基于小波-隐马尔可夫的波形异常扰动类型识别研究  

Research on Recognition of Waveform Abnormal Disturbance Types Using Wavelet-Hidden Markov Models

作　　者：苏寅生[1] 李智勇[1] 刘春晓[1] 李斌[1] 吴云亮[1] 张艳[2] SU Yinsheng;LI Zhiyong;LIU Chunxiao;LI Bin;WU Yunliang;ZHANG Yan(Power Dispatching and Control Center of China Southern Power Grid,Guangzhou 510623,Guangdong,China;School of Electronics and Information Engineering,Anhui University,Hefei 230601,Anhui,China)

机构地区：[1]中国南方电网电力调度控制中心,广东广州510623 [2]安徽大学电子信息工程学院,安徽合肥230601

出　　处：《电网与清洁能源》2021年第4期53-59,共7页Power System and Clean Energy

基　　金：国家自然科学基金青年科学基金项目(71601001);南网科技项目(适应南方电网电力市场的年、月安全校核系统生产采购)。

摘　　要：针对波形异常扰动的识别问题,提出一种基于小波-隐马尔可夫的分类方法,利用小波变换提取波形异常扰动特征。波形异常扰动利用小波变换被分解成多分辨率小波域,其小波系数由HMM构建模型。基于此模型,结合最大似然实现波形异常扰动的分类识别,并在7200 V配电线上的实际波形异常扰动数据进行分类,并通过后处理对结果进行调整。实验结果表明,该方法在电力工业的507个真实波形异常扰动事件分类的正确率为99%。The Wavelet-based HMM is designed to recognize waveform abnormal disturbances,where the wavelet transform is adopted to extract waveform abnormal disturbance features.The waveform abnormal disturbance is decomposed into multi-resolution wavelet domains,and the wavelet coefficients are modeled by the HMM.Based on this modeling,the maximum likelihood classification is applied to classify actual waveform abnormal disturbance data recorded on a 7200 V distribution line,and the result is tuned by post-processing.The experimental results show that the method has a correct rate of 99%for the classification of 507 real waveform abnormal disturbance in the power industry.

关 键 词：波形异常扰动 类型识别 小波 隐马尔可夫模型 

分 类 号：TM933[电气工程—电力电子与电力传动]