PJM is adding more items to the to-do list resulting from the September heat wave, during which officials ordered limited load sheds to prevent a wider system collapse.
A 104-page analysis of the operational events and market impacts resulted in 22 recommendations, including 11 not previously announced (see sidebar). RTO officials briefed members on the report last week — ironically amidst the arctic blast that set a new winter demand record.
The analysis reads a bit like a thriller, documenting PJM dispatchers’ minute-by-minute decision-making — and identifying mistakes and missed opportunities for reducing or eliminating some of the five load sheds Sept. 9 and 10.
The city of Sturgis, Mich., emerges as a hero in the drama, as the city’s behind-the-meter generator and conservation measures by residents combined to provide 8 MW of relief, preventing a third day of load shedding on Sept. 11.
The report attributed the load sheds in part to inaccurate transmission, weather and load forecast models and also cited errors in synchronized reserve estimates. Load sheds did not significantly affect prices, the report concludes. But the dispatch of demand response caused both price increases and decreases and shortfalls in Financial Transmission Rights funding.
The report also illustrates the limits of demand response in relieving transmission constraints and identified operator errors and communication lapses.
PJM has already taken steps to address half of the recommendations in the report. (See Big To-Do List from September Heat Wave.)
Among the previously undisclosed details in the report:
- Closing the South Akron-Clay 138 kV line might have prevented the Sept. 10 load shed in FirstEnergy’s Tod area near Warren, Ohio.
- PJM might have avoided the load shed in the AEP Summit area Sept. 10 by dispatching 395 MW of combustion turbines that were off line. It did not do so because of a modeling error and because it was not monitoring a 138 kV line not under RTO control.
- The Sept. 9 and 10 load sheds in AEP’s Pigeon River area in southern Michigan might have been avoided had a scheduled rebuilding of a 69-kV line been complete. PJM is working with AEP to “fast-track” the project, which is currently scheduled for completion in June 2017 under the Regional Transmission Expansion Plan.
- PJM should have ordered the Sept. 10 load shed in Erie South area of Penelec 40 minutes earlier than it did, immediately after an analysis indicated it was the only solution to prevent a potential cascading outage. The load shed was preceded by the unplanned outage of two hydropower units (Seneca #1 and #2) that were scheduled to run at full output, a combined 421 MW.
- The Environmental Protection Agency’s Mercury and Air Toxics Standards (MATS) indirectly contributed to one load shed. The planned outage of the South Canton 765 kV/345 kV transformer — required to support an upgrade needed prior to the retirement of five New Castle generators —contributed to less than 1 MW of the 16 MW FirstEnergy (ATSI) Tod area load shed.
- Access to recently retired generation would not have eliminated the load sheds, although the five Bay Shore and East Lake generators retired in September 2012 could have reduced the Tod load shed by 75% and the AEP Summit outage by almost half.
Modeling
Many of the report’s findings and recommendations deal with PJM’s transmission modeling:
- PJM’s contingency analysis of the Pigeon River area failed to include both a planned outage on the 69-kV Moore Park Tap-Industrial Park line and a relay limitation on the Lagrange-Howe (NIPSCO) section of the line. Because of the relay limitation, the most severe real-time contingency would automatically relay the Lagrange-Howe 69-kV line out of service. The Moore Park Tap-Industrial Park line was not modeled by PJM because it is below the 100-kV level; current PJM rules do not require reporting of outages below 100-kV.
- Ratings on the Summit-Industrial 138 kV line, which figured in the 25 MW load shed in the AEP Summit area, were incorrectly listed as 251 MVA for normal (24 hours), emergency (four hours), and maximum (15 minute) conditions. “The reason for having different ratings is to give the dispatcher time to trend and validate the flows as well as take action to reduce the flows on the line,” the report said. “The impact of all the ratings being the same is there is no time for the dispatcher to perform anything but the most extreme action that must be taken once the load dump rating is reached. In this case, it was to issue the PCLLRW [Post Contingency Local Load Relief] and ultimately shed load.”
- PJM incorrectly modeled a 138 kV series device, resulting in a 20 MVA difference between PJM and AEP’s state estimator solutions. PJM correctly compensated for the difference in real-time by conducting a cascading outage analysis at AEP’s lower threshold.
- Because of the modeling error and because the Industrial-Summit 138 kV line is classified as a monitored priority 2 (MP 2) facility — which is above the 100 kV NERC Bulk Electric System (BES) level but not turned over to PJM for control — PJM did not dispatch 395 MW of combustion turbines that were off line, “which may have eliminated the need for the load shed.”
Synchronized Reserve
The heat wave also exposed problems with PJM’s estimates of synchronized reserves.
PJM issued a call for synchronized reserves Sept. 10, believing it had 1,665 MW available. It never got more than 400 MW of relief, with only 200 MW showing up in the first 10 minutes. As a result, the spinning event — which normally last only 10 minutes — ran for more than an hour.
The report concluded that some generation operators do not respond to PJM requests to confirm their synchronized reserves — called an Instantaneous Reserve Check (IRC) — or “provide stale or unreliable data.”
It also cited errors by operators who manually reduced output from some generating units to relieve transmission constraints Sept. 10. Because they failed to log the units as “Manual Dispatch,” PJM’s Security Constrained Economic Dispatch (SCED) software returned the units to a higher output and calculated available Tier 1 reserves from some units on the sending end of transmission constraints, although those units could not increase their output.
Heat Wave Forecasting Errors
Forecasting temperatures also proved problematic. Temperature forecasts for 10 PJM zones missed actual conditions by an average of 2 to 3 degrees over the three days, with errors as high as 10 degrees.
These contributed to load forecasts that fell up to 3.6% short. “Backcasting” — rerunning the load forecast using actual temperatures to separate the effect of the weather forecast errors — still produced average errors of 2% to 4.5%.
PJM said this is because its “Neural Net” forecasting tool relies on the previous day’s temperature and load trends. “When temperatures change significantly from one day to the next, it takes time for the Neural Net to catch up. Therefore the model inherently does not handle this first day of change well.”
Communications
The report also raises questions about how PJM operators communicated their actions to others on the grid and within PJM headquarters.
It noted that operations management chose not to call a System Operations Subcommittee Transmission (SOS-T) conference call on Sept. 10, because although “temperatures were higher than normal there were no forecasted events that would adversely impact the bulk electric system.” The calls are scheduled on an as-needed basis during emergency events to share information.
After the five load sheds, on Sept. 9 and 10, a conference call was held on Sept. 11. “While most SOS-Transmission members agreed that the communications of the conference call were adequate, some conference call participants stated that they would have liked more detailed information provided for the operations issues being discussed.”
Generic Logging
Dispatch staff logged the load sheds as generic transmission events because they were unaware of a category in the Emergency Procedures application for a “Local Load Relief Action.” Officials said dispatchers were unaware of the category because it had rarely if ever been used before and because its name did not exactly match the “Post Contingency Local Load Relief Action” instructions in PJM Manual 13: Emergency Operations. “As a result, those parties who depended on the Emergency Procedures application for notification were not notified of the load shed events.”
Many PJM officials, including the State Government Policy, Member Relations, Federal Government Affairs, and Corporate Communications departments were not informed about the load sheds, delaying their ability to communicate with stakeholders. “Dispatch has no formal notification checklist to follow except for certain emergency procedures steps requiring specific notifications pursuant to DOE, FERC, NERC, or PJM Manual requirements.”
Demand Response
The September heat wave illustrated that demand response — which proved a valuable tool during capacity shortages during the July heat wave — is less useful in relieving transmission constraints.
PJM dispatched DR on Sept. 10 and 11 after load forecasts fell 4,000 to 5,000 MW short of actual load. Of 740 MW called in ATSI and the South Canton subzone, 695 MW responded (94%).
Curtailment service providers provide street addresses for their resources but this information is not mapped electrically to the nearest substation. “When using these resources for transmission constraints, it is important for the dispatchers to know precisely where the curtailment will occur so that they can better understand the impact on the observed constraint,” the report said. “Too many DR resources on the wrong side of a constraint can make a constraint worse.”
The report identified 11 MW of demand response in the Summit area, which it said could have reduced, but not eliminated, that Sept. 10 load shed. The exact impact of the DR is unknown because of the lack of electrical mapping.
In addition, the long lead time of most of the DR resources does not lend itself well to addressing transmission constraints, which often need controlling actions within 30 minutes.
Starting in delivery year 2014/15, DR is required to respond on a subzonal basis if PJM establishes subzone the day before issuing a dispatch order. Only seven subzones are currently defined.
In December, members approved changes that will allow PJM operators more flexibility in dispatching demand response, including a reduction in the lead time to 30 minutes. (See Members OK DR Dispatch Rules after Late Amendments.)
Price Impact
The load sheds “did not have significant impacts on market outcomes,” the report concluded. Demand response, however, set prices in the ATSI zone at about $1,800 for hours ending 15 through 20 on Sept. 11 while causing prices to crash from more than $200 to less than $70 in some other regions in HE 16.
PJM said such a price drop usually results from a sudden influx of imports coming into PJM as price-takers. In this instance, it resulted from operators calling for more DR than was ultimately needed.
The DR deployment was called based on an expected peak load of 153,000 MW — nearly 6,600 MW above the actual peak. “Since PJM does not account for these MW as additional reserves, LMP is set by the marginal resource and demand response did not … set price when dispatched because this volume of demand response was not ultimately required,” the report explained.
After dropping below $70 in HE 16, prices rebounded to more than $200/MWh in HE 17.
FTR Funding Shortfalls
The dispatch of DR contributed to large differences between day-ahead and real-time prices in the ATSI zone, increasing FTR funding shortfalls for the month.
September 10 and 11 showed $29.3 million in FTR underfunding, more than half of the $56.3 million shortfall for the month. “Under the current market rules, FTR holders can be adversely impacted significantly by such emergency procedures taken to maintain system reliability when they have no impact to the Real-Time Market or system operations. PJM believes that this is a flaw in the market design that needs to be addressed.”