摘要
针对电能质量扰动信号在强噪声下识别准确率低的问题,提出了一种基于小波降噪和深度学习的电能质量扰动信号识别方法。在信号输入前,采用分层自适应阈值函数HATF(hierarchical adaptive threshold function)降噪法对信号降噪处理;接着在卷积神经网络CNN(convolutional neural network)扰动分类方法之上,对网络加入扰动信号的时序性识别,构建了E-CNN(enhanced-conventional neural network)的融合网络模型提高对含噪信号的识别准确率。仿真结果显示,与信号未去噪时的卷积神经网络相比,引入降噪后的融合网络模型在强噪声环境下的识别准确率依然可以达到98.40%,可以有效分类6种单一扰动信号和4种复合扰动信号。
Aimed at the problem of low recognition accuracy of power quality disturbance signals under strong noise,a power quality disturbance signal recognition method is proposed based on wavelet denoising and deep learning.Before the signal is input,the hierarchical adaptive threshold function(HATF)denoising algorithm is used to reduce the noise in the signal.Then,based on the disturbance classification method of convolutional neural network(CNN),the timeseries recognition of disturbance signals is added to the network,and an enhanced-conventional neural network(E-CNN)fusion network model is constructed to improve the recognition accuracy of noisy signals.Simulation results show that compared with the CNN when the signal is not denoised,the fusion network model with noise reduction has a recognition accuracy of as high as 98.40%under strong noise,and it can effectively classify six kinds of single disturbance signal and four kinds of composite disturbance signals.
作者
刘烨
程杉
王瑞
左先旺
徐敬伟
LIU Ye;CHENG Shan;WANG Rui;ZUO Xianwang;XU Jingwei(College of Electrical Engineering&New Energy,China Three Gorges University,Yichang 443002,China;Yichang Key Laboratory of Intelligent Operation and Security Defense of Power System(China Three Gorges University),Yichang 443002,China;Dongying Power Supply Company,State Grid Shandong Electric Power Company,Dongying 257000,China)
出处
《电力系统及其自动化学报》
CSCD
北大核心
2022年第11期17-23,共7页
Proceedings of the CSU-EPSA
基金
国家自然科学基金资助项目(51607105)。
关键词
电能质量扰动信号
小波降噪算法
分层阈值函数
深度学习
卷积神经网络
power quality disturbance signal
wavelet denoising algorithm
hierarchical threshold function
deep learning
convolutional neural network(CNN)
