumped storage projects in its footprint. Since including storage in the optimization incurs a significant cost in terms of computational effort17, it is done only for the day-ahead time frame, and only for the largest storage projects18. None of the existing organized markets provides storage operators with a way to make beneficial adjustments to their day-ahead schedules in response to changes in grid conditions once the day-ahead market closes. The practical implication is that storage built for time-shifting or long-term storage will never be able to produce forecast levels of operating profits, because those forecasts implicitly assume the storage operator has perfect knowledge about market prices.
From the standpoint of a storage operator, the ideal market structure is one in which periodic or perhaps even continuous trading is allowed for each of the next 24 hours, and possibly longer for devices with multi-day storage capability. The grid operator's centralized market optimization software would be replaced by a system of decentralized decision-making and complex offer curves would be replaced with single-part bids and offers. Tradable commodities would include hourly energy, point-to-point or point-to-hub transportation (transmission), and options on energy (ancillary services). In 1996, the author and Dr. Edward Cazalet proposed this type of trading arrangement, which uses the same basic principles employed by commodity and equity markets around the world, as an alternative to the market ultimately adopted by California19.
Current prices for market services also act as an impediment to storage adoption. The amount of revenue a storage device can earn from selling capacity, ancillary (spinning reserves, non-spinning reserves and regulation) and time-shifting or long-term storage services is unlikely to be adequate to support most storage technologies at current market prices. With respect to capacity payments, prices for capacity in the centralized capacity markets20 have tended to clear well below the cost of new entry for a combustion turbine plant -- the highest price in PJM seems to be around $80/kW/year21 in a load pocket and the system-wide clearing prices is $45/kW/year or less22.
Although storage built for time-shifting and long-term storage can provide and receive standby payments to provide ancillary services, it can only claim those payments for hours in which ancillary services are less valuable because it is typically operating in either the charging or discharging mode in those hours when ancillary services are most valuable. Consequently, any added revenues from supplying ancillary services when the storage device is not otherwise engaged are likely to be relatively small unless the storage device is specifically built and operated only to provide ancillary services. However, even in this special case, ancillary services revenues are currently relatively meager and are likely to remain that way due to competition from flexible demand23 and fossil-fired generation. Beacon Power's recent bankruptcy filing was prompted in part by falling prices for frequency regulation in the New York ISO market that rendered its new Stephentown flywheel plant uneconomic24. While FERC's recent Order 755 changes the compensation mechanism for frequency regulation in ways that favor fast-acting resources like storage, increased competition is likely to maintain downward pressure on ancillary services prices even as demand increases. In California, for example, the peak need for frequency regulation in 2020 of around 1,100 MW can easily be met by more than 22,000 MW of regulation-capable fossil-fired, hydro and pumped storage capacity25.
Subsidies, Accountability and Performance
At least one analysis has justified subsidies for storage at levels that would make electricity consumers indifferent26, typically in the form of lump sum payments or tax credits once the project enters commercial operation. One of the drawbacks of this approach is that project developers and owners who are the likely recipients of any incentives can earn back their invested capital fairly quickly once the project is operational, while consumers are at risk for both the incentive payments and benefits that may not materialize in accordance with the developer's projections.
In fact, subsidies for storage may be unusually risky from the standpoint of the consumers who will be asked to fund it. Incentives tied to production create a perverse incentive to maximize production even if doing so produces operating losses (the efficiency -adjusted cost of off-peak charging is greater than the value of displaced on-peak generation). Incentives that are not tied to production may allow the developer to meet its financial return targets without ever actually running the storage device. To some extent, these same perverse incentives exist with subsidies for conventional generation, but they are rarely discussed.
Subsidizing storage also means too much of it is likely to be built, which will further weaken the economic rationale for storage and other resource options. In organized markets, subsidies for preferred resources depress market revenues for unsubsidized resources that may also be required to remain in the market for operational reasons, thereby triggering round after round of corrective subsidies to keep everyone whole.
Policymakers and regulators need to carefully consider the cost, performance and operational aspects of storage before they implement aggressive policies aimed at increasing deployment. Generally speaking, storage is still too expensive compared with available alternatives for time-shifting energy and for providing capacity and ancillary services. It is also a much more expensive alternative than fossil-fired generation for meeting demand during periods when renewable energy production simply isn't available. As currently structured, organized markets in the US don't provide the information and the trading opportunities storage operators to maximize their operating profits, and consequently the operating benefits of storage are unlikely to be realized.
Rather than providing direct subsidies to build storage, policymakers should work with manufacturers to develop technology roadmaps that focus on cost and performance improvements, and then fund research and development efforts with clearly defined cost and performance targets.
- Storage Technology for Renewable and Green Energy Act of 2011 (www.opencongress.org/bill/112-s1845/show)
- AB 2514 (http://www.carebs.org/CA_ab_2514_bill_20100929_chaptered.pdf)
- Lawrence Berkeley Labs cites an ARPA-E cost target of $100/kWh. See http://bestar.lbl.gov/arpa-e/files/ITRI-Talk-Final.pdf, slide 3.
- Current prices for frequency regulation
- Or in the case of flexible demand, decremental costs.
- For example, they could end up "winning" all of their trades to sell energy and none of their trades to buy, in which case they must buy energy in the real-time market at uncertain prices that might result in operating losses.
- Add reference to report.
- Market optimization software performs much the same role as a vertically integrated utility's economic dispatch algorithm. It uses bids and offers from buyers and sellers to matches supply and demand at least cost while observing a number of system constraints.
- http://elibrary.ferc.gov/idmws/common/OpenNat.asp?fileID=12702070 See slides 17-22.
- Since prices in all hours dictate storage operation and since storage operation influences prices, several iterations through the market software may be required to converge on an optimal solution.
- http://www.cazalet.com/images/Simplified_Bidding_for_WEPEX.pdf In addition to facilitating storage operation, the approach outlined in our paper has a number of advantages for market operators, other resource owners, retailers, market monitors and regulators.
- Centralized capacity markets are operated by PJM, the New York ISO and ISO New England.
- PJM reports capacity prices in $/MW/day. As reported, the price was $225/MW/day.
- For example, an aggregation of commercial buildings with HVAC plants powered by variable frequency drives could easily provide frequency regulation at a lower cost than storage.
- Conversation with Donald Tretheway at the California ISO.
- See California's Statewide Joint IOU Study on Permanent Load Shifting at http://ethree.com/public_projects/sce1.php