基于VMD和IAO-SVM的电压暂降源识别方法

Voltage Sag Source Identification Method Based on VMD and IAO-SVM

针对支持向量机(support vector machine,SVM)的惩罚因子、核函数参数选择困难和天鹰优化(aquila optimizer,AO)算法在寻优时容易陷入局部最优解的问题,利用改进的天鹰优化(improved aquila optimizer,IAO)算法对SVM的惩罚因子和核函数参数进行寻优,构建IAO-SVM分类器,利用变分模态分解(variational mode decomposition,VMD)提取电压暂降源信号三相电压的特征向量,并进行归一化处理之后输入到构造好的IAO-SVM分类器中对样本进行训练与识别,并与K近邻、极限学习机、SVM和AO-SVM这4种分类器进行对比。仿真结果表明,在对8种电压暂降源信号分别加入0 dB、10 dB、20 dB、30 dB、40 dB、50 dB和60 dB的高斯白噪声情况下,IAO-SVM分类器识别的准确率分别为99.5%、94%、99.25%、100%、99.25%、98.5%和97.25%,其识别准确率最高,验证了在对信号加入不同的高斯白噪声时,IAO-SVM分类器均具有较高的识别准确率和抗噪声能力,有助于解决电压暂降源的分类问题。

Aiming at the difficulty of selecting the penalty factors and kernel function parameters of support vector machine(SVM)and the problem that the aquila optimizer(AO)algorithm is easy to fall into the local optimal solution during optimization,the improved aquila optimizer(IAO)algorithm is used to optimize the penalty factors and kernel function parameters of SVM,and the IAO-SVM classifier is constructed.The feature vector of three-phase voltage of voltage sag source signal is extracted by using variational mode decomposition(VMD).After normalization processing,the feature vector is input into the constructed IAO-SVM classifier to train and identify the samples,and compared with K-nearest neighbor classifier,extreme learning machine classifier,SVM classifier and AO-SVM classifier.The simulation results indicate that when the Gaussian white noise of 0 dB,10 dB,20 dB,30 dB,40 dB,50 dB and 60 dB is added to eight kinds of voltage sag source signals,the recognition accuracy of the IAO-SVM classifier is 99.5%,94%,99.25%,100%,99.25%,98.5%and 97.25%respectively.The recognition accuracy is the highest among the five classifiers.It is verified that the IAO-SVM classifier has higher recognition accuracy and anti-noise ability when adding different Gaussian white noise to the signal,which is helpful to solve the problem of voltage sag source classification.