Density peaks algorithm based on information entropy and merging strategy for power load curve clustering

To solve the problems of density peaks clustering (DPC) algorithm sensitive to cutoff distance and subjectivity of clustering center selection, we propose an improved density peaks algorithm based on information entropy and merging strategy (DPC-IEMS) for realizing power load curve clustering. First, a cutoff distance optimization method based on information entropy is proposed. This method uses sparrow search algorithm (SSA) to find the minimum value of information entropy about the product of local density and relative distance to calculate the optimal cutoff distance suitable for the load datasets. Then, a merging strategy is proposed to realize the adaptive selection of clustering centers. This strategy first generates a large number of initial sub-clusters by DPC, and then merges the sub-clusters using the fusion condition until the final iteration condition is satisfied. The performance of DPC-IEMS algorithm is evaluated on the U.S. load datasets and the Chinese load datasets, and the effectiveness and practicality of DPC-IEMS algorithm for power load curve clustering are fully demonstrated.