The invention discloses a user short-term power consumption prediction method based on multi-table fusion data, and belongs to the technical field of electric power. Certain correlation exists betweenexisting user energy consumption. And if the correlation among the user energy consumption can be fully mined, the prediction precision can be effectively improved. According to the method, firstly,similar days are selected through a fuzzy clustering method, then the similar days are used as training samples of a support vector machine, parameters of the support vector machine are optimized, anda trained model is obtained and used for predicting the electricity consumption of a user; According to the method, the correlation between user energy consumption can be fully mined, multi-table fusion data is reasonably utilized, and the prediction precision is effectively improved. According to the method, the influence factors of various power consumption are fully considered, the predictionprecision is effectively improved, the number of training samples is reduced, and the method is of great significance to detection of abnormal power consumption behaviors of users.