To improve the accuracy of short-term load forecasting on the user side, a short-term load combination prediction method based on adaptive chirp mode decomposition (ACMD) and sparrow search algorithm (SSA) optimized bi-directional long short-term memory network (BiLSTM) is proposed. Given the problem of strong fluctuation and non-stationarity of short-term power load, ACMD is used to decompose the short-term load time series into several relatively simple sub-components, and BiLSTM is used to predict each sub-component. At the same time, in order to overcome the problem of unstable prediction results caused by different parameter values of BiLSTM, SSA is used to optimize the hyperparameters of the BiLSTM model. The prediction results of each sub-component are superimposed to obtain the final prediction results. Compared with single prediction model and multiple combination prediction models, the experimental results show that this method has higher prediction accuracy. This work is supported by the National Science Foundation of China (No. 52077146).