Low-voltage table front fault diagnosis method and system based on knowledge graph reasoning

Abstract

The application discloses a low-voltage table front fault diagnosis method and system based on knowledge graph reasoning, belongs to the technical field of low-voltage table front fault detection, and through construction of a fault knowledge graph, scattered related multi-source data of a low-voltage power supply network is converted into a semantic network related to fault types and fault correlation characteristics, without relying on a complex topology model and a large amount of computing power, the cost threshold of comprehensive monitoring is greatly reduced, the monitoring coverage is improved, meanwhile, a fault chain is obtained through fault reasoning according to the knowledge graph, so that the reasoning process is traceable, the reliability of the diagnosis result is improved, finally, the diagnosis result is obtained according to the fault chain, high-risk users and fault types can be accurately locked, passive repair is changed into active early warning, and the operation stability of the low-voltage power supply network is improved.

Description

Technical Field

[0001] This invention relates to the field of low-pressure meter upstream fault detection technology, specifically to a low-pressure meter upstream fault diagnosis method and system based on knowledge graph reasoning. Background Technology

[0002] As the final link in the power system, the low-voltage distribution network directly affects the stability of users' electricity supply. With increasing service life, latent faults such as oxidation and loosening of pre-meter connections become frequent, easily leading to voltage drops, line overheating, and even power outages, severely impacting power supply reliability. Traditional diagnostic methods have significant shortcomings: firstly, they rely on manual inspections and user reports, a reactive approach lacking the ability to proactively identify latent faults; secondly, they require precise pre-meter topology models, relying on comprehensive records and substantial computing power, which is cost-limited and hinders comprehensive monitoring; and thirdly, they are insufficient in identifying hidden hazards within the sealed meter boxes, resulting in vague fault location and low maintenance efficiency. Furthermore, traditional machine learning diagnostic models suffer from a "black box" defect, with unexplainable reasoning processes that are difficult for grassroots maintenance personnel to accept. Against this backdrop, there is an urgent need for an efficient and cost-effective diagnostic method.

[0003] Chinese Patent Publication No. CN115291035A discloses a method and system for detecting line faults in a low-voltage distribution transformer substation. The method involves setting up several line fault detection devices on the low-voltage distribution transformer substation line to detect faults. A fault acquisition unit then acquires fault information from the two nearest line fault detection devices on either side of the fault on the low-voltage distribution transformer substation line. A fault information analysis unit analyzes and verifies the fault information uploaded by the two line fault detection devices, and, based on different types of fault signals, analyzes and verifies the fault information uploaded by the two line fault detection devices. The system marks fault information; filters the fault information after the fault information analysis unit marks the signal to remove false alarms and erroneous fault information; makes a preliminary judgment on the information filtered out by the fault information filtering unit and gives the judgment result; and determines the fault information and gives a processing solution through the fault determination unit. Although it does not use a topology model and does not rely on a lot of computing power, it only compares the information detected by the nearest fault acquisition unit and performs multi-level marking and filtering. To a certain extent, it can locate the area where the fault occurs, but it cannot determine the specific fault type, nor can it directly locate a certain device or user. The diagnostic accuracy is not high enough. Summary of the Invention

[0004] This invention addresses the problems of existing low-voltage meter-front fault diagnosis methods, which mostly rely on meter-front topology models, resulting in high diagnostic costs due to excessive computing power, difficulty in large-scale monitoring, and inability to determine specific fault types and accurately locate faults. It provides a low-voltage meter-front fault diagnosis method and system based on knowledge graph reasoning. By constructing a fault knowledge graph, it transforms scattered, multi-source data related to the low-voltage power supply network into a semantic network associated with fault types and fault correlation features. This eliminates the need for complex topology models and massive computing power, significantly reducing the cost threshold for comprehensive monitoring and improving monitoring coverage. Furthermore, fault reasoning based on the knowledge graph yields fault chains, making the reasoning process traceable and improving the reliability of diagnostic results. Finally, the diagnostic results obtained from the fault chains can accurately identify high-risk users and fault types, realizing a shift from passive emergency repair to proactive early warning and improving the operational stability of the low-voltage power supply network.

[0005] In a first aspect, one technical solution provided in this embodiment of the invention is: a low-voltage meter pre-fault diagnosis method based on knowledge graph reasoning, comprising the following steps: S1. Obtain real-time operating data, historical fault data, and environmental data of the low-voltage power supply network; S2. Extract features from historical fault data and corresponding environmental data to obtain fault association features and fault performance. Construct a fault knowledge graph based on the relationship between fault association features and fault performance. S3. Extract features from real-time operating data and corresponding environmental data to obtain real-time associated features; perform fault reasoning on real-time associated features based on fault knowledge graph to obtain candidate fault chains; perform confidence analysis on candidate fault chains to obtain target fault chains; S4. Based on the information in the target fault chain, identify high-risk users and fault types as diagnostic results.

[0006] This solution integrates three types of heterogeneous data—real-time operation, historical faults, and environmental data—breaking the data silo limitations of traditional diagnostics. It eliminates the need for complex topology models and massive computing power, significantly reducing the cost threshold for comprehensive low-voltage distribution network monitoring and improving monitoring coverage. By extracting features from historical data and constructing a fault knowledge graph, it identifies explicit rules and implicit associations of fault correlation features and fault manifestations, enabling accurate identification of hidden hazards such as oxidation of connectors inside meter boxes, which are difficult to detect using traditional methods. Candidate fault chains are generated through knowledge graph reasoning, and target fault chains are then selected through confidence analysis, making the diagnostic process traceable and explainable. This overcomes the "black box" defect of traditional machine learning models and increases the confidence of grassroots maintenance personnel in the diagnostic results. By identifying high-risk users and fault types through target fault chains, targeted maintenance can be guided, promoting a shift in low-voltage meter fault diagnosis from passive emergency repair to proactive early warning, optimizing maintenance resource allocation, and improving power supply reliability.

[0007] Preferably, the real-time operating data includes transformer area measurement data and user measurement data; the environmental data includes ambient temperature, weather type, user geographic information, and device node information. The transformer area measurement data includes transformer area voltage, transformer area current, and transformer area active power; the user measurement data includes user voltage, user current, and user active power; the user geographic information includes user ID and user location; the device node information includes device ID and device access node location. The historical fault data refers to the operational data during the fault period; The fault manifestations refer to the electrical characteristics of various faults that have appeared in historical fault data.

[0008] This solution integrates measurement data such as voltage, current, and active power of transformer substations and users to provide core electrical basis for fault feature extraction; by incorporating environmental data such as ambient temperature and weather type, it can uncover the implicit correlation between environmental factors and fault occurrence; by linking user geographic information, equipment node information, and historical fault data, it provides accurate data support for entity relationship modeling and fault chain reasoning in knowledge graphs, helping to accurately determine fault type and location, thereby improving diagnostic efficiency and accuracy.

[0009] Preferably, in S2, feature extraction is performed on historical fault data and environmental data for the corresponding time period to obtain fault-related features, including the following steps: In historical fault data, the voltage difference between the user's low point voltage and the transformer area's low point voltage is calculated to obtain the low point voltage difference, and the voltage difference between the user's high point voltage and the low point voltage is calculated to obtain the high point voltage difference. When the change in user current within time t is greater than or equal to the current deviation threshold, the transient resistance is obtained by calculating the ratio of the change in user voltage to the change in current. The transient ratio is obtained by calculating the proportion of the change in current to the user's rated current. The low-point voltage difference, high-point voltage difference, transient resistance, and transient ratio are used as electrical characteristics. The local fluctuation characteristics and long-term dependence characteristics are obtained by analyzing the current curves of user voltage and user current as time series characteristics; The correlation between ambient temperature and weather type and electrical characteristics is calculated to obtain environmental correlation characteristics; Electrical characteristics, timing characteristics, and environmental characteristics are used as fault-related characteristics.

[0010] In this scheme, the core electrical characteristics at the time of fault occurrence are accurately captured by calculating the low-point voltage difference, high-point voltage difference, transient resistance, and transient ratio, providing a reliable basis for fault correlation analysis. Local fluctuations and long-term dependent time-series characteristics are extracted by analyzing voltage and current curves to distinguish between progressive faults such as joint oxidation and intermittent faults such as poor line contact. Environmental correlation characteristics are generated by calculating the correlation between environmental factors and electrical characteristics to uncover the implicit correlation between conditions such as high temperature and high humidity and faults. These three types of characteristics together constitute fault correlation characteristics, providing accurate support for entity relationship modeling of the fault knowledge graph, improving the scientific nature of subsequent fault chain reasoning, and helping to identify hidden hazards within the meter box seal.

[0011] Preferably, in S2, a fault knowledge graph is constructed based on the correlation between fault association features and fault manifestations, including the following steps: The device is treated as the detection entity, and user information is bound to the corresponding device based on the subordinate relationship between each device and the corresponding user. Based on all the fault types that have occurred in each device in the historical fault data, each fault type is treated as a fault entity. Detection entities are deployed based on the positional relationship between each device. The detection entities are associated with the fault entities based on the correspondence between each device and the fault type. Using fault association features as feature entities, feature entities are associated with fault entities based on the numerical threshold discrimination relationship between various fault types and different fault association features; Using fault manifestations as symptom entities, a fault knowledge graph is obtained by associating symptom entities with fault entities based on the fault manifestations corresponding to various faults that have appeared in historical fault data.

[0012] This solution constructs a clear semantic framework for a fault knowledge graph by building four core entities: detection, fault, symptoms, and features. This provides a structured carrier for subsequent fault reasoning. By establishing the location, value, and association relationships between entities, scattered information such as user-device location, fault type-associated features, and fault-manifestation is transformed into a visual semantic network, thus breaking down data silos. Through structured graph association, explicit rules and implicit associations of features, faults, and symptoms are linked, automating the entire diagnostic process without human intervention. This solves the pain point of traditional diagnosis relying on human experience and provides accurate knowledge support for real-time feature matching and fault chain generation. It also improves the identification accuracy of hidden hazards such as joint oxidation and facilitates the rapid determination of fault type and location.

[0013] Preferably, in S3, candidate fault chains are obtained by performing fault reasoning on real-time associated features based on the fault knowledge graph, including the following steps: Real-time associated features are mapped to feature entities. Starting from the mapped feature entities, the possible fault entities, symptom entities, and detection entities are traced based on numerical threshold discrimination relationships, fault manifestations corresponding to various faults, and the correspondence between each device and fault type to obtain candidate fault chains.

[0014] In this solution, by mapping real-time associated features to feature entities in a knowledge graph, accurate matching between real-time operational data and historical fault knowledge is achieved. By tracing the numerical, association, and positional relationships between features, faults, symptoms, and detection entities, a complete candidate fault chain is generated, making the fault reasoning process traceable and explainable. This breaks through the "black box" limitation of traditional diagnosis, clarifies the potential association paths of faults, lays a solid foundation for subsequent confidence analysis and accurate location, and significantly improves the efficiency of fault investigation.

[0015] As a preferred embodiment, in step S3, confidence analysis is performed on the candidate fault chains to obtain the target fault chain, including the following steps: Calculate the cosine similarity between the real-time correlation features and the fault correlation features in each candidate fault chain; Obtain the historical fault frequency and historical inference accuracy of each candidate fault chain in the historical maintenance records; The confidence level of the failure chain is calculated based on cosine similarity, historical failure frequency, and historical inference accuracy. The fault chain with the highest confidence among the candidate fault chains is selected as the target fault chain.

[0016] This solution combines the cosine similarity of real-time and historical features, historical fault frequency, and inference accuracy to calculate the confidence of candidate fault chains from multiple dimensions, and then selects the target fault chain with the highest confidence. This effectively solves the pain point of ambiguous traditional diagnostic results, improves the accuracy of fault judgment, reduces misjudgments and omissions, and provides a reliable basis for on-site maintenance, thus helping to accurately locate and efficiently handle faults before low-voltage meters.

[0017] Preferably, in S4, high-risk users and fault types are identified as diagnostic results based on information in the target fault chain, including the following steps: The information in the fault chain includes fault detection entities, fault entities, and symptom entities; Based on the fault detection entity, obtain the ID and location information of the faulty device and high-risk user; based on the fault entity, determine the corresponding fault type; based on the symptom entity, determine the fault manifestation. The diagnostic results are obtained by taking the ID and location information of the faulty device and high-risk user, as well as the fault type and fault manifestation, and outputting them in a structured form.

[0018] This solution analyzes the detection, fault, and symptom entities of the target fault chain to accurately pinpoint high-risk user IDs, locations, fault types, and manifestations. By providing structured diagnostic results, it allows maintenance personnel to understand the situation at a glance and directly carry out targeted repairs, avoiding indiscriminate inspections, significantly improving maintenance efficiency, and facilitating proactive early warning and handling of faults before low-voltage meters.

[0019] Secondly, one technical solution provided in this embodiment of the invention is: a low-voltage meter pre-fault diagnosis system based on knowledge graph reasoning, including a data acquisition module, a feature extraction module, a knowledge graph construction module, a reasoning module, a confidence analysis module, and a diagnosis module; The data acquisition module acquires real-time operating data, historical fault data, and environmental data of the low-voltage power supply network; The feature extraction module extracts features from historical fault data and environmental data of the corresponding time period to obtain fault-related features and fault performance, and extracts features from real-time operating data and environmental data of the corresponding time period to obtain real-time related features. The graph construction module constructs a fault knowledge graph based on the correlation between fault association features and fault manifestations. The reasoning module performs fault reasoning on real-time associated features based on the fault knowledge graph to obtain candidate fault chains. The confidence analysis module performs confidence analysis on the candidate fault chains to obtain the target fault chain; The diagnostic module identifies high-risk users and fault types based on information in the target fault chain as diagnostic results.

[0020] In this solution, a corresponding system is built to integrate the low-pressure meter fault diagnosis method, thereby realizing human-computer interaction and improving the user experience.

[0021] Thirdly, one technical solution provided in this embodiment of the invention is: a computer device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory is used to store computer programs; the processor is used to execute the program stored in the memory to implement the steps of the low-voltage meter pre-fault diagnosis method based on knowledge graph reasoning.

[0022] Fourthly, one technical solution provided in this embodiment of the invention is: a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the steps of a low-voltage meter pre-fault diagnosis method based on knowledge graph reasoning.

[0023] The beneficial effects of this invention are as follows: By constructing a fault knowledge graph, this invention transforms scattered, multi-source data related to low-voltage power supply networks into a semantic network associated with fault types and fault correlation characteristics. This eliminates the need for complex topology models and massive computing power, significantly reducing the cost threshold for comprehensive monitoring and improving monitoring coverage. Simultaneously, fault reasoning based on the knowledge graph yields fault chains, making the reasoning process traceable and improving the reliability of diagnostic results. Finally, based on the fault chains, diagnostic results can accurately identify high-risk users and fault types, realizing a shift from passive emergency repairs to proactive early warning, thereby improving the operational stability of low-voltage power supply networks.

[0024] The above description of the invention is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0025] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings. The drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings.

[0026] Figure 1 This is a flowchart of the low-pressure meter pre-fault diagnosis method based on knowledge graph reasoning of the present invention; Figure 2 This is a partial schematic diagram of the fault knowledge graph in this embodiment; Figure 3 This is a schematic diagram of the low-pressure meter pre-fault diagnosis system based on knowledge graph reasoning according to the present invention; Figure 4 This is a schematic diagram of a computer device provided by the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only one preferred embodiment of this invention and are only used to explain this invention. They do not limit the scope of protection of this invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0028] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations (or steps) as sequential processes, many of the operations (or steps) can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but it may also have additional steps not included in the figures; the process may correspond to a method, function, procedure, subroutine, subroutine, etc.

[0029] Example 1: To address the problem that most existing low-voltage meter-front fault diagnosis methods rely on meter-front topology models, resulting in high diagnostic costs and difficulty in large-scale monitoring due to excessive computing power, this example provides a low-voltage meter-front fault diagnosis method based on knowledge graph reasoning, such as... Figure 1 As shown, it includes the following steps: S1: Acquire real-time operating data, historical fault data, and environmental data of the low-voltage power supply network.

[0030] In this embodiment, the real-time operating data includes station area measurement data and user measurement data; the environmental data includes ambient temperature, weather type, user geographic information, and device node information. The transformer area measurement data includes transformer area voltage, transformer area current, and transformer area active power; the user measurement data includes user voltage, user current, and user active power; the user geographic information includes user ID and user location; the device node information includes device ID and device access node location. The historical fault data refers to the operational data during the fault period; The fault manifestations refer to the electrical characteristics of various faults that have appeared in historical fault data, including voltage sags, resistance jumps / drops, current surges, and multi-user voltage drops.

[0031] This embodiment integrates measurement data such as voltage, current, and active power of the transformer area and users to provide core electrical basis for fault feature extraction; by incorporating environmental data such as ambient temperature and weather type, it can uncover the implicit correlation between environmental factors and fault occurrence; by associating user geographic information, equipment node information, and historical fault data, it provides accurate data support for entity relationship modeling and fault chain reasoning in knowledge graphs, helping to achieve accurate determination of fault type and location, thereby improving diagnostic efficiency and accuracy.

[0032] S2: Extract features from historical fault data and corresponding environmental data to obtain fault association features and fault performance. Construct a fault knowledge graph based on the relationship between fault association features and fault performance.

[0033] In this embodiment, feature extraction is performed on historical fault data and environmental data for the corresponding time period to obtain fault-related features, including the following steps: In historical fault data, the voltage difference between the user's low point voltage and the transformer area's low point voltage is calculated to obtain the low point voltage difference, and the voltage difference between the user's high point voltage and the low point voltage is calculated to obtain the high point voltage difference. When the change in user current within time t is greater than or equal to the current deviation threshold, the transient resistance is obtained by calculating the ratio of the change in user voltage to the change in current. The transient ratio is obtained by calculating the proportion of the change in current to the user's rated current. The low-point voltage difference, high-point voltage difference, transient resistance, and transient ratio are used as electrical characteristics. The local fluctuation characteristics and long-term dependence characteristics are obtained by analyzing the current curves of user voltage and user current as time series characteristics; The correlation between ambient temperature and weather type and electrical characteristics is calculated to obtain environmental correlation characteristics; Electrical characteristics, timing characteristics, and environmental characteristics are used as fault-related characteristics.

[0034] Specifically, low point voltage difference High point voltage difference transient resistance , Transient ratio ;in, For users' low point voltage, This is the low point voltage in the transformer area. For users' high point voltage, Let be the change in current over time t, and let the change be greater than or equal to the current deviation threshold. Let be the user current at time t. The user current at time t=0, The rated current for the user.

[0035] Local fluctuation features can be extracted using one-dimensional convolution kernels of convolutional neural networks, such as the abrupt change slope at the onset of voltage transients and the stability during the recovery phase. Long-term dependent features can be extracted using the gating capture mechanism of long short-term memory neural networks, such as the gradual trend of voltage changes and the periodicity of current fluctuations.

[0036] Environmental correlation characteristics are used to calculate the correlation between environmental factors and electrical characteristics, such as the rate of change of transient resistance under high temperature and high humidity conditions (R0). 雨天 -R 晴天 ) / R 晴天 These are correlation factors that can demonstrate the relationship between environmental factors and electrical characteristics.

[0037] This embodiment calculates low-point voltage difference, high-point voltage difference, transient resistance, and transient ratio to accurately capture the core electrical characteristics at the time of fault occurrence, providing a reliable basis for fault correlation analysis. It extracts local fluctuations and long-term dependent time-series features by analyzing voltage and current curves to distinguish between progressive faults such as joint oxidation and intermittent faults such as poor line contact. It generates environmental correlation features by calculating the correlation between environmental factors and electrical characteristics to uncover implicit correlations between conditions such as high temperature and high humidity and faults. These three types of features together constitute fault correlation features, providing precise support for entity relationship modeling in the fault knowledge graph, improving the scientific rigor of subsequent fault chain reasoning, and helping to identify hidden hazards within the meter box seal.

[0038] In this embodiment, a fault knowledge graph is constructed based on the correlation between fault association features and fault manifestations, including the following steps: The device is treated as the detection entity, and user information is bound to the corresponding device based on the subordinate relationship between each device and the corresponding user. Based on all the fault types that have occurred in each device in the historical fault data, each fault type is treated as a fault entity. Detection entities are deployed based on the positional relationship between each device. The detection entities are associated with the fault entities based on the correspondence between each device and the fault type. Using fault association features as feature entities, feature entities are associated with fault entities based on the numerical threshold discrimination relationship between various fault types and different fault association features; Using fault manifestations as symptom entities, a fault knowledge graph is obtained by associating symptom entities with fault entities based on the fault manifestations corresponding to various faults that have appeared in historical fault data.

[0039] Specifically, the detection entities include public transformers, meters, connectors, wires, junction boxes, low-voltage users, and transformer substations; fault entities include connector oxidation, loose connectors, broken neutral wires, and terminal overloads; symptom entities include voltage dips, resistance surges / dips, current surges, and multi-user voltage drops; among them, voltage dips are defined as user voltage remaining below the rated voltage for a duration t1; resistance surges / dips are defined as transient resistance exceeding the normal range; current surges are defined as user current changes exceeding the current change threshold within a time period t2; and multi-user voltage drops are defined as voltage drops exceeding the voltage change threshold for more than n users simultaneously within a time period t3. In addition to fault-related features, auxiliary features can be added to improve the accuracy of fault type screening, such as connector materials, equipment operating time, and specific power supply equipment, for further screening of specific fault types.

[0040] This embodiment constructs four core entities—detection, fault, symptom, and feature—to build a clear semantic framework for a fault knowledge graph, providing a structured carrier for subsequent fault reasoning. By establishing the location, value, and association relationships between entities, it transforms scattered information such as user-device location, fault type-associated features, and fault-manifestation into a visual semantic network, thereby breaking down data silos. Through structured graph association, it automates the entire diagnostic process by linking explicit rules and implicit associations of features, faults, and symptoms, eliminating the need for human intervention. This addresses the pain point of traditional diagnosis relying on human experience and provides accurate knowledge support for real-time feature matching and fault chain generation, improving the accuracy of identifying hidden hazards such as joint oxidation and facilitating the rapid determination of fault type and location.

[0041] S3: Extract features from real-time operating data and corresponding environmental data to obtain real-time associated features; perform fault reasoning on real-time associated features based on fault knowledge graph to obtain candidate fault chains; perform confidence analysis on candidate fault chains to obtain target fault chains.

[0042] In this embodiment, real-time associated features are mapped to feature entities. Starting from the mapped feature entities, possible fault entities, symptom entities, and detection entities are traced based on numerical threshold discrimination relationships, fault manifestations corresponding to various faults, and the correspondence between each device and fault type to obtain candidate fault chains.

[0043] Specifically, each fault chain is composed of an arrangement of fault manifestations, fault types, and fault locations, for example: Figure 2 As shown, the symptom entities are single user voltage drop >15V and multiple user voltage drop >10V; the characteristic entities are transient resistance >6Ω, transient ratio >30%, and operating years; the fault entities are connector oxidation, loose connectors, and poor contact in the junction box; the detection entities are the inside of the meter box and the junction box. In one instance, a real-time fault characteristic was detected as a user voltage drop greater than 15V and transient resistance greater than 6Ω. This was imported into the characteristic entities, resulting in fault chain 1: user voltage drop greater than 15V + transient resistance greater than 6Ω → connector oxidation → inside the meter box. Further filtering by querying the equipment's operating years yielded fault chain 2: user voltage drop greater than 15V + transient resistance greater than 6Ω + operating years greater than 10 years → loose connector → inside the meter box. For another example, if multiple user voltage drops >10V and transient ratio >30% were detected, and the entire low-voltage distribution network uses junction box power supply, then fault chain 3 was obtained: multiple users voltage drops greater than 10V. +Distribution box power supply +Transient ratio >30% →Distribution box poor contact →Distribution box.

[0044] This embodiment achieves accurate matching between real-time operational data and historical fault knowledge by mapping real-time associated features to feature entities in a knowledge graph. By tracing the numerical, association, and positional relationships between features, faults, symptoms, and detection entities, a complete candidate fault chain is generated, making the fault reasoning process traceable and explainable. This breaks through the "black box" limitation of traditional diagnosis, clarifies the potential association paths of faults, lays a solid foundation for subsequent confidence analysis and accurate location, and significantly improves fault troubleshooting efficiency.

[0045] In this embodiment, the target fault chain is obtained by performing confidence analysis on the candidate fault chains, including the following steps: Calculate the cosine similarity between the real-time correlation features and the fault correlation features in each candidate fault chain; Obtain the historical fault frequency and historical inference accuracy of each candidate fault chain in the historical maintenance records; The confidence level of the failure chain is calculated based on cosine similarity, historical failure frequency, and historical inference accuracy. The fault chain with the highest confidence among the candidate fault chains is selected as the target fault chain.

[0046] Specifically, the formula for calculating the confidence level is: in, For cosine similarity, Historical failure frequency For historical reasoning accuracy, , and Weights are assigned to the confidence levels, which can be adjusted according to actual needs. Finally, based on the calculation, fault chain 1 has the highest confidence level, so the final diagnosis is that there is an oxidation fault at the copper-aluminum connector inside the meter box, manifested as a voltage drop of 15V and a transient resistance of 6Ω.

[0047] This embodiment combines the cosine similarity of real-time and historical features, historical fault frequency, and inference accuracy to calculate the confidence of candidate fault chains from multiple dimensions, and selects the target fault chain with the highest confidence. This effectively solves the pain point of vague traditional diagnostic results, improves the accuracy of fault judgment, reduces misjudgments and omissions, and provides a reliable basis for on-site maintenance, thus helping to accurately locate and efficiently handle faults before low-voltage meters.

[0048] S4: Identify high-risk users and fault types based on information in the target fault chain as diagnostic results.

[0049] In this embodiment, determining high-risk users and fault types based on information in the target fault chain as diagnostic results includes the following steps: The information in the fault chain includes fault detection entities, fault entities, and symptom entities; Based on the fault detection entity, obtain the ID and location information of the faulty device and high-risk user; based on the fault entity, determine the corresponding fault type; based on the symptom entity, determine the fault manifestation. The diagnostic results are obtained by taking the ID and location information of the faulty device and high-risk user, as well as the fault type and fault manifestation, and outputting them in a structured form.

[0050] This embodiment can accurately pinpoint high-risk user IDs, locations, fault types, and manifestations by analyzing the detection, faults, and symptom entities of the target fault chain. By outputting the diagnostic results in a structured manner, maintenance personnel can easily understand the situation and directly carry out targeted repairs, avoiding indiscriminate inspections, significantly improving maintenance efficiency, and helping to achieve proactive early warning and handling of faults before low-voltage meters.

[0051] Example 2: This example also provides a low-voltage meter pre-fault diagnosis system based on knowledge graph reasoning, such as... Figure 3 As shown, it includes a data acquisition module, a feature extraction module, a map construction module, an inference module, a confidence analysis module, and a diagnostic module; The data acquisition module acquires real-time operating data, historical fault data, and environmental data of the low-voltage power supply network; The feature extraction module extracts features from historical fault data and environmental data of the corresponding time period to obtain fault-related features and fault performance, and extracts features from real-time operating data and environmental data of the corresponding time period to obtain real-time related features. The graph construction module constructs a fault knowledge graph based on the correlation between fault association features and fault manifestations. The reasoning module performs fault reasoning on real-time associated features based on the fault knowledge graph to obtain candidate fault chains. The confidence analysis module performs confidence analysis on the candidate fault chains to obtain the target fault chain; The diagnostic module identifies high-risk users and fault types based on information in the target fault chain as diagnostic results.

[0052] This embodiment integrates the low-pressure meter pre-fault diagnosis method in this solution by constructing a corresponding system, realizing human-computer interaction and improving the user experience.

[0053] This embodiment also provides a computer device, such as... Figure 4 As shown, it includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory is used to store computer programs; When the processor executes the program stored in memory, it implements a low-voltage meter fault diagnosis method based on knowledge graph reasoning.

[0054] The communication bus mentioned in the above electronic devices can be a Peripheral Component Interconnect Standard (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0055] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0056] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0057] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0058] This application also provides a computer-readable storage medium storing a computer program, which is executed by a processor as a low-voltage meter front fault diagnosis method based on knowledge graph reasoning.

[0059] As can be seen from the above embodiments, it has at least the following substantial effects: (1) By integrating three types of heterogeneous data—real-time operation, historical faults, and environment—this invention breaks the data silo limitation of traditional diagnosis, eliminates the need for complex topology models and large computing power, thereby significantly reducing the cost threshold for comprehensive monitoring of low-voltage distribution networks and improving monitoring coverage. (2) By extracting features from historical data and constructing a fault knowledge graph, this invention has accumulated explicit rules and implicit associations of fault association features and fault manifestations, thereby accurately identifying hidden hazards that are difficult to detect by traditional methods, such as oxidation of connectors inside the meter box seal. (3) This invention generates candidate fault chains through knowledge graph reasoning, and then filters target fault chains through confidence analysis, making the diagnosis process traceable and explainable, solving the "black box" defect of traditional machine learning models, and improving the confidence of grassroots operation and maintenance personnel in the diagnosis results; (4) This invention identifies high-risk users and fault types by targeting fault chains, thereby guiding targeted maintenance, promoting the transformation of low-voltage meter fault diagnosis from passive emergency repair to proactive early warning, optimizing the allocation of operation and maintenance resources, and improving power supply reliability.

[0060] The specific embodiments described above are preferred embodiments of the present invention and are not intended to limit the specific scope of the present invention. The scope of the present invention includes, but is not limited to, these specific embodiments. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A method for diagnosing faults before low-voltage meters based on knowledge graph reasoning, characterized in that: Includes the following steps: S1. Obtain real-time operating data, historical fault data, and environmental data of the low-voltage power supply network; S2. Extract features from historical fault data and corresponding environmental data to obtain fault association features and fault performance. Construct a fault knowledge graph based on the relationship between fault association features and fault performance. S3. Extract features from real-time operational data and corresponding time-period environmental data to obtain real-time related features; Candidate fault chains are obtained by performing fault reasoning on real-time associated features based on fault knowledge graphs. The target fault chain is obtained by performing confidence analysis on the candidate fault chains; S4. Based on the information in the target fault chain, identify high-risk users and fault types as diagnostic results.

2. The low-voltage meter pre-fault diagnosis method based on knowledge graph reasoning according to claim 1, characterized in that: The real-time operational data includes transformer area measurement data and user measurement data; the environmental data includes ambient temperature, weather type, user geographic information, and device node information. The transformer area measurement data includes transformer area voltage, transformer area current, and transformer area active power; the user measurement data includes user voltage, user current, and user active power; the user geographic information includes user ID and user location; the device node information includes device ID and device access node location. The historical fault data refers to the operational data during the fault period; The fault manifestations refer to the electrical characteristics of various faults that have appeared in historical fault data.

3. The low-voltage meter pre-fault diagnosis method based on knowledge graph reasoning according to claim 2, characterized in that: In S2, feature extraction is performed on historical fault data and environmental data for the corresponding time period to obtain fault correlation features, including the following steps: In historical fault data, the voltage difference between the user's low point voltage and the transformer area's low point voltage is calculated to obtain the low point voltage difference, and the voltage difference between the user's high point voltage and the low point voltage is calculated to obtain the high point voltage difference. When the change in user current within time t is greater than or equal to the current deviation threshold, the transient resistance is obtained by calculating the ratio of the change in user voltage to the change in current. The transient ratio is obtained by calculating the proportion of the change in current to the user's rated current. The low-point voltage difference, high-point voltage difference, transient resistance, and transient ratio are used as electrical characteristics. The local fluctuation characteristics and long-term dependence characteristics are obtained by analyzing the current curves of user voltage and user current as time series characteristics; The correlation between ambient temperature and weather type and electrical characteristics is calculated to obtain environmental correlation characteristics; Electrical characteristics, timing characteristics, and environmental characteristics are used as fault-related characteristics.

4. The low-voltage meter pre-fault diagnosis method based on knowledge graph reasoning according to claim 2, characterized in that: In S2, a fault knowledge graph is constructed based on the relationship between fault association features and fault manifestations, including the following steps: The device is treated as the detection entity, and user information is bound to the corresponding device based on the subordinate relationship between each device and the corresponding user. Based on all the fault types that have occurred in each device in the historical fault data, each fault type is treated as a fault entity. Detection entities are deployed based on the positional relationship between each device. The detection entities are associated with the fault entities based on the correspondence between each device and the fault type. Using fault association features as feature entities, feature entities are associated with fault entities based on the numerical threshold discrimination relationship between various fault types and different fault association features; Using fault manifestations as symptom entities, a fault knowledge graph is obtained by associating symptom entities with fault entities based on the fault manifestations corresponding to various faults that have appeared in historical fault data.

5. The low-voltage meter pre-fault diagnosis method based on knowledge graph reasoning according to claim 4, characterized in that: In S3, candidate fault chains are obtained by performing fault reasoning on real-time associated features based on the fault knowledge graph, including the following steps: Real-time associated features are mapped to feature entities. Starting from the mapped feature entities, the possible fault entities, symptom entities, and detection entities are traced based on numerical threshold discrimination relationships, fault manifestations corresponding to various faults, and the correspondence between each device and fault type to obtain candidate fault chains.

6. The low-voltage meter pre-fault diagnosis method based on knowledge graph reasoning according to claim 1, characterized in that: In S3, confidence analysis is performed on the candidate fault chains to obtain the target fault chain, including the following steps: Calculate the cosine similarity between the real-time correlation features and the fault correlation features in each candidate fault chain; Obtain the historical fault frequency and historical inference accuracy of each candidate fault chain in the historical maintenance records; The confidence level of the failure chain is calculated based on cosine similarity, historical failure frequency, and historical inference accuracy. The fault chain with the highest confidence among the candidate fault chains is selected as the target fault chain.

7. The low-voltage meter pre-fault diagnosis method based on knowledge graph reasoning according to claim 4, characterized in that: In S4, high-risk users and fault types are identified as diagnostic results based on information in the target fault chain, including the following steps: The information in the fault chain includes fault detection entities, fault entities, and symptom entities; Based on the fault detection entity, obtain the ID and location information of the faulty device and high-risk user; based on the fault entity, determine the corresponding fault type; based on the symptom entity, determine the fault manifestation. The diagnostic results are obtained by taking the ID and location information of the faulty device and high-risk user, as well as the fault type and fault manifestation, and outputting them in a structured form.

8. A low-pressure meter pre-fault diagnosis system based on knowledge graph reasoning, applicable to the low-pressure meter pre-fault diagnosis method based on knowledge graph reasoning as described in any one of claims 1-7, characterized in that: It includes a data acquisition module, a feature extraction module, a graph construction module, an inference module, a confidence analysis module, and a diagnostic module; The data acquisition module acquires real-time operating data, historical fault data, and environmental data of the low-voltage power supply network; The feature extraction module extracts features from historical fault data and environmental data of the corresponding time period to obtain fault-related features and fault performance, and extracts features from real-time operating data and environmental data of the corresponding time period to obtain real-time related features. The graph construction module constructs a fault knowledge graph based on the correlation between fault association features and fault manifestations. The reasoning module performs fault reasoning on real-time associated features based on the fault knowledge graph to obtain candidate fault chains. The confidence analysis module performs confidence analysis on the candidate fault chains to obtain the target fault chain; The diagnostic module identifies high-risk users and fault types based on information in the target fault chain as diagnostic results.

9. A computer device, characterized in that: The system includes a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other via the communication bus. The memory is used to store computer programs. When the processor executes the program stored in the memory, it implements the steps of the low-voltage meter pre-fault diagnosis method based on knowledge graph reasoning as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the low-voltage meter pre-fault diagnosis method based on knowledge graph reasoning as described in any one of claims 1-7.

Images