Density Peak Clustering Load Curve Based on Mutual Local Density and Multi Cluster Merging

It is indispensable and important work in the context of electric power big data to extract and cluster analysis of electricity load users accurately, rapidly and efficiently, so as to obtain the user’s electricity consumption behavior and demand pattern. To address the problem that density peak clustering algorithms have difficulty in identifying the cluster centers of low-density datasets, which results in poor clustering results, and the problem that single-step assignment strategy is highly likely to lead to sample misallocation, this paper proposes a density peak load curve clustering algorithm based on mutual local density and multi-cluster merging. First, a new local density is defined based on the K-nearest neighbor algorithm. Next, the density peaks of all sample points are calculated, and the first m samples are selected as the initial density peaks. Then, the clusters are divided into micro-clusters according to the Pearson correlation between the sample points and the initial density peaks. Finally, the global and local features of each class cluster are considered and new fusion criteria are defined to achieve the fusion of micro-clusters. The algorithm analysis shows that the proposed algorithm can effectively identify the class cluster centers of low-density clusters. Moreover, compared with the single-step assignment strategy, the proposed multi-cluster fusion iterative strategy in this paper is superior in terms of cluster evaluation metrics.