Capacity planning for the smart grid.
James F. Wilson is an economist and principal of Wilson Energy Economics, and also is an affiliate of LECG, LLC. Email him at firstname.lastname@example.org. This article expresses the author’s views and not necessarily those of any client.
The one-day-in-10-years criterion for capacity planning is coming under scrutiny. Making the most of the smart grid and demand management requires a less conservative approach. Markets and prices rather than administrative rules will ensure resource adequacy in a more efficient way.
Resource adequacy planning in the United States has rested upon a very conservative criterion (“one day in 10 years;” perhaps an order of magnitude more conservative than can be justified by marginal benefits) that also has been applied conservatively, as explained in Part I of this article (see April 2010 Fortnightly). Why has electric resource planning been so conservative? If there was a rationale for this practice in the past, is the rationale still applicable today, and for the future? How must resource adequacy practices be adapted for the coming smart grid?
Why So Conservative?
The 1-in-10 criterion, and conservative approaches to its application, apparently became widely accepted decades ago when electricity demand in the United States grew at a fairly steady rate and continuously required generating capacity additions. Under such circumstances, if utility resource planning was conservative and targeted high reserve margins, the excess capacity was never excess for long. In addition, the power plants built to meet incremental needs in the past required years to build, and with rapidly growing demand, the risk and potential cost of not beginning construction in time was substantial. Under these circumstances, the costs and risks favored targeting large reserve margins and building needed capacity well in advance.
Resource adequacy might have been, and remains, very conservative for additional, very human, reasons. A much higher frequency of distribution system outages has been tolerated by planners and regulators, perhaps because of the inevitability of the acts of nature or component failures that are the proximate causes of such outages. While the frequency of distribution system outages can be reduced through more aggressive vegetation management and other practices, these outages generally aren’t under the control of utility planners or regulatory authorities. In contrast, outages due to inadequate resources seem more preventable—a few more megawatts would have reduced or eliminated the need for firm curtailment—and thus suggest a failure by the utility or RTO to cause enough capacity to be built in a timely manner, or by regulatory authorities to issue permits or approve cost-recovery mechanisms for new construction.
Providing a very high level of resource adequacy also might reflect greater concern on the part of utility planners over reliability, for which they are responsible, than its cost, which is passed on to consumers. Having abundant resources arranged well in advance also makes both planning and operation of the system easier.
The historical conditions that contributed to acceptance of highly conservative approaches to resource planning are changing. Growth in electricity demand has become more variable and generally slowed, both in absolute terms and relative to economic growth (see Figure 1). At present, peak demand growth and the need for capacity additions in the near term are more uncertain due to the downturn in the economy that began in 2008, rising electricity prices over the past several years, and recent accelerated efforts to achieve greater energy efficiency. State and federal energy policies have encouraged energy efficiency and demand response for many years, and programs have been strengthened in recent years, with many states implementing enabling legislation and setting specific targets for reductions in energy use. Increasing concerns about climate change and other environmental impacts of energy production have heightened interest in efficiency and renewable sources of energy.
These changes are contributing to slower growth in peak loads. On the PJM system, for example, weather-normalized peak load was nearly flat from 2005 to 2008, before declining sharply in 2009 due to the recession.1 The risk that capacity reserves built at this time might be unneeded, and might remain unneeded for years, is much higher than at any time over the past decades.
In addition, there’s greater flexibility today in capacity planning, due to the availability of incremental capacity resources with shorter lead times than the large fossil fuel plants that were the primary means for meeting incremental capacity needs in the past; these resources include demand response, incremental upgrades to increase the capacity or extend the life of existing plants, and deferred retirements, among other sources. For example, a summary of the incremental capacity additions from the first six PJM reliability pricing model (RPM) capacity auctions shows that short lead-time resources accounted for 82 percent of the 27,641 MW of incremental resources in these auctions (see Figure 2). The combination of declining peak load growth and the availability of short lead-time resources reduces the need for, and value of, building large reserve margins and of acquiring them well in advance.
The Coming Smart Grid
Part I of this article showed that traditional approaches and attitudes toward resource adequacy planning have been extremely conservative, and an adjustment to these approaches would have been justified and in the consumers’ interest. However, the changes happening today, looking forward to the second decade of the 21st century, provide additional reasons to reconsider the industry’s long-standing approaches to resource adequacy.
Very broad-based efforts are underway and accelerating to implement the complex of substantial changes to the electricity industry included under the heading of smart grid. The smart grid will involve upgraded control and communications as well as advanced metering infrastructure (AMI), which will allow real-time communications to and from customers and their smart end-use devices. Many pilot studies have been completed successfully, and utilities across the nation are proposing to install millions of advanced meters in the next few years. While some of the elements of the smart grid are years away, many are being implemented now, and some forecasts anticipate enormous investments in the smart grid over the next several years.
These changes will accommodate substantial increases in demand response and price-responsive demand, as documented in a recent FERC staff report,2 leading to future peak loads becoming much more manageable and price-responsive. While large industrial and commercial consumers have been compensated for standing ready to provide peak load reductions for a long time, smaller commercial and residential consumers also will be offered advanced metering and pricing programs that provide incentives to reduce or delay consumption when electricity is relatively scarce or expensive. Residential consumers will be able to control how their air conditioners, refrigerators, washers, dryers and other appliances respond to such signals, reflecting their individual preferences for comfort, convenience and savings. Longer term, in addition to price-driven reductions in consumption, energy storage devices, on-site generation, and electric vehicles increasingly will sell power back to the grid in peak hours, reducing capacity needs.
The strong push for advanced meters, demand response and price-responsive demand reflects the expectation that substantial benefits will result from these investments. The primary benefit of demand response and price-responsive demand is the reduction in peak load and peak capacity requirements; fewer generating plants and transmission lines will be needed.
Realizing the enormous potential for price-responsive electricity demand will require changes to wholesale and retail pricing approaches that encourage and reward reductions in consumption at peak times. Because reducing peak loads can obviate the need to build additional power plants, large incentives to reduce demand or sell back power at such times are economically justified. Wholesale “scarcity pricing” regimes are being implemented to ensure that prices for energy and short-term reserves rise to high levels if there is insufficient price-induced supply and demand response to maintain reserves close to target levels. More and more customers will be offered real-time pricing, critical peak pricing or critical peak rebate programs3 that include substantial inducements during peak periods to reduce or delay electricity use. These changes will occur in conjunction with consumer protections at the retail level (i.e., through critical peak rebate approaches, or bill protection provisions) and wholesale level (i.e., addressing the potential for exercise of market power in peak or near-peak periods).
Pilot programs have shown that these innovations can lead to substantial reductions in peak loads. The FERC staff report cited earlier estimates potential 2019 reductions in peak load ranging from 6 percent to 13 percent in various parts of the country even under “expanded business-as-usual” assumptions, with much higher reductions under “achievable participation” assumptions.4
Creating Price-Responsive Demand
In recent years, demand-side reductions (i.e., primarily direct-controlled load reductions by industrial and large commercial customers) have been treated as capacity resources for capacity planning purposes in some regions. Load reductions resulting from energy efficiency measures also can be treated as capacity resources in some regions.5 However, treating demand-side reductions as capacity resources requires identification of baseline consumption levels, measurement and verification procedures, and other administrative complexities. As the quantities grow, concerns may increase about the performance and reliability of these reductions. Peak load forecasting also becomes problematic, as load reductions treated as capacity resources (but not other reductions or efficiency improvements) in principle should be included in peak load forecasts to avoid double-counting reductions.
Longer term, the objective should be to de-emphasize or phase out the practice of treating demand reductions as capacity resources, which is unique to the electricity industry; ultimately, consumers should pay based on what and when they consume, as they do in other industries (price-responsive demand has been called “third-generation” demand response, with direct load control and RTO demand-response programs constituting the first and second generations, respectively6). However, this will require further development of wholesale and retail pricing approaches; in the meanwhile, treating demand response as a resource provides a strong incentive for further development of this valuable capability.
Transition to the Smart Grid
The next several years likely will be a transitional period during which price-responsive demand will be developing at a pace that will vary considerably from region to region, and will be difficult to predict in advance. During this transitional period, resource adequacy approaches should be adapted to fully anticipate, accommodate, encourage, and reward peak-reducing technologies and practices in a manner that does not jeopardize reliability. There’s a significant risk that if traditional, highly conservative resource adequacy policies are continued, they will lead to over-procurement and excess capacity during this period, which could discourage and delay the development of the smart grid and price-responsive demand.
Peak load forecasts for capacity planning should fully anticipate the potential for peak load reductions, including the reductions induced by the potential for prices to rise to very high levels (i.e., approaching the value of lost load, or VOLL) when operating reserves fall below target levels. A portion of the reductions may be treated as resources and recognized in supply plans, with the remainder recognized in peak load forecasts. The challenge during the transitional period is to accurately forecast, without double-counting, all reasonably anticipated future peak load reductions and energy efficiency improvements. Unfortunately, future load reductions are highly uncertain, even looking out just a few years. And forecasting approaches that primarily extrapolate past trends to predict future peak load levels generally won’t accurately project new and accelerating trends, such as efficiency improvements and increasingly price-responsive demand, except with a substantial lag.
While utilities and RTOs will play a large role in implementing the smart grid, they understandably will be cautious about anticipating, in their long-term peak load forecasts and capacity planning, significant impacts of unproven targets and programs regarding efficiency, demand response and price-responsive demand. RTOs especially will have little control over these developments and won’t be able to ensure that programs intended to realize peak load reductions will be implemented in a timely and effective manner. Therefore, there is a risk that peak load forecasting and capacity planning will continue to be conservative and to reflect a believe-it-when-I-see-it approach to much of the potential for peak load reductions.
Because the pace with which the various smart-grid elements will be put in place will vary from area to area and is largely under the control of states, utilities, and load-serving entities (LSEs), these entities should play an active role in forecasting the impacts of these programs and determining the capacity requirements they displace in each area.Where RTOs prepare peak load forecasts for the entire RTO region, affording a greater role in forecasting future peak loads (and load reductions) to states, utilities and LSEs potentially could reduce the risk that these forecasts will fail to reasonably anticipate future peak load reductions. Should emergency purchases or firm curtailments be required due to a shortage of capacity, the costs and curtailments should be allocated to those entities that are short on capacity in proportion to their shortages, to the extent this is feasible.
Because adequacy criteria focus on curtailment of firm load, reserve margins and capacity requirements should be based upon the peak load levels expected to result under the most extreme scarcity pricing conditions, with prices and incentives at the highest possible levels and leading to maximum load reductions.
At present, reserve margins and capacity requirements are driven by forecasts of extreme peak load conditions expected to occur very rarely. As peak loads become more manageable, capacity requirements will decline and be targeted to net load levels likely to occur with much greater frequency, due to peak shaving and also the shifting of load reductions to adjacent hours. If loads on the highest load days can be managed (as needed) down to the load levels that occur many more days each year (see Figure 3), much less capacity will be needed to satisfy the one-in-10 criterion, and the least-used capacity will be called upon to provide energy or ancillary services in many more hours.
As noted above, increases in peak load and capacity requirements are presently much more uncertain than they were in the past, due to uncertainties about economic growth, higher electricity prices, increasing efficiencies, and other demand-side developments. These circumstances suggest that maintaining flexibility in capacity procurement is especially valuable at this time and can reduce the risk and potential cost of procuring excess capacity, or paying excessive prices for capacity, years in advance.
The risk of procuring excessive amounts of capacity is greatest where capacity requirements are determined and committed to years ahead, for instance, where RTOs require fulfilling capacity obligations three years in advance (e.g., the ISO-NE and PJM RTOs). However, as noted above, many of the incremental resources available at this time have fairly short lead-times. When substantial short lead-time resources are available, three-year advance mandatory procurement isn’t necessary to ensure adequate lead time to acquire needed resources for reliability, and it also may exclude or impose risks on shorter lead-time resources, who may not be prepared to offer their capacity so far in advance. Nor does mandatory forward procurement contribute in a significant way to the ability to attract major new power plants; such facilities require long-term contracts or other long-term revenue assurance.
The risk of excess procurement associated with forward procurement obligations can be reduced by providing additional flexibility in fulfilling the forward purchase obligations. A small step in this direction is accomplished by reducing the fraction of anticipated requirements that must be purchased the full three years in advance to less than 100 percent, in recognition of the availability of short lead-time resources, and also the uncertainty of the peak load forecast. PJM has implemented this change to its RPM mechanism. Additional flexibility for market participants to shift purchases and sales between the years-forward auctions and those closer to each delivery year, subject to limits to protect against market power, would contribute to market efficiency and reduce the risk of excess procurement and excessive capacity prices in forward markets. For example, allowing virtual capacity offers in the forward markets would provide such flexibility.
Adapting capacity procurement rules to afford greater flexibility will reduce the risk of procuring unneeded capacity that would preempt short lead-time resources and peak load reductions, and which may be more desirable and cost-effective ways to balance supply and demand.
If utilities and RTOs continue existing, conservative approaches to forecasting and planning for future capacity needs, they will maintain adequate capacity, but they also likely will undermine the actual need for, and value of, smart-grid enhancements and peak-reducing capability. If reserves continue to be planned for the “dumb peak” (i.e., reflecting only contractually committed demand-response providers), there likely will be excess capacity and infrequent instances of low reserves and high prices, and, therefore, only weak price incentives for electricity consumers to invest in, and deploy, smart appliances and other peak-reducing technologies that realize the majority of their value at such times. Consumers could end up bearing both the cost of the excess capacity, and also the cost of the advanced meters and smart devices that could be of little value during the excess capacity conditions. In addition, the anticipated excess capacity, by reducing the need for, and value of, price response and energy efficiency, serves as a disincentive to achieving the targets set for their development.
Resource Adequacy on the Smart Grid
Longer term, when a substantial fraction of peak load has become price-sensitive and manageable, traditional resource adequacy planning approaches based on adequacy criteria (i.e., such as “one-in-10”) will become both technically problematic and also unnecessary.
The common assumption that peak load is independent of supply availability no longer will hold, because prices increasingly will link peak demands (and especially the extreme peaks traditionally associated with loss-of-load risk) to supply conditions. When prices and incentives can reach high levels, possibly approaching VOLL, to call forth the maximum price response, modeling the circumstances under which firm curtailment could occur (i.e., the basis of LOLE studies to determine one-in-10 reserve margins) becomes both more difficult and less meaningful. It becomes more difficult to accurately model firm curtailment circumstances because actual peak loads become highly dependent upon system conditions through the price link. And it becomes less meaningful to distinguish the circumstances under which firm curtailment might occur because, by definition, the average firm customer exposed to real-time prices is close to indifferent between paying prices close to VOLL or being curtailed.
To the degree price-responsive demand reduces peak loads and shifts load to adjacent hours, there will be many more hours with load near the maximum levels; as a result, the highest-cost peaking capacity that’s needed primarily for reliability will be called upon much more frequently. To accomplish these load reductions, prices will be somewhat elevated, and the marginal peaking generation should earn much more than under the present circumstances, when it runs so infrequently (“one-day- in-10 years”).
Under these circumstances, the symptom of inadequate capacity no longer will be unacceptably frequent involuntary firm curtailments, as assumed under the traditional approaches to calculating reserve requirements based on LOLE criteria. Instead, the symptom of inadequate capacity will be too many hours with high prices and substantial voluntary, price-induced customer reductions, imposing a cost on these customers that exceeds the incremental cost of additional peaking capacity.
Part I of this article explained that the level of capacity is optimal and efficient when the incremental cost of additional capacity equals its incremental benefit; and historically, the benefit had to do with reducing the risk of having to curtail firm customers. On a system with substantial price-responsive demand, the incremental benefit of capacity results from the additional service provided to price-responsive customers, as the need to involuntarily curtail firm loads due to inadequate resources becomes increasingly unlikely. Put differently, supply additions will result from the interplay of supply and demand, as they do for other goods and services, rather than administrative planning criteria and reliability rules.
Ultimately, the conservative one-in-10 resource adequacy criterion, to the extent it will remain possible to meaningfully apply it, will suggest levels of capacity lower than the optimal amounts, and less than the amounts that will be provided under market incentives. As markets and prices rather than reliability rules begin to determine capacity levels, the RTOs’ capacity mechanisms can be phased out, as such payments no longer will be needed to achieve acceptable levels of reliability (see illustrative example in sidebar). The shift away from revenue recovery through capacity payments also will help achieve goals for attracting renewable resources, which capacity values typically are deeply discounted.
The future holds more instances with (at least a chance of) high prices and low reserves.
In the past and at present, with limited ability to manage peak loads, providing a high level of reliability requires planning a capacity margin over extreme peak load levels that are unlikely to occur. As a result, there nearly always is abundant capacity and operating reserves rarely fall below desired levels. As peak loads become more manageable and price-driven, generating capacity should be planned assuming the capability actually will be used. Even during the transitional period, the planning outlook should include more frequent short periods of low reserves and rising prices—or at least the potential for these conditions, unless there is sufficient demand and supply response. If there is sufficient response, price spikes won’t occur, or will be brief and muted. And activation of demand-side reductions should not be considered indicative of a failure to plan and build adequate resources. These actions become part of the plan, to be expected, up to a point. In addition, instances of high prices and critical peak pricing and rebates—when consumers realize the value of their investments in smart meters and devices—will have a positive impact on sales and deployment of such devices, ultimately strengthening the ability of the smart grid to ensure resource adequacy.