A Novel Fault Diagnosis of GIS Partial Discharge Based on Improved Whale Optimization Algorithm

• Published in: IEEE access Vol. 12; p. 1
• Main Authors: Sun, Wei, Ma, Hongzhong, Wang, Sihan
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Correlation coefficients; Discharge; Discharges (electric); Entropy; Fault detection; Fault diagnosis; Feature extraction; Gas insulation; GIS; improved whale optimization algorithm; Local optimization; Optimization algorithms; Optimization models; Parameters; Partial discharge; Permutations; Principal components analysis; Signal resolution; Support vector machines; Switchgear; VMD
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2024.3349410

Partial discharge (PD) seriously affects the operational safety of power equipment. In order to effectively diagnose the PD in gas insulated switchgear (GIS), a GIS PD fault diagnosis method based on improved whale optimization algorithm (IWOA) is proposed, which optimizes variational mode decomposition (VMD) and support vector machine (SVM) to adaptively determine the appropriate parameters and further enhance performance. A laboratory GIS PD platform is built to collect four types of PD fault signals (point discharge, particle discharge, floating discharge, and air-gap discharge). Firstly, a nonlinear arctangent convergence factor and adaptive weight are proposed to address the issue of local optimization in the WOA optimization process. Then, IWOA is used to optimize parameters of VMD (mode parameter K and penalty factor α). Next, effective intrinsic mode functions (IMFs) are screened through correlation coefficients which are greater than 0.2. Because a single scale cannot fully reflect all signal information, and more important information is distributed in other scales, multiscale permutation entropy (MPE) is introduced for feature extraction. Furthermore, the principal component analysis (PCA) method is employed for dimension reduction of initial feature vectors, which reduces the dimension of 33 feature vectors to 7. Finally, SVM based on IWOA is applied to train and test the experimental data to identify different types of PD faults, and achieve diagnosis of GIS PD. Through experimental analysis and comparison with other methods such as EMD-MPE, WOA-VMD-MSE, etc., the proposed method has good diagnostic effects. Also, it proves the robustness and feasibility of the presented solution. The optimization model provides a reference for solving fault diagnosis of GIS PD problems.