基于改进PSO对卷积层核数量优化的电能质量扰动分类  

Classification of power quality disturbances based on improvedPSO to optimizethe number of convolution cores

作　　者：程志友 姜帅[1] 胡杰 汪德胜 CHENG Zhiyou;JIANG Shuai;HU Jie;WANG Desheng(School of Internet,Anhui University,Hefei 230601,China;Power Quality Engineering Research Center,Ministry of Education,Anhui University,Hefei 230601,China)

机构地区：[1]安徽大学互联网学院,安徽合肥230601 [2]安徽大学教育部电能质量研究中心,安徽合肥230601

出　　处：《电工电能新技术》2023年第11期40-49,共10页Advanced Technology of Electrical Engineering and Energy

基　　金：国家自然科学基金项目(61672032);安徽省科技重大专项(18030901018);安徽省自然科学基金项目(2108085QE237)。

摘　　要：电能质量扰动分类方法目前存在人工提取特征困难、训练模型参数冗余等缺点,本文针对以上问题提出一种将改进的粒子群优化算法和卷积神经网络相结合的电能质量扰动类型分类方法。首先将差分预处理过的一维扰动波形送入卷积神经网络进行特征提取,接着利用改进的粒子群算法对卷积层中的卷积核数量矩阵进行自动迭代,通过该算法减少模型冗余的参数,提高模型的扰动分类效果。仿真和实测结果表明,基于改进粒子群优化卷积神经网络(IPSO-CNN)方法具有扰动类型分类准确度高、损失率低且模型训练参数更少的优点,是一种有效的电能质量扰动分类方法。At present,the power quality disturbance classification method has some shortcomings,such as difficulty in manual feature extraction and redundancy of training model parameters.In this paper,a power quality disturbance classification method combining improved particle swarm optimization algorithm and convolutional neural network is proposed to solve the above problems.First,the differential pretreated one-dimensional disturbance waveform is sent to the convolutional neural network for feature extraction.Then the improved particle swarm optimization algorithm is used to automatically iterate the number matrix of convolution kernels in the convolution layer.Through this algorithm,the redundant parameters of the model are reduced,and the disturbance classification effect of the model is improved.The simulation and measurement results show that IPSO-CNN method has the advantages of high accuracy,low loss rate and fewer model training parameters,and is an effective power quality disturbance classification method.

关 键 词：电能质量 扰动类型分类 粒子群优化 线性权值递减 卷积神经网络 

分 类 号：TM711[电气工程—电力系统及自动化]