Fault diagnosis method for oil-immersed transformer based on XGBoost optimized by genetic algorithm

In order to improve the accuracy and reliability of fault diagnosis for oil-immersed transformer, the fault diagnosis method for oil-immersed transformer based on XGBoost(eXtreme Gradient Boosting) optimized by GA(Genetic Algorithm) is studied. Firstly, based on DGA(Dissolved Gas Analysis), the non-coding method is used to extract the 9-dimensional fault characteristics of oil-immersed transformer, and the data samples are normalized. The fault diagnosis model based on XGBoost is built with the normalized samples as inputs, and the hyperparameters in the model are simultaneously optimized by GA. In case study, 547 samples of DGA data determined by fault types are collected for comparison experiments. Results show that the diagnosis accuracy and stability of the proposed method are significantly improved compared with the existing traditional methods. The optimization effect of GA on the fault diagnosis model is verified at the same time. © 2021, Electric Power Automation Equipment Press. All right reserved.