Accurate spatial and temporal distribution prediction model for electric vehicle charging load is an important basis for dealing with the impact of electric vehicle connected to the grid and researching charging facility planning. Due to the limited number and the unreasonable layout of existing charging facilities, the historical electric charging load data can not reflect the actual charging demand of electric vehicles explicitly. Under the background, a load forecasting method of power system is proposed based on the grid division. Firstly, the prediction region is divided into block that is taken as the spatial prediction unit. Then, the charging load of block with charging facilities and the history data of prediction indicators are utilized to assign historical charging data in blocks with charging facilities to each block. Secondly, a relationship between charging load and influencing factors is established by employing Bayesian regularization BP neural network algorithm. Finally, Haidian District of Beijing is considered for simulation to verify the proposed prediction method. It is shown that this method can accurately predict the spatial and temporal distribution of the electric vehicle charging load.