Classification Model of Abnormal Heating Composite Insulator Based on Random Forest Algorithm
There are usually different correspondenct maintenance decisions and emergency treatments for various types of abnormal heating composite insulators.Therefore,it is urgent to establish a method for the identification of the abnormal heating composite insulators.In this paper,the surface axis temperature curves of the contaminated composite insulators,the composite insulators with aging and damp sheath and the decay-like composite insulators are analyzed,and seven temperature characteristic quantities are defined to characterize the abnormal heating states of the composite insulators.Then,a classification model of the abnormal heating composite insulators based on the random forest algorithm is proposed,and a sample set of the temperature feature quantities is constructed by using the infrared spectra of the abnormal heating composite insulators taken in the field and the laboratory.The SMOTE algorithm is used to perform the data augmentation and the 10-fold cross-validation on the temperature features of the decay-like composite insulator to select the number of the decision trees.The model test results show that the classification precision of the contaminated composite insulators,the composite insulators with aging and damp sheath and the decay-like composite insulators reaches 96％,93％and 94％,respectively.
abnormal heatingcomposite insulatorrandom forestdecay-like rod
万方数据
维普
