Day-Ahead Residential Customer Load Forecasting Using Prophet

Short-term load forecasting at the customer level is challenging due to its high variability and uncertainty. In this paper, a forecasting library for Python and R, namely Prophet, is applied to this application. Historical load consumption patterns with 15-min time intervals, extracted from a local customer's premise, are used for training purposes. The tool is then applied to predict 24-hour ahead load forecast at any day of the year. Weather data in the form of ambient temperature is employed as an exogenous regressor to further improve load forecasting during the hot summer months. Various indices as used to measure the performance of the forecasting tool.