Research On Low-Voltage Power Line Channel Characteristics And Its Noise Model

Low-voltage power line communication(PLC)technology is widely used in various fields with its unique advantages.In order to improve the anti-interference ability of PLC,high-precision modeling of its noise is necessary.The high-precision modeling of low voltage power line background noise is presented in this paper.The main research contents and works in this paper are as follows:Firstly,this paper summarizes the noise characteristics of PLC channels.The colored background noise and narrowband noise are simulated by MATLAB respectively,which can be used as noise source in the simulation models in the rest of this thesis.Secondly,the colored background noise is modeled by wavelet packet transform and peak-type Markov chain respectively based on different wavelet basis functions,then we calculate their PSDs and corresponding root-mean-square errors(RMSEs)to determine the optimal wavelet basis.Thirdly,this paper gives a novel background noise model based on the wavelet neural network(WNN).The colored background noise and narrowband noise are modeled by the WNN respectively,then we compare its modeling result with the traditional model.Fourthly,a novel background noise model based on the least square support vector machine(LS-SVM)has been studied since WNN has the disadvantage of difficult decision for the number of hidden layer nodes.The LS-SVM models the colored background noise and narrowband noise respectively,we compare its modeling result with the traditional model too.The results show that modeling accuracy of wavelet-markov chain using the Daubecies wavelet basis function is the highest;both the output noise waveforms and its PSDs simulated by the WNN and LS-SVM models have good agreements with the test noise and its PSDs shown as smaller RMSEs than the traditional model.In conclusion,the Daubecies wavelet is the optimal wavelet basis function of the wavelet packet transform and peak-type Markov chain;both the WNN and the LS-SVM are effective to model the background noise,their modeling accuracies are higher than the traditional model.