为提升变压器故障预测的准确性,提出了一种基于灰狼(Grey Wolf Optimization,GWO)算法优化支持向量机(Support Vector Machine,SVM)的变压器故障预测方法.采用GWO算法对SVM进行优化,建立了基于GWO-SVM变压器油中溶解特征气体预测模型,根据油中溶解特征气体随时间变化的特点,通过求取嵌入维数确定模型输入量.文章采用实际运行变压器的油中溶解气体分析(Dissolved Gas Analysis,DGA)数据进行仿真分析,并与其他预测方法对比,结果表明,GWO-SVM模型对H2预测平均相对误差和均方根误差分别为4.38％和9.48μL/L,预测精度高于其他方法.在变压器油中溶解特征气体含量预测的基础上,利用IEC三比值法进行变压器故障诊断,诊断结果与变压器实际故障一致,验证了变压器故障预测方法的实用性和有效性.
Transformer Fault Prediction Based on Grey Wolf Algorithm Optimizing Support Vector Machine
To improve the accuracy of transformer fault prediction,a method based on grey wolf optimiza-tion ( GWO) optimized support vector machine ( SVM) is proposed.The GWO algorithm was used to opti-mize SVM,and a prediction model for dissolved characteristic gases in transformer oil based on GWO-SVM was established.Based on the characteristics of dissolved characteristic gases in oil over time,the input of the model was determined by calculating the embedding dimension.Simulation analysis was conducted u-sing dissolved gas analysis ( DGA) data from actual operating transformers,and compared with other pre-diction methods.The results showed that the GWO-SVM model had an average relative error and root mean square error of 4.38％and 9.48μL/L for H2 prediction respectively,with higher prediction accuracy than other methods.On the basis of predicting the dissolved characteristic gas content in transformer oil,the IEC three ratio method was used for transformer fault diagnosis.The diagnostic results were consistent with the actual faults of the transformer,verifying the practicality and effectiveness of the transformer fault pre-diction method.
transformerfault predictionsupport vector machinegrey wolf optimization algorithmcharacteristic gas
万方数据
维普
