ERCOT’s summer peak demand forecast: new investment, generator profits, no blackouts

A version of this article is available on Greentech Media

By Eric Gimon

Wholesale electricity markets are in the news, as oversupplied markets drive prices down and force early retirements of coal and nuclear units.  Companies like FirstEnergy have petitioned the federal government for salvation through regulatory changes and new market rules to drive up prices and restore profit margins – but the picture is different in Texas.

The Electric Reliability Council of Texas’ (ERCOT) “energy-only” market model exposes the value of flexible resources without capacity markets, testing market design in a high-renewables future.

Texas’ market model is working so far: market forces are accelerating the transition from dirty, expensive plants to cleaner, cheaper resources including variable renewables, demand response, and batteries.  Avoiding capacity markets has saved ERCOT customers billions and the system has remained reliable.

But recent coal retirements and increased load forecasts are putting ERCOT’s energy-only market model to the test.  ERCOT’s reserve margin, a key measure of resource adequacy, is expected to be significantly below its target level this summer, prompting fears of electricity service disruptions.  Capacity market debate observers will be watching closely, as capacity subsidies, the most common alternative for ensuring reliability, have been criticized for slowing the economic transition away from uneconomic coal and nuclear while suppressing price signals for more flexible units that complement cheaper, cleaner energy resources.

To a naïve observer, the energy-only market structure’s test will be whether ERCOT can avoid shortfalls, i.e. a loss-of-load event, this summer.  But no level of investment or reserve margin can entirely eliminate all risk or protect the grid from ever falling short.  The true test of ERCOT’s market design is whether strong investment signals, i.e. higher prices, spur investment to drive the system back from acceptable risk to a more desirable level of risk.  Fortunately ERCOT looks capable of meeting this subtler test, and it should stay the course to avoid expensive capacity markets.

Will a disruption happen? Putting ERCOT’s planning reserve margin in context

A December 2017 planning report prompted concerns that an energy-only market might not ensure adequate reliability, pegging ERCOT’s summer 2018 expected planning reserve margin (PRM) at 9.3 percent.  ERCOT’s PRM has since increased to 11 percent, meaning expected generation fleet capacity exceeds expected summer peak load, minus emergency load management tools (including types of demand response) by 11 percent.

This is below ERCOT’s 13.75 percent minimum target reserve margin, which reflects ERCOT’s desired risk threshold in line with a “one event in ten years” resource adequacy standard.  However, this summer’s dip below the target PRM doesn’t mean Texas is taking unacceptable system-wide service disruption risks.

Because plenty of uncertainty exists about exactly what level of reserve margin corresponds to a given system risk level, the target PRM is not a magic number, shown by a recent Brattle and Astrapé study.  In 2014, the Public Utilities Commission of Texas (PUCT) asked them to estimate ERCOT’s economically-optimal PRM to inform their ongoing review of market design for resource adequacy.  The PUCT wanted to know if ERCOT’s energy-only market design could deliver their desired reliability level. Brattle’s top line results were eye-opening, with values for possible target reserve margins all over the place:

Source: Samuel Newell et al., Estimating the Economically Optimal Reserve Margin in ERCOT, The Brattle Group & Astrape Consulting, prepared for the Public Utilities Commission of Texas (2014)

The Economically-Optimal Reserve Margin (EORM) is the reserve margin emerging from a least-cost system, already below this summer’s forecasted 11 percent PRM.  According to Brattle’s modelling, with that PRM one can expect a “loss of load event” (LOLE), defined as a system deficit triggering rotating outages, happening slightly less than once every two years (0.44 annual probability of a LOLE – Table 13).  This is significantly worse than ERCOT’s desired reliability standard of one event in ten years (0.1 LOLE), which Brattle’s base case indicates a 14.1 percent PRM is required to achieve.

The report also includes a wide range of results (12.6 percent-16.1 percent) for a reliability standard of 0.1 LOLE in sensitivity cases.  This wide range reflects how sensitive an estimate of the desirable PRM is to model assumptions, especially assumptions about the frequency of extreme events.  The PRM cannot be a precise measure of reliability risk beyond an accuracy of a couple percentage points.

Whatever happens in Texas this summer, the probabilities in this study probably won’t correspond exactly to the level of reliability risk ERCOT is running at an 11 percent PRM.

Is the risk of system reliability as bad as it seems?

Should ERCOT be headed towards an undesirable level of reliability this summer, the natural follow-up question is: How close to the Sun will Texas’ grid fly?  In other words, does an 11 percent PRM equal at least an acceptable level of system risk if new investment will push it back up in coming years?

One fact should lead policymakers to conclude that Texas is still facing an acceptable level of system risk this summer: compared to metrics from other jurisdictions, 11 percent seems like an adequate PRM.

The one in ten years resource adequacy standard is a historical construct adopted by the electric power industry that grid operators can interpret differently.  For example, “one event in ten years” could be thought of as one day or 24 hours in ten years, i.e. 2.4 loss-of-load-hours (LOLH) per year.  According to figure ES-1, ERCOT only needs a 9.1 percent PRM to achieve this LOLH standard, so it would be compliant this summer.  International jurisdictions often use the 0.001 percent expected unserved energy (EUE) – the report lists 9.6 percent as the minimum PRM to achieve that metric.  Hence, plenty of evidence shows 11 percent could be considered an acceptable PRM.

Furthermore, reliability from all three of these metrics is more stringent than what customers experience due to distribution-related outages.  Suppose this summer ERCOT has an unusually high peak load and reserves dip too low.  After a progressive series of steps allowing system operators to add generation from other grids and enlisting large customers who voluntarily are paid to be curtailed during emergencies, ERCOT and the PUCT will ask the public to conserve electricity.  Once all avenues are played out, ERCOT can institute rotating outages to preserve the entire grid’s integrity – no system-wide blackout would occur.

Rotating outages have only happened three times in ERCOT history, yet even then customers actually experience relatively little disruption compared to what they are already used to because of problems on the distribution grid.

An earlier 2012 Brattle report explains that even at the lower 2.4 LOLH reliability standard, the possibility of rotating outages means customers can expect in a given year to be without power for “only three minutes per customer; this compares to an average of a few hundred minutes per customer per year from distribution outages.”  The slight possibility of more system-deficit issues is a blip compared to much more common distribution outages, and likely won’t meaningfully degrade customer service.

Finally, it is also quite possible that the risk of a reliability event this summer is smaller than Brattle’s 50 percent probability because of strong economic incentives in ERCOT’s market design.  Energy-only markets reward economic self-interest (greed) in a socially productive way when prices exactly spike at times of maximum system stress.  Much higher summer prices when reserves run short mean that price-responsive customers are more likely to reduce their load to take advantage of money-saving opportunities.   On the supply-side, combustion-turbines and old gas steam plants modeled as having a 19-20 percent chance of outage are already gearing up to run during peak demand.  The PRM itself has crept up from 9.3 to 11 percent over six months partly through market response to forecasted higher prices.

The real debate: how can energy-only markets adequately mitigate reliability risk?

So, even if a single loss-of-load event in ERCOT happens this summer, such an occurrence would not give particularly precise information on whether system risk was at 1-in-3 year or 1-in-10 year level.  While this summer’s planning reserve margin is one indicator of system reliability risk, understanding the exact connection requires detailed modelling and depends on key assumptions.  Because LOLEs happen against a background of much more frequent distribution outages, customers aren’t likely to detect any material changes to service reliability compared to prior years with higher PRMs.

Instead of focusing on PRM, regulators and interested observers should examine other metrics to evaluate ERCOT’s near-term health.  This summer, they should watch how frequently the grid is calling on all its available resources or facing higher demand than expected.  More or less frequent periods of stress correspond to a higher or lower risk of a loss-of-load event and are important data for modelers evaluating system reliability risk.  Regulators should look at actual resource performance during periods of stress to see if market participants are responding to the immediate performance incentives inherent in an energy-only market design.  Most importantly, regulators should pay attention to the economic signals for increased market participation that the low PRM is sending through forward and real-time markets.

By choosing an energy-only market, Texas regulators have accepted that prices will be too low some years to stimulate significant new investment, but that this market will also foster new investment as prices spike when resources become short.  The big questions become: as the PRM moves up and down over various business cycles, where does it average out?  And does this average correspond to an acceptable level of reliability?

The PRM resulting from market forces is called the market equilibrium reserve margin (typically a bit higher than the economically optimal reserve margin). This equilibrium matters to policymakers because if it is lower than the benchmark PRM they think is needed to reach their desired reliability level, than the energy-only market design will not achieve their reliability goals.  Policymakers must then make further tweaks to the energy-only markets to increase generator revenues during times of stress, like changes to the operating reserve demand curve or, as a last resort, drastic measures like a capacity market.

According to Brattle’s 2014 report, ERCOT needs a capacity market because its equilibrium is around 11.5 percent PRM, well short of its 13.75 percent target and the 14.1 percent PRM that achieves a 1-in-10 LOLE.  Brattle estimated long-term incremental costs of a capacity market to customers at $400 million/year, a one percent bill premium.  Even so, the PUCT continued trusting the market and Texas customers have benefited greatly given the high PRMs over the last few years.

Texas should let the market work as designed

Despite the PRM dropping below benchmark, facts on the ground argue Texas should stay the course.  Apart from the shortcomings and inefficiencies of a capacity market, the Brattle modelling may be underestimating the market response to a low PRM, meaning the energy-only market should provide financial support for an adequate level of risk.

Forward prices (ICE) are very high for this summer ($120/MWh and $220/MWh average 7am-10pm for July and August as of May 10th).  Based on such prices, combined-cycle gas (CCGT) plants in the market could easily make revenues net of short-term expenses more than 200-250 percent the level necessary to recover annualized long-term capital expenses – a significant incentive to invest in new gas resources.  Furthermore, the marginal new build may not be a CCGT or gas peaker plant, with associated build-time lag.  Resources like wind, batteries, solar, reciprocating engines, and demand response can – and will – come online a lot quicker in this pricing environment.

Instead of revisiting fixes like capacity markets, Texas policymakers should hang tight and give markets a chance to show their stuff while focusing on continual improvement to energy-market efficiency, all the while at an arguably acceptable risk.  For example, finding ways to improve the availability of price responsive demand (lowers costs and the need for reserves), or promoting policies to increase the finances of new clean resources by improving credit-worthiness of load counter-parties (like retailers) that use them to hedge exposure to soaring prices like those expected this summer.

By doubling down on its faith in markets, Texas can continue to be a great example of a market-driven transition to a cleaner, cheaper, and more reliable grid.


Thanks to Lenae Shirley and Rob Gramlich for their feedback on this piece.  The article and any errors therein are solely the responsibility of America’s Power Plan.