Abstract: The domestic power load has high randomness, and using a single forecasting model to forecast will result in low accuracy and long prediction time. A combined forecasting model of ensemble empirical mode decomposition JP2(EEMD)-Automatic encoder (AE)-long short-term memory network (LSTM) is established for short-term load forecasting of domestic power consumption. The EEMD algorithm was used to decompose the load data into a finite number of intrinsic mode components (IMF) and a residual component, and then combined with the feature sequences trained by the automatic encoder, and the LSTM model was established to predict the linear weighting to generate the final prediction results. The experimental results show that EEMD-AE-LSTM model is more accurate than other models, and it is an effective short-term load forecasting method for domestic power.