基于参数优化支持向量机的矿用变压器故障诊断

Mine Transformer Fault Diagnosis Based on the Parameter Optimization of Support Vector Machine

支持向量机SVM(Support Vector Machines)理论在变压器故障诊断中得到了越来越多的应用,但在参数的优化选择方面还存在理论支持问题。为及时监测矿用变压器潜伏性故障和提高故障诊断效率,根据支持向量机原理,采用变压器故障时产生的氢气、甲烷、乙烷、乙烯、乙炔的浓度数据,提出了支持向量机的参数c和参数g的交叉验证算法,寻找最佳的参数c和参数g,利用优化后的参数对训练集进行训练,最终得到最佳的支持向量机模型,并对测试集进行分类,从而诊断出矿用变压器的故障类型。实例研究结果表明,该方法可行,故障诊断准确率为88.4615%,,具有较高的故障诊断准确率。

Support Vector Machines（SVM） theory is used widely in the transformer fault diagnosis, but there are prob- lems about parameter optimization selection which are not theoretical support. For monitoring mine transformer latent fault in time and improving the efficiency of fault diagnosis, Dissolved Gases Analysis in transformer oil is applied in this paper,training set and testing set are established the concentration data what are hydrogen, methane, ethane, ethylene, acetylene in the failure of the transformer, because transformer fault data are limited, parameters c and g are optimized which based on cross validation algorithm optimization and Support vector machine （SVM）. In the training set,parameters best c and best g are found by Cross validation algorithm, then using the optimized parameters for training of the training set, the best support vector machine （SVM） model is found, then classifying test set, so the transformer fault types are diagnosed. The transformer fault diagnosis example analysis results show that the method is feasible and effective,fault diagnosis accuracy rate is 88.4615% ,it has the high accuracy of fault diagnosis.