Transformer fault diagnosis based on Fisher criterion and BKA-optimized KELM

Abstract

To improve the quality of extracting characteristic parameters of dissolved gases in transformer oil and the accuracy of fault diagnosis, this paper proposes a new method for transformer fault diagnosis based on the Fisher criterion and the Black-winged Kite Algorithm (BKA) to optimize the Kernel Extreme Learning Machine (KELM). First, based on the Dissolved Gas Analysis (DGA) method, the 5-dimensional set of fault characteristic gases is expanded to 30 dimensions using the ratio method. The Fisher criterion is then applied for dimensionality reduction and feature selection to reduce feature redundancy, and a new feature set is constructed using the selected features. The KELM method was employed for transformer fault classification. To address the issue of certain model parameters significantly influencing diagnostic results, the BKA algorithm was applied to optimize relevant KELM parameters using training set samples, thereby establishing the BKA-KELM fault diagnosis model. Test set samples were then input into the BKA-KELM model for fault diagnosis. Experimental results show that the proposed method achieves a fault diagnosis accuracy of 99.67%. Compared to the PSO-KELM model optimized using the Particle Swarm Optimization (PSO) algorithm, the GWOKELM model optimized using the Grey Wolf Optimizer (GWO) algorithm, and the WOA-KELM model optimized by the Whale Optimization Algorithm (WOA), the fault diagnosis accuracy of the proposed method improved by 3%, 2.67%, and 2%, respectively. This validates the effectiveness of the proposed method and demonstrates its practical engineering value.