Optical transmission network performance prediction method, system, electronic device, and storage medium

Abstract

The embodiments of the present application relate to the field of communication and disclose an optical transmission network performance prediction method, a system, an electronic device, and a storage medium. In the present application, the optical transmission network performance prediction method includes the steps of: acquiring a performance prediction model; acquiring model update information of the performance prediction model by a horizontal federated learning technique, the model update information being obtained after training the performance prediction model based on performance data of each network element in a single domain, and being used to update the performance prediction model; reporting the model update information and the single domain topology information to a multi-domain management server, and having the multi-domain management server generate an optical transmission network digital twin model for making performance prediction for the optical transmission network according to the model update information and the single domain topology information.

Description

[Technical field]

[0001] (CROSS REFERENCE TO RELATED APPLICATIONS) This application claims priority to a Chinese patent application filed on March 10, 2022 and bearing application number 202210234032.2.

[0002] The embodiments of the present application relate to the field of communications, and in particular to an optical transmission network performance prediction method, system, electronic device, and storage medium. [Background technology]

[0003] With the development of communication technology, the global communication industry is moving from the Internet era and cloud era to the smart era, and new opportunities and new challenges will drive the acceleration of comprehensive network transformation and upgrade. In order to realize the digital transformation of carrier networks and provide vertical industries and consumer users with zero-latency, zero-contact, zero-fault services and user experiences, autonomous network technology is proposed to build self-configuration, self-repair, and self-optimization network capabilities in the entire life cycle of carrier network operation. To realize the digital transformation of networks, it is necessary to establish digital mirroring of the internal and external environments of the network and integrate simulation and preventive prediction and other capabilities. How to build such a simulation network is an issue that needs to be resolved quickly. Summary of the Invention [Problem to be solved by the invention]

[0004] A main objective of the embodiments of the present application is to propose an optical transmission network performance prediction, a system, an electronic device, and a storage medium that can realize a simulation network construction in which an optical transmission network is constructed. [Means for solving the problem]

[0005] In order to achieve the above object, an embodiment of the present application provides an optical transmission network performance prediction method for use in a single-domain management server, the method including: obtaining a performance prediction model; obtaining model update information for the performance prediction model through a horizontal federated learning technique, where the model update information is obtained after training the performance prediction model based on performance data of each network element in a single domain, and is used to update the performance prediction model; and reporting the model update information and the single-domain topology information to a multi-domain management server, and having the multi-domain management server generate an optical transmission network digital twin model for performing performance prediction for the optical transmission network based on the model update information and the single-domain topology information.

[0006] In order to achieve the above objective, an embodiment of the present application further provides an optical transmission network performance prediction method for use in a multi-domain management server, the method including: obtaining model update information and single-domain topology information reported by a single-domain management server, where the model update information is obtained after training the performance prediction model of the single-domain management server based on the performance data of each network element in the single domain through a horizontal federated learning technology, and is used to update the performance prediction model; and generating an optical transmission network digital twin model for making performance prediction for the optical transmission network based on the model update information and the single-domain topology information.

[0007] In order to achieve the above object, an embodiment of the present application further provides an optical transmission network performance prediction system, the system including: a single-domain management server; and a multi-domain management server, the single-domain management server being communicatively connected to the multi-domain management server, wherein the single-domain management server obtains a performance prediction model; and obtains model update information of the performance prediction model through a horizontal federated learning technique, the model update information being obtained after training the performance prediction model based on the performance data of each network element in the single domain; and reports the model update information and the single-domain topology information to the multi-domain management server, the multi-domain management server being used to generate an optical transmission network digital twin model, the multi-domain management server obtains the model update information and the single-domain topology information reported by the single-domain management server, the model update information being obtained after training the performance prediction model of the single-domain management server based on the performance data of each network element in the single domain through a horizontal federated learning technique; and is used to generate an optical transmission network digital twin model for performance prediction for the optical transmission network based on the model update information and the single-domain topology information.

[0008] An embodiment of the present application further provides an electronic device, the electronic device including at least one processor and a memory communicatively connected to the at least one processor, the memory storing instructions executable by the at least one processor, the instructions being executable by the at least one processor to cause the at least one processor to execute the optical transport network performance prediction method described above.

[0009] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, realizes the above optical transport network performance prediction method.

[0010] The optical transmission network performance prediction method proposed in this application is such that the single-domain management server obtains model update information and reports it to the multi-domain management server, and the multi-domain management server finally generates an optical transmission network digital twin model, that is, the training difficulty problem caused by data distribution in each single domain through horizontal associative learning is overcome. Since the model update information is obtained after training the performance prediction model based on the performance data of each network element in the single domain, the update of the performance prediction model based on the model update information can be more consistent with the actual processing capacity of the actual network element. Therefore, the single-domain management server reports the model update information and the single-domain topology information to the multi-domain management server, and the multi-domain management server can generate an optical transmission network digital twin model based on the updated performance prediction model. This optical transmission network digital twin model is also a model that is consistent with the actual processing and transmission capacity of the optical transmission network, thereby realizing the simulation network construction in which the optical transmission network is constructed. [Brief description of the drawings]

[0011] [Figure 1] 1 is a flow diagram illustrating an optical transport network performance prediction method according to an embodiment of the present application used in a single domain management server; [Diagram 2] FIG. 1 is a schematic diagram of an optical transport network performance prediction system architecture according to an embodiment of the present application. [Diagram 3] FIG. 2 is a schematic diagram of an optical transport network performance prediction system architecture according to an embodiment of the present application. [Figure 4] FIG. 2 is a schematic diagram of a network element structure according to an embodiment of the present application; [Diagram 5] FIG. 2 is a schematic diagram of a performance prediction model according to an embodiment of the present application. [Figure 6] FIG. 2 is a schematic diagram of performance data sampling according to an embodiment of the present application. [Figure 7]FIG. 1 is a schematic flow diagram of a method for training a performance prediction model according to an embodiment of the present application. [Figure 8] FIG. 2 is a schematic flow diagram of a method for training a performance prediction model according to an embodiment of the present application; [Figure 9] FIG. 3 is a schematic flow diagram of a performance prediction model training method according to an embodiment of the present application; [Figure 10] FIG. 4 is a schematic flow diagram of a performance prediction model training method according to an embodiment of the present application. [Figure 11] 1 is a flow diagram illustrating an optical transport network performance prediction method according to an embodiment of the present application used in a multi-domain management server; [Figure 12] FIG. 1 is a schematic structural diagram of an optical transmission network performance prediction system according to an embodiment of the present application. [Figure 13] FIG. 2 is a schematic diagram of an electronic device structure according to an embodiment of the present application. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the following describes each embodiment of the present application in detail in conjunction with the accompanying drawings. However, in order to be understood by those skilled in the art, many technical details are provided in each embodiment of the present application to allow the reader to better understand the present application. However, the technical solution claimed for protection in the present application can be realized without these technical details and various changes and modifications based on the following embodiments. The division of the following embodiments is for convenience of description and does not constitute any limitation on the specific implementation of the present application, and the embodiments can be combined and referenced with each other on the premise that there is no contradiction.

[0013] An embodiment of the present application relates to a method for predicting performance of an optical transport network, which, as shown in FIG. 1, includes the following steps:

[0014] Step 101, obtain a performance prediction model; Step 102, by using a horizontal federated learning technique to obtain model update information of a performance prediction model, where the model update information is obtained after training the performance prediction model based on the performance data of each network element in a single domain, and is used to update the performance prediction model; Step 103: Report the model update information and the single-domain topology information to a multi-domain management server, and have the multi-domain management server generate an optical transmission network digital twin model for performance prediction of the optical transmission network based on the model update information.

[0015] The optical transmission network performance prediction method of this embodiment is used in a single domain management server, and the single domain management server is a management server that manages each network element in one domain in the optical transmission network, and each single domain management server can run a single domain management control system of a manufacturer corresponding to the single domain. An optical transport network (referred to as "OTN") can be horizontally divided into different management domains, and a single management domain among them may be configured by OTN equipment of a single equipment vendor, or may be configured by a network or sub-network of an operator. A single management domain is a single domain, and each single domain has a single domain management server for communicating with a multi-domain management server, and a multi-domain organization system can be run on the multi-single domain management server to generate an optical transmission network digital twin model. The optical transmission network digital twin model can perform performance prediction for the optical transmission network, or perform other data simulation demands.

[0016] The global communications industry is moving from the Internet era and cloud era to the smart era, and new opportunities and new challenges are driving the acceleration of the comprehensive transformation and upgrade of networks. Against this background, the concept of autonomous networks (AN) was put forward in 2019. The AN concept aims to enable operators to transform their networks digitally through the integration of network technology and digital technology, provide zero-latency, zero-contact, zero-fault services and user experiences for vertical industries and consumer users, and build self-configuring, self-repairing, and self-optimizing network capabilities throughout the entire life cycle of operator network operation. To realize AN, digital twin (DT) technology has been established and is considered by the industry as the foundation for digital transformation and an important support and component for realizing AN architecture and technology. Network digital twins can accurately sense their own and external environmental conditions, establish digital mirroring of internal and external environments, and integrate simulation and preventive prediction, playing an important enabling technological role in cases such as "low-cost trial and error," "smart decision-making," and "predictive maintenance."

[0017] Digital twin, or DT technology, may be understood as constructing a digital (virtual) model of a physical object in a digital space, continuously modifying the model using data from the physical object, updating the state of the model, matching it with the physical object throughout its entire life cycle, and faithfully mirroring the real-time operating state of the physical object, and this model becomes a digital twin of the physical object (called a digital twin). Monitoring, analysis, prediction, diagnosis, training, and simulation can be performed based on the digital twin, and the simulation results can be fed back to the physical object to help optimize and make decisions for the physical object. Technologies related to digital twin model structure, real-time updates on the digital model state, simulation analysis and control decisions by digital twin, etc. may be collectively referred to as digital twin DT technology. From this, it can be seen that how to build a DT model by linking it with the structural features, state features, and DT application scenarios of the physical object mirrored in the twin is an important technology of DT.

[0018] In the autonomous network system architecture, in the complex network environment of OTN multi-domain, it is necessary to build a DT network model and generate an OTN DT network layer to realize the analysis ability of the entire network OTN through DT technology. The DT network layer is an abstraction of the entire inter-domain physical network model, so that the communication between DT network elements is not limited by the physical network space, and the visibility and operability of DT network elements is not limited by the space restrictions such as physical network domain division and management control restrictions. This requires that the OTN DT model has the general abstract description ability for the physical network elements and network element device function mechanisms of each domain.

[0019] Take the modeling of an OTN performance prediction algorithm belonging to a DT sensing system algorithm model in a single-domain OTN network domain as an example. Although the network element equipment comes from different equipment manufacturers, and each network element equipment has the same network element operation function in the same network domain environment, the attributes of each network element itself such as device lot, model number, service life and so on are not the same, and the optical performance of equipment devices, optical fibers, etc. changes in actual operation use according to different rules. The DT performance prediction model obtained by sampling and training only some network elements or a single network element device only reflects the performance change characteristics of this part of network elements or this network element, and does not have ubiquity for other network element equipment in a single domain or network element equipment in other inter-domain networks, has poor generalization ability, and is prone to problems such as overfitting during inference, resulting in low accuracy and lack of authority in OTN network performance prediction analysis. However, when the performance data of each network element in each domain is reported to the multi-domain organization system for unified training, there is a problem of data privacy leakage: each OTN physical network domain is based on the single-domain management and control system of different equipment manufacturers, so in the construction and training of the inter-domain digital twin layer model, each physical single domain has data privacy requirements, and it is not easy to directly report the OTN network performance data within each domain to the multi-domain organization system for unified training. How to solve the problem of overfitting of performance prediction function model caused by local training and at the same time solve the problem of sample data privacy leakage caused by centralized training has become a bottleneck in OTNDT case function model modeling and needs to be solved urgently. The emergence of federated learning technology has provided an effective route to solve the above problems.

[0020] The modeling problems that federated learning can overcome are as follows: 1. Data does not leave the local area: Due to data security and privacy protection, data does not leave the local area, which makes it difficult to converge. 2. Model generalization ability: The model generalization ability is poor due to the small size of the training data set. 3. Model training efficiency: The model training efficiency problem is caused by placing too much emphasis on centralized cloud computing training and ignoring the computing power of edge devices.

[0021] The advantages of federated learning are as follows: 1. Only local data training is used, and the data itself is not exchanged, but the model parameters updated by encryption are exchanged. 2. Model training is performed using data from different environments (time, location) of the devices, and the common model is updated and then transmitted to the devices, improving the model generalization ability. 3. Parallel training is performed using the computing power of edge devices, improving the efficiency of model training.

[0022] The flow of the horizontal federated learning includes the following steps: Step 1, each participant calculates the model gradient locally, and uses encryption techniques such as homomorphic encryption, differential privacy or secret sharing to camouflage the gradient information, and sends the camouflaged result (called encrypted gradient) to the aggregation server. Step 2, the server performs a secure aggregation operation, for example, using weighted average based on homomorphic encryption. Step 3, the server sends the aggregated result to each participant. Step 4, each participant decrypts the received gradient and uses the decrypted gradient result to update its respective model parameters. As described above, the present application has established a method mechanism for constructing an OTN DT network function model by adopting a horizontal federated learning technique.

[0023] In the optical transmission network performance prediction method of the present application, the single-domain management server obtains model update information and reports it to the multi-domain management server, and the multi-domain management server finally generates an optical transmission network digital twin model, that is, the training difficulty problem caused by data distribution in each single domain in horizontal associative learning is overcome. Since the model update information is obtained after training the performance prediction model based on the performance data of each network element in the single domain, the update of the performance prediction model based on the model update information can be more consistent with the actual processing capacity of the actual network element. Therefore, the single-domain management server reports the model update information and the single-domain topology information to the multi-domain management server, and the multi-domain management server can generate an optical transmission network digital twin model based on the updated performance prediction model. This optical transmission network digital twin model is also a model that matches the actual processing and transmission capacity in the optical transmission network, thereby realizing the simulation network construction in which the optical transmission network is constructed.

[0024] The implementation details of the optical transmission network performance prediction method of this embodiment will be described in detail below. The following content is only for facilitating understanding of the implementation details provided, and is not essential for implementing this method.

[0025] In step 101, the single-domain management server obtains a performance prediction model. Here, the performance prediction model may be stored in the single-domain management server in advance, or may be obtained by the single-domain management server from the multi-domain management server or another electronic device.

[0026] In one example, as shown in FIG. 2, the optical transmission network performance prediction method may be used in a scenario in which each network element in a single domain has training capability (hereinafter referred to as Scheme A). In Scheme A, the performance prediction model may be a single-domain performance prediction model, and the model update information may be a single-domain digital twin model obtained based on the updated single-domain performance prediction model. The single-domain management server obtains the model update information of the performance prediction model in the following manner: send the single-domain performance prediction model to each network element in the single domain, and obtain parameter update amounts of the single-domain performance prediction model reported by each network element, where each network element trains its single-domain performance prediction model respectively based on local performance data, calculates parameter update amounts, updates the parameters of the single-domain performance prediction model based on the parameter update amounts reported by each network element, to obtain the updated single-domain performance prediction model, and obtains the single-domain digital twin model based on the updated single-domain performance prediction model and the single-domain topology information of the single domain.

[0027] In this embodiment, the single-domain performance prediction model is used as the performance prediction model, and the single-domain digital twin model obtained based on the updated single-domain performance prediction model is used as model update information. The single-domain management server sends the single-domain performance prediction model to each network element in this single domain, and each network element trains its own single-domain performance prediction model, so that the updated single-domain performance prediction model can be adapted to the performance characteristics of each network element. This solves the problems that the network elements themselves have different attributes such as device lot, model number, and service life, resulting in poor generalization ability of the performance prediction model training and easy overfitting during inference, and achieves effects such as strong generalization ability of the performance prediction model and high prediction accuracy.

[0028] In another example, the optical transmission network performance prediction method may be used in a scenario where only a single-domain management server in a single domain has training capability, but other network elements do not have training capability (hereinafter referred to as Scheme B). In Scheme B, the performance prediction model may be a network performance prediction model, and the model update information may be a parameter update amount of the network performance prediction model. The single-domain management server can achieve obtaining the model update information of the performance prediction model in the following manner: train the network performance prediction model according to the performance data of each network element, and obtain the parameter update amount of the network performance prediction model, where the parameter update amount of the network performance prediction model is reported to a multi-domain management server, and is used to allow the multi-domain management server to update the network performance prediction model according to the parameter update amount of the network performance prediction model.

[0029] In this embodiment, the network performance prediction model is taken as the performance prediction model, and the parameter update amount of the network performance prediction model is taken as the model update information, so that the single-domain management server can train the network performance prediction model by itself according to the performance data of each network element, and obtain the parameter update amount of the network performance prediction model; the multi-domain management server integrates the parameter update amount of the network performance prediction model of each single domain and updates the network performance prediction model; the single-domain management server performs model training, which can reduce the number of devices required for training and save computational resources, and can realize the optical transmission network performance prediction method even in the scene where computational resources are relatively scarce. At the same time, the network elements themselves have different attributes such as device lot, model number, and service life, so that the problems such as poor generalization ability of the performance prediction model training and easy overfitting during inference can be solved, and the effects such as strong generalization ability of the performance prediction model and high prediction accuracy can be realized.

[0030] In step 102, the single-domain management server obtains model update information of the performance prediction model, where the model update information is obtained after training the performance prediction model based on the performance data of each network element in the single domain, and is used to update the performance prediction model. The model update information may be calculated by the single-domain management server, or may be calculated by other network elements in the single domain where the single-domain management server is located.

[0031] JPEG2025509385000002.jpg140170

[0032] JPEG2025509385000003.jpg236170

[0033] In step 103, the single-domain management server reports the model update information and the single-domain topology information to the multi-domain management server, and makes the multi-domain management server generate an optical transmission network digital twin model for performance prediction of the optical transmission network based on the model update information and the single-domain topology information. Here, after receiving the model update information and the single-domain topology information reported by the multiple single-domain management servers, the multi-domain management server generates an optical transmission network digital twin model, and the optical transmission network digital twin model is a digital twin model of the entire optical transmission network, i.e., a digital twin model including the multiple single domains reported, and performs a digital-to-analog simulation of the optical transmission network.

[0034] In one example, the parameters of the single-domain performance prediction model are updated according to the parameter update amount reported by each network element, and the single-domain performance prediction model after the update is obtained, and the single-domain management server specifically needs to iteratively update the single-domain performance prediction model, where the iterative update includes: sending the updated single-domain performance prediction model to each network element in the single domain, obtaining the parameter update amount reported again by each network element, updating the single-domain performance prediction model according to the parameter update amount reported again, and starting the next iterative update until an iterative stop condition is met, where the iterative stop condition includes that the parameter update amount of the performance prediction model converges.

[0035] In this embodiment, by repeatedly updating the single-domain performance prediction model, the prediction effect of the single-domain performance prediction model is made to be more consistent with the actual performance of the equipment, and the update of the single-domain performance prediction model is realized, and the optical transmission network digital twin model generated by the multi-domain management server has a better simulation effect.

[0036] JPEG2025509385000004.jpg103170

[0037] JPEG2025509385000005.jpg187170

[0038] JPEG2025509385000006.jpg138170

[0039] JPEG2025509385000007.jpg81170

[0040] JPEG2025509385000008.jpg131170

[0041] JPEG2025509385000009.jpg89170

[0042] JPEG2025509385000010.jpg53170

[0043] JPEG2025509385000011.jpg142170

[0044] JPEG2025509385000012.jpg40170

[0045] Step 2: Each manufacturer's single domain builds a DT network layer of the domain based on the domain network topology model (i.e., single domain topology information), the basic model of each network element node, the functional model of the domain network performance prediction, etc., and reports according to the requirements of the multi-domain organization system.

[0046] Step 3: The multi-domain orchestration system connects the DT network layers of each domain together, and then constructs a DT network layer model (i.e., an optical transport network digital twin model) that generates the entire network.

[0047] JPEG2025509385000013.jpg80170

[0048] JPEG2025509385000014.jpg94170

[0049] The characteristics of method A are as follows: 1. Each network element node within a single domain is equipped with AI training capability, and can timely detect and predict the performance changes of this network element through performance sampling and model training for this node.

[0050] 2. The single-domain management control system also has AI training capabilities, and plays the role of an edge server in this solution from the perspective of federated learning technology. Each network element node in the single domain reports the performance prediction model parameter gradient (or parameter update amount) trained by its own AI algorithm to this domain management control system, and this domain management control system aggregates and processes the model parameter gradient (or parameter update amount) of all network elements reported, and at the same time updates the parameters of this domain performance prediction common model (i.e., single-domain performance prediction model) built by this domain management control system.

[0051] 3. The single-domain management control system broadcasts the updated common model parameters for performance prediction in the single domain to each network element node in the single domain.

[0052] 4. Each network element node updates the network element performance prediction model parameters using the common model parameters, and initiates the next model training and interaction with the single-domain management control system based on this iteration.

[0053] 5. If a communication failure occurs between a network element node and a single-domain management control system, or a failure occurs in the node itself, causing the model parameter gradient (or parameter update amount) of the network element node to be unable to be uploaded, this situation will not affect the update of the single-domain management control system to its own common model parameters or its interaction with other network element nodes.

[0054] 6. This solution can solve problems such as poor generalization ability of performance prediction model training due to differences in device lots, model numbers, service lives, etc. of each network element node in this domain, and overfitting that is likely to occur in model prediction inference.

[0055] 7. Each single domain management control system generates the OTN DT network layer of each network element of this domain (including the DT basic structure model of this network element, the DT performance prediction function model, etc.) according to the common performance prediction model of this domain and the network topology information of this domain obtained by the final training, and sends it to the multi-domain composition system. The multi-domain composition system stitches together the OTN DT network layers reported by each single domain, and further obtains the entire network OTN DT network layer. Because it is the entire network DT network layer obtained by stitching, the performance prediction models of each single domain part included are different.

[0056] 8. When a single-domain management control system and each network element within the domain are constructed by a network from the same equipment manufacturer, the network element does not need to encrypt the performance prediction model parameter gradient of the network element before reporting the model parameter gradient (or parameter update amount) of the network element.

[0057] 9. Adopt a horizontal association learning training common performance prediction model, and ensure that the RNN (here taking RNN as an example) training model of each network element node distributed within a single domain has the same structure as the RNN training model of the single-domain management and control system that plays the role of edge server in training, and ensure unified training and synchronous refresh of RNN model parameters, including all vector element attributes input from the RNN model, the number of vector elements, the number of layers of the RNN model, the number of neurons in each layer, the activation functions between layers, the connection relationships, the output vector attributes, the number of output vector elements, etc.

[0058] The technical effects of this solution are embodied in solving problems such as poor generalization ability of DT performance prediction model training and tendency of overfitting during inference due to different attributes of each network element such as device lot, model number, and service life, and achieving effects such as strong generalization ability of OTN DT performance prediction model and high prediction accuracy.

[0059] In one example, the optical transmission network performance prediction system is applied to Scheme B, and the specific steps include: Step 1: The multi-domain organization and each single domain are trained by horizontal federated learning technique to obtain a multi-domain OTN network performance prediction function model (i.e., network performance prediction model).

[0060] Step 2: The multi-domain orchestration system predicts the common function model and the encrypted topology information reported by each single domain based on the multi-domain OTN network performance, and generates a network OTN DT network layer overall model (i.e., optical transport network digital twin model).

[0061] JPEG2025509385000015.jpg66170

[0062] JPEG2025509385000016.jpg71170

[0063] JPEG2025509385000017.jpg98170

[0064] The characteristics of method B are as follows: 1. Each network element node within a single domain does not have AI training modeling capability, each single-domain management control system has AI training modeling capability, and the multi-domain orchestration system across manufacturers also has AI training modeling capability.

[0065] 2. From the perspective of federated learning technology, the multi-domain composition system in this solution plays the role of an edge server. Each single-domain management control system reports the performance prediction model parameter gradient (or parameter update amount) trained by its own AI algorithm to the multi-domain composition system through homomorphic encryption technology, and the multi-domain composition system performs homomorphic encryption aggregation processing on the model parameter gradient (or parameter update amount) of all the single-domain management control systems reported, and at the same time updates the multi-domain performance prediction common model parameters built by the multi-domain composition system.

[0066] 3. The multi-domain orchestration system broadcasts the updated multi-domain performance prediction common model parameters to each single-domain management control system.

[0067] 4. Each single-domain management and control system refreshes its domain performance prediction model parameters using the common model parameters, and initiates the next model training and interaction with the multi-domain orchestration system based on this iteration.

[0068] 5. If a communication failure occurs between a single-domain management control system and a multi-domain organization system, or a failure occurs in the node itself, causing the model parameter gradient (or parameter update amount) of this domain management control system to be unable to be uploaded, this situation will not affect the update of the multi-domain organization system's own common model parameters and its interaction with other single-domain management control systems.

[0069] 6. This solution can solve problems such as poor generalization ability of performance prediction model training due to differences in device lots, model numbers, service lives, etc. of each network element node between different domains, and overfitting that is prone to occur in model prediction inference, so that the performance prediction model can be generalized across domains and has stronger generalization ability.

[0070] 7. The multi-domain orchestration system generates an inter-domain OTN DT network layer for the entire network based on the multi-domain common performance prediction model obtained in the final training and the encrypted network topology information reported by each domain, and the performance prediction models of each OTN domain are all the same.

[0071] 8. Since each domain is constructed by a network of different equipment manufacturers, each domain management control system needs to encrypt the performance prediction model parameter gradient (or parameter update amount) of this domain before reporting the domain model parameter gradient (or parameter update amount), and the domain network topology information reported by each domain to the multi-domain management control system is encrypted as necessary.

[0072] 9. A common performance prediction model for horizontal associative learning training is adopted to ensure that the RNN (here taking RNN as an example) training model of each single-domain management control system has the same structure as the RNN training model of the multi-domain orchestration system that plays the role of edge server in training, including all vector element attributes input from the RNN model, the number of vector elements, the number of layers of the RNN model, the number of neurons in each layer, the activation functions between layers, the connection relationships, the output vector attributes, the number of output vector elements, etc., to ensure unified training and synchronous refresh of RNN model parameters.

[0073] The technical effects of this solution are achieved by solving problems such as poor generalization ability of the inter-domain OTN DT performance prediction model training and the tendency for overfitting to occur during inference due to different attributes of each manufacturer's network elements, such as device lot, model number, and service life, and by achieving effects such as strong generalization ability of the inter-domain OTN DT performance prediction model and high prediction accuracy.

[0074] The step divisions of the various methods above are only for clarity of description, and they can be integrated into one step or divided into any steps in implementation, as long as they contain the same logical relationship, they are all within the scope of protection of this patent; making insignificant modifications to the algorithm or flow or introducing insignificant designs does not change the core design of the algorithm and flow.

[0075] An embodiment of the present application further relates to an optical transmission network performance prediction method, in which a multi-domain management server, as shown in FIG. 11 , includes the following steps: Step 1101: obtain model update information and single-domain topology information reported by a single-domain management server, where the model update information is obtained after training the performance prediction model of the single-domain management server according to the performance data of each network element in the single domain through horizontal federated learning technology, and is used to update the performance prediction model; Step 1102: Based on the model update information and the single domain topology information, an optical transmission network digital twin model is generated for performance prediction of the optical transmission network.

[0076] The multi-domain management server of this embodiment is used in the optical transmission network performance prediction method, and the multi-domain management server is a server that manages all single-domain management servers in the optical transmission network, and the multi-domain management server runs a multi-domain organization system to realize the optical transmission network performance prediction method.

[0077] In one example, the performance prediction model is a single-domain performance prediction model, and the model update information is a single-domain digital twin model obtained based on the updated single-domain performance prediction model, and the multi-domain management server is realized in the following manner: based on the model update information and the single-domain topology information, an optical transmission network digital twin model is generated, and based on the single-domain topology information, the single-domain digital twin models reported by the single-domain management servers of each single domain are combined to obtain the optical transmission network digital twin model.

[0078] In another example, the performance prediction model is a network performance prediction model, the model update information is a parameter update amount of the network performance prediction model, and the multi-domain management server is realized in the following manner: based on the model update information and the single-domain topology information, generate an optical transmission network digital twin model, update the network performance prediction model based on the parameter update amount of the network performance prediction model, and generate an optical transmission network digital twin model based on the updated network performance prediction model and the single-domain topology information.

[0079] JPEG2025509385000018.jpg159170

[0080] This embodiment is an embodiment corresponding to the above embodiment, and it is clear that this embodiment can be implemented by combining with the above embodiment. The details of the related art mentioned in the above embodiment are still valid in this embodiment, and will not be described further here in order to reduce duplication. Accordingly, the details of the related art mentioned in this embodiment can be applied to the above embodiment.

[0081] An embodiment of the present application further relates to an optical transmission network performance prediction system, as shown in FIG. The system includes a single domain management server 1201 and a multi-domain management server 1202, and the single domain management server 1201 is connected to the multi-domain management server 1202 for communication; Here, the single-domain management server 1201 obtains a performance prediction model, and obtains model update information of the performance prediction model through horizontal federated learning technology, where the model update information is obtained after training the performance prediction model based on the performance data of each network element in the single domain; and reports the model update information and the single-domain topology information to the multi-domain management server 1202, which is used by the multi-domain management server 1202 to generate an optical transmission network digital twin model; The multi-domain management server 1202 obtains model update information and single-domain topology information reported by the single-domain management server 1201, and the model update information is obtained after training the performance prediction model of the single-domain management server 1201 based on the performance data of each network element in the single domain through horizontal federated learning technology, and is used to generate an optical transmission network digital twin model for performance prediction for the optical transmission network based on the model update information and the single-domain topology information.

[0082] It is clear that the present embodiment is a system embodiment corresponding to the above embodiment, and the present embodiment can be implemented by combining with the above embodiment. The details of the related art mentioned in the above embodiment are still valid in the present embodiment, and will not be described further here in order to reduce duplication. Accordingly, the details of the related art mentioned in the present embodiment may be applied to the above embodiment.

[0083] An embodiment of the present application further relates to an electronic device, as shown in FIG. 13, including at least one processor 1301 and a memory 1302 in communication with the at least one processor, where the memory 1302 stores instructions that can be executed by the at least one processor 1301, and the instructions cause the at least one processor 1301 to execute the method of any of the above embodiments.

[0084] Here, the memory 1302 and the processor 1301 are connected in a bus manner, which may include any number of interconnected buses and bridges, and the bus connects various electrical paths of one or more processors 1301 and the memory 1302. The bus may further connect various other circuits, such as peripherals, voltage regulators and power management circuits, all of which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and a transceiver. The transceiver may be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Information processed by the processor 1301 is transmitted over a wireless medium via an antenna, which further receives information and transmits information to the processor 1301.

[0085] The processor 1301 is responsible for managing the bus and general processing and may provide a variety of functions including timing, peripheral interfaces, voltage regulation, power management, and other control functions, however, the memory 1302 may also be used to store information used in performing the operations of the processor.

[0086] An embodiment of the present application relates to a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the above method embodiments.

[0087] That is, as can be understood by those skilled in the art, all or some of the steps of implementing the methods of the above embodiments may be instructed to be completed by related hardware by a program, and the program is stored in one storage medium and includes some instructions for making one device (which may be a one-chip microcomputer, chip, etc.) or a processor execute all or some of the steps of the methods of each embodiment of the present application. However, the above-mentioned storage medium includes various media capable of storing program codes, such as U disks, removable hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

Claims

1. An optical transmission network performance prediction method for use in a single domain management server, comprising: Obtaining a performance prediction model; Obtaining model update information of the performance prediction model through a horizontal federated learning technique, the model update information being obtained after training the performance prediction model based on each network element performance data in a single domain, and being used to update the performance prediction model; reporting the model update information and the single-domain topology information to a multi-domain management server, and causing the multi-domain management server to generate an optical transport network digital twin model for performing performance prediction for the optical transport network based on the model update information and the single-domain topology information; A method for predicting optical transport network performance.

2. The performance prediction model includes a single-domain performance prediction model; The model update information includes obtaining a single-domain digital twin model based on the updated single-domain performance prediction model; The obtaining of model update information of the performance prediction model includes: disseminating the single-domain performance prediction model to each network element in the single domain; Obtaining parameter updates of the single-domain performance prediction model reported by each network element, where each network element trains the single-domain performance prediction model respectively based on local performance data and calculates the parameter updates; updating a parameter of the single-domain performance prediction model based on the parameter update amount reported by each network element, and obtaining the updated single-domain performance prediction model; and obtaining the single domain digital twin model based on the updated single domain performance prediction model and single domain topology information of the single domain. The method for predicting optical transmission network performance according to claim 1 .

3. The step of updating a parameter of the single-domain performance prediction model based on the parameter update amount reported by each network element and obtaining the updated single-domain performance prediction model includes: iteratively updating the single-domain performance prediction model; Wherein, the iterative update includes: sending the updated single-domain performance prediction model to each network element in the single domain, obtaining a parameter update amount reported by each network element again, updating the single-domain performance prediction model according to the parameter update amount reported again by the network element, and starting a next iterative update until an iteration stop condition is met; The method for predicting optical transmission network performance according to claim 2.

4. wherein the performance prediction model includes a network performance prediction model; The model update information includes a parameter update amount of the network performance prediction model, The obtaining of model update information of the performance prediction model includes: training the network performance prediction model based on each network element performance data; and obtaining parameter updates for the network performance prediction model; Wherein, the parameter update amount of the network performance prediction model is reported to the multi-domain management server, and is used to make the multi-domain management server update the network performance prediction model according to the parameter update amount of the network performance prediction model. The method for predicting optical transmission network performance according to claim 1 .

5.

6.

7. An optical transmission network performance prediction method for use in a multi-domain management server, comprising: Obtaining model update information and single-domain topology information reported by a single-domain management server, the model update information being obtained after training a performance prediction model of the single-domain management server based on the performance data of each network element in a single domain through a horizontal federated learning technique, and being used to update the performance prediction model; generating an optical transport network digital twin model for performing performance prediction for the optical transport network based on the model update information and the single domain topology information; A method for predicting optical transport network performance.

8. The performance prediction model includes a single-domain performance prediction model; The model update information includes obtaining a single-domain digital twin model based on the updated single-domain performance prediction model; The generating of an optical transport network digital twin model based on the model update information and the single domain topology information includes: Based on the single domain topology information, the single domain digital twin models reported by the single domain management servers of each single domain are connected together to obtain the optical transport network digital twin model. The method for predicting optical transmission network performance according to claim 7.

9. the performance prediction model includes a network performance prediction model; The model update information includes a parameter update amount of the network performance prediction model, The generating of an optical transport network digital twin model based on the model update information and the single domain topology information includes: updating the network performance prediction model based on a parameter update amount of the network performance prediction model; generating an optical transport network digital twin model based on the updated network performance prediction model and the single domain topology information; The method for predicting optical transmission network performance according to claim 7.

10. An optical transmission network performance prediction system, comprising: a single domain administration server and a multi-domain administration server, the single domain administration server being in communication with the multi-domain administration server; Wherein, the single-domain management server obtains a performance prediction model, and obtains model update information of the performance prediction model through a horizontal federated learning technique, where the model update information is obtained after training the performance prediction model based on the performance data of each network element in a single domain; and reports the model update information and the single-domain topology information to a multi-domain management server, which is used by the multi-domain management server to generate an optical transport network digital twin model; The multi-domain management server acquires the model update information and the single-domain topology information reported by the single-domain management server, and the model update information is obtained after training a performance prediction model of the single-domain management server based on performance data of each network element in the single domain using the horizontal federated learning technology; and the optical transmission network performance prediction system is used to generate the optical transmission network digital twin model for performance prediction for the optical transmission network based on the model update information and the single-domain topology information.

11. An electronic device, At least one processor; a memory in communication with the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions, when executed by the at least one processor, cause the at least one processor to execute the optical transmission network performance prediction method according to any one of claims 1 to 6 or the optical transmission network performance prediction method according to any one of claims 7 to 9. electronic equipment.

12. A computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to realize the optical transmission network performance prediction method according to any one of claims 1 to 6 or the optical transmission network performance prediction method according to any one of claims 7 to 9. A computer-readable storage medium.

Images