Health Index Analysis of XLPE Cable Insulation using Machine Learning Technique

Deterioration due to aging and partial discharge is the primary cause in cable insulation failure, however, replacement and maintenance of underground cable circuits, during the period of excavation, is very expensive. The information regarding the severity of the insulation level assists to make smarter informed decisions for system planning and repair prediction. The application of machine learning (ML) towards the prediction of the insulation health condition of high voltage XLPE cable was emphasized in this work. The interpretation and recognition of the insulation health condition analysed with the help of different machine learning algorithms like Support vector machine (SVM), K-Nearest Neighbour (KNN), Artificial Neural Network (ANN), and Naïve Bayes. The classification based on different ML classifier requires a pre-processing of the input data obtained from the test results. The test result provided information about each sample's Partial Discharge (PD) magnitude, Aging, Neutral corrosion, Loading, Visual condition, etc. This work mainly focused on the classification of the insulation dataset, i.e. the multiclass classification of five different health index classes based on the acquired dataset. So that a comparative study of performance parameter or classification score in each classifier was easily analysed. In this work SVM with hyper parameter tuning provided the best result, i.e. 98% accuracy or 2% error as compared to other classifiers.