The invention discloses an optimization method for processing load data. By calculating a clustering validity index, an optimal cluster number k is given in a cluster number search range, so that theshortcoming that a conventional k-means algorithm cannot give the optimal cluster number is overcome; when a PFS value is maximal, the corresponding cluster number k reaches the optimum, and k types of outputs are selected, namely, historical load data is classified into k types to be processed, so that an optimal load historical data processing method is obtained; an improved k-means clustering algorithm is in the cluster number search range; a clustering center is stable and clustering results are subjected to clustering validity index calculation; and the cluster number corresponding to themaximal PFS index value is selected as the optimal cluster number, so that compared with a clustering algorithm in the past, a great improvement is achieved.