By Suzanne Herel

PJM is reducing its load forecast for 2018 by 2.6%, due in part to a temporary change in modeling that aims to address over-forecasting in recent years.

Acknowledging criticism that its forecasts have overestimated economic growth and failed to capture energy efficiency and behavioral changes that have dampened demand, PJM officials will use a “binary variable” to reduce next year’s forecast.

“There are things outside our model that our model is not picking up,” PJM’s Andrew Gledhill told the Planning Committee last week in a briefing on its draft load forecast.

Before applying the variable, PJM was projecting a 1.5% reduction in its 2018 summer peak load compared with the projection it made last year.

In addition to reducing the forecast for summer 2018 — the delivery year for next year’s capacity auction — the draft report reduces the summer peak load forecast for 2015 by 4,716 MW (-2.9%).

Peak load for 2020, the next Regional Transmission Expansion Plan (RTEP) study year, was cut by 4,152 MW (-2.5%) versus last year’s projection.

The workaround is a short-term fix. A more comprehensive solution is undergoing testing and expected to be implemented next year. (See Planning Committee Briefs: PJM Seeking Improved Load Forecasts.)

Economist James Wilson, a consultant to consumer advocates, questioned the use of the binary variable, saying it overcorrects in the short term and results in too high a rate of growth in later years. “It’s not a very good approach,” he said.

PJM Vice President of Planning Steve Herling said the debate would soon be moot. “I’m less concerned about the long-term implications of [this year’s fix] because we’re not going to be doing it next year,” he said.

Wilson also questioned why forecasters continue to add years to their historical period instead of dropping some of the earlier years.

Wilson said the first four years of PJM’s 1998-2014 historical base was a period when peak demand was growing in about a 1-to-1 relationship with growth in PJM’s economic variable, an elasticity that hasn’t been seen since and which may not return because of increased energy efficiency and demand response.

“A better way to move the forecast in the right direction would be to drop some of those now-anomalous early years from the forecast period,” he said.

Gledhill said the data from those previous years remains valid. “When you start shortening the estimation period, you’re shortening the period where you can measure how load reacts to economics,” he said.

Direct Energy’s David “Scarp” Scarpignato backed Wilson’s argument. “Something’s changed that’s making that data way-back-when less useful in the forecast,” he said. “Do you really want more data points if some of the data points are garbage?”

Herling said a final forecast report will be presented by the end of this month.