基于BO-CatBoost的变压器故障诊断

Transformer Fault Diagnosis Based on BO-CatBoost

针对现有的利用经验判断变压器故障准确率过低的问题,提出一种基于BO-CatBoost算法的模型进行变压器故障诊断。根据油中溶解气体分析(DGA)数据,结合编码比值方法构建11个新的变压器故障特征,并对所有特征进行标准化处理;将标准化后的特征作为CatBoost故障诊断模型的输入。所采用的贝叶斯优化算法能充分利用历史调参信息对模型中的多个超参数进行寻优,避免陷入局部最优。仿真结果表明:在引入了贝叶斯参数寻优后,CatBoost算法的诊断准确率提高到了92.93%,优于贝叶斯优化支持向量机和随机森林模型,该方法在变压器故障诊断中有良好的应用前景。

Aiming at the existing empirical method’s poor accuracy in judging transformer faults,a method based on BO-CatBoost algorithm model for transformer fault diagnosis was proposed.Based on analyzing the dissolved gas(DGA)in oil,11 new transformer fault feature vectors were constructed in combination with the coding ratio method,and all feature vectors were standardized and then were input into the CatBoost fault diagnosis model.The adopted Bayesian optimization algorithm can make full use of historical tuning information to optimize multiple hyperparameters in the model so as to avoid the falling into local optimality.The simulation results show that,after the introduction of Bayesian parameter optimization,the diagnostic accuracy of the Cat Boost algorithm can be improved to 92.93%,which outperforms both the Bayesian optimization support vector machine and the random forest model.This method has good application prospects in transformer fault diagnosis.