Interval-specific building load forecasting models for demand resource planning

This paper presents a method of load forecasting specifically for predicting a building's electrical load for demand resource planning. This forecasting method allows the manager of a controllable load to assess his or her risk and capabilities when participating in the energy market. Correlation studies are performed using demand data collected for the main library building at Drexel University in order to identify important factors that drive the electrical demand of the building. A general problem formulation for building-specific load forecasting is first presented. The remainder of the paper focuses on the variability of the model type for each forecast interval and its impact on the forecast. It is shown that unlike in other forecasting methods, the best-fit model varies with time of day. Using the collected demand data, the forecasting method presented here is tested and the results compared to that of a basic forecaster.