Short-term probabilistic household electricity load forecasting based on trend decomposition

With the rapid development of home energy management systems (HEMS), forecasting electricity loads is increasingly important for the operations of energy storage and usage in residential households. However, household electricity loads can be extremely uncertain due to the complex consumption behaviors of residents, posing challenges for existing hours-ahead forecasting methods. Moreover, there is a scarcity of research considering loads and weather conditions at different time scales for probabilistic load forecasting (PLF) methods, thus forecasts are dependent on trends of a limited spectrum. In this study, a decomposition-based method is proposed for the short-term probabilistic load forecasting (STPLF) of household electricity loads. Specifically, historical data are decomposed into trend components of different time scales using a novel structure. Based on these components, quantile forecasts are produced respectively and synthesized into total probabilistic forecasts for loads hours ahead. The effectiveness and superiority of the proposed method are demonstrated by the simulation results on household electricity loads.