Takagi–Sugeno Fuzzy Based Power System Fault Section Diagnosis Models Via Genetic Learning Adaptive GSK Algorithm
Changsong Li,Guojiang Xiong,Xiaofan Fu,Ali Wagdy Mohamed,Xufeng Yuan,Mohammed Azmi Al-Betar,Ponnuthurai Nagaratnam Suganthan
DOI: https://doi.org/10.1016/j.knosys.2022.109773
IF: 8.139
2022-01-01
Knowledge-Based Systems
Abstract:To effectively deal with the operating uncertainties of protective relays and circuit breakers existing in the power system faults, an improved fault section diagnosis (FSD) method is proposed by using Takagi–Sugeno fuzzy neural networks (T–S FNN). In this method an optimal T–S FNN-based diagnosis model is built with the idea of distributed parallel processing for each section instead of the whole power system. To obtain accurate T–S FNN-based diagnosis models, a genetic learning adaptive gaining-sharing knowledge-based algorithm (GLAGSK) is designed to optimize their structure parameters and consequent parameters. GLAGSK combines an adaptive knowledge ratio and a genetic learning strategy to balance population diversity and convergence speed to boost the optimization ability. After a fault occurs, selective optimal T–S FNN-based diagnosis models are triggered according to the alarm information. They work in parallel to improve the fault diagnosis efficiency. Simulation results of three test systems including an actual fault event show that, compared with other peer algorithms, GLAGSK can obtain more accurate T–S FNN-based diagnosis models with faster global convergence. Besides, compared with the BP and RBF neural networks and other FSD methods, the proposed FSD method based on optimal T–S FNN-based diagnosis models can diagnose different complex faults successfully with higher fault credibility. • Improved fault section diagnosis (FSD) method based on T–S FNN is proposed. • T–S FNN-based diagnosis model is designed for each power system section. • Enhanced metaheuristic named GLAGSK is presented to optimize T–S FNNs. • GLAGSK is boosted by adaptive knowledge ratio and genetic learning strategy. • Proposed FSD method can diagnose different complex fault scenarios accurately.