A V2G power response model based on the prediction, evaluation, and correction of charging stations real-time frequency regulation capability

Abstract

With the charging stations (CSs) construction and the vehicle-to-grid (V2G) development, electric vehicles (EVs) have become an important load-side controllable resource. Therefore, a V2G power response model based on the prediction, evaluation, and correction of CSs real-time frequency regulation (FR) capability is proposed in this paper. Firstly, a hierarchical control framework for large-scale EVs aggregated to participate in power grid dispatching/FR service is proposed. Secondly, an extreme gradient boosting (XGBoost)-convolutional neural network (CNN)-bidirectional long-term and short-term memory (BiLSTM)-attention prediction model for CSs FR capability is proposed, which combines the advantages of CNN and BiLSTM to strengthen the mining of multi-dimensional features. Meanwhile, a rolling evaluation-correction model for CSs FR capability based on the EV CC–CV charging process is proposed, which improves the evaluation fineness and aggregation fitness. Furthermore, a V2G power response model considering the EV battery loss is established. Finally, the simulation results show that compared with LSTM, XGBoost-CNN-BiLSTM, support vector machine, and other prediction models, the proposed XGBoost-CNN-BiLSTM-attention CSs FR capability prediction model with improvement has a better prediction accuracy. In addition, the V2G power response model can achieve the coordination between the EV users’ charging demands and FR tasks.