Prediction of DC Corona Onset Voltage for Rod-Plane Air Gaps by a Support Vector Machine

This paper proposes a new method to predict the corona onset voltage for a rod- plane air gap, based on the support vector machine （SVM）. Because the SVM is not limited by the size, dimension and nonlinearity of the samples, this method can realize accurate prediction with few training data. Only electric field features are chosen as the input; no geometric parameter is included. Therefore, the experiment data of one kind of electrode can be used to predict the corona onset voltages of other electrodes with different sizes. With the experimental data obtained by ozone detection technology, and experimental data provided by the reference, the efficiency of the proposed method is validated. Accurate predicted results with an average relative less than 3% are obtained with only 6 experimental data.