Font Size: a A A

Application Of Machine Learning Technology In Coherent Optical OFDM System

Posted on:2021-09-05Degree:MasterType:Thesis
Country:ChinaCandidate:H J ZhangFull Text:PDF
GTID:2518306308472714Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
With the advent of the era of big data,large-capacity,high-speed,long-distance transmission systems are bound to be the mainstream of the next time.Coherent Optical Orthogonal Frequency Division Multiplexing(CO-OFDM)technology is a joint product of coherent optical technology and orthogonal frequency division multiplexing(OFDM)technology.It has the advantages of both and will play an important role in the new generation of high-speed optical networks.In recent years,machine learning technology has achieved very successful applications in the field of computer science,and naturally it will play an import role in the field of optical communications.In CO-OFDM,because the signal peak-to-average ratio(PAPR)is relatively large,under the effect of the non-linear characteristics of the components,system performance will be seriously deteriorated.Therefore,in this paper,theoretical analysis and simulation research are performed for the nonlinear compensation of coherent optical OFDM,First,a VPI simulation software and MATLAB were used to build a CO-OFDM system based on the CO-OFDM principle.Then proposed a non-linear compensation scheme based on multilayer perceptron neural network(MLP-ANN)technology.In a hexadecimal quadrature amplitude modulation(16-QAM)CO-OFDM system,compared to a linear equalizer,the MLP-ANN nonlinear equalizer(NLE)can improve the Q value by 1.4dB,0.6dB through single-mode fibre at 600km and 2000km,respectively.Meanwhile it proved strong robustness.In addition,in CO-OFDM,due to channel fading,the sub-channel state conditions are different.The traditional power allocation schemes either blindly equal all subchannels without considering the specific status conditions of different subchannels,or ignores the nonlinear effects and interactions between subcarriers,and assume that the signal-to-noise ratio and transmit power of the subcarriers have a simple linear relationship.This inevitably affects the transmission performance of the channel,so that the transmission capacity of the system cannot be maximized.Therefore,this paper also conducts theoretical and simulation research on coherent optical OFDM subcarrier power allocation.Simulation results show that the proposed power allocation scheme based on neural network and genetic algorithm can match more suitable power for each sub-channel than the traditional power allocation scheme,thereby improving the system's error performance.
Keywords/Search Tags:Coherent optical orthogonal frequency division multiplexing, neural network, nonlinear equalizer, adaptive power allocation
PDF Full Text Request
Related items