CNN-LSTM Neural Network Based On Time Variable Weight Optimization For Electric Vehicle Charging Load Prediction In Commercial District

The distribution of electric vehicle charging load is random in time and space. The commercial center is located in the core of the city, the range of electric powered motors is large, and the influencing elements are various. Aiming at the fact that the traditional forecasting model does not consider the intensity of the influencing factors at different time periods, a LSTM neural network based on time variable weight optimization is proposed to predict the charging load of electric powered motors. The relevant information of man-vehicle-pile-network was collected, and the forward and reverse data were analyzed and processed by TOPSIS entropy weight method. The correlation weights were assigned to the influencing factors of man-vehicle-pile-network to realize multi-period factor differentiation. Finally, CNN-LSTM neural network is used to simulate and predict the charging load of a commercial power station in Nanchang in one day, and the proposed approach is in contrast with the traditional method to verify the effectiveness of the proposed method. The empirical results show that the electric vehicle charging load forecasting method is feasible.