WO2023045565A1 - 网络管控方法及其系统、存储介质 - Google Patents

网络管控方法及其系统、存储介质 Download PDF

Info

Publication number
WO2023045565A1
WO2023045565A1 PCT/CN2022/109218 CN2022109218W WO2023045565A1 WO 2023045565 A1 WO2023045565 A1 WO 2023045565A1 CN 2022109218 W CN2022109218 W CN 2022109218W WO 2023045565 A1 WO2023045565 A1 WO 2023045565A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
cognitive
physical
target object
change value
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2022/109218
Other languages
English (en)
French (fr)
Inventor
王大江
王其磊
周瑾
贾逵
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
ZTE Corp
Original Assignee
ZTE Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by ZTE Corp filed Critical ZTE Corp
Priority to EP22871623.9A priority Critical patent/EP4390783A4/en
Priority to US18/684,040 priority patent/US12574299B2/en
Priority to KR1020247011359A priority patent/KR20240061646A/ko
Publication of WO2023045565A1 publication Critical patent/WO2023045565A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/16—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0803—Configuration setting
    • H04L41/0813—Configuration setting characterised by the conditions triggering a change of settings
    • H04L41/0816—Configuration setting characterised by the conditions triggering a change of settings the condition being an adaptation, e.g. in response to network events
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0866—Checking the configuration
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0894—Policy-based network configuration management
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/147—Network analysis or design for predicting network behaviour
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00—Arrangements for monitoring or testing data switching networks
    • H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0895—Configuration of virtualised networks or elements, e.g. virtualised network function or OpenFlow elements
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00—Arrangements for monitoring or testing data switching networks
    • H04L43/20—Arrangements for monitoring or testing data switching networks the monitoring system or the monitored elements being virtualised, abstracted or software-defined entities, e.g. SDN or NFV
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00—Arrangements for monitoring or testing data switching networks
    • H04L43/50—Testing arrangements

Definitions

  • the present application relates to but is not limited to the field of optical communication networks, and in particular relates to a network management and control method, system, and storage medium.
  • Cognitive optical network technology is a new generation of intelligent optical network management and control technology based on machine learning. It can automatically perceive, understand and learn the network environment, adjust network configuration in real time, intelligently adapt to changes in the network environment, and realize rapid fault detection and monitoring. Positioning, real-time optical path performance monitoring and quality prediction, automatic optimization of transmission parameters, realization of traffic prediction and routing planning, root-finding of faults, reduction of optical layer recovery time, etc., improve the overall quality of all optical networks, and provide autonomous driving for optical networks
  • New intelligent optical network management and control technologies such as Optical Network (ADON), Intent-Based Optical Network (IBON), and cloud-optical integration provide the technical basis for network management and control.
  • ADON Optical Network
  • IBON Intent-Based Optical Network
  • cloud-optical integration provide the technical basis for network management and control.
  • Embodiments of the present application provide a network management and control method, a system thereof, and a storage medium.
  • the embodiment of the present application provides a network management and control method, including: obtaining the parameter change value of the target object, wherein the parameter change value comes from the DT virtual model, and the DT virtual model is constructed according to the physical model, and the physical The model is composed of entity objects of the physical network, and the parameter change value represents the change of the transmission performance of the target object; the parameter change value is input into the pre-trained perception model, and the state prediction result output by the perception model is obtained ; Input the state prediction result into the pre-trained cognitive model, and obtain the configuration adjustment information output by the cognitive model; when the configuration adjustment information is verified by simulation, adjust the physical model according to the configuration adjustment information .
  • an embodiment of the present application provides a network management and control system, including: a memory, a processor, and a computer program stored on the memory and operable on the processor, and the processor implements the following when executing the computer program: The network management and control method described in the first aspect.
  • FIG. 1 is a flowchart of a network management and control method provided by an embodiment of the present application
  • Fig. 2 is the structural diagram of the DT case model and physical model that another embodiment of the present application provides;
  • Fig. 3 is the structural diagram of the DT case model that another embodiment of the present application provides;
  • Fig. 4 is a flow chart of determining a target object provided by another embodiment of the present application.
  • Fig. 5 is a flow chart of obtaining an environmental data set provided by another embodiment of the present application.
  • Fig. 6 is a flow chart of constructing a signal generation model provided by another embodiment of the present application.
  • FIG. 7 is a flow chart for generating parameter change values provided by another embodiment of the present application.
  • FIG. 8 is a flow chart of obtaining configuration adjustment information provided by another embodiment of the present application.
  • Fig. 9 is a flowchart of application configuration adjustment information provided by another embodiment of the present application.
  • FIG. 10 is an example flowchart of a network management and control method provided by another embodiment of the present application.
  • Fig. 11 is a flow chart of training a cognitive model and a perception model provided by another embodiment of the present application.
  • Fig. 12 is an example diagram of a cognitive model provided by another embodiment of the present application.
  • FIG. 13 is a schematic topology diagram of Example 1 provided by another embodiment of the present application.
  • Fig. 14 is a flowchart of Example 1 provided by another embodiment of the present application.
  • Fig. 15 is a structural diagram of a network management and control system provided by another embodiment of the present application.
  • the present application provides a network management and control method, its system, and a storage medium.
  • the method includes: acquiring a parameter change value of a target object, wherein the parameter change value comes from a DT virtual model, and the DT virtual model is constructed according to a physical model.
  • the physical model is composed of entity objects of the physical network, and the parameter change value represents the change of the transmission performance of the target object; the parameter change value is input into the pre-trained perception model, and the state prediction output by the perception model is obtained Result; the state prediction result is input into the pre-trained cognitive model, and the configuration adjustment information output by the cognitive model is obtained; when the configuration adjustment information is verified by simulation, the physical device is adjusted according to the configuration adjustment information.
  • Model is composed of entity objects of the physical network, and the parameter change value represents the change of the transmission performance of the target object;
  • the parameter change value is input into the pre-trained perception model, and the state prediction output by the perception model is obtained Result;
  • the state prediction result is input into the pre-
  • the DT virtual model can be used as the data basis to perceive and predict the change and change trend of the physical model through the perception model, and use the perception and prediction results as the input of the cognitive model to obtain the configuration of the physical network Adjust information and realize automatic management and control for physical networks after simulation verification.
  • FIG. 1 is a network management and control method provided by an embodiment of the present application, including but not limited to step S110 , step S120 , step S130 and step S140 .
  • Step S110 obtain the parameter change value of the target object, wherein the parameter change value comes from the digital twin DT virtual model, the DT virtual model is constructed according to the physical model, the physical model is composed of physical network entity objects, and the parameter change value represents the transmission of the target object performance changes.
  • the physical model can be composed of any physical object, such as an optical transport network (Optical Transport Network, OTN) topology environment composed of multiple routing nodes and transmission links.
  • OTN optical Transport Network
  • This embodiment does not limit the specific composition of the physical model .
  • the target object can be any entity object in the physical model, which can be selected according to the analysis requirements.
  • the parameter corresponding to the parameter change value may be any performance parameter of the target object.
  • the parameter type is not limited too much, as long as it can represent the transmission performance of the target object.
  • the parameter change value can be the change range of the value of the performance parameter, such as the increase or decrease of the bandwidth, the increase or decrease of the number of ports, or the change of the state, etc., which can reflect the change of the performance parameter of the entity object.
  • a DT case usually includes a basic model and a functional model.
  • the hierarchical structure of the DT case can be referred to as shown in Figure 2.
  • the basic model is a DT virtual model
  • the functional model includes a perception model and a cognitive model.
  • the DT virtual model is obtained through abstract virtual processing for the physical model, and the DT virtual model is used as the basic model to serve as the model foundation and mathematical background for subsequent perception analysis and cognitive analysis;
  • the abstract virtual models are perception algorithms Model, cognitive algorithm model construction, training and inference basis.
  • Step S120 input the parameter change value into the pre-trained perceptual model, and obtain the state prediction result output by the perceptual model.
  • the perception model belongs to the functional model of the DT case. Based on the DT case application scenario, relying on and extracting the virtual model information of the target object, and sampling the traffic or performance parameters of the target object in the physical model, object status and other data as the input and output, training and generating the corresponding perception model.
  • the perception model can be implemented by any artificial intelligence (AI) technology that can achieve prediction, such as convolutional neural network (Convolution Neural Network, CNN), deep neural network (Deep Neural Network, DNN), deep reinforcement learning (Deep Reinforcement Learning, DRL), etc., this embodiment does not make too many limitations on the algorithm specifically adopted by the perception model.
  • AI artificial intelligence
  • the perceptual model can be used to infer and predict the impact of changes in the network environment where the DT case target object is located on the current network state and network services involved in the target object, and obtain the state prediction results.
  • the state of the link is For light traffic status, just adjust different status prediction results for different object types.
  • the DT virtual model describes the parameters of the environment in which the physical model is located.
  • the perception model Through the perception model, the environmental changes caused by parameter changes can be actively predicted and perceived, thus providing trigger signals and models for further network cognition. Input to realize automatic detection of OTN.
  • Step S130 input the state prediction result into the pre-trained cognitive model, and obtain configuration adjustment information output by the cognitive model.
  • the cognitive model belongs to the functional model of the DT case, which can rely on and extract the virtual model information of the target object based on the application scenario of the DT case, and use the state prediction results obtained from the analysis of the perception model as input, combined with AI technology Train and generate the corresponding cognitive model, and infer the configuration adjustment information of the DT case target object through the algorithm model.
  • the cognitive model can use any algorithm related to cognitive technology, as long as it can realize cognitive technology, such as the actor-evaluator (Actor-Critic) algorithm, this embodiment does not go into details about the specific algorithm and its implementation. Those skilled in the art have an incentive to choose the appropriate algorithm to configure the cognitive model according to the needs of the occasion.
  • the DT virtual model is used as the basic structure model to provide the environmental data of the physical model for the cognitive model and the perception model, and the output of the perception model
  • the state prediction result can be connected in series between the cognitive model and the cognitive model. From data acquisition to active perception of changes to determination of configuration adjustment information, automatic and intelligent control of the network is realized in a closed-loop form.
  • each model can be split and multiple sub-instances can be created separately. Here I won't go into details.
  • Step S140 when the configuration adjustment information passes the simulation verification, adjust the physical model according to the configuration adjustment information.
  • the configuration adjustment information configuration obtained by using cognitive technology can be any available configuration information, such as the adjustment of performance parameters, or a configuration change strategy, such as adjusting the transmission path, etc.
  • the environment changes.
  • simulation verification can be performed before application. This embodiment does not limit the specific process of simulation verification too much. In this case, those skilled in the art are familiar with how to perform simulation verification, so details are not repeated here.
  • the overall process of this embodiment is a cyclic and iterative process, which can span the entire network autonomy life cycle, by real-time monitoring and perception of changes in the network environment related to the target object, inferring and predicting the relevant state changes of the target object, And use AI technology and simulation technology to obtain cognitive judgment and analyzed configuration adjustment information for the target object, and then implement the configuration adjustment information in the physical network environment.
  • the cognitive process includes the simulation of network changes, and
  • the timeliness and accuracy of optimizing the configuration change can be used as the quality evaluation criteria to measure the performance of the network management and control system modeling and algorithms.
  • step S110 shown in the embodiment of FIG. 1 is executed, the following steps S410 to S420 are also included but not limited to.
  • Step S410 determining the physical object whose performance parameter changes in the physical model as the target object.
  • step S420 the analysis requirement information is acquired, and the object to be analyzed in the analysis requirement information is determined as the target object.
  • the performance parameters of an entity object can be in a state of change or remain unchanged.
  • the number of link ports is usually fixed, while the traffic of a transmission link changes in real time.
  • the target object can also be selected according to the analysis requirements, for example, for the survivability analysis requirements, the analysis chain
  • the impact of changes in the number of ports or transmission link failures on the physical network are motivated to select a method for determining the target object according to actual needs, and no more limitations are made here.
  • step S510 is also included but not limited to.
  • step S510 the environment data set corresponding to the physical model is acquired from the DT virtual model, and the environment data set includes performance parameters and state parameters of each entity object in the physical model.
  • the network environment of the physical model can be monitored in real time and the corresponding environmental data can be collected to form an environmental data set composed of various environmental data.
  • the environmental data set can include each entity object The performance parameters and state parameters of , the specific collection method will not be repeated here.
  • step S510 shown in the embodiment of FIG. 5 is executed, the following steps S610 to S630 are also included but not limited to.
  • Step S610 acquiring a preset time series.
  • Step S620 determining the probability distribution of performance parameters and state parameters in time series.
  • Step S630 constructing a signal occurrence model according to the probability distribution and time series.
  • the kernel smoothing method can be used to obtain the corresponding parameters in the specified time series The probability distribution on the above, and build a signal generation model based on this, so that the signal generation model can generate data that does not appear in the real environment according to the probability distribution, and use it as a sample parameter for perception and cognition, and realize network analysis and control.
  • the time series can be a period of time of any length.
  • the link traffic on the first day of each month can be collected in units of months, and the link traffic can be determined by kernel smoothing
  • the constructed signal generation model can determine the link traffic on the first day in several months in the future according to the probability distribution.
  • the step S110 shown in the embodiment of FIG. 1 also includes but not limited to the following steps S710 to S720.
  • Step S710 determining the target analysis sequence.
  • step S720 the parameter change value of the target object in the target analysis time series is constructed according to the signal generation model.
  • target analysis time series can be any time period, for example, the first day of the next three months, which can be determined according to actual analysis requirements.
  • the parameter change values of the target object at the target analysis time series can be generated through the signal generation model, and the data can be generated by simulating the environment , to improve the simulation efficiency.
  • step S130 shown in the embodiment of FIG. 1 also includes but not limited to the following steps S810 to S820.
  • Step S810 obtaining a preset target state and constraint conditions.
  • Step S820 input the environment data set, target state, state prediction result and constraint conditions into the cognitive model, and obtain configuration adjustment information for the target object output by the cognitive model.
  • the target state can be aimed at the target object or the entire physical model.
  • the target object can be a transmission link
  • the target state can be set as "'maximum load balance" as the track The final return", so as to obtain the configuration adjustment information that can achieve the above-mentioned target state through the cognitive model.
  • constraints can be set in advance, or can be determined according to the state prediction results output by the perception model. For example, if a link failure needs to be simulated, the constraints can be set so that the transmission The model can make cognitive predictions based on constraints.
  • path A is a congested path and path B is a light-loaded path
  • path A can be used as a must-avoid constraint
  • path B can be used as a must-pass constraint
  • the specific selection method of constraint conditions can be determined according to actual analysis requirements, and will not be limited here.
  • the cognitive model can use Actor-Critic’s DRL technology to input environmental data sets, target states, state prediction results and constraints into the cognitive model, and can perform global or local cognitive optimization through DRL technology, thereby Obtain the configuration adjustment information after cognitive optimization, so as to realize network optimization.
  • step S140 shown in the embodiment of FIG. 1 also includes but not limited to the following steps S910 to S920.
  • Step S910 performing simulation verification according to the configuration adjustment information, and obtaining a simulation result.
  • Step S920 when the simulation result indicates that the running state of the physical model meets the preset standard, it is determined that the configuration adjustment information has passed the simulation verification.
  • the configuration adjustment information is obtained based on the current physical environment and the prediction results of the perception model. After the configuration adjustment information is applied to the physical model, there may be derivative effects, such as the traffic of the transmission link as the target object. After the limitation, the flow of other transmission links increases, resulting in link congestion.
  • the simulation tool can be used to simulate after obtaining the configuration adjustment information. As long as those skilled in the art have configuration adjustment information, they will be motivated to select a suitable simulation tool according to actual needs.
  • the preset standard can be determined according to the performance requirements of the physical network. For example, after the adjustment, the transmission link will not be congested, or the traffic meets the preset threshold, and the preset standard can be adjusted according to the actual demand.
  • the network management and control method of this example includes but is not limited to the following steps S1010 to S1080.
  • Step S1010 monitor the physical network environment in real time, and collect network environment data related to the DT case.
  • Step S1021 build a DT case for the specified physical model, wherein, the DT case includes a basic model and a functional model, the basic model is a DT virtual model, and the functional model includes a perception model and a cognitive model, and execute step S1031.
  • Step S1022 build an environment data set according to the DT virtual model, and execute step S1032 or step S1033.
  • Step S1031 determine the parameter change value of the physical network environment according to the environmental data set, if prediction is required and the perception model is available, execute step S1042 or step S1043, if no prediction is required, execute step S1041.
  • Step S1032 build a signal generator according to the environment data set, and if the perception model is available, go to step S1042.
  • Step S1033 using AI technology to train, build, and refresh the corresponding perception model.
  • Step S1041 judge and analyze changes in the physical network environment based on the DT virtual model, and obtain state prediction results. If the perception model is available, execute step S1060.
  • step S1042 the parameter change value is used as an input to obtain a state prediction result, and if the perceptual model is available, execute step S1060.
  • Step S1043 predict the state prediction result according to the time series parameter change value of the network performance generated by the signal generator, and if the perception model is available, execute step S1060.
  • Step S1050 constructing and training a cognitive model according to the state prediction results of the environmental data set.
  • step S1060 the environmental data set and the state prediction result are used as input of the cognitive model to obtain configuration change information through prediction.
  • Step S1070 call the emulator simulation tool to simulate the configuration change information, determine the feasibility of the configuration change and the derivative impact on the network environment, if the simulation result meets the network operation requirements, execute step S1080, otherwise execute step S1060 again.
  • Step S1080 applying the configuration change information to the physical network.
  • step S520 shown in the embodiment of FIG. 5 also includes but not limited to the following step S1110.
  • Step S1110 perform training from the cognitive model to the perception model through the backpropagation algorithm, wherein the performance parameter is the training input of the perception model, the state parameter is the training output of the perception model, and the state prediction result and the environmental data set are the cognitive model
  • the performance parameter is the training input of the perception model
  • the state parameter is the training output of the perception model
  • the state prediction result and the environmental data set are the cognitive model
  • the training input, the target state is determined as the output of the cognitive model.
  • the performance parameter is used as the input and the state parameter is used as the output, so that the trained perception model can predict the corresponding state parameter according to the change of the performance parameter, for example, using the recurrent neural network (Recurrent Neural Network, RNN) as the AI algorithm of the perception model, those skilled in the art are familiar with how to train the deep learning model, so I won’t go into details here.
  • RNN recurrent Neural Network
  • the perceptual model and the cognitive model are combined in series, and the model parameters in the cognitive model and perceptual model can be controlled by gradient descent/ Ascending method, perform iterative update until the gradient of the serial model parameters approaches to 0, the whole iterative update ends, and then complete the unified training of the end-to-end AI model from the cognitive model to the perception model, the backpropagation algorithm and the chain calculation
  • the guiding method is a technique well known to those skilled in the art, and will not be repeated here. Training through backpropagation can realize the integrated series connection of the cognitive model to the perception model, which is conducive to the automatic detection, automatic optimization and automatic troubleshooting of the physical network, and improves the stability of the physical network.
  • the perception model uses RNN technology, which can take the distribution of network traffic at different times in the entire time series cycle as RNN input to predict the status of key links of the OTN network at different times in the entire time series cycle, and Combined with the topological state of the current network, as the input of the cognitive model, the cognitive model can use the Actor-Critic DRL technology, and after training, it can be used to reason and obtain the best OTN network traffic optimization change configuration scheme, such as the one shown in Figure 12 structure, the performance parameter is the network traffic distribution, where t is the number of performance parameters, input X1 to Xt into the RNN unit for prediction, where X1 is input into the RNN unit 1, the output state prediction result is the key link in the whole time The state at different times in the sequence cycle, where the output corresponding to the input X1 is Z1, and so on, after the perception model outputs the state prediction result, Z1 to Zt and the environmental data set are used as the training input of the cognitive model, and the target state As the objective function
  • the physical model takes the OTN Mesh networking topology shown in Figure 13 as an example.
  • this network there are 6 data unit ODUflex service paths, namely path 1: A-S, path 2: A-K, and path 3: K-C , Path 4: A-B-C, Path 5: T-C, Path 6: S-N-T;
  • the physical model also includes the key link 1: S-N, and the key link 2: A-B.
  • this example takes 15 minutes as the target to analyze the timing .
  • the method in this example includes but is not limited to the following steps S1410 to S1450.
  • Step S1410 determine the current OTN network traffic distribution as the target object of the DT case, and construct the DT virtual model of the corresponding DT case.
  • Step S1420 based on the OTN network traffic abstraction virtual model of DT case and the OTN network traffic perception and analysis requirements, construct an OTN network traffic perception algorithm model.
  • the structure of the perceptual model can refer to Figure 12.
  • the input at each moment t is It is in the form of a vector, and each element of the vector is a continuous variable, indicating the average traffic value of each ODUflex service every 15 minutes at time t, in Gbit/s.
  • Step S1430 predict the trend state of the key link through the perception model.
  • Z t is in the form of a vector, and each element is a discrete variable, which represents the trend state value of n key links of the OTN network topology at time t, and the trend state is described in a numerical way, for example, 0 represents the normal state of traffic, and 1 represents the tide tension status, 2 means traffic congestion, 3 means ebb and flow state, 4 means traffic light load,
  • Step S1440 based on the abstract virtual model of OTN network traffic of DT case, the inferred output of OTN network traffic perception algorithm model, and the requirements of OTN network traffic cognitive analysis, construct the OTN network traffic cognitive algorithm model of DT case.
  • the cognitive algorithm model of DT case to implement the cognitive analysis process of OTN network traffic mainly adopts the Actor-Critic algorithm of DRL technology.
  • the input of this Actor-Critic algorithm instance includes: the current network topology of OTN , business distribution, and the congestion trend status of each key link at time 0 to t. Based on the above input, it is also possible to use congested links as must-avoid constraints and light-loaded links as must-pass constraints.
  • Step S1450 obtain the adjustment policy output by the OTN network traffic cognitive algorithm model of the DT case, and adjust the physical model according to the adjustment policy.
  • the network management and control system 1500 includes: a memory 1510 , a processor 1520 , and a computer stored in the memory 1510 and capable of running on the processor 1520 program.
  • the processor 1520 and the memory 1510 may be connected through a bus or in other ways.
  • the non-transitory software programs and instructions required to implement the network management and control method of the above-mentioned embodiment are stored in the memory 1510, and when executed by the processor 1520, the network management and control method in the above-mentioned embodiment is executed, for example, executing the above-described Figure 1 Method step S110 to step S140 in, method step S410 to step S420 in Fig. 4, method step S510 in Fig. 5, method step S610 to step S630 in Fig. 6, method step S710 to step S720 in Fig. 7, The method step S810 to step S820 in FIG. 8 , the method step S910 to step S920 in FIG. 9 , and the method step S1110 in FIG. 11 .
  • the device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • an embodiment of the present application also provides a computer-readable storage medium, the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are executed by a processor or a controller, for example, by the above-mentioned Execution by a processor in the embodiment of the network management and control system can cause the above-mentioned processor to execute the network management and control method in the above-mentioned embodiment, for example, execute the method steps S110 to S140 in FIG. 1 described above, and the method steps in FIG. 4 S410 to step S420, method step S510 in FIG. 5, method step S610 to step S630 in FIG. 6, method step S710 to step S720 in FIG. 7, method step S810 to step S820 in FIG.
  • Method step S910 to step S920, method step S1110 in FIG. 11 can be implemented as software, firmware, hardware and an appropriate combination thereof.
  • Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit .
  • Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media).
  • computer storage media includes both volatile and nonvolatile media implemented in any method or technology for storage of information, such as computer readable instructions, data structures, program modules, or other data. permanent, removable and non-removable media.
  • Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cartridges, tape, magnetic disk storage or other magnetic storage devices, or can Any other medium used to store desired information and which can be accessed by a computer.
  • communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media .
  • the embodiment of the present application includes: obtaining the parameter change value of the target object, wherein the parameter change value comes from the DT virtual model, and the DT virtual model is constructed according to the physical model, the physical model is composed of physical network entity objects, and the parameter The change value represents the change of the transmission performance of the target object; the parameter change value is input into the pre-trained perception model, and the state prediction result output by the perception model is obtained; the state prediction result is input into the pre-trained
  • the cognitive model of the cognitive model is used to obtain configuration adjustment information output by the cognitive model; when the configuration adjustment information is verified by simulation, the physical model is adjusted according to the configuration adjustment information.
  • the DT virtual model can be used as the data basis to perceive and predict the change and change trend of the physical model through the perception model, and use the perception and prediction results as the input of the cognitive model to obtain the configuration of the physical network Adjust information and realize automatic management and control for physical networks after simulation verification.
  • the embodiment of the present application can combine the DT technology and the cognitive optical network technology to realize the automatic management and control of the optical network.

Landscapes

  • Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Environmental & Geological Engineering (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

一种网络管控方法及其系统、存储介质，包括：获取目标对象的参数变化值，其中，参数变化值来自于DT虚拟模型，DT虚拟模型根据物理模型构建，物理模型由物理网络的实体对象构成，参数变化值表征目标对象的传输性能的变化（S110）；将参数变化值输入至感知模型，获取感知模型输出的状态预测结果（S120）；将状态预测结果输入至认知模型，获取认知模型输出的配置调整信息（S130）；在仿真验证后根据配置调整信息调整物理模型（S140）。

Description

网络管控方法及其系统、存储介质
相关申请的交叉引用
本申请基于申请号为202111108462.1、申请日为2021年9月22日的中国专利申请提出，并要求该中国专利申请的优先权，该中国专利申请的全部内容在此引入本申请作为参考。
技术领域
本申请涉及但不限于光通信网络领域，尤其涉及一种网络管控方法及其系统、存储介质。
背景技术
认知光网络技术是一种基于机器学习的新一代智能光网络管控技术，能够自动感知、理解和学习网络环境，实时调整网络配置，智能地适应网络环境的变化，还可以实现快速故障检测和定位、实时光路性能监测和质量预判、自动优化传输参数、实现流量预测和路由规划、进行故障寻根、减少光层恢复时间等，提高全光网的整体质量，为光网络自动驾驶(Autonomous Driving Optical Network，ADON)、意图光网络(Intent-Based Optical Network，IBON)、云光融合等新的智能化光网络管控技术提供了网络管控的技术基础。
对光网络的认知分析和仿真，需要在不影响物理网络环境和运行的前提下进行，通常会采用数字孪生(Digital Twin，DT)技术创建物理对象的高保真数字虚拟模型，模拟物理对象的行为，但是DT生成的虚拟模型只能作为数据基础，目前并没有相关技术能够实现DT与认知光网络技术的协同。
发明内容
以下是对本文详细描述的主题的概述。本概述并非是为了限制权利要求的保护范围。
本申请实施例提供了一种网络管控方法及其系统、存储介质。
第一方面，本申请实施例提供了一种网络管控方法，包括：获取目标对象的参数变化值，其中，所述参数变化值来自于DT虚拟模型，DT虚拟模型根据物理模型构建，所述物理模型由物理网络的实体对象构成，所述参数变化值表征所述目标对象的传输性能的变化；将所述参数变化值输入至预先训练好的感知模型，获取所述感知模型输出的状态预测结果；将所述状态预测结果输入至预先训练好的认知模型，获取所述认知模型输出的配置调整信息；当所述配置调整信息通过仿真验证，根据所述配置调整信息调整所述物理模型。
第二方面，本申请实施例提供了一种网络管控系统，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如第一方面所述的网络管控方法。
本申请的其它特征和优点将在随后的说明书中阐述，并且，部分地从说明书中变得显而易见，或者通过实施本申请而了解。本申请的目的和其他优点可通过在说明书、权利要求书以及附图中所特别指出的结构来实现和获得。
附图说明
附图用来提供对本申请技术方案的进一步理解，并且构成说明书的一部分，与本申请的实施例一起用于解释本申请的技术方案，并不构成对本申请技术方案的限制。
图1是本申请一个实施例提供的网络管控方法的流程图；
图2是本申请另一个实施例提供的DT case模型与物理模型的结构图；
图3是本申请另一个实施例提供的DT case模型的结构图；
图4是本申请另一个实施例提供的确定目标对象的流程图；
图5是本申请另一个实施例提供的获取环境数据集的流程图；
图6是本申请另一个实施例提供的构建信号发生模型的流程图；
图7是本申请另一个实施例提供的生成参数变化值的流程图；
图8是本申请另一个实施例提供的得到配置调整信息的流程图；
图9是本申请另一个实施例提供的应用配置调整信息的流程图；
图10是本申请另一个实施例提供的网络管控方法的示例流程图；
图11是本申请另一个实施例提供的训练认知模型和感知模型的流程图；
图12是本申请另一个实施例提供的认知模型的示例图；
图13是本申请另一个实施例提供的示例一的拓扑示意图；
图14是本申请另一个实施例提供的示例一的流程图；
图15是本申请另一个实施例提供的网络管控系统的结构图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本申请，并不用于限定本申请。
需要说明的是，虽然在装置示意图中进行了功能模块划分，在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于装置中的模块划分，或流程图中的顺序执行所示出或描述的步骤。说明书、权利要求书或上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。
本申请提供了一种网络管控方法及其系统、存储介质，方法包括：获取目标对象的参数变化值，其中，所述参数变化值来自于DT虚拟模型，DT虚拟模型根据物理模型构建，所述物理模型由物理网络的实体对象构成，所述参数变化值表征所述目标对象的传输性能的变化；将所述参数变化值输入至预先训练好的感知模型，获取所述感知模型输出的状态预测结果；将所述状态预测结果输入至预先训练好的认知模型，获取所述认知模型输出的配置调整信息；当所述配置调整信息通过仿真验证，根据所述配置调整信息调整所述物理模型。根据本申请实施例提供的方案，能够以DT虚拟模型作为数据基础，通过感知模型感知、预测物理模型的变化及变化趋势，并将感知、预测结果作为认知模型的输入，得到物理网络的配置调整信息，并经过仿真验证后实现针对物理网络的自动管控。
以下对本申请的网络管控方法进行进一步阐述。
如图1所示，图1是本申请一个实施例提供的一种网络管控方法，包括但不限于有步骤S110、步骤S120、步骤S130和步骤S140。
步骤S110，获取目标对象的参数变化值，其中，参数变化值来自于数字孪生DT虚拟模型，DT虚拟模型根据物理模型构建，物理模型由物理网络的实体对象构成，参数变化值表征目标对象的传输性能的变化。
需要说明的是，物理模型可以由任意实体对象构成，例如由多个路由节点和传输链路构成的光传送网(Optical Transport Network，OTN)拓扑环境，本实施例对物理模型的具体构成不作限定。目标对象可以是物理模型中的任意实体对象，根据分析需求选取即可。
需要说明的是，参数变化值所对应的参数可以是目标对象的任意性能参数，本实施例对参数类型不作过多限定，能够表征目标对象的传输性能即可。参数变化值可以是性能参数的数值的变化幅度，例如带宽的增加或减少，端口数的增加或者减少，也可以是状态的变化等，能够体现出实体对象的性能参数变化即可。
值得注意的是，DT case通常包括基础模型和功能模型，在本实施例中，DT case的层级结构可以参考图2所示，基础模型为DT虚拟模型，功能模型包括感知模型和认知模型，其中，DT虚拟模型通过针对物理模型进行抽象化的虚拟处理得到，以DT虚拟模型作为基础模型，以作为后续感知分析、认知分析的模型基础与数学背景；抽象化的虚拟模型分别是感知算法模型、认知算法模型构建、训练与推理的基础。
步骤S120，将参数变化值输入至预先训练好的感知模型，获取感知模型输出的状态预测结果。
需要说明的是，感知模型属于该DT case的功能模型。基于DT case应用场景，依托、抽取目标对象的虚拟模型信息，并采样目标对象在物理模型的流量或性能参数、对象状态等数据作为该输入、输出，训练生成对应的感知模型。感知模型可以通过任意可实现预测的人工智能(Artificial Intelligence，AI)技术实现，例如卷积神经网络(Convolution Neural Network，CNN)、深度神经网络(Deep Neural Network，DNN)、深度强化学习(Deep Reinforcement Learning，DRL)等，本实施例对感知模型所具体采用的算法不作过多限定。
值得注意的是，通过感知模型可以推断、预测DT case目标对象所处的网络环境变化对该目标对象涉及的当前的网络状态、网络服务的波及影响，得到状态预测结果，例如链路的状态为流量轻载状态，针对不同对象类型调整不同的状态预测结果即可。
可以理解的是，DT虚拟模型描述了物理模型所处的环境的参数，通过感知模型能够对参数变化所造成的环境变化进行主动预测和感知，从而为进一步的网络认知提供了触发信号和模型输入，实现对OTN的自动化检测。
步骤S130，将状态预测结果输入至预先训练好的认知模型，获取认知模型输出的配置调整信息。
需要说明的是，认知模型属于该DT case的功能模型，能够基于DT case应用场景，依托、抽取目标对象的虚拟模型信息，并根据感知模型分析得出的状态预测结果作为输入，结合AI技术训练并生成对应的认知模型，并通过该算法模型推断获得对该DT case目标对象的配置调整信息。认知模型可以采用任意认知技术相关的算法，能够实现认知技术即可，例如演员-评价者(Actor-Critic)算法，本实施例对具体的算法及其实现方式不作过多赘述，本领域技术人员有动机根据时机需求选取合适的算法以配置认知模型。
需要说明的是，认知模型、感知模型和DT虚拟模型的关系可以参考图3所示，DT虚拟模型作为基础结构模型，为认知模型和感知模型提供物理模型的环境数据，感知模型输出的状态预测结果作为认知模型的输入，能够感知模型和认知模型的串联，实现了从数据获取到主动感知变化到确定配置调整信息，以闭环的形式实现了网络的自动化和智能化管控。
值得注意的是，在同一DT case中的三种模型，在感知与认知分析的处理中，视对象结构、功能分析需要，还可对每种模型分别拆分和创建多个子实例，在此不多做赘述。
步骤S140，当配置调整信息通过仿真验证，根据配置调整信息调整物理模型。
需要说明的是，利用认知技术得到的配置调整信息配置可以是任意可用的配置信息，例 如性能参数的调整，也可以是配置变更策略，例如调整传输路径等，能够对物理模型所处的实体环境产生变化即可。
值得注意的是，为了确保配置调整信息不会对物理模型产生不必要的衍生影响，可以在应用之前进行仿真验证，本实施例对仿真验证的具体过程不作过多限定，在具备配置调整信息的情况下，本领域技术人员熟知如何进行仿真验证，在此不多作赘述。
需要说明的是，本实施例的整体流程是一个循环迭代的过程，可以跨越整个网络自治生命周期，通过实时监控并感知与目标对象相关的网络环境变化，推断、预测目标对象的相关状态变化，并通过AI技术、仿真技术获得认知判断、分析后的针对目标对象的配置调整信息，进而在物理网络环境中实施该配置调整信息，同时，认知流程中包含了对网络变更的仿真、与配置实施，对于认知模型，可以以优化实现变更配置的及时性、准确率作为衡量该网络管控系统建模与算法性能的质量评估标准。
通过本实施例的技术方案，能够实现自动化和智能化的认知分析，有效减少变更配置物理网络环境的尝试次数、并提高变更配置解决问题的精准度，减少参数调整更对网络服务正常运营的影响。
另外，在一实施例中，参照图4，在执行完图1实施例所示的步骤S110之前，还包括但不限于有以下步骤S410至步骤S420。
步骤S410，将物理模型中性能参数发生变化的实体对象确定为目标对象。
或者，步骤S420，获取分析需求信息，将分析需求信息中的待分析对象确定为目标对象。
需要说明的是，对于物理网络而言，实体对象的性能参数可以是处于变化状态，也可以是维持不变的，例如链路端口数通常是固定的，而传输链路的流量是实时变化的，因此，可以将性能参数发生变化的实体对象作为目标对象，通过DT虚拟模型获取其参数变化值进行感知和认知；也可以根据分析需求选取目标对象，例如针对生存性分析需求，需要分析链路端口数变化或者传输链路出现故障对物理网络造成的影响。本领域技术人员有动机根据实际需求选取确定目标对象的方法，在此不多作限定。
另外，在一实施例中，参照图5，在执行完图1实施例所示的步骤S120之前，还包括但不限于有以下步骤S510。
步骤S510，从DT虚拟模型获取与物理模型相对应的环境数据集，环境数据集包括物理模型中各个实体对象的性能参数和状态参数。
需要说明的是，在DT case构建完成后，可以实时监控物理模型的网络环境，并且收集对应的环境数据，从而形成由多种环境数据构成的环境数据集，环境数据集可以包括每一个实体对象的性能参数和状态参数，具体采集方式在此不多作赘述。
另外，在一实施例中，参照图6，在执行完图5实施例所示的步骤S510之后，还包括但不限于有以下步骤S610至步骤S630。
步骤S610，获取预先设定的时间序列。
步骤S620，确定性能参数和状态参数在时间序列的概率分布。
步骤S630，根据概率分布和时间序列构建信号发生模型。
需要说明的是，为了避免物理数据采样的非及时性、意外中断、或者采样条件不具备等情况，在获取到性能参数和状态参数之后，可以采用核平滑的方法得到对应参数在指定的时间序列上的概率分布，并以此构建信号发生模型，使得信号发生模型能够根据概率分布生成 的真实环境中未出现的数据，作为样本参数用于感知和认知，实现网络分析和管控。
需要说明的是，时间序列可以是任意长度的一段时间，例如以链路流量为例，可以以月为单位，采集每个月第一天的链路流量，通过核平滑的方式确定链路流量在数值上的概率分布，构建出的信号发生模型能够根据该概率分布确定未来若干个月内第一天的链路流量。
另外，在一实施例中，参照图7，图1实施例所示的步骤S110，还包括但不限于有以下步骤S710至步骤S720。
步骤S710，确定目标分析时序。
步骤S720，根据信号发生模型构建出目标对象在目标分析时序的参数变化值。
需要说明的是，目标分析时序可以是任意时长的时间段，例如未来3个月的第一天，根据实际分析需求确定即可。
需要说明的是，由于信号发生模型根据性能参数和状态参数的概率分布确定，因此在确定目标分析时序之后，可以通过信号发生模型生成目标对象在目标分析时序的参数变化值，通过模拟环境产生数据，提高仿真效率。
另外，在一实施例中，参照图8，图1实施例所示的步骤S130，还包括但不限于有以下步骤S810至步骤S820。
步骤S810，获取预先设定好的目标状态和约束条件。
步骤S820，将环境数据集、目标状态、状态预测结果和约束条件输入至认知模型，获取认知模型输出的针对目标对象的配置调整信息。
需要说明的是，目标状态可以是针对目标对象，也可以是针对整个物理模型的，例如对于OTN网络，目标对象可以是传输链路，目标状态可以设定为“‘负载均衡度最大’作轨迹最终回报”，从而通过认知模型得到能够实现上述目标状态的配置调整信息，。
需要说明的是，约束条件可以预先设定，也可以根据感知模型输出的状态预测结果确定，例如需要模拟一条链路故障的情况，则约束条件可以设置为该传输链路不可用，使得认知模型能够根据约束条件进行认知预测，又如，在得到的状态预测结果中，路径A为拥塞路径，路径B为轻载路径，则可以将路径A作为必避约束，路径B作为必经约束，具体的约束条件选取方式可以根据实际分析需求确定，在此不多作限定。
值得注意的是，认知模型可以采用Actor-Critic的DRL技术，将环境数据集、目标状态、状态预测结果和约束条件输入至认知模型，能够通过DRL技术进行全局或者局部认知优化，从而获得认知优化后的配置调整信息，从而实现网络优化。
另外，在一实施例中，参照图9，图1实施例所示的步骤S140，还包括但不限于有以下步骤S910至步骤S920。
步骤S910，根据配置调整信息进行仿真验证，并获取仿真结果。
步骤S920，当仿真结果表征物理模型的运行状态符合预设标准，确定配置调整信息通过仿真验证。
需要说明的是，配置调整信息根据当前的物理环境和感知模型的预测结果得到，在将配置调整信息应用到物理模型之后，可能会出现衍生影响，例如对作为目标对象的传输链路的流量进行限定之后，导致其他传输链路的流量增加，从而出现链路拥塞的，为了确保优化效果，在得到配置调整信息后可以先通过仿真工具进行仿真，本实施例并不对仿真工具和具体仿真方法做出限定，本领域技术人员在具备配置调整信息的情况下，有动机根据实际需求选 取合适的仿真工具。
值得注意的是，预设标准可以根据对物理网络的性能需求确定，例如调整之后传输链路不会出现拥塞，或者流量满足预先设定的阈值，根据实际需求调整预设标准即可。
为了更好地说明本申请实施例的DT case的协同关系，以下提出一个具体示例，如图10所示，本示例的网络管控方法包括但不限于有以下步骤S1010至步骤S1080。
步骤S1010，实时监控物理网络环境，收集与该DT case相关的网络环境数据。
步骤S1021，针对指定的物理模型构建DT case，其中，DT case包括基础模型和功能模型，基础模型为DT虚拟模型，功能模型包括感知模型和认知模型，执行步骤S1031。
步骤S1022，根据DT虚拟模型构建环境数据集，执行步骤S1032或者步骤S1033。
步骤S1031，根据环境数据集确定物理网络环境的参数变化值，若需要进行预测并而且感知模型可用，执行步骤S1042或者步骤S1043，若不需要进行预测，执行步骤S1041。
步骤S1032，根据环境数据集构建信号发生器，若感知模型可用，执行步骤S1042。
步骤S1033，采用AI技术训练、构建、刷新对应的感知模型。
步骤S1041,基于DT虚拟模型判断分析物理网络环境变化，得到状态预测结果,若感知模型可用，执行步骤S1060。
步骤S1042，将参数变化值作为输入，得到状态预测结果,若感知模型可用，执行步骤S1060。
步骤S1043，根据信号发生器产生网络性能的时间序列的参数变化值，预测出状态预测结果，若感知模型可用，执行步骤S1060。
步骤S1050，根据环境数据集合状态预测结果构建、训练认知模型。
步骤S1060，将环境数据集、状态预测结果作为认知模型的输入，预测得到配置变更信息。
步骤S1070，调用emulator仿真工具对配置变更信息进行仿真，确定配置变更的可行性以及对网络环境的衍生影响,若仿真结果满足网络运营需求，执行步骤S1080，否则重新执行步骤S1060。
步骤S1080，将配置变更信息应用至物理网络。
另外，在一实施例中，参照图11，图5实施例所示的步骤S520，还包括但不限于有以下步骤S1110。
步骤S1110，通过反向传播算法进行从认知模型到感知模型的训练，其中，性能参数为感知模型的训练输入，状态参数为感知模型的训练输出，状态预测结果和环境数据集为认知模型的训练输入，目标状态确定为认知模型的输出。
需要说明的是，在具备环境数据集的基础上，根据性能参数作为输入，状态参数作为输出，使得训练后的感知模型能够根据性能参数的变化预测出对应的状态参数，例如，利用循环神经网络(Recurrent Neural Network，RNN)作为感知模型的AI算法，本领域技术人员熟知如何进行深度学习模型的训练，在此不多作赘述。
值得注意的是，在本实施例中，将感知模型与认知模型串联组合，采用反向传播算法对认知模型、感知模型中的模型参数，可以通过基于链式求导法则的梯度下降/上升法，进行迭代更新，直至对串联模型参数的梯度逼近到0，整个迭代更新结束，进而完成从认知模型到感知模型的端到端AI模型的统一训练，反向传播算法和链式求导的方法为本领域技术人员熟 知的技术，在此不多作赘述。通过反向传播的方式进行训练，能够实现认知模型到感知模型的一体化串联，有利于实现对物理网络的自动检测、自动优化和自动排障，提高物理网络的稳定性。
值得注意的是，感知模型采用RNN技术，可以把在整个时间序列周期中不同时刻网络业务流量分布作为RNN输入，用以预测OTN网络的关键链路在整个时间序列周期中不同时刻的状态，并结合当前网络的拓扑状态，作为认知模型的输入，认知模型可以采用Actor-Critic的DRL技术，训练后用以推理获得最佳的OTN网络流量优化变更配置方案，例如图12所示的的结构，性能参数为网络业务流量分布，其中t为性能参数的数量，将X1至Xt输入至RNN单元中进行预测，其中X1输入RNN单元1后，输出的状态预测结果为键链路在整个时间序列周期中不同时刻的状态，其中，输入X1所对应的输出为Z1，以此类推，在感知模型输出状态预测结果后，Z1至Zt以及环境数据集作为认知模型的训练输入，以目标状态作为认知模型的目标函数，从而实现认知模型的训练，本领域技术人员在具备输入和输出的情况下，熟知如何进行深度学习网络的训练，在此不多作赘述。
为了更好地说明本申请的技术方案，以下提出一个具体示例。
在本示例中，物理模型以图13所示的OTN Mesh组网拓扑为例，在该网络中具有6条数据单元ODUflex业务路径，分别为路径1：A-S、路径2：A-K、路径3：K-C、路径4：A-B-C、路径5：T-C、路径6：S-N-T；其中，在该物理模型中还包括关键链路1：S-N，关键链路2：A-B，其中，本示例以15分钟为目标分析时序。
参照图14，本示例的方法包括但不限于有以下步骤S1410至步骤S1450。
步骤S1410，将当前的OTN网络流量分布确定为DT case的目标对象，构建对应的DT case的DT虚拟模型。
步骤S1420，基于DT case的OTN网络流量抽象化虚拟模型及OTN网络流量感知分析需求，构建OTN网络流量感知算法模型。
需要说明的是，感知模型的结构可以参考图12所示，以RNN为例，每一时刻t的输入为
Figure PCTCN2022109218-appb-000001
为向量形式，该向量的每个元素为连续变量，表示t时刻每条ODUflex业务的每15分钟平均流量值，单位Gbit/s。
步骤S1430，通过感知模型预测出关键链路的趋势状态。
需要说明的是，根据上述X t的输入，预测出对应的每一时刻t的输出为
Figure PCTCN2022109218-appb-000002
其中，Z t为向量形式，每个元素为离散变量，表示t时刻n条OTN网络拓扑关键链路的趋势状态值，趋势状态以数值的方式描述，例如0表示流量正常状态、1表示潮张状态、2表示流量拥塞、3表示潮落状态、4表示流量轻载，
步骤S1440，基于DT case的OTN网络流量抽象化虚拟模型、OTN网络流量感知算法模型 的推断输出、OTN网络流量认知分析需求，构建DT case的OTN网络流量认知算法模型。
需要说明的是，在本示例中，DT case实现OTN网络流量认知分析流程的认知算法模型主要采用DRL技术的Actor-Critic算法，该Actor-Critic算法实例的输入包括：OTN当前的网络拓扑、业务分布、各关键链路在0至t时刻的拥塞趋势状态，基于上述输入，还可以将拥塞链路作为必避约束、将轻载链路作为必经约束。
步骤S1450，获取DT case的OTN网络流量认知算法模型输出的调整策略，根据调整策略调整物理模型。
另外，参照图15，本申请的一个实施例还提供了一种网络管控系统，该网络管控系统1500包括：存储器1510、处理器1520及存储在存储器1510上并可在处理器1520上运行的计算机程序。
处理器1520和存储器1510可以通过总线或者其他方式连接。
实现上述实施例的网络管控方法所需的非暂态软件程序以及指令存储在存储器1510中，当被处理器1520执行时，执行上述实施例中的网络管控方法，例如，执行以上描述的图1中的方法步骤S110至步骤S140、图4中的方法步骤S410至步骤S420、图5中的方法步骤S510、图6中的方法步骤S610至步骤S630、图7中的方法步骤S710至步骤S720、图8中的方法步骤S810至步骤S820、图9中的方法步骤S910至步骤S920、图11中的方法步骤S1110。
以上所描述的装置实施例仅仅是示意性的，其中作为分离部件说明的单元可以是或者也可以不是物理上分开的，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。
此外，本申请的一个实施例还提供了一种计算机可读存储介质，该计算机可读存储介质存储有计算机可执行指令，该计算机可执行指令被一个处理器或控制器执行，例如，被上述网络管控系统实施例中的一个处理器执行，可使得上述处理器执行上述实施例中的网络管控方法，例如，执行以上描述的图1中的方法步骤S110至步骤S140、图4中的方法步骤S410至步骤S420、图5中的方法步骤S510、图6中的方法步骤S610至步骤S630、图7中的方法步骤S710至步骤S720、图8中的方法步骤S810至步骤S820、图9中的方法步骤S910至步骤S920、图11中的方法步骤S1110。本领域普通技术人员可以理解，上文中所公开方法中的全部或某些步骤、系统可以被实施为软件、固件、硬件及其适当的组合。某些物理组件或所有物理组件可以被实施为由处理器，如中央处理器、数字信号处理器或微处理器执行的软件，或者被实施为硬件，或者被实施为集成电路，如专用集成电路。这样的软件可以分布在计算机可读介质上，计算机可读介质可以包括计算机存储介质(或非暂时性介质)和通信介质(或暂时性介质)。如本领域普通技术人员公知的，术语计算机存储介质包括在用于存储信息(诸如计算机可读指令、数据结构、程序模块或其他数据)的任何方法或技术中实施的易失性和非易失性、可移除和不可移除介质。计算机存储介质包括但不限于RAM、ROM、EEPROM、闪存或其他存储器技术、CD-ROM、数字多功能盘(DVD)或其他光盘存储、磁盒、磁带、磁盘存储或其他磁存储装置、或者可以用于存储期望的信息并且可以被计算机访问的任何其他的介质。此外，本领域普通技术人员公知的是，通信介质通常包含计算机可读指令、数据结构、程序模块或者诸如载波或其他传输机制之类的调制数据信号中的其他数据，并且可包括任何信息递送介质。
本申请实施例包括：获取目标对象的参数变化值，其中，所述参数变化值来自于DT虚拟 模型，DT虚拟模型根据物理模型构建，所述物理模型由物理网络的实体对象构成，所述参数变化值表征所述目标对象的传输性能的变化；将所述参数变化值输入至预先训练好的感知模型，获取所述感知模型输出的状态预测结果；将所述状态预测结果输入至预先训练好的认知模型，获取所述认知模型输出的配置调整信息；当所述配置调整信息通过仿真验证，根据所述配置调整信息调整所述物理模型。根据本申请实施例提供的方案，能够以DT虚拟模型作为数据基础，通过感知模型感知、预测物理模型的变化及变化趋势，并将感知、预测结果作为认知模型的输入，得到物理网络的配置调整信息，并经过仿真验证后实现针对物理网络的自动管控。本申请实施例能够结合DT技术和认知光网络技术，实现光网络的自动管控。
以上是对本申请的一些实施进行了具体说明，但本申请并不局限于上述实施方式，熟悉本领域的技术人员在不违背本申请精神的前提下还可作出种种的等同变形或替换，这些等同的变形或替换均包含在本申请权利要求所限定的范围内。

Claims (11)

  1. 一种网络管控方法，包括：
    获取目标对象的参数变化值，其中，所述参数变化值来自于数字孪生DT虚拟模型，DT虚拟模型根据物理模型构建，所述物理模型由物理网络的实体对象构成，所述参数变化值表征所述目标对象的传输性能的变化；
    将所述参数变化值输入至预先训练好的感知模型，获取所述感知模型输出的状态预测结果；
    将所述状态预测结果输入至预先训练好的认知模型，获取所述认知模型输出的配置调整信息；
    当所述配置调整信息通过仿真验证，根据所述配置调整信息调整所述物理模型。
  2. 根据权利要求1所述的方法，其中，在所述获取目标对象的参数变化值之前，所述方法还包括:
    将所述物理模型中性能参数发生变化的实体对象确定为所述目标对象；
    或者，
    获取分析需求信息，将所述分析需求信息中的待分析对象确定为所述目标对象。
  3. 根据权利要求1所述的方法，其中，在所述将所述参数变化值输入至预先训练好的感知模型之前，所述方法还包括：
    从所述DT虚拟模型获取与所述物理模型相对应的环境数据集，所述环境数据集包括所述物理模型中各个所述实体对象的性能参数和状态参数。
  4. 根据权利要求3所述的方法，其中，在所述从所述DT虚拟模型获取与所述物理模型相对应的环境数据集之后，所述方法还包括：
    获取预先设定的时间序列；
    确定所述性能参数和所述状态参数在所述时间序列的概率分布；
    根据所述概率分布和所述时间序列构建信号发生模型。
  5. 根据权利要求4所述的方法，其中，所述获取目标对象的参数变化值，包括：
    确定目标分析时序；
    根据所述信号发生模型构建出所述目标对象在所述目标分析时序的所述参数变化值。
  6. 根据权利要求3所述的方法，其中，所述将所述状态预测结果输入至预先训练好的认知模型，获取所述认知模型输出的配置调整信息，包括：
    获取预先设定好的目标状态和约束条件；
    将所述环境数据集、所述目标状态、所述状态预测结果和所述约束条件输入至所述认知模型，获取所述认知模型输出的针对所述目标对象的配置调整信息。
  7. 根据权利要求1所述的方法，其中，根据当所述配置调整信息通过仿真验证，包括：
    根据所述配置调整信息进行仿真验证，并获取所述仿真结果；
    当所述仿真结果表征所述物理模型的运行状态符合预设标准，确定所述配置调整信息通过所述仿真验证。
  8. 根据权利要求6所述的方法，其中，在所述从所述DT虚拟模型获取与所述物理模型相对应的环境数据集之后，包括：
    通过反向传播算法进行从所述认知模型到所述感知模型的训练，其中，所述性能参数为所述感知模型的训练输入，所述状态参数为所述感知模型的训练输出，所述状态预测结果和所述环境数据集为所述认知模型的训练输入，所述目标状态确定为所述认知模型的输出。
  9. 根据权利要求8所述的方法，其中，所述通过反向传播算法进行所述认知模型和所述感知模型的训练，包括：
    按照由输出端到输入端的顺序，确定针对所述认知模型和所述感知模型中每一层网络的输出结果的梯度值；
    当所述梯度值满足预设训练条件，确定从所述认知模型的输出端至和所述感知模型的输入端的训练完成。
  10. 一种网络管控系统，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，其中，所述处理器执行所述计算机程序时实现如权利要求1至9中任意一项所述的网络管控方法。
  11. 一种计算机可读存储介质，存储有计算机可执行指令，其中，所述计算机可执行指令用于执行如权利要求1至9中任意一项所述的网络管控方法。
PCT/CN2022/109218 2021-09-22 2022-07-29 网络管控方法及其系统、存储介质 Ceased WO2023045565A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
EP22871623.9A EP4390783A4 (en) 2021-09-22 2022-07-29 NETWORK MANAGEMENT AND CONTROL METHOD, AND RELATED SYSTEM, AND STORAGE MEDIUM
US18/684,040 US12574299B2 (en) 2021-09-22 2022-07-29 Network management and control method and system thereof, and storage medium
KR1020247011359A KR20240061646A (ko) 2021-09-22 2022-07-29 네트워크 관리 방법 및 그 시스템, 저장 매체

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111108462.1 2021-09-22
CN202111108462.1A CN115865679A (zh) 2021-09-22 2021-09-22 网络管控方法及其系统、存储介质

Publications (1)

Publication Number Publication Date
WO2023045565A1 true WO2023045565A1 (zh) 2023-03-30

Family

ID=85652122

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/109218 Ceased WO2023045565A1 (zh) 2021-09-22 2022-07-29 网络管控方法及其系统、存储介质

Country Status (5)

Country Link
US (1) US12574299B2 (zh)
EP (1) EP4390783A4 (zh)
KR (1) KR20240061646A (zh)
CN (1) CN115865679A (zh)
WO (1) WO2023045565A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116132337A (zh) * 2023-04-04 2023-05-16 深圳行云创新科技有限公司 一种基于服务网格技术的接口流量异常检测方法
CN117201325A (zh) * 2023-09-07 2023-12-08 之江实验室 一种网络仿真脚本的生成方法及装置
WO2025020808A1 (zh) * 2023-07-21 2025-01-30 中兴通讯股份有限公司 网络能效优化方法、计算机设备和计算机可读存储介质
WO2025124747A1 (en) * 2023-12-15 2025-06-19 Telefonaktiebolaget Lm Ericsson (Publ) Acquiring updated values of parameters of a digital twin

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119696983A (zh) * 2023-09-22 2025-03-25 华为技术有限公司 网络管理操作指令执行方法、装置、系统及存储介质
CN121234539A (zh) * 2024-06-30 2025-12-30 中兴通讯股份有限公司 数据采集方法、电子设备及存储介质
EP4704391A1 (en) * 2024-09-03 2026-03-04 Nokia Solutions and Networks Oy Digital twin of a network
CN119766691B (zh) * 2024-12-31 2025-11-18 中国移动通信集团广东有限公司 中断感知识别方法、装置、电子设备、存储介质及产品

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103151049A (zh) * 2013-01-29 2013-06-12 武汉大学 一种面向移动音频的服务质量保障方法及系统
CN109787928A (zh) * 2019-01-25 2019-05-21 中国人民解放军国防科技大学 基于决策树twsvm的高速光信号调制格式识别方法
CN111208759A (zh) * 2019-12-30 2020-05-29 中国矿业大学(北京) 矿井无人化综采工作面数字孪生智能监控系统
TWI700907B (zh) * 2019-12-02 2020-08-01 中華電信股份有限公司 光纖迴路障礙診斷方法
CN111835565A (zh) * 2020-07-06 2020-10-27 重庆金美通信有限责任公司 一种基于数字孪生的通信网络优化方法、装置和系统
CN112187547A (zh) * 2020-10-09 2021-01-05 南京邮电大学 一种基于数字孪生的网络模型
CN113111473A (zh) * 2021-04-25 2021-07-13 北京邮电大学 传输系统功率谱的优化方法、装置及电子设备

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11038769B2 (en) * 2017-11-16 2021-06-15 Verizon Patent And Licensing Inc. Method and system for virtual network emulation and self-organizing network control using deep generative models
CN108763398B (zh) 2018-05-22 2021-08-17 腾讯科技（深圳）有限公司 数据库配置参数处理方法、装置、计算机设备和存储介质
US10938666B2 (en) * 2018-08-23 2021-03-02 Hewlett Packard Enterprise Development Lp Network testing simulation
US11074149B2 (en) * 2019-04-30 2021-07-27 At&T Intellectual Property I, L.P. Cloud simulation and validation system
US11088921B2 (en) * 2019-09-06 2021-08-10 Eagle Technology, Llc Systems and method for providing an ontogenesis emergence and confidence engine
US11595271B2 (en) * 2021-07-20 2023-02-28 EMC IP Holding Company LLC Digital twin architecture for multi-access edge computing environment

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103151049A (zh) * 2013-01-29 2013-06-12 武汉大学 一种面向移动音频的服务质量保障方法及系统
CN109787928A (zh) * 2019-01-25 2019-05-21 中国人民解放军国防科技大学 基于决策树twsvm的高速光信号调制格式识别方法
TWI700907B (zh) * 2019-12-02 2020-08-01 中華電信股份有限公司 光纖迴路障礙診斷方法
CN111208759A (zh) * 2019-12-30 2020-05-29 中国矿业大学(北京) 矿井无人化综采工作面数字孪生智能监控系统
CN111835565A (zh) * 2020-07-06 2020-10-27 重庆金美通信有限责任公司 一种基于数字孪生的通信网络优化方法、装置和系统
CN112187547A (zh) * 2020-10-09 2021-01-05 南京邮电大学 一种基于数字孪生的网络模型
CN113111473A (zh) * 2021-04-25 2021-07-13 北京邮电大学 传输系统功率谱的优化方法、装置及电子设备

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4390783A4 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116132337A (zh) * 2023-04-04 2023-05-16 深圳行云创新科技有限公司 一种基于服务网格技术的接口流量异常检测方法
CN116132337B (zh) * 2023-04-04 2023-06-13 深圳行云创新科技有限公司 一种基于服务网格技术的接口流量异常检测方法
WO2025020808A1 (zh) * 2023-07-21 2025-01-30 中兴通讯股份有限公司 网络能效优化方法、计算机设备和计算机可读存储介质
CN117201325A (zh) * 2023-09-07 2023-12-08 之江实验室 一种网络仿真脚本的生成方法及装置
WO2025124747A1 (en) * 2023-12-15 2025-06-19 Telefonaktiebolaget Lm Ericsson (Publ) Acquiring updated values of parameters of a digital twin

Also Published As

Publication number Publication date
CN115865679A (zh) 2023-03-28
US20240396810A1 (en) 2024-11-28
KR20240061646A (ko) 2024-05-08
EP4390783A4 (en) 2024-12-04
EP4390783A1 (en) 2024-06-26
US12574299B2 (en) 2026-03-10

Similar Documents

Publication Publication Date Title
WO2023045565A1 (zh) 网络管控方法及其系统、存储介质
CN120416056A (zh) 深度学习及拓扑分析的网络拓扑智能生成方法及系统
CN106953802A (zh) 一种基于深度学习的网络最优路径选择方法
CN116304089A (zh) 一种融合权重与时态信息的知识图谱补全方法
Liu et al. Integrating artificial bee colony algorithm and BP neural network for software aging prediction in IoT environment
CN119420691A (zh) 动态多链路智能管理与调度系统
CN120811853A (zh) 一种基于认知数字孪生技术的计算机网络管理系统
CN118945100A (zh) 一种基于深度强化学习的路由调度方法
CN120710881A (zh) 一种基于人工智能的网络处理系统
Duran et al. Digital twin-native ai-driven service architecture for industrial networks
Wang et al. Topology-aware graph reinforcement learning for dynamic routing in cloud networks
Li et al. Automatic data generation and optimization for digital twin network
CN120811605B (zh) 面向城市干线网络的可重构量子通信路由同步系统
CN115550233A (zh) 分布式路由确定方法、电子设备及存储介质
CN120821727B (zh) 基于神经网络的多源异构能源数据融合方法及系统
CN116260709A (zh) 一种基于故障传播关系的通信网络故障定位方法及系统
CN121000613A (zh) 一种基于数字孪生的网络仿真方法、装置及系统
CN115422302A (zh) 一种基于数据仓库的分布式灾备演练数据优化及预警方法
CN119182654A (zh) SRv6网络故障的智能诊断方法以及装置
CN117792987A (zh) 基于边缘计算的应急通信通道分流方法
CN121462529B (zh) 网络资源分配方法、电子设备及存储介质
CN121462429B (zh) 基于大模型智能体的跨环境泛化网络流量优化方法
Bolodurina et al. Development and research of model for ensuring reliability of operation of network infrastructure objects in cyber-physical system located in the cloud platform
CN121585553A (zh) 基于数据分析的网络性能优化系统及方法
Kumar AI-Based Network Traffic Optimization

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22871623

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 18684040

Country of ref document: US

WWE Wipo information: entry into national phase

Ref document number: 2401001880

Country of ref document: TH

WWE Wipo information: entry into national phase

Ref document number: 2022871623

Country of ref document: EP

ENP Entry into the national phase

Ref document number: 2022871623

Country of ref document: EP

Effective date: 20240322

ENP Entry into the national phase

Ref document number: 20247011359

Country of ref document: KR

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

WWG Wipo information: grant in national office

Ref document number: 18684040

Country of ref document: US