Short-Term Load Forecast Error Distributions and Implications for Renewable Integration Studies

Abstract

Load forecasting at the day-ahead timescale is a critical aspect of power system operations in the unit commitment process. It is also an important factor in renewable energy integration studies, where the combination of load and wind or solar forecasting techniques create the net load uncertainty that must be managed by the economic dispatch process or with suitable reserves. An understanding of the load forecasting errors that may occur in this process can lead to better decisions about the amount of reserves necessary to compensate for the errors that do occur. In this work, we performed a statistical analysis of the day-ahead (and two-day-ahead) load forecasting errors observed in two independent system operators for a one-year period. Comparisons were made with the normal distribution commonly assumed in power system operation simulations used for renewable power integration studies. Further analysis identified time periods when the load is more likely to be under- or overforecast.