A Load Forecasting Method of Electric Vehicles Charging Station Group Based on GAN-RF Model
Gang Wang, Lizhen Wu, Gan Xuan
Abstract
Aiming at the stochastic and fluctuant problem of load in electric vehicles (EVs) charging stations, and considering the influence of human behavior on load forecasting, an information-physics-society system (CPSS) oriented load forecasting method for electric vehicles charging stations is proposed. Firstly, the generative adversarial networks (GAN) is used to handle the artificial load data with historical data. A huge amount of data including social behavior is generated and time-series load forecasting data is extracted by using stochastic forest (RF) algorithm to complete load forecasting for different charging stations, and the relative position information of each charging station is recorded by establishing spatio-temporal matrix, the load forecasting of electric vehicle charging station group is finished by filling the forecasting data into the space-time Matrix. The GAN-RF model is compared with support vector regression (SVR), BP neural network model. The experimental results show that the proposed method based on GAN-RF has high prediction accuracy and strong generalization ability.