摘要: According to the data density of the data, select the initial cluster center to fully reflect the distribution characteristics of the data; In the iterative calculation of the new cluster center, the average distance be-tween the data and the center of the class is used as the cluster center of the new iteration to eliminate the influence of the noise point. The cluster evaluation Index Hubert Index and D Index select the optimal cluster number to reflect the spatial distribution of data more accurately. By using the improved k-means algorithm to cluster the data of a variable table, the typical daily load characteristic data of four quarters were obtained, and the short-term load prediction was carried out on this basis. The analysis shows that the load modeling method based on improved k-means is feasible and effective for grid simulation analysis.