Electric Vehicle Charging Load Prediction Model Based on Improved LSTM Neural Network

With the rising penetration rate of electric vehicles, charging facilities and charging load will become the key growth point of urban power grid development. However, the centralized charging of a large number of electric vehicles leads to severe load fluctuations and intensifies the heteroscedasticity of the load series. At the same time, considering that the current research on electric vehicle (EV) charging load prediction lacks real data support, and the consideration scenario is too simple and the influencing factors are not fully considered. Therefore, an electric vehicle charging load prediction model considering heteroscedasticity and based on improved LSTM neural network is proposed. First, the heteroscedasticity of the original series is stripped by time series decomposition. Secondly, the correlation between load components and climate and temperature is explored. Finally, a medium and long term charge load forecasting framework for heteroscedasticity is constructed, and the LSTM neural network forecasting model is integrated to decompose and reconstruct the load. The simulation results show that the proposed model improves the accuracy of load prediction and provides important support for the rational planning and stable operation of urban power grid.