Data Augmentation Based Anomaly Data Detection for Charging Piles

Abstract

As electric vehicle (EV) charging facilities continue to grow in size, the proper operation of EV charging posts is of particular importance. However, certain non-human factors can lead to data anomalies in charging posts, thus hindering the normal operation of EV charging posts, as well as the daily operation and profitability of charging stations. Therefore, this paper lectures on the features of generative adversarial networks (GAN) that can retain the original data features and random forests that can detect anomalous data, and performs anomaly detection on the anomalous data detected by the EV charging station management system. Finally, the experimental results show that the GAN used in this paper can generate more anomalous data to augment the original dataset and that the model trained from the data-augmented dataset has higher data anomaly detection capability than the model trained from the dataset with less anomalous data without data augmentation.