Semi-Supervised Learning based Power Transformer Health Monitoring Index using Deep Learning

The proposed work is used to define the type of fault in power transformer using dissolved gas analysis (DGA). It is based on a semi-supervised deep neural network (DNN) architecture and hyperparameter-tuned transfer learning algorithms (DGA). To address the significant issue that the major pseudo-labeling encountered in available data are unlabelled in nature, which causes the impact of semi-supervised learning importance in this proposed work. Second, various demi-supervised learning approaches are used to determine the proposed work's effectiveness. Third, hyperparameters in the DNN are optimized using the Whale Optimization Algorithm (WOA) for different training optimizers. The suggested scenarios are based on 500 lab-gathered samples and literature samples. The word “results” refers to how the initial decomposing material affected the transformer fault's severity and demonstrates the suggested solution's viability. This model has a superior accuracy (97.8
^%
) for diagnosing transformer fault types compared to other DGA techniques.