Large consumer short-term electricity consumption prediction method and device and medium

The invention relates to a large user short-term electricity consumption prediction method and device and a medium. The method comprises the following steps: S1, acquiring historical daily load data of a large user; s2, clustering the historical daily load data of the large users by using a K-means clustering method to obtain clustered daily load data; s3, respectively inputting the N groups of daily load data into the corresponding N large user short-term electricity consumption prediction sub-models to obtain N prediction results; and S4, calculating the similarity between the input of the large user short-term electricity consumption prediction sub-models and the center of each class in the clustered daily load data, determining the weight of each large user short-term electricity consumption prediction sub-model, multiplying the weight of each large user short-term electricity consumption prediction sub-model by a corresponding prediction result, adding all obtained products, and obtaining the large user