基于小波神经网络的低压电力线背景噪声建模  

Research on Modeling of Low-Voltage Power Line Background Noise by Wavelet Neural Networks

作　　者：索超男 赵雄文[1] 张慧[1] 卢文冰[1] 

机构地区：[1]华北电力大学电气与电子工程学院,北京102206

出　　处：《中国电力》2017年第9期89-94,共6页Electric Power

基　　金：国网重庆市电力公司资助项目(KH15010158)~~

摘　　要：为增强电力线通信的抗噪能力,针对低压电力线信道噪声中的有色背景噪声和窄带噪声,提出一种基于小波神经网络的建模方法。首先,将背景噪声进行小波神经网络建模,对比所建模型输出噪声与测试噪声的时域波形及功率谱密度,计算两者功率谱密度的均方根误差;然后,对同一组背景噪声分别进行基于传统的小波马尔科夫链和小波神经网络的建模,并计算2种模型输出噪声与测试噪声的功率谱密度及其均方根误差。仿真结果表明,小波神经网络输出噪声与测试噪声的时域波形及功率谱密度均有着较一致的变化趋势,因此小波神经网络对低压电力线信道背景噪声的建模是有效的,对宽带噪声的建模效果更好。In order to improve anti-interference ability of power line communications, a new modeling method for colored background noise and narrowband noise based on wavelet neural network is proposed. Firstly, the background noise is modeled by wavelet neural network. The output noise waveforms and power spectrum densities （PSD） obtained from model are compared with test noise by calculating root mean square error（RMSE）. Moreover, the background noise is also modeled by traditional wavelet packet transform and peak typed Markov chain. RMSEs of PSDs before and after modeling are also calculated. Simulation results show that both output noise waveforms and PSD obtained by proposed model have good agreements with test noise. The RMSE is smaller than the value generated using wavelet packet transform and peak typed Markov chain. Therefore, the proposed wavelet neural network model is effective in modeling background noise, especially for the wideband colored background noise.

关 键 词：有色背景噪声 窄带背景噪声 小波神经网络 小波马尔科夫链 

分 类 号：TN913.6[电子电信—通信与信息系统]