On the Persistence Forecast of Single Home Electricity Power Consumption

This work analyzes an electricity power consumption forecast for a single home using a various type of generalized persistence models. The predictor composes of parallel banks of simple averaging models (SAM) for each hour within a day. Fitting SAM is performed separately for each day of the week using appropriate training sets of past electricity power consumption. The performance of the predictor was evaluated using real data which represents power consumption per minute measured over almost 4 years for a single home near Paris, France (approximately 2 million data points). Experiments show that proposed system for predicting power consumption one day ahead gave mean absolute value of relative percentage error (MARPE) lower more than 18% comparing to ordinary persistent prediction, without significant computational cost.