Open AccessModelling and comparison for low‐voltage broadband power line noise using LS‐SVM and wavelet neural networks
Author(s) -
Zhao Xiongwen,
Zhang Hui,
Li Liang,
Lu Wenbing,
Ding Yi,
Liu Junyu
Publication year - 2019
Publication title -
iet communications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.355
H-Index - 62
eISSN - 1751-8636
pISSN - 1751-8628
DOI - 10.1049/iet-com.2018.5282
Subject(s) - computer science , noise (video) , support vector machine , autoregressive model , gaussian noise , power line communication , artificial neural network , wavelet , voltage , algorithm , artificial intelligence , electronic engineering , power (physics) , mathematics , engineering , statistics , electrical engineering , physics , quantum mechanics , image (mathematics)
This study is to construct the autoregressive models for the low‐voltage broadband power line communication (PLC) channel noise by two machine learning algorithms, namely the least square support vector machine (LS‐SVM) and wavelet neural networks. The main work is to compare the two classical machine learning algorithms and also compare them with the traditional Markovian–Gaussian method. To verify their availability and ability to adapt to the time‐variant PLC channels, noise measurements for low‐voltage PLC channels in indoor and outdoor scenarios are carried out. The accuracy and efficiency of the two models are studied and compared based on a large amount of measurement data. The results show that both of the noise models can simulate and adapt to the time‐variant low‐voltage broadband PLC channels very well. The LS‐SVM model is found to have shorter simulation time and higher accuracy. Moreover, the proposed noise models are also compared with the traditional Markovian–Gaussian model. The results show that both the proposed noise models exhibit higher accuracy and lower complexity, especially that the LS‐SVM is more appropriate to be applied as a noise generator in PLC link and network level simulations instead of the current Markovian–Gaussian model.