EMERGENCY-READY EV CHARGING: OPTIMIZING LOAD PREDICTIONS FOR CRITICAL SITUATIONS
Keywords:
deep learning; long short-term memory artificial neural network; short term load prediction; K-means clustering algorithmAbstract
The use of electric vehicles not only drives economic growth but also contributes to the goal of "carbon peak and carbon neutrality". However, large-scale electric vehicle charging station access impacts the security and stability of the power grid. To effectively predict electric vehicle charging load, this paper builds LSTM neural network model based on PyTorch. The historical load data of a charging pile in Beijing from December 12, 2019 to February 1, 2020 was selected for analysis and prediction. Firstly, quantificat the epidemic’s impact on charging load, vectorize and normalize the historical load data. Then, the BPTT algorithm was used to train the neural network to predict the data of February 1, 2020, and the error was finally 4.17%. The prediction accuracy is much higher than that of the RNN and the CNN.