A method for monitoring the on-line state of a power cable sheath insulation
By deploying multi-node synchronous acquisition units along the transmission cable, constructing a dynamic graph neural network and an electromagnetic-thermal-current three-field coupled simulation engine, and combining it with a multi-agent game model, continuous online assessment of the insulation status of the transmission cable sheath was achieved. This solved the problems of assessment distortion and response lag in traditional monitoring methods and improved the ability to capture early signs of insulation degradation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ELECTRIC POWER RESEARCH INSTITUTE OF STATE GRID SHANDONG ELECTRIC POWER COMPANY
- Filing Date
- 2026-04-14
- Publication Date
- 2026-06-26
AI Technical Summary
Existing methods for monitoring the insulation of power transmission cable sheaths rely on the acquisition of a single physical quantity, which cannot achieve synchronous perception of the electrical-thermal-mechanical coupling state. Furthermore, most systems cannot achieve continuous online evaluation, resulting in evaluation distortion and response lag.
By deploying multi-node synchronous acquisition units along the cable, multi-dimensional sensing data of grounding wire current, surface temperature and ambient temperature are obtained. A dynamic graph neural network is constructed to drive a physical simulation engine that couples electromagnetic, thermal and current fields, simulates the transient charge migration path of partial discharge pulses, and uses a multi-agent game model to assess the insulation health status.
It enables continuous online sensing of the electrical-thermal-environmental three-field coupling parameters of the insulation status of power transmission cable sheaths, improves the ability to capture early signs of insulation degradation, overcomes the problems of assessment distortion and response lag in traditional methods, and provides high-confidence insulation health status decision signals.
Smart Images
Figure CN122017505B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of online condition monitoring technology, and in particular to a method for online condition monitoring of the insulation of power transmission cable sheaths. Background Technology
[0002] Currently, monitoring of transmission cable sheath insulation generally relies on traditional methods such as grounding current measurement, surface temperature detection, infrared thermal imaging, and offline partial discharge testing. These methods are typically based on the acquisition of a single physical quantity, such as measuring the sheath grounding circulation current using a clamp meter, or obtaining the cable surface temperature distribution using an infrared thermometer. Partial discharge is often detected at fixed points using pulse current methods or ultrasonic methods under power outage conditions. Monitoring nodes are sparsely distributed, lacking the ability to synchronously perceive the electrical-thermal-mechanical coupling state of the entire cable, and most systems can only provide static or periodic data, failing to achieve continuous online assessment.
[0003] Existing technologies have limitations: single-parameter monitoring is difficult to reflect the multi-field coupling mechanism of insulation degradation; grounding current is easily affected by soil resistivity, stray current and multi-point grounding interference, leading to evaluation distortion; partial discharge detection is difficult to capture nanosecond-level transient pulses due to insufficient sampling frequency and weak noise resistance, and the positioning error generally exceeds 1 meter; infrared thermometry is affected by environmental radiation and surface emissivity and cannot penetrate the sheath to identify internal defects; offline detection requires power outage operation, has a delayed response, and cannot support condition-based maintenance requirements. Summary of the Invention
[0004] This application provides an online monitoring method for the insulation status of power transmission cable sheaths to solve the problems of existing solutions, such as single-parameter monitoring being unable to reflect the multi-field coupling mechanism of insulation degradation, infrared thermometry being affected by environmental radiation and surface emissivity, offline detection requiring power outage operation, and response lag.
[0005] In a first aspect, this application provides a method for online monitoring of the insulation condition of a power transmission cable sheath, the method comprising:
[0006] Based on the length of the transmission cable sheath insulation, determine the specific data acquisition node and obtain the grounding current, surface temperature, and ambient temperature of the data acquisition node;
[0007] Each data acquisition node along the cable is considered as a graph node. The physical distance and electrical coupling relationship are used as edge weights to obtain a multidimensional sensor data graph. The multidimensional sensor data graph is used as input to construct a dynamic graph neural network and output the graph structure state flow.
[0008] Using graph-structured state flow as input, a physical simulation engine based on electromagnetic-thermal-current three-field coupling is driven to simulate the transient charge migration path of partial discharge pulses in the XLPE sheath; by solving the modified Maxwell equations, the transient electromagnetic fingerprint corresponding to each node is generated.
[0009] The electromagnetic fingerprint of insulation degradation is input into a multi-agent game model based on information entropy flow. Each cable node is regarded as an agent with state assessment capability, and the dynamic decision signal of insulation health status is output.
[0010] In one implementation of this application, the specific data acquisition node is determined based on the length of the transmission cable sheath insulation, specifically including:
[0011] Obtain information on the total length of the insulation sheath of the power transmission cable to be monitored;
[0012] The theoretical number of data acquisition nodes is calculated by dividing the total cable length by the preset node spacing; if the result is not an integer, it is rounded up.
[0013] Starting from the beginning of the cable, the specific physical location of each data acquisition node is determined sequentially along the length of the cable according to the preset node spacing.
[0014] After completing the theoretical calculations for uniform point distribution, data acquisition nodes were added at cable joints, bends, previously known vulnerable points, and pre-set key monitoring sections.
[0015] In one implementation of this application, acquiring the grounding current, surface temperature, and ambient temperature of the acquisition node specifically includes:
[0016] The grounding current of the acquisition node is obtained non-contactly through a clamp-on current transformer, the surface temperature of the acquisition node is monitored through an integrated temperature sensor, and the ambient temperature corresponding to the acquisition node is obtained through a preset environmental sensor.
[0017] In one implementation of this application, each data acquisition node along the cable is considered as a graph node, and a multi-dimensional sensor data graph is obtained using physical distance and electrical coupling relationship as edge weights, specifically including:
[0018] Each data acquisition node deployed along the insulated line of the power transmission cable is considered as a node in a graph structure; wherein, the attribute vector of each node includes at least: grounding current, surface temperature, and ambient temperature;
[0019] Calculate the actual physical distance between any two data acquisition nodes along the cable; use the physical distance as the first weight of the edge;
[0020] Obtain the electrical topology of the power transmission cable, assign the second weight of the first echelon to the edges between nodes located in the same power supply circuit, phase or with direct electrical connection, and assign the second weight of the second echelon to the edges between nodes not located in the same power supply circuit, phase or with direct electrical connection.
[0021] The first weight and the second weight are normalized respectively; the normalized first weight and the second weight are input into a preset fusion function to obtain the final weight value;
[0022] All nodes, multidimensional sensing attributes, and the connection relationships between all edges, along with their final weight values, are organized together to form a complete multidimensional sensing data graph.
[0023] In one implementation of this application, a multidimensional sensor data graph is used as input to construct a dynamic graph neural network, and the output is a graph structure state flow, specifically including:
[0024] Based on the number of nodes, node attribute dimensions, and edge weight range of the multidimensional sensor data graph, the number of network layers, node update iterations, and feature propagation step size of the dynamic graph neural network are set.
[0025] The grounding current, surface temperature, and ambient temperature corresponding to each acquisition node are used as node state characteristics; all node state characteristics are uniformly normalized.
[0026] Through multi-layer network iterative calculation, the local state features and overall topological features of all nodes along the cable are extracted layer by layer, and the changes in sensor data in the time dimension are fused to form a continuous dynamic feature sequence.
[0027] The extracted and integrated node state features and time series features are integrated into a unified graph structure state flow output.
[0028] In one implementation of this application, a physical simulation engine based on electromagnetic-thermal-current three-field coupling is driven by a graph-structured state flow as input to simulate the transient charge migration path of a partial discharge pulse in an XLPE sheath, specifically including:
[0029] Based on the material parameters, geometric structure, and electrical characteristics of the XLPE sheath of the cable, a three-dimensional refined physical simulation model of the cable segment is established; a multi-physics coupling simulation engine integrating electromagnetic field, thermal field, and current field is built.
[0030] The node state characteristics in the graph structure state flow are converted into boundary conditions and excitation source parameters that the simulation engine can recognize, and the node state characteristics are mapped to the spatial location of the physical simulation model.
[0031] In the simulation model, the preset potential defect points of the cable insulation are used as partial discharge sources to simulate the preset transient high voltage pulses emitted by the partial discharge sources.
[0032] Based on electromagnetic-thermal coupling, the local Joule heat generated by pulsed discharge is calculated and used as the input for updating the thermal field;
[0033] Based on thermal-current coupling, the current field is updated according to the effect of material temperature change on conductivity; based on current-electromagnetic coupling, the updated current field is used as the excitation source of the electromagnetic field to form a closed-loop iterative calculation.
[0034] By solving Maxwell's equations modified according to actual physical conditions, the entire coupled system is driven to calculate the propagation of electromagnetic waves in XLPE insulating media containing defects and inhomogeneities.
[0035] Through simulation at several time steps, the entire transient propagation and evolution path of the transient electromagnetic disturbance generated by the partial discharge pulse inside the cable insulation sheath was finally depicted.
[0036] In one implementation of this application, a transient electromagnetic fingerprint corresponding to each node is generated by solving the modified Maxwell's equations, specifically including:
[0037] Through simulation at several time steps, the entire transient propagation and evolution path of the transient electromagnetic disturbance generated by the partial discharge pulse inside the cable insulation sheath is depicted.
[0038] Based on path calculation, the preset transient electromagnetic response signal characterized by each simulated node on the output cable is used as a transient electromagnetic fingerprint reflecting the local microstructure state of the insulation at that point.
[0039] In one implementation of this application, the electromagnetic fingerprint of insulation degradation is input into a multi-agent game model based on information entropy flow. Each cable node is regarded as an agent with state evaluation capabilities, and a dynamic decision signal of insulation health status is output, specifically including:
[0040] Each data acquisition node equipped with sensors along the power transmission cable is defined as an independent intelligent agent. The insulation degradation electromagnetic fingerprint corresponding to each intelligent agent is used as the initial state information input model for that agent. An information entropy flow is input to each intelligent agent, describing the disorder of the system. Transient electromagnetic fingerprints are shared among the agents. Each intelligent agent engages in a game-like interaction based on its own electromagnetic fingerprint state and the input information entropy flow. The goal of each agent is to assess the insulation health state of the cable segment it represents. During the game, agents can cooperate or compete, adjusting their state assessment results through pre-defined strategies. After a pre-defined round of game interaction and interaction with the electromagnetic fingerprint state, a quantitative assessment signal regarding the insulation health state of the corresponding cable segment is obtained from the final output of each intelligent agent.
[0041] As can be seen from the above technical solutions, this application has the following advantages:
[0042] By deploying multi-node synchronous acquisition units along the cable line, multi-dimensional sensing data of grounding wire current, surface temperature, and ambient temperature are acquired in real time, enabling continuous online sensing of the electrical-thermal-environmental three-field coupled parameters of the transmission cable sheath insulation condition across the entire line. Compared to traditional single-parameter offline detection methods, this structure eliminates the problems of missed detections and response lags caused by sparse monitoring points and long sampling cycles, providing a continuous, dynamic, and comprehensive real-time data foundation for condition-based maintenance and improving the ability to detect early signs of insulation degradation.
[0043] By constructing a dynamic graph structure with physical distance and electrical coupling as edge weights from multidimensional sensor data and inputting it into a dynamic graph neural network, the nonlinear correlation and state propagation characteristics between nodes along the cable are modeled. The output graph structure state flow accurately reflects the spatial evolution trend of insulation degradation. This mechanism overcomes the evaluation distortion problem caused by neglecting the multi-physics coupling mechanism in traditional methods. Especially under complex conditions where the grounding current is affected by soil resistivity or stray current interference, it can still achieve adaptive filtering and effective feature extraction of interference signals through graph topology, thus improving the evaluation robustness.
[0044] A graph-structured state flow-driven electromagnetic-thermal-current three-field coupled physical simulation engine, combined with the solution of modified Maxwell's equations, can generate a transient electromagnetic fingerprint for each node, realistically reproducing the charge migration path of partial discharge pulses in the XLPE sheath. This fingerprint information is further input into a multi-agent game model based on information entropy flow, allowing each node to act as an independent evaluation agent. Through the game mechanism, it dynamically weighs local anomalies against the global state, outputting a high-confidence insulation health status decision signal. This architecture does not rely on offline testing or manual intervention, achieving a closed-loop online evaluation from data perception to intelligent diagnosis, solving the core shortcomings of traditional partial discharge detection such as insufficient sampling frequency, large positioning errors, and the need for power outages. Attached Figure Description
[0045] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0046] Figure 1 This is a flowchart of an online monitoring method for the insulation status of a power transmission cable sheath provided in an embodiment of this application. Detailed Implementation
[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0048] Those skilled in the art should understand that the embodiments described below are merely preferred embodiments of this disclosure and do not imply that this disclosure can only be implemented through these preferred embodiments. These preferred embodiments are merely used to explain the technical principles of this disclosure and are not intended to limit the scope of protection of this disclosure. Based on the preferred embodiments provided by this disclosure, all other embodiments obtained by those skilled in the art without creative effort should still fall within the scope of protection of this disclosure.
[0049] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0050] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.
[0051] The embodiment provides a method for online monitoring of the insulation status of power transmission cable sheaths, such as... Figure 1 As shown in the embodiments of this application, the method mainly includes the following steps:
[0052] Step 110: Determine the specific data acquisition node based on the length of the power transmission cable sheath insulation, and obtain the grounding current, surface temperature, and ambient temperature of the data acquisition node.
[0053] In some embodiments, the specific data acquisition node is determined based on the length of the transmission cable sheath insulation, specifically including:
[0054] Obtain the total length information of the sheath insulation of the transmission cable to be monitored; calculate the theoretical number of acquisition nodes by dividing the total cable length by the preset node spacing; if the result is not an integer, round it up; starting from the beginning of the cable, determine the specific physical layout location of each acquisition node along the length of the cable according to the preset node spacing; after completing the theoretical calculation of uniform distribution, add acquisition nodes at the joints, bends, previously known vulnerable points, and preset key monitoring sections of the cable.
[0055] As an example, the total length of the cable sheath insulation is 3.2 km, with a preset node spacing of 200 m, resulting in a theoretical number of 16 nodes (3200 ÷ 200 = 16). Along the cable route, starting from the initial terminal, a data collection point is set up every 200 m, located near the cable supports for easy installation and away from areas with strong electromagnetic interference. Additional data collection nodes are added at cable joints (5 locations), 90° bends (3 locations), and sections with frequent historical faults (2 locations, at K1+450 and K2+820), increasing the total number of nodes to 26. Each node is fixed to the outer surface of the cable sheath, and the sensors are protected by stainless steel clamps and waterproof sealing boxes to ensure long-term environmental adaptability.
[0056] Specifically, acquiring the grounding current, surface temperature, and ambient temperature of the data acquisition node includes:
[0057] The grounding current of the acquisition node is obtained non-contactly through a clamp-on current transformer, the surface temperature of the acquisition node is monitored through an integrated temperature sensor, and the ambient temperature corresponding to the acquisition node is obtained through a preset environmental sensor.
[0058] As an example, each acquisition node is equipped with three types of sensors: the grounding wire current is measured non-contactly using a clamp-on current transformer with a range of 0~50A and an accuracy of ±0.5%, installed below the connection point between the sheath grounding wire and the grounding grid; the surface temperature is measured using a PT100 platinum resistance temperature sensor, encapsulated in a weather-resistant epoxy resin housing, directly attached to the sheath surface, with the heat conduction path optimized by thermally conductive silicone grease; the ambient temperature is acquired by an independent digital temperature and humidity sensor (model SHT35), installed in a protective box 1.5m above the ground, sheltered from sun and rain, with a horizontal distance of no more than 5m from the corresponding acquisition point. All sensor signals are transmitted to the nearest data acquisition unit via shielded twisted-pair cables, with the sampling frequency uniformly set to 1Hz and the synchronization clock error controlled within ±10ms.
[0059] Step 120: Treat each data acquisition node along the cable as a graph node, and use physical distance and electrical coupling relationship as edge weights to obtain a multi-dimensional sensor data graph. Use the multi-dimensional sensor data graph as input to construct a dynamic graph neural network and output the graph structure state flow.
[0060] Specifically, each data acquisition node along the cable is considered a graph node, and physical distance and electrical coupling relationship are used as edge weights to obtain a multi-dimensional sensor data graph, which includes:
[0061] Each data acquisition node deployed along the insulated sheath of the power transmission cable is considered a node in the graph structure. Each node's attribute vector includes at least: grounding current, surface temperature, and ambient temperature. The actual physical distance between any two acquisition nodes along the cable is calculated. This physical distance is used as the first weight of the edge. The electrical topology of the power transmission cable is obtained. Edges between nodes located in the same power supply circuit, phase, or with direct electrical connection are assigned a second weight to the first echelon, and edges between nodes not located in the same power supply circuit, phase, or with direct electrical connection are assigned a second weight to the second echelon. The first and second weights are normalized. The normalized first and second weights are input into a preset fusion function to obtain the final weight value. All nodes, multi-dimensional sensing attributes, and the connection relationships between all edges, along with the final weight value, are collectively organized into a complete multi-dimensional sensing data graph.
[0062] As an example, 26 data acquisition nodes were deployed in the cable line. The actual physical distance between nodes was determined by verifying the cable laying drawings and using a laser rangefinder. The waveform propagation method was used for verification: based on the wave propagation speed of oil-impregnated paper-insulated cables (160 m / s) and a sampling frequency of 25 MHz, the distance represented by each point was calculated to be 3.2 meters. Combined with the waveform grid count recorded by the monitoring system (5 points per grid), the error in the distance between adjacent nodes was estimated to be controlled within ±0.5%. The electrical topology was derived from the substation SCADA system. Based on the switch status and feeder affiliation, nodes in the same power supply circuit (such as all nodes of phase A in a three-phase cable) were identified and assigned to the first tier of weights; nodes in different circuits (such as different phases or different feeders) were assigned to the second tier of weights. The physical distance weight and electrical coupling weight were respectively subjected to minimum-maximum normalization. The normalization formula is as follows:
[0063] The physical distance ranges from 0 to 500m, and the electrical coupling weight ranges from 0 to 1.
[0064] The fusion function uses linear weighting: Finally, a directed weighted graph with 26 nodes, each with attributes of [grounding current, surface temperature, ambient temperature], is constructed, with 325 edges (fully connected) and weights ranging from [0.12, 0.98].
[0065] This involves using multidimensional sensor data graphs as input to construct a dynamic graph neural network, outputting a graph-structured state flow, specifically including:
[0066] Based on the number of nodes, node attribute dimensions, and edge weight range of the multidimensional sensor data graph, the number of network layers, node update iterations, and feature propagation step size of the dynamic graph neural network are set. The grounding current, surface temperature, and ambient temperature corresponding to each acquisition node are used as node state features. All node state features are uniformly normalized. Through multi-layer network iterative calculation, the local state features and overall topological features of all nodes along the cable are extracted layer by layer, and the changes in sensor data over time are fused to form a continuous dynamic feature sequence. The extracted and integrated node state features and time series features are integrated into a unified graph structure state flow output.
[0067] As an example, the dynamic graph neural network employs a two-layer graph convolutional structure, with each layer updating nodes 3 times and a feature propagation stride of 1 to match the state evolution time series at a sampling frequency of 1Hz. The network input consists of 26 nodes with 3 node feature dimensions (current, surface temperature, ambient temperature) and a hidden layer dimension of 16. Network training uses fixed parameters: a learning rate of 0.01, a ReLU activation function, and no Dropout. After each round of graph neural network computation, the output is a 16-dimensional hidden state vector for each node, concatenated with the state from the previous time step to form a time series feature. Finally, the 26×16-dimensional state matrices of all nodes are sorted by node number and output as a graph-structured state flow in JSON array format.
[0068] Step 130: Using the graph structure state flow as input, drive the physical simulation engine based on electromagnetic-thermal-current three-field coupling to simulate the transient charge migration path of the partial discharge pulse in the XLPE sheath; generate the transient electromagnetic fingerprint corresponding to each node by solving the modified Maxwell equations.
[0069] The simulation engine, driven by a graph-structured state flow as input, simulates the transient charge migration path of a partial discharge pulse within the XLPE sheath. Specifically, it includes:
[0070] Based on the material parameters, geometric structure, and electrical characteristics of the XLPE sheath of the cable, a three-dimensional refined physical simulation model of the cable segment is established. A multi-physics coupled simulation engine integrating electromagnetic field, thermal field, and current field is built. The node state characteristics in the graph structure state flow are converted into boundary conditions and excitation source parameters that the simulation engine can recognize, and the node state characteristics are mapped to the spatial location of the physical simulation model. In the simulation model, the preset potential defect points of the cable insulation are used as partial discharge sources to simulate the preset transient high-voltage pulse emitted by the partial discharge sources. Based on electromagnetic-thermal coupling, the generation of pulse discharge is calculated. The local Joule heating is used as the input for updating the thermal field; based on thermal-current coupling, the current field is updated according to the effect of material temperature change on conductivity; based on current-electromagnetic coupling, the updated current field is used as the excitation source of the electromagnetic field to form a closed-loop iterative calculation; by solving Maxwell's equations modified according to actual physical conditions, the entire coupled system is driven to calculate the propagation of electromagnetic waves in XLPE insulation medium containing defects and inhomogeneities; through simulation at several time steps, the entire transient propagation and evolution path of transient electromagnetic disturbances generated by partial discharge pulses inside the cable insulation sheath is finally depicted.
[0071] It should be noted that the preset transient high voltage pulse range is 5kV–30kV.
[0072] As an example, in a cable line, based on the 26 node graph structure state flow generated in the previous steps, the [grounding current, surface temperature, ambient temperature] state vector of each node is mapped to the corresponding spatial location in the 3D cable simulation model. The simulation model is constructed using COMSOL Multiphysics, and its geometry includes a conductor, a 3mm thick XLPE insulation layer, and an outer sheath. Five 1mm diameter spherical voids are pre-defined within the insulation layer as candidate regions for partial discharge sources. The XLPE material parameters are taken from industry measured values: relative permittivity 2.3±0.1, thermal conductivity 0.3 W / (m·K), initial conductivity 5×10⁻⁶. - ¹ 4 S / m. The effect of temperature on conductivity is modeled using the Arrhenius model:
[0073] ;in, Let be the Boltzmann constant, and T be the absolute temperature (K). The surface temperature of each node in the graph-structured state flow is used as the boundary heat source input, driving the thermal field update; changes in the thermal field are fed back to the conductivity calculation, updating the current field; the current field serves as the excitation source driving the electromagnetic field solution. The simulation time step is set to 0.5 ns, with a total duration of 100 ns. The time-varying Maxwell's equations are used, embedding a temperature-dependent conductivity term to achieve a three-field closed-loop iteration.
[0074] Specifically, by solving the modified Maxwell's equations, a transient electromagnetic fingerprint corresponding to each node is generated, including:
[0075] Through simulation at several time steps, the entire transient propagation and evolution path of the transient electromagnetic disturbance generated by the partial discharge pulse inside the cable insulation sheath is depicted; based on the path calculation, the preset transient electromagnetic response signal characterized by each simulation node on the cable is output as a transient electromagnetic fingerprint reflecting the local microstructure state of the insulation at that point.
[0076] As an example, the simulation system completes the electromagnetic-thermal-current three-field coupling calculation in each iteration: first, the Joule heat density is calculated from the transient current field. The thermal field module updates the temperature distribution; then the conductivity is corrected based on the temperature change. The updated current density is fed back to the current field; finally, the corrected Maxwell's equations are solved using the updated current density as the source term. The spatiotemporal distributions of electric field E(t) and magnetic flux density B(t) are obtained. After 1200 time steps of simulation, the time-domain waveform of electric field intensity E(t) at each node is extracted as a transient electromagnetic fingerprint: rise time ≤ 5 ns, main pulse width 20–50 ns, amplitude range 0.1–5 V / m, and waveform morphology modulated by local defect geometry and material inhomogeneity. The output format is a JSON array containing node ID, timestamp, and E(t) sampling sequence (sampling rate 1 GHz, duration 200 ns).
[0077] Step 140: Input the electromagnetic fingerprint of insulation degradation into the multi-agent game model based on information entropy flow, treat each cable node as an agent with state assessment capability, and output the dynamic decision signal of insulation health status.
[0078] In some embodiments, the electromagnetic fingerprint of insulation degradation is input into a multi-agent game model based on information entropy flow, each cable node is regarded as an agent with state evaluation capabilities, and a dynamic decision signal of insulation health status is output, specifically including:
[0079] Each data acquisition node equipped with sensors along the power transmission cable is defined as an independent intelligent agent. The insulation degradation electromagnetic fingerprint corresponding to each intelligent agent is used as the initial state information input model for that agent. An information entropy flow is input to each intelligent agent, describing the disorder of the system. Transient electromagnetic fingerprints are shared among the agents. Each intelligent agent engages in a game-like interaction based on its own electromagnetic fingerprint state and the input information entropy flow. The goal of each agent is to assess the insulation health state of the cable segment it represents. During the game, agents can cooperate or compete, adjusting their state assessment results through pre-defined strategies. After a pre-defined round of game interaction and interaction with the electromagnetic fingerprint state, a quantitative assessment signal regarding the insulation health state of the corresponding cable segment is obtained from the final output of each intelligent agent.
[0080] As an example, in a 110kV XLPE cable line, the transient electromagnetic fingerprints (sampling rate 1GHz, duration 200ns) corresponding to each of the 26 data acquisition nodes are normalized and used as the initial state input for each agent. Each agent shares its electromagnetic fingerprint sequence to the local area network every second via the MQTT protocol, forming a distributed information interaction environment. The information entropy flow is defined as: calculating the Shannon entropy for the time-domain signal of the electromagnetic fingerprint of each node, using the formula... ,in Let represent the probability distribution of the electric field amplitude within 256 quantization intervals, reflecting the disturbance disorder of the local insulating medium. In each round of the game, each agent adjusts its strategy based on its own entropy and the average entropy of its neighborhood (three nodes before and after it): the cooperative strategy minimizes the mean squared error of its entropy relative to the neighborhood, while the competitive strategy aims to make its own entropy lower than the neighborhood mean. The objective function is... Where α=0.7 and β=0.3. The game consists of 3 rounds, with a 200ms interval between each round to match the electromagnetic fingerprint update cycle. After three rounds of interaction, each agent outputs a normalized health index as the final evaluation signal, ranging from [0,1], where 0 represents normal and 1 represents severe degradation with no performance improvement indication. The output format is consistent with the previous steps, being a JSON array containing the node ID, timestamp, and quantified evaluation value, which is directly called by the state decision engine.
[0081] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for online monitoring of the insulation status of power transmission cable sheaths, characterized in that, The method includes: Based on the length of the transmission cable sheath insulation, determine the specific data acquisition node and obtain the grounding current, surface temperature, and ambient temperature of the data acquisition node; Each data acquisition node along the cable is considered as a graph node. The physical distance and electrical coupling relationship are used as edge weights to obtain a multidimensional sensor data graph. The multidimensional sensor data graph is used as input to construct a dynamic graph neural network and output the graph structure state flow. Using graph-structured state flow as input, a physical simulation engine based on electromagnetic-thermal-current three-field coupling is driven to simulate the transient charge migration path of partial discharge pulses in the XLPE sheath. By solving the modified Maxwell's equations, a transient electromagnetic fingerprint corresponding to each node is generated. Specifically, through simulation at several time steps, the entire transient propagation and evolution path of the transient electromagnetic disturbance generated by the partial discharge pulse inside the cable insulation sheath is depicted. Based on the path calculation, the preset transient electromagnetic response signal characterized by each simulated node on the cable is output as a transient electromagnetic fingerprint reflecting the local microstructure state of the insulation of that node. The electromagnetic fingerprint of insulation degradation is input into a multi-agent game model based on information entropy flow. Each cable node is treated as an agent with state evaluation capabilities, and the dynamic decision signal of insulation health status is output, specifically including: Each data acquisition node equipped with sensors along the power transmission cable is defined as an independent intelligent agent. The insulation degradation electromagnetic fingerprint corresponding to each intelligent agent is used as the initial state information input model for that agent. An information entropy flow is input to each intelligent agent, describing the disorder of the system. Transient electromagnetic fingerprints are shared among the agents. Each intelligent agent engages in a game-like interaction based on its own electromagnetic fingerprint state and the input information entropy flow. The goal of each agent is to assess the insulation health state of the cable segment it represents. During the game, agents can cooperate or compete, adjusting their state assessment results through pre-defined strategies. After a pre-defined round of game interaction and interaction with the electromagnetic fingerprint state, a quantitative assessment signal regarding the insulation health state of the corresponding cable segment is obtained from the final output of each intelligent agent.
2. The method for online monitoring of the insulation status of transmission cable sheaths according to claim 1, characterized in that, The specific data collection nodes are determined based on the length of the transmission cable sheath insulation, including: Obtain information on the total length of the insulation sheath of the power transmission cable to be monitored; The theoretical number of data acquisition nodes is calculated by dividing the total cable length by the preset node spacing; if the result is not an integer, it is rounded up. Starting from the beginning of the cable, the specific physical location of each data acquisition node is determined sequentially along the length of the cable according to the preset node spacing. After completing the theoretical calculations for uniform point distribution, data acquisition nodes were added at cable joints, bends, previously known vulnerable points, and pre-set key monitoring sections.
3. The method for online monitoring of the insulation status of power transmission cable sheaths according to claim 1, characterized in that, Acquire the grounding current, surface temperature, and ambient temperature of the data acquisition node, specifically including: The grounding current of the acquisition node is obtained non-contactly through a clamp-on current transformer, the surface temperature of the acquisition node is monitored through an integrated temperature sensor, and the ambient temperature corresponding to the acquisition node is obtained through a preset environmental sensor.
4. The method for online monitoring of the insulation status of transmission cable sheaths according to claim 1, characterized in that, Each data acquisition node along the cable is considered a graph node, and physical distance and electrical coupling relationship are used as edge weights to obtain a multi-dimensional sensor data graph, specifically including: Each data acquisition node deployed along the insulated line of the power transmission cable is considered as a node in a graph structure; wherein, the attribute vector of each node includes at least: grounding current, surface temperature, and ambient temperature; Calculate the actual physical distance between any two data acquisition nodes along the cable; use the physical distance as the first weight of the edge; Obtain the electrical topology of the power transmission cable, assign the second weight of the first echelon to the edges between nodes located in the same power supply circuit, phase or with direct electrical connection, and assign the second weight of the second echelon to the edges between nodes not located in the same power supply circuit, phase or with direct electrical connection. The first weight and the second weight are normalized respectively; the normalized first weight and the second weight are input into a preset fusion function to obtain the final weight value; All nodes, multidimensional sensing attributes, and the connection relationships between all edges, along with their final weight values, are organized together to form a complete multidimensional sensing data graph.
5. The method for online monitoring of the insulation status of transmission cable sheaths according to claim 1, characterized in that, Using multidimensional sensor data graphs as input, a dynamic graph neural network is constructed to output a graph-structured state flow, specifically including: Based on the number of nodes, node attribute dimensions, and edge weight range of the multidimensional sensor data graph, the number of network layers, node update iterations, and feature propagation step size of the dynamic graph neural network are set. The grounding current, surface temperature, and ambient temperature corresponding to each acquisition node are used as node state characteristics; all node state characteristics are uniformly normalized. Through multi-layer network iterative calculation, the local state features and overall topological features of all nodes along the cable are extracted layer by layer, and the changes in sensor data in the time dimension are fused to form a continuous dynamic feature sequence. The extracted and integrated node state features and time series features are integrated into a unified graph structure state flow output.
6. The method for online monitoring of the insulation status of transmission cable sheaths according to claim 1, characterized in that, Using graph-structured state flow as input, a physical simulation engine based on electromagnetic-thermal-current three-field coupling is driven to simulate the transient charge migration path of partial discharge pulses in the XLPE sheath, specifically including: Based on the material parameters, geometric structure, and electrical characteristics of the XLPE sheath of the cable, a three-dimensional refined physical simulation model of the cable segment is established; a multi-physics coupling simulation engine integrating electromagnetic field, thermal field, and current field is built. The node state characteristics in the graph structure state flow are converted into boundary conditions and excitation source parameters that the simulation engine can recognize, and the node state characteristics are mapped to the spatial location of the physical simulation model. In the simulation model, the preset potential defect points of the cable insulation are used as partial discharge sources to simulate the preset transient high voltage pulses emitted by the partial discharge sources. Based on electromagnetic-thermal coupling, the local Joule heat generated by pulsed discharge is calculated and used as the input for updating the thermal field; Based on thermal-current coupling, the current field is updated according to the effect of material temperature change on conductivity; based on current-electromagnetic coupling, the updated current field is used as the excitation source of the electromagnetic field to form a closed-loop iterative calculation. By solving Maxwell's equations modified according to actual physical conditions, the entire coupled system is driven to calculate the propagation of electromagnetic waves in XLPE insulating media containing defects and inhomogeneities. Through simulation at several time steps, the entire transient propagation and evolution path of the transient electromagnetic disturbance generated by the partial discharge pulse inside the cable insulation sheath was finally depicted.