Transmission Performance Estimation Of Open Submarine Cable System Based On Machine Learning

Optical fiber is the most popular used communication medium in the space of communication.As an important industry pillar of sustainable development strategy in China,submarine optical cable communication has the following advantages:communication quality is efficient and reliable,anti-interference ability is very strong,hidden performance is strong,and safety effect is obvious.In recent years,the use of submarine optical cable as a means of communication has become more and more popular,and its line laying has reached millions of kilometers.With the continuous development of information technology,the submarine optical cable communication system has a very high and irreplaceable position.For open submarine optical cable transmission system,different manufacturers can access through each shore station,so the transmission link of submarine optical cable will be greatly improved.But in the same submarine optical fiber link system,such as baud rate of transmission signal,input optical power and modulation format of signal,etc.,the relevant parameters are very different.Therefore,it is very important to evaluate the performance of the system under different baud rate,power,modulation format and other parameters.This research is mainly based on the analysis of the research results at home and abroad.Based on the full understanding of the characteristics of optical fiber communication transmission system and the structure of open submarine optical cable system,the transmission performance evaluation model of open submarine optical cable system based on machine learning is designed and studied,so as to evaluate the transmission performance of open submarine optical cable system more accurately.The machine learning model can indirectly estimate the performance parameters which cannot be measured directly,so as to optimize the design of the transmission system.Machine learning model can replace complex integral theory calculation,improve the calculation speed,reduce the cost of users of submarine optical cable,so as to meet the rapid growth of the application demand of submarine open optical cable communication system.The main work of this paper includes:Firstyly,this paper analyzes the communication scenarios and modeling requirements of the open submarine cable communication system,establishes the communication model and basic architecture of the open submarine cable communication system,calculates the transmission performance of the open submarine cable communication system,and obtains the training data set and test data set required by the corresponding machine learning.Secondly,the transmission performance of open submarine optical cable communication system with different signal baud rate,input optical power,modulation format and other parameters is accurately evaluated,and the cooperative transmission model of open submarine optical cable communication system based on machine learning technology is established.The results show that the model can replace the complex integral theoretical calculation.The overall average error between the theoretical value of machine learning model and the predicted Q value is 0.27dB,and the overall average error of OSNR value is 0.33dB.The prediction time of machine learning model is stable at about 4.4s.With the increase of the number of interference channels,when the number of disturbances is more than 5,the calculation speed predicted by machine learning model is obviously better than the theoretical calculation time.Thirdly,a non cooperative transmission model of open submarine cable communication system based on machine learning technology is established.By training different scene models for prediction,the unknown link and system parameters can be estimated more accurately.The results show that when 10000 data sets are taken,the best prediction accuracy can be achieved by the number of spans,input optical power,span length and baud rate,and the average prediction accuracy is 93.2%,96.9%,94%,86.6%and 87.6%,respectively.The machine learning model can predict the transmission parameters of open cable transmission system accurately when the number of cross segments,length of the cross segments,input optical power,baud rate,modulation format and other parameters are unknown.