Energy demand forecasting: industry practices and challenges

Abstract

Accurate forecasting of energy demand plays a key role for utility companies, network operators, producers and suppliers of energy. Demand forecasts are utilized for unit commitment, market bidding, network operation and maintenance, integration of renewable energy sources, and for novel dynamic pricing mechanisms, e.g., demand response. In order to achieve accurate forecasts with high spatial and temporal resolution, data from various sources needs to be integrated: Smart meters, SCADA, weather forecasts, physical, statistical and geographical models. In this talk I will give an overview of recent work within IBM Research on an intelligent large-scale energy demand forecasting solution which provides forecasts at different aggregation levels, quantifies uncertainty in demand, and estimates the amount of distributed renewable energy behind the meters. The solution can be seamlessly integrated with external applications for network planning and decision support, and has been validated with leading electric utility companies world-wide.