Short-term electric vehicles charging load forecasting based on deep learning in low-quality data environments

Abstract

The accurate prediction of electric vehicles (EVs) load is the research basis for evaluating the impact of EVs on the power grid and optimizing the operation of the power grid. However, because the accumulated data of the newly operated EV charging stations are scarce, it is very challenging to use scarce data to obtain accurate prediction results. On the one hand, the missing values and outliers in the scarce dataset have a greater impact on the prediction results. On the other hand, a model with high accuracy cannot be trained using scarce datasets. To obtain accurate EV prediction results based on scarce datasets, a data generation method based on a generative adversarial network (GAN) is proposed. The proposed model is used to alleviate the influence of low-quality EVs load datasets on the prediction results. In addition, the performance of the prediction model is critical for improving the accuracy. In this study, a new gating mechanism called the Mogrifier is adopted in the long short-term memory (LSTM) network to improve its performance. Finally, the effectiveness of the proposed method is verified by experiments.