Research on distribution network load forecasting considering the impact of EVs integration based on TCN-EAD method

The charging/discharging behavior of electric vehicles (EVs) constitutes a highly flexible load resource, playing a significant role in distribution network regulation and control. This integration alters the fundamental characteristics of the basic power load, which presents limitations in traditional load forecasting methods. In order to enhance the accuracy of distribution network load predictions, this paper proposes a novel forecasting technique, which leverages a temporal convolutional network combined with an encoder-attention-decoder (TCN-EAD) architecture, explicitly accounting for the influence of EV integration operating in vehicle-to-grid (V2G) mode. Firstly, the Markov Chain Monte Carlo (MCMC) technique is employed to simulate EV charging/discharging profiles based on trip chain analysis. Then, a comprehensive forecasting strategy utilizing the TCN-EAD model is applied to train on and predict the distribution network load. This analysis examines the impact of EVs charging /discharging load on the basic power demand. Additionally, a comparison is conducted among various forecasting methods and models to illustrate the effectiveness and accuracy of this proposed method, particularly under scenarios involving substantial EV grid penetration.