基于ESMD和SSA-PNN的电能质量扰动信号识别分类

PQD signal recognition and classification based on ESMD and SSA-PNN

针对传统概率神经网络(PNN)分类器中平滑因子依靠人工经验赋值,导致电能质量扰动信号识别分类精度不高的问题,文中提出一种基于极点对称模态分解和麻雀搜索算法优化概率神经网络(SSA-PNN)的电能质量扰动信号识别分类方法。首先,添加含噪声的电能质量扰动信号;其次,利用极点对称模态分解算法对扰动信号进行分解,得到不同频率的本征模态函数;再根据原信号与本征模态函数分量的相关系数选取有代表性的分量,对代表性分量提取能量值和样本熵并将其作为特征向量;最后,创新性地利用麻雀搜索算法优化概率神经网络中的平滑因子,寻找最优平滑因子构建SSA-PNN分类器,将特征向量输入传统PNN分类器和SSA-PNN分类器中进行识别分类。仿真结果表明,相较于传统PNN分类器,SSAPNN分类器的准确率较高,可为电能质量扰动信号识别分类提供一种新的解决方案。

In allusion to the problem that the smoothing factor of traditional probabilistic neural network(PNN)classifier is assigned by means of the artificial experience,which leads to the low accuracy of power quality disturbance(PQD)signal recognition and classification,a method of PQD signal recognition and classification based on extreme-point symmetric mode decomposition(ESMD)and sparrow search algorithm optimization probabilistic neural network(SSA-PNN)is proposed.The PQD signal with noise is added,and the ESMD algorithm is used to decompose the disturbance signal to obtain the intrinsic mode function of different frequencies.The representative components are selected according to the correlation coefficients between the original signal and the intrinsic mode function components,and the energy value and sample entropy are extracted as eigenvectors for representative components.The SSA is innovatively used to optimize the smoothing factor in PNN,and the optimal smoothing factor is found to build the SSA-PNN classifier.The feature vector is input into the traditional PNN classifier and SSA-PNN classifier for recognition and classification.The simulation results show that in comparison with the traditional PNN classifier,SSA-PNN classifier has higher accuracy and can provide a new solution for PQD signal recognition and classification.