Striving for Best Data for Load Forecasting


Executive Column: Steve Herling, vice president of Planning

The goal of load forecasting is to narrow the margin between what we predict the electric load will be and what it actually turns out to be. Because key variables that drive the forecast can be difficult to predict—weather and economic conditions, for example—PJM produces a distribution of possible loads, not just a single forecast value.

Planning studies and the Reliability Pricing Model capacity auction then use the median (also known as the “50/50”) load from this distribution.  The “50/50” refers to the fact that it is equally probable the actual load will be above the forecast as it will be below the forecast.

Through the stakeholder process this year, we are looking to make a number of changes that will impact the new 15-year forecast that comes out in January 2016. The first major change is to bring more precision into modeling customer behavior in future load forecasts. The second major change will reduce the number of years of weather history to be considered in predicting future weather.

Customer behavior includes, for example, energy-efficiency programs, behind-the-meter generation and rooftop solar. These developments were not explicitly captured in previous PJM load forecasts.  Now, however, based on more granular information we are receiving from the U.S. Energy Information Administration and from stakeholders, we are in the process of adjusting the model inputs to more explicitly capture these customer behavior changes.

Another key component to load forecasting is weather—looking at historical weather as a predictor for future weather. PJM has always used a 40-year weather history. However, in recent years, scientists have argued that 20 years is a more relevant period to consider. We currently are investigating this issue and will be sharing analysis results with the Planning Committee in September.

PJM has been working on these changes since the beginning of this year and is in the final stages of testing and analyzing the results to ensure the changes result in a more accurate load forecast model on both an RTO-wide and zonal basis.  PJM has taken its preliminary results to the Planning Committee and will review its final proposed load forecasting model enhancements with the Planning Committee and the Markets and Reliability Committee in the coming months.

Given the financial impacts of the RPM capacity procurement,  some have challenged PJM to apply the new thinking to the load forecast that’s to be used in August for the  annual capacity auction for the 2018/2019 delivery year.  As explained earlier, the modeling changes are still in development and some changes have yet to be reviewed by stakeholders.  In addition, there is work to be done to make sure that energy efficiency will not be double-counted in the load forecast and in the auction procurement.

PJM explains these factors in more detail in its answer to the complaint filed with the Federal Energy Regulatory Commission filed on July 20.