A real-time analysis and early warning method for gas composition inside a ring main unit

Abstract

The application provides a real-time analysis and early warning method for gas components in a ring main unit, comprising: obtaining initial gas concentrations and initial temperature distribution of multiple spatial positions in the ring main unit, identifying local areas with obvious temperature rise in the cabinet through the initial temperature distribution, locating heat-prone parts such as cable terminals to obtain abnormal heat source coordinates; evaluating a first concentration change rate and a second concentration change rate, identifying that a first sampling node with a slow concentration rise is interfered by backflow gas dilution to obtain an interfered node marker and an effective sampling node set; removing data of the first sampling node according to the interfered node marker, tracking a concentration change trajectory of a second sampling node in the effective sampling node set to obtain a gas diffusion path.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for real-time analysis and early warning of gas composition inside a ring main unit. Background Technology

[0002] As the core hub of the power distribution network, the internal operating status of ring main units directly affects the safety and stability of the power system. During long-term operation, when solid insulation media such as cable terminals and epoxy-cast insulation components experience thermal aging, moisture degradation, or partial discharge, they decompose and release specific gaseous components. These include ozone generated in the air by partial discharge and characteristic gases produced by the thermal decomposition of solid insulation. Real-time analysis and early warning of these gaseous components are a crucial line of defense against major power outages. In the prior art, a partial discharge monitoring system for a ring main unit, with publication number CN103412247A, discloses a technology that collects and analyzes ozone concentration and temperature inside the ring main unit through an exhaust pipe, regulating valve, gas sampling unit, temperature sensor, and signal analysis unit to achieve partial discharge monitoring. The disclosed technology still mainly relies on a fixed exhaust channel and preset sampling position to obtain gas information inside the unit. It does not combine the movement of the thermal plume inside the unit, the backflow dilution effect caused by the obstruction and reversal at the top, and the change in the slope of gas concentration rise at different spatial locations to adjust the early warning judgment conditions. Therefore, when abnormal gas is flushed and diluted by the backflow air or fault gas preferentially accumulates in the lateral area away from the fault source, problems such as delayed concentration response and untimely early warning may easily occur.

[0003] Traditional monitoring methods generally follow the principle of proximity, assuming that placing the gas sampling port as close as possible to heat-prone areas such as cable terminals will quickly detect abnormal gases. However, this judgment method based on absolute spatial advantage completely ignores the complex fluid dynamics changes within the enclosed cabinet, leading to frequent slow responses from early warning systems in actual operation. The root cause of this slowness lies in the conflicting relationship between natural air convection and gas diffusion within the cabinet. When a cable terminal abnormally heats up and releases decomposition gases, the heat drives the surrounding air to form an upward thermal plume. This airflow is inevitably obstructed and deflected upon reaching the top of the cabinet, forming downward or lateral backflow gas. At this point, the sampling port, originally positioned near the heat-prone area, is precisely in the path of this backflow gas. The newly generated abnormal decomposition gases are forcibly diluted by a large amount of backflow air that does not carry fault information before they can fully enter the sensing channel. For example, when a cable joint experiences a minor overheating fault, the concentration of the fault gas in the gas mixture drawn in from the sampling port immediately adjacent to the joint rises extremely slowly. Conversely, in the lateral region some distance from the heating center, because it avoids the direct impact of the strong backflow, the fault gas accumulates more quickly along the flow direction, resulting in a significantly steeper increase in concentration. This severe mismatch between spatial distance and the rate of concentration change renders conventional early warning mechanisms ineffective.

[0004] How to accurately capture the true correlation between the movement trajectory of the thermal plume and the changes in gas concentration under complex natural convection and backflow dilution interference, so as to dynamically adjust the early warning judgment conditions for abnormal gases and completely solve the contradiction between sampling location and gas capture speed, has become a key issue for achieving accurate early warning of internal faults in ring main units. Summary of the Invention

[0005] This invention provides a method for real-time analysis and early warning of gas composition inside a ring main unit, the method comprising: The initial gas concentration and initial temperature distribution at multiple spatial locations inside the ring main unit are obtained. The local areas with rising temperature inside the unit are identified by the initial temperature distribution, the heat source is located, and the coordinates of the abnormal heat source are obtained. Based on the coordinates of the abnormal heat source, and combined with the natural convection pattern formed by the heated gas rising along the heat source and falling back along the cabinet wall after cooling, the scouring path of the return gas near the heat source and the range of the lateral guide zone that the return gas does not flow directly through are determined. The first sampling node located on the backflow gas flushing path and the second sampling node located in the lateral guide zone are obtained. The first concentration change rate of the first sampling node and the second concentration change rate of the second sampling node are separated from the initial gas concentration at each spatial location. The rates of change of the first and second concentrations were evaluated, and the first sampling node with a slower concentration increase was identified as being affected by dilution interference from the backflow gas. The disturbed node was marked and the set of effective sampling nodes was obtained. Based on the data of the first sampling node that is removed by the interference node marker, the concentration change trajectory of the second sampling node in the set of valid sampling nodes is tracked to obtain the diffusion path of the decomposition gas. By decomposing the gas diffusion path to extract the true concentration rise slope and the spatial distance between the effective sampling node set and the abnormal heat source, the true concentration rise slope and spatial distance are integrated into a comprehensive basis for identifying the severity of insulation faults. The warning triggering threshold is determined by comprehensively identifying the severity of insulation faults and matching the strength of return current dilution. The internal operating conditions of the ring main unit are assessed to obtain the final fault warning result, which is then output to the terminal equipment.

[0006] Furthermore, obtaining the initial gas concentration and initial temperature distribution at multiple spatial locations inside the ring main unit includes: Multiple spatial locations are divided along the vertical and horizontal directions inside the ring main unit. Temperature sensing nodes and gas sampling nodes are arranged around the cable terminals, busbar connections, and epoxy-cast insulation components. Thermistor temperature measuring elements are used to synchronously collect temperature readings at each node. Electrochemical gas sensors are used to collect the initial gas concentrations of ozone and carbon monoxide characteristic gases. The initial gas concentrations are used as the concentration baseline for subsequent determination.

[0007] Furthermore, the step of identifying localized areas of temperature rise within the cabinet through the initial temperature distribution, locating the heat source, and obtaining the coordinates of the abnormal heat source includes: The initial temperature distribution is plotted along the three-dimensional coordinate direction inside the cabinet based on the readings of each temperature sensing node. The temperature rise gradient is calculated for the temperature difference between adjacent nodes. The area covered by adjacent nodes whose temperature rise gradient exceeds a preset threshold is marked as a set of candidate heat-generating areas. The location of the node with the highest reading and the steepest surrounding temperature rise gradient is determined as the coordinates of the abnormal heat source.

[0008] Furthermore, based on the coordinates of the abnormal heat source and the natural convection pattern formed by the heated gas rising along the heat source and falling back along the cabinet wall after cooling, the scouring path of the return gas near the heat source and the lateral guide zone range through which the return gas does not flow directly are determined, including: A buoyancy driving condition is applied at the coordinate position of the abnormal heat source. The steady-state solution of the air flow field inside the cabinet is obtained by computational fluid dynamics. The trajectory of the hot plume formed by the upward floating of the heated gas from the heat source is traced in the vertical direction. The upward main axis of the hot plume trajectory is used as the center reference line of the flow field partition. Along the rising main axis, identify the gas top obstruction area on the inner wall of the cabinet top, obtain the return flow components that fall down along the cabinet wall and diffuse in all directions, spatially compare the fall trajectory of the return flow components with the horizontal projection of the coordinates of the abnormal heat source, and determine the spatial range traversed by the fall trajectory as the flushing path of the return gas near the heat source. Within the remaining space, the spatial region where the flow velocity is lower than a preset threshold and the flow direction deviates from the main direction of the return flow component is defined as the lateral flow guidance zone.

[0009] Furthermore, acquiring the first sampling node located on the return gas flushing path and the second sampling node located in the lateral guide zone includes: Within the space covered by the return gas flushing path, the gas sampling nodes within that range are marked as first sampling nodes, and within the lateral guide zone, the gas sampling nodes within that range are marked as second sampling nodes. The initial gas concentration time series data of each spatial location are read according to a preset sampling time interval, and the concentration change rate is determined by the ratio of the concentration difference between adjacent sampling times to the sampling time interval.

[0010] Furthermore, the identification of the first sampling node with a slower concentration increase is affected by dilution interference from the backflow gas, resulting in a marked interference node and a set of valid sampling nodes, including: The first concentration change rate and the second concentration change rate are synchronously compared within the same time window. The first concentration change rate is subtracted from the second concentration change rate to obtain the rate difference. The rate difference is compared with a preset dilution interference threshold. The corresponding first sampling node that exceeds the threshold is marked as an interfered node. The remaining sampling nodes are summarized into a set of valid sampling nodes.

[0011] Furthermore, the step of removing the data from the first sampling node based on the interference node marker, tracing the concentration change trajectory of the second sampling node in the set of valid sampling nodes, and obtaining the diffusion path of the decomposition gas includes: Based on the interference node marking, the concentration records corresponding to the nodes marked as interference are removed from the concentration time series data of the first sampling node, and the concentration time series data and corresponding spatial coordinates of the second sampling node in the set of valid sampling nodes are retained to obtain the concentration observation data set after interference removal. The concentration values ​​of each second sampling node are read in the order of sampling time. The second sampling nodes whose concentration values ​​exceed the preset rise threshold at the same sampling time are sorted from high to low according to the concentration values ​​to obtain the decomposition gas concentration rise sequence. The spatial coordinates of adjacent nodes in the sequence of increasing concentration of the decomposition gas are connected sequentially along the direction of decreasing concentration values, and the spatial direction through which the connecting lines pass is determined as the diffusion path of the decomposition gas.

[0012] Furthermore, the extraction of the true concentration rise slope through decomposition of gas diffusion paths, and the spatial distance between the effective sampling node set and the abnormal heat source, are integrated into a comprehensive basis for identifying the severity of insulation faults, including: Concentration time-series data are sequentially read from the second sampling node with the highest concentration value to subsequent nodes along the diffusion path of the decomposed gas. The concentration rise slope of each node is determined by the ratio of the concentration difference between adjacent sampling times to the sampling time interval, and the maximum value is taken as the true concentration rise slope. The spatial distance is calculated based on the Euclidean distance formula in three-dimensional space; The slope of the actual concentration increase is assigned a positive weight, and the spatial distance is assigned a negative weight. The weighted sum is then used as a comprehensive basis for identifying the severity of insulation faults.

[0013] Furthermore, the determination of the early warning trigger threshold, which is adapted to the strength of return current dilution, based on a comprehensive identification criterion of the severity of insulation faults, includes: The proportion of the interfered node marker in the first sampling node and the second sampling node is read, and the proportion is used as a quantitative indicator of the backflow dilution strength. The preset warning trigger limit benchmark value is corrected according to the quantitative indicator, and the downward adjustment range of the warning trigger limit benchmark value is determined according to the quantitative indicator.

[0014] Furthermore, the assessment of the internal operating conditions of the ring main unit, obtaining the final fault warning result, and outputting the final fault warning result to the terminal equipment includes: The numerical values ​​of the comprehensive identification criteria for the severity of the insulation fault are compared step by step with the warning triggering limits. If the level exceeds the first limit, the internal operating condition of the ring main unit is determined to be at the attention level. If the level exceeds the second limit, it is determined to be at the alarm level. The final fault warning result is obtained and encapsulated into an alarm signal containing the coordinates of the heated part, the warning level and the concentration rise slope according to the preset message format. The signal is then output to the terminal device through the communication link between the ring main unit and the monitoring terminal.

[0015] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a method for real-time analysis and early warning of gas composition inside a ring main unit (RNB), proposing a solution to the problem of natural convection dilution interference in RNB gas concentration monitoring. During RNB operation, decomposition gases generated by heating components such as cable terminals are diluted by natural convection formed by the rising of heated gases, causing sampling nodes on the return flow path to fail to accurately reflect the true degree of insulation degradation. This invention identifies and eliminates sampling nodes affected by return flow dilution interference by locating the coordinates of abnormal heating sources and determining the natural convection path and lateral flow guidance zone. It focuses on tracking the concentration change trajectory of effective sampling nodes located within the lateral flow guidance zone that are not affected by dilution, thereby extracting the true concentration rise slope. Combining the true concentration rise slope with spatial distance forms a comprehensive identification criterion for the severity of insulation faults. The early warning trigger limit is dynamically determined based on the strength of return flow dilution, achieving accurate early warning of insulation faults in the RNB and effectively avoiding missed or false alarms caused by convection dilution. Attached Figure Description

[0016] Figure 1 This is a flowchart of a method for real-time analysis and early warning of gas composition inside a ring main unit according to the present invention.

[0017] Figure 2 This is a schematic diagram of a method for real-time analysis and early warning of gas composition inside a ring main unit according to the present invention.

[0018] Figure 3 This is another schematic diagram of a method for real-time analysis and early warning of gas composition inside a ring main unit according to the present invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification. Furthermore, the ring main unit disclosed in this application is not suitable for SF6 environmentally friendly structures, using solid insulating components and air as the insulating medium, and the cabinet is not filled with insulating gases such as SF6.

[0020] like Figures 1-3 This embodiment of a method for real-time analysis and early warning of gas composition inside a ring main unit may specifically include: S101. Obtain the initial gas concentration and initial temperature distribution at multiple spatial locations inside the ring main unit. Identify local areas with significantly increased temperature inside the unit through the initial temperature distribution, locate easily heated parts such as cable terminals, and obtain the coordinates of abnormal heat sources.

[0021] Multiple spatial locations are divided vertically and horizontally within the ring main unit. Temperature sensing nodes and gas sampling nodes are deployed at cable terminals, busbar connections, and around epoxy-cast insulation components. Thermistor temperature sensors synchronously collect temperature readings from each node, while electrochemical gas sensors collect the initial gas concentrations of ozone and carbon monoxide, which serve as the baseline for subsequent determination. Based on the readings from each temperature sensing node, an initial temperature distribution is plotted along the three-dimensional coordinates within the unit. The temperature rise gradient is calculated for the temperature difference between adjacent nodes. If the temperature rise gradient exceeds a preset threshold, the area covered by the adjacent nodes is marked as a locally elevated temperature region, resulting in a set of candidate heat-generating regions. Within this set of candidate heat-generating regions, a secondary comparison is performed on the node positions at cable terminals and insulation component joints. Combining this with the absolute temperature readings of each node in the initial temperature distribution, the node position with the highest reading and the steepest surrounding temperature rise gradient is determined as the coordinates of the abnormal heat source.

[0022] It should be noted that the ring main unit has a relatively enclosed spatial structure, housing key components such as cable terminals, busbar connections, epoxy-cast insulation components, and circuit breakers. These components are prone to localized overheating during long-term operation. Overheating can trigger the pyrolysis of the solid insulation medium, releasing characteristic gases such as ozone and carbon monoxide. In one embodiment, to detect the coupling of temperature and gas anomalies in the early stages of a fault, the ring main unit is divided vertically from bottom to top into three levels: bottom, middle, and top. Horizontally, it is divided into four directional zones around the main energized components: front, rear, left, and right, thus forming a grid-like layout structure with multiple spatial locations.

[0023] For example, temperature sensing nodes use Pt100 platinum resistance thermometers or NTC thermistors, respectively attached to heat-prone locations such as the stress cone outer sheath of cable terminals, the fastening surface of busbar connection bolts, and the windward side of epoxy-cast insulation components. Three to five nodes are arranged in each ring main unit. Gas sampling nodes are integrated gas detection units, arranged one-to-one with the temperature sensing nodes, installed in the cabinet cavity 50 to 150 mm above each temperature sensing node. Each gas sampling node integrates two types of electrochemical gas sensors in parallel: one for detecting ozone concentration and the other for detecting carbon monoxide concentration. Both types of sensors share the same power supply and communication interface, and the gas sampling nodes uniformly output concentration data. During the steady-state period after the ring main unit is put into operation, each temperature sensing node synchronously collects temperature readings at its location, and each gas sampling node synchronously outputs the ozone and carbon monoxide concentrations at its location. The collected concentration data is used as the concentration baseline for subsequent judgments.

[0024] Specifically, the concentration baseline reflects the background gas level inside the ring main unit under fault-free steady-state operation. A set of data is stored for each spatial location corresponding to the concentration baseline, serving as a reference starting point for subsequent concentration change determination.

[0025] In one possible implementation, the readings of each temperature sensor node are mapped to a three-dimensional coordinate system within the cabinet according to their spatial location, and an initial temperature distribution is plotted along the vertical and horizontal directions. This initial temperature distribution presents the temperature values ​​at each spatial location within the cabinet in a gridded format, allowing the temperature difference between adjacent spatial locations to be visually apparent.

[0026] Preferably, the temperature rise gradient is calculated based on the temperature difference between adjacent nodes. The temperature rise gradient is defined as the ratio of the difference in temperature readings between two adjacent temperature sensing nodes to the spatial distance between the two nodes, and its dimension is the amount of temperature change per unit spatial distance.

[0027] For example, if the node reading at the stress cone location of the cable terminal is 65 degrees Celsius, and the reading at an adjacent node is 45 degrees Celsius, with a spatial distance of 0.1 meters between the two nodes, then the temperature rise gradient between the two nodes is 200 degrees Celsius per meter. This temperature rise gradient reflects the steepness of the temperature field inside the cabinet; a larger temperature rise gradient indicates a significant heat source concentration near that location. Furthermore, the calculated temperature rise gradients for all adjacent node pairs are compared with a preset temperature rise gradient threshold. If the temperature rise gradient exceeds the preset threshold, the spatial area covered by the adjacent nodes is marked as a localized area with a significantly increased temperature. This localized area often covers components such as cable terminals, busbar connections, and epoxy-cast insulation parts. All localized areas meeting the criteria are summarized to obtain a set of candidate heat-generating areas.

[0028] In one embodiment, the candidate heat-generating region set may simultaneously contain several locations with abnormal temperature rises. The temperature rise gradient alone is insufficient to uniquely determine the precise location of the abnormal heat source. Based on the temperature rise gradient distribution data obtained in the preceding steps, all nodes within the candidate heat-generating region set with a temperature rise gradient greater than 2°C per meter are extracted as the initial set of abnormal heat source coordinates. To this end, a secondary comparison is performed between the cable terminal node locations and the insulation joint node locations within the candidate heat-generating region set. This secondary comparison includes two aspects: firstly, reading the absolute temperature reading T of each node; secondly, checking the temperature rise gradient value G around each node. The specific fusion rule is as follows: calculate the comprehensive anomaly index S = 0.6 × T normalized + 0.4 × G normalized for the candidate nodes, where T normalized = (TT minimum) / (T maximum - T minimum), and G normalized = (GG minimum) / (G maximum - G minimum). Select the node with the largest comprehensive anomaly index S as the final abnormal heat source coordinate. When the difference in S values ​​among multiple nodes is less than 0.05, the node with the higher absolute temperature reading is preferentially selected as the abnormal heat source location.

[0029] Specifically, the temperature values ​​T of each measuring point are first sorted, and the top 10% of nodes with the highest temperatures are selected as initial candidates. Then, among these initial candidates, the comprehensive anomaly index S = 0.6 × T (standardized) + 0.4 × G (standardized) is calculated, and the node with the largest S value is selected as the final coordinate of the abnormal heat source. When the difference in S is less than 0.05, the node with the higher absolute temperature reading is selected. The specific coordinates of the selected node in the three-dimensional coordinate direction within the cabinet, i.e., the X-axis, Y-axis, and Z-axis coordinates of the node, are directly recorded as the coordinates of the abnormal heat source for subsequent heat source location analysis.

[0030] It is understandable that the coordinates of the abnormal heat source reflect both the horizontal orientation of the heat-generating part and its vertical height level, so that the positioning accuracy of easily heated parts such as cable terminals and insulation joints meets the requirements of the subsequent concentration determination process.

[0031] For example, in a typical implementation scenario, all candidate heat-generating areas are first traversed in the three-dimensional node temperature field established in the previous steps within the cabinet. For each node, the temperature rise gradient between it and its adjacent nodes is calculated. The calculation method is the temperature difference ΔT between two adjacent nodes divided by the spatial distance ΔL between the two nodes, i.e., G = ΔT / ΔL, where G is the temperature rise gradient in that direction, in degrees Celsius per meter. The maximum value of the gradient magnitude in the x, y, and z directions of that node is taken as the representative value of its surrounding temperature rise gradient. The temperature reading of the stress cone outer sheath node of a certain incoming cable terminal in the cabinet is 68 degrees Celsius. Its distance from its adjacent node is 0.05 meters, and the temperature difference is approximately 11 degrees Celsius. Therefore, the surrounding temperature rise gradient reaches 220 degrees Celsius per meter, significantly higher than the corresponding values ​​of nodes in other candidate heat-generating areas within the cabinet. The three-dimensional coordinates corresponding to this stress cone outer sheath node in the aforementioned three-dimensional temperature field are then used as the coordinates of the abnormal heat source located in this step and passed to step S106 as input parameters for subsequent processing.

[0032] S102. Based on the coordinates of the abnormal heat source, and combined with the natural convection pattern formed by the heated gas rising along the heat source and falling back along the cabinet wall after cooling, determine the scouring path of the return gas near the heat source, as well as the lateral guide zone range through which the return gas does not flow directly.

[0033] Based on the coordinates of the abnormal heat source, and considering the geometric boundaries and the arrangement of energized components within the ring main unit, a buoyancy driving condition is applied at the heat source coordinates. Computational fluid dynamics is used to solve the steady-state airflow field within the unit. The trajectory of the thermal plume formed by the upward movement of heated gas from the heat source is traced vertically, and the ascending axis of this thermal plume trajectory is used as the center reference line for subsequent flow field partitioning. The thermal plume trajectory is continued to be traced along the ascending axis, and a top obstruction area for the gas is identified on the inner wall of the top of the unit. For the streamline direction within this top obstruction area, components of the return flow that fall downwards along the unit wall and diffuse outwards are obtained. The trajectory of the return flow component is spatially compared with the horizontal projection of the abnormal heat source coordinates. If the distance between the two is less than a preset threshold, the spatial range traversed by the return trajectory is determined as the scouring path of the return gas near the heat source. In the steady-state solution of the airflow field inside the cabinet, the space occupied by the scouring path and the hot plume channel corresponding to the rising main axis are removed. For the velocity magnitude and direction of each location in the remaining space, the space region where the velocity magnitude is lower than the preset threshold and the direction of the flow deviates from the main direction of the return flow component is determined as the lateral guide zone range through which the return gas does not flow directly.

[0034] It should be noted that the ring main unit's internal structure is a relatively enclosed metal shell. Besides live components such as cable terminals, busbar connections, epoxy-cast insulation, and circuit breakers, the cabinet also includes structural components such as supporting beams, insulating partitions, observation windows, and pressure relief channels. These structural components collectively constitute the geometric boundary for airflow within the cabinet. The heat-generating areas in this design mainly refer to the areas of live components within the cabinet that generate heat due to current flow. Specifically, these include busbar contact surfaces, cable terminal connections, circuit breaker contacts, and the surfaces of current-carrying conductors. These areas can reach temperatures of 60 to 90 degrees Celsius under rated load operation, constituting the main heat source within the cabinet. Under natural convection conditions, the air inside the cabinet is heated by the heat-generating areas, creating a density difference and forming an upward-rising airflow. This airflow is obstructed upon reaching the inner top wall of the cabinet and then deflects downwards along the wall, ultimately forming a closed convection loop within the cabinet.

[0035] In one possible implementation, the coordinates of the abnormal heat source are used as input conditions for solving the flow field, and the geometric boundaries of the cabinet interior and the outlines of the energized components are used as the solid boundaries of the solution region. The solid boundaries are set to no-slip conditions, and the top inner wall, bottom plate, and side walls of the cabinet are discretized into mesh cells according to their actual geometric shapes. Based on the abnormal heat source coordinates identified in the previous step—that is, the coordinates of the center position of the component whose measured temperature rise inside the cabinet exceeds the ambient temperature by 8 degrees Celsius—a buoyancy driving condition is applied to the mesh cells corresponding to the abnormal heat source coordinates. The heat-generating area refers to the surface region of the electrical components inside the cabinet where the temperature rises abnormally, including critical areas such as circuit breaker contacts, busbar connections, and cable terminal joints. The mesh cells are obtained by dividing the three-dimensional space of the distribution cabinet into structured hexahedral meshes, with mesh sizes ranging from 5 mm to 20 mm, and densified to 2 mm to 5 mm near the heat-generating areas to improve calculation accuracy. The buoyancy driving condition uses the surface temperature of the heated part as the heat source boundary input. The relationship between the air density ρ inside the cabinet and the temperature T is approximated, i.e., ρ = ρ0 × [1 minus β × (T minus T0)], where ρ0 is the reference density, β is the volume expansion coefficient taken as 0.0033 Kelvin, and T0 is the reference temperature taken as 20℃. This causes the heated air to generate upward buoyancy due to the decrease in density. During the iteration process, the solver simultaneously solves the continuity equation, momentum equation, and energy equation until it converges when the residuals of each physical quantity in the flow field are less than 0.0001, thus obtaining the steady-state solution of the air flow field inside the cabinet.

[0036] In one embodiment, streamlines are extracted from the steady-state solution, and the streamlines are traced upwards from the coordinate position of the abnormal heat source in a vertical direction. The streamlines gradually converge above the heat-generating part to form an upward air column, which is the thermal plume trajectory of the heated gas.

[0037] For example, if the coordinates of the abnormal heat source obtained from the previous steps are located in the middle layer of the cabinet, assuming this heat source is a cable termination stress cone component, the heat plume trajectory extends from this location towards the top of the cabinet. Multiple streamlines contained in the heat plume trajectory are collected and projected onto a horizontal plane. The coordinates of each projected streamline are then calculated using a centroid-weighted average based on the temperature gradient to obtain the projection centerline. This centerline is defined as the ascending principal axis and serves as the central reference line for the flow field partition. The heat plume trajectory continues to be traced upwards along this ascending principal axis. When the streamline reaches within 0.1 meters of the inner wall of the top of the cabinet, the angle between the streamline direction vector and the vertically upward direction is monitored. When the angle exceeds 30 degrees, it is determined that the streamline direction has significantly deflected, indicating that the airflow is obstructed by the inner wall of the top. The spatial range satisfying this deflection criterion is identified as the top obstructed area.

[0038] Preferably, for each streamline within the obstructed area at the top, its flow direction after obstruction is read, and the streamline component that falls downward along the cabinet wall and the streamline component that diffuses horizontally in all directions are extracted separately. The two are collectively referred to as the return flow component. The return flow component extends downward along the inner surfaces of the left, right, front, and rear side walls of the cabinet, gradually approaching the bottom of the cabinet and the area below the heating element.

[0039] It should be noted that during the downward fall of the return flow component, its trajectory is similar to the horizontal projection position of the coordinates of the abnormal heat source to varying degrees.

[0040] Specifically, the heat-generating area refers to a localized region inside the cabinet where the temperature rises abnormally due to the operation of components, including but not limited to the surface of the transformer windings, the heat dissipation surface of the IGBT module, and the busbar connection points. The three-dimensional coordinates P0=x0,y0,z0 of its geometric center point are obtained by registering infrared thermal imaging with the three-dimensional point cloud of the cabinet in the previous step S104, and are directly used as input parameters for this step. The fall trajectory of the return current component is discretely sampled at a time step of 0.05s to obtain the trajectory point sequence Pi=xi,yi,zi, i=1,2,...,n. Its horizontal projection Qi=xi,yi is taken, and the Euclidean distance di between each Qi and the horizontal projection Q0=x0,y0 of the heat source is calculated. The minimum value dmin=mindi in the sequence is taken as the comparison value. If dmin is lower than the preset threshold D0, it is determined that the fall trajectory directly passes through the spatial range near the heat-generating area, that is, falls within the horizontal circular domain with Q0 as the center and D0 as the radius. The threshold D0 is determined based on the cabinet size and the power level of the heat source. For a conventional 0.6m×0.8m distribution cabinet, the value ranges from 0.05m to 0.15m. For every 500W increase in heat generation power, the value can be appropriately increased by 0.02m. In this embodiment, D0=0.1m is used.

[0041] In one embodiment, the heat-generating area refers to the localized region where the measured temperature rise of energized components such as busbar contact surfaces, cable terminal connection points, and circuit breaker contacts exceeds the ambient temperature by 8 degrees Celsius. Fluid simulation software is used to track the movement trajectories of particles in the return gas. When a gas particle moves away from the heat-generating area and then approaches it again, the particle's path is considered its fall trajectory. A distance threshold d is set to 0.1 meters, consistent with the aforementioned threshold D0. All fall trajectories with a closest distance to the surface of the heat-generating area less than or equal to d are counted. The spatial area occupied by all fall trajectories meeting this distance condition is summarized and determined as the scouring path of the return gas near the heat-generating area. The heat-generating area refers to the localized region where the measured temperature rise by temperature sensors exceeds the ambient temperature by 8 degrees Celsius. It is located by infrared temperature measurement arrays arranged at busbar connection points, circuit breaker contacts, and cable joints, and its range is defined by a spherical space with a radius of 100 millimeters centered on the peak temperature rise point. The gas sampling nodes are miniature suction-type sensor probes uniformly distributed on the top, side walls, and bottom of the cabinet. Six to nine probes are arranged in a single cabinet to collect the concentration of decomposed gases in real time. Based on the aforementioned determination of the heating parts and sampling arrangement, the airflow field inside the cabinet is solved by steady-state CFD using the finite volume method, with the three-dimensional structure of the cabinet and the measured boundary conditions as input. The k-epsilon two-equation model is used for the turbulence model, and the mesh is generated using tetrahedral unstructured meshing. The mesh is locally refined to 2 to 5 mm near the heating parts and sampling ports, and controlled to 10 to 15 mm in other areas, with a total mesh size of approximately 800,000 to 1,500,000. After convergence, the following two types of spaces are removed from the computational domain: First, the space occupied by the scouring path, i.e., the area covered by the return flow trajectory formed by the heated gas falling downwards along the cabinet wall after being obstructed at the top of the cabinet and flowing back along the outer side of the heating part, categorized as a velocity greater than 0.2 m / s and a negative vertical component of the direction vector within this area; second, the thermal plume channel corresponding to the rising main axis, i.e., the columnar upward airflow area formed by buoyancy directly above the heating part, categorized as a positive vertical component of velocity greater than 0.3 m / s and a horizontal offset not exceeding 30 mm. The remaining grid cells after removal constitute the screening area. For each grid cell within this screening area, its velocity magnitude and velocity direction vector are extracted as the basis for subsequent identification of concentration anomalies.

[0042] For example, if the flow velocity of a certain grid cell is lower than a preset threshold, and the angle between its flow velocity direction vector and the main direction of the return flow component exceeds a preset angle threshold, then the grid cell is determined to be in a position where the flow is relatively gentle and deviates from the main return flow.

[0043] Preferably, the velocity threshold is 0.05 m / s, and the included angle threshold is 60°. When the deviation angle α between the fluid velocity vector within a grid cell and the main return flow direction is greater than 60°, the cell is determined to be a flow deviation region. A uniform grid generation method is used, with the grid cell side length set to 0.1 meters to ensure both capturing flow field details and maintaining computational efficiency. All grid cells that meet the above criteria are aggregated, and the aggregated region is determined as the low-velocity flow deviation region where the return gas does not flow directly. Within this region, the fluid velocity is low, and the flow direction deviates significantly from the main return flow direction, forming a relatively independent flow boundary layer.

[0044] It is understandable that the air in the lateral flow guidance zone mainly migrates slowly horizontally, avoiding the direct upward channel of the thermal plume and the direct scouring of the return flow. This allows the decomposition gas released from the heating part to gradually accumulate in this range during the horizontal diffusion process, providing a key spatial positioning basis for the subsequent concentration change determination.

[0045] S103. Obtain the first sampling node located on the backflow gas flushing path and the second sampling node located in the lateral guide zone, and separate the first concentration change rate of the first sampling node and the second concentration change rate of the second sampling node from the initial gas concentration at each spatial location.

[0046] Within the spatial range covered by the flushing path, gas sampling nodes located within this range are marked as first sampling nodes, and within the lateral guide zone, gas sampling nodes located within this range are marked as second sampling nodes, thus obtaining a spatial classification of the two types of sampling nodes. For the first and second sampling nodes, initial gas concentration time-series data at each spatial location are read according to a preset sampling time interval. The time-series data is separated into two groups according to the node category. For each group of data, the concentration change rate is determined by the ratio of the concentration difference between adjacent sampling times to the sampling time interval, thus obtaining the first concentration change rate of the first sampling node and the second concentration change rate of the second sampling node.

[0047] It should be noted that among the multiple gas sampling nodes already deployed inside the ring network cabinet, the spatial assignment of each sampling node is determined according to the aforementioned flow field zoning results.

[0048] Specifically, gas sampling nodes are pre-defined in the numerical simulation grid, uniformly arranged along the return channel with a spacing of 0.2 to 0.5 meters, to monitor gas flow characteristics at different locations. For the three-dimensional coordinates (x, y, z) of each gas sampling node, the following rules are used for spatial classification: If the node coordinates satisfy |x - xc| ≤ rc and |y - yc| ≤ rc, where xc and yc are the coordinates of the scour path centerline and rc is the radius of the scour path, then the coordinates fall within the spatial range covered by the scour path, and the gas sampling node is marked as the first sampling node. If the node coordinates do not meet the above conditions, but its distance d from the scour path boundary is within the range of 0.5 to 2 meters, and the flow velocity v at that location is less than 30% of the mainstream flow velocity, then the coordinates fall within the lateral slow flow zone, and the gas sampling node is marked as the second sampling node, thus completing the spatial classification of the two types of sampling nodes.

[0049] In one possible implementation, initial gas concentration data of the spatial locations of the first and second sampling nodes are synchronously read according to a preset sampling time interval, which can be set to 1 to 10 seconds. The sampling time of the second sampling node is aligned with the sampling clock of the first sampling node, so that the two time series data correspond one-to-one at the same timestamp. For sample points where the timestamp deviation exceeds 0.1 seconds due to transmission delay, the missing values ​​are filled in by linear interpolation. The final output is a two-channel time series in the form of X1=[x1(t1), x1(t2), ..., x1(tn)] and X2=[x2(t1), x2(t2), ..., x2(tn)], where x1 and x2 represent the gas concentration values ​​of the first and second sampling nodes at time ti, respectively, and n is the total number of sampling points, which are used for subsequent concentration gradient analysis and diffusion direction determination.

[0050] For example, the sampling time interval is 5 seconds. After each sampling time interval, each gas sampling node synchronously reports the ozone concentration value and the carbon monoxide concentration value, thereby forming continuous concentration time series data.

[0051] Specifically, the concentration time-series data is separated into two sets of data corresponding to the first sampling node and the second sampling node, based on their node categories. The concentration values ​​are expressed in milligrams per cubic meter (mg / m³), and the sampling time interval is expressed in seconds (e.g., 1 second). For each set of concentration time-series data, the concentration difference Δc is obtained by subtracting the concentration values ​​read at two adjacent sampling times. The concentration difference Δc is then divided by the sampling time interval Δt to obtain the concentration change rate v corresponding to that sampling time, i.e., v = Δc / Δt, with units of milligrams per cubic meter per second. Since multiple first and second sampling nodes can be deployed within the monitoring area, let m be the number of first sampling nodes and n be the number of second sampling nodes, where m and n range from 1 to 8. The aggregation process is as follows: The arithmetic mean of the concentration change rates of all nodes at the first sampling node at the same sampling time is taken to obtain the first concentration change rate V1 at that sampling time; the arithmetic mean of the concentration change rates of all nodes at the second sampling node at the same sampling time is taken to obtain the second concentration change rate V2 at that sampling time, i.e., V1 = (v11 + v12 + ... + v1m) / m, V2 = (v21 + v22 + ... + v2n) / n. When a node has missing data or abnormal readings, the data for that node at that sampling time is removed before averaging to ensure the stability of the aggregation result.

[0052] S104. Evaluate the first and second concentration change rates, identify the first sampling node with a slower concentration increase that is affected by dilution interference from the backflow gas, and obtain the markers of the interfered nodes and the set of effective sampling nodes.

[0053] For the first and second concentration change rates, they are synchronously compared within the same time window. The values ​​of the first and second concentration change rates are read, and the rate difference is obtained by subtracting the first concentration change rate from the second rate change rate. This rate difference is used as a relative drop index of the concentration slope. This relative drop index is compared with a preset dilution interference threshold. If the relative drop index exceeds the preset dilution interference threshold, it is determined that the concentration time-series data at the first sampling node is affected by the backflow gas's scouring and dilution interference, and the corresponding first sampling node is marked as an interfered node. Based on the interfered node markings, the nodes marked as interfered are removed from all sampling nodes formed by the first and second sampling nodes. The remaining sampling nodes are summarized into a valid sampling node set, which retains the sampling nodes not affected by backflow gas dilution interference and their corresponding concentration time-series data.

[0054] It should be noted that the first rate of concentration change reflects the rate of increase in the concentration of decomposed gas at the gas sampling node located on the path of the backflow gas, while the second rate of concentration change reflects the rate of increase in the concentration of decomposed gas at the gas sampling node located in the lateral flow guide zone. After the decomposed gas is released from the abnormally heated area, the concentration rise is slow at the location being swept by the backflow gas because air without fault information is constantly passing by; while in the lateral flow guide zone, the airflow is relatively gentle, allowing the decomposed gas to gradually accumulate, resulting in a faster rate of concentration increase.

[0055] In one possible implementation, the first concentration change rate and the second concentration change rate are compared synchronously within the same time window. The time window refers to a fixed-length observation period, such as a sliding window of 30 seconds. The window slides sequentially along the time axis on the concentration time-series data, allowing the concentration change rates at the two sampling nodes to be compared within the same time period.

[0056] Specifically, within the time window, six concentration sampling points are collected at the low-speed sideflow zone and the scouring path location, respectively. A least-squares linear fit is performed on each set of sampling points according to the time series, and the slope of the fitted line represents the concentration change rate at that location within the window. The difference between the concentration change rate at the low-speed sideflow zone and the concentration change rate at the scouring path location is called the velocity difference. This velocity difference is the relative drop index of the concentration slope, reflecting the degree to which the low-speed sideflow zone leads the scouring path location in the rate of decomposition gas accumulation. The original region is named the low-speed sideflow zone because the airflow velocity in this region is lower than that in the main return channel and it is located in the lateral region of the main path, where the gas exhibits a slow lateral flow. A larger relative drop index value indicates a more significant degree of dilution interference from the return gas at the scouring path location. Furthermore, the relative drop index is compared with a preset dilution interference threshold. The process for setting the dilution interference threshold is as follows: During the factory commissioning phase or the initial operation phase of the ring main unit, steady-state background concentration data for no less than 72 hours is continuously collected. The concentration change rate of the two types of locations is calculated using the same 30-second window and 5-second sampling interval. The mean m and standard deviation s of the rate difference corresponding to all windows are statistically analyzed. The dilution interference threshold is set to m + 3 × s. In this embodiment, the threshold is calculated to be 0.02 mg / m³ / s, which reflects the allowable natural concentration fluctuation difference between the two types of locations under fault-free conditions.

[0057] In one embodiment, when the concentration unit is milligrams per cubic meter, the dilution interference threshold is set to 0.02 milligrams per cubic meter per second. This threshold is determined based on the degree of influence of backflow gas on the sampling point in a typical industrial environment, and can effectively identify abnormal concentration drops caused by airflow swirl. When the relative drop index exceeds the dilution interference threshold, it is determined that the concentration time series data at the first sampling node is affected by the backflow gas and dilution interference, and the sampling node is marked as being in an interfered state. Through the marking mechanism, subsequent data processing can identify and remove the concentration data of the interfered nodes, avoiding their impact on the accuracy of the overall analysis results.

[0058] Preferably, for the first sampling node that has not exceeded the dilution interference threshold, no interference node marker is applied, so that it continues to remain within the effective observation range.

[0059] Specifically, based on the interference node markings, nodes marked as interference are removed from all sampling nodes formed by the first and second sampling nodes. After removal, the remaining sampling nodes include: the first sampling node not marked as an interference node, and all the second sampling nodes. These remaining sampling nodes are then aggregated into a valid sampling node set. This valid sampling node set retains the spatial coordinate information and corresponding concentration time-series data of sampling nodes not affected by backflow gas dilution, ensuring that the concentration changes of the decomposition gas released from the heating site are accurately reflected.

[0060] S105. Based on the data of the first sampling node that is removed according to the interference node mark, the concentration change trajectory of the second sampling node in the set of effective sampling nodes is tracked to obtain the diffusion path of the decomposition gas.

[0061] Based on the interference node markings, the concentration records corresponding to the nodes marked as interference are removed from the concentration time-series data of the first sampling nodes. Only the concentration time-series data of the second sampling nodes in the effective sampling node set and their corresponding spatial coordinates are retained, resulting in a concentration observation data set after interference removal. For the concentration observation data set after interference removal, the concentration values ​​of each second sampling node are read in chronological order of sampling time. The second sampling nodes whose concentration values ​​exceed a preset rise threshold at the same sampling time are sorted from high to low concentration values ​​to obtain a decomposition gas concentration rise sequence. Based on the spatial coordinates of each second sampling node in the decomposition gas concentration rise sequence, the spatial coordinates of adjacent nodes are connected sequentially along the direction of concentration values ​​from high to low. The lines connecting adjacent nodes extend segment by segment as the concentration values ​​decrease, and the spatial direction traversed by the connecting lines is determined as the decomposition gas diffusion path.

[0062] It should be noted that the interference node marker reflects the fact that the concentration time series data at the first sampling node is affected by the backflow gas and dilution interference. In the concentration time series data of the first sampling node, the concentration record corresponding to the node marked with the interference node has lost its ability to truly reflect the decomposition gas released from the heating part. The concentration record is removed from the original concentration data, and only the concentration time series data of the second sampling node in the effective sampling node set and the corresponding spatial coordinates are retained. The spatial coordinates include the position value of the second sampling node in the three-dimensional coordinate direction inside the cabinet, thereby obtaining the concentration observation data set after removing interference.

[0063] In one possible implementation, the set of concentration observation data after removing interference is stored in tabular form, and each record contains the number of a second sampling node, three-dimensional spatial coordinates, sampling time, and the ozone concentration and carbon monoxide concentration values ​​collected at that sampling time.

[0064] Specifically, for the concentration observation data set after removing interference, the concentration values ​​of each second sampling node are read sequentially according to the sampling time sequence. At each sampling time, the current concentration value of each second sampling node is compared with the concentration baseline. If the difference exceeds a preset increase threshold, it is determined that the second sampling node is in a concentration increase state at that sampling time.

[0065] Preferably, the preset rise threshold is 0.01 mg / m³, and the preset rise threshold is set based on the natural fluctuation range of the concentration baseline under steady-state operation of the ring mains unit. Further, all second sampling nodes that are in a concentration rise state at the same sampling time are aggregated, and the nodes are sorted from high to low according to their current concentration values.

[0066] For example, at a certain sampling time, there are four second sampling nodes in the state of rising concentration. Arranging the four nodes in descending order of concentration value yields the concentration rise sequence of the decomposition gas at that sampling time. By sequentially forming the concentration rise sequence at each time along the sampling time axis, the distribution of high and low concentration values ​​of the decomposition gas in the cabinet space can be clearly presented.

[0067] In one embodiment, based on the spatial coordinates of each second sampling node in the rising concentration sequence of the decomposed gas, a line is pieced together segment by segment along the direction of decreasing concentration values. Specifically, the process is as follows: taking the second sampling node with the highest concentration value as the starting point, the neighborhood of this node is traversed, with a neighborhood radius of 2.0 meters. One of the second sampling nodes with the second highest concentration value in the neighborhood is selected as the next node, and the coordinates of the two nodes are connected according to a Euclidean straight line segment in three-dimensional space to form the first segment. If there is no unconnected second sampling node in the 2.0-meter neighborhood of the starting point, the search continues in the extended neighborhood from 2.0 to 3.5 meters, still prioritizing nodes with higher concentration values. If there are still no candidate nodes in the extended neighborhood, a concentration gradient transition point is inserted between the two nodes. The coordinates of the transition point are taken as the midpoint of the line connecting the two ends and offset from the geometric center of the connected node group by 0.3 times the line segment length, avoiding the connection line crossing areas without gas diffusion. This method uses a newly accessed node as the new starting point, and continues to search for the second sampling node with the second lowest concentration value in its neighborhood as the subsequent node. This process is repeated segment by segment until all second sampling nodes in the concentration increase sequence are accessed or the spatial distance between the remaining nodes exceeds an extended neighborhood of 3.5 meters, forming a complete line trajectory that fits the diffusion direction. Adjacent nodes must simultaneously meet two conditions: first, they must be sequentially adjacent in the concentration increase sequence; second, their spatial Euclidean distance must not exceed 3.5 meters. Both conditions are indispensable.

[0068] It is understandable that the lines connecting adjacent nodes extend segment by segment as the concentration decreases, and the overall direction of the lines reflects the spatial distribution characteristic of the decomposition gas concentration gradually decreasing from the source location to the distant location. The spatial direction traversed by the lines is defined as the decomposition gas diffusion path. The decomposition gas diffusion path exists in the form of a broken line segment in the three-dimensional coordinate direction within the cabinet, with its first end close to the second sampling node with the highest concentration value and its end located at the second sampling node with the lower concentration value.

[0069] For example, in a typical implementation scenario, insulation pyrolysis occurs near the terminal of an incoming cable inside the cabinet. The decomposition gas rises through the thermal plume and accumulates in the lateral flow guide zone. The second sampling node with the highest concentration is located in the middle layer above the cable terminal, the second highest node is located in the middle layer behind the cabinet, and the node with the next lowest concentration is located in the bottom layer behind the cabinet. The broken line segment formed by connecting the spatial coordinates of the three second sampling nodes in sequence is the diffusion path of the decomposition gas in this fault.

[0070] S106. Extract the true concentration rise slope by decomposing the gas diffusion path, and the spatial distance between the effective sampling node set and the abnormal heat source. Integrate the true concentration rise slope and spatial distance into a comprehensive basis for identifying the severity of insulation faults.

[0071] According to the diffusion path of the decomposed gas, concentration time-series data is sequentially read from the second sampling node with the highest concentration value to subsequent nodes along the diffusion path. For each second sampling node, the concentration rise slope is determined by the ratio of the concentration difference between adjacent sampling times to the sampling time interval. The concentration rise slopes of all second sampling nodes along the diffusion path are summed, and the maximum value is taken as the true concentration rise slope. For the spatial coordinates of each second sampling node in the effective sampling node set, the distance between the effective sampling node set and the abnormal heat source is calculated using the three-dimensional Euclidean distance formula, obtaining the spatial distance between the effective sampling node set and the abnormal heat source. The true concentration rise slope and the spatial distance are weighted and summed, with the true concentration rise slope assigned a positive weight and the spatial distance assigned a negative weight. The weighted sum is used as the comprehensive identification basis for the severity of insulation faults.

[0072] The heating site refers to a localized area inside the ring main unit where the temperature rises abnormally due to insulation degradation, increased contact resistance, or partial discharge, accompanied by pyrolysis of the solid insulation medium. This area is typically located at busbar joints, circuit breaker moving and stationary contacts, disconnector contacts, cable joints, and bushing roots. The spatial range is generally a region with a radius of 2 to 5 centimeters centered on the fault source. Identifying characteristics include a surface temperature at least 15 degrees Celsius higher than the surrounding normal area, and the continuous release of ozone and carbon monoxide decomposition products that can be captured by electrochemical gas sensors. The decomposition gas diffusion path reflects the actual spatial orientation of the decomposition gases released from the heating site within the ring main unit. Concentration time-series data unaffected by backflow dilution can be obtained from the second sampling node along the diffusion path. Concentration time-series data for each second sampling node is sequentially read from the second sampling node with the highest concentration value towards subsequent nodes along the diffusion path.

[0073] Specifically, for each second sampling node, the concentration rise slope of that node is determined by the ratio of the concentration difference between adjacent sampling times to the sampling time interval.

[0074] In one embodiment, the sampling time interval is 5 seconds. If the ozone concentration value collected by a second sampling node at an adjacent sampling time increases from 0.04 mg / m³ to 0.06 mg / m³, then the concentration rise slope of that node is 0.004 mg / m³ / s. Further, the concentration rise slopes corresponding to each second sampling node along the diffusion path are summarized, and the concentration rise slope values ​​of each node are compared. The concentration rise slope with the largest value is taken as the true concentration rise slope. The true concentration rise slope reflects the actual accumulation rate of the decomposition gas released from the heating site after it has moved away from the backflow dilution interference.

[0075] In one possible implementation, the effective sampling node set is derived from the second sampling nodes retained after concentration rise slope filtering in step S105, and the coordinates of the abnormal heat source are obtained through reverse tracing positioning in step S104. For each second sampling node in the effective sampling node set, its spatial coordinates are extracted, and the distance between its coordinates and the coordinates of the abnormal heat source is calculated using the three-dimensional Euclidean distance formula, which is: Where d represents the spatial distance, x1, y1, and z1 are the three-dimensional coordinate components of the second sampling node, and x0, y0, and z0 are the three-dimensional coordinate components of the abnormal heat source. After traversing and calculating all nodes in the set of effective sampling nodes, a distance array is formed. This distance array contains the spatial distance values ​​between each second sampling node and the abnormal heat source, which are used for subsequent diffusion rate analysis. In another implementation, Manhattan distance or Chebyshev distance can be used for calculation to adapt to the calculation requirements of different spatial grid layouts.

[0076] It is understandable that the actual concentration rise slope and the spatial distance reflect the severity of the insulation fault from two different dimensions: the faster the concentration rises, the more intense the release of decomposition gases, and the closer the spatial distance, the tighter the coupling between the heating location and the gas accumulation location.

[0077] Specifically, the weighted summation is performed by the actual concentration rise slope and the spatial distance. The integrated relationship of the weighted summation is expressed as S = a × kb × d, where S is the value of the comprehensive identification basis for the severity of insulation faults, k is the actual concentration rise slope in milligrams per cubic meter per second, d is the spatial distance in meters, a ranges from 0.6 to 0.8, b ranges from 0.2 to 0.4, and satisfies a + b = 1, with typical values ​​of a = 0.7 and b = 0.3.

[0078] Preferably, the weighting coefficients a and b are pre-calibrated based on the steady-state operating data collected during the factory commissioning phase or the initial operation phase of the ring main unit. Specifically, the calibration method is as follows: during steady-state operation of the ring main unit, different degrees of temperature rise are artificially applied at the known heat source location, and the gas concentration change rate detected by each gas sensor and the distance between each sensor and the heat source are recorded simultaneously to obtain multiple sets of concentration change rate data, distance data, and corresponding comprehensive identification criteria values. The least squares method is used to fit the above multiple sets of data to solve for the values ​​of weighting coefficients a and b that minimize the fitting error. After calibration, weighting coefficient a is set to a value between 0.6 and 0.8, and weighting coefficient b is set to a value between 0.2 and 0.4, satisfying that a plus b equals 1, so that the faster the concentration rises and the closer to the heat source, the larger the comprehensive identification criteria value.

[0079] In one embodiment, the normalized form S = a·k′ is used. b·d′, where the weighting coefficients a and b are dimensionless, with a=0.7 and b=0.3, and a+b=1. The value obtained from the above weighted summation process is used as the comprehensive identification criterion for the severity of insulation faults. This comprehensive identification criterion integrates information from two dimensions—concentration change rate and spatial distance—into a numerical judgment criterion, enabling a unified numerical expression of the severity of insulation faults corresponding to heating sites.

[0080] S107. Based on the comprehensive identification criteria of the severity of insulation faults, determine the early warning triggering limit that is compatible with the strength of return current dilution, assess the internal operating conditions of the ring main unit, obtain the final fault early warning result, and output the final fault early warning result to the terminal equipment.

[0081] Based on the comprehensive identification criteria for the severity of the insulation fault, the proportion of the affected node marker in the first and second sampling nodes is read. This proportion is used as a quantitative indicator of the backflow dilution strength. The preset warning trigger threshold benchmark value is corrected based on this quantitative indicator; the larger the quantitative indicator, the greater the downward adjustment of the warning trigger threshold benchmark value, resulting in a warning trigger threshold adapted to the backflow dilution strength. The value of the comprehensive identification criteria for the severity of the insulation fault is compared step-by-step with the warning trigger threshold. If the value exceeds the first-level threshold of the warning trigger threshold, the internal operating condition of the ring main unit is determined to be at the concern level; if the value further exceeds the second-level threshold, the internal operating condition of the ring main unit is determined to be at the alarm level, resulting in the final fault warning result. The final fault warning result is encapsulated into an alarm signal containing the coordinates of the heating location, the warning level, and the concentration rise slope according to a preset message format. The alarm signal is then output to the terminal device via the communication link between the ring main unit and the monitoring terminal.

[0082] The comprehensive identification criterion for the severity of the insulation fault is a numerical value derived from the fusion of two parameters: the rate of concentration change obtained previously and the spatial distance from the sensor to the suspected discharge point. This value is denoted as S, and its calculation method follows the formula S=a·k′ determined in step S106. b·d′, where k′ is the normalized true concentration rise slope, d′ is the normalized spatial distance, and a and b are the dimensionless weighting coefficients calibrated in step S106 (a=0.7, b=0.3). The larger the value of S, the higher the probability of insulation abnormality inside the ring main unit. The comparability of S will change under different return current dilution conditions, so the warning trigger limit cannot use a fixed value, but should be corrected according to the degree of return current dilution. The warning trigger limit is divided into two levels: a primary benchmark value and a secondary benchmark value. The primary benchmark value is 5.0, and the secondary benchmark value is 8.0. The actual limit is obtained by multiplying the benchmark value by the correction factor. The correction factor decreases as the return current dilution quantification index R increases, and the value range is 0.6 to 1.0, that is, the stronger the return current dilution, the greater the downward adjustment of the limit. When the S calculated in real time exceeds the corrected primary or secondary limit, a warning at the attention level or alarm level is triggered respectively.

[0083] In one possible implementation, regarding the marked interference nodes, the first and second sampling nodes are judged one by one: according to the criterion determined in step S104, when the relative drop index of a certain first sampling node exceeds the dilution interference threshold of 0.02 mg / m³ / s, the node is marked as an interference node. The number of nodes marked as interference, N1, is counted, and the total number N of all sampling nodes is read. The ratio R = N1 / N is used as a quantitative indicator of the strength of backflow dilution, where R is the proportion of interference nodes, N1 is the number of interference nodes, and N is the total number of sampling nodes. When R is less than 0.2, it is judged as weak dilution; when R is between 0.2 and 0.5, it is judged as moderate dilution; when R is greater than 0.5, it is judged as strong dilution. The judgment result is output to the subsequent backflow control stage as the basis for adjusting the backflow parameters.

[0084] For example, if there are 20 sampling nodes in total, and 8 nodes are marked as disturbed, then the quantification index is set to 0.4. A higher quantification index indicates a more significant backflow dilution interference and a greater proportion of the faulty gas being flushed away.

[0085] Specifically, the preset warning trigger threshold benchmark value is corrected based on the quantitative indicators. The warning trigger threshold benchmark value is preset based on steady-state operating data collected during the factory commissioning phase of the ring main unit, and is divided into two levels: a primary benchmark value and a secondary benchmark value. The correction process adopts a linear downward adjustment method, multiplying the benchmark value by a correction factor determined by the quantitative indicators. The correction factor decreases as the quantitative indicators increase, so that the warning trigger threshold shifts downward accordingly when the backflow dilution is stronger.

[0086] Preferably, the primary benchmark value is 5.0, the secondary benchmark value is 8.0, and the correction factor ranges from 0.6 to 1.0. Further, the two values ​​obtained after correction are used as the warning trigger limits, denoted as the primary limit and the secondary limit, respectively. The primary limit corresponds to the threshold for determining mild anomalies, and the secondary limit corresponds to the threshold for determining significant anomalies.

[0087] In one embodiment, the numerical value used to comprehensively identify the severity of the insulation fault is compared step-by-step with the warning trigger limits. If the value exceeds the first-level limit of the warning trigger limits, the operating condition inside the ring main unit is determined to be at the concern level, indicating that the increase in the concentration of decomposition gas released from a certain heat-generating part inside the unit has become a coupling sign with the heat-generating location, requiring continuous observation. If the value further exceeds the second-level limit of the warning trigger limits, the operating condition inside the ring main unit is determined to be at the alarm level, indicating that the insulation pyrolysis process inside the unit has reached a level that cannot be ignored.

[0088] It is understood that the attention level and the alarm level together constitute the final fault warning result. The final fault warning result is recorded in the monitoring database in the form of a warning level label and a trigger timestamp.

[0089] Specifically, the heat-generating areas refer to key component areas inside the ring main unit that may experience abnormal temperature rises due to poor contact, insulation aging, or arc discharge. These include circuit breaker contacts, busbar connection points, cable terminal joints, and the roots of insulating bushings, located in the cabinet's three-dimensional coordinate system using x, y, and z axes. The coordinates of the abnormal heat source are taken from the three-dimensional coordinates of the highest temperature point output by the infrared thermal imager positioning module in the previous step. The actual concentration rise slope is taken from the unit-time concentration change rate obtained by least-squares fitting of the ozone and carbon monoxide concentration time-series data in the previous step, with units of milligrams per cubic meter per second. Based on the final fault warning result, the warning level, the coordinates of the abnormal heat source, the actual concentration rise slope, and the trigger timestamp are filled into the corresponding byte positions of the message according to the preset message format, encapsulating them into an alarm signal containing the coordinates of the abnormal heat source, the warning level, and the concentration rise slope. The message format adopts a frame structure commonly used in fieldbus communication, including a 2-byte frame header, an 18-byte data area, and a 2-byte frame trailer. The data area is arranged in the following order: 1 byte for the warning level, 3 bytes for the x-coordinate, 3 bytes for the y-coordinate, 3 bytes for the z-coordinate, 4 bytes for the concentration rise slope, and 4 bytes for the trigger timestamp. Based on the communication link between the ring network cabinet and the monitoring terminal, the alarm signal is transmitted to the terminal device via an optical fiber communication channel or a wireless communication channel. The terminal device includes a human-machine interface workstation in the substation control center and a mobile terminal held by maintenance and inspection personnel. Upon receiving the alarm signal, the terminal device presents the alarm entry in different colors and with different audible alerts according to the warning level, and illuminates the corresponding marker on the cabinet's 3D schematic diagram based on the coordinates of the abnormal heat source.

[0090] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for real-time analysis and early warning of gas composition inside a ring main unit, characterized in that, The method includes: The initial gas concentration and initial temperature distribution at multiple spatial locations inside the ring main unit are obtained. The local areas with rising temperature inside the unit are identified by the initial temperature distribution, the heat source is located, and the coordinates of the abnormal heat source are obtained. Based on the coordinates of the abnormal heat source, and combined with the natural convection pattern formed by the heated gas rising along the heat source and falling back along the cabinet wall after cooling, the scouring path of the return gas near the heat source and the range of the lateral guide zone that the return gas does not flow directly through are determined. The first sampling node located on the backflow gas flushing path and the second sampling node located in the lateral guide zone are obtained. The first concentration change rate of the first sampling node and the second concentration change rate of the second sampling node are separated from the initial gas concentration at each spatial location. The rates of change of the first and second concentrations were evaluated, and the first sampling node with a slower concentration increase was identified as being affected by dilution interference from the backflow gas. The disturbed node was marked and the set of effective sampling nodes was obtained. Based on the data of the first sampling node that is removed by the interference node marker, the concentration change trajectory of the second sampling node in the set of valid sampling nodes is tracked to obtain the diffusion path of the decomposition gas. By decomposing the gas diffusion path to extract the true concentration rise slope and the spatial distance between the effective sampling node set and the abnormal heat source, the true concentration rise slope and spatial distance are integrated into a comprehensive basis for identifying the severity of insulation faults. The warning triggering threshold is determined by comprehensively identifying the severity of insulation faults and matching the strength of return current dilution. The internal operating conditions of the ring main unit are assessed to obtain the final fault warning result, which is then output to the terminal equipment.

2. The method for real-time analysis and early warning of gas composition inside a ring main unit according to claim 1, characterized in that, The acquisition of the initial gas concentration and initial temperature distribution at multiple spatial locations inside the ring main unit includes: Multiple spatial locations are divided along the vertical and horizontal directions inside the ring main unit. Temperature sensing nodes and gas sampling nodes are arranged around the cable terminals, busbar connections, and epoxy-cast insulation components. Thermistor temperature measuring elements are used to synchronously collect temperature readings at each node. Electrochemical gas sensors are used to collect the initial gas concentrations of ozone and carbon monoxide characteristic gases. The initial gas concentrations are used as the concentration baseline for subsequent determination.

3. The method for real-time analysis and early warning of gas composition inside a ring main unit according to claim 1, characterized in that, The process of identifying localized areas of temperature rise within the cabinet through initial temperature distribution, locating the heat source, and obtaining the coordinates of the abnormal heat source includes: The initial temperature distribution is plotted along the three-dimensional coordinate direction inside the cabinet based on the readings of each temperature sensing node. The temperature rise gradient is calculated for the temperature difference between adjacent nodes. The area covered by adjacent nodes whose temperature rise gradient exceeds a preset threshold is marked as a set of candidate heat-generating areas. The location of the node with the highest reading and the steepest surrounding temperature rise gradient is determined as the coordinates of the abnormal heat source.

4. The method for real-time analysis and early warning of gas composition inside a ring main unit according to claim 1, characterized in that, Based on the coordinates of the abnormal heat source, and combined with the natural convection pattern of heated gas rising along the heat source and falling back along the cabinet wall after cooling, the scouring path of the return gas near the heat source and the lateral guidance zone range through which the return gas does not flow directly are determined, including: A buoyancy driving condition is applied at the coordinate position of the abnormal heat source. The steady-state solution of the air flow field inside the cabinet is obtained by computational fluid dynamics. The trajectory of the hot plume formed by the upward floating of the heated gas from the heat source is traced in the vertical direction. The upward main axis of the hot plume trajectory is used as the center reference line of the flow field partition. Along the rising main axis, identify the gas top obstruction area on the inner wall of the cabinet top, obtain the return flow components that fall down along the cabinet wall and diffuse in all directions, spatially compare the fall trajectory of the return flow components with the horizontal projection of the coordinates of the abnormal heat source, and determine the spatial range traversed by the fall trajectory as the flushing path of the return gas near the heat source. Within the remaining space, the spatial region where the flow velocity is lower than a preset threshold and the flow direction deviates from the main direction of the return flow component is defined as the lateral flow guidance zone.

5. The method for real-time analysis and early warning of gas composition inside a ring main unit according to claim 1, characterized in that, The acquisition of the first sampling node located on the reflux gas flushing path and the second sampling node located in the lateral guide zone includes: Within the space covered by the return gas flushing path, the gas sampling nodes within that range are marked as first sampling nodes, and within the lateral guide zone, the gas sampling nodes within that range are marked as second sampling nodes. The initial gas concentration time series data of each spatial location are read according to a preset sampling time interval, and the concentration change rate is determined by the ratio of the concentration difference between adjacent sampling times to the sampling time interval.

6. The method for real-time analysis and early warning of gas composition inside a ring main unit according to claim 1, characterized in that, The first sampling node, whose concentration rises slowly, is identified as being affected by dilution interference from the backflow gas. This process yields a marker for the affected node and a set of valid sampling nodes, including: The first concentration change rate and the second concentration change rate are synchronously compared within the same time window. The first concentration change rate is subtracted from the second concentration change rate to obtain the rate difference. The rate difference is compared with a preset dilution interference threshold. The corresponding first sampling node that exceeds the threshold is marked as an interfered node. The remaining sampling nodes are summarized into a set of valid sampling nodes.

7. The method for real-time analysis and early warning of gas composition inside a ring main unit according to claim 1, characterized in that, The step of removing the data from the first sampling node based on the interference node marker, tracing the concentration change trajectory of the second sampling node in the set of valid sampling nodes, and obtaining the diffusion path of the decomposition gas includes: Based on the interference node marking, the concentration records corresponding to the nodes marked as interference are removed from the concentration time series data of the first sampling node, and the concentration time series data and corresponding spatial coordinates of the second sampling node in the set of valid sampling nodes are retained to obtain the concentration observation data set after interference removal. The concentration values ​​of each second sampling node are read in the order of sampling time. The second sampling nodes whose concentration values ​​exceed the preset rise threshold at the same sampling time are sorted from high to low according to the concentration values ​​to obtain the decomposition gas concentration rise sequence. The spatial coordinates of adjacent nodes in the sequence of increasing concentration of the decomposition gas are connected sequentially along the direction of decreasing concentration values, and the spatial direction through which the connecting lines pass is determined as the diffusion path of the decomposition gas.

8. The method for real-time analysis and early warning of gas composition inside a ring main unit according to claim 1, characterized in that, The method involves extracting the true concentration rise slope by decomposing the gas diffusion path, and the spatial distance between the effective sampling node set and the abnormal heat source. Integrating the true concentration rise slope and spatial distance into a comprehensive basis for identifying the severity of insulation faults includes: Concentration time-series data are sequentially read from the second sampling node with the highest concentration value to subsequent nodes along the diffusion path of the decomposed gas. The concentration rise slope of each node is determined by the ratio of the concentration difference between adjacent sampling times to the sampling time interval, and the maximum value is taken as the true concentration rise slope. The spatial distance is calculated based on the Euclidean distance formula in three-dimensional space. The slope of the actual concentration increase is assigned a positive weight, and the spatial distance is assigned a negative weight. The weighted sum is then used as a comprehensive basis for identifying the severity of insulation faults.

9. The method for real-time analysis and early warning of gas composition inside a ring main unit according to claim 1, characterized in that, The method of determining the early warning trigger threshold that is compatible with the strength of return current dilution based on a comprehensive identification criterion of insulation fault severity includes: The proportion of the interfered node marker in the first sampling node and the second sampling node is read, and the proportion is used as a quantitative indicator of the backflow dilution strength. The preset warning trigger limit benchmark value is corrected according to the quantitative indicator, and the downward adjustment range of the warning trigger limit benchmark value is determined according to the quantitative indicator.

10. A method for real-time analysis and early warning of gas composition inside a ring main unit according to claim 1, characterized in that, The assessment of the internal operating conditions of the ring main unit, obtaining the final fault warning result, and outputting the final fault warning result to the terminal equipment includes: The numerical values ​​of the comprehensive identification criteria for the severity of the insulation fault are compared step by step with the warning triggering limits. If the level exceeds the first limit, the internal operating condition of the ring main unit is determined to be at the attention level. If the level exceeds the second limit, it is determined to be at the alarm level. The final fault warning result is obtained and encapsulated into an alarm signal containing the coordinates of the heated part, the warning level and the concentration rise slope according to the preset message format. The signal is then output to the terminal device through the communication link between the ring main unit and the monitoring terminal.

Images