Residential DC Load Forecasting Using Long Short-term Memory Network (LSTM)

In modern power systems, load forecasting has become a crucial aspect. These advanced power systems demand flexibility, efficient operation, scalability, and better resource management. Future smart homes will incorporate both AC and DC distribution systems for energy efficiency. Therefore, accurate load forecasting is imperative. However, it can be a difficult task as it includes various factors such as the number of devices in a household, their diversity, time, location, season, and the behavior of the occupants. In this study, a short-term residential DC load forecasting model based on a deep learning algorithm called long-short-term memory networks (LSTM) has been developed. The model is trained using data from a household in Estonia and is evaluated by forecasting the next day’s load. The simulation results provide quite an accurate forecasting with root mean square error (RMSE) is around 0.15 kW.