Policy Advisor To Former Texas PUC Commissioner Says ORDC Contains Errors In LOLP Calculation

October 08,2018

Richard E. Wakeland, P.E., former Policy Advisor to former Texas PUC Commissioner Kenneth W. Anderson, has submitted comments to the Texas PUC concerning the PUC's review of summer 2018 market performance, in which Wakeland says that the Operating Reserve Demand Curve suffers from two errors in the calculation of the loss of load probability (LOLP)

Wakeland writes, "the ORDC did not perform as expected through the summer peak demand period, nor has it ever performed as expected. The loss of load probability (LOLP) is incorrectly calculated in two ways: (1) The application of the hour-ahead forecasted reserve level error distribution can only predict the chance of real-time reserve level being less than the minimum contingency level (MCL or X) if it is applied to the hour-ahead forecasted reserve level. The ERCOT ORDC LOLP determination has never calculated the LOLP correctly. The hour-ahead forecasted reserve level error distribution has always been incorrectly applied to the real-time reserve level. There is no mathematical basis for such an application, and the LOLP calculated by the methodology does not represent the probability that the real-time reserve level will equal or be less than MCL over the next hour as it has been purported to do. (2) When this incorrect LOLP is calculated by integrating the seasonal and time of day hour-ahead forecasted reserve level error distribution, the LOLP is undercalculated by a factor of 2. This is because of a misapplication of normal distribution probability theory."

In summation, Wakeland writes, "ERCOT has an ORDC LOLP that (1) calculates a fictitious LOLP because it uses the hour-ahead forecasted reserve level distribution with the real-time reserve level, (2) has to use 6 distributions a day, four seasons of the year to get something that remotely approaches a normal distribution, and this introduces pricing reliability unit commitments into the market six times a day (3) underestimates the LOLP when it does calculate it because of a mis-application of normal distribution probability theory and (4) now has created a missing money appeal to introduce a perturbation of a multiplying factor into the gooned up mathematics in an attempt to correct a problem that doesn't exist because the distribution itself is better because of improved forecasting."

"The entire problem is because ERCOT's determination of LOLP is based upon the overall accuracy of a forecast of the system, not based upon how the market perceives risk through the price of energy. Market forces drive and control price formation, not the forecast. If it is desired to have a scarcity pricing mechanism reflect a true valuation of operating reserves, the scarcity pricing mechanism must be structured around the price of energy, and how that price changes as operating reserves increase or decrease. The market doesn't give a damn about ERCOT's forecasting ability. Couple this foundational problem with mathematical errors that calculate a fictitious LOLP and underestimate the fictitious LOLP by a factor of 2.0 and you get market participants that are victims of the poor design looking for any relief possible, in this case a multiplying factor for [sigma] greater than 1.0. It is like tossing morphine syrettes to badly wounded shipmates instead of getting them to the fleet surgeon as soon as possible. The victims in ERCOT crave the morphine, but they need the surgeon. Correct the foundational problem (don't use forecasts to determine a scarcity pricing mechanism) and use correct mathematics, and you will fix the problems. To do that, the hour-ahead forecasted reserve level error distribution needs to be thrown away and replaced with a model that satisfies first principles and ideally has a basis in market pricing."

Wakeland's filing includes two working papers discussing the errors, which can be found here