Intelligent diagnosis of partial discharge in mining dry-type transformers based on hybrid optimization XGBoost

The partial discharge signal of mining dry-type transformers is weak and susceptible to interference, resulting in information redundancy or improper feature selection in the process of feature extraction and classification, which affects the accuracy and stability of diagnostic results. Therefore, a partial discharge intelligent diagnosis method for mining dry-type transformers based on hybrid optimization XGBoost is proposed. Firstly, the partial discharge signal of the transformer is denoised and its key features are extracted. Adopting the hybrid optimization XGBoost algorithm to screen and optimize the extracted features, in order to remove redundant information and retain the most influential feature subset. Use hierarchical clustering method to perform clustering analysis on features. Finally, based on the clustering results, the characteristic category of partial discharge is determined to achieve intelligent diagnosis. The experimental results show that this method improves the accuracy of partial discharge diagnosis for mining dry-type transformers.