基于神经网络与改进马尔可夫链中压背景噪声建模研究

Research on Modeling of Middle Pressure Background Noise Based on Neural Network and Improved Markov Chain

在中压电力线通信中,信道噪声构成复杂,需根据不同类型噪声单独分析建模。针对一段特定中压线路的背景噪声,提出了一种基于小波包变换的噪声模型,将得到的小波包系数分别进行神经网络训练和改进马尔可夫链计算转移概率矩阵,得到新的小波包系数重构噪声信号,并进行仿真验证及去噪,同时将2种方法与传统直接神经网络训练比较分析。结果表明,基于改进马尔可夫链方法所建噪声比传统马尔可夫链方法更加准确,基于小波包变换的神经网络方法所建噪声与原噪声相似度更高,去噪效果更好,且优于传统神经网络训练方法,为进一步研究中压电力线通信提供了可行性方案。

In medium voltage power line communication,the channel noise composition is complex,and it needs to be analyzed and modeled separately according to different types of noise.Aiming at the background noise of a specific medium voltage line,a noise model based on wavelet packet transformation is proposed.The obtained wavelet packet coefficients are respectively trained by neural networks and the transfer probability matrix of the improved Markov chain is obtained to reconstruct the noise signals.The simulation verification and denoising are carried out,and the two methods are compared with the traditional direct neural network training.The results show that the noise built by the improved Markov chain method is more accurate than the traditional Markov chain method.The noise built by the neural network method based on the wavelet packet transform is more similar to the original noise,and the noise reduction effect is better than that of the traditional neural network training method,which provides a feasible scheme for the further study of medium voltage power line communication.