A method for optimizing the transmission capacity of dynamic multi-band optical networks based on GSNR sensing

Abstract

This invention provides a dynamic multi-band optical network transmission capacity optimization method based on GSNR sensing. The method constructs a nonlinear mapping model between the GSNR characteristics of the input signal and the GSNR characteristics of the output signal, using Raman amplifier pump configuration parameters. This model senses the GSNR characteristics of the input signal in real time and predicts changes in the GSNR distribution of the output signal, thus accurately predicting the GSNR distribution under dynamic input scenarios. By combining intelligent optimization algorithms to dynamically adjust the pump configuration parameters of the Raman amplifier, real-time sensing and precise control of the transmission system's GSNR are achieved, thereby improving the average GSNR and flatness of the channel in the optical network and significantly increasing transmission capacity. This method improves the transmission capacity of dynamic multi-band optical networks, effectively solves the adaptability problem of traditional methods in dynamic scenarios, and significantly improves the performance and stability of optical networks.

Description

Technical Field

[0001] This invention belongs to the field of optical communication technology, specifically relating to the optimization of optical network transmission performance, and in particular to a method for optimizing transmission capacity in multi-band optical networks based on generalized signal-to-noise ratio (GSNR) sensing. Background Technology

[0002] With the development of new technologies such as the Internet, cloud computing, and 5G / 6G, the demand for data traffic is exploding, and existing optical communication networks are facing enormous pressure to expand their capacity. Traditional single-band optical networks are limited in terms of capacity improvement, while multi-band optical networks have become an important solution for future transmission capacity expansion. By introducing new bands, optical networks can utilize the entire communication window of standard single-mode fiber (SSMF), significantly expanding transmission bandwidth. However, in multi-band optical networks, the transmission performance of the signal channel is affected by complex factors such as nonlinear effects, amplified spontaneous emission noise (ASE), and stimulated Raman scattering (SRS), resulting in significant unevenness in the generalized signal-to-noise ratio, thus limiting further improvements in overall transmission capacity.

[0003] Raman amplifiers, as optical amplifiers based on nonlinear optical effects, have become an important technology choice in multi-band optical networks due to their wide gain bandwidth and low noise figure. Compared with erbium-doped fiber amplifiers (EDFAs), Raman amplifiers offer greater flexibility, allowing for arbitrary gain profile configurations across multiple bands by adjusting the wavelength and power of the pump light. However, traditional Raman amplifier configuration methods are typically based on fixed signal input conditions, failing to consider the real-time changes in signal transmission in dynamic multi-band optical networks. Existing research often employs numerical modeling and simulation methods to optimize the gain profile of Raman amplifiers, but these methods primarily focus on improving amplifier gain performance, neglecting the importance of GSNR optimization for increasing transmission capacity.

[0004] In recent years, with the development of machine learning technology, modeling methods based on artificial neural networks (ANNs) have been increasingly applied to the pump design of Raman amplifiers. By training models, researchers can establish a mapping relationship between pump parameters and Raman gain profiles, providing support for amplifier optimization. However, these methods still have limitations: on the one hand, they fail to directly optimize transmission capacity for multi-band optical networks; on the other hand, traditional models are usually trained on static input signals, lacking adaptability to dynamic signal scenarios. When signal conditions change, the model's predictive ability and optimization performance will significantly decrease.

[0005] In multi-band optical networks, the flatness and average value of GSNR are crucial factors affecting transmission capacity. Uneven GSNR distribution across channels can lead to a bottleneck effect on system performance, making it difficult to increase the transmission rate of some channels and thus limiting the overall capacity growth. Therefore, optimizing the GSNR distribution of multi-band optical networks under dynamic input conditions to improve system transmission capacity is a significant challenge in current optical communication research.

[0006] To address the aforementioned issues, a method is urgently needed that can sense GSNR distribution and dynamically optimize the transmission performance of multi-band optical networks. This method should comprehensively consider the synergistic effects of dynamically changing input signal characteristics and pump power configuration, and utilize an efficient optimization mechanism to achieve real-time adjustment of Raman amplifier configuration, thereby improving the overall transmission capacity of the multi-band optical network. Such methods should overcome the limitations of traditional technologies in static scenarios, possessing good adaptability and scalability to meet the practical needs of efficient operation and capacity expansion of optical networks in various dynamic transmission environments. Summary of the Invention

[0007] Objective: This invention aims to address the problem of limited overall transmission capacity in multi-band optical networks due to uneven generalized signal-to-noise ratio (GSNR) distribution and insufficient optimization. By proposing a dynamic multi-band optical network transmission capacity optimization method based on GSNR sensing, this invention aims to optimize the Raman amplifier configuration in dynamic transmission scenarios, achieving real-time sensing and precise control of GSNR, thereby improving the average GSNR and flatness of each channel in the optical network and significantly increasing transmission capacity. This invention constructs a GSNR sensing model to accurately predict GSNR distribution changes under dynamic input signal conditions and combines it with intelligent optimization algorithms to dynamically adjust the Raman amplifier pump configuration, ensuring globally optimal transmission performance in various dynamic scenarios. This invention aims to reduce the complexity of the optimization process while improving optimization efficiency, providing efficient support for the rapid deployment and operation of multi-band optical networks.

[0008] Technical solution: This invention provides a method for optimizing the transmission capacity of dynamic multi-band optical networks based on GSNR sensing, the method comprising:

[0009] By constructing a dynamic signal transmission model for a multi-band optical network, input signal GSNR curves and their corresponding pump configuration parameters under different transmission scenarios are collected to generate a dataset for modeling and optimization. The dataset includes dynamically changing input GSNR curves, Raman amplifier pump power configurations, and output GSNR curves, which are used for subsequent model training and algorithm optimization. To improve the efficiency of modeling and optimization, the data generation process uses feature dimensionality reduction to group the channels, mapping the multi-band GSNR data to several key channel feature sets, ensuring reasonable data dimensionality while covering the typical characteristics of GSNR data in multi-band dynamic scenarios.

[0010] An artificial neural network model for dynamic GSNR sensing is constructed, which realizes a nonlinear mapping relationship between the input signal GSNR curve and the pump configuration parameters to the output GSNR curve. The modeling process includes model architecture optimization and parameter training. First, in terms of model architecture, by scanning different numbers of hidden layers and node sizes, and combining different activation functions, a model structure capable of accurately predicting the GSNR distribution in multi-band signal transmission is designed. Second, by optimizing training parameters (such as learning rate and optimizer), the model is trained and validated using a dataset generated from multi-band optical network experimental simulations, ultimately obtaining a GSNR sensing model suitable for dynamic input scenarios. The model training process calculates the loss function based on the predicted output GSNR curve and the true value. The training objective is to minimize the error between the two to ensure that the model has good generalization ability in dynamic multi-band optical networks.

[0011] Based on the constructed GSNR sensing model, this invention proposes a pump configuration optimization method for dynamic scenarios, aiming to improve transmission capacity. Specifically, an optimization objective function is designed to maximize the average value and distribution flatness of GSNR, and a smart optimization algorithm is used to perform global search and parameter optimization of the Raman amplifier pump power configuration. The optimization process includes the following steps: In the initialization phase, multiple candidate pump configurations are generated as an initial solution set based on a multi-dimensional parameter space; during the iterative optimization process, the search direction and step size of the candidate solutions are dynamically adjusted to guide the solution set towards the optimal region; in each iteration, the objective function value is calculated in real time based on the GSNR sensing model, the fitness of the candidate solutions is evaluated, and finally, the optimal pump configuration that maximizes the objective function value is selected. Through the adaptive search mechanism of this algorithm, the coordinated optimization of pump parameters and GSNR distribution in dynamic scenarios is achieved.

[0012] This optimization method is applied to the specific transmission scenario of a dynamic multi-band optical network. A GSNR sensing model is used to analyze the real-time input signal GSNR distribution, and a particle swarm optimization algorithm is combined to dynamically adjust the pump power configuration, thereby effectively improving the generalized signal-to-noise ratio (SNR). During the optimization process, the model provides real-time feedback, enabling the pump configuration to dynamically adapt to changes in the input signal, achieving optimization and flattening of the GSNR distribution between channels in the multi-band optical network, thus improving the overall transmission capacity of the system.

[0013] This invention provides a dynamic multi-band optical network transmission capacity optimization method based on GSNR sensing, applicable to various optical network scenarios, including but not limited to ultra-long-distance optical transmission systems. Due to the large number of transmission channels and strong spectral unevenness, dynamic multi-band optical networks have extremely high demands for transmission capacity optimization. This invention, by combining GSNR sensing with intelligent optimization algorithms, provides an efficient and highly adaptable optimization method for these transmission scenarios.

[0014] In some embodiments, the optimization algorithm employs a particle swarm optimization (PSO) method, where each particle in the swarm corresponds to a set of pump power configurations, and the fitness function is the objective function value. Through dynamic updates of particle position and velocity, the swarm gradually converges to the optimal solution. During the optimization process, the algorithm utilizes real-time feedback provided by the GSNR sensing model to dynamically combine the adjustment of pump power with the distribution characteristics of GSNR between multi-band optical network channels, thereby achieving global optimization of inter-channel transmission performance.

[0015] Beneficial Effects: Compared with the prior art, the present invention has the following advantages: (1) Dynamic optimization and high transmission capacity improvement: The present invention, through the combination of GSNR sensing model and intelligent optimization algorithm, can adapt to the signal change conditions of dynamic multi-band optical networks in real time and optimize the GSNR distribution between channels. Under limited resource conditions, it significantly improves the average value and flatness of GSNR between channels, overcomes the problem of insufficient adaptability of traditional methods in dynamic transmission scenarios, and achieves higher transmission capacity for multi-band optical networks. (2) Real-time feedback and efficient configuration: The method of the present invention analyzes the dynamic input conditions in real time through GSNR sensing model and uses the feedback results to guide the optimization algorithm to adjust the pump configuration parameters. Users can quickly adjust the configuration according to the optimization results, realize efficient optimization in dynamic transmission scenarios, and provide a guarantee for the stable operation of multi-band optical networks.

[0016] This invention discloses a dynamic multi-band optical network transmission capacity optimization method based on GSNR sensing. By constructing an accurate GSNR sensing model and optimization algorithm, it achieves global optimization of inter-channel transmission performance in dynamic multi-band optical networks, effectively improving the system's transmission capacity. Compared with existing technologies, this method overcomes the limitation of being applicable only to static transmission scenarios. It can not only optimize the pump configuration of Raman amplifiers in real time under dynamic input conditions, but also significantly improve the flatness and average value of GSNR distribution, providing strong technical support for the efficient deployment and operation of multi-band optical networks. Attached Figure Description

[0017] Figure 1 This is a flowchart of the dynamic multi-band optical network transmission capacity optimization method based on GSNR sensing of the present invention.

[0018] Figure 2 This is a schematic diagram of the dynamic multi-band optical network transmission capacity optimization framework based on GSNR sensing of the present invention.

[0019] Figure 3 This is a comparison chart of the GSNR prediction results of the GSNR sensing model of this invention and its comparative model in a dynamic transmission scenario.

[0020] Figure 4 This is a comparison chart showing the capacity optimization results of the GSNR-based sensing optimization method and the comparative scheme in a dynamic transmission scenario. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative effort are all within the scope of protection of the present invention.

[0022] In the context of the rapid development of information technology, the widespread application of cloud computing, the Internet of Things, and 5G / 6G communications has led to an exponential increase in data traffic. Optical networks, as the core carrier platform of the information superhighway, face unprecedented demands for expanded transmission capacity. Traditional single-band optical networks are gradually failing to meet this trend, prompting multi-band optical networks to become the mainstream. To further improve the transmission efficiency and capacity of optical networks, it is necessary to overcome the limitations caused by the uneven transmission performance between channels. Generalized signal-to-noise ratio (GSNR), as an important indicator for measuring the transmission performance of optical channels, directly affects the upper limit of the overall capacity of the optical network system due to its balanced distribution. Therefore, achieving accurate sensing and dynamic optimization of GSNR in multi-band optical networks has become crucial for improving transmission capacity and system stability.

[0023] This invention focuses on the dynamic sensing and optimization of the generalized signal-to-noise ratio (GSNR) distribution in multi-band optical networks. Specifically, by adjusting the pump configuration parameters of the Raman amplifier, it aims to achieve GSNR equalization for multi-band signals during fiber optic transmission, thereby effectively improving the overall system transmission capacity and stability. In multi-band fiber optic transmission systems, due to nonlinear effects and the influence of stimulated Raman scattering (SRS) and amplified spontaneous emission noise (ASE), the transmission performance of different band channels exhibits significant differences, leading to uneven GSNR distribution, which becomes a major bottleneck restricting system performance optimization. The Raman amplifier, as an important means of achieving gain compensation in multi-band transmission, possesses high flexibility and programmability; its pump power and wavelength configuration determine the system gain profile. However, due to the dynamic changes in signal conditions, traditional optimization methods based on fixed configurations or static modeling struggle to quickly respond to changes in the actual transmission environment, resulting in poor adaptability and unstable optimization effects. To address this, this invention provides a dynamic multi-band optical network transmission capacity optimization method based on GSNR sensing. By constructing a high-precision sensing model, it senses input signal characteristics in real time and predicts output GSNR changes. Combined with an intelligent optimization algorithm, it dynamically adjusts pump configuration parameters to achieve optimal control of transmission performance. This invention effectively improves the optimization efficiency and adaptability of multi-band optical networks in dynamic transmission scenarios. The relevant technical process is as follows: Figure 1 As shown, the specific steps include S100 to S400.

[0024] Step S100: Construct a dataset for GSNR sensing modeling; the dataset is used to train an artificial neural network model to achieve high-precision prediction of the channel transmission performance of dynamic multi-band optical networks.

[0025] In the specific implementation process, the construction of the dataset includes the following steps: building a multi-band optical network simulation transmission platform to simulate different dynamic transmission scenarios; randomly setting the GSNR distribution of the input signal and the pump power configuration of the Raman amplifier, and successively collecting the corresponding output GSNR distribution curves to form a training dataset and a test dataset containing input GSNR features, pump configuration parameters and corresponding output GSNR distributions.

[0026] In the specific implementation process, the multi-band optical network transmission scenarios built by the simulation platform include, but are not limited to, the following scenarios: multiplexing transmission scenarios with different numbers of bands, transmission scenarios with different fiber optic link lengths, and scenarios with different input signal power levels.

[0027] In the specific implementation process, in the dataset construction step, in order to reduce the model training complexity and improve the model's generalization ability, the multi-band GSNR data is processed by channel grouping and feature dimensionality reduction to extract a typical set of GSNR distribution features in order to retain the key transmission characteristics of the data.

[0028] In this embodiment, the simulated link is a hybrid amplification link structure covering the C+L band. This link structure is configured with a Raman amplifier as the pumping mode, using a reverse pumping configuration. The fiber optic link connected to the Raman amplifier is 80 km long. To fully cover the typical characteristics of dynamic multi-band signal transmission, the signal covers 80 channels with a channel spacing of 150 GHz, a signal baud rate of 128 GBaud, and a flat transmit power spectrum.

[0029] In this embodiment, the dynamic transmission scenario is constructed by setting 1 to 10 EDFA (Erbium-doped Fiber Amplifier) ​​segments as amplification units to simulate the optical signal amplification process under various actual transmission conditions. The simulation parameters include fiber link lengths of 40km to 120km, a spacing of 40km, a fiber attenuation coefficient of 0.2dB / km, a dispersion coefficient of 16.7ps / (nm·km), and a nonlinearity coefficient of 1.265(nm·km). -1 The dataset was collected using the open-source optical network simulation platform GNPy. A total of 25,100 data samples were generated using the simulation configuration described above, of which 80% were used for neural network model training and 20% for model validation.

[0030] In this embodiment, during the data dimensionality reduction process, in order to reduce the dimensionality of the input features to meet the modeling requirements of the artificial neural network, the GSNR features of the 80 channels are divided into groups of 5 adjacent channels, thereby obtaining 16 groups of GSNR dimensionality reduction features, which constitute the input of the neural network model.

[0031] Step S200: Construct a GSNR sensing model and train the model; the GSNR sensing model is implemented using an artificial neural network (ANN) and aims to establish a nonlinear mapping relationship between the dynamic input GSNR distribution, Raman pump configuration parameters and the output GSNR distribution.

[0032] In its implementation, the artificial neural network model structure comprises an input layer, multiple hidden layers, and an output layer. The input layer primarily receives GSNR feature data generated in step S100, including the mean and standard deviation of the GSNR of the dynamic signal, as well as Raman pump configuration parameters. The number of hidden layers and the node size are optimized based on prediction errors and generalization capabilities during training. Each hidden layer processes the input signal using an activation function (such as ReLU or Sigmoid) to help the model capture complex nonlinear relationships in the data. The output layer generates the model's predicted output GSNR distribution, aiming to optimize the performance of the optical network through this distribution.

[0033] In this embodiment, the training process employs the standard backpropagation (BP) algorithm to iteratively optimize the model parameters. In each training step, the error value is calculated based on the loss function, which is the mean squared error between the predicted and actual values. The weights and biases in the model are adjusted using the gradient descent algorithm. During training, the optimization objective is to minimize the value of the loss function, thereby making the model's predictive ability increasingly accurate and achieving a better fit. To further improve the model's performance, the Adam optimizer is chosen, which combines the advantages of gradient descent and automatically adjusts the learning rate of each parameter, thus improving training efficiency and accuracy.

[0034] In this embodiment, the root mean square error (RMSE) is used as the error evaluation criterion, but other error measures such as mean absolute error (MAE) or coefficient of determination (R²) are also used. 2 It can also be used for model performance evaluation. MAE is suitable for datasets with relatively uniform bias, while R... 2 Used to measure the model's fit. Depending on the specific application requirements, technicians can choose an appropriate error metric to optimize the model's predictive accuracy.

[0035] In the specific implementation process, to avoid overfitting and improve the model's generalization ability, regularization and Dropout strategies were introduced during training. Regularization constrains the model's weights, preventing overfitting on the training dataset and thus improving the model's adaptability to unknown data. Dropout further prevents overfitting by randomly discarding some neurons during training and effectively enhances the model's robustness. To ensure the model's training stability and prediction accuracy, hyperparameter optimization was performed during training; parameters such as the learning rate and optimizer type were determined through cross-validation and grid search.

[0036] Step S300: Optimize dynamic pump configuration based on optimization algorithm; the optimization algorithm adopts particle swarm optimization (PSO), each particle in the particle swarm corresponds to a set of pump power configurations, and the fitness function is defined as the optimization objective function of GSNR distribution.

[0037] The optimization process includes the following steps: First, the current particle configuration is evaluated using the GSNR perception model. Fitness values ​​are calculated based on model feedback, and the configuration effectiveness is assessed. Then, in each iteration, the particle's position and velocity are dynamically adjusted to guide it closer to the global optimum. Finally, the optimal particle in the current population is evaluated based on the optimization objective function, and the pump configuration corresponding to the particle with the highest fitness value is output as the optimal solution. An optimization threshold and termination condition are set during the iteration process. Optimization stops and the current optimal solution is output when the fitness value reaches the preset threshold or the set number of iterations is completed.

[0038] The specific optimization objective function is as follows:

[0039]

[0040] Where i and j are the channel under consideration and the exponent of the Raman amplifier pump, respectively. Gin,i is the GSNR of the input signal, and Pj is the pump power of the Raman amplifier. and These are the mean and standard deviation of the output GSNR, respectively, in dB.

[0041] Step S400: Apply the optimal pump power configuration parameters obtained by the particle swarm optimization algorithm to the actual dynamic transmission scenario of multi-band optical network.

[0042] In practical implementation, the dynamic GSNR distribution characteristics of the input signal are monitored in real time. A GSNR sensing model is used to predict the output GSNR distribution under the current input conditions. Combined with the feedback results of the optimization algorithm, the Raman amplifier pump configuration parameters are dynamically adjusted to achieve real-time optimized control of the inter-channel GSNR. In actual transmission verification, the optimized pump configuration obtained using this method is compared with the traditional fixed configuration method. Experimental results show that the method of this invention can significantly improve the average GSNR and distribution flatness of the dynamic multi-band optical network transmission system, thereby effectively improving the overall transmission capacity and stability of the system.

[0043] In this embodiment, 100 sets of representative dynamic input signals were randomly generated. These dynamic input signals covered multiple channels within the C+L band and included different signal power levels and fiber link length parameters to simulate the typical characteristics of a real multi-band optical network under different transmission scenarios. Using these signals, the output GSNR distribution was predicted using the optimization methods described in this invention based on the GSNR sensing model and the traditional static model, respectively. The corresponding root mean square error (RMSE) was calculated to evaluate the prediction performance of the two models under dynamic conditions. Experimental results show that under the conditions of these 100 sets of dynamic input signals, the prediction RMSE of the GSNR sensing model described in this invention is 0.58 dB, while the prediction RMSE of the traditional static model is 6.50 dB.

[0044] In an optimization case study of a dynamic transmission scenario, 200 different sets of dynamic input signals were randomly generated based on the constructed multi-band optical network simulation platform. Each set of input signals contained multi-channel GSNR features covering the C+L band range, used to comprehensively evaluate the adaptability and effectiveness of the optimization method of this invention under variable scenarios. For each set of input signals, comparative experiments were conducted using the optimization method based on the GSNR sensing model described in this invention and the traditional static configuration method.

[0045] During the experiment, for each input signal, the output GSNR distribution was first predicted based on the GSNR sensing model, and the optimal pump power configuration was determined using the particle swarm optimization (PSO) algorithm. The optimized configuration was then applied to the simulation transmission platform to obtain the actual output GSNR distribution, which was compared with the prediction results of the sensing model to verify the accuracy of the model's prediction. Simultaneously, pump configuration was performed using traditional methods, and the output GSNR distribution was recorded.

[0046] To quantify the performance differences between the two methods in GSNR optimization, statistical analysis was performed on the mean and standard deviation of the output GSNR distribution. The average GSNR improvement and flatness improvement rate were used as evaluation indicators. In this embodiment, GSNR flatness is characterized by the difference between the maximum and minimum GSNR values. Experimental results show that the optimization method based on the GSNR sensing model outperforms the traditional method in all 200 sets of input signals.

[0047] Specifically, the optimization method described in this invention significantly improved the average output GSNR in all test scenarios, with an average improvement of 1.08 dB. Simultaneously, the flatness of the inter-channel GSNR was also significantly improved, with a reduction in standard deviation corresponding to an average flatness improvement of 40.02%. Figure 3 The results demonstrate the prediction performance of the GSNR perception model in dynamic scenes, showing higher fitting accuracy and generalization ability compared to traditional models. Figure 4 The comparison of optimization results for 200 sets of input signals clearly demonstrates the advantages of the method of the present invention in improving GSNR performance.

[0048] Based on the above experimental results, it can be seen that the dynamic multi-band optical network transmission capacity optimization method based on GSNR sensing proposed in this invention can effectively cope with the adaptive adjustment requirements of pump configuration under complex dynamic input conditions. It not only significantly improves the transmission performance of the system, but also has good generalization ability and adaptability, and is suitable for various practical optical network deployment scenarios.

[0049] Those skilled in the art can implement the various exemplary components and methods described in the embodiments disclosed in this invention in hardware, software, or a combination of both. They can also adjust and optimize the technical solutions according to specific application requirements; however, these adjustments and optimizations should not be considered as exceeding the scope of protection of this invention. The technical features defined in the claims of this invention can be applied individually or in any combination.

[0050] It should be noted that the terms "comprising" or "including" as used in this specification do not exclude the presence of other unlisted elements or steps, and their scope should have the same meaning as "including but not limited to". Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

[0051] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of the invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, and improvements made without departing from the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing the transmission capacity of a dynamic multi-band optical network based on GSNR sensing, characterized in that, By using the GSNR distribution of the input signal as the input dimension of a data-driven prediction model, combined with the pump power configuration parameters of the Raman amplifier, the signal transmission performance in dynamic transmission scenarios is optimized, thereby improving the transmission capacity of the entire optical network. The method includes the following steps: (1) Construct a dataset containing the dynamic signal GSNR distribution and Raman amplifier pump configuration parameters, extract GSNR features of specified channels, and generate the feature dataset required to construct the prediction model; (2) Establish a GSNR sensing and prediction model. Based on the dynamic signal GSNR distribution and pump power configuration parameters in the input feature data, output the predicted GSNR distribution. This model provides real-time feedback for the optimization process. (3) Based on the real-time feedback of the GSNR sensing and prediction model, an intelligent optimization algorithm is used to perform a global search of the multi-dimensional pump power parameter space to generate the optimal configuration scheme for Raman pumping under the current transmission scenario. (4) Apply the optimized pump configuration to the actual transmission scenario, and improve the balance of GSNR between channels through closed-loop feedback control, thereby achieving optimized improvement of transmission capacity.

2. The method for optimizing the transmission capacity of a dynamic multi-band optical network based on GSNR sensing according to claim 1, characterized in that, The designated channel refers to the target channel that needs to be prioritized for optimization in dynamic transmission scenarios, ensuring that the optimization process focuses on the set of key channels that have the greatest impact on the overall capacity in the current transmission scenario.

3. The method for optimizing the transmission capacity of a dynamic multi-band optical network based on GSNR sensing according to claim 1, characterized in that, The feature dataset includes multi-band signal GSNR distribution data, Raman amplifier pump power configuration vector and corresponding output GSNR distribution data under dynamic transmission scenarios, which are obtained from actual transmission scenarios or simulated transmission scenarios.

4. The method for optimizing the transmission capacity of a dynamic multi-band optical network based on GSNR sensing according to claim 1, characterized in that, The GSNR sensing and prediction model is constructed based on a data-driven modeling method. The input dimension expands the GSNR distribution information of the signal to improve the model's ability to sense and predict generalized signal-to-noise ratio changes in multi-band optical networks, thereby ensuring the optimization effect of transmission capacity in dynamic scenarios.

5. The method for optimizing the transmission capacity of a dynamic multi-band optical network based on GSNR sensing according to claim 1, characterized in that, The GSNR sensing prediction model establishes a mapping relationship between the GSNR distribution of the system input and output signals and the pump configuration parameters, providing a high-precision prediction basis for subsequent optimization.

6. The method for optimizing the transmission capacity of a dynamic multi-band optical network based on GSNR sensing according to claim 1, characterized in that, In the step of calculating the loss function based on the predicted output GSNR curve and the true value in the GSNR perception prediction model, the loss function adopts mean square error, root mean square error, mean absolute error or maximum error.

7. The method for optimizing the transmission capacity of a dynamic multi-band optical network based on GSNR sensing according to claim 1, characterized in that, The intelligent optimization algorithm achieves the optimal pump power configuration to improve transmission capacity by dynamically updating the search strategy in the parameter space, combining real-time feedback from the GSNR sensing model, and using the objective function as the evaluation criterion.

8. The intelligent optimization algorithm according to claim 1, characterized in that, The objective function for optimization is defined as a combination of the average GSNR and flatness between channels, where flatness is represented by the standard deviation of GSNR, specifically as follows: Where i and j are the channels under consideration and the exponent of the Raman amplifier pump, respectively; Gin,i is the GSNR of the input signal, and Pj is the pump power of the Raman amplifier; and These are the mean and standard deviation of the output GSNR, respectively, in dB.

9. The method for optimizing the transmission capacity of a dynamic multi-band optical network based on GSNR sensing according to claim 1, characterized in that, The method constructs a closed-loop feedback control system to monitor the changes in the GSNR distribution of the input signal in dynamic transmission scenarios in real time, and periodically triggers the optimization algorithm to update the pump configuration parameters, ensuring that the multi-band optical network maintains high transmission capacity and stability during long-term operation.

10. The closed-loop feedback control system according to claim 9, characterized in that, The system dynamically combines the prediction results of the GSNR sensing model with the output of the intelligent optimization algorithm through iterative optimization and real-time configuration adjustment, forming an adaptive transmission capacity optimization closed loop.

Images