Dominant Node Selection Method Based on Fast Search and Density Peak Clustering

Abstract

The invention discloses a pilot node selection method based on clustering by fast search and density peaks. A sensitivity method is used for sequentially quickly calculating voltage deviation of other PQ nodes after each PQ node eliminates disturbance voltage deviation, the calculated voltage deviation serves as a feature vector for the corresponding PQ node, and a clustering feature matrix is obtained; a clustering algorithm is used for calculating the local density of each node, the shortest distance with a high density point and the average shortest distance of all nodes according to the clustering feature matrix, a candidate clustering center is determined, and the number of pilot nodes is determined according to a system voltage reactive power sensitivity correlation matrix in the current state; and through reduced order optimization, the pilot nodes in the current three-stage voltage control cycle are determined. The fast clustering algorithm has an extremely short operation time, the calculation based on the sensitivity can avoid multiple times of power flow calculation, and thus, the provided algorithm overcomes limitations of the traditional large-scale exhaustive combination optimization, the operation amount is reduced, and the selection result has better control effects.

Description

technical field [0001] The invention relates to a dominant node selection method based on fast search and density peak clustering. Background technique [0002] The selection of the dominant node is an important part of the implementation process of the three-level voltage control of the large power grid. As the control target of the second-level voltage control link, its selection is very important for the control effect of the whole network optimization. [0003] The traditional dominant node selection method is very mature, which mainly includes directly taking the node with the largest electrical distance center or regional voltage and reactive power sensitivity as the dominant node, based on the sensitivity and aiming at the minimum voltage deviation of the load node to optimize the dominant node through genetic algorithms and other intelligent algorithms , on the basis of partitioning, consider node controllability factors and select dominant nodes based on node voltag...

AI Technical Summary

Problems solved by technology

Traditional clustering methods such as K-means, agglomerative clustering, etc. also use iterative search in the process of finding the cluster center, which is not suitable for online computing

Images