Method and system for generating OTN digital twin network based on vertical federation learning

Abstract

This application discloses a method and system for generating an OTN digital twin network based on vertical federated learning, which exchanges data of each single-domain physical network after homomorphically encrypting the data by vertical federated learning, constructs a single-domain training set according to the fault root cause flag and the associated alarm information corresponding to the fault root cause flag, trains a cross-domain fault root cause identification model by the single-domain training set, judges convergence according to the common model parameters and the model parameter update amount during training, and generates an OTN digital twin network based on the trained cross-domain fault root cause identification model and the topology information of each single-domain physical network.

Description

[Technical field]

[0001] This application is filed based on a Chinese patent application bearing application number 202210286945.9 and filing date March 23, 2022, and claims priority to the Chinese patent application, the entire contents of which are incorporated herein by reference.

[0002] The present application relates to the application field of digital twin technology, and in particular to a method and system for generating an OTN digital twin network based on Vertical Federated Learning. [Background technology]

[0003] Based on digital twins, monitoring, analysis, prediction, diagnosis, training, and simulation can be performed, and the simulation results can be fed back to physical objects to support optimization and decision-making of the physical objects. Related technologies such as building digital twin models, updating the status of digital models in real time, and simulation analysis and control decision-making based on digital twins are collectively referred to as digital twin (DT) technology.

[0004] When DT technology is applied to the telecommunications field, it can realize the analysis ability of the entire network, the optical transport network (OTN). The DT network layer is a model that abstracts the entire cross-domain OTN physical network, and the communication between DT network elements is not limited by the physical network space, and the visibility and operability of the DT network elements are not affected by the spatial restrictions such as physical network division, and the management and control restrictions. Therefore, as the OTN DT network layer of the cross-domain OTN physical network, the functional model of this OTN DT network layer needs to globally analyze the entire network of the cross-domain OTN physical network, and also collect sample data from each domain to train the functional model of this OTN DT network layer.

[0005] Generally, each single domain in the entire cross-domain OTN physical network has a large amount of training data collection and processing. In addition, each single domain has privacy protection requirements for related data information such as alarms, network topology, and user traffic, and it is not appropriate to open everything to the public and report it centrally. Therefore, how each single domain management and control system collects and trains complete training sample data for the cross-domain fault root cause identification model, enhances the generalization ability of the model inference, and protects the data privacy of each single domain has become a bottleneck in modeling the digital twin function model of the cross-domain OTN network, and needs to be solved urgently. Summary of the Invention [Problem to be solved by the invention]

[0006] The following is a summary of the subject matter described in detail herein. This summary is not intended to limit the scope of the claims.

[0007] An embodiment of the present application provides a method and system for creating an OTN digital twin network based on vertical federation learning. [Means for solving the problem]

[0008] In a first aspect, the present embodiment comprises: A method for generating an OTN digital twin network based on vertical federation learning, comprising: The present invention is applicable to any single-domain management and control system in an OTN multi-domain physical network system, the OTN multi-domain physical network system further includes a multi-domain orchestration system, and the multi-domain orchestration system and the single-domain management and control system have a cross-domain fault root cause identification model with the same structure; homomorphically encrypting the local fault root cause flag to obtain an encrypted fault root cause flag; receiving all encrypted alarm sample sequences corresponding to the encrypted fault root cause flags, the encrypted alarm sample sequences being obtained by the single-domain management control system of the corresponding single domain homomorphically encrypting related alarm information; generating a single-domain training set based on the encrypted fault root cause flags and the encrypted alarm sample sequences; Training a local cross-domain fault root cause identification model based on the single-domain training set, and obtaining model parameter updates for the cross-domain fault root cause identification model; and reporting the model parameter update amount to the multi-domain orchestration system, so that the multi-domain orchestration system generates an OTN digital twin network based on the model parameter update amount and topology information of each of the single-domain management control systems.

[0009] In a second aspect, the present embodiment comprises: A method for generating an OTN digital twin network based on vertical federation learning, comprising: Applied to a multi-domain orchestration system in an OTN multi-domain physical network system, the OTN multi-domain physical network system further includes a single-domain management and control system, the multi-domain orchestration system and the single-domain management and control system have a cross-domain fault root cause identification model with the same structure; receiving model parameter updates generated by the single-domain management control system; and generating an OTN digital twin network based on the model parameter updates and topology information of each of the single-domain physical networks; The model parameter update amount is obtained by the single-domain management control system training the cross-domain fault root cause identification model of the corresponding single domain based on a single-domain training set, the single-domain training set is generated by the single-domain management control system based on an encrypted fault root cause flag and an encrypted alarm sample sequence corresponding to the encrypted fault root cause flag, the encrypted fault root cause flag is obtained by homomorphically encrypting the fault root cause flag of the single domain in which the single-domain management control system exists, and the encrypted alarm sample sequence is obtained by homomorphically encrypting related alarm information of the single domain in which the single-domain management control system exists.

[0010] In a third aspect, an embodiment of the present application provides a single domain management control system 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 executed by the at least one processor to cause the at least one processor to perform the OTN digital twin network generation method described in the first aspect.

[0011] In a fourth aspect, an embodiment of the present application provides a multi-domain orchestration system 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 executed by the at least one processor to cause the at least one processor to perform the method for generating an OTN digital twin network described in the second aspect.

[0012] In a fifth aspect, embodiments of the present application provide an OTN multi-domain physical network system including a single-domain management and control system as set forth in the third aspect and a multi-domain orchestration system as set forth in the fourth aspect, wherein the multi-domain orchestration system and the single-domain management and control system are connected.

[0013] Other features and advantages of the present application will be set forth in the following specification, and in part will be obvious from the specification, or may be learned by the practice of the present application. The objectives and other advantages of the present application may be realized and obtained by the structure particularly pointed out in the written description, claims and drawings. The drawings are used to provide a further understanding of the technical solution of the present application, constitute a part of the specification, and are used to interpret the technical solution of the present application together with the embodiments of the present application, and are not intended to limit the technical solution of the present application. [Brief description of the drawings]

[0014] [Figure 1] 1 is an overall flowchart of an OTN digital twin network generation method performed by a single domain management control system according to one embodiment of the present application. [Diagram 2] FIG. 1 is an architecture diagram of an OTN cross-domain physical network according to one embodiment of the present application. [Diagram 3] FIG. 1 is an architecture diagram of a cross-domain fault root cause identification model according to one embodiment of the present application. [Figure 4] 1 is a flowchart of a method for constructing a single-domain training set according to an embodiment of the present application. [Diagram 5] 4 is a flowchart of a method for constructing a single-domain training set based on fault root cause flags and associated alarm information according to an embodiment of the present application. [Figure 6] 1 is a method flow diagram of an iterative training process according to one embodiment of the present application. [Figure 7]1 is a method flowchart of a single iteration training process of a single-domain management control system and a multi-domain orchestration system according to an embodiment of the present application. [Figure 8] 1 is an overall flowchart of a method for generating an OTN digital twin network performed by a multi-domain orchestration system according to one embodiment of the present application. [Figure 9] 1 is a method flow diagram of an iterative training process according to one embodiment of the present application. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] In order to make the purpose, technical solution and advantages of the present application clearer, the present application will be described in more detail below with reference to the drawings and examples. The specific examples described in this specification are only used to interpret the present application, and are not used to limit the present application.

[0016] The global communications industry is moving from the Internet and cloud era to the smart era, and new opportunities and new challenges are accelerating the comprehensive transformation and upgrading of networks. Against this background, TM Forum put forward the concept of autonomous networks (AN) in 2019. After more than two years, the AN concept has been agreed upon by the industry, aiming to enable the digital transformation of carrier networks through the integration of network technology and digital technology, provide vertical industries and consumer users with innovative ICT services and user experiences with zero waiting, zero contact, and zero failure, and build self-configuring, self-repairing, and self-optimizing network capabilities for the entire life cycle of carrier network operation. In addition, emerging digital twin technologies have also entered a period of rapid development and have been widely applied in fields such as manufacturing and supply chain, supporting the digital transformation of industry. The digital transformation of the telecommunications industry also has a strong demand for digital twin technology, and the industry considers it to be the foundation of the digital transformation and an important support and component for realizing the AN autonomous network architecture and technology. By accurately sensing the state of itself and the external environment, the network digital twin constructs a digital mirror of the internal and external environment, and combines capabilities such as simulation and preventive prediction, playing an important enabling technology role in scenarios such as "low-cost trial and error", "smart decision-making" and "predictive maintenance".

[0017] When applied to the carrier's communication network scenario, under the AN autonomous network architecture, in the complex network environment of OTN multi-domain, it is necessary to build an OTN DT network model, generate an OTN DT network layer, and realize the analysis ability for OTN of the entire network through DT technology. The DT network layer is a model that abstracts the entire cross-domain OTN physical network, and the communication between DT network elements is not limited by the physical network space, and the visibility and operability of DT network elements are not affected by spatial restrictions such as physical network division or management and control restrictions. Therefore, as the OTN DT network layer of the cross-domain OTN physical network, the functional model of this OTN DT network layer needs to globally analyze the entire network of the cross-domain OTN physical network, and also collect sample data from each domain to train the functional model of this OTN DT network layer. Take the construction of a cross-domain OTN fault root cause identification functional model belonging to the cross-domain OTN network layer DT case sensing class algorithm model as an example, and usually use supervised learning to build a cross-domain OTN fault root cause identification algorithm model. One complete training sample is composed of the input part (each domain alarm), the output fault root cause identification value of the recurrent neural network (RNN), and the root cause flag provided by the domain where the fault root cause exists. Furthermore, the complexity of model training sample collection for cross-domain fault root cause identification is as follows:

[0018] a. Uncertainty of the occurrence of the root cause of failure The fault root causes of different training samples are from different domains. The true fault root cause of the previous training sample occurs in A domain, and the fault root cause flag is provided by A domain. The true fault root cause of the next training sample may occur in B domain, and the fault root cause flag is provided by B domain.

[0019] b. The alarms and fault root cause flags belonging to the input part of the same training sample are from different domains, and a single domain cannot obtain a complete training sample without obtaining the relevant alarm information and fault root cause flags provided by other domains, otherwise it cannot train the RNN model of the domain.

[0020] Therefore, for model training of cross-domain OTN fault root cause identification, the management and control system of a single OTN domain cannot collect all training samples and complete training within the scope of the domain alone. When model training is performed in a unified manner by a multi-domain orchestration system that unifies the cooperation of each single-domain management and control system, the following problems exist:

[0021] 1. The collection and processing of large amounts of training data for each single domain increases the computational load of the multi-domain orchestration system and is inconsistent with the functional positioning of the multi-domain orchestration system.

[0022] 2. Each single domain has privacy protection requirements for related data information such as alarms, network topology, and user traffic, and it is not appropriate to make everything public and report it centrally to the multi-domain orchestration system.

[0023] In short, how each single-domain management and control system can collect and train complete training sample data for the cross-domain fault root cause identification model, enhance the generalization ability of the model inference, and protect the privacy of the data in each domain has become a bottleneck in modeling the DT case function model for cross-domain OTN networks, and needs to be resolved urgently.

[0024] Based on this, the embodiments of the present application provide a method and system for generating an OTN digital twin network, which uses vertical federated learning technology to construct training sample data for a cross-domain fault root cause identification model, solves problems related to data privacy, and improves the generalization ability of the model.

[0025] With reference to FIG. 1, an embodiment of the present application provides a method for generating an OTN digital twin network. The OTN digital twin network is mapped to an OTN multi-domain physical network system including a plurality of single-domain physical networks, and the OTN multi-domain physical network system further includes a single-domain management control system and a multi-domain orchestration system, and the single-domain management control system corresponds to the single-domain physical network in a one-to-one manner, and the single-domain management control system and the multi-domain orchestration system have a cross-domain fault root cause identification model with the same structure. The steps of the method performed by the single-domain management control system in the OTN multi-domain physical network system and the method performed by the multi-domain orchestration system are described in detail in the following two parts.

[0026] For any single-domain management control system in an OTN multi-domain physical network system, the method includes, but is not limited to, the following steps S110, S120, S130, S140, and S150.

[0027] Step S110: homomorphically encrypt the local fault root cause flag to obtain an encrypted fault root cause flag.

[0028] Step S120: receive all encrypted alarm sample sequences corresponding to encrypted fault root cause flags, where the encrypted alarm sample sequences are obtained by the single-domain management and control system of the corresponding single-domain homomorphically encrypting related alarm information.

[0029] Step S130: Generate a single-domain training set based on the encrypted fault root cause flags and the encrypted alarm sample sequences.

[0030] Step S140: Train a local cross-domain fault root cause identification model based on the single-domain training set, and obtain model parameter update amounts of the cross-domain fault root cause identification model.

[0031] Step S150: Report the model parameter update amount to the multi-domain orchestration system, so that the multi-domain orchestration system generates an OTN digital twin network according to the model parameter update amount and the topology information of each single-domain management control system.

[0032] The main architecture of vertical federated learning involves three kinds of entities: entity A, entity B, and coordinator C. In the application scenario of vertical federated learning, the samples of A and B are from a common data-side user, and the sample IDs on the data side are the same, but the feature dimensions of different samples are different. The simple process of vertical federated learning is as follows:

[0033] Step 1: Coordinator C creates a key pair and sends the common key to Party A and Party B.

[0034] Step 2: Party A and Party B encrypt and exchange intermediate results, which are used to calculate gradients and losses.

[0035] Step 3: Party A and Party B calculate the encryption gradients and add an additional mask, respectively. Party B also calculates the encryption loss. Party A and Party B send the encryption results to Party C.

[0036] Step 4: Coordinator C decrypts the gradient information and loss information and sends the results back to Party A and Party B. Party A and Party B unmask the gradient information and update their model parameters based on these gradient information.

[0037] Based on vertical federation learning, the construction scheme of the cross-domain DT network layer based on vertical federation learning in the embodiment of the present application is a DT modeling scheme of homogeneous cross-domain VFML. Take the construction of a cross-domain fault root cause identification model of the cross-domain OTN network DT case as an example, the structure of the OTN multi-domain physical network is shown in FIG. 2. The OTN multi-domain physical network includes multiple single-domain physical networks, and these single-domain physical networks may be constructed based on the switching technology of the same vendor or based on the switching technology of different vendors. A single-domain management control system (OMC: Operations & Maintenance Center) is set for each single-domain physical network, and each single-domain management control system is connected to a multi-domain orchestration system, which outputs a trained cross-domain fault root cause identification model and performs DT modeling based on the topology information of each single-domain physical network. The concept and features of the embodiment of the present application are as follows:

[0038] 1. Each network element node in a single domain does not have AI training capability, each single domain management and control system has AI training modeling capability, and the cross-vendor multi-domain orchestration system also has AI training modeling capability. A cross-domain fault diagnosis analysis case is taken as a training sample. A fault root cause flag occurs in domain k, and model training infers that the relevant input alarms of this root cause flag are scattered in other domains, and the fault root cause flag and the relevant alarm information of each domain all belong to this cross-domain fault analysis case. Only by collecting all the relevant alarms and root cause flags of each domain can model training be realized, which can meet the required application scenarios and training conditions of vertical federated learning.

[0039] 2. In this embodiment, the multi-domain orchestration system plays the role of an edge server. Each single-domain management control system homomorphically encrypts the cross-domain fault root cause identification model parameter update amount trained by its own AI algorithm, and then reports it to the multi-domain orchestration system. The multi-domain orchestration system then performs decryption and aggregation processing on the reported model parameter update amounts of all single-domain management control systems, while updating the cross-domain fault root cause identification common model parameters built by the multi-domain orchestration system.

[0040] 3. The multi-domain orchestration system distributes the updated cross-domain fault root cause identification common model parameters to each single-domain management control system by broadcasting.

[0041] 4. Each single-domain management control system updates the cross-domain fault root cause identification model parameters of its domain using the common model parameters, and then iteratively initiates the next model training and interaction with the multi-domain orchestration system based on the updated common model parameters.

[0042] 5. If a communication failure occurs between a single-domain management control system and a multi-domain orchestration system, or if a node itself fails, and the model parameter gradients of the management control system of this domain cannot be uploaded, the multi-domain orchestration system will not be affected in updating its own common model parameters or interacting with other single-domain management control systems.

[0043] 6. In this embodiment, the relevant OTN domains participating in the training of vertical federated learning can be selected based on the lot of fault root cause training samples. For example, if the fault root cause flag of this lot is related to the alarms of K OTN domains, the model training and update of this lot occurs between the multi-domain orchestrator and the K single-domain management control systems. If the fault root cause flag of the next lot is related to the alarms of M OTN domains, the model training and update of this lot occurs between the multi-domain orchestrator and the M single-domain management control systems. The single-domain management control systems participating in the training of two lots may or may not overlap, but the multi-domain orchestration system always participates in the training of the cross-domain fault root cause identification model of all lots, thereby ensuring that the trained cross-domain fault root cause identification common model has stronger generalization ability and robustness.

[0044] 7. The multi-domain orchestration system generates a cross-domain OTN DT network layer for the entire network based on the trained cross-domain fault root cause identification model, the encrypted network topology information reported by each domain, and other function model information.

[0045] 8. Since each domain is constructed by different equipment manufacturers in a network, the management and control system of each domain needs to encrypt the cross-domain fault root cause identification model parameter update of this domain before reporting the model parameter update amount of this domain. In addition, the network topology information of this domain reported by each domain to the multi-domain orchestration system also undergoes encryption processing as necessary.

[0046] 9. When using vertical federated learning to train a common cross-domain fault root cause identification model, in order to ensure the unified training and synchronous refresh of the RNN+Softmax model parameters, it is necessary to ensure that the structure of the RNN+Softmax (here we take RNN+Softmax as an example) algorithm model of each single-domain management control system and the RNN+Softmax algorithm model of the multi-domain orchestration system that plays the role of edge server during training are the same. This structure includes the input vector attributes of the RNN+Softmax model, the number of vector parameters, the number of layers of the RNN+Softmax 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 parameters, etc.

[0047] Referring to FIG. 3, the modeling method of the embodiment of the present application is as follows.

[0048] JPEG2025509870000002.jpg100170

[0049] JPEG2025509870000003.jpg38170

[0050] JPEG2025509870000004.jpg34170

[0051] JPEG2025509870000005.jpg39170

[0052] JPEG2025509870000006.jpg58170

[0053] JPEG2025509870000007.jpg52170

[0054] JPEG2025509870000008.jpg40170

[0055] Here, the 1{·} operation indicates a value rule that takes the value 1 if the expression inside the parentheses is true, and 0 if it is false.

[0056] JPEG2025509870000009.jpg33170

[0057] JPEG2025509870000010.jpg67170

[0058] JPEG2025509870000011.jpg13170

[0059] Referring to FIG. 4, for the training set of the cross-domain fault root cause identification model, the construction method can be realized in the following steps.

[0060] Step S210: Merge each encrypted alert sample sequence to obtain a complete encrypted alert training sample.

[0061] Step S220: Construct a single-domain training set based on the correspondence between the encrypted fault root cause flags and the complete encrypted alarm training samples.

[0062] Based on the method in which A and B encrypt and exchange intermediate data in vertical federated learning, the fault root cause flag of a single-domain management control system and the related alarm information of all other single-domain management control systems except for the single-domain management control system are homomorphically encrypted and exchanged, the encrypted fault root cause flag and the encrypted alarm sample sequence corresponding to the encrypted fault root cause flag are obtained, and a single-domain training set is constructed from the above two sets of encrypted information, and this single-domain training set belongs to the single domain in which the fault root cause flag exists.

[0063] Referring to FIG. 5 , a single domain in which a single-domain supervisory control system having a fault root cause flag exists is defined as domain k, and for any single-domain supervisory control system other than domain k, constructing a single-domain training set of the single domain to which it belongs includes the following steps:

[0064] Step S221: According to the fault root cause flag of the domain k, determine a first alarm sample vector associated with the fault root cause flag.

[0065] Step S222: homomorphically encrypt the first alert sample vector to obtain a first encrypted alert sample vector, and send the first encrypted alert sample vector to another single domain.

[0066] Step S223: receive a second encrypted alarm sample vector after homomorphic encryption of another single domain, and a second encrypted alarm sample vector corresponding to the second encrypted alarm sample vector is associated with a fault root cause flag.

[0067] Step S224: Merge the first encrypted alert sample vector and the second encrypted alert sample vector to obtain a current single-domain encrypted alert sample sequence.

[0068] Step S225: Construct a single-domain training set based on the encrypted fault root cause flag and the encrypted alarm sample sequence provided by domain k, where the encrypted fault root cause flag is obtained by domain k homomorphically encrypting the fault root cause flag.

[0069] Referring to FIG. 2, take as an example an OTN single domain 1 of vendor 1 obtains one complete encrypted training sample, and assume that the fault root cause flag of this sample is provided by domain k.

[0070] JPEG2025509870000012.jpg27170

[0071] JPEG2025509870000013.jpg50170

[0072] JPEG2025509870000014.jpg23170

[0073] JPEG2025509870000015.jpg127170

[0074] JPEG2025509870000016.jpg40170

[0075] JPEG2025509870000017.jpg55170

[0076] Based on the above cross-domain fault root cause identification model and the constructed single-domain training set, we propose a training procedure for the cross-domain fault root cause identification model in the cross-domain OTN case.

[0077] The overall procedure of the vertical federation learning training of the cross-domain fault root cause identification RNN+Softmax model of the multi-domain orchestration system and the RNN model corresponding to each single-domain management control system is as follows:

[0078] JPEG2025509870000018.jpg50170

[0079] Referring to FIG. 6, the training process of above step S150 may include the following steps.

[0080] Step S310: Report the model parameter update amount to the multi-domain orchestration system.

[0081] Step S320: Receive common model parameters delivered from the multi-domain orchestration system, and the common model parameters are obtained based on the model parameter update amount and the initial common model parameters delivered by the multi-domain orchestration system to the single-domain management control system before iterative training.

[0082] Step S330: Update model parameters of the local cross-domain fault root cause identification model according to the common model parameters.

[0083] Step S340: Iteratively train the cross-domain failure root cause identification model based on the model parameter update amount and the common model parameters, so that the multi-domain orchestration system generates an OTN digital twin network based on the trained cross-domain failure root cause identification model and the topology information of each single-domain management control system, until the cross-domain failure root cause identification model meets a termination condition.

[0084] The single-domain management control system and the multi-domain orchestration system have the same structure of the cross-domain fault root cause identification model, so that in the training process, iterative training is performed by transmitting the model parameters. The single-domain management control system calculates the gradient every time training is completed, and sends the model parameter update amount to the multi-domain orchestration system. The multi-domain orchestration system judges convergence based on the model parameter update amount, and if not, calculates new common model parameters based on the model parameter update amount, and delivers the new common model parameters to the single-domain management control system, and thus performs iterations to finally converge the parameters.

[0085] Referring to FIG. 7, based on the above procedure of iterative training, the procedure of one iteration in the embodiment of the present application is as follows:

[0086] Step S410: The single-domain management control system receives the updated common model parameters.

[0087] Step S420: The single-domain management control system obtains gradients for the updated common model parameters based on the cross-domain fault root cause identification model of the single-domain to which it belongs, and performs homomorphic encryption.

[0088] Step S430: Determine a model parameter update amount of the cross-domain fault root cause identification model according to the gradient calculation result, and report the model parameter update amount to the multi-domain orchestration system.

[0089] Step S440: The multi-domain orchestration system updates the common model parameters according to the model parameter update amount, and delivers the common model parameters to the single-domain management control system.

[0090] JPEG2025509870000019.jpg59170

[0091] JPEG2025509870000020.jpg33170

[0092] JPEG2025509870000021.jpg26170

[0093] JPEG2025509870000022.jpg30170

[0094] JPEG2025509870000023.jpg23170

[0095] JPEG2025509870000024.jpg30170

[0096] JPEG2025509870000025.jpg32170

[0097] JPEG2025509870000026.jpg32170

[0098] Referring to FIG. 8, for a multi-domain orchestration system in an OTN multi-domain physical network system, the method includes, but is not limited to, the following step S510.

[0099] Step S510: Receive the model parameter updates generated by the single-domain management control system, and generate an OTN digital twin network based on the model parameter updates and the topology information of each single-domain physical network.

[0100] Wherein, the model parameter update amount is obtained by the single-domain management control system training a cross-domain fault root cause identification model of the corresponding single domain based on the single-domain training set. The single-domain training set is generated by the single-domain management control system based on the encrypted fault root cause flag and the encrypted alarm sample sequence corresponding to the encrypted fault root cause flag. The encrypted fault root cause flag is obtained by homomorphically encrypting the fault root cause flag of the single domain in which the single-domain management control system exists. The encrypted alarm sample sequence is obtained by homomorphically encrypting the relevant alarm information of the single domain in which the single-domain management control system exists.

[0101] Similarly, referring to the above-mentioned steps S110 to S150, the multi-domain orchestration system performs iterative training based on the model parameter update amount uploaded by the single-domain management control system, and based on the result of the iterative training, combines the topology information of each single-domain physical network to generate an OTN digital twin network.

[0102] Referring to FIG. 9, on the multi-domain orchestration system side, performing iterative training in step S510 to generate an OTN digital twin network may include the following steps.

[0103] Step S511: Generate common model parameters based on the model parameter update amount and the initial common model parameters, and deliver the common model parameters, where the initial common model parameters are delivered by the multi-domain orchestration system to the single-domain management control system before iterative training.

[0104] Step S512: Iteratively train the cross-domain fault root cause identification model according to the model parameter update amount and the common model parameters until the cross-domain fault root cause identification model meets a termination condition.

[0105] Step S513: Generate an OTN digital twin network based on the trained cross-domain fault root cause identification model and the topology information of each single-domain management control system.

[0106] The termination conditions for the iterative training of the above multi-domain orchestration system are the same as the termination conditions for the iterative training of the above single-domain management control system, and the description will not be repeated here.

[0107] In the above embodiment, the data of each single-domain physical network is homomorphically encrypted by vertical federation learning technology, a single-domain training set is constructed according to the fault root cause flag and the associated alarm information corresponding to the fault root cause flag, a cross-domain fault root cause identification model is trained by the single-domain training set, and the convergence is judged according to the common model parameters and the model parameter update amount during training, and an OTN digital twin network is generated according to the trained cross-domain fault root cause identification model and the topology information of each single-domain physical network. The embodiment of the present application meets the privacy protection requirements of associated data such as alarm information, user traffic data, etc. of each single domain in a multi-domain network, and when the multi-domain orchestration system is an edge server in an OTN multi-domain physical network, the method of the embodiment of the present application can be applied to utilize the computing power of the edge device to perform parallel training, thereby improving the model training efficiency.

[0108] It should be noted that the embodiments of the present application are not only applicable to OTN networks, but also applicable to other homogeneous networks such as PTN, POTN, IP networks, etc. The above cross-domain fault identification models are all described using the RNN+softmax structure as an example, but obviously, in the field of neural network algorithms, there are multiple variations and interchangeable algorithms in the RNN, and there are other types of classification algorithms in the classification layer besides softmax, so no examples are given here, and those skilled in the art can select an appropriate algorithm according to the actual situation to construct a cross-domain fault identification model.

[0109] An embodiment of the present application further provides a single-domain management control system, the single-domain management control system 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 executed by the at least one processor to cause the at least one processor to execute the above-mentioned method for generating an OTN digital twin network on the single-domain management control system side.

[0110] An embodiment of the present application further provides a multi-domain orchestration system, the multi-domain orchestration system 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 executed by the at least one processor to cause the at least one processor to perform the above-mentioned method for generating an OTN digital twin network on the multi-domain orchestration system side.

[0111] An embodiment of the present application further provides an OTN multi-domain physical network system, which includes the above-mentioned single-domain management control system and multi-domain orchestration system that execute the OTN digital twin network generation method, and in which the single-domain management control system and the multi-domain orchestration system are connected to exchange data.

[0112] The method for generating an OTN digital twin network according to the embodiment of the present application has at least the following beneficial effects: The data of each single-domain physical network is homomorphically encrypted by vertical federation learning technology, and then data is exchanged; a single-domain training set is constructed according to the fault root cause flag and the associated alarm information corresponding to the fault root cause flag; a cross-domain fault root cause identification model is trained by the single-domain training set; during training, convergence is judged according to the common model parameters and the model parameter update amount; and an OTN digital twin network is generated according to the trained cross-domain fault root cause identification model and the topology information of each single-domain physical network. The embodiment of the present application meets the privacy protection requirements of associated data such as alarm information, user traffic data, etc. of each single-domain in a multi-domain network; and when the multi-domain orchestration system is an edge server in an OTN multi-domain physical network, the method according to the embodiment of the present application can utilize the computing power of the edge device to perform parallel training, thereby improving the model training efficiency.

[0113] All or part of the steps in the methods and systems disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processor, digital signal processor, microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on a computer readable medium, which may include computer storage media (or non-transitory media) and communication media (or transitory media). As known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (e.g., computer readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cartridge, magnetic tape, magnetic disk storage or other magnetic storage device, or any other medium that can be used to store the desired information and that can be accessed by a computer. Additionally, communication media typically includes computer readable instructions, data structures, program modules or other data in a modulated data signal, such as a carrier wave or other transmission mechanism, and may include any information delivery media as known to those skilled in the art.

[0114] Although several implementations of the present application have been specifically described above, the present application is not limited to the above embodiments, and a person skilled in the art may make various equivalent modifications or substitutions without departing from the scope of the present application, and all of these equivalent modifications or substitutions are intended to be included within the scope defined by the claims of the present application.

Claims

1. A method for generating an optical transport network (OTN) digital twin network based on vertical federation learning, comprising: The present invention is applicable to any single-domain management control system in an OTN multi-domain physical network system, the OTN multi-domain physical network system further includes a multi-domain orchestration system, and the multi-domain orchestration system and the single-domain management control system have a cross-domain fault root cause identification model with the same structure; homomorphically encrypting the local fault root cause flag to obtain an encrypted fault root cause flag; receiving all encrypted alarm sample sequences corresponding to the encrypted fault root cause flags, the encrypted alarm sample sequences being obtained by the single-domain management control system of the corresponding single domain homomorphically encrypting related alarm information; generating a single-domain training set based on the encrypted fault root cause flags and the encrypted alarm sample sequences; Training a local cross-domain fault root cause identification model based on the single-domain training set, and obtaining model parameter updates for the cross-domain fault root cause identification model; and reporting the model parameter update amount to the multi-domain orchestration system so that the multi-domain orchestration system generates an OTN digital twin network based on the model parameter update amount and topology information of each of the single domain management control systems.

2. The method for generating an OTN digital twin network as described in claim 1, wherein each of the encrypted alarm sample sequences is an l-dimensional column vector, and the remaining element items of the encrypted alarm sample sequences are zero-padded except for element items of related alarm information.

3. generating a single-domain training set based on the encrypted fault root cause flags and the encrypted alarm sample sequences, merging each of the encrypted alert sample sequences to obtain a complete encrypted alert training sample; The method for generating an OTN digital twin network as described in claim 2, comprising a step of constructing a single-domain training set based on a correspondence between the encryption failure root cause flags and the complete encryption alarm training samples.

4. The step of reporting the model parameter update amount to the multi-domain orchestration system so that the multi-domain orchestration system generates an OTN digital twin network based on the model parameter update amount and topology information of each of the single-domain management control systems, reporting the model parameter updates to the multi-domain orchestration system; receiving common model parameters delivered from the multi-domain orchestration system, the common model parameters being obtained based on the model parameter update amount and initial common model parameters delivered by the multi-domain orchestration system to the single-domain management control system before iterative training; updating model parameters of a local cross-domain fault root cause identification model according to the common model parameters; and iteratively training the cross-domain failure root cause identification model based on the model parameter update amount and the common model parameters until the cross-domain failure root cause identification model satisfies a termination condition, so that the multi-domain orchestration system generates an OTN digital twin network based on the trained cross-domain failure root cause identification model and topology information of each of the single-domain management control systems.

5.

6.

7. prior to reporting said topology information; The method for generating an OTN digital twin network as described in claim 1, further comprising a step of encrypting the topology information.

8. A method for generating an OTN digital twin network based on vertical federation learning, comprising: Applied to a multi-domain orchestration system in an OTN multi-domain physical network system, the OTN multi-domain physical network system further includes a single-domain management control system, and the multi-domain orchestration system and the single-domain management control system have a cross-domain fault root cause identification model with the same structure; receiving model parameter updates generated by the single-domain management control system; and generating an OTN digital twin network based on the model parameter updates and topology information of each of the single-domain physical networks; A method for generating an OTN digital twin network, wherein the model parameter update amount is obtained by the single-domain management control system training the cross-domain fault root cause identification model of the corresponding single domain based on a single-domain training set, the single-domain training set is generated by the single-domain management control system based on an encrypted fault root cause flag and an encrypted alarm sample sequence corresponding to the encrypted fault root cause flag, the encrypted fault root cause flag is obtained by homomorphically encrypting the fault root cause flag of the single domain in which the single-domain management control system exists, and the encrypted alarm sample sequence is obtained by homomorphically encrypting related alarm information of the single domain in which the single-domain management control system exists.

9. The step of generating an OTN digital twin network based on the model parameter update amount and topology information of each of the single domain physical networks includes: generating common model parameters based on the model parameter update amount and an initial common model parameter, and distributing the common model parameters, the initial common model parameters being distributed by the multi-domain orchestration system to the single-domain management control system before iterative training; Iteratively training the cross-domain fault root cause identification model according to the model parameter update amount and the common model parameter until the cross-domain fault root cause identification model satisfies a termination condition; The method for generating an OTN digital twin network as described in claim 8, further comprising: generating an OTN digital twin network based on the trained cross-domain fault root cause identification model and topology information of each of the single-domain management control systems.

10.

11.

12. The method for generating an OTN digital twin network as described in claim 8, wherein the cross-domain fault root cause identification model is composed of a plurality of recurrent neural network (RNN) units and a softmax classification layer.

13. A single domain management control system comprising 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 executed by the at least one processor to cause the at least one processor to execute an OTN digital twin network generation method described in any one of claims 1 to 7.

14. A multi-domain orchestration system comprising 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 executed by the at least one processor to cause the at least one processor to perform the OTN digital twin network generation method described in any one of claims 8 to 12.

15. An OTN multi-domain physical network system comprising: a single-domain management control system according to claim 13; and a multi-domain orchestration system according to claim 14, wherein the multi-domain orchestration system and the single-domain management control system are connected to each other.

Images