Probability Prediction of Electric Vehicle Schedulable Capacity Based on Improved Informer with Copula

Inhalt:
Under the background of "carbon peaking" and "carbon neutralization", the scale of electric vehicles (EVs) in China have developed rapidly. EV aggregator schedulable capacity (EVASC) prediction is important for large-scale EV participation in ancillary services for the power system. In this paper, a probability prediction algorithm for the EVASC, the Improved Informer (IInformer) with Copula (IInformer- Copula), is proposed. By the proposed IInformer-Copula algorithm, the IInformer algorithm is designed to capture the change trend information of the input sequence, which overcomes the disadvantage of the traditional informer algorithm that the sparse attention cannot pay attention to the change trend information of the sequence. The Copula function is used to establish the joint probability distribution of the EVASC prediction values vs prediction errors, so as to obtain the EVASC probability prediction results. The proposed model is validated with the actual historical charging record data of a city. The results show that the proposed method can significantly improve the prediction accuracy and reduce the prediction time of a long-time sequence.