A Comparative Study of Predictive Approaches for Load Forecasting in Smart Buildings

Predicting electricity consumption represents one of most important information for efficient energy management in smart buildings. It is mainly used for occupancy prediction and the development of optimized control approaches of building's appliances (e.g., lighting and heating/air conditioning systems). Recently, several approaches have been proposed for load profiling, prediction and forecasting. The work presented in this paper is towards the development of load forecasting approaches for being integrated for occupancy prediction and context-driven control of building's appliances. We mainly investigated the accuracy of various machine learning and statistical methods for forecasting energy consumption. An IoT and Big Data based platform was deployed for gathering near-real time data about electricity/load consumption. Recorded data were used to deploy predictive models using ARIMA, SARIMA, XGBoost, Random Forest (RF), and Long Short-Term Memory. Experiments have been conducted and results are reported to shed more light on the accuracy of these methods for load forecasting.