组合神经网络在变压器故障诊断中的应用

Application of Compound Networks in Fault Diagnosis of Power Transformer

针对油中溶解气体分析数据的归一化预处理,利用可靠性数据分析特征气体浓度和累积频率的概念,提出了两种新的归一化方法:特征浓度归一化法和混合归一化法,引入Fisher准则函数来评价两种预处理方法的效果。检验结果表明,这两种归一化的数据预处理方法可获得类间均值差值较大、类内离散度小的效果。运用不同的归一化预处理方法对故障变压器的色谱数据进行处理后作为训练样本,对CP算法的组合神经网络进行训练。检验样本的诊断结果表明,新的归一化预处理方法能够提高网络诊断的准确率。

Using the concepts of typical gas＇s concentration and cumulative frequency in analysis of the reliability data for dealing with the pretreatment of data of dissolved gas analysis （DGA）, two new normalized methods which named characteristic normalization and mix normalization were presented in this paper. The Fisher rule to evaluate the results of the two pretreatment methods was also introduced. The evaluation of the results indicated that both of the two data pretreatment methods could achieve the purpose of big difference in the value of mean between classes and small difference in dispersion of a class. The DGA data of the failure transformers were treated by different normalization methods as the training samples, and then the samples were trained in the compound neural networks which use the CP algorithm. The diagnosis results of the test samples indicated that the new methods may help to improve the precision of network diagnosis.