基于改进的杜鹃搜索算法优化支持向量机的10kV并联电容器组故障诊断和预警研究

Fault Diagnosis and Early Warning Study of 10kV Parallel Capacitor Bank Based on Improved Cuckoo Search Algorithm Optimizing SVM

为提高电力电容器组故障诊断的精度,针对支持向量机(Support Vector Machine,SVM)预测结果易受惩罚因子c和核函数参数g参数选择的影响,为避免杜鹃搜索算法陷入局部最优,将自适应步长和最优解高斯变异引入杜鹃搜索算法,提出一种改进的杜鹃搜索算法优化支持向量机的10kV并联电容器组故障诊断和识别模型,实现10kV并联电容器组故障的高精度诊断和识别。实验结果表明,与GA_SVM、PSO_SVM和CSA_SVM相比,提出的算法ICSA_SVM可以有效提高电容器组故障诊断的准确率,具有收敛速度快的优点,为电容器组的诊断和识别提供新的方法和途径。

In order to improve the accuracy of fault diagnosis, for power capacitor set, the prediction results of support vector machine (Support Vector Machine, SVM) are easily affected by the selection of the penalty factor and the parameter parameter of the kernel function. In order to avoid the local optimum of the azalea search algorithm, the adaptive step length and the optimal solution Gauss variation are introduced to the azalea search algorithm. An improved Rhododendron search algorithm is developed to optimize the fault diagnosis and recognition model of the 10kV shunt capacitor bank for support vector machines (SVM), and the high precision diagnosis and recognition of the 10kV shunt capacitor bank fault is realized. The experimental results show that, compared with GA_SVM, PSO_SVM and CSA_SVM, the proposed algorithm ICSA_SVM can effectively improve the accuracy of fault diagnosis of capacitor banks, and has the advantages of fast convergence speed, and provides a new method and way for the diagnosis and recognition of capacitor banks.