Intelligent operation and maintenance system for passive optical network based on dynamic topology modeling

Abstract

This invention discloses an intelligent operation and maintenance system for passive optical networks (PONs) based on dynamic topology modeling. The invention relates to the field of communication network management and industrial control system technology. The dynamic topology modeling module updates the adjacency matrix of the network in real time, constructing a dynamic topology model reflecting the current physical connection relationships. Elements in the adjacency matrix represent the physical connection state between nodes at any given time. When the state of the optical splitter port changes, the adjacency matrix is ​​instantly reconstructed. An intelligent diagnostic analysis module is used to input the real-time updated adjacency matrix and the collected operational data feature matrix into a pre-trained deep learning model. The key advantage of this invention is that by updating the adjacency matrix in real time according to the optical splitter port state, the diagnostic model always infers based on real physical connections, eliminating fault location errors caused by ODN changes and achieving self-awareness and accurate fault location.

Description

Technical Field

[0001] This invention relates to the field of communication network management and industrial control system technology, specifically to a passive optical network intelligent operation and maintenance system based on dynamic topology modeling. Background Technology

[0002] Passive optical networks (PONs) are the mainstream technology for broadband access networks and are widely used in home broadband, enterprise leased lines and industrial internet scenarios. A typical PON network consists of an optical line terminal (OLT), an optical distribution network (ODN), and optical network units (ONUs), forming a point-to-multipoint (P2MP) topology. With the explosive growth of network scale and the deepening of "fiber to the home" and "copper to the home", the operation and maintenance of PON networks face huge challenges.

[0003] In existing technologies, the operation and maintenance management of PON networks mainly relies on the network management system (EMS) or software-defined networking (SDN) controller. For example, Chinese patent document CN111865419B discloses a "Smart Optical Access Network Central Office Cloud System Based on Modular Architecture for 5G". This solution introduces OpenFlow virtual switches and Network Function Virtualization (NFV) technology to pool resources such as OLT and BBU, achieving separation of the control plane and forwarding plane, greatly improving the flexibility of resource configuration and the openness of the system. However, this existing technology still has significant technical defects in actual operation and maintenance scenarios: First, lagging topology awareness leads to fault location drift. Existing systems (such as CN111865419B) typically rely on pre-configured static databases or periodic SNMP (Simple Network Management Protocol) polling to obtain network topology. In actual engineering projects, ODN networks often contain a large number of passive optical splitters. Changes in their physical connections (such as fiber optic cable splicing or splitter port replacement) cannot be reported to the control layer in real time. When the physical topology changes but the logical topology is not updated synchronously, once a fault occurs, the diagnostic algorithm based on the old topology will locate the fault point to the wrong node or link (i.e., fault location drift), causing maintenance personnel to "go to the wrong station," which seriously affects the efficiency of emergency repair.

[0004] Second, there is a lack of a tiered self-healing mechanism. Existing fault handling strategies are usually binary: either automatically restart the device or report it for manual handling. For example, Chinese patent document CN103281117B discloses a "Method and Device for Automatic Detection and Adaptation of ONU to EPON and GPON". Although it solves the problem of mismatched ONU access mode parameters, this is only an atomic solution for a single specific fault (parameter configuration error). Existing systems lack a progressive (multi-level) self-healing strategy that integrates parameter adaptation, device reset, link protection switching and other means. When the network is in a sub-healthy state without hard breaks (such as bit error rate fluctuations), a single restart strategy may be ineffective or even aggravate service interruption. The system lacks an intelligent judgment mechanism based on ONU online time threshold verification to trigger different levels of response strategies.

[0005] In summary, existing technologies urgently need an intelligent operation and maintenance system that can perceive physical layer topology changes in real time and perform deep learning-based fault diagnosis and multi-level self-healing based on dynamic topology. Summary of the Invention

[0006] To address the problems of fault location drift and single-level self-healing strategy failure caused by network topology changes in existing technologies, this invention provides an intelligent operation and maintenance system for passive optical networks based on dynamic topology modeling.

[0007] The technical solution adopted in this invention is as follows: A passive optical network intelligent operation and maintenance system based on dynamic topology modeling includes: a data acquisition module, a dynamic topology modeling module, an intelligent diagnosis and analysis module, and a multi-level self-healing execution module.

[0008] The data acquisition module is used to collect real-time operating data of the passive optical network, including optical splitter port status information, ONU online time information, optical link status (LOS) alarms, bit error rate and optical power data. The dynamic topology modeling module is used to update the network's adjacency matrix in real time based on the optical splitter port status information. Construct a dynamic topology model that reflects the current physical connectivity relationships; The adjacency matrix elements in Represents a node With nodes At any moment The physical connection status is determined, and when the status of the optical splitter port changes, the adjacency matrix is ​​reconstructed immediately. The intelligent diagnostic analysis module is used to update the adjacency matrix in real time. With the feature matrix of the collected running data Input a pre-trained deep learning model, which adopts a graph convolutional neural network (GCN) architecture, aggregates the features of neighboring nodes through graph convolution operations, and outputs the probability distribution and fault type of faulty nodes; The multi-level self-healing execution module is used to receive fault diagnosis results and trigger a multi-level self-healing strategy based on the ONU online time threshold verification mechanism. (1) Obtain the current online duration of the faulty ONU. ; (2) If Less than the preset jitter threshold This is determined to be network jitter, and a silent observation is initiated. (3) If Greater than Then, the following multi-level self-healing process will be executed sequentially: Level 1 self-healing: Call the parameter adaptation subroutine to detect and reset the ONU's access mode parameters and frame delimitation state machine parameters; Second-level self-healing: If the first level fails, a reset command is sent to the faulty ONU through the ONU management and control interface to restart it; Level 3 self-healing: If Level 2 fails and the network supports protection switching, trigger Type B or Type C link protection switching; Level 4 self-healing: If all the above levels fail, a dispatch instruction is generated and sent to the maintenance personnel's terminal.

[0009] Furthermore, the deep learning model in the intelligent diagnostic analysis module utilizes dual convolutional channels to process the trend component, seasonal component, and residual component of the data respectively, and incorporates the dynamic adjacency matrix. Introducing a graph convolutional layer, its inter-layer propagation formula is: ; in, To introduce a dynamic adjacency matrix with self-loops, For the first The feature input of the layer, This is a learnable weight matrix.

[0010] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows: 1. Solved the topology drift problem: By updating the adjacency matrix in real time according to the port status of the optical splitter, it is ensured that the diagnostic model is always based on the real physical connection for reasoning, eliminating the fault location error caused by ODN changes, and realizing the self-sensing and accurate location of faults.

[0011] 2. Achieved refined operation and maintenance: Introduced ONU online time threshold verification and multi-level self-healing mechanism to avoid overreaction to instantaneous jitter. At the same time, it organically combined parameter adaptation, hard reset and protection switching to form a closed-loop self-healing system, which greatly reduced the invalid order dispatch rate and manual operation and maintenance costs.

[0012] 3. Improved diagnostic accuracy: Combining the graph convolution concept in existing technologies, it is innovatively applied to dynamically changing PON topologies. Spatial correlation (neighbor node status) is used to assist in judging the fault nature of isolated nodes, significantly improving the diagnostic accuracy in complex fault scenarios. Attached Figure Description

[0013] Figure 1 This is an overall architecture diagram of the intelligent operation and maintenance system for passive optical networks based on dynamic topology modeling provided in this embodiment of the invention; Figure 2 This is a schematic diagram of the dynamic topology modeling and adjacency matrix update process in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the principle of the intelligent diagnostic algorithm based on GCN in this embodiment of the invention; Figure 4 This is a flowchart of the multi-level self-healing triggering mechanism and execution logic in an embodiment of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0015] Example 1: System Overall Architecture and Dynamic Topology Modeling like Figure 1 As shown, this embodiment provides an intelligent operation and maintenance system for passive optical networks based on dynamic topology modeling. The system runs on an industrial-grade server or cloud virtualization environment (refer to the modular architecture of CN 111865419 B), and is logically divided into a data acquisition layer, a core processing layer, and an execution control layer.

[0016] 1. Data Acquisition Layer: This layer is responsible for obtaining raw data from physical devices, including: OLT device status: Collect the transmit optical power, receive optical power, bit error rate (BER), and round-trip time (RTT) of the PON port via SNMP or Telemetry protocol.

[0017] Intelligent ODN Status: The occupancy status (Port_Status) of the splitter port and the serial number (SN) of the connected ONU can be sensed in real time through the electronic tag contact or fiber optic sensing unit of the intelligent optical splitter (eODN).

[0018] ONU Status: Collects the ONU's transmit and receive power, voltage, current, temperature, and online time (System_Up_Time).

[0019] 2. Core Processing Layer (Dynamic Topology Modeling Module): The core innovation of this invention lies in abandoning the static configuration table and adopting dynamic adjacency matrix modeling.

[0020] Define the topology of the PON network as follows: ,in It is a set of nodes (including OLT PON ports, splitter ports, and ONUs). for The set of edges at time points.

[0021] Adjacency matrix construction method: The system maintains a Adjacency Matrix ,in This represents the total number of network nodes.

[0022] Initialization: Generate a baseline matrix based on initial planning data. .

[0023] Real-time update mechanism: When the data acquisition layer detects the first... The port status changed from "idle" to "occupied", and the OLT side resolved to the newly connected ONU (SN number is...). When this happens, the system automatically determines that the ONU is connected to the splitter's first branch. port.

[0024] Matrix update logic: Let node For splitter ports ,node For ONU .

[0025] = 1 (indicating the existence of a physical connection); = 1 (symmetric in undirected graphs); If a port disconnection is detected or the ONU comes online on another PON port, then set... =0.

[0026] This process solves the "topology drift" problem mentioned in the background technology. For example, when maintenance personnel move the ONU from port 1 of splitter A to port 2 of splitter B, the traditional system still considers the ONU to be at A, making subsequent troubleshooting of splitter A ineffective. This system, however, will update the ONU in real time. Set to 0, Set to 1 to ensure that the logical topology is synchronized with the physical world within milliseconds.

[0027] 3. Core Processing Layer (Intelligent Diagnostic Analysis Module): This module introduces deep learning algorithms, specifically Graph Convolutional Neural Networks (GCN).

[0028] Input features: Constructing the feature matrix ,in The feature dimensions include optical power deviation, historical average bit error rate, and LOS alarm indicators.

[0029] Spatial dependency extraction: Utilizing the publicly available graph convolution concept, but applying it to the dynamic topology of this case.

[0030] The calculation formula is: ; in This is the normalized dynamic adjacency matrix.

[0031] Innovation Analysis: Due to It changes dynamically over time. GCN can aggregate the real neighbor node information at the current moment. For example, when all ONUs connected to a certain optical splitter simultaneously experience a decrease in optical power (characteristic). (changes), and in the matrix All these ONUs are connected to the same parent node (splitter). The GCN network can output the "splitter node" as the root cause of the fault with a very high probability, rather than falsely reporting multiple independent ONU faults.

[0032] Example 2: A multi-level self-healing mechanism based on time threshold verification like Figure 4 As shown in the figure, this embodiment describes in detail the closed-loop processing flow after a fault occurs, which overcomes the shortcomings of the existing technology being simple and rigid.

[0033] Step S1: Fault Triggering and Preliminary Characterization When the intelligent diagnostic module outputs that a certain ONU or link has a fault (such as LOS or LOF), the system first reads the current status parameters of that ONU.

[0034] Step S2: Time Threshold Verification Read the ONU's System_Up_Time (online duration), and record it as... .

[0035] Set jitter threshold (e.g., 60 seconds) and stability threshold (e.g., 24 hours).

[0036] like < If the network is identified as experiencing a momentary outage or an unstable state during initialization, self-healing is not triggered; only a log is recorded. This avoids misoperation caused by minor adjustments to fiber optic connectors or during device restarts.

[0037] like > If the fault is confirmed to be substantial, a multi-level self-healing process will be initiated.

[0038] Step S3: Multi-level Self-healing Level 1: Logic Layer Self-Healing - Parameter Adaptation and Refactoring This step draws upon and improves upon the method in CN 103281117 B.

[0039] The system automatically checks the matching degree between the ONU template configured on the OLT side and the actual capabilities of the ONU.

[0040] Action 1: Detect the optical link parameters (such as FEC switch, DBA bandwidth type) sent by the OLT.

[0041] Action 2: If a mode mismatch is found (e.g., the OLT is configured in GPON mode, but the line characteristics show an EPON frame header), a configuration script will be automatically sent to modify the access mode parameters to a matching value.

[0042] Action 3: Reset the framing method and frame delimitation state machine of the Media Access Control (MAC) layer.

[0043] Effect verification: Waiting If the fault is cleared within a certain timeframe, the process ends.

[0044] Level 2: Device-level self-healing -- Hard reset If Level 1 is invalid, the fault may be due to a deadlock in the device firmware or hardware.

[0045] Action: Send a reboot command via OMCI (ONU management and control interface); Alternatively, the port can be physically powered off and restarted (Power Cycle) in milliseconds by controlling the power control unit of the smart splitter.

[0046] This strategy can resolve over 90% of electronic component failures that appear to be dead.

[0047] Level 3: Network Layer Self-Healing - Protection Switching If Level 2 is invalid, and the topology database shows that the node is in a Type B or Type C protection group.

[0048] Action: The OLT issues a switchover command, forcing the service to switch from the primary PON port to the backup PON port.

[0049] At this point, the dynamic topology modeling module will update the adjacency matrix synchronously. This maps the connection relationship to an alternative path, ensuring normal subsequent monitoring.

[0050] Level 4: Human intervention If all three automatic strategies fail, the system will automatically generate a work order, which includes: the location of the fault (precise coordinates based on the dynamic topology), possible causes (output by AI, such as "fiber optic cable break"), and recommended spare parts.

[0051] Example 3: Typical Application Scenario - Troubleshooting Caused by Optical Splitter Replacement This embodiment describes a real-world operation and maintenance scenario, demonstrating how the present invention comprehensively utilizes dynamic topology and multi-level self-healing.

[0052] Scene background: In an old residential community undergoing fiber optic upgrades, the engineers replaced the original 1:32 splitter (primary splitter) with two cascaded 1:8 splitters (secondary splitters) without notifying the network management center.

[0053] Process deduction: 1. Topology change awareness: After the replacement is completed, the ONU comes back online, and the intelligent ODN device detects the port connection change or the link layer topology changes as resolved by LLDP / OAM messages.

[0054] The dynamic topology modeling module responds immediately, updating the original star topology (1 center corresponding to 32 leaves) to a tree-like cascaded topology (1 center corresponding to 2 nodes, each node corresponding to several leaves), and the adjacency matrix... It was rewritten.

[0055] 2. Fault Occurrence: Due to construction quality issues, the optical attenuation of the secondary optical splitter was too high, causing frequent disconnections (flapping) of the 16 ONUs connected to it.

[0056] 3. Intelligent Diagnosis: GCN model reads new And low optical power alarms for 16 ONUs, due to It has been updated to a cascaded structure. The model uses convolution operations to discover that these 16 ONUs have a common parent node (the new secondary optical splitter), thus accurately outputting the diagnostic conclusion that "the secondary optical splitter port attenuation is too high".

[0057] Compared to existing technologies: Without dynamic topology, the old system still assumes that all 32 ONUs are under the same splitter, which may lead to a misjudgment of a first-level splitter failure, causing maintenance personnel to check the normal upper-level equipment.

[0058] 4. Self-healing trigger: The system detected that the ONU's online time was extremely short (less than) (and repeatedly reconnected).

[0059] Level 1 triggered: Attempt to adjust the OLT transmit optical power, ineffective.

[0060] Level 2 triggered: Attempt to restart the ONU, ineffective.

[0061] The final diagnosis was a physical link optical attenuation hard fault, which was not within the scope of self-healing. This directly triggered a Level 4 dispatch, along with an accurate description of the fault location (secondary optical splitter).

[0062] Example 4: Hardware and Software Implementation Details of the System This section supplements the specific implementation parameters of the system.

[0063] 1. Controller hardware: It adopts a high-performance server based on x86 architecture, configured with dual Intel Xeon processors, 256GB of memory, and equipped with an NVIDIA Tesla T4 accelerator card for GCN inference acceleration. This is consistent with the concept of general server resource pool in CN 111865419 B, but this embodiment emphasizes the role of AI accelerator card in dynamic matrix operations.

[0064] 2. Software Stack: Operating system: CentOS 7.9.

[0065] Control platform: Custom-developed based on ONOS (Open Network Operating System).

[0066] AI framework: PyTorch 1.8, used for loading and running pre-trained graph neural network models.

[0067] Database: Redis cluster (used to store real-time topology matrix) InfluxDB (used to store time-series performance data) ).

[0068] 3. Algorithm hyperparameters: GCN layers: set to 3 layers to capture the influence of neighboring nodes within a three-hop range.

[0069] Self-healing detection cycle: 100ms.

[0070] Adjacency matrix dimension: Supports a maximum of 10,000 nodes (covering a large OLT site).

[0071] With the above hardware and software configuration, this system can complete the closed-loop process from topology awareness to fault self-healing in milliseconds, truly realizing the "autonomous driving" of the PON network.

[0072] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A passive optical network intelligent operation and maintenance system based on dynamic topology modeling, characterized in that, include: The system includes a data acquisition module, a dynamic topology modeling module, an intelligent diagnostic analysis module, and a multi-level self-healing execution module. The data acquisition module is used to collect the operation data of the passive optical network in real time. The operation data includes optical splitter port status information, ONU online time information, optical link status alarm, bit error rate and optical power data. The dynamic topology modeling module is used to update the network's adjacency matrix in real time based on the optical splitter port status information. Construct a dynamic topology model that reflects the current physical connectivity relationships; Wherein, the adjacency matrix elements in Represents a node With nodes At any moment The physical connection status is determined, and when the status of the optical splitter port changes, the adjacency matrix is ​​reconstructed immediately. The intelligent diagnostic analysis module is used to update the adjacency matrix in real time. With the feature matrix of the collected running data Input a pre-trained deep learning model and output the probability distribution and fault type of faulty nodes; The multi-level self-healing execution module is used to receive fault diagnosis results and trigger multi-level self-healing strategies based on the ONU online time threshold verification mechanism.

2. The intelligent operation and maintenance system for passive optical networks based on dynamic topology modeling according to claim 1, characterized in that, The dynamic topology modeling module maintains a Adjacency Matrix ,in This represents the total number of network nodes. When the data acquisition module detects the first intelligent splitter When the port status changes from idle to occupied, and the OLT side resolves a newly connected ONU, the system determines that the ONU is connected to the splitter's first... Port, and perform matrix update: Let node For splitter ports ,node For newly launched ONUs, =1 and =1; If a port is detected to be disconnected, the corresponding matrix element is set to 0.

3. The intelligent operation and maintenance system for passive optical networks based on dynamic topology modeling according to claim 1, characterized in that, The deep learning model in the intelligent diagnostic analysis module adopts a graph convolutional neural network architecture, which aggregates the features of neighboring nodes through graph convolution operations. The feature matrix ,in The total number of network nodes. The feature dimension includes optical power deviation value, historical average bit error rate, and optical link status alarm identifier.

4. The intelligent operation and maintenance system for passive optical networks based on dynamic topology modeling according to claim 3, characterized in that, The inter-layer propagation formula of the graph convolutional neural network is: ； in, To introduce a dynamic adjacency matrix with self-loops, Let be the adjacency matrix at the current time. for An identity matrix of order 1; for Measure matrix; For the first Layer feature input; For the first Layer-learnable weight matrix; This is the activation function.

5. The intelligent operation and maintenance system for passive optical networks based on dynamic topology modeling according to claim 1, characterized in that, The multi-level self-healing execution module triggers a multi-level self-healing strategy based on the ONU online time threshold verification mechanism, specifically including: Get the current online duration of the faulty ONU ; like Less than the preset jitter threshold This is determined to be network jitter, and a silent observation is initiated. like Greater than the preset jitter threshold Then, a multi-level self-healing process will be executed sequentially.

6. The intelligent operation and maintenance system for passive optical networks based on dynamic topology modeling according to claim 5, characterized in that, The multi-level self-healing process includes first-level self-healing, second-level self-healing, third-level self-healing and fourth-level self-healing executed sequentially. The first level of self-healing involves calling a parameter adaptation subroutine to detect and reset the ONU's access mode parameters and frame delimitation state machine parameters.

7. The intelligent operation and maintenance system for passive optical networks based on dynamic topology modeling according to claim 6, characterized in that, The second level of self-healing is executed after the first level of self-healing fails. Specifically, it sends a reset command to the faulty ONU through the ONU management control interface to restart it.

8. The intelligent operation and maintenance system for passive optical networks based on dynamic topology modeling according to claim 6, characterized in that, The third level of self-healing is executed after the second level of self-healing fails, and is triggered when the network supports protection switching. Specifically, the OLT issues a switching instruction to switch the service from the primary PON port to the backup PON port. During the third level of self-healing, the dynamic topology modeling module synchronously updates the adjacency matrix. This maps the connection relationships to alternative paths.

9. The intelligent operation and maintenance system for passive optical networks based on dynamic topology modeling according to claim 6, characterized in that, The fourth level of self-healing is executed after the first, second and third levels of self-healing have failed. Specifically, it generates a dispatch instruction to the maintenance personnel's terminal. The dispatch instruction includes the fault location, fault cause and recommended spare parts based on the dynamic topology.

10. The intelligent operation and maintenance system for passive optical networks based on dynamic topology modeling according to claim 1, characterized in that, The data collected by the data acquisition module includes: The optical power transmitted and received by the PON port of the OLT device is collected via the SNMP protocol; The electronic tag contacts of the intelligent optical splitter can be used to sense the occupancy status of the splitter ports and the serial number of the connected ONUs in real time. Collect the ONU's transmit and receive power, voltage, current, temperature, and online time.

Images