The invention provides an electric vehicle load prediction method. The method comprises the steps of dividing historical data samples into three parts, wherein the first two parts are training sampledata, and detection sample data of the third part is used for detecting a network and a generalization ability of the network; by using the training sample data of the two parts and a genetic algorithm, optimizing the neural network twice; and finally obtaining a group of weight threshold values and a network structure for prediction. According to the method, the shortcomings of the neural networkare overcome, and the advantages of the genetic algorithm and the neural network are combined, so that the precision of electric vehicle load prediction is improved.