Power system short-term load prediction method and related device

The invention discloses an electric power system short-term load prediction method and a related device, and the method comprises the steps: obtaining load related data in an electric power system, the load related data comprising date, time, temperature, humidity and an electric power load value; carrying out preprocessing operation on the load related data, wherein the preprocessing operation comprises data cleaning, abnormal value elimination and missing value filling; the load related data are input into a preset KELM model for feature extraction operation, a plurality of optimized load feature vectors are obtained, and the preset KELM model comprises an XGBoost layer, a KELM layer and an Adaboost layer; and performing fusion prediction according to the plurality of optimized load feature vectors based on Stacking-NN, and calculating the difference degree among the plurality of optimized load feature vectors by adopting a preset Pearson correlation coefficient in the fusion prediction process, thereby obt