基于油色谱超立方映射的电力变压器缺陷援例诊断模型

文中提出了一种援例缺陷诊断模型,针对油色谱特征气体数据的分布特点提出了归一化超立方映射方法,将油色谱数据映射到可直接应用的超立方空间域中;同时,针对性地提出援例相似度算法和基于计权选举的诊断结果判定方法。并通过仿真实验确定了模型中参数的选取和优化。该模型在案例库交叉验证中表现出较高的正确率,平均正确率达到88.53%,高于现有BP神经网络和支持向量机技术,能正确诊断运行中充油设备的缺陷,在工程上具有重要的实际应用价值。

Case Based Power Transformer Defeats Diagnose Model Using Hypercube Mapping of Oil Chromatography

A case based defect diagnosis model is proposed in this paper. A normalized hypercube mapping method is proposed according to the distribution characteristics of oil chromatogram data. Oil chromatogram data in the hypercube space domain can be applied for diagnosing directly. Meanwhile, the case similarity degree method and the judging method of diagnosis result based on weighting election are put forward, and the selection and optimization of parameters in the model are confirmed by simulation experiments. The model shows high correct rate in the cross validation of the case database. The average accuracy rate was 88.53%, higher than the existing BP neural network and support vector machine technology. It can diagnose the defects of the oil-immersed equipment. It is verified that the model proposed in this paper has significant practical application value in engineering.