基于MPA优化PNN的电能质量扰动识别方法仿真  

Simulation of Power Quality Disturbance Identification Method Based on MPA Optimized PNN

作　　者：陈一镖 倪陈义 陈浩 李彬彬 CHEN Yi-biao;NI Chen-yi;CHEN Hao;LI Bin-bin(College of Mechanical and Electrical Engineering,Wenzhou University,Wenzhou Zhejiang 325035,China;School of Computer and Artificial Intelligence,Changzhou University,Changzhou Jiangsu 213164,China;Zhejiang Yide Technology Co.,Ltd,Wenzhou Zhejiang 325000,China)

机构地区：[1]温州大学机电工程学院,浙江温州325035 [2]常州大学计算机与人工智能学院,江苏常州213164 [3]浙江亿德科技有限公司,浙江温州325000

出　　处：《计算机仿真》2023年第6期107-113,共7页Computer Simulation

基　　金：温州市科技局项目(ZG2019019);温州市科技局项目(ZG2020015)。

摘　　要：电能质量扰动的识别精度直接影响着电网电能质量的治理手段及方法。为了解决复杂电网环境下的复合电能质量扰动识别问题,提出了一种基于改进小波阈值法消噪和海洋捕食者算法优化概率神经网络(MPA-PNN)的电能质量扰动识别方法。首先采用改进小波阈值法对8种典型的电能质量扰动信号进行消噪处理,并利用小波变换对消噪完成的信号进行多尺度分解,以其中3个区分度较为明显的维度能量构成输入特征向量,最后利用MPA优化PNN的平滑参数σ,完成电能质量扰动信号识别模型的训练。仿真结果表明,与单一PNN、改进小波阈值法-PNN、改进小波阈值法-GA-PNN等方法进行比较,改进小波阈值法-MPA-PNN方法可以有效降低噪声影响,在识别精度及模型优化速度方面均有一定的提升。The identification accuracy of power quality disturbances has a significant impact on the power quality management in the power grid.In order to solve the problem of identifying composite power quality disturbance in a complex power grid,this paper proposes an identification method of power quality disturbance based on improved wavelet threshold method and marine predator algorithm optimized probabilistic neural network(MPA-PNN).Firstly,the improved wavelet thresholding method was used to denoise eight typical power quality disturbance signals,and decomposed signals were decomposed at multiple scales using wavelet transform.The input feature vector,which consists of the three most distinctive energy dimensions,was used to train the identification model,optimizing smoot-hing parameter o by MPA method.The simulation results show that,compared with single PNN,improved wavelet thresholding-PNN,and improved wavelet thresholding-GA-PNN,the improved wavelet thresholding-MPA-PNN method can effectively reduce the influence of noise and has an improvement in identification accuracy and model op-timization speed.

关 键 词：电能质量扰动 小波变换 海洋捕食者算法 概率神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]