Power system load is a nonlinear, non-stationary time series of load with typical periodicity and randomness.In order to reduce the nonlinearity of the load sequence and improve the prediction accuracy, a lumped empirical mode decomposition method (EEMD)and a modified artificial neural network (GABP)short-term load forecasting method are proposed.Using EEMD to decompose the load sequence into several stationary components of different frequencies, highlighting the local characteristics of the original load data, solving the classification fuzzy problem in EMD decomposition, and using GABP network to predict, solving the problem that BP is easy to fall into the local optimal solution.The appropriate parameters are used to construct different EEMD-GABP prediction models for each component, and the meteorological factors are introduced to predict each component separately, and the final predicted value is obtained after reconstruction.The example shows that the accuracy prediction high stability of load based on EMD-GABP prediction model is higher than that of traditional models such as ARIMA model and SVM model.