基于最小二乘支持向量机和小波神经网络的电力线通信信道噪声建模研究

Noise Modeling for Power Line Communication Channel Using the LS-SVM and Wavelet Neural Networks

电力线通信是智能电网中的一种重要通信方式,电网中噪声干扰复杂,建立电力线通信信道噪声模型对于深入研究智能电网中低压电力线通信性能至关重要。针对低压电力线通信信道噪声特性,分别提出基于最小二乘支持向量机(LS-SVM)模型和小波神经网络模型在电力线信道噪声中的应用。为了验证并比较LS-SVM和小波神经网络模型对时变的低压电力线信道噪声建模的有效性,在室内和室外环境下对低压电力线通信信道的噪声进行测量,基于大量的测量数据,研究两个模型的准确度和效率。结果表明,两个噪声模型能够很好地仿真和适应时变的低压电力线通信信道,LS-SVM模型有更高的精度和更短的仿真时间。此外,提出的两个模型与传统的Markovian-Gaussian模型进行比较,结果表明,两个噪声模型有更高的精度和更低的复杂度,尤其是LS-SVM模型能够代替传统的Markovian-Gaussian模型,更适合用作低压电力线通信信道噪声发生器。该噪声模型的提出对研究在电力线通信系统和无线通信系统中内部和外部电磁源的电磁干扰有重要意义。

Power line communication(PLC)is an important communication way in smart grid.PLC channel noise is complex in such environment.It is essential to establish PLC channel noise model for in-depth study of the performance of low-voltage PLC in smart grid.This paper proposes two PLC channel noise models based on the least square support vector machine(LS-SVM)and wavelet neural network,respectively aiming at characterizing low-voltage PLC channel noise.To validate and compare their applicability to the time-variant PLC channels,noise measurements of low-voltage PLC channels in indoor and outdoor scenarios were carried out,the accuracy and efficiency of two models were studied based on large amount of measurement data.The results show that both models can simulate and adapt to the time-varying low-voltage PLC channels very well,while LS-SVM model has shorter simulation time and higher accuracy.Moreover,the proposed noise models are compared with traditional Markovian-Gaussian model.The results show that our proposed noise models have higher accuracy and lower complexity,especially the LS-SVM model is more appropriate to be applied as a noise generator instead of current Markovian-Gaussian model.The proposed models are helpful for investigating EMI on internal and external electromagnetic sources in the PLC and wireless.