In order to solve the problem that EV charging load is susceptible to “time shift” due to meteorological factors, a load prediction model BiGRU-Attention is proposed to take key meteorological factors into account and to obtain data-dependent charging load key influencing factors by utilizing the Random Forest method. We propose a load prediction model BiGRU-Attention that takes into account the key meteorological factors and obtains the key influencing factors of charging load by using the random forest method. The results show that the deep learning method can achieve better prediction results for charging load data with nonlinear characteristics. Aiming at the characteristics of periodicity of load data and time shift of meteorological factors, BiGRU-Attention can fully explore the complex relationship between charging load and key meteorological factors, and it has better feature extraction ability, which can obtain higher load prediction accuracy to meet the operational needs of real power grids. Demand.