Power Fault Collaborative Sensing and Precise Location System Based on Data Fusion

Abstract

This invention relates to the field of power fault monitoring technology, providing a power fault collaborative perception and precise location system based on data fusion. This system integrates topology and terrain risk to generate fault feature maps offline, serving as a judgment knowledge base. When a power fault occurs, each node completes a preliminary diagnosis on-site based on the map; an evidence network is constructed to coordinate the conclusions of each node with meteorological priors, forming a reliable global fault segment determination; within the locked segment, precise fault location is achieved based on traveling wave feature matching. This invention forms a closed-loop process from risk modeling and map pre-generation to rapid on-site judgment by monitoring nodes, ultimately achieving precise fault location. This significantly improves the reliability, speed, and accuracy of fault location in complex mountainous distribution networks with a high proportion of distributed power sources in weak communication environments.

Description

Technical Field

[0001] This invention relates to the field of power fault monitoring technology, and in particular to a power fault collaborative sensing and precise location system based on data fusion. Background Technology

[0002] With the advancement of new power system construction, the high proportion of distributed power sources connected to mountainous distribution networks, coupled with the weak communication environment caused by complex terrain, poses a severe challenge to fault location. Traditional fault location methods that rely on fixed criteria and centralized processing are ill-suited to the complex challenges of disordered fault current characteristics and incomplete and asynchronous monitoring data under communication constraints after the connection of multiple distributed power sources. This leads to location failures or delays, seriously affecting power restoration and system security.

[0003] Existing technologies primarily focus on optimizing single data sources or algorithms, failing to systematically coordinate multi-dimensional heterogeneous data such as electrical quantities, equipment status, geographical topology, and meteorological environment. They also lack mechanisms for achieving rapid local assessment and efficient global evidence fusion under weak communication constraints, thus failing to meet the urgent needs of complex mountainous distribution networks for highly reliable and adaptive fault location. Therefore, there is a pressing need for a method and system that can adapt to the weak communication environment of mountainous areas, deeply integrate multi-source heterogeneous data, and achieve precise fault location through collaborative sensing and evidence fusion, thereby enhancing the fault self-healing capability and operational resilience of complex mountainous distribution networks. Summary of the Invention

[0004] To overcome the defects and shortcomings of existing technologies, this invention provides a power fault collaborative sensing and precise location system based on data fusion.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] In a first aspect, the present invention provides a power fault collaborative sensing and precise location system based on data fusion, comprising the following modules:

[0007] Multi-source acquisition module: Simultaneously acquires electrical monitoring data, equipment status data, meteorological and environmental monitoring data, and power grid topology and terrain coupling data of the target power grid;

[0008] Risk modeling module: Constructs a dynamic topology model based on coupled power grid topology and terrain data, and generates a fault theoretical response map through terrain risk weighted simulation;

[0009] Collaborative analysis module: Each monitoring node extracts theoretical response features from the fault theoretical response spectrum, integrates electrical monitoring data and equipment status data, and generates local fault diagnosis conclusions through feature matching and credibility weighted analysis;

[0010] Evidence fusion module: Constructs an evidence network based on the local fault diagnosis conclusions and dynamic topology model of each monitoring node; and uses meteorological and environmental monitoring data as prior evidence to generate fault section determination results by coordinating conflicts of multi-source evidence.

[0011] Location execution module: Based on the fault section determination results, extract the theoretical traveling wave characteristics of the fault section, and realize the power fault location of the target power grid through traveling wave feature matching.

[0012] According to the above technical solution, the steps of synchronously collecting electrical monitoring data, equipment status data, meteorological and environmental monitoring data, and power grid topology and terrain coupling data of the target power grid include:

[0013] S110. By collecting distribution monitoring data from each monitoring node of the target power grid, construct electrical monitoring data that includes three-phase current waveform data and high-frequency traveling wave transient data, specifically including:

[0014] By deploying distribution automation terminals at each monitoring node of the target power grid, three-phase current waveform data reflecting the changes in the path and amplitude of fault current are collected using built-in current transformers; high-frequency traveling wave transient data at the time of power faults are simultaneously collected through the transient waveform recording module integrated in the distribution automation terminal.

[0015] S120. By collecting equipment monitoring data of power equipment within the target power grid, construct equipment status data that includes equipment temperature data and equipment partial discharge data, specifically including:

[0016] Wireless temperature measurement units deployed at each monitoring node collect equipment temperature data reflecting the thermal state of the equipment connection points and the equipment body; ultra-high frequency partial discharge monitoring devices deployed synchronously at the monitoring nodes collect equipment partial discharge data reflecting internal insulation defects.

[0017] S130. By collecting external environmental monitoring data of the target power grid area, construct meteorological and environmental monitoring data that includes lightning location data and icing monitoring data, specifically including:

[0018] By connecting to a lightning monitoring system covering the target power grid area, lightning location data with accurate time, geographical coordinates, and lightning current intensity information is collected; by deploying online line icing monitoring devices on transmission lines, icing monitoring data with line location, icing thickness, and growth rate information is collected.

[0019] S140. By calling the power grid dispatching system and geographic information system of the target power grid area, construct real-time power grid topology data and distributed power generation operation data reflecting the real-time operating status of the power grid, as well as terrain slope data reflecting geographical environmental risks, specifically including:

[0020] By calling the data interfaces of the power grid dispatching system and geographic information system of the target power grid area, two types of data are collected synchronously: one is real-time topology data of the power grid, including the physical equipment list of the distribution network, the real-time connection relationship between physical equipment, and the switching status. In the physical equipment list, a unique monitoring node identifier is defined for each distribution monitoring terminal (i.e., monitoring node) and associated with the physical equipment installed on the distribution monitoring terminal. The second is distributed power generation operation data, which reflects the real-time output and switching status of distributed power generation, collected according to the equipment identifier of the power generation grid connection point. By accessing the geographic information system and based on the spatial path coordinates of the target power grid line corridor, terrain slope data, including spatial coordinate range and corresponding average slope value, is extracted with line segments as the basic unit.

[0021] According to the above technical solution, the steps of constructing a dynamic topology model based on coupled power grid topology and terrain data, and generating a fault theoretical response map through terrain risk-weighted simulation include:

[0022] S210. By analyzing the real-time power grid topology data and distributed power generation operation data in the power grid topology and terrain coupling data, a dynamic topology model is constructed, specifically including:

[0023] Based on the list of physical devices and their connections in the real-time topology data of the power grid, each physical node in the distribution network is instantiated as a vertex in the dynamic topology model, and each line connecting the physical node is instantiated as an edge connecting two corresponding vertices, thereby generating the basic dynamic topology model skeleton.

[0024] Based on the continuously acquired switch opening and closing status from the real-time power grid topology data, the corresponding edges are dynamically enabled or disabled to update the electrical connectivity between vertices. Additionally, distributed generation operation data is read, and the corresponding distributed generation vertex is located in the dynamic topology model based on the device identifier of the power generation grid connection point; the real-time output value and switching status of the distributed generation are then written into the attributes of that distributed generation vertex.

[0025] S220. Based on terrain slope data, add geographical risk labels to the dynamic topology model, specifically including:

[0026] The terrain slope data is read, which is divided into segments and records the spatial coordinate range and corresponding average slope value of each segment. Each edge in the dynamic topology model is traversed, and a matching slope value is searched in the terrain slope data based on the spatial coordinates of the associated line identifier. If the found slope value is greater than a preset high slope threshold, a high slope risk label is added to the edge.

[0027] S230. Based on the dynamic topology model, perform power system transient simulation, and differentiate the simulation parameters based on geographical risk labels. Output theoretical electrical response data for each monitoring node, including theoretical data of three-phase current waveforms and theoretical data of high-frequency traveling wave transients, specifically including:

[0028] In step S220, the edges marked with high slope risk labels in the dynamic topology model are used as one type of virtual fault point, and a single-phase ground fault is injected. At the same time, the vertices directly adjacent to these high-risk edges are used as another type of virtual fault point, and a three-phase short-circuit fault is injected.

[0029] Based on the current dynamic topology model skeleton and distributed power source vertex attributes, power system transient simulation is performed for each virtual fault point. The electrical response data of each vertex in the dynamic topology model under the fault scenario (i.e., the simulation conditions jointly determined by the virtual fault point and the corresponding fault type) are derived. On this basis, the electrical response data of each vertex corresponding to the monitoring node is extracted as the theoretical electrical response data of each monitoring node under the fault scenario. During the calculation, the electrical parameters of the line segment where the virtual fault point is located are corrected by terrain weighting based on the terrain slope data associated with the virtual fault point. The simulation is completed based on the corrected electrical response parameters, and the theoretical electrical response data of each monitoring node under each virtual fault simulation, including the theoretical data of three-phase current waveform and the theoretical data of high-frequency traveling wave transient, are output.

[0030] S240. Integrate the theoretical electrical response data of each monitoring node under power system transient simulation to generate a fault theoretical response map, specifically including:

[0031] Using the monitoring node identifier as the main index, the theoretical electrical response data of the monitoring node under different fault scenarios are aggregated to form a fault theoretical response map centered on the monitoring node that supports fast retrieval and matching.

[0032] According to the above technical solution, the steps for each monitoring node to extract theoretical response features from the fault theoretical response spectrum, integrate electrical monitoring data and equipment status data, and generate local fault diagnosis conclusions through feature matching and confidence weighted analysis include:

[0033] S310. Each monitoring node retrieves and obtains the theoretical response characteristic data corresponding to its own monitoring node from the fault theoretical response map, specifically including:

[0034] Each monitoring node obtains the corresponding theoretical response characteristic data from the fault theoretical response spectrum based on its own monitoring node identifier; the theoretical response characteristic data refers to the three-phase current waveform theoretical data and high-frequency traveling wave transient theoretical data corresponding to the monitoring node stored in the fault theoretical response spectrum.

[0035] S320. The waveform similarity of the three-phase current waveform data in the electrical monitoring data and the theoretical three-phase current waveform data in the theoretical response characteristic data is calculated to obtain a steady-state matching index; the propagation timing and waveform consistency of the high-frequency traveling wave transient data and the high-frequency traveling wave transient theoretical data are calculated to obtain a transient matching index; the steady-state matching index and the transient matching index are fused differentially based on the fault electrical characteristics represented by the theoretical electrical response data to generate a comprehensive matching score corresponding to each theoretical electrical response data, and the comprehensive matching score is used for sorting and filtering to generate a preliminary fault hypothesis sequence, specifically including:

[0036] The steady-state matching index is obtained by calculating the waveform similarity after time-series alignment of the three-phase current waveform data in the electrical monitoring data with the theoretical three-phase current waveform data of each monitoring node. The transient matching index is obtained by calculating the time difference of arrival of the high-frequency traveling wave transient data in the electrical monitoring data with the theoretical high-frequency traveling wave transient data of each monitoring node and matching the waveform correlation. The steady-state matching index and the transient matching index are then fused to generate a comprehensive matching score for each fault scenario. The fusion calculation assigns fusion weights to the steady-state matching index and the transient matching index based on the fault type indicated by the fault scenario; for short-circuit faults, the steady-state matching index is emphasized; for ground faults, the transient matching index is emphasized.

[0037] Based on the comprehensive matching score, all associated fault scenarios in the fault theory response map are sorted in descending order. Fault scenarios with scores higher than the set matching threshold are selected. The virtual fault point, fault type and corresponding comprehensive matching score of the fault scenario are used as output items to form a preliminary fault hypothesis sequence.

[0038] S330: Each monitoring node retrieves device status data and, in conjunction with associated distributed power supply operation data, performs a confidence-weighted correction on the initial fault hypothesis sequence to generate a local fault diagnosis conclusion, specifically including:

[0039] Each monitoring node locates the topology vertex corresponding to the virtual fault point in the dynamic topology model based on the virtual fault points assumed in the preliminary fault hypothesis sequence, and queries the distributed power source vertex directly electrically associated with that vertex in the network topology, and reads the real-time output value and switching status flag in the attributes of that distributed power source vertex.

[0040] The system calls upon the equipment temperature data and partial discharge data of key monitoring vertices to calculate a basic reliability factor characterizing the reliability of the equipment's own state. Simultaneously, it calculates the output fluctuation rate based on the real-time output values ​​of the associated distributed power sources. Based on the output fluctuation rate and switching status, it calculates the power source influence weight: when the distributed power source is in grid-connected state, the power source influence weight is positively correlated with the output fluctuation rate; the greater the fluctuation rate, the lower the weight value (indicating that the power source operation has a greater interference with monitoring). When the distributed power source is in off-grid state, it is assigned a fixed high weight.

[0041] The basic credibility factor is multiplied by the power supply influence weight to obtain a device observation weight coefficient in the range of 0 to 1; the device observation weight coefficient is used to correct the corresponding comprehensive matching score to obtain the credibility score of the fused device and power supply status.

[0042] All fault hypotheses are reordered based on the revised confidence scores. The top-ranked hypotheses, along with their fault locations, fault types, and confidence scores, are then structured and encapsulated to generate and upload the local fault diagnosis conclusions for this monitoring node.

[0043] According to the above technical solution, the steps of constructing an evidence network based on the local fault diagnosis conclusions and dynamic topology model of each monitoring node, and using meteorological and environmental monitoring data as prior evidence to generate fault segment determination results by coordinating conflicts of multi-source evidence include:

[0044] S410. Based on the dynamic topology model and the local fault diagnosis conclusions of each monitoring node, construct an evidence network for topological constraints, specifically including:

[0045] Using the vertices in the dynamic topology model as evidence network nodes and the connection relationships (edges) between vertices as evidence propagation paths, an evidence network is constructed that is isomorphic to the physical topology of the power grid.

[0046] The local fault diagnosis conclusions uploaded by each monitoring node are mapped to the corresponding evidence network node in the evidence network according to the corresponding monitoring node identifier, and serve as the initial evidence for that node; wherein, the virtual fault point, fault type and credibility score carried in the local fault diagnosis conclusion are converted into the initial support of the evidence node for the corresponding fault scenario.

[0047] Based on the real-time connection state (edge ​​enable / disable) described by the dynamic topology model, the propagable paths of evidence are defined, and based on the operating state (real-time output and switching state) of the associated distributed power source vertices, a dynamic propagation weight is assigned to each path to complete the initialization of the evidence network.

[0048] S420. In the evidence network, each evidence network node iteratively exchanges support information with adjacent evidence network nodes along the connection path, and dynamically adjusts the weight of the support information in the propagation based on the dynamic topology model and associated distributed power source operation data to generate preliminary coordinated evidence, specifically including:

[0049] In the evidence network, each evidence network node transmits a message containing its initial support to its neighboring evidence network nodes along the effective propagation path, and simultaneously receives messages from all neighboring evidence network nodes. Each evidence network node calculates its updated support for each fault scenario according to the update rules of the belief propagation algorithm, based on all received messages and the propagation weight of each message's propagation path.

[0050] The above message passing and support update process is executed iteratively. For messages originating from evidence network nodes that are electrically distant or have low propagation weights on their paths, and which conflict with local evidence, the weight of such messages in the support update calculation of the receiving evidence network nodes will be reduced. When the support changes of all evidence network nodes for the main failure scenario are less than a set threshold, the iteration terminates. At this point, the support distribution of all evidence network nodes in the entire evidence network is the preliminary coordinated global evidence distribution, which serves as preliminary coordinated evidence.

[0051] S430. Generate prior meteorological events based on meteorological and environmental monitoring data; correct the support information of associated line edges in the dynamic topology model based on the prior meteorological events, and generate a corrected global evidence distribution, specifically including:

[0052] First, characteristic parameters of prior meteorological events are extracted from meteorological and environmental monitoring data: for prior meteorological events such as lightning, the lightning current intensity value, the precise geographical coordinates of the lightning strike point and the timestamp are extracted; for prior meteorological events such as icing, the measured value of icing thickness, the icing thickness growth rate and the timestamp of the line monitoring section are extracted.

[0053] Next, the risk quantification index of prior meteorological events on the associated line sides in the dynamic topology model is calculated: for prior meteorological events such as lightning, the lightning risk quantification index is calculated by combining the spatial distance between the lightning strike point and the target line side and the terrain shielding effect; for prior meteorological events such as icing, the icing risk quantification index is calculated by combining the ratio of icing thickness to the design value and the growth rate.

[0054] Subsequently, spatial mapping and evidence correction are performed: each prior meteorological event is mapped to one or more corresponding line edges in the dynamic topology model according to its geographical location or monitoring segment identifier; for each line edge associated with a prior meteorological event, the current support of all its related fault hypotheses is obtained, and the support is quantitatively corrected using the calculated risk quantification index.

[0055] By traversing and processing all prior meteorological events, a systematic correction of the evidence support is completed, generating a global evidence distribution that fully integrates and quantifies the assessment results of prior meteorological events.

[0056] S440. Identify the fault segment with the highest support information from the global evidence distribution and generate a fault segment determination result, specifically including:

[0057] From the global evidence distribution, the final support of each fault scenario is extracted, and the fault scenario with the highest support exceeding the set confidence threshold is selected. According to the dynamic topology model, the virtual fault point corresponding to the selected fault scenario is parsed and converted into a specific physical segment description in the power grid. This description covers the faulty line segment and the physical equipment associated with the faulty line segment. The physical segment description and the corresponding highest support are used as the fault segment determination result of the system consensus, providing a target range for subsequent location execution.

[0058] According to the above technical solution, the step of extracting the theoretical traveling wave characteristics of the fault section based on the fault section determination result and realizing the power fault location of the target power grid through traveling wave feature matching includes:

[0059] S510. Based on the fault section determination result, extract the high-frequency traveling wave transient theoretical data of each monitoring node in the corresponding section from the fault theoretical response spectrum, as the theoretical traveling wave characteristics of the fault section, specifically including:

[0060] From the fault theoretical response spectrum, retrieve and extract the high-frequency traveling wave transient theoretical data of the located monitoring node under the fault scenario corresponding to the fault section, as the theoretical traveling wave feature of the fault section.

[0061] S520. Based on the theoretical traveling wave characteristics of the fault section, match and analyze the high-frequency traveling wave transient data of the monitoring nodes within the fault section to determine the location coordinates of the fault point, specifically including:

[0062] Obtain electrical monitoring data of relevant monitoring nodes within the fault-determined section, and extract high-frequency traveling wave transient data from them; perform time-series alignment and waveform consistency matching between the extracted high-frequency traveling wave transient data of each monitoring node and the theoretical traveling wave characteristics of the corresponding node extracted in S510.

[0063] Based on the time difference of arrival of traveling waves and waveform matching results of each monitoring node, and combined with the line parameters in the fault section, the precise location coordinates of the fault point in the section are calculated and determined; the location coordinates of the fault point are output to complete the precise location of the power fault in the target power grid.

[0064] Secondly, this application provides an electronic device, including: a processor and a memory, wherein the memory stores a computer program that can be called by the processor, and the processor executes a power fault collaborative sensing and precise positioning system based on data fusion by calling the computer program stored in the memory.

[0065] Thirdly, this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to execute a data fusion-based power fault collaborative sensing and precise location system.

[0066] Compared with the prior art, this application has the following advantages and beneficial effects:

[0067] This application achieves accurate pre-characterization and efficient knowledge storage of fault characteristics in complex power grids in mountainous areas by deeply fusing data and pre-generating fault theoretical response maps covering high-risk scenarios based on terrain risk-weighted simulation. Through local rapid feature matching and diagnosis by each monitoring node based on the map, it realizes on-site fault assessment and rapid response under weak communication constraints. By constructing an evidence network based on power grid topology constraints and integrating node diagnostic conclusions with high-deterministic meteorological prior evidence for conflict coordination, it achieves reliable determination of fault sections under multiple power source injection.

[0068] This application forms a complete technical closed loop from "risk prediction, local rapid judgment, global arbitration to precise point location", which significantly improves the reliability, speed and accuracy of fault location in mountainous power distribution networks with a high proportion of distributed power sources in weak communication environments. Attached Figure Description

[0069] Figure 1 This is a schematic diagram of the structure of a power fault collaborative sensing and precise location system based on data fusion provided in an embodiment of this application;

[0070] Figure 2 This is a schematic diagram of the process for generating a fault theoretical response map provided in an embodiment of this application;

[0071] Figure 3 This is a schematic diagram of the process for generating local fault diagnosis conclusions provided in an embodiment of this application;

[0072] Figure 4 This is a schematic diagram of the process for generating fault segment determination results provided in an embodiment of this application. Detailed Implementation

[0073] The technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solution of this application, rather than limitations thereof. In the absence of conflict, the embodiments and technical features in the embodiments can be combined with each other.

[0074] Please see Figure 1 , Figure 1 This is a schematic diagram of the structure of a power fault collaborative sensing and precise location system based on data fusion provided in an embodiment of this application, which specifically includes the following modules:

[0075] Multi-source acquisition module: Simultaneously acquires electrical monitoring data, equipment status data, meteorological and environmental monitoring data, and power grid topology and terrain coupling data of the target power grid.

[0076] In this embodiment, the synchronous acquisition of electrical monitoring data, equipment status data, meteorological and environmental monitoring data, and power grid topology and terrain coupling data of the target power grid includes the following specific contents:

[0077] S110. By collecting distribution monitoring data from each monitoring node of the target power grid, construct electrical monitoring data that includes three-phase current waveform data and high-frequency traveling wave transient data.

[0078] By deploying distribution automation terminals at each monitoring node of the target power grid, the current transformers built into the distribution automation terminals are used to collect three-phase current waveform data in real time, which reflects the changes in the path and amplitude of the fault current. The three-phase current waveform data is continuously recorded with time as the horizontal axis and current value as the vertical axis. It includes the power frequency period waveform and its corresponding harmonic components. The power frequency period is 20 milliseconds corresponding to 50Hz AC. The original amplitude data of each phase current is divided by the rated current value of the line segment to obtain the normalized three-phase current waveform data.

[0079] Meanwhile, the transient waveform recording module integrated into the power distribution automation terminal synchronously collects transient voltage and current signals when a power fault occurs, and generates high-frequency traveling wave transient data containing the arrival time, amplitude and polarity information of the traveling wave front after analog-to-digital conversion.

[0080] S120. By collecting equipment monitoring data of power equipment in the target power grid, construct equipment status data including equipment temperature data and equipment partial discharge data.

[0081] By deploying wireless temperature measurement units at various monitoring nodes within the target power grid, real-time temperature data reflecting the thermal state of the equipment connection points and the equipment itself is collected. The equipment temperature data includes the measured temperature value, the measurement point location identifier, and the corresponding timestamp. The normalized temperature index characterizing the deviation of the real-time thermal state of the measurement point is obtained by subtracting the average temperature obtained by the wireless temperature measurement unit under historical normal operation from the temperature measured value collected by each wireless temperature measurement unit, and then dividing by the corresponding temperature standard deviation. The temperature standard deviation is calculated based on the standard deviation of the temperature value sequence continuously collected by the wireless temperature measurement unit during historical normal operation, and is used to quantify the dispersion (i.e., the natural fluctuation amplitude) of the temperature value of the wireless temperature measurement unit around the historical mean.

[0082] Meanwhile, by deploying UHF partial discharge monitoring devices at each monitoring node, partial discharge data reflecting internal insulation defects of the equipment is collected. The partial discharge data is captured by UHF sensors to capture partial discharge signals, and after processing, discharge pulse frequency information is extracted. The discharge pulse frequency collected by each monitoring node is divided by the upper limit of pulse frequency statistics under historical defect-free conditions to obtain the normalized partial discharge frequency index.

[0083] S130. By collecting external environmental monitoring data of the target power grid area, construct meteorological and environmental monitoring data that includes lightning location data and icing monitoring data.

[0084] By connecting to a lightning monitoring system covering the target power grid area, lightning location data with accurate timestamps, geographic coordinates, and raw lightning current intensity information is collected. This raw lightning current intensity information represents the initial energy parameters of the lightning strike event in kiloamperes. At the same time, through online icing monitoring devices deployed along the target power grid lines (or towers), icing monitoring data with line section location identification, measured icing thickness, and growth rate information is collected.

[0085] S140. By calling the power grid dispatching system and geographic information system of the target power grid area, construct real-time power grid topology data and distributed power source operation data that reflect the real-time operation status of the power grid, as well as terrain slope data that reflects geographical environmental risks.

[0086] By calling the real-time data interface of the power grid dispatch system's energy management system (EMS) or distribution automation system (DAS), real-time power grid topology data is collected, including a list of physical devices, real-time connection relationships between physical devices, and switch opening and closing status. Each physical device in the target power grid area is assigned a unique device identifier, which follows the unified coding standard for power grid assets and is used to identify a transformer, a switch, or a line segment in this embodiment.

[0087] Simultaneously, the real-time data interface collects distributed power source operation data reflecting the real-time output values ​​and switching status of each distributed power source. Each distributed power source is associated and indexed through the device identifier corresponding to the physical device at the distributed power source's grid connection point. In addition, the power distribution monitoring terminal deployed on the physical device is logically defined as a monitoring node, and the monitoring node inherits or is associated with the device identifier of the physical device, which serves as the unique identifier of the monitoring node.

[0088] By calling geographic information system data services, based on the spatial path coordinates of the target power grid line corridor, the line is divided into segments along the line direction with adjacent towers as boundaries. The spatial coordinate range (start and end coordinates) of each line segment and the average slope value corresponding to the line segment are extracted to form structured terrain slope data.

[0089] Risk modeling module: Constructs a dynamic topology model based on coupled power grid topology and terrain data, and generates a fault theoretical response map through terrain risk weighted simulation.

[0090] In this embodiment, as Figure 2 As shown, a dynamic topology model is constructed based on coupled power grid topology and terrain data, and a fault theoretical response map is generated through terrain risk-weighted simulation. The specific steps include the following:

[0091] S210. By analyzing the real-time topology data of the power grid and the operation data of distributed power sources in the coupled data of power grid topology and terrain, a dynamic topology model is constructed.

[0092] Traverse the list of physical devices in the real-time topology data of the power grid, and create each physical device (including substation outgoing switches, sectionalizing switches, tie switches, transformers and load points) as a vertex in the dynamic topology model, and assign each vertex with the device identifier extracted from the physical device list; according to the real-time connection relationship between physical devices, create an edge for two related vertices in the dynamic topology model, and assign the corresponding line identifier to the edge, thereby generating the basic skeleton of the dynamic topology model;

[0093] Meanwhile, the switch opening and closing status is continuously read. If the edge corresponding to a switch is marked as disconnected, the edge is set to a disabled state in the dynamic topology model to update the electrical connectivity between vertices.

[0094] Finally, the distributed power source operation data is read, which includes the device identifier, real-time output value, and switching status of the distributed power source grid connection point. Based on the device identifier of the grid connection point, the vertex with the same device identifier is searched in the basic skeleton of the constructed dynamic topology model. The real-time output value and switching status of the distributed power source are written into the attribute of the vertex, thereby completing the construction of the dynamic topology model.

[0095] S220. Based on terrain slope data, add geographical risk labels to the dynamic topology model.

[0096] Read the terrain slope data obtained in step S140. The terrain slope data has been organized into segments according to the route, and the spatial coordinate range and corresponding average slope value of each segment are recorded.

[0097] The algorithm iterates through each edge in the dynamic topology model (corresponding to a physical line segment) and obtains the spatial coordinate information of the edge based on the actual line identifier associated with that edge in the power grid topology. Then, it searches for line segment records in the terrain slope data that match the spatial coordinate range of the edge information and reads the average slope value of that line segment.

[0098] The average slope value read is compared with a preset high slope threshold. In this embodiment, the default high slope threshold is 25 degrees, which is a critical value set based on relevant power industry standards and experience in the operation and maintenance of power lines in mountainous areas.

[0099] If the average slope value corresponding to a certain edge is greater than this high slope threshold, then a high slope risk label is added to that edge.

[0100] S230. Based on the dynamic topology model, perform power system transient simulation and perform differentiated weighting on the simulation parameters based on geographical risk labels. Output theoretical electrical response data for each monitoring node, including theoretical data of three-phase current waveforms and theoretical data of high-frequency traveling wave transients.

[0101] In step S220, the edges marked with high slope risk labels are treated as one type of virtual fault point and a single-phase ground fault is injected. At the same time, the vertices directly connected to these high slope risk edges are treated as another type of virtual fault point and a three-phase short-circuit fault is injected.

[0102] For each virtual fault point, based on the slope data of the associated line segment, the electrical parameters of the corresponding line segment are differentially weighted and corrected according to the following correction function:

[0103] ;

[0104] in, Represents the corrected vertex With vertex The equivalent impedance of the line segment between them, in ohms; This represents the original power frequency equivalent impedance of the line segment calculated based on the conductor type, length, and distance to ground, in ohms. The normalized impact coefficient represents the risk label of high gradient on this section of the line. The calculation method is as follows: In the formula This is the actual gradient value for this section of the line. The preset high slope threshold (25 degrees) is used. The reference slope value used for normalization (in this embodiment, it is taken as...) (65 degrees represents an extreme terrain condition). The preset weighting factor is used to adjust the correction strength for slope risk. The specific value is obtained by comparing and analyzing historical fault data and simulation verification results under typical mountainous terrain where the target power grid is located. The aim is to match the statistical regularity of the theoretical response data generated by the simulation with the actual fault characteristics (in this embodiment, the default values ​​are taken separately). =0.15).

[0105] After parameter correction, the determined virtual fault points, fault types, and the entire network parameter set including the line impedance after terrain risk weighting correction are input into an electromagnetic transient simulation program (such as EMTP or PSCAD) to perform power system transient simulation calculations to deduce the fault transient voltage and current waveforms at each vertex. The core input parameters for the simulation include: the location and type of the virtual fault point (three-phase short circuit or single-phase grounding), line resistance, inductance, and ground capacitance parameters based on the line design ledger and considering terrain correction, and the power source model and real-time output values ​​reflecting the real-time status of distributed power sources. The power source model (such as the controlled current source model of a photovoltaic inverter or the controlled voltage source model of an energy storage converter) is set according to the standard control architecture of grid-connected converters known in the field. The core static parameters of this power source model (including rated capacity and rated voltage) are derived from the equipment nameplate data.

[0106] The simulation is performed under preset operating conditions (in this embodiment, the simulation time window is 20 milliseconds after the fault starts, the calculation step is 1 microsecond, and the output sampling rate is 1MHz). The electromagnetic transient simulation program solves the power grid time-domain differential equations based on this full network parameter set, thereby calculating the theoretical electrical response data of each monitoring node under the corresponding fault scenario.

[0107] Finally, the three-phase current time-domain waveforms at the vertices corresponding to the identifiers of each monitoring node are extracted from the simulation results as theoretical data of the three-phase current waveforms, and the traveling wave voltage propagation waveform at the vertices is extracted as theoretical data of the high-frequency traveling wave transient. Together, they constitute the theoretical electrical response data of the monitoring node under this specific fault scenario. After generating the theoretical electrical response data, the theoretical data of the three-phase current waveforms are normalized. The amplitude of each phase current in the theoretical data of the three-phase current waveforms is divided by the rated current value of the corresponding line segment. Finally, the above process is completed by traversing all virtual fault points, and all theoretical electrical response data are output.

[0108] S240. Integrate the theoretical electrical response data of each monitoring node under the transient simulation of the power system to generate a fault theoretical response map.

[0109] The system iterates through all theoretical electrical response data output after performing transient simulation for each virtual fault point in step S230. Using the unique monitoring node identifier of each monitoring node as the main index, it aggregates all three-phase current waveform theoretical data and high-frequency traveling wave transient theoretical data belonging to the same monitoring node but from different virtual fault points to form a data set centered on that monitoring node. Finally, it systematically organizes the data sets of all monitoring nodes to generate a fault theoretical response map.

[0110] Collaborative analysis module: Each monitoring node extracts theoretical response features from the fault theoretical response spectrum, integrates electrical monitoring data and equipment status data, and generates local fault diagnosis conclusions through feature matching and credibility weighted analysis.

[0111] In this embodiment, as Figure 3 As shown, each monitoring node extracts theoretical response features from the fault theoretical response spectrum and integrates electrical monitoring data and equipment status data. Through feature matching and confidence weighted analysis, a local fault diagnosis conclusion is generated, including the following specific steps:

[0112] S310. Each monitoring node retrieves and obtains the theoretical response characteristic data corresponding to its own monitoring node from the fault theoretical response map.

[0113] Each monitoring node initiates a query request to the fault theoretical response map; based on the monitoring node identifier, a matching search is performed in the fault theoretical response map to extract the three-phase current waveform theoretical data and high-frequency traveling wave transient theoretical data corresponding to the monitoring node, which are used as the theoretical response features for the monitoring node to perform local fault diagnosis.

[0114] S320. Calculate the waveform similarity between the three-phase current waveform data in the electrical monitoring data and the theoretical three-phase current waveform data in the theoretical response characteristic data to obtain a steady-state matching index; calculate the propagation timing and waveform consistency between the high-frequency traveling wave transient data and the high-frequency traveling wave transient theoretical data to obtain a transient matching index; perform differential fusion calculation between the steady-state matching index and the transient matching index based on the fault electrical characteristics represented by the theoretical electrical response data to generate a comprehensive matching score corresponding to each theoretical electrical response data, and sort and filter according to the comprehensive matching score to generate a preliminary fault hypothesis sequence.

[0115] First, steady-state characteristic matching is performed. The collected normalized three-phase current waveform data is time-aligned with the corresponding normalized three-phase current waveform theoretical data extracted from the fault theoretical response spectrum. That is, the time axis of the two sets of waveform data sequences is calibrated with the fault initiation time as the reference zero point.

[0116] Based on the aligned two sets of current waveform data sequences, the data segment from the 2nd to the 5th power frequency cycle after the fault starts (i.e., from the 20th millisecond to the 100th millisecond) is taken as the steady-state calculation window. Within this steady-state calculation window, the three-phase current waveform data and the theoretical three-phase current waveform data are regarded as two discrete signal vectors. The ratio of the product of the covariance and the standard deviation of each vector is calculated to obtain the Pearson correlation coefficient, which is used as a steady-state matching index to characterize the similarity of steady-state waveforms.

[0117] Secondly, transient feature matching is performed. High-frequency traveling wave transient data is compared with corresponding high-frequency traveling wave transient theoretical data. Using the wavelet transform modulus maxima method, the arrival time of the traveling wave front is detected from both types of data, and the absolute time difference is calculated. Then, using the arrival time of the traveling wave front as the alignment benchmark, the same analysis time window is extracted for both transient waveforms (20 microseconds after the arrival of the wave front in this embodiment), and the envelope of the waveform within this analysis time window is extracted using Hilbert transform. The Pearson correlation coefficient of the two envelope sequences is calculated. Finally, the time difference and the envelope correlation coefficient are fused to generate a transient matching index. In this embodiment, the fusion calculation method is as follows: the transient matching index is equal to the envelope correlation coefficient divided by (the sum of the time difference and a very small constant), so that the smaller the time difference and the more similar the waveforms, the larger the index value, comprehensively representing the consistency of transient features.

[0118] Next, a comprehensive score is generated based on the fault type through differentiated fusion. According to the fault type corresponding to the current virtual fault point (three-phase short-circuit fault or single-phase ground fault), the above steady-state matching index and transient matching index are weighted and fused to calculate a comprehensive matching degree score.

[0119] To address the differences in electrical response mechanisms across different fault types, the contribution weights of the two types of indicators in the comprehensive score are dynamically adjusted. When the fault type is a three-phase short-circuit fault, the fault current is large and the steady-state characteristics are significant; therefore, the comprehensive score should emphasize the steady-state matching indicator. When the fault type is a single-phase ground fault, the fault current is small but the traveling wave transient characteristics are prominent; therefore, the comprehensive score should emphasize the transient matching indicator. (In this embodiment, the default weight allocation is: for three-phase short-circuit faults, the steady-state matching indicator accounts for 70% and the transient matching indicator accounts for 30%; for single-phase ground faults, the steady-state matching indicator accounts for 30% and the transient matching indicator accounts for 70%). This weight allocation scheme ensures that the comprehensive matching score can more accurately reflect the degree of agreement between the current fault hypothesis and the observed data across different dominant characteristic dimensions.

[0120] Finally, after completing the above matching and scoring calculations, the fault scenarios with scores higher than the matching threshold (defined by virtual fault points, fault types, and theoretical electrical response data) are selected and the corresponding virtual fault points, fault types, and scores are used as output items to form a preliminary fault hypothesis sequence.

[0121] The matching degree judgment threshold is determined through the following statistical analysis process: collect historical fault case data of the target power grid area over the past three years, calculate the comprehensive matching degree score of all correct and incorrect fault assumptions generated by each monitoring node for each real power fault; calculate the comprehensive matching degree score distribution of correct and incorrect fault assumptions respectively; based on the comprehensive matching degree score distribution data, determine the optimal judgment threshold with the optimization objective of maximizing the correct recognition rate and minimizing the false alarm rate.

[0122] S330: Each monitoring node calls up the device status data and combines it with the associated distributed power supply operation data to perform a confidence-weighted correction on the preliminary fault hypothesis sequence and generate a local fault diagnosis conclusion.

[0123] Construct a device observation weighting coefficient This system is used to fuse equipment status data and distributed power source operation data to perform physical-level reliability calibration on preliminary fault hypothesis sequences based on electrical feature matching. The construction of the equipment observation weight coefficients is based on the fact that the reliability of the monitoring node's observation data is simultaneously affected by the equipment's own health status (characterized by equipment temperature and partial discharge) and the interference from nearby distributed power sources (characterized by output fluctuations), and these two effects are independent of each other. Equipment observation weight coefficients This is obtained by multiplying the following two core factors:

[0124] First, calculate the basic credibility factor. This basic reliability factor kernel reflects the impact of the health status of the associated device on the reliability of the monitoring data, and the calculation function is:

[0125] ;

[0126] in: and These are the normalized temperature index and the normalized partial discharge frequency index corresponding to this monitoring node, respectively. Both are dimensionless quantities, and the larger the value, the more significant the abnormality of the equipment status. and The preset contribution weight satisfies =1, where and The weighting allocation is based on the technical consensus of power equipment condition monitoring, namely that abnormal temperature at equipment connection points directly affects circuit impedance and current measurements, while partial discharge reflects the long-term state of insulation. The two have different immediate impacts on the reliability of current electrical measurements. To quantify this difference, this embodiment pre-sets differentiated contribution weights for temperature and partial discharge indicators, set as follows: =0.6, =0.4;

[0127] This function comprehensively evaluates multi-dimensional device status anomalies through weighted summation and subtracts them from 1, resulting in a healthier device status. The closer it is to 1, the higher the level of trust support provided at the device level.

[0128] Secondly, calculate the power supply influence weights. This weighted kernel quantifies the comprehensive interference caused by the operating states of multiple associated distributed power sources on the electrical measurement environment of the monitoring nodes. The calculation function (power source influence weighting function) is as follows:

[0129] ;

[0130] in: It represents the natural exponential function, that is, the exponential operation with the natural constant e (approximately equal to 2.71828) as the base; The aggregate output volatility is a dimensionless quantity. The calculation method is as follows: First, extract all grid-connected distributed power sources whose electrical distance (e.g., shortest path hop count) from the current monitoring node in the dynamic topology model is less than or equal to 2 from the distributed power source operation data; second, calculate the ratio of the standard deviation of the output value of each such distributed power source within a time window (e.g., 10 seconds) before the fault to the rated output of such distributed power sources to obtain the individual volatility. Finally, all of them Electrical distance from the monitoring node (Number of jumps) are inversely weighted and summed: ; The larger the value, the more unstable the overall operation of the nearby distributed power source;

[0131] The fluctuation attenuation coefficient is a constant greater than 0. Its specific value is determined by fitting and analyzing the correlation between power fluctuations and diagnostic errors in historical monitoring data. The aim is to ensure that the power supply influence weighting function remains within the typical fluctuation range (e.g., ...). The function exhibits a clear decay gradient (between 0 and 1), thus achieving a balance between sensitively distinguishing the degree of interference and avoiding premature saturation of the function value.

[0132] The power source influence weighting function reflects the nonlinear decay effect of power output fluctuation on the observation reliability: when the aggregate volatility of distributed power sources is very low, the weight is close to 1; when the aggregate volatility of distributed power sources increases, the weight drops rapidly, indicating that the power source operation has increased interference with the monitoring data.

[0133] Ultimately, the equipment observation weighting coefficient Multiplying the two above, we get: = This multiplication relationship reflects the health status of the equipment itself. Interference with external power supply The common and coupled impact on the reliability of monitoring data means that a low value in any factor will lead to a significant reduction in the overall equipment observation weight coefficient, which is consistent with the judgment logic of series risks in engineering practice.

[0134] The device is used to observe the weighting coefficients and correct the overall matching score of each item to obtain the final credibility score. Based on the corrected final credibility score, all fault hypotheses are sorted in descending order, and the top N fault hypotheses are selected (N=3 by default in this embodiment, and this number is determined based on the distribution statistics of effective fault hypotheses of a single node in typical historical fault data). The virtual fault point, fault type, and final credibility score of each adopted fault hypothesis are encapsulated together with the identifier of this monitoring node and the timestamp to generate a structured local fault diagnosis conclusion and upload it.

[0135] Evidence fusion module: Constructs an evidence network based on the local fault diagnosis conclusions and dynamic topology model of each monitoring node; and uses meteorological and environmental monitoring data as prior evidence to generate fault section determination results by coordinating conflicts of multi-source evidence.

[0136] In this embodiment, as Figure 4 As shown, an evidence network is constructed based on the local fault diagnosis conclusions of each monitoring node and the dynamic topology model; meteorological and environmental monitoring data are used as prior verification evidence, and fault segment determination results are generated by coordinating conflicts among multi-source evidence, including the following specific steps:

[0137] S410. Based on the dynamic topology model and the local fault diagnosis conclusions of each monitoring node, construct an evidence network of topological constraints.

[0138] Each vertex in the dynamic topology model is instantiated as an evidence network node in the evidence network; each edge representing an electrical connection in the dynamic topology model is instantiated as an evidence propagation path connecting two corresponding evidence network nodes in the evidence network. Subsequently, the fault hypothesis (i.e., each item in the preliminary fault hypothesis sequence, including the virtual fault point, fault type, and final confidence score) contained in the local fault diagnosis conclusion uploaded by each monitoring node is used as the initial evidence for that evidence network node; based on the monitoring node identifier in the local fault diagnosis conclusion, the fault hypothesis is mapped to the corresponding evidence network node in the evidence network, and the final confidence score of the fault hypothesis is directly set as the initial support of that evidence network node.

[0139] Simultaneously, based on the real-time connection status (enabled or disabled) of each edge in the dynamic topology model, the validity of the evidence propagation path at the current moment is defined; and based on the real-time switching status and output fluctuation of the distributed power source vertices associated with each edge, an initial propagation weight is assigned to each valid evidence propagation path.

[0140] The specific method for setting the propagation weight is as follows: Extract all distributed power sources associated with both ends of the currently valid evidence propagation path, and calculate the output fluctuation rate of each distributed power source in the last 5 minutes, i.e., the ratio of (maximum real-time output value minus minimum real-time output value) to the rated output of the distributed power source; where the rated output is the rated capacity on the nameplate of each distributed power source. Take the average of the output fluctuation rates of all associated power sources as the comprehensive fluctuation rate of the evidence propagation path; calculate according to the formula "initial propagation weight = 1 - 0.9 × comprehensive fluctuation rate" (if there are no associated distributed power sources in the evidence propagation path, the comprehensive fluctuation rate is 0); finally, adjust the calculated weight value to the interval [0.1, 1] (take 0.1 when it is below 0.1, and take 1 when it is above 1).

[0141] The aforementioned propagation weights are based on the fact that output volatility directly characterizes the stability of distributed power source output. Higher volatility indicates greater uncertainty in the electrical state of the path. The reverse correlation weights can reduce the propagation interference of unreliable data, while the coefficient of 0.9 and the range of [0.1,1] ensure both the effect of volatility suppression and avoid completely blocking the transmission of valid evidence, thus conforming to the actual operating rules of the power grid. Through the above operations, an evidence network is constructed that is isomorphic to the physical topology of the power grid and is loaded with initial evidence states and propagation weights.

[0142] Through the above steps, an evidence network is constructed that is isomorphic to the physical topology of the power grid and is loaded with initial evidence states and propagation weights.

[0143] S420. In the evidence network, each evidence network node iteratively exchanges support information with adjacent evidence network nodes along the connection path, and dynamically adjusts the weight of support information in the propagation based on the dynamic topology model and associated distributed power source operation data to generate preliminary coordinated evidence.

[0144] In the constructed evidence network, each node iterates through multiple rounds of message passing and support updates with its neighboring nodes along the effective propagation path. By introducing physical constraints of the power grid, local evidence is dynamically coordinated to generate preliminary coordinated evidence. The core algorithm of this process consists of a dynamic propagation weight function and a conflict evidence attenuation rule.

[0145] First, a dynamic propagation weight function is constructed to quantify the propagation reliability of evidence under the constraints of the power grid's physical topology. In each iteration of the evidence network, when node i transmits a message to its neighbor j, the propagation effectiveness of that message is determined by the dynamic weight function. The adjustment, specifically the calculation formula, is as follows:

[0146] ;

[0147] in, The dynamic propagation weight is a dimensionless coefficient that ranges from 0 to 1. The larger the value, the stronger the evidence network node. Send to evidence network nodes The higher the weight of the message in the recipient's support update calculation; The reference distance is 1 (unit: kilometers) and is a constant used for distance normalization; For evidence network nodes and The electrical distance between them is measured in kilometers (km). The specific value is obtained by querying the dynamic topology model and calculating the shortest path length among all valid connection paths between the corresponding two vertices, reflecting the physical topology constraints of the power grid.

[0148] For evidence network nodes and The average output fluctuation rate of the associated distributed power sources is a dimensionless quantity. The calculation method is as follows: Extract the output values ​​of all grid-connected distributed power sources that are electrically closest (e.g., within 2 hops) to evidence network nodes i and j within a short time window (e.g., 10 seconds) before the fault. Calculate the ratio of the standard deviation of each output value to the rated output, and then take the average. This average value reflects the stability of the operation of the relevant distributed power sources. The topology weight percentage is a preset constant (default in this embodiment). =0.6), the value is determined based on statistical analysis of the correlation between historical communication quality and electrical propagation of faults, and is used to balance the relative importance of electrical distance and power fluctuation in the weight calculation.

[0149] Secondly, in the evidence network nodes When aggregating messages from all neighboring nodes to update its own support, the conflict evidence decay rule is executed. Evidence network nodes For from neighboring nodes When fusing support information for a specific fault hypothesis, the actual weighting coefficients used are... Determined by the following formula:

[0150] ;

[0151] in, These are effective weighting coefficients used to ultimately calculate the contribution of neighboring node messages; Let be the consistency function, where and Representing nodes respectively Self and Nodes The consensus function for the most supported fault hypothesis (i.e., the core hypothesis) is defined as follows: if two core hypotheses point to the same path edge or vertex in the dynamic topology model, then... =1, otherwise =0; therefore, when the core assumptions point to the same fault location (evidence network node), the message weight is not decayed; when the core assumptions contradict each other, the message weight is completely decayed (the coefficient is 0).

[0152] Each evidence network node updates its own support based on its current support for each failure hypothesis and the messages received from neighboring nodes. The update process employs a probability propagation mechanism based on Bayesian theory. Specifically, each evidence network node calculates its posterior probability for each failure hypothesis. This posterior probability is obtained by normalizing the product of the node's initial support and the likelihood information from all neighboring nodes regarding that failure hypothesis. The likelihood information from neighboring nodes consists of messages sent by those neighboring nodes and weighted by dynamic propagation. The support value after adjustment and decay by the consistency function.

[0153] During the iteration process, each evidence network node synchronously executes the above update and sends the updated support to its neighboring nodes again. The iteration termination condition is as follows: among all evidence network nodes, the maximum absolute value of the difference between the support value of the fault hypothesis with the highest current support and the corresponding support value in the previous round is less than the iteration termination threshold for three consecutive times (in this embodiment, the default iteration termination threshold is 0.01, which is set based on the statistical analysis of the convergence process of the evidence network in historical fault cases of the target power grid, aiming to balance iteration efficiency and evidence stability, and the value is set so that the network can reach a convergence state within 5-10 rounds under typical fault scenarios); when this condition is met, the iteration stops, and the support distribution formed by each evidence network node in the evidence network at this time is the preliminary coordination evidence.

[0154] S430. Generate prior meteorological events based on meteorological and environmental monitoring data; correct the support information of associated line edges in the dynamic topology model based on the prior meteorological events, and generate the corrected global evidence distribution.

[0155] For a priori meteorological event of lightning, the risk intensity value of a certain line edge L in the dynamic topology model is... Calculated using the following function:

[0156] ;

[0157] in: The lightning current intensity is derived from the raw lightning current intensity information in lightning location data. This is a reference value used to normalize the lightning current intensity. The specific value is set based on the upper limit of the current range of the distribution monitoring terminal in the target power grid, and is introduced... The purpose is to convert the original physical quantity with the dimension of current into a ratio of a dimensionless quantity; The shortest spatial distance between the lightning strike point and the line edge L is calculated using a spatial analysis algorithm based on the geographic coordinates of the lightning location data and the spatial coordinates of the line edge L in the power grid geographic information system (usually represented as a continuous coordinate sequence or line segment vector between line nodes). The empirical attenuation distance constant is determined by collecting historical lightning strike fault case data in the target area and extracting the distance data between the lightning strike point and the faulty line location in each lightning strike fault. Using distance as the independent variable and fault occurrence as the dependent variable, an exponential function curve characterizing the attenuation of lightning strike impact with distance is fitted. The exponential attenuation constant that optimizes the statistical relationship between distance and fault probability in this attenuation curve is determined as follows: The value obtained through this step; The value ensures that the calculated lightning strike risk intensity matches the statistical regularity of historical faults in the area;

[0158] The terrain occlusion coefficient is a dimensionless quantity ranging from 0 to 1. It is calculated as follows: based on digital elevation model data from a geographic information system, elevation points are sampled at fixed intervals along the path connecting the lightning strike point and the target route. It is then determined whether the view from each sampling point is obstructed by terrain. The proportion of unobstructed sampling points along the entire path is calculated; this proportion is the terrain occlusion coefficient. .

[0159] For the prior meteorological event of lightning during icing, the corresponding risk intensity value along the line. Calculated using the following function:

[0160] ;

[0161] in: The ratio of icing thickness is a dimensionless quantity, and its value is the measured value of icing thickness from the icing monitoring data of the line monitoring section. (Unit: mm) and the design ice thickness for this section of the line. The ratio, i.e. ; The growth rate is derived from the growth rate information of icing monitoring data; The growth rate impact coefficient is specifically derived from a set of icing fault cases extracted from historical faults in the target area. Each set of icing fault cases includes the measured icing thickness of the line section before the fault. With the rate of increase in ice thickness ; Calculate the icing thickness ratio for each icing failure case set; Using this set of historical data, quantify the growth rate through regression analysis. The degree of independent contribution to failure risk; the coefficients are determined based on the regression analysis results. The specific values ​​are used to ensure that the calculated value of the icing event function is consistent with the risk distribution in historical cases.

[0162] To obtain the comprehensive meteorological risk intensity value R (i.e., the lightning risk intensity value is fused through linear superposition), With icing risk intensity value After obtaining the results, based on the engineering principle that multiple independent meteorological stresses have a cumulative effect on the risk of power line faults, the support for the relevant fault assumptions in the preliminary coordination evidence is revised. The revision function is as follows:

[0163] ;

[0164] in, and These represent the support level of a certain line edge for a specific fault hypothesis before and after the correction. To verify the environmental data fusion weighting coefficients, specific values ​​were determined by collecting historical case data including comprehensive meteorological risk intensity values, initial support of the evidence network, and final confirmed fault locations. The optimization objective was to ensure consistency between the fault determination results and the actual location. These values ​​were then adjusted during simulation. The optimal solution is obtained by fitting the values. This represents the comprehensive meteorological risk intensity value mapped to the edge of the route.

[0165] The above function iterates through all meteorological events, performs quantitative correction of the evidence support, and generates a global evidence distribution that integrates environmental disaster-causing factors.

[0166] S440. Identify the fault segment with the highest support information from the global evidence distribution and generate the fault segment determination result.

[0167] The final support values ​​of all fault hypotheses in the global evidence distribution are traversed, and the fault hypothesis with the highest support that exceeds the set confidence threshold is selected. The confidence threshold is set based on the support distribution characteristics of the evidence network in historical fault case simulation outputs: the final support data of all correct and incorrect fault hypotheses output by the evidence network in historical cases are collected, and the confidence threshold is determined through distribution analysis with the goal of achieving the best fault segment identification accuracy on the validation set (in this embodiment, the default confidence threshold value is 0.7).

[0168] Based on the virtual fault point identifier and fault type corresponding to the fault hypothesis, the specific power grid physical device (vertices or edges) is located in the dynamic topology model and parsed and converted into a clear physical segment description, which includes the faulty line segment, associated equipment, and location information. Finally, this physical segment description, the corresponding fault type, and the final support score are encapsulated together into a structured fault segment determination result.

[0169] Location execution module: Based on the fault section determination results, extract the theoretical traveling wave characteristics of the fault section, and realize the power fault location of the target power grid through traveling wave feature matching.

[0170] In this embodiment, the extraction of theoretical traveling wave features of the fault section based on the fault section determination results, and the realization of power fault location of the target power grid through traveling wave feature matching, includes the following specific contents:

[0171] S510. Based on the fault section determination results, extract the high-frequency traveling wave transient theoretical data of each monitoring node in the corresponding section from the fault theoretical response spectrum, and use it as the theoretical traveling wave characteristics of the fault section.

[0172] First, based on the physical segment (line segment and associated equipment) described in the fault segment determination result, all monitoring nodes covered or adjacent to the physical segment are located in the dynamic topology model, and the identifiers of these monitoring nodes are obtained. Then, based on the power grid physical segment (line segment) determined by the fault segment determination result, the high-frequency traveling wave transient theoretical data corresponding to all monitoring nodes located within or adjacent to the power grid physical segment when the segment is marked as a virtual fault point are retrieved and extracted from the fault theoretical response map. Finally, the extracted high-frequency traveling wave transient theoretical data of each monitoring node are organized according to the node identifier to form a systematic theoretical traveling wave feature set, which serves as the benchmark data for subsequent traveling wave feature matching.

[0173] S520. Based on the theoretical traveling wave characteristics of the fault section, match and analyze the high-frequency traveling wave transient data of the monitoring nodes in the fault section to determine the location coordinates of the fault point.

[0174] Based on the theoretical traveling wave feature set of the fault section, precise fault location is achieved through traveling wave feature matching. The specific process is as follows: Based on the target line section determined by the fault section assessment results, measured high-frequency traveling wave transient data from both ends or all relevant monitoring nodes of the section are acquired, and the precise arrival time of the initial traveling wave front of the fault is detected in each data set; according to the wave propagation velocity per unit length of the line in the section (determined by the line type and structure) and the section topology length, the specific location coordinates of the fault point are calculated using the two-end traveling wave ranging formula; assuming the total length of the section... wave speed is The time difference of arrival of the traveling wave detected at both monitoring points is The location of the fault point from one end The calculation formula is Finally, the precise location coordinates of the fault point are output, completing the accurate location of the power fault in the target power grid.

[0175] Embodiments of the present invention also provide an electronic device, including a memory, a processor, and a communication bus; the memory and the processor are connected via the communication bus. The memory stores a data fusion-based power fault collaborative sensing and precise location system that can be loaded by the processor and executed as provided in the above embodiments.

[0176] The memory can be used to store instructions, programs, code, code sets, or instruction sets. The memory may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for at least one function, and instructions for implementing the data fusion-based power fault collaborative sensing and precise location system provided in the above embodiments. The data storage area may store data involved in the data fusion-based power fault collaborative sensing and precise location system provided in the above embodiments.

[0177] A processor may include one or more processing cores. The processor executes instructions, programs, code sets, or instruction sets stored in memory, and calls data stored in memory to perform various functions and process data as described in this application. The processor may be at least one of a specific application-specific integrated circuit, a digital signal processor, a digital signal processing device, a programmable logic device, a field-programmable gate array, a central processing unit, a controller, a microcontroller, and a microprocessor. It is understood that, for different devices, the electronic devices used to implement the above-described processor functions may also be other types, and the embodiments of this application do not specifically limit the specific implementation.

[0178] A communication bus may include a pathway for transmitting information between the aforementioned components. The communication bus can be a PCI bus or an EISA bus, etc. Communication buses can be categorized into address buses, data buses, control buses, etc.

[0179] This application provides a computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described in the above embodiments, which is a data fusion-based power fault collaborative sensing and precise location system.

[0180] In this embodiment, a computer-readable storage medium can be a tangible device that holds and stores instructions used by an instruction execution device. A computer-readable storage medium can be, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof. Specifically, a computer-readable storage medium can be a portable computer disk, a hard disk, a USB flash drive, a random access memory, a read-only memory, an erasable programmable read-only memory, a podium random access memory, a portable compressed disk read-only memory, a digital multifunction disk, a memory stick, a floppy disk, an optical disk, a magnetic disk, a mechanical encoding device, or any combination thereof.

[0181] The terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0182] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing application concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions claimed in this application.

Claims

1. A power fault collaborative sensing and precise location system based on data fusion, characterized in that, Includes the following modules: Multi-source acquisition module: Simultaneously acquires electrical monitoring data, equipment status data, meteorological and environmental monitoring data, and power grid topology and terrain coupling data of the target power grid; S110. By collecting distribution monitoring data from each monitoring node of the target power grid, construct electrical monitoring data that includes three-phase current waveform data and high-frequency traveling wave transient data. S120. By collecting equipment monitoring data of power equipment in the target power grid, construct equipment status data including equipment temperature data and equipment partial discharge data; S130. By collecting external environmental monitoring data of the target power grid area, construct meteorological and environmental monitoring data that includes lightning location data and icing monitoring data; S140. By calling the power grid dispatching system and geographic information system of the target power grid area, construct real-time power grid topology data and distributed power source operation data that reflect the real-time operation status of the power grid, as well as terrain slope data that reflects geographical environmental risks. Risk modeling module: Constructs a dynamic topology model based on coupled power grid topology and terrain data, and generates a fault theoretical response map through terrain risk weighted simulation; S210. By analyzing the real-time power grid topology data and distributed power generation operation data in the power grid topology and terrain coupling data, a dynamic topology model is constructed. S220. Based on terrain slope data, add geographical risk labels to the dynamic topology model; S230. Based on the dynamic topology model, perform power system transient simulation and perform differentiated weighting on the simulation parameters based on geographical risk labels, and output theoretical electrical response data for each monitoring node, including theoretical data of three-phase current waveforms and theoretical data of high-frequency traveling wave transients. S240. Integrate the theoretical electrical response data of each monitoring node under power system transient simulation to generate a fault theoretical response map; Collaborative analysis module: Each monitoring node extracts theoretical response features from the fault theoretical response spectrum, integrates electrical monitoring data and equipment status data, and generates local fault diagnosis conclusions through feature matching and credibility weighted analysis; Evidence fusion module: Constructs an evidence network based on the local fault diagnosis conclusions of each monitoring node and the dynamic topology model; Meteorological and environmental monitoring data are used as preliminary verification evidence, and the results of fault section determination are generated by coordinating conflicts of multi-source evidence. Location execution module: Based on the fault section determination results, extract the theoretical traveling wave characteristics of the fault section, and realize the power fault location of the target power grid through traveling wave feature matching.

2. The power fault collaborative sensing and precise location system based on data fusion according to claim 1, characterized in that, Each monitoring node extracts theoretical response features from the fault theoretical response spectrum and integrates electrical monitoring data and equipment status data. Through feature matching and confidence weighted analysis, a local fault diagnosis conclusion is generated, including the following steps: S310. Each monitoring node retrieves and obtains the theoretical response characteristic data corresponding to its own monitoring node from the fault theoretical response map. S320. Calculate the waveform similarity between the three-phase current waveform data in the electrical monitoring data and the theoretical three-phase current waveform data in the theoretical response characteristic data to obtain a steady-state matching index; calculate the propagation timing and waveform consistency between the high-frequency traveling wave transient data and the high-frequency traveling wave transient theoretical data to obtain a transient matching index; perform differential fusion calculation between the steady-state matching index and the transient matching index based on the fault electrical characteristics represented by the theoretical electrical response data to generate a comprehensive matching score corresponding to each theoretical electrical response data, and sort and filter according to the comprehensive matching score to generate a preliminary fault hypothesis sequence; S330: Each monitoring node calls up the device status data and combines it with the associated distributed power supply operation data to perform a confidence-weighted correction on the preliminary fault hypothesis sequence and generate a local fault diagnosis conclusion.

3. The power fault collaborative sensing and precise location system based on data fusion according to claim 2, characterized in that, The evidence network is constructed based on the local fault diagnosis conclusions of each monitoring node and the dynamic topology model, including the following steps: S410. Based on the dynamic topology model and the local fault diagnosis conclusions of each monitoring node, construct an evidence network of topological constraints; S420. In the evidence network, each evidence network node iteratively exchanges support information with adjacent evidence network nodes along the connection path, and dynamically adjusts the weight of the support information in the propagation based on the dynamic topology model and the associated distributed power source operation data to generate preliminary coordinated evidence.

4. The power fault collaborative sensing and precise location system based on data fusion according to claim 3, characterized in that, Using meteorological and environmental monitoring data as prior evidence, and by coordinating conflicts among multi-source evidence, a fault section determination result is generated, including the following steps: S430. Generate prior meteorological events based on meteorological and environmental monitoring data; correct the support information of associated line edges in the dynamic topology model based on the prior meteorological events, and generate the corrected global evidence distribution. S440. Identify the fault segment with the highest support information from the global evidence distribution and generate the fault segment determination result.

5. The power fault collaborative sensing and precise location system based on data fusion according to claim 4, characterized in that, Based on the fault section determination results, theoretical traveling wave characteristics of the fault section are extracted, and power fault location of the target power grid is achieved through traveling wave characteristic matching, including the following steps: S510. Based on the fault section determination result, extract the high-frequency traveling wave transient theoretical data of each monitoring node in the corresponding section from the fault theoretical response spectrum, and use it as the theoretical traveling wave feature of the fault section. S520. Based on the theoretical traveling wave characteristics of the fault section, match and analyze the high-frequency traveling wave transient data of the monitoring nodes in the fault section to determine the location coordinates of the fault point.

Images