组合核支持向量机在放电模式识别中的优化策略

传统的单一核函数SVM无法实现局部放电中的多特征空间向量的映射分类，且目前SVM大都采用不同尺度的径向基函数作为核函数，核的调整空间较为有限，无法针对不同特征空间的放电参数达到普适的效果。针对上述问题，本文提出一种基于多分组特征的组合核多分类SVM的局部放电识别方法。该方法首先完成多类特征空间在不同类型SVM核函数中的映射分类，再采用骨干粒子群(BBPSO)优化方法选取最佳核参数，并求解核函数权值系数，最终形成最优核函数组合分类模型。实验结果表明，该方法对多个特征空间数据具有普适性，且融合效果理想，分类精度高于误差反向传播神经网络(BPNN)和SVM识别方法。

Optimization Strategy Research on Combined-Kernel Support Vector Machine for Partial Discharge Pattern Recognition

Conventional single kernel SVM has natural defects on mapping multiple partial discharge(PD) feature spaces and classifying multiple PD types. Most popular SVM classifiers adopt RBF with different parameters as kernel functions that limited the adjustment space; moreover the universality to process multiple feature spaces is missed. Aiming at these problems, a grouped-feature based combined-kernel multiclass support vector machine(CKM-SVM) is proposed. Multiple PD feature spaces are constructed and mapped to different SVM kernel functions; then each kernel function is optimized via bare-bone particle swarm optimization(BBPSO), and then the weight coefficients for CKM-SVM model are calculated. Tests show that CKM-SVM performs good feature spatial fusion; additionally the recognition accuracy precedes BPNN and SVM.