Raman amplification system simulation method and device

Abstract

The present invention discloses a method for simulating a Raman amplification system: using a simplified power coupling equation and a Gaussian noise model to calculate the output signal gain, output noise power and output optical signal-to-noise ratio; using the calculated output signal gain as an input parameter, and using a Gaussian process regression model to reversely calculate the required pump power; comparing the required pump power calculated by the Gaussian process regression model with the actual input pump power, using the root mean square error as the objective function, and using the gradient descent method to optimize the output signal gain, thereby achieving correction of the output signal gain. The simulation system established according to the method of the present invention has a simple structure and is easy to implement; the present invention can solve the problems that the traditional theoretical model is too complex, difficult to solve, and time-consuming, and is more practical. The present invention also provides a corresponding Raman amplification system simulation device.

Description

Technical Field

[0001] The present invention belongs to the technical field of optical fiber communication, and more specifically, relates to a Raman amplification system simulation method and device. Background Art

[0002] With the continuous development of video services and cloud technologies, communication traffic has continued to grow explosively, and high-speed, large-capacity transmission and low network construction costs have been put forward for optical transmission networks. In order to further expand the system capacity, it is necessary to use dense wavelength division multiplexing (DWDM) to expand the number of channels and expand the wavelength range of channel operation as much as possible. Distributed Raman amplifiers (DRA) are an effective solution. However, DRA also has its shortcomings. Since the transmission fiber is used as the medium, its gain and noise characteristics are closely related to the transmission fiber, pump wavelength and power, channel wavelength and power, etc., and they change with the transmission distance. Compared with the traditional lumped erbium-doped fiber amplifier (EDFA), the working performance of DRA is more likely to change with the changes of the system, and more factors need to be considered in the process of network design and system optimization. In addition, since Raman amplification requires a high-power pump laser, the material cost of DRA is much higher than that of EDFA. It is impossible to build a physical system first and then verify and optimize it in engineering. At the same time, the high pump power will also cause safety hazards during repeated debugging in the project. The best way is to be able to predict the performance of the product in advance, make it work in the best state, and debug it once.

[0003] To this end, it is necessary to establish a fast and accurate simulation method to simulate and calculate the power changes of signals and pumps when they are transmitted along the optical fiber, so that the complex debugging process can be implemented in simulation rather than in engineering, thereby ensuring the reliability of the transmission system. Summary of the invention

[0004] In view of the above defects or improvement needs of the prior art, the present invention proposes a Raman amplification system simulation method, which aims to simulate and calculate the signal gain and noise power of each channel in the Raman amplification system as quickly as possible with high accuracy, so as to evaluate the performance of the Raman amplification system.

[0005] To achieve the above object, according to one aspect of the present invention, a Raman amplification system simulation method is provided, comprising the following steps:

[0006] The output signal gain, output noise power and output optical signal-to-noise ratio are calculated using the simplified power coupling equation and Gaussian noise model;

[0007] The calculated output signal gain and output noise power are used as input parameters, and the required pump power is inversely calculated using the Gaussian process regression model;

[0008] The required pump power calculated by the Gaussian process regression model is compared with the actual input pump power. The root mean square error is used as the objective function, and the output signal gain is optimized using the gradient descent method to achieve the correction of the output signal gain.

[0009] In one embodiment of the present invention, the simplified Raman power coupling equation is:

[0010]

[0011] in Indicates that the jth forward pump is at position z The power at Indicates that the jth pump in the backward direction is at position z The power at That is, the input pump power of the jth pump in the forward direction, That is, the input pump power of the jth pump in the backward direction; α pj represents the loss coefficient of the jth pump in the optical fiber.

[0012] In one embodiment of the present invention, the establishment of the Gaussian process regression model specifically includes:

[0013] The number of pumps, pump wavelength, fiber type, and fiber length tested in the experiment are consistent with the Raman amplification system to be simulated, and the pump power is any configuration that can be achieved in the experiment;

[0014] When establishing a Gaussian process regression model, the signal gain and noise power are used as inputs, and the pump power is used as output. The experimental test data is divided into a training set and a test set, and a suitable kernel function is selected and the initial values ​​of the corresponding hyperparameters are set, thereby determining the prior model of the Gaussian process regression model.

[0015] By training the prior model, the optimal hyperparameters are obtained, and then the posterior model of the Gaussian process regression model, i.e., the regression prediction model, is determined;

[0016] Use the regression prediction model to predict the input of the test set and obtain the distribution of the predicted output points, that is, the mean and covariance with the ability to express uncertainty;

[0017] The predicted output of the test set is compared with the actual output, and the corresponding root mean square error and mean absolute error are calculated to analyze and evaluate the prediction performance of the Gaussian process regression model.

[0018] In one embodiment of the present invention, the step of optimizing the output signal gain by using the gradient descent method specifically includes:

[0019] At each iteration, the corrected value is the output signal gain G i and the noise power P ni , so that it moves in the direction where the root mean square error between the predicted pump power and the actual pump power becomes smaller. The iterative correction process can be expressed as:

[0020] G i,1 =G i,0 +dx1 (13)

[0021] P n,1 =P n,0 +dx2 (14)

[0022]

[0023]

[0024] Among them G i represents the switching gain of the ith channel in the simulation, P ni represents the noise power of the i-th channel at the output end, the subscripts 0 and 1 represent the initial value and the first iteration result, new and old correspond to the next iteration and the previous iteration respectively, γ1 and γ2 represent G i and P ni The step length, and are the differentials of the root mean square error with respect to the switching gain and noise power of the ith channel, respectively. dx1 and dx2 represent the independent variables G i and P ni The amount of change.

[0025] In one embodiment of the present invention, the output signal gain is calculated as follows:

[0026] Calculate the signal power P at the fiber length L si (L), and the input signal power P si (0) Compare and get the gain G of each channel i , which is in the form of

[0027]

[0028] In one embodiment of the present invention, the output noise power is calculated as follows:

[0029] The noise is decomposed into two parts: amplified spontaneous radiation noise and nonlinear noise, so as to calculate the signal-to-noise ratio. The form of the signal-to-noise ratio is:

[0030]

[0031] Where SNR i represents the signal-to-noise ratio of the i-th channel; Psi (0) is the optical power of the fiber-input signal of the ith channel, which is given by the input signal power; P ni is the noise power of the i-th channel at the output, including the amplified spontaneous radiation noise and nonlinear noise, P ASEi is the amplified spontaneous emission noise power of the i-th channel at the output end, calculated by the conventional Raman power coupling equation; η ni is the nonlinear crosstalk coefficient of the ith channel, is the nonlinear noise of the ith channel, P si (0) to the power of 3;

[0032] In one embodiment of the present invention, the nonlinear noise is calculated as follows:

[0033] The nonlinear noise is calculated by the Gaussian noise model, where the nonlinear crosstalk coefficient η of the i-th channel is ni Dependent on the signal frequency, its form is:

[0034]

[0035] Where f i is the center frequency of the ith channel, B ch is the signal bandwidth, G(f i ) is the power spectral density of the nonlinear crosstalk of the i-th channel, P si (0) is the optical power of the fiber-input signal of the i-th channel.

[0036] In one embodiment of the present invention, before calculating the output signal gain, output noise power, and output optical signal-to-noise ratio using the simplified power coupling equation and the Gaussian noise model, the method further includes:

[0037] Determine the parameters of the system to be simulated, including the variables in the simulation process: input pump power, input signal power and input end signal optical signal-to-noise ratio, as well as the constants of the simulation system: fiber type and fiber length.

[0038] In one embodiment of the present invention, after optimizing the output signal gain by using the gradient descent method, the method further includes:

[0039] The adjusted output signal gain and output noise power are input as input parameters into the Gaussian process regression model, and repeated iterations are performed to reduce the root mean square error between the required pump power predicted by the Gaussian process regression model and the actual input pump power to less than a preset value.

[0040] According to another aspect of the present invention, a Raman amplification system simulation device is also provided, comprising a simplified power coupling calculation module, a Gaussian process regression calculation module and a gradient descent optimization adjustment module, wherein:

[0041] The simplified power coupling calculation module is used to calculate the output signal gain, output noise power and output optical signal-to-noise ratio using the simplified power coupling equation and the Gaussian noise model;

[0042] The Gaussian process regression calculation module is used to use the calculated output signal gain and output noise power as input parameters and reversely calculate the required pump power using a Gaussian process regression model;

[0043] The gradient descent optimization adjustment module is used to compare the required pump power calculated by the Gaussian process regression model with the actual input pump power, and use the root mean square error as the objective function to optimize the output signal gain using the gradient descent method, so as to achieve the correction of the output signal gain.

[0044] In general, the above technical solution conceived by the present invention has the following beneficial effects compared with the prior art:

[0045] (1) The simulation system established according to the method of the present invention has a simple structure and is easy to implement;

[0046] (2) The present invention can solve the problems that traditional theoretical models are too complex, difficult to solve, and time-consuming, and is more practical;

[0047] (3) The parameters required for the model established according to the method of the present invention are easy to determine, and many parameters can be directly obtained by referring to the product manual;

[0048] (4) Compared with other simplified simulation methods, the present invention has significant advantages in accuracy, but the complexity is not particularly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A schematic diagram of a flow chart of a Raman amplification system simulation method provided by an embodiment of the present invention;

[0050] Figure 2 This is a scenario where an embodiment of the present invention is applied to a 200km bidirectional Raman amplification system;

[0051] Figure 3 Comparison of simulation and experimental results of the embodiments of the present invention;

[0052] Figure 4 It is a structural schematic diagram of a Raman amplification system simulation system in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0054] Figure 1 FIG. 4 is a flow chart of the Raman amplification system simulation method of the present invention. Figure 1 As shown, the method of the embodiment of the present invention is: the input parameters are as shown in the figure, including input pump power, input signal power, fiber type, fiber length, input end signal optical signal-to-noise ratio, firstly, the simplified power coupling equation and Gaussian noise model (Gaussian Noise Model, GN model) are used to calculate the output signal gain, output noise power, and output optical signal-to-noise ratio; then, the calculated output signal gain is used as the input parameter, and the required pump power is reversely calculated using the Gaussian process regression model; then, the required pump power calculated by the Gaussian process regression model is compared with the actual input pump power, and the root mean square error is used as the objective function, and the output signal gain is optimized by the gradient descent method, so as to realize the correction of the output signal gain.

[0055] Specifically, the method of the present invention comprises the following steps:

[0056] Step 1, determine the parameters of the system to be simulated, including the variables in the simulation process: input pump power, input signal power and input end signal optical signal-to-noise ratio, and the constants of the simulation system: fiber type and fiber length;

[0057] Among them, the variables include input pump power, input signal power, and input end signal light signal-to-noise ratio; the fiber type and fiber length need to be determined, and once the Gaussian process regression model is trained, it will not change;

[0058] Furthermore, when determining the type of optical fiber, its loss characteristics and Raman gain characteristics for light of different frequencies are also clarified, namely the loss coefficient α(f) and the Raman coupling coefficient C R (f pj ,f si ) should also be determined accordingly;

[0059] In step 1, various parameters that affect the system performance are considered when simulating the Raman amplification system. Accurate determination of these parameters ensures subsequent accurate calculations:

[0060] Step 2: According to the parameters obtained in step 1, the output signal power and amplified spontaneous radiation noise of each channel are solved by using the simplified Raman power coupling equation in combination with the Raman power coupling equation, and the nonlinear noise of each channel is solved by using the Gaussian noise model; the output signal gain of each channel is calculated according to the output signal power and signal input power of each channel; the output noise power of each channel is obtained according to the amplified spontaneous radiation noise and nonlinear noise of each channel.

[0061] Specifically, to calculate the output signal gain of each channel, it is necessary to first calculate the pump power evolution, ignoring the consumption caused by the pump providing gain to the signal, and obtain the simplified Raman power coupling equation, which is in the form of:

[0062]

[0063] in Indicates that the jth forward pump is at position z The power at Indicates that the jth pump in the backward direction is at position z The power at That is, the input pump power of the jth pump in the forward direction, That is, the input pump power of the jth pump in the backward direction, determined by step 1; α pj represents the loss coefficient of the jth pump in the optical fiber.

[0064] After obtaining the pump power evolution by simplifying the Raman power coupling equation, the conventional Raman power coupling equation is used to calculate the output signal power and output noise power of each channel, which is in the form of

[0065]

[0066]

[0067] Where P si (z) represents the power of the ith channel at the transmission distance z, P ASEi (z) represents the amplified spontaneous emission noise of the i-th channel at the transmission distance z. The signal power and the amplified spontaneous emission noise power at position 0 are determined by the input signal power and the input end signal optical signal-to-noise ratio in step 1. si represents the loss coefficient of the ith channel in the optical fiber, C R (f pj ,f si ) indicates the frequency is f pj The pump frequency is f si The Raman coupling coefficient between the signals of , the superscript + in the pump power indicates forward pumping, the superscript - indicates backward pumping, h is the Planck constant, k is the Boltzmann constant, and T is the Kelvin temperature.

[0068] The signal power P at the fiber length L is calculated using the conventional Raman power coupling equation si (L) and input signal power P si (0) The gain G of each channel can be obtained by comparing i , which is in the form of

[0069]

[0070] When solving the conventional Raman power coupling equation, since the signal input power is known, that is, the initial value is known, and the pump power has been calculated using a simplified model, the differential equation can be directly solved numerically using the Runge-Kutta method.

[0071] When calculating noise, the noise is decomposed into two parts: amplified spontaneous radiation noise and nonlinear noise, so as to calculate the signal-to-noise ratio. The form of the signal-to-noise ratio is:

[0072]

[0073] Where SNR i represents the signal-to-noise ratio of the i-th channel; P si (0) is the optical power of the fiber-input signal of the ith channel, which is given by the input signal power; P ni is the noise power of the i-th channel at the output, including the amplified spontaneous radiation noise and nonlinear noise, P ASEi is the amplified spontaneous emission noise power of the i-th channel at the output end, calculated by the conventional Raman power coupling equation; η ni is the nonlinear crosstalk coefficient of the ith channel, is the nonlinear noise of the ith channel, P si (0) to the power of 3.

[0074] The nonlinear noise is calculated by the GN model, where the nonlinear crosstalk coefficient η of the i-th channel is ni Dependent on the signal frequency, its form is:

[0075]

[0076] Where f i is the center frequency of the ith channel, B ch is the signal bandwidth, G(f i ) is the power spectral density of the nonlinear crosstalk of the i-th channel, P si (0) is the optical power of the fiber-input signal of the i-th channel.

[0077] The power spectral density G(f) of nonlinear crosstalk is calculated by integration and its form is

[0078]

[0079] Where γ is the nonlinear coefficient, G Tx is the input signal power spectral density, Φ=-4π 2 (f1-f)(f2-f)[β2+πβ3(f1+f2)]z, β2 is the group velocity dispersion parameter, β3 is the slope of the group velocity dispersion parameter, and ρ(z, f) is the normalized signal power distribution.

[0080] The normalized signal power distribution ρ(z, f) can be calculated based on the Raman gain in the form of

[0081]

[0082] Where P tot is the total power of the input signal, α is the loss coefficient of the signal in the optical fiber, z is the transmission distance, C R is the Raman coupling coefficient, is the effective transmission distance, f is the center frequency of the channel, G Tx is the input signal power spectral density.

[0083] In step 2, the simplified Raman power coupling equation and GN model used are low-complexity methods with short calculation time and good accuracy:

[0084] Step 3, inputting the output signal gain and output noise power of each channel obtained in step 2 into the Gaussian process regression model, outputting the required pump power, comparing the required pump power predicted by the Gaussian process regression model with the actual input pump power to obtain a root mean square error, and adjusting the output signal gain and output noise power by using a gradient descent method based on the root mean square error;

[0085] In the embodiment of the present invention, it is first necessary to establish a Gaussian process regression model through experimental tests to obtain the relationship between signal gain, noise power and required pump power; in this step, it is necessary to conduct more experimental tests to obtain the signal gain and noise power of different channels under different pump power configurations, and establish a Gaussian process regression model through a large number of experimental test results to obtain the relationship between signal gain, noise power and required pump power; the input is signal gain and noise power, the output is pump power, and the fiber type, fiber length, etc. remain unchanged by default, and the model training is completed under this premise;

[0086] Specifically, the configurations of the experimental test such as the number of pumps, pump wavelength, fiber type, and fiber length are consistent with the Raman amplification system to be simulated, and the pump power is any configuration that can be achieved in the experiment. When establishing a Gaussian process regression model, the signal gain and noise power are used as inputs, and the pump power is used as output. The experimental test data is divided into a training set and a test set, and a suitable kernel function is selected and the initial values ​​of the corresponding hyperparameters are set to determine the prior model of the Gaussian process regression model. By training the prior model, the optimal hyperparameters are obtained, and then the posterior model of the Gaussian process regression model, that is, the regression prediction model, is determined; the input of the test set is predicted using the regression prediction model to obtain the distribution of the predicted output points, that is, the mean and covariance with the ability to express uncertainty; the predicted output of the test set is compared with the actual output, and the corresponding root mean square error and mean absolute error are calculated to analyze and evaluate the prediction performance of the Gaussian process regression model; in this step, the established Gaussian process regression model relies on the experimental measured data and can ensure sufficient accuracy;

[0087] Specifically, the form of the root mean square error is:

[0088]

[0089] Where P pi , GPR is the pump power predicted by the Gaussian process regression model, and m is the number of pumps.

[0090] In step 3, we use the Gaussian process regression model to reversely predict the pump power and compare it with the actual known true value, and use the gradient descent method to feedback and adjust the signal gain and noise power according to the root mean square error;

[0091] Step 4, using the output signal gain and output noise power adjusted in step 3 as input parameters, repeating step 3, iteratively reducing the root mean square error between the required pump power predicted by the Gaussian process regression model and the actual input pump power to less than a preset value;

[0092] Specifically, the gradient descent method is used to adjust the output signal gain and output noise power, and the minimum value of the root mean square error is solved or made lower than the preset value in the direction of decreasing the root mean square error. In each iteration, the corrected value is the output signal gain G i and the output noise power P ni , so that it moves in the direction where the root mean square error between the predicted pump power and the actual pump power becomes smaller. The iterative correction process can be expressed as:

[0093] G i,1 =G i,0 +dx1 (13)

[0094] P n,1 =Pn,0 +dx2 (14)

[0095]

[0096] Among them G i represents the switching gain of the ith channel in the simulation, P ni represents the noise power of the i-th channel at the output end, the subscripts 0 and 1 represent the initial value and the first iteration result, new and old correspond to the next iteration and the previous iteration respectively, γ1 and γ2 represent G i and P ni The step size is determined by G i and P ni The value of γ0 and γ1 is determined by the size of G (these two values ​​are empirical values, usually proportional to G i and P ni ). and are the differentials of the root mean square error with respect to the switching gain and noise power of the ith channel, respectively. dx1 and dx2 represent the independent variables G i and P ni The change in value should be as small as possible.

[0097] Specifically, according to the initial G i,0 and P ni,0 We can use formula (12) to calculate a set of RMSE, and then according to formulas (13) and (14) we can get the G of the first iteration i,1 and P ni,1 , and use formula (12) to calculate a new set of RMSE, and substitute the two calculated RMSE into formulas (16) and (18) to obtain G i and P ni Substituting their respective gradients into formulas (15) and (17) yields the new G i and P ni Then, the error can be reduced by repeatedly iterating formulas (15) to (18).

[0098] In step 4, we repeat step 3 to gradually reduce the error, reduce the error of the simplified calculation method in step 2, and make the calculation have sufficient accuracy; in this embodiment, there are 4 pump sources at the front and back, 47 signals, pump wavelengths of 1425; 1435; 1455; 1465nm, signal frequencies of 191.4 to 196THz, and intervals of 100GHz. Figure 2 The scenario of the 200km bidirectional Raman amplification system shown in FIG. Figure 3The experimental and simulation results in this embodiment are compared, where (a) is the switch gain of 47 signals, (b) is the output OSNR of the 4 monitored signals, and the frequencies of the monitored signals are 195.9, 194.5, 192.7, and 191.5 THz, respectively. After iterative correction, the simulation errors are only 0.37, 0.39, 0.47, and 0.01 dB, which is about 4 dB lower than the error before iterative correction. The method proposed in the present invention only takes about 6 seconds to calculate, while the traditional method of calculating the power coupling equation takes more than ten seconds. Under the premise of ensuring the calculation accuracy, the calculation speed of the present invention has a significant advantage.

[0099] It needs to be further explained that if Figure 1 As shown, the process of calculating the output signal gain, output noise power, and output optical signal-to-noise ratio using the simplified power coupling equation and the Gaussian noise model only needs to be executed once, while the subsequent reverse calculation of the required pump power using the Gaussian process regression model, and the comparison of the required pump power calculated by the Gaussian process regression model with the actual input pump power, using the root mean square error as the objective function, and using the gradient descent method to optimize the output signal gain, thereby realizing the process of correcting the output signal gain, needs to be executed multiple times until the root mean square error is reduced to less than a preset value.

[0100] Furthermore, if Figure 4 As shown, the present invention also provides a Raman amplification system simulation device, including a simplified power coupling calculation module, a Gaussian process regression calculation module and a gradient descent optimization adjustment module, wherein:

[0101] The simplified power coupling calculation module is used to calculate the output signal gain, output noise power and output optical signal-to-noise ratio using the simplified power coupling equation and the Gaussian noise model;

[0102] The Gaussian process regression calculation module is used to use the calculated output signal gain and output noise power as input parameters and reversely calculate the required pump power using a Gaussian process regression model;

[0103] The gradient descent optimization adjustment module is used to compare the required pump power calculated by the Gaussian process regression model with the actual input pump power, and use the root mean square error as the objective function to optimize the output signal gain using the gradient descent method, so as to achieve the correction of the output signal gain.

[0104] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A Raman amplification system simulation method, characterized in that: The steps include: The output signal gain, output noise power and output optical signal-to-noise ratio are calculated using a simplified power coupling equation and a Gaussian noise model; the simplified power coupling equation is: in represents the power of the jth forward pump at position z, represents the power of the jth pump at position z in the backward direction, That is, the input pump power of the jth pump in the forward direction, That is, the input pump power of the jth pump in the backward direction; α pj represents the loss coefficient of the jth pump in the optical fiber; The output signal gain is calculated as follows: Calculate the signal power P at the fiber length L si (L), and the input signal power P si (0) Compare and get the gain G of each channel i , which is in the form of The output noise power is calculated as: The noise is decomposed into two parts: amplified spontaneous radiation noise and nonlinear noise, so as to calculate the signal-to-noise ratio. The form of the signal-to-noise ratio is: Where SNR i represents the signal-to-noise ratio of the i-th channel; P si (0) is the optical power of the fiber-input signal of the ith channel, which is given by the input signal power; P ni is the noise power of the i-th channel at the output, including the amplified spontaneous radiation noise and nonlinear noise, P ASEi is the amplified spontaneous emission noise power of the i-th channel at the output, calculated by the conventional Raman power coupling equation; η ni is the nonlinear crosstalk coefficient of the ith channel, is the nonlinear noise of the ith channel, P si (0) to the power of 3; The calculated output signal gain and output noise power are used as input parameters, and the required pump power is inversely calculated using the Gaussian process regression model; The required pump power calculated by the Gaussian process regression model is compared with the actual input pump power. The root mean square error is used as the objective function, and the output signal gain is optimized using the gradient descent method to achieve the correction of the output signal gain.

2. The Raman amplification system simulation method according to claim 1, characterized in that: The establishment of the Gaussian process regression model specifically includes: The number of pumps, pump wavelength, fiber type, and fiber length tested in the experiment are consistent with the Raman amplification system to be simulated, and the pump power is any configuration that can be achieved in the experiment; When establishing a Gaussian process regression model, the signal gain and noise power are used as inputs, and the pump power is used as output. The experimental test data is divided into a training set and a test set, and a suitable kernel function is selected and the initial values ​​of the corresponding hyperparameters are set, thereby determining the prior model of the Gaussian process regression model. By training the prior model, the optimal hyperparameters are obtained, and then the posterior model of the Gaussian process regression model, that is, the regression prediction model, is determined; Use the regression prediction model to predict the input of the test set and obtain the distribution of the predicted output points, that is, the mean and covariance with uncertainty expression capabilities; The predicted output of the test set is compared with the actual output, and the corresponding root mean square error and mean absolute error are calculated to analyze and evaluate the prediction performance of the Gaussian process regression model.

3. The Raman amplification system simulation method according to claim 1, characterized in that: The method of optimizing the output signal gain by using the gradient descent method specifically includes: At each iteration, the corrected value is the output signal gain G i and noise power P ni , so that it moves in the direction where the root mean square error between the predicted pump power and the actual pump power becomes smaller. The iterative correction process can be expressed as: G i,1 =G i,0 +dx1(13) P n,1 =P n,0 +dx2(14) Among them G i represents the switching gain of the ith channel in the simulation, P ni represents the noise power of the i-th channel at the output end, the subscripts 0 and 1 represent the initial value and the first iteration result, new and old correspond to the next iteration and the previous iteration respectively, γ1 and γ2 represent G i and P ni The step length, and are the differentials of the root mean square error with respect to the switching gain and noise power of the ith channel, respectively. dx1 and dx2 represent the independent variables G i and P ni The amount of change.

4. The Raman amplification system simulation method according to claim 1, characterized in that: The nonlinear noise is calculated as follows: The nonlinear noise is calculated by the Gaussian noise model, where the nonlinear crosstalk coefficient η of the i-th channel is ni Dependent on the signal frequency, its form is: Where f i is the center frequency of the ith channel, B ch is the signal bandwidth, G(f i ) is the power spectral density of the nonlinear crosstalk of the i-th channel, P si (0) is the optical power of the fiber-input signal of the i-th channel.

5. The Raman amplification system simulation method according to claim 1, characterized in that: Before calculating the output signal gain, output noise power, and output optical signal-to-noise ratio using the simplified power coupling equation and the Gaussian noise model, the method further includes: Determine the parameters of the system to be simulated, including the variables in the simulation process: input pump power, input signal power and input end signal optical signal-to-noise ratio, as well as the constants of the simulation system: fiber type and fiber length.

6. The Raman amplification system simulation method according to claim 1, characterized in that: After optimizing the output signal gain using the gradient descent method, it also includes: The adjusted output signal gain and output noise power are input as input parameters into the Gaussian process regression model, and repeated iterations are performed to reduce the root mean square error between the required pump power predicted by the Gaussian process regression model and the actual input pump power to less than a preset value.

7. A Raman amplification system simulation device, characterized in that: The method comprises a simplified power coupling calculation module, a Gaussian process regression calculation module and a gradient descent optimization adjustment module, and is used to complete the Raman amplification system simulation method according to any one of claims 1 to 6, wherein: The simplified power coupling calculation module is used to calculate the output signal gain, output noise power and output optical signal-to-noise ratio using the simplified power coupling equation and the Gaussian noise model; The Gaussian process regression calculation module is used to take the calculated output signal gain and output noise power as input parameters, and use the Gaussian process regression model to reversely calculate the required pump power; the gradient descent optimization adjustment module is used to compare the required pump power calculated by the Gaussian process regression model with the actual input pump power, use the root mean square error as the objective function, and use the gradient descent method to optimize the output signal gain, thereby realizing the correction of the output signal gain.

Images