Fault section location in resonant grounding distribution systems based on feature subset optimization of phase current variation
Hao Bai,Mu‐Yan Chen,Mou‐Fa Guo,Yi‐Peng Liu,Jian‐Hong Gao
DOI: https://doi.org/10.1002/cta.3954
IF: 2.378
2024-02-17
International Journal of Circuit Theory and Applications
Abstract:In this paper, a single‐phase grounding fault section location method is proposed by using genetic algorithm and support vector machine (SVM) to achieve feature subset optimization. In the process of section identification, the empirical error caused by artificially selecting features is avoided, and cumulative errors of multiple links such as feature selection, parameter optimization, and model training are reduced. Additionally, the communication and measurement requirements are reduced since feeder terminal unit only needs to measure local three‐phase current information. Summary The existing single‐phase grounding (SPG) fault section location methods typically suffer from difficulty in feature selection, limited feeder terminal units (FTUs) configuration, and excessive dependence on communication, which weaken their generalization and robustness. To overcome these challenges, an SPG fault section location approach based on feature subset optimization is proposed. First, the relation between the position of FTU and its three‐phase current variation is analyzed, and its fault features are extracted to construct the candidate feature sets as feature subset optimization objects. Then, genetic algorithm and support vector machine (SVM) are combined to select the optimal feature subset with small dimensions and recognition error, which avoids the empirical errors of artificial feature selection. To reduce the cumulative errors, the SVM hyperparameters are simultaneously optimized. Finally, the SVM model is trained based on the optimal feature subset and hyperparameters. In the absence of zero‐sequence current measurement, three‐phase currents measured by FTU are locally processed to locate the fault section by the trained SVM. The experimental results verified the effectiveness and feasibility of the proposed method.
engineering, electrical & electronic