应用多分类多核学习支持向量机的变压器故障诊断方法

A Fault Diagnosis Method for Power Transformer Based on Multiclass Multiple-kernel Learning Support Vector Machine

提出一种基于多分类多核学习支持向量机的变压器故障诊断方法,相对于传统的2分类支持向量机,该方法有如下特点:算法针对单一的优化目标函数求解,只需设计1组参数,降低了支持向量机在解决多类问题中模型构造和参数选择的难度;核函数是多个基核函数的组合,提高了分类的精度;将模型分解为2个凸优问题进行求解,问题的复杂度低,求解速度快。诊断实例表明,该方法能保证较高的诊断准确率,具有较好的实用性和推广性。

A novel support vector machine （SVM）, i.e. multiclass multiple-kernel learning support vector machine （MMKL-SVM）, for the fault diagnosis of power transformers is proposed in this paper. Unlike traditional SVM that may fail under some circumstances, the fault diagnosis method based on MMKL-SVM has some good theoretical properties, e.g. it only deals with a simple objective function, and the classification results can be obtained by dil：ect calculation on the basis of a simple decision function; it can conduct calculation with an optimal kernel function composed of linear combinations of basic kernels, further boosting the overall performance; the solutions for it can be efficiently gained by iteratively solving two convex optimization functions with a low computation cost and high speed. Diagnosis test results show that the MMKL-SVM method has high classification accuracy, which proves its effectiveness and usefulness.