Distributed PV Cluster Partition Based on KNN-DPC Clustering Algorithm

This paper addresses the issue of voltage violation caused by numerous distributed PV systems connected to distribution networks. Considering that the DPC algorithm is greatly influenced by subjective definition of the cutoff distance during its application, a K-nearest neighbor optimization DPC algorithm (KNN-DPC) is proposed. By utilizing the comprehensive sensitivity matrix to characterize electrical distances as inputs and accelerating the search for density peak points, the impact of cutoff distance can be avoided, and the precision of the cluster division can be efficiently enhanced. The IEEE33 node and IEEE69 node distribution system with distributed PV integration is used to verify the cluster division, and the modularity function is adopted as an evaluation index for the division results. The evaluation index for different K values is calculated, and the implementation effects of different algorithms are compared correspondingly, demonstrating the efficacy and superiority of the proposed approach.