低壓電力線信道特性與噪聲模型的研究
發(fā)布時間：2019-03-02 17:39
【摘要】：低壓電力線通信(Power Line Communication,PLC)技術具有無需重新布線等獨特的優(yōu)點,被廣泛應用于各個領域。噪聲是影響PLC通信系統(tǒng)可靠性的主要因素之一,它會惡化通信質量,甚至會造成整個通信過程的中斷。因此,研究電力線通信系統(tǒng)內噪聲的高精度模型對提高該系統(tǒng)的抗噪能力意義深遠。本文主要研究了低壓PLC信道中背景噪聲的高精度建模問題。相關研究內容與成果如下:1、著重介紹了電力線信道的噪聲特性。在MATLAB上分別對有色背景噪聲及窄帶噪聲進行仿真,得到其時域波形及功率譜密度(Power Spectrum Density,PSD),作為本論文后續(xù)噪聲建模問題研究的源數(shù)據(jù)使用。2、在對有色背景噪聲進行小波峰式馬爾科夫鏈建模時,研究了不同小波基函數(shù)對建模效果的影響。通過計算建模前后噪聲功率譜密度的均方根誤差確定了具有最高建模精度的小波基函數(shù)。3、給出一種基于小波神經(jīng)網(wǎng)絡的新型背景噪聲模型。對有色背景噪聲及窄帶噪聲分別進行小波神經(jīng)網(wǎng)絡建模,對比所建模型輸出噪聲與測試噪聲的時域波形及PSD,計算兩者功率譜密度的均方根誤差,并將該模型的建模效果與傳統(tǒng)的小波峰式馬爾科夫鏈模型相對比。4、針對小波神經(jīng)網(wǎng)絡具有隱層節(jié)點個數(shù)難以確定的缺點,給出一種基于LS-SVM的新型背景噪聲模型。對有色背景噪聲及窄帶噪聲分別開展基于LS-SVM模型的建模研究,對比所建模型輸出噪聲與測試噪聲的時域波形及功率譜密度,計算兩者功率譜密度的均方根誤差,并將該模型的建模效果與小波峰式馬爾科夫鏈模型進行對比,驗證LS-SVM模型的優(yōu)劣。研究結果表明,Daubecies、Biorthogonal和Haar小波基函數(shù)中,使用Daubecies小波基函數(shù)的小波峰式馬爾科夫鏈的建模精度最高;小波神經(jīng)網(wǎng)絡和LS-SVM模型輸出噪聲與測試噪聲的時域波形及功率譜密度均有著較一致的變化趨勢;兩種模型的建模誤差均小于小波峰式馬爾科夫鏈模型。綜上所述,Daubecies小波可選為有色背景噪聲進行小波峰式馬爾科夫鏈建模的最佳小波基函數(shù);小波神經(jīng)網(wǎng)絡和LS-SVM模型對背景噪聲的建模均是有效的,它們的建模精度均高于傳統(tǒng)的小波馬爾科夫鏈。
[Abstract]:Low voltage power line communication (Power Line Communication,PLC (low voltage power line communication) technology is widely used in various fields because of its unique advantages such as no re-wiring and so on. Noise is one of the main factors affecting the reliability of PLC communication system. It will worsen the communication quality and even cause the interruption of the whole communication process. Therefore, the study of the high-precision model of internal noise in power line communication system is of great significance to improve the anti-noise capability of the system. In this paper, the high-precision modeling of background noise in low-voltage PLC channel is studied. The related research contents and achievements are as follows: 1. The noise characteristics of power line channel are emphatically introduced. The colored background noise and narrow band noise are simulated on MATLAB, and their time domain waveforms and power spectral density (Power Spectrum Density,PSD) are obtained, which can be used as the source data for further research of noise modeling in this paper. In the modeling of colored background noise by wavelet peak Markov chain, the influence of different wavelet basis functions on the modeling effect is studied. The wavelet basis function with the highest modeling accuracy is determined by calculating the root mean square error of noise power spectral density before and after modeling. 3. A new background noise model based on wavelet neural network is presented. The colored background noise and narrow band noise are modeled by wavelet neural network, and the RMS error of power spectral density between the output noise and the test noise in time domain and the PSD, calculation are compared. The modeling effect of the model is compared with the traditional Markov chain model. 4. Aiming at the disadvantage that the number of hidden layer nodes in the wavelet neural network is difficult to determine, a new background noise model based on LS-SVM is proposed. The modeling of colored background noise and narrow band noise based on LS-SVM model is studied. The time domain waveform and power spectral density of output noise and test noise are compared, and the root mean square error of power spectral density of the two models is calculated. The modeling effect of this model is compared with that of wavelet peak Markov chain model, and the advantages and disadvantages of the LS-SVM model are verified. The results show that in Daubecies,Biorthogonal and Haar wavelet basis functions, the modeling accuracy of wavelet peak Markov chain using Daubecies wavelet basis function is the highest. The output noise of wavelet neural network and LS-SVM model is consistent with the time domain waveform and power spectral density of test noise, and the modeling error of the two models is smaller than that of wavelet peak Markov chain model. To sum up, Daubecies wavelet can be selected as the best wavelet basis function for wavelet peak Markov chain modeling with colored background noise. Both wavelet neural network and LS-SVM model are effective in modeling background noise, and their modeling accuracy is higher than that of traditional wavelet Markov chain.
【學位授予單位】：華北電力大學(北京)
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TN913.6
[Abstract]:Low voltage power line communication (Power Line Communication,PLC (low voltage power line communication) technology is widely used in various fields because of its unique advantages such as no re-wiring and so on. Noise is one of the main factors affecting the reliability of PLC communication system. It will worsen the communication quality and even cause the interruption of the whole communication process. Therefore, the study of the high-precision model of internal noise in power line communication system is of great significance to improve the anti-noise capability of the system. In this paper, the high-precision modeling of background noise in low-voltage PLC channel is studied. The related research contents and achievements are as follows: 1. The noise characteristics of power line channel are emphatically introduced. The colored background noise and narrow band noise are simulated on MATLAB, and their time domain waveforms and power spectral density (Power Spectrum Density,PSD) are obtained, which can be used as the source data for further research of noise modeling in this paper. In the modeling of colored background noise by wavelet peak Markov chain, the influence of different wavelet basis functions on the modeling effect is studied. The wavelet basis function with the highest modeling accuracy is determined by calculating the root mean square error of noise power spectral density before and after modeling. 3. A new background noise model based on wavelet neural network is presented. The colored background noise and narrow band noise are modeled by wavelet neural network, and the RMS error of power spectral density between the output noise and the test noise in time domain and the PSD, calculation are compared. The modeling effect of the model is compared with the traditional Markov chain model. 4. Aiming at the disadvantage that the number of hidden layer nodes in the wavelet neural network is difficult to determine, a new background noise model based on LS-SVM is proposed. The modeling of colored background noise and narrow band noise based on LS-SVM model is studied. The time domain waveform and power spectral density of output noise and test noise are compared, and the root mean square error of power spectral density of the two models is calculated. The modeling effect of this model is compared with that of wavelet peak Markov chain model, and the advantages and disadvantages of the LS-SVM model are verified. The results show that in Daubecies,Biorthogonal and Haar wavelet basis functions, the modeling accuracy of wavelet peak Markov chain using Daubecies wavelet basis function is the highest. The output noise of wavelet neural network and LS-SVM model is consistent with the time domain waveform and power spectral density of test noise, and the modeling error of the two models is smaller than that of wavelet peak Markov chain model. To sum up, Daubecies wavelet can be selected as the best wavelet basis function for wavelet peak Markov chain modeling with colored background noise. Both wavelet neural network and LS-SVM model are effective in modeling background noise, and their modeling accuracy is higher than that of traditional wavelet Markov chain.
【學位授予單位】：華北電力大學(北京)
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TN913.6
【參考文獻】
相關期刊論文 前10條
1 葉君;孫洪亮;王毅;文成亮;;一種適用于低壓電力線通信信道的背景噪聲建模方法[J];重慶郵電大學學報(自然科學版);2015年06期
2 姚海燕;張靜;留毅;潘姝;徐貝;李題印;周念成;;基于多尺度小波判據(jù)和時頻特征關聯(lián)的電纜早期故障檢測和識別方法[J];電力系統(tǒng)保護與控制;2015年09期
3 邵天宇;王立欣;白瑾s，
本文編號：2433299
本文鏈接：http://www.sikaile.net/kejilunwen/xinxigongchenglunwen/2433299.html
