Health assessment of circuit breaker considering external environment factor based on three-stage extreme gradient boosting

In order to achieve the on-line evaluation of circuit breaker operation state accurately and quickly, a three-stage health assessment approach based on extreme gradient boosting (XGBoost) is proposed in this paper. The presented method assesses the health condition of circuit breaker using data from its electrical, mechanical, insulation, temperature and external environmental characteristics in three stages. The first stage is applied to resolve if there is any equipment abnormality during the circuit breaker operation. Then, if there is an abnormality detected, the second stage is utilized to settle the category of the abnormality based on the multi-dimensional characteristic data. The third stage evaluates the degree of abnormality of the circuit breaker quantitatively. In the three-stage of health assessment, XGBoost algorithm is adopted to establish the health assessment model, and the optimal parameters of XGBoost model are obtained using both grid search method and cross validation method. In addition, to address the disadvantage of unbalanced sample data categories, Gaussian noise method are employed to enrich the operational database. The simulation result of l0kV circuit breakers shows that evaluation accuracy of the model in the paper is 92.9%, which can accurately evaluate the operational state of circuit breaker and facilitate the maintenance and early warning of the equipment.