VALLEY FORGE, Pa. — Unpredictable rainstorms throughout PJM in the middle of July resulted in overestimated load forecasts last month, the RTO’s Chris Pilong told members during an Aug. 8 Operating Committee meeting.
“Just about every afternoon, we had rainstorms that were possible … and just about every day, those storms did hit,” he said, noting that they reduced temperatures and caused daily forecasts to exceed actual load by as much as 6,000 MW.
One example: While the RTO forecast peak demand of approximately 150,000 MW on July 20, early afternoon storms capped load at about 145,000, Pilong said. That allowed the summer peak so far to remain the previous day, July 19, when demand hit 146,635 MW, he said.
PJM’s peak load forecasting error for the month was 4.58%, 0.79 percentage points above July 2016, PJM’s Joe Ciabattoni said. The overall forecasting error was 3.13%, just above PJM’s 3% target. Ciabattoni attributed the deviation to the “pop-up storms” on the western side of the footprint. The errors were highest in the distribution territories of Dayton Power & Light, Duquesne Light, FirstEnergy’s American Transmission Systems Inc. and Duke Energy Ohio/Kentucky.
The monthly balancing authority area control error limit, which measures how well the RTO maintains constant frequency control, was 99.9%, with total excursions and excursion minutes at their lowest levels in at least the past year.
The Anatomy of a LMP Spike
LMPs jumped to about $900/MWh around 10:30 a.m. on July 27, Pilong said. An outage the previous evening on the Black Oak-Hatfield 500-kV line created increased flow on the Conastone-Peach Bottom 500-kV line. PJM had been controlling flow over the latter segment to roughly 93% of its limit, but flow levels on the line can be volatile, as they are sensitive to load or generation movements. Moves from hydro units and shifting load increased flow over the line by about 5%, Pilong said.
Operators directed PJM’s security-constrained economic dispatch (SCED) engine to reduce flow over the line by 6%, but the demand was too great and sent the engine into a “relaxation mode” in which it doesn’t attempt to control the flow. Pilong said that in those instances, SCED just fails to solve the case and moves on to resolve the rest of the system.
Operators reduced their request to 2%, which SCED could perform but only by raising the LMP to about $900/MWh. The price spike caused generation to respond, which reduced the strain on the line and allowed SCED to realign prices.
GT Power Group’s Dave Pratzon asked if such short-term and unexpected fluctuations were caused by PJM’s move from 15-minute to 10-minute dispatch settlements. He also questioned whether generators will be expected to follow PJM’s dispatch signals during those periods and how signal deviations will be handled.
Pilong said that with the shorter lookaheads, larger transmission changes will create greater price separation. Part of the issue, he said, is the response time of low-cost resources.
“If the low-cost generation isn’t able to move as rapidly as we need it to, we may need to reach out to more costly resources for a short period time,” he said.
Reserve Differences Explained
In response to a stakeholder data request, PJM’s Lisa Morelli explained why the real-time SCED engine has not priced for shortages that stakeholders have observed in published data.
The public data come from PJM’s emergency management system (EMS), Morelli said, which has more conservative estimates than SCED. The most significant difference for this was a 2% “back off” in which the EMS system assumes resources will only achieve 98% of their stated capability. This assumption was removed in changes made on July 11, Morelli said, and immediately resulted in a 300-MW increase in the EMS synch reserve.
There are also differences between real-time SCED’s 10-minute lookahead and EMS’ real-time data. Additionally, real-time SCED uses units’ SpinMax (the reserve maximum) to estimate reserves, while the EMS uses the lesser of the SpinMax or the EcoMax (the economic maximum).
Exelon’s Sharon Midgley noted that the data request also asked for all of the unapproved real-time SCED cases, which she said would provide more clarity on whether PJM operators are being presented with real-time SCED cases that include shortages but are declining to select them.
Midgley and Old Dominion Electric Cooperative’s Adrien Ford said they had not been aware of the removal of the 2% back off and asked that PJM be more communicative with such changes in the future.
Morelli said there has been a 12% improvement in the alignment between the two measurements since PJM moved to the 10-minute settlement.
— Rory D. Sweeney