Energy Consumption Forecasting in Home Energy Management System using Deep Learning Techniques

Smart Grid is a self-sufficient system that allows any sort of generation source to be integrated into the grid, resulting in consistent, feasible, intact, and high-quality electrical power for end consumers. Among the many smart grid applications, the most important one to consider is the Home Energy Management System (HEMS). Energy forecasting plays a vital role in HEMS, because of inconsistent availability of generation and non-uniform and highly dynamic loading conditions. This article presents a brief introduction of HEMS and the techniques used for load forecasting. A novel technique based on long short-term memory (LSTM) and Deep Neural Network (DNN) is proposed for predicting periodic power consumption. Load forecasting is performed for a next seven-day period. When a microgrid is set up, several parameters such as energy consumption of various home appliances, solar irradiance, and total load are taken into account when building a dataset. This article gives a brief summary of HEMS and the load forecasting methodologies. The efficiency of proposed algorithms is illustrated and compared.