Abstract: A short-term prediction model of power load based on bidirectional long short-term memory（BiLSTM）optimized by empirical mode decomposition（EMD）and hunter-prey optimization（HPO）algorithm was proposed,which can solve the problem that the prediction accuracy of power load is not high due to randomness and nonlinearity. Firstly,the load time series was decomposed into multiple intrinsic mode function components and a residual component by empirical mode decomposition.Then,the hunter-prey optimization algorithm was used to optimize the bidirectional long-term and short-term memory neural network to construct the HPO-BiLSTM prediction model. And each intrinsic mode function component and residual component were normalized and input into the HPO-BiLSTM prediction model for prediction. The predicted values of each component were inversely normalized and directly added to obtain the final prediction result.Finally,the load data of a certain area from March 1 to11,2018 were selected for analysis.The simulation results show that compared with BiLSTM,HPO-BiLSTM,EMD-BiLSTM,EMD-GA-BiLSTM and EMD-PSO-BiLSTM prediction models,the EMD-HPO-BiLSTM prediction model has higher prediction accuracy and better fitting effect.