A Partitioning Method for Reactive Power and Voltage Control Based on AP Clustering

Abstract

The invention discloses a reactive voltage control zoning method based on AP clustering, comprising the following steps: obtaining voltage reactive power sensitivity based on a Jacobian matrix obtained through power flow calculation, and constructing a PQ node electrical distance matrix based on the voltage sensitivity between PQ nodes; Negate the elements of the PQ node electrical distance matrix to construct a similarity matrix; obtain the optimal cluster midpoint and the number of clusters, obtain the partition clustering results, and complete the PQ node clustering; calculate each PV node pair based on the perturbation method The voltage regulation sensitivity of each PQ area divides the PV nodes into the PQ partition with the most sensitive voltage regulation. The AP clustering algorithm of the present invention is applied to the voltage and reactive power partitions to automatically obtain the number of clusters, and conducts unsupervised learning based on deterministic evidence propagation, which effectively reduces the subjective factors in the clustering process and effectively solves the random search problem. Contains random factors.

Description

technical field [0001] The invention relates to an AP clustering-based reactive voltage control partition method. Background technique [0002] Grid node partitioning is the basis of three-level voltage control, so an effective partitioning method is the basic subject of voltage control. At present, the research results of partition methods are relatively rich, and the traditional voltage partition methods can be summarized into the following four categories: fuzzy clustering, graph theory, heuristic algorithms (such as simulated annealing method, tabu search, etc.), and learning methods (such as K-means). Existing methods can be better applied to traditional power grids, but rely on experience to manually specify the number of partitions, which will increase the subjective factors in the partitioning process; at the same time, when the clustering algorithm starts, its initial clustering center and search direction are random It is determined that the clustering result is e...

Images