Abstract: With the increasing proportion of cooling load in load structure year by year, it is of great significance to estimate the cooling load for the load medium and short term load forecasting. Influenced by the new normal economy and capacity elimination policies, the basic load varies greatly from month to month, and traditional cooling load calculation methods will be no longer applicable. This paper constructs a combined cooling load estimation model based on support vector regression and K-means clustering, including SVR-Winters-based variable-scale basic load forecasting and secondary stripping of EMD-Kmeans cooling load. Based on the actual data of a certain province in Northwest China, the results show that the method can effectively solve the problems such as large monthly differences in the basic load and intra-day random fluctuations, and has higher accuracy and better adaptability.