Abstract: In view of the accuracy of short-term load forecasting,a short-term load forecasting method based on improved grey relational analysis and back propagation（BP）neural network optimized by bat algorithm（IGRA-BA-BP）is proposed.On the basis of traditional grey relational analysis,comprehensive grey correlation degree associated with distance proximity and shape similarity is introduced to select the similar days of higher similarity.In order to reduce the difference of training samples and improve the accuracy of prediction,the samples of similar day set are used to train BP neural network prediction model which is optimized by bat algorithm.Taking historical data in a region of southern China as an actual example,the prediction results of simple BP neural network,BP neural network optimized by bat algorithm（BA-BP）and the traditional grey relational analysis and BP neural network optimized bat algorithm（GRA-BA-BP）are compared with the short-term load forecasting method based on the improved grey relational analysis and BP neural network optimized by bat algorithm,the results show that the prediction accuracy of the proposed method is better.