Condition Monitoring Method for Oil-Immersed Power Transformer based on Convolutional Neural Network and Metaheuristic OOBO Algorithm

Abstract

In this paper, a novel condition monitoring method is presented for an oil-immersed power transformer based on the Convolutional Neural Network (CNN) and metaheuristic One-to-One-Based-Optimizer (OOBO) algorithm. The CNN algorithm is employed for classifying the various faults, whereas metaheuristic OOBO is utilized to fine-tune the parameters of the CNN algorithm to enhance the classification accuracy. In addition, the unbalanced database is balanced using the SMOTE algorithm and normalized using the min-max normalization algorithm. The simulation evaluation is performed for the standard dataset, and various parameters are measured to evaluate the proposed method. The result shows the proposed method achieves a high value of accuracy for classifying the various faults. The proposed method achieves an accuracy value of 0.99 for discharge fault, 0.98 for partial discharge, 0.98 for thermal, and 0.99 for no fault.