Integrating Distributed PV Forecasts with Load Forecasts
Electrical load depends critically on weather variables, in addition to human usage elements.
Recently, the growing penetration distributed solar power production “behind the meter” has been making a substantial impact on net load, reducing demand during sunny periods and increasing net load variability under rapidly changeable weather conditions. The National Center for Atmospheric Research has developed a forecast system that leverages weather information and historical data to provide short-range net electrical load forecasts. This system utilizes forecasts of key meteorological variables from the NCAR-developed Dynamic Integrated foreCast (DICast®) system to produce hourly 0-168 hour load forecasts. An historical database of observed load, solar production, distributed solar installations, and weather variables is analyzed using statistical learning methods to create the forecast models. Such forecasts will facilitate utilities’ integration of an increasing base of distributed solar installations. NCAR’s net load forecast uses a statistical learning approach based on regression trees that has proved to provide forecasts within a couple percent of accuracy on average.