The Estimation of Breakdown Voltage of Vegetable Oil using Support Vector Machine

Oil as an insulating medium is widely used in power apparatus and it is important to have knowledge about its breakdown characteristics. Support Vector Machine (SVM) can be a fruitful tool for breakdown voltage (BDV) estimation. In this work, the objective is to explore the application of Support Vector Machine (SVM) to estimate breakdown voltage. Experiments are carried out on vegetable oil to obtain its breakdown voltage using Weibull distribution. Further, the experiments are done for different electrode gap and electrode shape. The experimental results are fed to SVM to train, test and estimate BDV. At the breakdown condition, the electric field features are extracted from the simulated electric field distribution. The electric field features are processed using Principal Component Analysis (PCA) to obtain the significant electric field parameter that contributes to breakdown characteristics of the oil medium. The output of PCA is used to construct a classification model to predict breakdown voltage by Support Vector Machine (SVM). The optimum value of SVM parameters are obtained using grid search and K – fold cross validation technique. The tuned SVM model is used to estimate the breakdown voltage of the oil medium under different electrode gap and shape. It is seen that the estimated BDV fairly matches with the experimental results. This shows the significance of using electric field features to predict breakdown voltage using SVM.