A Power System Network Partition Framework for Data-driven Regional Voltage Control

This paper investigates an efficient modal method for reducing weak transmission stability boundaries and identifying voltage control areas. This method divides the power system network into regions to eventually reduce the control candidates for each controller and minimize the interaction between each voltage control area. To determine the optimized number of areas, proper selection of threshold parameter ff associated with this method is required. A new approach using kmeans and Bayesian Information criterion to cluster identical Q-V minimum and the set of generators reaching their Q-limit are presented to determine the ff value. The data-driven voltage controller is implemented by parallel decision trees via offline training; it provides Voltage Security Assessment for each control combinations. The selected control combination will execute the control by switching on and off of each discrete control candidate. The effectiveness of the advocate division approach is demonstrated in IEEE 118 bus system. The results show the performance and ability of proposed system division approach to provide voltage control areas and reduce the control candidates for the data-driven regional voltage controller.