Abstract:

In order to solve the problems of single feature selection and low prediction accuracy of residual electrical life prediction of AC contactor,a prediction model was proposed based on genetic algorithm support vector regression (GA-SVR).Firstly,the voltage and current information was collected using the AC contactor full-life test platform and the feature parameters which can characterize the operation state effectively were extracted.Then,the neighborhood component analysis was used to calculate the feature importance,and the optimal feature subset was selected combining with the Pearson correlation coefficient.Finally,the residual breaking times of the AC contactor were used as the life prediction label,and the residual electrical life of the AC contactor was predicted by the GA-SVR method.The case analysis shows that the method has a high accuracy and can meet the actual engineering needs.

Key words: AC contactor, feature selection, genetic algorithm, support vector regression (SVR)