Charging Pile Load Forecasting using Long Short-Term Memory and Recurrent Neural Network

Nowadays, the development of Electric Vehicles (EVs) is probably leading to threaten the power system. As new EV gain popularity, cities are increasingly focusing on installing EV charging infrastructure. But, forecasting the demand for charging stations is one of the serious issues facing while mitigating growth of $\mathbf{E V}$ infrastructure, caused by the increase of diffusion of EVs. RNN (Recurrent Neural Network) makes difficulty to learn long-term dependencies in charging load patterns. To overcome these issues of load forecasting, the method of Long Short-Term Memory and Recurrent Neural Network (LSTM-RNN) is proposed to predicts the charging demand at the certain point of charging. Also, it ensures accurate forecasts and optimized energy management with infrastructure planning for EV. For the analysis, the proposed LSTM-RNN employs with a two-stage data preprocessing technique on charging load dataset. Then the refined data is fed into the load forecasting which predicts charging power requirements to optimize charging schedules and ensure reliable with efficient energy supply. The proposed LSTM-RNN presents greater results than the existing methods of Enhanced Attention-Based Long Short-term Memory (EA-LSTM) on load forecasting of charging pile in terms of MAE (0.987), RMSE (8.345) and $R^{2}(0.986)$ respectively.