Research On Intelligent Gain Control Of Hybrid Raman-Erbium Doped Fiber Amplifiers In Optical Communication Systems

With the explosive growth of data traffic,optical fiber communication systems are facing the pressure of increasing transmission capacity.In order to reduce the cost of system expansion,the transmission wavelength of a single optical fiber is usually increased.However,the gain and noise performance of erbium-doped fiber amplifiers(EDFAs)degrade as the communication frequency band expands from the traditional C-band.Raman Fiber Amplifier(RFA)is used in long-distance transmission systems in combination with EDFA to form a hybrid fiber amplifier that combines performance and cost with flexible gain spectrum,gain bandwidth,and good noise performance.On the other hand,before operating the wavelength in a dynamic optical network,it is necessary to estimate the transmission quality and optimize the channel power to ensure the normal operation of all services.When optimizing the channel power of the transmission system using the hybrid fiber amplifier,the flexible and adjustable gain spectrum of the RFA is usually used,and the power dynamics of the EDFA are considered at the same time to realize the specific power spectrum output of the hybrid amplifier node.However,the traditional RFA gain spectrum design method cannot meet the real-time requirements of dynamic optical networks.At the same time,the number of pumps in commercial RFA is limited,and there are often gain design errors.Moreover,there is a lack of accurate EDFA forward power prediction and backward power configuration models.This makes hybrid amplifiers unable to take full advantage of dynamic optical networks.In this paper,the following innovative solutions are proposed for the intelligent gain control of hybrid fiber amplifiers to achieve the specific power spectrum output required for power optimization strategies in dynamic channel load scenarios:First,due to the lack of accurate EDFA forward power prediction and backward power configuration models,and for intelligent gain control of hybrid amplifiers,we use neural networks to model EDFAs.98%of the test errors of the forward and backward models are less than 0.11dB;and due to the lack of an accurate forward power prediction model and a realtime Raman gain spectrum design model,a neural network was used to establish a forward power prediction and reverse gain design model based on any channel load and pump power.The forward average error is 0.23dB,and the pump power design error is less than 11mW.Second,due to the power dynamics of EDFA in the multi-span transmission link and the complex SRS effect under dynamic channel load conditions,the previous method cannot accurately predict power evolution,so we use neural networks to establish a power evolution model,with a 98%test error of less than 0.14dB;in order to compensate for the Raman gain spectrum design error,the autoencoder is built using the power evolution model to complete the optimization of each channel power.