Research On Quality Of Transmission Estimation And Optical Performance Monitoring Techniques In Optical Networks

In recent years,with the continuous rise of high-bandwidth services,backbone optical networks are evolving toward high-capacity,dynamically reconfigurable and transparent to meet the growing data transmission needs of users.Due to the lack of optical-electrical-optical regeneration function in the intermediate nodes of dynamically reconfigurable and transparent optical networks,physical impairments to optical signals during transmission will continue to accumulate,resulting in continuous deterioration of the quality of transmission(QoT)of optical signals,which cannot meet the service requirements at the destination node.Therefore,before deploying a new lightpath,the QoT of the lightpath to be deployed must be estimated to fully utilize the network physical layer resources and reduce the network margin.On the other hand,optical performance monitoring(OPM)is a technique used to evaluate optical network performance and constraints,especially physical impairments.Monitoring data from OPM enables optical network management systems to implement impairment-aware routing wavelength assignment algorithms for better QoT by considering real-time signal distortion in various links between source and destination nodes during routing wavelength assignment.In addition,the physical state of a particular channel provided by OPM helps to automatically negotiate the optimal configuration of network components between the two transmission ends for more flexible,scalable,reliable and efficient network operations.Based on the above background,this thesis conducts a series of researches around QoT estimation techniques and OPM techniques in optical networks for the related working principles,simulations and experimental validation,and the main innovations are as follows:(1)A machine learning-based QoT estimation technique is proposed,which uses the span number,span length,modulation format,bit rate,and channel incident optical power as the input feature vector of the machine learning model to estimate the QoT of the lightpath to be deployed.The performance of the K-nearest neighbor,logistic regression,and support vector machine based QoT estimation techniques is verified by building a wavelength division multiplexing transmission simulation system.The simulation results show that the support vector machine based QoT estimation technique has the best performance.(2)An optical signal-to-noise ratio(OSNR)monitoring technique based on Gaussian process regression is proposed,which utilizes a broadband,central wavelength-tunable optical bandpass filter to filter the signal and uses the collected optical power sequence as the input feature vector to achieve OSNR monitoring.Experimental validation was performed by building 9×32GBaud PDM-16 QAM and 9×32GBaud PDM-QPSK/16QAM/64 QAM transmission systems.Experimental results show that the OSNR monitoring technique can achieve high accuracy and modulation-format-transparent OSNR monitoring with robustness to chromatic dispersion,polarization mode dispersion,nonlinear effects,and cascade filtering effects,and it is not necessary to provide any prior information about the transmission distance,incident optical power,and the number of wavelength selective switches experienced of the signal.Besides,this technique has the advantage of being low-cost and suitable for intermediate node monitoring.(3)Based on the(2)above,an OPM technique based on multi-task artificial neural network is further proposed,which similarly utilizes a broadband,central wavelength-tunable optical bandpass filter for filtering and uses the collected optical power sequence as the input feature vector to achieve OSNR monitoring,incident optical power monitoring,and baud rate identification simultaneously.Experimental verification is carried out by building 9×10GBaud QPSK and 9×32GBaud PDM-16 QAM transmission systems.Experimental results show that the OPM technique uses one multi-task artificial neural network instead of three single-task models to achieve the simultaneous monitoring of three OPM parameters,which greatly reduces the complexity.Furthermore,the OPM technique can achieve high accuracy OSNR and incident optical power monitoring as well as high accuracy baud rate identification without providing any priori information about transmission distance of the signal,and has the advantages of transparency to modulation format,robustness to chromatic dispersion,polarization mode dispersion and nonlinear effects,low cost,and applicability to intermediate nodes of the network.