SUMMARY: Economist and veteran of the Northwest power scene, Robert McCullough gives his view of the root-cause analysis of the rolling blackouts California invoked during two days in August, and presents some solutions.

It is difficult to read the Oct. 6 report "Preliminary Root Cause Analysis: Mid-August 2020 Heat Storm" from the California ISO, California PUC and California Energy Commission without a sense of humor. The report blames the innocent (global warming) and rewards the uninvolved (Gov. Gavin Newsom).

However, it does shed light on why the best funded balancing authority on the West Coast was the only organization that required rolling blackouts during hot weather on Aug. 14 and 15.

A good place to start is with the facts. What went wrong?

The August rolling blackouts were policy decisions, not operational ones. A combination of poor reliability planning, high loads, high resource outages, and computer programming issues led CAISO to initiate rolling blackouts.

Overall, CAISO reports a reserve margin of 15 percent, within a state with a 20.9 percent reserve margin and a region with a 22.7 percent reserve margin. This is, by historical operating standards, an excellent reserve margin for facing a hot summer day. The reserve margin required for the state by NERC was only 13.7 percent.

Planned and forced outages at CAISO were a serious challenge. The loss from outages for Aug. 14 was 5,424 MW and for Aug. 15 was 6,490 MW. The grid operator predicted that their August capacity was 46,903 MW. This would correspond to an outage rate of 11.6 percent on Aug. 14, increasing to 13.8 percent the next day. These are very high loss rates, as the NERC database of outage rates shows.

The Aug. 15 outage levels are unusually high for a large diversified system.

On the other hand, CAISO's forecasted peak loads on the rolling blackout days were relatively close to actuals.

The forecasts for the days with rolling blackouts were good. However, their forecasted peak load was 45,907 MW. This is based on an unusual "1 in 2" peak forecasting methodology. This, as is addressed in the preliminary report, is not so good.

A traditional utility planner would not view this as acceptable. CAISO adds in expected imports in order to fix the problem; the utility term for this is "leaning on the neighbors." There are a number of reasons why assuming that imports will be available during hot weather isn't a common practice in utility reliability planning:

  • The assumption that warm conditions would not affect neighboring systems is incorrect.
  • Transmission outages—especially those interties that pass through mountainous and forested terrain—are subject to significant outages. The massive intertie north to the Pacific Northwest is specifically a concern given increasing wildfire risks.
  • The economics of reliability planning that are not based on firm contracts is doubtful at best.
  • Although the ISO performs a massive modelling effort with thousands of runs, the accuracy of the model is only as good as its assumptions. The assumptions concerning Pacific Northwest hydro operations are inadequate.

Simply put, CAISO's existing reliability planning methodology would not be acceptable anywhere outside of California.

The Preliminary Root Cause Analysis admits as much on page 43:

"The climate change-induced extreme heat storm across the western United States resulted in the demand for electricity exceeding the existing electricity resource planning targets. The existing resource planning processes are not designed to fully address an extreme heat storm like the one experienced in mid-August."

Translated into normal English, this paragraph effectively says that leaning on the neighbors does not work if the weather is bad. A better paragraph might have said "leaning on neighboring systems without firm capacity and transmission contracts is expensive and undependable."

Page 43 goes on to address two other "root causes":

"In transitioning to a reliable, clean, and affordable resource mix, resource planning targets have not kept pace to lead to sufficient resources that can be relied upon to meet demand in the early evening hours."

True. Every utility in North America is addressing the challenge of integrating low cost intermittent renewable resources into their systems. On Aug. 14 and 15, only one system failed to do so. Any number of studies have indicated that CAISO needs more energy storage and/or quick start gas turbines. This paragraph might well have been rewritten as "global warming snuck up on us." So, indeed, it did.

The most interesting paragraph on page 43 states:

"Some practices in the day-ahead energy market exacerbated the supply challenges under highly stressed conditions."

A recommended rewrite of this paragraph might well be "we have some very embarrassing problems to report."

The report indicates two specific problems and may be referencing, depending on a careful reading of the tortuous text, an even more significant problem of market manipulation.

First, a central problem in the summer of 2000 was a practice by the three investor-owned utilities to underbid their loads. By submitting bids less than actual loads, this created an apparent surplus in the day-ahead market and produced lower day-ahead prices. The shortfall would then need to be addressed in the real-time market. Surprisingly, this practice appears to still be present in the ISO's California marketplace.

It is very difficult to adjust generation schedules—especially for large shortfalls—unless a fleet of quick start turbines are available or substantial unloaded hydroelectric capacity is available. The right answer does not require new resources, however. The right answer is to identify the underbidding behavior and solve it prospectively.

Second, it appears that a computer programming error at the ISO treated financial transactions as being physical transactions.

Physical transactions involve a commitment to deliver an actual commodity at a specific time and place. Ordering a pair of shoes over the internet is a physical transaction. The buyer can depend on actually wearing the shoes when they are delivered.

Alternatively, a financial transaction would allow you to gamble on the price of shoes falling or rising in the near future. If successful, the counterparty would send you a payment for the difference in the price offered today and the price that was actually charged later. You cannot wear these shoes. If you are planning on wearing financial shoes, you may well be barefoot later.

The market in question is called the convergence market. It allows "fixed for floating" transactions where the price in the day-ahead market is compared with the price in the real-time market on the following day. The result is either a windfall or a bill. No actual energy is involved in such a transaction. Like the financial shoes above, the financial transaction will not drive air conditioners or illuminate homes.

A very opaque portion of the preliminary report can be found on pages 101 through 107. The reader can judge for themselves whether this passage communicates the problem adequately:

"CAISO had previously applied the price inconsistency market enhancements (PIMEs) to the residual unit commitment (RUC) as a matter of applying PIME to all its markets. The PIME in the other markets is necessary because it is necessary to have consistency between energy schedules and prices. The lack of energy schedules in RUC obviates the need for PIME in the RUC process. As a result, starting from the day-ahead market for Sept. 5, the CAISO stopped applying the PIME functionality to RUC process, which enabled it to use the scheduling run results for RUC schedules and awards instead of the pricing run results."

This translates to "we applied an energy fix to the capacity calculations which overstated available capacity in the real time operations." It appears that the ISO created a computer programming adjustment so arcane that only when it contributed to rolling blackouts was the error corrected.

The solution appears to have worked. Although the peak loads on Aug. 14 and 15 were high, even higher loads on Aug. 18 and Sept. 6 did not lead to rolling blackouts.

The computer problem might also have contributed to market manipulation.

Similar to under-scheduled load, during conditions in which physical supply is scarce, cleared virtual supply can mask physical supply shortages and allow more demand including low-priority exports to clear than what can be physically supported.

Market participants can profit—enormously—from artificial shortages. This was the raison d'etre for a number of the trading schemes that crashed the California markets in 2000 and 2001. Convergence bidding—especially with the aforementioned computer problem—would provide an incentive to place unrealistic bids. If the bids helped create a shortage, the day-ahead offer would be paid at high real time prices and the trader would receive a windfall. Best of all, no real energy is involved, so the only constraint on such a market scheme would be the appetite for risk of the trader involved.

Even worse, if the convergence bid was submitted by a generator whose unit could contribute to such a shortage if it went off line, this would be a powerful incentive to report unnecessary outages. Given the very high outage rates experienced by CAISO, this seems like a serious concern.

In a horse race, when you bet against your own horse and it loses the race, the authorities step in and test the horse for drugs. This seems like a good reason for FERC to carefully review convergence trades over this period. The decision by CAISO to close the convergence market after Aug. 15 indicates that similar concerns might have occurred to them as well.