基于支持向量机的棒-板空气间隙击穿电压预测方法及其应用

Breakdown Voltage Prediction Method of Rod-Plane Air Gaps Based on Support Vector Machine and Its Applications

空气间隙的击穿电压与放电起始前的电场分布特征存在多维非线性关系。为了实现空气间隙的击穿电压预测,以电场特征集作为输入,以间隙在加载电压下是否击穿作为输出,采用支持向量分类机建立击穿电压预测模型。针对极不均匀电场空气间隙的击穿特性受电晕影响的问题,提出两种修正方法:通过增加受电晕影响的训练样本数据,提高预测模型的泛化性能;或基于"电晕云"的思想进行二次电场计算及特征量提取,对预测模型的输入特征进行修正。采用修正后的模型对极不均匀电场下棒-板间隙的工频击穿电压及棒-板长空气间隙的操作冲击放电电压进行预测,预测值与试验值吻合良好。该方法有利于减少试验次数,降低试验成本。

There is a multidimensional nonlinear relationship between the breakdown voltage of air gap and the electric field distribution features before discharge inception. In order to predict the breakdown voltage of air gap,taking the electric field feature set as input,and whether the gap will breakdown under the applied voltage as output,the breakdown voltage prediction model was established by support vector classifier（ SVC）. Since the breakdown characteristics of air gap in an extremely non-uniform electric field are affected by corona, two correction methods were proposed. The training sample data affected by corona were added to improve the generalization performance of the prediction model. Alternatively, based on the idea of ＂corona cloud ＂, the electric field distribution was calculated secondly and the electric field features were extracted again,so as to modify the input features of the prediction model. The modified model was used to predict the power frequency breakdown voltage of rod-plane gap in extremely non-uniform electric field and the switching impulse discharge voltage of long rod-plane air gap. The predicted results are in good agreement with the experimental values. This method is conducive to decrease test work and reduce cost.