EV Load Forecasting Methods That Consider the Dispatchable Potential in V2G System

The EV charging loads have a strong spatial and temporal randomness, which makes it more difficult to control the grid and affects power quality. Considering the differences in the time series characteristics and influencing factors of different types of EV charging loads, a prediction model of EV charging loads considering day types are constructed; the DPC clustering algorithm based on principal component analysis (PCA) is used to cluster the charging loads and extract similar daily loads by mining the characteristic attributes of the data; the CNN-LSTM is used to predict the similar daily loads after clustering. The prediction results were compared with the test set, and the results showed that the prediction accuracy based on this model was higher than that of the non-clustered CNN-LSTM method, which verified the validity of the prediction model.