In view of the low accuracy and low efficiency of load forecasting, sub-sequence undergoing processing of complementary ensemble empirical mode decomposition (CEEMD) and fuzzy entropy (FE) is predicted using a model of arithmetic optimization algorithm (AOA) and least squares support vector machines (LSSVM). The CEEMD-FE-AOA-LSSVM prediction model is then constructed. First, the FE algorithm is used to reconstruct the entropy value of each sub-sequence after CEEMD processing. This improves the anti-interference ability and computing efficiency of the model. Then, the AOA-LSSVM model is used to predict each sub-sequence after comprehensive treatment, and the prediction is superimposed on the output. Finally, the model is compared horizontally and longitudinally by the error function, and its performance is tested using the two comparison results. Through experiments, compared with CEEMD-LSSVM, AOA-LSSVM and CEEMD-AOA-LSSVM, CEEMD-FE-AOA-LSSVM combination model can take into account both prediction accuracy and efficiency, and improve the overall performance. At the same time, it is verified that the model processed by CEEMD or AOA can effectively improve prediction accuracy. This work is supported by the National Natural Science Foundation of China (No. 51807133).