Abstract: A substation load clustering algorithm based on an improved Gaussian mixture model（GMM） is proposed t o a d d r e s s t h e d r a w b a c k s o f t h e c o n v e n t i o n a l G M M clustering algorithm, such as complex calculation, slow rate of convergence, blindness and subjectivity in manually determining the number of clusters. Based on the conventional GMM clustering algorithm, the initial clustering center is chosen by the k-means algorithm. The reliability of the output results is increased while the number of iterations of the GMM clustering algorithm is decreased by using this algorithm.Under various clustering numbers, the algorithm generates the Davies-Boldin（DB） index, Calinski-Harabasz（CH） index, and silhouette coefficient（SC） of clustering results. The cluster evaluation mixed（CEM） index is created by using the entropy weight approach to calculate the weights of various evaluation indicators. The optimal number of clusters is calculated as the number of clusters that correspond to the maximum value of the CEM. This optimal number of clusters is then used as input into the improved GMM clustering algorithm to obtain the clustering results and clustering centers of the substation load. The illustration demonstrates that the suggested algorithm improves the computational efficiency and stability of the conventional GMM clustering algorithm and has good cluster synthesis capability for substation loads, which aids in clustering result optimization.