A fast self-healing method and system based on AI prediction and T-RMR

Abstract

This application provides a rapid self-healing method and system based on AI prediction and T-RMR, relating to the field of self-healing protection for power communication networks. The method includes: collecting real-time indicator data through front-end sensing devices and performing AI prediction based on this data to obtain the node health of each communication node, the link health of each communication link, and predicted alarm events for the next moment; acquiring network topology information, device capabilities, node health, link health, and predicted alarm events through a digital twin platform, generating the T-RMR topology and the latest time slot mapping relationship between the main ring and the redundant ring, and distributing the latest time slot mapping relationship to each network node device through an SDN controller; and performing time slot switching based on the latest time slot mapping relationship when a communication link fault or node fault is detected. This method enables low-latency, seamless switching during faults.

Description

Technical Field

[0001] This application relates to the field of self-healing protection of power communication networks, and in particular to a rapid self-healing method and system based on AI prediction and T-RMR. Background Technology

[0002] With the large-scale integration of smart grids and new energy sources, power communication networks bear the responsibility of real-time transmission of massive amounts of data and command issuance. Their network reliability and latency performance directly affect the safe and stable operation of the power grid. In traditional architectures, most systems employ single-ring or dual-ring SONET / SDH ring network protection mechanisms, such as Unidirectional Path Switched Ring (UPSR) and Bidirectional Line Switched Ring (BLSR). Their fault detection and switching latency is typically between 10ms (detection) and 30-50ms (switching), with a total switching latency of 40-60ms, insufficient to meet the low-latency, rapid self-healing requirements of modern power grids. Furthermore, traditional ring network protection relies heavily on statically preset protection paths and manual configuration. After a link or node failure, manual or semi-automated maintenance is required, resulting in a time-consuming recovery process, low resource utilization, and further reliability degradation when facing complex topologies or multi-point faults. Summary of the Invention

[0003] This application provides a rapid self-healing method and system based on AI prediction and T-RMR, which can achieve seamless switching with low latency in the event of a fault. The technical solution is as follows:

[0004] According to one aspect of this application, a rapid self-healing method based on AI prediction and T-RMR is provided. This method is applied to a power communication network self-healing system, which includes at least front-end sensing devices, a digital twin platform, an SDN controller, and network node devices. The method includes:

[0005] Real-time indicator data is collected by the front-end sensing device, and AI prediction is performed based on the real-time indicator data to obtain the node health of each communication node, the link health of each communication link, and the predicted alarm events at the next moment.

[0006] The digital twin platform acquires the network topology information, device capabilities, node health, link health, and predicted alarm events, generates the T-RMR topology and the latest time slot mapping relationship between the main ring and the redundant ring, and distributes the latest time slot mapping relationship to each network node device through the SDN controller.

[0007] When a communication link failure or node failure is detected, the network node device performs time slot switching based on the latest time slot mapping relationship.

[0008] According to another aspect of this application, a power communication network self-healing system is provided, which includes at least a front-end sensing device, a digital twin platform, an SDN controller, and network node devices:

[0009] The front-end sensing device is used to collect real-time indicator data and perform AI prediction based on the real-time indicator data to obtain the node health of each communication node, the link health of each communication link, and the predicted alarm events at the next moment.

[0010] The digital twin platform is used to acquire network topology information, device capabilities, node health, link health and predicted alarm events, generate T-RMR topology and the latest time slot mapping relationship between the main ring and the redundant ring, and distribute the latest time slot mapping relationship to each network node device through the SDN controller.

[0011] The network node device is used to perform time slot switching based on the latest time slot mapping relationship when a communication link failure or node failure is detected.

[0012] According to another aspect of this application, a computer device is provided, comprising: a processor and a memory storing a program, the program including instructions that, when executed by the processor, cause the processor to perform the rapid self-healing method based on AI prediction and T-RMR as described above.

[0013] According to another aspect of this application, a non-transitory computer-readable storage medium is provided storing computer instructions for causing the computer to perform the rapid self-healing method based on AI prediction and T-RMR as described above.

[0014] According to another aspect of this application, a computer program product is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the aforementioned rapid self-healing method based on AI prediction and T-RMR.

[0015] The beneficial effects of the technical solutions provided in this application include at least the following:

[0016] This application proposes a rapid self-healing method and system for multi-ring networks based on AI prediction and time-series redundancy. AI prediction and hierarchical alarms are implemented: lightweight neural network models are deployed at each node to continuously monitor and predict key indicators such as link latency, optical power, and jitter; alarms are automatically graded (normal / warning / fault) based on the prediction results, achieving a closed loop of "0 → warning → re-fault". Furthermore, outside the main ring, multiple logically redundant rings are divided according to time slots, without adding physical optical cables, achieving multiple backups only through time slot multiplexing and time slot switching; dynamic reconfiguration using a time slot mapping table (TS-MAP) allows switching to be completed within a sub-1.25μs time slot granularity when a single link or node fails. In addition, a digital twin platform for the communication network is established to aggregate network status and AI prediction data in real time. The SDN controller runs online optimization algorithms based on the twin, and NFV nodes deploy self-healing service chains, automatically issuing commands such as rerouting and time slot table updates. During the recovery process, bypass traffic can be seamlessly migrated in the virtualized switching plane. Fault warning is achieved through AI models, fine-grained physical layer protection is provided by multiple time-slot-level logic loops, and seamless switching with low latency is achieved by combining digital twins and SDN / NFV control. Attached Figure Description

[0017] Further details, features, and advantages of this application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0018] Figure 1 A flowchart of a rapid self-healing method based on AI prediction and T-RMR according to an exemplary embodiment of this application is shown;

[0019] Figure 2 This is a schematic diagram of T-RMR provided in an exemplary embodiment of this application;

[0020] Figure 3 This is an overall flowchart of a digital twin platform execution method provided in an exemplary embodiment of this application;

[0021] Figure 4 A flowchart is shown for another rapid self-healing method based on AI prediction and T-RMR according to an exemplary embodiment of this application;

[0022] Figure 5 A flowchart is shown for another rapid self-healing method based on AI prediction and T-RMR according to an exemplary embodiment of this application;

[0023] Figure 6 This is a schematic diagram of the structure of a power communication network self-healing system provided in an embodiment of this application;

[0024] Figure 7 A structural block diagram of an exemplary computer device that can be used to implement embodiments of this application is shown. Detailed Implementation

[0025] Embodiments of this application will now be described in more detail with reference to the accompanying drawings. While some embodiments of this application are shown in the drawings, it should be understood that this application can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this application. It should be understood that the drawings and embodiments of this application are for illustrative purposes only and are not intended to limit the scope of protection of this application.

[0026] It should be understood that the steps described in the method embodiments of this application may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this application is not limited in this respect.

[0027] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc., mentioned in this application are only used to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies. It should be noted that the modifications "a" and "a plurality" mentioned in this application are illustrative and not restrictive, and those skilled in the art should understand that unless explicitly indicated in the context, they should be understood as "one or more". The names of messages or information exchanged between multiple devices in the embodiments of this application are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0028] The present invention will now be described with reference to the accompanying drawings. The technical solutions provided by the embodiments of this application will be explained in detail through specific examples and application scenarios.

[0029] With the large-scale integration of smart grids and new energy sources, power communication networks bear the responsibility of real-time transmission of massive amounts of data and command issuance. Their network reliability and latency performance directly affect the safe and stable operation of the power grid. In traditional architectures, most systems employ single-ring or dual-ring SONET / SDH ring network protection mechanisms, such as unidirectional path protection rings and bidirectional line protection rings. Their fault detection and switching latency is typically between 10ms (detection) and 30-50ms (switching), with a total switching latency of 40-60ms, insufficient to meet the low-latency, rapid self-healing requirements of modern power grids. Furthermore, traditional ring network protection relies heavily on statically pre-configured protection paths and manual configuration. After a link or node failure, manual or semi-automated processes are required for maintenance, resulting in long recovery times, low resource utilization, and further reliability degradation when facing complex topologies or multi-point faults. For the energy internet and multi-agent collaborative control, self-healing power grid systems have proposed schemes based on real-time scheduling of multiple agents. These schemes ensure the timely execution of fault recovery operation sequences through fault chains, safety trees, and deadline constraints, but do not achieve ultra-low latency switching at the physical layer.

[0030] In recent years, Software-Defined Networking (SDN) and Network Functions Virtualization (NFV) technologies have been introduced into power communication networks. Through centralized control and virtualized service chains, they have enabled dynamic orchestration of network traffic and policy distribution. However, their protection switching is mainly concentrated at Layer 2 and above, lacking support for fine-grained time-slot-level backup and switching of physical links, making it difficult to overcome the latency bottleneck of traditional ring networks. Meanwhile, digital twins, as an emerging paradigm, provide virtualized images of network entities for power communication networks. They can aggregate link status, predictive data, and topology information in real time, providing global visualization support for automated decision-making. However, existing digital twin platforms mostly focus on performance monitoring and resource scheduling, lacking deep integration with physical layer time-slot-level redundancy mechanisms, and cannot achieve sub-µs ultra-low latency switching when faults occur. Furthermore, AI-driven fault prediction and anomaly detection have been widely applied in other fields, such as infrastructure digital twins combining rules and machine learning models to achieve early fault warning and self-healing optimization. However, in power communication networks, this is still in the exploratory stage, and a complete solution for fine-grained time-slot-level backup switching and multi-level autonomous collaboration has not yet been formed.

[0031] To address the switching latency bottleneck in related technologies, this application provides a rapid self-healing method and system based on AI prediction and T-RMR, enabling low-latency, seamless switching and recovery in the event of power communication network link or node failures. Please refer to... Figure 1 The document illustrates a flowchart of a rapid self-healing method based on AI prediction and T-RMR, according to an exemplary embodiment of this application. The method is described using an example of its application in a power communication network self-healing system. Figure 1 As shown, the method includes:

[0032] Step 101: Collect real-time indicator data through front-end sensing devices, and make AI predictions based on the real-time indicator data to obtain the node health of each communication node, the link health of each communication link, and the predicted alarm events at the next moment.

[0033] The power communication network self-healing system includes at least front-end sensing devices, a digital twin platform, an SDN controller, and network node devices.

[0034] In one possible implementation, front-end sensing devices (also known as data acquisition agents) are deployed at each network node device (also known as a communication node). The front-end sensing devices collect real-time indicator data of each communication node in real time, and perform AI prediction based on the real-time indicator data to obtain the node health of each communication node, the link health of each communication link, and the predicted alarm events at the next moment.

[0035] Specifically, the process of front-end sensing devices collecting data and making AI predictions can include a data collection agent deployment stage, a real-time indicator collection stage, a local AI model prediction stage, and a tiered alarm push stage.

[0036] During the data acquisition agent deployment phase: Install a lightweight data acquisition agent on each communication node, such as a switch or OLT / ONU, supporting gNMI / NetConf or SNMP Pull / Push mode to ensure compatibility with the existing network management system; configure the agent to maintain a long connection (TLS+gRPC) with the upper-layer platform and report the acquisition status and heartbeat periodically (e.g., 1ms~10ms).

[0037] During the real-time data acquisition phase: the deployed data acquisition agent reads system metrics (i.e., real-time metrics data) such as optical power, bit error rate (BER), latency, jitter, CPU, memory, and port utilization from the physical layer and the device itself; it then performs sliding window aggregation (e.g., within 100ms), noise reduction, and normalization on the raw data. Sliding window aggregation refers to summarizing, denoising, and normalizing the raw sampling points within a window every 100ms as a processing unit. The main purpose of this step is to reduce the data volume (gigabit sampling can be reduced to hundreds of megabytes of statistical value), improve data quality by suppressing noise, and standardize the input scale for subsequent AI model processing.

[0038] In the local AI model prediction stage, the steps of making AI predictions based on real-time indicator data to obtain the node health of each communication node and the link health of each communication link at the next moment may include steps 101A and 101B.

[0039] Step 101A: Input the real-time indicator data into the short-term prediction model to obtain the predicted indicator data for the next time step output by the short-term prediction model.

[0040] During the offline phase, the server aggregates historical data, trains a Transformer short-term prediction model, and exports lightweight weights. During the deployment phase, the trained short-term prediction model (such as in ONNX format) is distributed to the data acquisition agent of each communication node and loaded into the embedded TensorRT runtime. During the online phase, the data acquisition agent performs forward inference once per sliding window cycle to generate the link health prediction for the next moment (i.e., the node health of each communication node and the link health of each communication link at the next moment).

[0041] In one possible implementation, real-time indicator data collected during the current sliding window period is input into a short-term prediction model, which outputs predicted indicator data for the next time step. The real-time indicator data may include bit error rate, latency, jitter, optical power, CPU utilization, and memory utilization; the output is the predicted indicator data for the next time step, which may include predicted values ​​for bit error rate, latency, jitter, optical power, CPU utilization, and memory utilization.

[0042] Each communication link corresponds to a unique physical port on the node. Real-time metrics such as bit error rate, latency, and jitter are collected through the interface and directly attributed to a specific communication link. Within each communication node, real-time metrics such as optical power, bit error rate, and latency are statistically analyzed for each physical port.

[0043] Step 101B: Based on the predicted index data and index calculation formula, determine the node health of each communication node and the link health of each communication link at the next moment.

[0044] For example, predefined formulas (1) and (2) are used to calculate the indicators of node health and link health:

[0045] (1)

[0046] (2)

[0047] Where (i, j) represents the link endpoints, For node health, For CPU utilization, For memory usage, For bit error rate, For link health, For time delay, The jitter value. For optical power, , , , , , As weight.

[0048] Optionally, , , , , , The weights are obtained through principal component analysis (PCA) based on historical data.

[0049] In one possible implementation, the predicted index data for the next moment output by the short-term prediction model is substituted into the index calculation formula to convert the predicted index data for the next moment into the predicted health status for the next moment. This determines the node health status of each communication node and the link health status of each communication link for the next moment, so as to predict whether the system will enter an "abnormal" or "faulty" state within the next T seconds based on the predicted health status.

[0050] In the tiered alarm push phase, the predicted alarm events are first determined based on node health, link health, and preset alarm thresholds, and then pushed to the digital twin platform. The predicted alarm threshold can be expressed as shown in formula (3):

[0051] (3)

[0052] Among them, with Taking the node health status as an example, a preset alarm threshold is set. and ,like Greater than If so, it means the communication node is normal. Less than and greater than This indicates that there is a fault warning in the communication node. Less than This indicates a fault in the communication node. Similarly, for... The (link health) also has preset alarm thresholds, which are divided into three levels: normal, warning and fault.

[0053] After determining the predicted alarm events based on node health and link health, the front-end sensing devices can push "early warning" events and "fault" events to the digital twin platform via the message bus (Kafka / AMQP). The digital twin platform is deployed in a regional edge cloud (not a central cloud), and because it needs to meet the requirement of less than 10ms latency for self-healing actions, it requires GPU resources to run reinforcement learning algorithms.

[0054] Optionally, when the local AI determines an emergency fault, the following logic can be executed: 1) Bypass the regular process without going through the digital twin platform; 2) Direct hardware intervention: activate the backup fiber optic link, perform a hard reboot of the faulty node to trigger the watchdog reset, and start local cache forwarding; 3) Accelerate detection: shorten the heartbeat interval from 10 ms to 1 ms. The goal is to complete the first level of protection within 3 ms, avoiding waiting for a response from the cloud.

[0055] Step 102: Obtain the network topology information, device capabilities, node health, link health and predicted alarm events through the digital twin platform, generate the T-RMR topology and the latest time slot mapping relationship between the main ring and the redundant ring, and distribute the latest time slot mapping relationship to each network node device through the SDN controller.

[0056] The digital twin platform can aggregate real-time network topology information, device capabilities, node health of each communication node, link health of each communication link, and predict alarm events to generate a T-RMR (Time Redundant Multi-Ring) topology and the latest time slot mapping relationship (TS-MAP) between the main ring and redundant rings. This latest time slot mapping relationship is then distributed to each network node device via the SDN controller. Outside the main ring, multiple logically redundant rings are divided according to time slots, without adding physical optical cables; multiple backups are achieved only through time slot multiplexing and time slot switching. Dynamic reconfiguration using the time slot mapping table (TS-MAP) allows for switching within a 1.25μs time slot granularity when a single link or node fails.

[0057] Please refer to Figure 2 This is a schematic diagram of T-RMR provided in an exemplary embodiment of this application. Figure 2 As shown, the physical ring network consists of four communication nodes: node A, node B, node C, and node D. On the same physical ring network, three independent logical rings are constructed through time slot division: the main service ring (…). - Occupied time slots 1, 4, 7, 10; Backup ring 1 ( - Occupied time slots 2, 5, 8, and 11; Backup ring 2 ( - Occupies time slots 3, 6, 9, and 12.

[0058] For example, with Figure 2 Based on this, the constructed time slot mapping relationship can be shown in Table 1:

[0059] Table 1

[0060]

[0061] If a fault exists: BC fiber breakage (affecting main ring services), the TS-MAP (time slot mapping table) will be dynamically updated as shown in Table 2.

[0062] Table 2

[0063]

[0064] Specifically, generating the T-RMR topology and the latest time slot mapping relationship between the main ring and the redundant ring may also include steps 102A to 102D.

[0065] Step 102A: Construct the T-RMR topology based on the network topology information and device capabilities;

[0066] Specifically, the entire network topology and device capabilities are retrieved in real time through interfaces such as NETCONF / RESTCONF and BGP LS. Device capabilities include hardware capabilities, number / rate of ports, cache size, supported time slot segmentation granularity, protocol support and virtualization capabilities, etc., to construct a T-RMR topology, representing the topological relationship between physical entities and logical time slots.

[0067] Step 102B: Determine the network failure risk value for the entire network based on node health and link health.

[0068] To avoid meaningless updates to the time slot mapping relationship, one possible implementation is to set the update trigger condition as follows: when the network fault risk value of the entire network exceeds the warning threshold, even if there is no actual fault at this time, a new time slot mapping relationship will be generated in advance so as to be sent to the network node device in advance before the fault occurs, thereby achieving the purpose of early intervention.

[0069] The digital twin platform includes a network-wide risk assessment model for comprehensive quantitative evaluation of the network's operational status. This model serves three main functions: First, it acts as a trigger mechanism for early intervention; when the calculated risk level (R) exceeds the warning threshold, TS-MAP switching can be initiated even without an actual fault. Second, the model serves as a reference for scheduling priorities in self-healing strategies, allowing the controller to determine which link to prioritize for backup and which main ring segment to switch first. Finally, it can be used to optimize the dynamic parameters of the objective function.

[0070] For example, the calculation formula for the risk assessment model can be shown in formula (4):

[0071] (4)

[0072] Where R is the network failure risk value, Indicates the health of the node. Indicates the health of the link. This represents the weight of a node's health status in relation to the overall network security, i.e., the risk coefficient. This represents the risk coefficient corresponding to the link's health. Optionally, the risk coefficient is generally different for each communication node or communication link. In practical applications, a tiered strategy can be adopted, for example, high-risk nodes / links: =0.03 =0.05; Normal node / link: =0.01 =0.01; Low-risk access point: =0.001 =0.001.

[0073] Alternatively, the risk coefficient can also be obtained from historical failure records. For example, network failure records from the past 12 months can be obtained, and the number of major failures each communication node n participated in can be counted and the node failure rate can be calculated. Similarly, the number of major failures each communication link l participated in can be counted and the link failure rate can be calculated. Then, based on the node failure rate and the link failure rate, the corresponding risk coefficient can be determined.

[0074] For example, the risk coefficient calculation formula for a communication node can be shown in formula (5), and the risk coefficient calculation formula for a communication link can be shown in formula (6).

[0075] (5)

[0076] (6)

[0077] in, A risk coefficient representing the health of a node; This represents the risk coefficient corresponding to the link health status; This represents the number of major failures that communication node n participated in within a preset historical time period; This indicates the number of major failures that occurred in communication link l within a preset historical time period. Nodes represents the number of nodes, and Links represents the number of links.

[0078] In one possible implementation, after the digital twin platform receives the node health of each communication node and the link health of each communication link pushed by the front-end sensing device, it substitutes them into formula (4) to calculate the network fault risk value of the entire network.

[0079] Step 102C: When the network fault risk value exceeds the preset risk threshold, the node health, link health and predicted alarm events are mapped to the nodes and time slot entities in the T-RMR topology.

[0080] If the network failure risk value exceeds a preset risk threshold, the subsequent step of updating the time slot mapping relationship is triggered. In one possible implementation, the digital twin platform can subscribe to the Kafka topic of the front-end sensing device to receive node health, link health, and predicted alarm events. After this data enters the twin platform, it updates the status, risk value, color identifier, and other attributes of the corresponding node / link in the T-RMR topology to map the access data to the corresponding node and time slot entity in the T-RMR topology.

[0081] The key to the mapping process is the binding between the "source address + interface information" of the front-end data and the "topology model + time slot distribution" of entities in the T-RMR topology. This can be divided into two steps: Step 1: Node / link binding. The digital twin platform maintains a mapping table, mainly containing the source device, source port, corresponding node, and corresponding link. The digital twin platform looks up and matches the binding based on the source IP or device identifier + port field in the data packet. Step 2: Mapping to the TS-MAP association mechanism, i.e., determining which time slots a certain link / service in the data is currently transmitting on, and which logical rings are affected when a problem occurs. In other words, the mapped T-RMR topology can indicate the spatial topology location of the current alarm event, the affected services, and predict the failure time.

[0082] Step 102D: Based on the mapped T-RMR topology, generate the latest time slot mapping relationship between the main ring and the redundant ring for each time slot.

[0083] Furthermore, the digital twin platform can generate the latest time slot mapping relationship between the main ring and the redundant ring based on the mapped T-RMR topology. For example, if the health index of the main ring is lower than the threshold, the time slot of its corresponding business flow will be switched to the backup time slot of the redundant ring.

[0084] Optionally, different implementation methods for the update time slot mapping relationship can be selected based on different predicted failure times. The predicted failure time quantifies the urgency of the alarm by predicting the remaining time before a node or link fails.

[0085] In one possible implementation, if the predicted failure time is less than a preset value, the digital twin platform can quickly generate a temporary time slot mapping table based on a rule engine. Its response time is less than 5ms, ensuring basic survivability. If the predicted failure time is greater than the preset value, the digital twin platform can also calculate the globally optimal solution based on the objective function and constraints to output the latest time slot mapping table. This calculation takes 50-100ms, maximizing resource utilization. For example, the preset value could be 10ms.

[0086] Specifically, generating the latest time slot mapping relationship between the main ring and the redundant ring based on the mapped T-RMR topology may also include steps 102D1 to 102D2.

[0087] Step 102D1: Based on the fault prediction frequency and link utilization, construct a time slot allocation model. The time slot allocation model consists of an objective function and constraints.

[0088] To minimize the total cost, the network load risk and handover cost are reduced by adjusting the time slot allocation of each communication link. Therefore, the objective function of the optimal time slot allocation problem is defined as shown in formula (7):

[0089] (7)

[0090] Where E represents the set of all communication links in the network, each element is a link pair, and (i, j) represents the communication link between node i and node j. Let (i, j) be the cost weight of the communication link. The cost weight represents a comprehensive value considering factors such as the link's physical distance, transmission delay, link congestion level, and fiber optic aging. A larger value indicates a higher cost for allocating resources on that link. Allocate time slots for communication link (i, j).

[0091] The constraints of the objective function include:

[0092] , (8)

[0093] (9)

[0094] (10)

[0095] in, This represents the total number of time slots allocated for different services k on the communication link (i, j); Let (i, j) be the maximum number of time slots in the communication link. Let k be the minimum time slot requirement for service k on the communication link (i, j); The minimum number of redundant time slots reserved for each communication link. Constraints , This indicates the maximum number of time slots that can be allocated to each link, ensuring that the physical link capacity is not exceeded; constraints. This means that the number of time slots allocated to service k on the link must meet the minimum time slot requirement of that service on that link, that is, the time slot allocation must match the service demand to guarantee the basic bandwidth of each service; constraint conditions. To ensure redundant time slots and reserve necessary protection resources for self-healing.

[0096] In one possible implementation, when constructing the objective function, the cost weights in the objective function can be dynamically determined based on the fault prediction probability and link utilization. The fault prediction frequency can be represented by the number of AI alerts at the same location in the past 5 minutes divided by 5; specific data can be obtained through Kafka alarm stream windows. Link utilization can be calculated by dividing the port's used bandwidth by the physical bandwidth.

[0097] Specifically, the formula for determining the cost weight is:

[0098] (11)

[0099] As cost weight, For fault prediction frequency, For link utilization, The weighting coefficients for fault prediction frequency, This is a weighting coefficient for link utilization. Based on the cost weight. The default value is 0.5. The default value is 0.3.

[0100] 102D2, based on the mapped T-RMR topology and time slot allocation model, generates the latest time slot mapping relationship between the main ring and the redundant ring for each time slot.

[0101] After constructing the time slot allocation model, the mapped T-RMR topology and time slot allocation model can be used to generate the latest time slot mapping relationship between the main ring and the redundant ring. Specifically, for small-scale networks, an exact algorithm, branch and bounds, can be used with the CPLEX / Gurobi solver to output the latest time slot mapping relationship; for large-scale networks, a heuristic algorithm, tabu search, with custom iterative optimization, can be used to output the latest time slot mapping relationship.

[0102] Please refer to Figure 3This is an overall flowchart of a digital twin platform execution method provided in an exemplary embodiment of this application. It includes: Step 301, constructing a digital twin model (i.e., T-RMR topology); Step 302, data and alarm access; Step 303, generating an SDN / NFV linkage self-healing strategy; specifically, the SDN controller uses the PCE or a custom path calculation module to read the alarm context in the digital twin (the alarm context refers to: spatial topology location, affected services, and predicted failure time), and generates new time slot mappings and flow table rules; the NFV orchestrator issues a self-healing service chain, dynamically starts or migrates bypass VNFs, ensuring that the logical loop is not interrupted. Step 304, dynamically calculating and issuing the TS-MAP. Step 305, bypass routing and time slot backup. Specifically, the controller simultaneously generates a bypass routing strategy, and after issuance, triggers the virtual switching plane device to immediately load the bypass flow table. The triggering condition is that when a primary path failure or high risk is detected, the SDN controller executes: pre-calculating the backup path: the backup path is based on the T-RMR topology, selecting a low-risk path. And low Communication links; reserved resources: pre-allocate redundant time slots on the backup path to satisfy formula (10); generation rules: the main path flow table forwards normally, and when the bypass flow table matches the fault label, it is redirected to the backup path. Optionally, the virtual switching plane deployment can be deployed on the edge server.

[0103] The digital twin platform outputs a TS-MAP containing primary and backup path mappings through the model, then converts it into a standard JSON structure through the SDN controller, and sends it through the distribution channel to the edge devices to receive the updated policies.

[0104] Step 103: When a communication link failure or node failure is detected, the network node device performs time slot switching based on the latest time slot mapping relationship.

[0105] In one possible implementation, when a network node device receives the latest timeslot mapping (TS-MAP) forwarded by the SDN controller, it immediately performs timeslot switching based on the latest timeslot mapping when it detects a communication link failure or node failure locally.

[0106] Specifically, such as Figure 4 As shown, the process of network node devices performing time slot switching and verifying whether time slot services have been restored may also include steps 401 to 403.

[0107] Step 401, Policy Analysis and Local Deployment.

[0108] Specifically, the edge device (the edge device is the network access node, responsible for performing time slot mapping and traffic switching) receives the TS-MAP and flow table rules, writes them into the TCAM using the pipeline processor, and updates the local time slot switching table. The local time slot switching table is updated according to the TS-MAP issued by the SDN controller and the globally optimal time slot allocation calculated by the digital twin platform. The update method is as follows: first, the latest time slot mapping relationship is parsed, then the pipeline processor directly writes it into the TCAM (tri-state content addressing memory), updates the local switching table index, and finally marks the time slot resources, for example: main ring time slot (state = activa) and redundant ring time slot (state = standby). The backup ring time slot mapping is preloaded. The backup ring time slot mapping is obtained from the pre-calculation of the digital twin platform: the global TS-MAP generated by 102D2 contains the main / backup path, and the redundant resources are forcibly reserved through formula (10) to obtain the backup ring pre-configured by the bypass routing strategy (step 305).

[0109] Step 402: Quickly switch time slot levels.

[0110] Specifically, when a physical link or node failure is detected locally, the affected time slots are immediately switched to the healthy ring according to TS-MAP. This process is completed within a 1.25µs granularity. At the same time, the time slot mapping of the faulty link is automatically cleared to avoid packet loss. The specific operation is as follows: only the mapping entry of the faulty link is deleted, the corresponding flow table entry in TCAM is cleared synchronously, and the topology information is retained for root cause analysis.

[0111] Optionally, a three-layer detection fusion mechanism can be used to detect physical link or node faults: the physical layer uses optical power monitoring to determine whether the received optical power is less than -28dBm; the data layer uses BFD fast detection to determine whether the packet loss rate within 3ms is greater than 40%; and the service layer uses time slot CRC check detection to determine whether there are more than or equal to 5 consecutive erroneous frames.

[0112] When switching, you can also consider switching path priorities. For example, switching path priorities... The formula for determining can be shown in formula (12):

[0113] (12)

[0114] in, Link distance, The number of time slots already allocated. This refers to the node's health status. The real-time health status of a node is calculated as follows: =1 - (Current number of alarms / Maximum tolerable number of alarms). Link distance can be measured using instruments such as an OTDR, and the number of allocated time slots can be obtained from a local counter.

[0115] Backup paths are prioritized according to switching path priority, with high-quality paths selected first. Edge devices load the new TS-MAP, and a virtual switching plane is triggered to complete seamless traffic migration. Specifically, the alternative path selection mechanism can be as follows: obtain available alternative paths and determine the health of nodes on the available alternative paths. If the value is greater than 0.8, it is determined as a candidate path; or, if the link distance of the available alternative path is less than 100km, it is determined as a candidate path; or, if the idle time slot of the available alternative path is greater than the business requirement, it is determined as a candidate path; then, the link distance, the number of allocated time slots and the node health of each candidate path are substituted into formula (12) to calculate the switching path priority of each candidate path, and the candidate path with the highest switching path priority is determined as the high-quality path for priority switching.

[0116] Step 403: Confirm the switch based on heartbeat detection.

[0117] Specifically, the edge device enables the heartbeat channel Bypass TAP mode on the control plane and sends heartbeat packets to the bypass ring at intervals of 1ms to 5ms. After receiving the return, it confirms that the switchover was successful. If the switchover is not restored after N consecutive heartbeats (N=3), it reports the failure to the control layer and triggers a secondary protection strategy (such as a backup satellite link).

[0118] Alternatively, please refer to Figure 5 The rapid self-healing method based on AI prediction and T-RMR provided in this application may also include steps 501 to 503.

[0119] Step 501: After the fault recovery is completed, obtain the node event logs of each network node device through the digital twin platform;

[0120] Specifically, after the fault recovery is completed, the digital twin platform will automatically pull the node event logs of each network node device (such as fault timestamps, time slot mapping before and after the switchover, heartbeat results, etc.), integrate them into training samples, and label the recovery latency, packet loss rate and business impact, so that the adjustment strategy of time slot mapping relationship can be optimized based on the training samples and labeled data.

[0121] Step 502: Determine the current network topology, current risk assessment results, and current time slot usage status based on node event logs using the digital twin platform;

[0122] Specifically, the replay data is input into the RL environment, and the state (current network topology, risk assessment results, and time slot usage status), action (TS-MAP adjustment strategy), and reward (reciprocal switching latency and reduced packet loss rate) are defined. The Deep Q-Learning or Policy Gradient algorithm is used for iterative training to update the parameters of the TS MAP generation module and the local AI model.

[0123] The recovery replay data, generated in step 501 by the digital twin platform through end-to-end log integration, is multi-dimensional training data that mainly includes fault context, handover process, performance metrics, and resource status. This data is integrated into training samples for the reinforcement learning training process. It's called "replay" because reinforcement learning typically uses an "experience replay" mechanism to store historical experiences (i.e., data from these recovery processes) and then sample from them for training, breaking down correlations between data and improving learning efficiency.

[0124] The current network topology refers to the network's topology when a fault occurs, including the connection relationships between nodes and links, as well as the status of each node and link. The current risk assessment result is calculated using the risk assessment model in the above embodiment; the time slot usage status refers to the time slot allocation of each link in the network, which can reflect the real-time distribution of time slot resources across the entire network, including allocated time slots, real-time utilization, and redundancy status.

[0125] Step 503: Update the current adjustment strategy through the digital twin platform based on the current network topology, current risk assessment results, current time slot usage status, current adjustment strategy of time slot mapping relationship and recovery performance reward.

[0126] During reinforcement learning, the system learns a strategy to select the optimal action in a given state (updating the current adjustment strategy) based on the state (composed of the three parts mentioned above: topology, risk assessment, and time slot state), the action (the current adjustment strategy of TS-MAP), and the reward (a reward function set according to recovery performance). Through repeated training, the system learns this strategy (updating the current adjustment strategy) to achieve better recovery in subsequent faults.

[0127] Specifically, once the new adjustment strategy and AI model weights have been verified to improve performance, they are automatically distributed to each layer through the CI / CD pipeline to complete closed-loop self-learning, enabling the system to have better recovery performance in subsequent failures.

[0128] In summary, this application proposes a rapid self-healing method and system for multi-ring networks based on AI prediction and time-redundant technology. AI prediction and hierarchical alarms are implemented: lightweight neural network models are deployed at each node to continuously monitor and predict key indicators such as link latency, optical power, and jitter; alarms are automatically graded (normal / warning / fault) based on the prediction results, achieving a closed loop of "0 → warning → re-fault". Furthermore, outside the main ring, multiple logically redundant rings are divided according to time slots, without adding physical optical cables, achieving multiple backups only through time slot multiplexing and time slot switching; dynamic reconfiguration using a time slot mapping table (TS-MAP) allows switching to be completed within a sub-1.25μs time slot granularity when a single link or node fails. In addition, a digital twin platform for the communication network is established to aggregate network status and AI prediction data in real time. The SDN controller runs online optimization algorithms based on the twin, and NFV nodes deploy self-healing service chains, automatically issuing commands such as rerouting and time slot table updates. During the recovery process, bypass traffic can be seamlessly migrated in the virtualized switching plane. Fault warning is achieved through AI models, fine-grained physical layer protection is provided by multiple time-slot-level logic loops, and seamless switching with low latency is achieved by combining digital twins and SDN / NFV control.

[0129] Please refer to Figure 6 This is a schematic diagram of the structure of a power communication network self-healing system provided in an embodiment of this application. For example, as shown... Figure 6 As shown, the system 600 includes at least a front-end sensing device 601, a digital twin platform 602, an SDN controller 603, and a network node device 604.

[0130] The front-end sensing device 601 is used to collect real-time indicator data and perform AI prediction based on the real-time indicator data to obtain the node health of each communication node, the link health of each communication link, and the predicted alarm events at the next moment.

[0131] The digital twin platform 602 is used to acquire network topology information, device capabilities, node health, link health and predicted alarm events, generate T-RMR topology and the latest time slot mapping relationship between the main ring and the redundant ring, and distribute the latest time slot mapping relationship to each network node device through the SDN controller.

[0132] The network node device 604 is used to perform time slot switching based on the latest time slot mapping relationship when a communication link failure or node failure is detected.

[0133] Optionally, the front-end sensing device 601 is further used for

[0134] The real-time indicator data is input into the short-term prediction model to obtain the predicted indicator data for the next time step output by the short-term prediction model.

[0135] Based on the predicted index data and index calculation formula, the node health of each communication node and the link health of each communication link are determined at the next moment.

[0136] The predicted alarm event is determined based on the node health, the link health, and the preset alarm threshold;

[0137] The formula for calculating the indicator is as follows:

[0138]

[0139] The health status of the node. For CPU utilization, For memory usage, For bit error rate, The health of the link, For time delay, The jitter value. For optical power, , , , , , As weight.

[0140] Optionally, the digital twin platform 602 is further used for:

[0141] Based on the network-wide topology information and the device capabilities, the T-RMR topology is constructed;

[0142] Based on the node health and link health, determine the network failure risk value of the entire network;

[0143] When the network failure risk value is higher than a preset risk threshold, the node health, the link health, and the predicted alarm event are mapped to the nodes and time slot entities in the T-RMR topology.

[0144] Based on the mapped T-RMR topology, the latest time slot mapping relationship between the main ring and the redundant ring is generated.

[0145] Optionally, the digital twin platform 602 is further used for:

[0146] Based on the fault prediction frequency and link utilization, a time slot allocation model is constructed, which consists of an objective function and constraints.

[0147] Based on the mapped T-RMR topology and the time slot allocation model, the latest time slot mapping relationship between the main ring and the redundant ring is generated for each time slot.

[0148] The objective function is:

[0149]

[0150] E represents the set of all communication links in the network, where each element is a link pair, and (i, j) represents the communication link between node i and node j. Let (i, j) be the cost weight of the communication link. Allocate time slots for communication link (i, j);

[0151] The constraints include:

[0152] ,

[0153]

[0154]

[0155] This represents the total number of time slots allocated for different services k on the communication link (i, j); Let (i, j) be the maximum number of time slots in the communication link. Let k be the minimum time slot requirement for service k on the communication link (i, j); The minimum number of redundant time slots reserved for each communication link.

[0156] Optionally, the digital twin platform 602 is further used for:

[0157] Based on the fault prediction frequency and the link utilization rate, the cost weight in the objective function is determined;

[0158] The formula for determining the cost weight is as follows:

[0159]

[0160] As cost weight, For fault prediction frequency, For link utilization, The weighting coefficients for fault prediction frequency, This is a weighting coefficient for link utilization. Based on the cost weight.

[0161] Optionally, the digital twin platform 602 is further used for:

[0162] After the fault recovery is completed, obtain the node event logs of each of the network node devices;

[0163] Based on the node event logs, the current network topology, current risk assessment results, and current time slot usage status are determined;

[0164] Based on the current network topology, the current risk assessment results, the current time slot usage status, the current adjustment strategy of the time slot mapping relationship, and the recovery performance reward, update the current adjustment strategy.

[0165] An exemplary embodiment of this application also provides a computer device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, the computer program, when executed by the at least one processor, causing the computer device to perform a rapid self-healing method based on AI prediction and T-RMR according to an embodiment of this application.

[0166] An exemplary embodiment of this application also provides a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a computer's processor, is used to cause the computer to perform a rapid self-healing method based on AI prediction and T-RMR according to an embodiment of this application.

[0167] An exemplary embodiment of this application also provides a computer program product, including a computer program, wherein, when executed by a computer's processor, the computer program is used to cause the computer to perform a rapid self-healing method based on AI prediction and T-RMR according to an embodiment of this application.

[0168] refer to Figure 7 The following is a structural block diagram of a computer device 700 that can serve as a server or client of this application, which is an example of a hardware device that can be applied to various aspects of this application. The term "computer device" is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. A computer device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the application described and / or claimed herein.

[0169] like Figure 7As shown, the computer device 700 includes a computing unit 701, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. The RAM 703 may also store various programs and data required for the operation of the computer device 700. The computing unit 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0170] Multiple components in the computer device 700 are connected to the I / O interface 705, including: an input unit 706, an output unit 707, a storage unit 708, and a communication unit 709. The input unit 706 can be any type of device capable of inputting information into the computer device 700. The input unit 706 can receive input numerical or character information and generate key signal inputs related to user settings and / or function control of the computer device. The output unit 707 can be any type of device capable of presenting information and may include, but is not limited to, a monitor, speaker, video / audio output terminal, vibrator, and / or printer. The storage unit 708 may include, but is not limited to, a hard disk and an optical disk. The communication unit 709 allows the computer device 700 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network interface cards, infrared communication devices, wireless communication transceivers, and / or chipsets, such as Bluetooth devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0171] Optionally, the computer device 700 also includes a single-channel EEG signal acquisition module (not shown in the figure). This module is used to acquire EEG signals and transmit them to the signal processor of the computer device 700 for EEG signal processing.

[0172] The computing unit 701 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 performs the various methods and processes described above. For example, in some embodiments, Figure 1 , Figure 3 , Figure 4 , Figure 5The method shown can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed on computer device 700 via ROM 702 and / or communication unit 709. In some embodiments, computing unit 701 can be configured to execute by any other suitable means (e.g., by means of firmware). Figure 1 , Figure 3 , Figure 4 , Figure 5 The method shown.

[0173] The program code used to implement the methods of this application may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0174] In the context of this application, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0175] As used in this application, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0176] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0177] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0178] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other.

Claims

1. A rapid self-healing method based on AI prediction and T-RMR, characterized in that, The method is applied to a power communication network self-healing system, which includes at least front-end sensing devices, a digital twin platform, an SDN controller, and network node devices. The method includes: Real-time indicator data is collected by the front-end sensing device, and AI prediction is performed based on the real-time indicator data to obtain the node health of each communication node, the link health of each communication link, and the predicted alarm events at the next moment. The digital twin platform acquires network topology information, device capabilities, node health, link health, and predicted alarm events. It then generates a T-RMR topology and the latest time slot mapping relationship between the main ring and redundant rings. This latest time slot mapping relationship is distributed to each network node device via the SDN controller. The T-RMR topology is a time-redundant multi-ring network, representing the topological relationship between physical entities and logical time slots. Multiple independent logical rings, including one main ring and multiple redundant rings, are constructed on the same physical ring network through time slot segmentation. This is achieved by using the network topology information and device capabilities. The device capabilities include hardware capabilities, the number / rate of ports, cache size, supported time slot segmentation granularity, protocol support, and virtualization capabilities. When a communication link failure or node failure is detected, the network node device performs time slot switching based on the latest time slot mapping relationship.

2. The method according to claim 1, characterized in that, The AI ​​prediction based on the real-time indicator data, to obtain the node health of each communication node, the link health of each communication link, and the predicted alarm events for the next moment, includes: The real-time indicator data is input into the short-term prediction model to obtain the predicted indicator data for the next time step output by the short-term prediction model. Based on the predicted index data and index calculation formula, the node health of each communication node and the link health of each communication link are determined at the next moment. The predicted alarm event is determined based on the node health, the link health, and the preset alarm threshold; The formula for calculating the indicator is as follows: The health status of the node. For CPU utilization, For memory usage, For bit error rate, The health of the link, For time delay, The jitter value. For optical power, , , , , , For weights.

3. The method according to claim 1, characterized in that, The generation of the T-RMR topology and the latest time slot mapping relationship between the main ring and the redundant ring includes: Based on the network-wide topology information and the device capabilities, the T-RMR topology is constructed; Based on the node health and link health, determine the network failure risk value of the entire network; When the network failure risk value is higher than a preset risk threshold, the node health, the link health, and the predicted alarm event are mapped to the nodes and time slot entities in the T-RMR topology. Based on the mapped T-RMR topology, the latest time slot mapping relationship between the main ring and the redundant ring is generated.

4. The method according to claim 3, characterized in that, The process of generating the latest time slot mapping relationship between the main ring and the redundant ring based on the mapped T-RMR topology includes: Based on the fault prediction frequency and link utilization, a time slot allocation model is constructed, which consists of an objective function and constraints. Based on the mapped T-RMR topology and the time slot allocation model, the latest time slot mapping relationship between the main ring and the redundant ring is generated for each time slot. The objective function is: E represents the set of all communication links in the network, where each element is a link pair, and (i, j) represents the communication link between node i and node j. Let (i, j) be the cost weight of the communication link. Allocate time slots for communication link (i, j); The constraints include: ， This represents the total number of time slots allocated for different services k on the communication link (i, j); Let (i, j) be the maximum number of time slots in the communication link. Let k be the minimum time slot requirement for service k on the communication link (i, j); The minimum number of redundant time slots reserved for each communication link.

5. The method according to claim 4, characterized in that, The time slot allocation model, based on fault prediction frequency and link utilization, includes: Based on the fault prediction frequency and the link utilization rate, the cost weight in the objective function is determined; The formula for determining the cost weight is as follows: As cost weight, For fault prediction frequency, For link utilization, The weighting coefficients for fault prediction frequency. This is a weighting coefficient for link utilization. Based on the cost weight.

6. The method according to claim 1, characterized in that, The method further includes: After the fault recovery is completed, the node event logs of each of the network node devices are obtained through the digital twin platform; The digital twin platform determines the current network topology, current risk assessment results, and current time slot usage status based on the node event logs. The digital twin platform updates the current adjustment strategy based on the current network topology, the current risk assessment results, the current time slot usage status, the current adjustment strategy of time slot mapping relationship, and the recovery performance reward.

7. A self-healing system for power communication networks, characterized in that, The power communication network self-healing system includes at least front-end sensing devices, a digital twin platform, an SDN controller, and network node devices: The front-end sensing device is used to collect real-time indicator data and perform AI prediction based on the real-time indicator data to obtain the node health of each communication node, the link health of each communication link, and the predicted alarm events at the next moment. The digital twin platform is used to acquire network topology information, device capabilities, node health, link health, and predicted alarm events; generate the T-RMR topology and the latest time slot mapping relationship between the main ring and the redundant ring; and distribute the latest time slot mapping relationship to each network node device through the SDN controller. The T-RMR topology is a time-redundant multi-ring network, representing the topological relationship between physical entities and logical time slots. On the same physical ring network, multiple independent logical rings are constructed through time slot division, including one main ring and multiple redundant rings, which are constructed from the network topology information and device capabilities. The device capabilities include hardware capabilities, number / rate of ports, cache size, supported time slot division granularity, protocol support, and virtualization capabilities. The network node device is used to perform time slot switching based on the latest time slot mapping relationship when a communication link failure or node failure is detected.

8. The system according to claim 7, characterized in that, The front-end sensing device is also used for The real-time indicator data is input into the short-term prediction model to obtain the predicted indicator data for the next time step output by the short-term prediction model. Based on the predicted index data and index calculation formula, the node health of each communication node and the link health of each communication link are determined at the next moment. The predicted alarm event is determined based on the node health, the link health, and the preset alarm threshold; The formula for calculating the indicator is as follows: The health status of the node. For CPU utilization, For memory usage, For bit error rate, The health of the link, For time delay, The jitter value. For optical power, , , , , , For weights.

9. A computer device, comprising: processor; as well as Stored program memory, The program includes instructions that, when executed by the processor, cause the processor to perform the rapid self-healing method based on AI prediction and T-RMR according to any one of claims 1-6.

10. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to execute the rapid self-healing method based on AI prediction and T-RMR according to any one of claims 1-6.

Images