Method and system for rapidly calculating power evolution in third-order Raman amplifier

Abstract

The invention provides a method and a system for quickly calculating power evolution in a third-order Raman amplifier. The method comprises the following steps: S1, calculating power evolution of light with various frequencies in an optical fiber based on parameters of the third-order Raman amplifier; s2, fixing link parameters, combining signal parameters, pumping parameters and initial values of power evolution into a data set, and training a neural network; s3, predicting a power evolution initial value of a new link by using the trained neural network, and taking the power evolution initial value as a convergence starting point of a targeting method; s4, enabling the targeting method to converge the initial power evolution value based on a continuity method and a gradient descent method, and obtaining a final power evolution value in the optical fiber. According to the method, the known link, signal and pumping parameters are utilized, a more reasonable power evolution initial value is predicted by training the neural network and serves as the starting point of the targeting method, the high-precision power evolution initial value is adopted, the number of iterations of a gradient descent method in the targeting method is reduced, the convergence speed is increased, and the precision of a final solution is improved.

Description

Technical Field

[0001] The present invention relates to the field of optical fiber communication technology, and in particular to a method and system for rapidly calculating power evolution in a third-order Raman amplifier. Background Art

[0002] In fiber-optic communication systems, Raman amplifiers play a vital role due to their ability to provide distributed gain over long distances. Their operating principle is based on stimulated Raman scattering (SRS), which is the nonlinear interaction between pump light and signal light. In order to accurately model and optimize the performance of such amplifiers, the computational problem of power evolution must be solved.

[0003] The calculation of power evolution involves solving the boundary value problem of Raman coupled differential equations, and traditional numerical methods such as the shooting method are usually used to solve such problems. However, in the third-order Raman amplifier, due to the increase in system complexity, the initial value setting of power evolution obtained by two one-way Raman coupled differential equations in the traditional method has a large error, which makes it difficult for the shooting method to converge; and the entire solution process takes too long, which not only makes the calculation inefficient, but also affects the accuracy of the final result. Summary of the invention

[0004] In view of the defects in the prior art, an object of the present invention is to provide a method and system for quickly calculating the power evolution in a third-order Raman amplifier.

[0005] A method for rapidly calculating power evolution in a third-order Raman amplifier provided by the present invention comprises:

[0006] Step S1: based on the parameters of the third-order Raman amplifier, the power evolution of each frequency light in the optical fiber is calculated;

[0007] The parameters of the third-order Raman amplifier include link parameters, signal parameters and pump parameters;

[0008] Step S2: fix the link parameters, combine the signal parameters, pump parameters and initial values ​​of power evolution into a data set, and train the neural network;

[0009] Step S3: Use the trained neural network to predict the initial value of the power evolution of the new link, and use it as the convergence starting point of the shooting method;

[0010] Step S4: The targeting method is used to converge the initial value of the power evolution based on the continuity method and the gradient descent method, and the final value of the power evolution inside the optical fiber is obtained.

[0011] Preferably, the initial value of the power evolution includes an initial value of the real input optical power and an initial value of the noise optical power generated by the combined effects of stimulated Raman scattering, Rayleigh reflection and end face reflection.

[0012] Preferably, the input features of the neural network are signals and pump parameters, and the output features are the initial values ​​of the evolution of optical power at all frequencies; the input features and output features of the neural network are connected using a fully connected layer, and the structure is a traditional feedforward neural network.

[0013] Preferably, step S4 includes the following sub-steps:

[0014] Step S4.1: The shooting method uses the trained neural network to predict the initial value of power evolution;

[0015] Step S4.2: forward calculating the Raman coupled differential equation using the initial value of the power evolution to obtain the final value of the power evolution;

[0016] Step S4.3: Determine the sequence of boundary conditions required for the stepwise convergence of the Raman coupled differential equations according to the final value of the power evolution and the continuity method;

[0017] Step S4.4: Use the gradient descent method to gradually solve the corresponding boundary value problem in the order of the boundary condition sequence from small to large, and use the result of each convergence as the convergence starting point for the next solution.

[0018] Preferably, the maximum value of the boundary condition sequence is the actual injected pump light power, the minimum value does not exceed the final value of the power evolution, and the interval is adaptively determined by a continuity method.

[0019] According to the present invention, a system for rapidly calculating power evolution in a third-order Raman amplifier comprises:

[0020] Module M1: Based on the parameters of the third-order Raman amplifier, the power evolution of each frequency light in the optical fiber is calculated;

[0021] The parameters of the third-order Raman amplifier include link parameters, signal parameters and pump parameters;

[0022] Module M2: fix the link parameters, combine the initial values ​​of signal parameters, pump parameters and power evolution into a data set, and train the neural network;

[0023] Module M3: Use the trained neural network to predict the initial value of the power evolution of the new link and use it as the convergence starting point of the shooting method;

[0024] Module M4: Make the shooting method converge the initial value of power evolution based on the continuity method and the gradient descent method, and obtain the final value of power evolution inside the optical fiber.

[0025] Preferably, the initial value of the power evolution includes an initial value of the real input optical power and an initial value of the noise optical power generated by the combined effects of stimulated Raman scattering, Rayleigh reflection and end face reflection.

[0026] Preferably, the input features of the neural network are signals and pump parameters, and the output features are the initial values ​​of the evolution of optical power at all frequencies; the input features and output features of the neural network are connected using a fully connected layer, and the structure is a traditional feedforward neural network.

[0027] Preferably, the module M4 includes the following submodules:

[0028] Module M4.1: Target shooting method uses the trained neural network to predict the initial value of power evolution;

[0029] Module M4.2: Forward calculation of the Raman coupled differential equation using the initial value of power evolution to obtain the final value of power evolution;

[0030] Module M4.3: Determine the sequence of boundary conditions required for the stepwise convergence of the Raman coupled differential equations based on the terminal value and continuity methods of power evolution;

[0031] Module M4.4: Use the gradient descent method to gradually solve the corresponding boundary value problem in the order of boundary conditions from small to large, and use the result of each convergence as the starting point of the next solution.

[0032] Preferably, the maximum value of the boundary condition sequence is the actual injected pump light power, the minimum value does not exceed the final value of the power evolution, and the interval is adaptively determined by a continuity method.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. The present invention utilizes known link, signal and pump parameters, and trains a neural network to predict a more reasonable initial value of power evolution as the starting point of the shooting method. By adopting a high-precision initial value of power evolution, the number of iterations of the gradient descent method in the shooting method is reduced, which not only accelerates the convergence speed, but also improves the accuracy of the final solution.

[0035] 2. The present invention achieves accurate prediction of the initial value of power evolution by gradually adjusting the boundary conditions using the continuity method and the gradient descent method, thereby improving the convergence success rate of the targeting method, being able to effectively handle complex Raman coupled differential equations, and providing strong support for the optimal design of optical fiber communication systems.

[0036] Other beneficial effects of the present invention will be explained in the specific implementation manner through the introduction of specific technical features and technical solutions. Through the introduction of these technical features and technical solutions, those skilled in the art should be able to understand the beneficial technical effects brought about by the technical features and technical solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0038] Figure 1 The figure is a flow chart of the method of the present invention.

[0039] Figure 2 It is a technical flow chart of the present invention. DETAILED DESCRIPTION

[0040] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0041] Reference Figure 1 and Figure 2 As shown, a method for quickly calculating the power evolution in a third-order Raman amplifier comprises:

[0042] Step 1: In a third-order Raman amplifier, with known link parameters, signal parameters and pump parameters, the target shooting method of solving boundary value problems of ordinary differential equations is used to numerically solve the power evolution of each frequency light in the optical fiber.

[0043] Step 2: Fix the link parameters and combine the signal parameters, pump parameters and initial values ​​of power evolution into a data set to train the neural network.

[0044] Step 2.1: The initial value of power evolution includes the initial value of the real input optical power and the initial value of the noise optical power generated by stimulated Raman scattering, Rayleigh reflection and end face reflection. The input features of the neural network are the signal and pump parameters, and the output features are the initial values ​​of the optical power evolution of all frequencies.

[0045] Step 2.2: The input features and output features of the neural network are connected using a fully connected layer, and the structure is a traditional feedforward neural network.

[0046] Step 3: Use the neural network to predict the initial value of the power evolution of the new link and use it as the convergence starting point of the targeting method.

[0047] Step 4: The targeting method uses the continuity method and the gradient descent method to converge the initial value of the power evolution to meet the boundary conditions of the Raman coupled differential equation. After successful convergence, the power evolution inside the optical fiber is obtained with high precision.

[0048] Step 4.1: The shooting method uses the previously trained neural network to predict the initial value of power evolution.

[0049] Step 4.2: Using the initial value of the power evolution, the Raman coupled differential equation is forward calculated by the fourth-order Runge-Kutta method to obtain the final value of the power evolution, which is expressed as:

[0050] P n+1 =P n +h(a1k1+a2k2+a3k3+a4k4)

[0051] in,

[0052] k1=f(z n ,P n )

[0053]

[0054] k4=f(z n +h,P n +k3h)

[0055] In the above formula, P n and P n+1 They are respectively z away from the starting point of the optical fiber n and z n+1 The real-time power corresponding to each frequency light at the time. h is the step size of each iteration in the fourth-order Runge-Kutta method, that is, h = z n+1 -z n . f(z n ,P n ) is the optical power P at each frequency n Relative to the distance z from the starting point of the optical fiber n The derivative of , that is, the Raman coupled differential equation.

[0056] Step 4.3: According to the final value of the power evolution and the continuity method, determine the boundary condition sequence required for the gradual convergence of the Raman coupled differential equation. The maximum value of the sequence is the actual injected pump light power, and the minimum value does not exceed the final value of the power evolution. The initial value of the interval needs to be set manually and then adjusted by the adaptive algorithm. For any pump power in the sequence, when the number of iterations in the target shooting process corresponding to the value is less than 20 times, the interval is increased; when the target shooting process corresponding to the value converges 5 times and still does not fall below the threshold, the interval is reduced. As the pump power in the sequence gradually increases to the actual injected pump light power, the continuity method converges.

[0057] Step 4.4: Use the gradient descent method to gradually solve the corresponding boundary value problem in the order of the boundary condition sequence from small to large, and use the result of each convergence as the starting point of the next solution.

[0058] The present invention also provides a system for rapidly calculating the power evolution in a third-order Raman amplifier. The system for rapidly calculating the power evolution in a third-order Raman amplifier can be implemented by executing the process steps of the method for rapidly calculating the power evolution in a third-order Raman amplifier, that is, those skilled in the art can understand the method for rapidly calculating the power evolution in a third-order Raman amplifier as a preferred implementation of the system for rapidly calculating the power evolution in a third-order Raman amplifier.

[0059] Specifically, a system for rapidly calculating power evolution in a third-order Raman amplifier includes:

[0060] Module M1: Based on the parameters of the third-order Raman amplifier, the power evolution of each frequency light in the optical fiber is calculated;

[0061] The parameters of the third-order Raman amplifier include link parameters, signal parameters and pump parameters;

[0062] Module M2: fix the link parameters, combine the initial values ​​of signal parameters, pump parameters and power evolution into a data set, and train the neural network;

[0063] Module M3: Use the trained neural network to predict the initial value of the power evolution of the new link and use it as the convergence starting point of the shooting method;

[0064] Module M4: Make the shooting method converge the initial value of power evolution based on the continuity method and the gradient descent method, and obtain the final value of power evolution inside the optical fiber.

[0065] The initial value of the power evolution includes the initial value of the real input optical power and the initial value of the noise optical power generated by the combined effects of stimulated Raman scattering, Rayleigh reflection and end face reflection.

[0066] The input features of the neural network are signals and pump parameters, and the output features are the initial values ​​of the evolution of optical power at all frequencies; the input features and output features of the neural network are connected using a fully connected layer, and the structure is a traditional feedforward neural network.

[0067] The module M4 includes the following submodules:

[0068] Module M4.1: Target shooting method uses the trained neural network to predict the initial value of power evolution;

[0069] Module M4.2: Forward calculation of the Raman coupled differential equation using the initial value of power evolution to obtain the final value of power evolution;

[0070] Module M4.3: Determine the sequence of boundary conditions required for the stepwise convergence of the Raman coupled differential equations based on the terminal value and continuity methods of power evolution;

[0071] Module M4.4: Use the gradient descent method to gradually solve the corresponding boundary value problem in the order of boundary conditions from small to large, and use the result of each convergence as the starting point of the next solution.

[0072] The maximum value of the boundary condition sequence is the actual injected pump light power, the minimum value does not exceed the final value of the power evolution, and the interval is adaptively determined by the continuity method.

[0073] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0074] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A method for rapidly calculating power evolution in a third-order Raman amplifier, characterized in that: include: Step S1: based on the parameters of the third-order Raman amplifier, the power evolution of each frequency light in the optical fiber is calculated; The parameters of the third-order Raman amplifier include link parameters, signal parameters and pump parameters; Step S2: fix the link parameters, combine the signal parameters, pump parameters and initial values ​​of power evolution into a data set, and train the neural network; Step S3: Use the trained neural network to predict the initial value of the power evolution of the new link, and use it as the convergence starting point of the shooting method; Step S4: The targeting method is used to converge the initial value of the power evolution based on the continuity method and the gradient descent method, and the final value of the power evolution inside the optical fiber is obtained.

2. The method for rapidly calculating power evolution in a third-order Raman amplifier according to claim 1, characterized in that: The initial value of the power evolution includes the initial value of the real input optical power and the initial value of the noise optical power generated by the combined effects of stimulated Raman scattering, Rayleigh reflection and end face reflection.

3. The method for rapidly calculating power evolution in a third-order Raman amplifier according to claim 2, characterized in that: The input features of the neural network are signals and pump parameters, and the output features are the initial values ​​of the evolution of optical power at all frequencies; the input features and output features of the neural network are connected using a fully connected layer, and the structure is a traditional feedforward neural network.

4. The method for rapidly calculating power evolution in a third-order Raman amplifier according to claim 1, characterized in that: The step S4 comprises the following sub-steps: Step S4.1: The shooting method uses the trained neural network to predict the initial value of power evolution; Step S4.2: forward calculating the Raman coupled differential equation using the initial value of the power evolution to obtain the final value of the power evolution; Step S4.3: Determine the sequence of boundary conditions required for the stepwise convergence of the Raman coupled differential equations according to the final value of the power evolution and the continuity method; Step S4.4: Use the gradient descent method to gradually solve the corresponding boundary value problem in the order of the boundary condition sequence from small to large, and use the result of each convergence as the convergence starting point for the next solution.

5. The method for rapidly calculating power evolution in a third-order Raman amplifier according to claim 4, characterized in that: The maximum value of the boundary condition sequence is the actual injected pump light power, the minimum value does not exceed the final value of the power evolution, and the interval is adaptively determined by the continuity method.

6. A system for rapidly calculating the power evolution in a third-order Raman amplifier, characterized in that: include: Module M1: Based on the parameters of the third-order Raman amplifier, the power evolution of each frequency light in the optical fiber is calculated; The parameters of the third-order Raman amplifier include link parameters, signal parameters and pump parameters; Module M2: fix the link parameters, combine the initial values ​​of signal parameters, pump parameters and power evolution into a data set, and train the neural network; Module M3: Use the trained neural network to predict the initial value of the power evolution of the new link and use it as the convergence starting point of the shooting method; Module M4: Make the shooting method converge the initial value of power evolution based on the continuity method and the gradient descent method, and obtain the final value of power evolution inside the optical fiber.

7. The system for rapidly calculating power evolution in a third-order Raman amplifier according to claim 6, characterized in that: The initial value of the power evolution includes the initial value of the real input optical power and the initial value of the noise optical power generated by the combined effects of stimulated Raman scattering, Rayleigh reflection and end face reflection.

8. The system for rapidly calculating power evolution in a third-order Raman amplifier according to claim 7, characterized in that: The input features of the neural network are signals and pump parameters, and the output features are the initial values ​​of the evolution of optical power at all frequencies; the input features and output features of the neural network are connected using a fully connected layer, and the structure is a traditional feedforward neural network.

9. The system for rapidly calculating power evolution in a third-order Raman amplifier according to claim 6, characterized in that: The module M4 includes the following submodules: Module M4.1: Target shooting method uses the trained neural network to predict the initial value of power evolution; Module M4.2: Forward calculation of the Raman coupled differential equation using the initial value of power evolution to obtain the final value of power evolution; Module M4.3: Determine the sequence of boundary conditions required for the stepwise convergence of the Raman coupled differential equations based on the terminal value and continuity methods of power evolution; Module M4.4: Use the gradient descent method to gradually solve the corresponding boundary value problem in the order of boundary conditions from small to large, and use the result of each convergence as the starting point of the next solution.

10. The system for rapidly calculating power evolution in a third-order Raman amplifier according to claim 9, characterized in that: The maximum value of the boundary condition sequence is the actual injected pump light power, the minimum value does not exceed the final value of the power evolution, and the interval is adaptively determined by the continuity method.

Images