Power distribution cabinet fault area detection method and system based on AI

Abstract

The invention discloses an AI-based power distribution cabinet fault area detection method and system, and relates to the field of power distribution cabinet detection. The method is applied to a power distribution cabinet monitoring system, and comprises the following steps: acquiring multi-modal monitoring data of a target power distribution cabinet, and generating an operation condition vector of the target power distribution cabinet; inputting the operation condition vector into a preset typical condition vector library for condition decomposition to obtain a plurality of sub-operation condition vectors; inputting the plurality of sub-operation condition vectors into a preset power distribution cabinet fault detection model to obtain fault information corresponding to the plurality of sub-operation condition vectors; inputting the plurality of pieces of fault information into a preset fault reasoning knowledge graph to obtain a plurality of fault paths; screening from the plurality of fault paths to obtain a plurality of target fault paths, and identifying starting nodes in the plurality of target fault paths; and starting nodes in the multiple target fault paths are output, and fault early warning is generated, so that maintenance personnel can be reminded of timely maintenance. The problem of low fault detection precision of a current power distribution cabinet is solved.

Description

Technical Field

[0001] This application relates to the field of power distribution cabinet testing, specifically to an AI-based method and system for detecting fault areas in power distribution cabinets. Background Technology

[0002] As the core equipment for power distribution, control, and protection in power systems, distribution cabinets are widely used in various scenarios such as industrial production, residential buildings, and public facilities. Their operational stability directly affects the safe and reliable power supply of the power system, as well as the safety of personnel and equipment. Therefore, achieving accurate and rapid detection of fault areas in distribution cabinets is of great practical significance.

[0003] Currently, with the rapid development of artificial intelligence technology, AI, with its powerful self-learning ability and efficient fault detection capability, is gradually replacing traditional power distribution cabinet fault detection methods such as manual inspection and single parameter threshold judgment. Specifically, the existing AI-based power distribution cabinet fault area detection method is based on deploying multimodal sensors at key locations in the power distribution cabinet to collect operating data such as temperature, current, vibration, and partial discharge. Then, using an AI model trained with a large amount of historical data under typical operating conditions, the collected data is analyzed and processed to identify the fault type and locate the fault area.

[0004] However, in practical applications, the operating conditions of power distribution cabinets are complex and continuously change dynamically with factors such as load changes, changes in environmental conditions, and increases in operating time. As a result, the typical operating conditions trained by the AI ​​model gradually become difficult to match the operating environment of the power distribution cabinet, which in turn leads to a decrease in the accuracy of fault detection results. Summary of the Invention

[0005] To address the issue of low fault detection accuracy in current power distribution cabinets, this application provides an AI-based method and system for detecting fault areas in power distribution cabinets.

[0006] Firstly, this application provides an AI-based method for detecting fault areas in power distribution cabinets, applied in a power distribution cabinet monitoring system. The method includes:

[0007] Acquire multimodal monitoring data of the target power distribution cabinet, extract operating condition features from the multimodal monitoring data, and generate an operating condition vector of the target power distribution cabinet;

[0008] The operating condition vector is input into a preset typical operating condition vector library for operating condition decomposition to obtain multiple sub-operating condition vectors.

[0009] Multiple sub-operating condition vectors are input into a preset power distribution cabinet fault detection model to obtain fault information corresponding to each of the multiple sub-operating condition vectors. The fault information includes faulty components, fault types, and fault confidence levels.

[0010] Multiple fault information entries are input into a preset fault reasoning knowledge graph to obtain multiple fault paths. Each fault path includes multiple fault nodes, and each fault node corresponds to a fault information entry.

[0011] Multiple target fault paths are obtained by filtering from the multiple fault paths, and the starting node in the multiple target fault paths is identified;

[0012] Output the starting node of multiple target fault paths and generate a fault warning to remind maintenance personnel to carry out timely repairs.

[0013] Optionally, the preset typical working condition vector library is constructed from multiple typical working condition vectors, wherein each typical working condition vector corresponds to a working condition matching degree threshold determined by the number of working condition dimensions and the working condition type, specifically as follows:

[0014] The multiple typical operating condition vectors are divided into multiple operating condition type groups according to multiple preset operating condition types, including load type, environment type and equipment status type;

[0015] Obtain historical working condition data of multiple typical working condition vectors in the first working condition type group, wherein the first working condition type group is any one of the multiple working condition type groups;

[0016] Calculate the pairwise similarity between historical working condition data of various typical working condition vectors in the first working condition type group to obtain the historical working condition data similarity set of the first working condition type group.

[0017] Calculate the mean similarity of the historical working condition data similarity set, and use the mean similarity as the matching degree benchmark threshold for any typical working condition vector in the first working condition type group;

[0018] Based on the number of working condition dimensions of the first typical working condition vector, the matching degree benchmark threshold is adjusted using a preset working condition matching degree threshold adjustment formula. This yields the working condition matching degree threshold for the first typical working condition vector, which is any one of multiple typical working condition vectors in the first working condition type group.

[0019] Optionally, the step of inputting the operating condition vector into a preset typical operating condition vector table for operating condition decomposition to obtain multiple sub-operating condition vectors specifically involves:

[0020] The matching degree between the operating condition vector and the multiple typical operating condition vectors is calculated to obtain the operating condition matching degree between the operating condition vector and the multiple typical operating condition vectors.

[0021] The operating condition matching degree between the second typical operating condition vector and the operating condition vector is compared with the operating condition matching degree threshold corresponding to the second typical operating condition vector, wherein the second typical operating condition vector is any one of the multiple typical operating condition vectors;

[0022] If the working condition matching degree between the second typical working condition vector and the operating working condition vector is greater than or equal to the working condition matching degree threshold corresponding to the second typical working condition vector, then based on the multiple feature dimensions of the second typical working condition vector, working condition features consistent with the multiple features of the second typical working condition vector are extracted from the operating working condition vector and reconstructed into a sub-operating working condition vector.

[0023] Optionally, the preset fault reasoning knowledge graph consists of multiple fault nodes, causal association edges between the multiple fault nodes, and a conditional probability table on each causal association edge. Each fault node represents a fault mode of an electrical component in the target power distribution rule, each causal association edge represents the fault propagation causal relationship between the two fault nodes, and the conditional probability table stores the conditional probability information of the fault nodes at both ends of the causal association edge.

[0024] Optionally, inputting multiple fault information entries into a preset fault reasoning knowledge graph to obtain multiple fault paths specifically involves:

[0025] Based on the fault type and fault component of multiple fault information, the multiple fault information is mapped to the preset fault reasoning knowledge graph to obtain multiple mapped fault nodes;

[0026] Taking multiple mapped fault nodes as the starting point for tracing the source, and based on the causal association edges between each fault node in the preset fault reasoning knowledge graph and the conditional probability table on each causal association edge, the fault source is traced along the causal association edges to obtain multiple fault propagation links between multiple mapped fault nodes.

[0027] The multiple fault propagation links are output to obtain multiple fault paths.

[0028] Optionally, the step of filtering multiple target fault paths from the multiple fault paths specifically involves:

[0029] The multiple fault nodes in the first fault path are divided into multiple measured fault nodes and multiple hypothetical fault nodes, and the first fault path is any one of the multiple fault paths.

[0030] Calculate the total number of fault nodes in the first fault path and the total number of measured fault nodes;

[0031] Calculate the mean fault confidence of multiple measured fault nodes in the first fault path;

[0032] The reliability score of the first fault path is determined based on the total number of fault nodes, the total number of measured fault nodes, and the average fault confidence.

[0033] If the reliability score of the first fault path is greater than or equal to a preset reliability score threshold, then the first fault path is determined to be the target fault path.

[0034] Optional,

[0035] Secondly, this application provides an AI-based power distribution cabinet fault area detection system. The system is a power distribution cabinet monitoring system, comprising an acquisition module, a processing module, and an output module, wherein:

[0036] The acquisition module is used to acquire multimodal monitoring data of the target power distribution cabinet, extract operating condition features from the multimodal monitoring data, and generate an operating condition vector of the target power distribution cabinet.

[0037] The processing module is used to input the operating condition vector into a preset typical operating condition vector library for operating condition decomposition to obtain multiple sub-operating condition vectors.

[0038] Multiple sub-operating condition vectors are input into a preset power distribution cabinet fault detection model to obtain fault information corresponding to each of the multiple sub-operating condition vectors. The fault information includes faulty components, fault types, and fault confidence levels.

[0039] Multiple fault information entries are input into a preset fault reasoning knowledge graph to obtain multiple fault paths. Each fault path includes multiple fault nodes, and each fault node corresponds to a fault information entry.

[0040] Multiple target fault paths are obtained by filtering from the multiple fault paths, and the starting node in the multiple target fault paths is identified;

[0041] The output module is used to output the starting node of multiple target fault paths and generate fault warnings to remind maintenance personnel to carry out timely repairs.

[0042] Thirdly, this application provides an electronic device including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of the first aspects.

[0043] Fourthly, this application provides a computer-readable storage medium storing instructions that, when executed, perform the method described in any one of the first aspects.

[0044] In summary, one or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0045] This application extracts operating condition features from the multimodal monitoring data of the target distribution cabinet under test to obtain operating condition vectors, thereby clarifying the current operating state of the target distribution cabinet. Then, the operating condition vectors are input into a preset typical operating condition vector space for operating condition decomposition, thereby breaking down the complex current operating state of the target distribution cabinet into sub-operating condition vectors of multiple typical operating conditions. At this point, the sub-operating condition vectors of multiple typical operating conditions are respectively input into a preset distribution cabinet fault detection model. Since the distribution cabinet fault detection model is trained on a large amount of historical data under typical operating conditions, relatively accurate fault information corresponding to each sub-operating condition vector can be obtained. At this time, the fault information includes the faulty component, fault type, and fault details. Confidence level. Then, since multiple fault information only characterizes the fault form, it is possible that a certain faulty component itself is not faulty, but is affected by the fault of the upstream faulty component, causing the faulty component to exhibit a faulty phenomenon. Therefore, in order to further determine the cause of the fault, this application inputs multiple fault information into a preset fault reasoning knowledge graph, sorts out the causal relationship of multiple fault information, obtains multiple fault paths, and then identifies the starting node in each fault path to determine the root cause node of each fault path. At this time, the root cause node is the cause of the fault in the distribution cabinet, thereby improving the fault detection accuracy of the distribution cabinet. Finally, multiple starting nodes are output and a fault warning is generated to remind maintenance personnel to carry out timely maintenance. Attached Figure Description

[0046] Figure 1 This is a flowchart illustrating an AI-based method for detecting fault areas in a power distribution cabinet, as provided in an embodiment of this application.

[0047] Figure 2 This is a schematic diagram of the structure of an AI-based power distribution cabinet fault area detection system provided in an embodiment of this application.

[0048] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0049] Explanation of reference numerals in the attached drawings: 1. Acquisition module; 2. Processing module; 3. Output module; 300. Electronic device; 301. Processor; 302. Communication bus; 303. User interface; 304. Network interface; 305. Memory. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0051] This application provides an AI-based method for detecting fault areas in power distribution cabinets. This method is applied to power distribution cabinet monitoring systems, such as... Figure 1 As shown, the method includes steps S101 to S106, which are as follows:

[0052] S101. Obtain multimodal monitoring data of the target power distribution cabinet, extract operating condition features from the multimodal monitoring data, and generate the operating condition vector of the target power distribution cabinet.

[0053] In the above steps, monitoring devices such as voltage sensors, current transformers, temperature and humidity sensors, and vibration sensors deployed in various areas of the target distribution cabinet are used to continuously collect multimodal monitoring data such as voltage, current, cabinet temperature and humidity, cabinet vibration intensity, and partial discharge. Then, the collected raw data is preprocessed by format conversion, deduplication, noise reduction, and smoothing. Then, the time-domain features (such as the effective value of current and the maximum temperature), frequency-domain features (such as the main frequency of the vibration signal), and statistical features (such as the variance of partial discharge) within a preset time window are extracted using the sliding window method. Finally, all extracted operating features are mapped into a fixed-length high-dimensional vector, which is the operating condition vector of the target distribution cabinet. This vector represents the current electrical load status, environmental status, and equipment health status of the target distribution cabinet.

[0054] S102. Input the operating condition vector into the preset typical operating condition vector library to decompose the operating condition and obtain multiple sub-operating condition vectors.

[0055] In the above steps, due to the complex operating conditions of the distribution cabinet, directly inputting the operating condition vector into the fault detection model would cause different types of operating condition features to interfere with each other, thus reducing the model's fault identification accuracy. Therefore, this application pre-sets a typical operating condition vector library, which contains a large number of typical operating condition vectors related to the target distribution cabinet. Then, the operating condition vector is input into the pre-set typical operating condition vector library for operating condition decomposition. Specifically, the matching degree (this application uses cosine similarity to characterize) is calculated one by one between the current operating condition vector and all typical operating condition vectors in the typical operating condition vector library. Multiple typical operating condition vectors with a matching degree greater than or equal to the pre-set operating condition matching degree threshold are selected, and then selected from the current operating condition vector... Feature values ​​consistent with the dimensional attributes of these typical operating condition vectors are extracted from the previous operating condition vectors, and multiple sub-operating condition vectors are reconstructed. This decomposes the current complex operating condition into multiple independent typical operating conditions, thereby reducing the mutual interference between different types of operating condition features and improving the recognition accuracy of the subsequent fault detection model. It should be noted that the current operating condition vector covers all operating condition features of the target distribution cabinet. At this time, the multiple typical operating condition vectors that match the current operating condition vector may not involve some operating condition features in the current operating condition vector. Therefore, in order to ensure the completeness of the subsequent fault analysis, this application uses these operating condition features as supplementary sub-operating condition vectors for subsequent fault analysis.

[0056] In one possible implementation, since the typical operating conditions involved in the target distribution cabinet may have different types and varying numbers of operating condition dimensions, using a uniform operating condition matching threshold would distort the operating condition matching results, thereby affecting the accuracy of subsequent operating condition decomposition and ultimately leading to a decrease in fault detection accuracy. Specifically, the typical operating conditions involved in the target distribution cabinet can be categorized into load-related, environmental, and equipment status-related types. Among these, environmental-related typical operating conditions have simple characteristics, strong stability, and fewer operating condition dimensions, while load-related and equipment status-related typical operating conditions have complex characteristics. The fluctuations are large, and the number of operating condition dimensions is large. In this case, if a uniform operating condition matching degree threshold is used, if the threshold is set too high, typical operating condition vectors such as load and equipment status with many dimensions and complex features will be misjudged as mismatched because the matching degree calculation value is difficult to reach the threshold, resulting in the omission of core operating condition features and incomplete operating condition decomposition. If the threshold is set too low, typical operating condition vectors such as environment with few dimensions and simple features will be misjudged as matched because the matching degree calculation value is too high, introducing redundant operating condition features, which will interfere with subsequent fault cause analysis and reduce the accuracy of fault detection in the distribution cabinet.

[0057] Therefore, to improve the accuracy and completeness of the operating condition decomposition, this application first divides the multiple typical operating condition vectors used to construct the typical operating condition vector library into multiple operating condition type groups according to multiple preset operating condition types. These preset operating condition types include, but are not limited to, load type, environmental type, and equipment status type. Then, historical operating condition data for multiple typical operating condition vectors is obtained. For each operating condition type group, the similarity of historical operating condition data between all pairs of typical operating condition vectors is calculated, resulting in a historical operating condition data similarity set corresponding to that operating condition type group. Then, the mean similarity of the historical operating condition data similarity set is calculated. Here, the mean similarity represents the commonality in similarity among most typical operating conditions under that operating condition type. Thus, the operating condition type can be... The average similarity of the corresponding groups is used as the baseline threshold for the operating condition matching degree of the operating condition type group. Then, for typical operating condition vectors with different numbers of operating condition dimensions in the same operating condition type group, as the number of operating condition dimensions increases, it indicates that the typical operating condition vector provides more fault detection evidence and has a larger fault tolerance space during fault detection. Therefore, its corresponding operating condition matching degree threshold can be appropriately lowered based on the baseline threshold for matching degree of the type group. Conversely, as the number of operating condition dimensions decreases, the typical operating condition vector provides less fault detection evidence and has a smaller fault tolerance space during fault detection. Its corresponding operating condition matching degree threshold needs to be appropriately increased based on the baseline threshold for matching degree of the type group. Thus, the operating condition matching degree threshold adapted to each typical operating condition vector is obtained. Specifically, adjusting the baseline threshold for operating condition matching degree based on the number of operating condition dimensions of the typical operating condition vector to obtain its adapted operating condition matching degree can be achieved using the following operating condition matching degree threshold adjustment formula:

[0058]

[0059] in, Let be the threshold value for the condition matching degree of the j-th typical condition vector in the i-th type of condition group. The benchmark threshold for the matching degree of the i-th type of working condition is... Let be the number of working condition dimensions of the j-th typical working condition vector in the i-th working condition type group. Let be the average number of work condition dimensions in the i-th work condition type group. This is the coefficient that influences the number of working condition dimensions on the working condition matching threshold.

[0060] In the above formula, when Greater than At that time, exist Based on the previous adjustment, it was lowered to adapt to the fault tolerance requirements of high-dimensional working condition vectors. Less than At that time, exist Based on this, it is adjusted upwards to meet the requirements of accurate matching of low-dimensional working condition vectors. equal At that time, and Maintain consistency.

[0061] Finally, each typical working condition vector is assigned a matching degree benchmark threshold. When the running working condition vector is input into the preset typical working condition vector library, the matching degree of the current running working condition vector is calculated one by one with all the typical working condition vectors in the typical working condition vector library. At this time, a variety of typical working condition vectors with matching degrees greater than or equal to their corresponding working condition matching degree thresholds are selected from the typical working condition vector library, thereby improving the accuracy and completeness of working condition decomposition.

[0062] S103. Input multiple sub-operating condition vectors into the preset power distribution cabinet fault detection model to obtain fault information corresponding to each of the multiple sub-operating condition vectors. The fault information includes faulty components, fault types, and fault confidence.

[0063] In the above steps, the preset power distribution cabinet fault detection model is a set of multiple models trained for each type of typical operating condition. It includes a load-type fault detection sub-model, an environmental-type fault detection sub-model, an equipment status-type fault detection sub-model, and an AI classification model for operating condition classification. The sub-models for fault detection are trained from a large amount of historical fault data and normal operation data under the corresponding operating condition type. Each fault detection sub-model is trained with typical operating condition vectors as input features and faulty components, fault types, and fault confidence as output labels. The AI ​​classification model for operating condition classification is trained from a large amount of historical typical operating condition vectors labeled with operating condition types. It can quickly identify the operating condition type to which the input sub-operating operating condition vector belongs, thereby achieving accurate matching between the sub-operating operating condition vector and the corresponding fault detection sub-model.

[0064] Therefore, when a sub-operating condition vector is input into a preset distribution cabinet fault detection model, the AI ​​classification model first identifies its corresponding operating condition type. Then, based on the identified operating condition type, the sub-operating condition vector is automatically routed to the matching fault detection sub-model. For example, a load-type sub-operating condition vector is routed to the load-type fault detection sub-model. Then, the corresponding fault detection sub-model performs feature analysis on the input sub-operating condition vector, combines the operating condition-fault correspondence learned during training, and outputs the fault information corresponding to the sub-operating condition vector. The fault information includes the fault component, fault type, and fault confidence. The fault component clearly points to the specific electrical component in the distribution cabinet that may fail, such as circuit breakers, contactors, cable joints, etc. The fault type clearly defines the specific manifestation of the fault, such as overload, short circuit, insulation aging, poor contact, etc. The fault confidence quantifies the reliability of the fault information, with a value range of [0,1]. The higher the confidence, the greater the probability of the corresponding fault occurring.

[0065] It should be noted that during the training process, each fault detection sub-model optimizes its generalization ability through methods such as sample expansion and cross-validation to avoid misjudgment of faults due to insufficient samples under a single operating condition. At the same time, the fault detection sub-model learns the feature distribution of the normal operating condition vector during training. When the input sub-operating condition vector meets the characteristics of the normal operating condition (i.e., the operating condition confidence is less than the threshold), it is determined that there is no fault corresponding to the sub-operating condition vector, and only a fault-free label is output to ensure the accuracy and rigor of the fault information output, providing reliable data support for subsequent fault reasoning and early warning.

[0066] S104. Input multiple fault information into a preset fault reasoning knowledge graph to obtain multiple fault paths. Each fault path includes multiple fault nodes, and each fault node corresponds to a fault information.

[0067] In the above steps, since multiple fault information only characterizes the fault form, it is possible that a certain faulty component itself is not faulty but is affected by the fault of the upstream faulty component, causing the faulty component to exhibit a faulty phenomenon. Therefore, in order to further determine the cause of the fault, this application inputs multiple fault information into a preset fault reasoning knowledge graph. Through the associative reasoning capability of the fault reasoning knowledge graph, the causal relationship between each fault information is sorted out, thereby generating multiple fault paths. Specifically, the pre-defined fault reasoning knowledge graph is constructed from the spatial structure and operating principle of the power distribution cabinet. It consists of fault nodes, causal links between fault nodes, and conditional probability tables on each causal link. Each fault node stores the name of an electrical component and its fault mode. Each causal link represents the fault propagation causal relationship between fault nodes. Specifically, it can be understood as the direct causal relationship where the fault of the electrical component corresponding to the upstream fault node causes the fault of the corresponding electrical component of the downstream fault node in a one-way direction along the electrical structure, signal transmission, or physical connection relationship of the power distribution cabinet. Each conditional probability table stores the conditional probability information of the fault nodes at both ends of the causal link. Specifically, it is the conditional probability value of the occurrence of the fault of the downstream fault node when the fault of the upstream fault node occurs. The conditional probability value is determined by the ratio of the frequency of the downstream fault occurring synchronously after the upstream fault occurs to the total frequency of the upstream fault in the historical operating data of power distribution cabinets of the same model and scenario as the target power distribution cabinet. Furthermore, it is dynamically updated as historical fault cases accumulate to ensure the timeliness of the conditional probability value.

[0068] When multiple fault information entries are input into a pre-defined fault reasoning knowledge graph, each fault information entry is first mapped to a corresponding fault node in the fault reasoning knowledge graph. This means that based on the fault component and fault type in the fault information, a fault node in the fault reasoning knowledge graph that stores the corresponding component name and fault mode is matched. Then, starting from each mapped fault node, based on the conditional probability table stored on the causal association edges, the fault propagation links between multiple fault nodes are traced upstream and downstream along the causal association edges. For example, fault nodes A, B, C, and D on the same association path are connected sequentially, where A, B, and D are mapped fault nodes, and C is an unmapped fault node. When fault a occurs in A, according to the conditional probability table between A and B, the probability of fault b occurring in B at this time is the highest. Then, identify the fault type of B. If the fault type of B is b, then connect A and B as a fault propagation link. If the fault type of B is not b, then do not connect A and B as a fault propagation link. Then, since C is an unmapped fault node, after determining that the fault type of B is b, according to the conditional probability table between B and C, assume that C has a fault c. At this time, according to the fault probability table between C and D, when C has a fault c, the probability of D having a fault d is the highest. Then identify the fault type of D. If the fault type of D is d, it means that the assumed fault c of C actually exists. Thus, A, B, C, and D can be constructed as a fault propagation link. If the fault type of D is not d, it means that C does not have a fault. Then only A and B are constructed as a fault propagation link. Therefore, the identified multiple fault transmission links are output as multiple fault paths. This process can not only clearly present the transmission process between various fault information, but also uncover hidden fault nodes that are not directly detected during the fault transmission process. This provides accurate and complete link support for subsequent identification of fault root causes, and further improves the accuracy of power distribution cabinet fault detection.

[0069] S105. Select multiple target fault paths from multiple fault paths and identify the starting node in the multiple target fault paths.

[0070] In the above steps, due to the different number of nodes and confidence levels of multiple fault paths, the reliability of different fault paths in reflecting the actual fault propagation process varies. Specifically, some fault paths have few nodes and low mean confidence levels, which may only capture isolated fault phenomena or spurious correlations, failing to accurately reflect the source and propagation pattern of the fault. Conversely, some fault paths have a sufficient number of nodes and high mean confidence levels, fully presenting the entire process of fault propagation from the source to the downstream, thus exhibiting higher reliability. Therefore, it is necessary to screen all fault paths, eliminating paths with low reliability and retaining paths with high reliability as target fault paths, providing a reliable basis for the accurate identification of subsequent fault causes.

[0071] This application first calculates the number of faulty nodes corresponding to each of the multiple fault paths, and then normalizes the number of faulty nodes corresponding to each of the multiple fault paths to obtain the evidence coefficients corresponding to each of the multiple fault paths, specifically:

[0072]

[0073] in, Let be the evidence coefficient for the i-th fault path. Let be the total number of faulty nodes in the i-th fault path. This represents the maximum number of faulty nodes across multiple fault paths.

[0074] Among them, the evidence coefficient represents the amount of evidence information provided by the failure path to explain the cause of the failure. The higher the amount of evidence information, the more evidence information the failure path can provide when explaining the cause of the failure, and the higher its reliability.

[0075] Then, for each fault path, the multiple fault nodes are divided into measured fault nodes and hypothetical fault nodes. Measured fault nodes are fault nodes that have been detected by the fault detection model and have clear fault information. Hypothetical fault nodes are fault nodes that are assumed to exist during the fault reasoning knowledge graph tracing process to improve the fault transmission link and connect with measured fault nodes. At this time, measured fault nodes have high reliability because they are directly derived from the model's measured analysis. Hypothetical fault nodes have relatively low reliability because they have no corresponding measured output from the fault detection model and no clear fault confidence. Therefore, this application counts the proportion of the number of measured fault nodes in the fault path to the total number of fault nodes in the fault path to determine the effective information content in the fault path. The more effective information content, the higher the reliability of the fault path.

[0076] Secondly, the average fault confidence of multiple measured fault nodes in each fault path is calculated to determine the overall reliability of each fault path. The higher the average fault confidence, the higher the reliability of the fault path.

[0077] Finally, a weighted sum is calculated for the evidence coefficient, effective information content, and mean fault confidence of each fault path to obtain a reliability score for each fault path. Then, the fault paths with a reliability score greater than or equal to a preset reliability score threshold are taken as target fault paths.

[0078] After filtering out multiple target fault paths, for each target fault path, the starting node of the target fault path is identified according to the direction of its causal relationship edge. The starting node corresponds to the upstream fault in the fault path and is not affected by the propagation of other fault nodes. Therefore, it is the fault cause corresponding to the fault path.

[0079] S106. Output the starting node of multiple target fault paths and generate a fault warning to remind maintenance personnel to carry out timely repairs.

[0080] In the above steps, the starting nodes of all target fault paths are summarized, and the starting nodes of the same faulty component and the same fault type are merged and deduplicated. Finally, multiple different starting nodes are output and a fault warning is generated, which is then pushed to the terminal equipment of maintenance personnel to remind them to carry out timely maintenance. The warning information includes the distribution cabinet number, the location of the fault, the details of the root cause fault, and the suggested maintenance measures.

[0081] Reference Figure 2 This application also provides an AI-based power distribution cabinet fault area detection system. The system is a power distribution cabinet monitoring system, comprising an acquisition module 1, a processing module 2, and an output module 3, wherein:

[0082] Module 1 is used to acquire multimodal monitoring data of the target power distribution cabinet, extract operating condition features from the multimodal monitoring data, and generate an operating condition vector of the target power distribution cabinet.

[0083] Processing module 2 is used to input the operating condition vector into a preset typical operating condition vector library for operating condition decomposition to obtain multiple sub-operating condition vectors.

[0084] Multiple sub-operating condition vectors are input into a preset power distribution cabinet fault detection model to obtain fault information corresponding to each sub-operating condition vector. The fault information includes faulty components, fault types, and fault confidence.

[0085] Multiple fault information is input into a preset fault reasoning knowledge graph to obtain multiple fault paths. Each fault path includes multiple fault nodes, and each fault node corresponds to a fault information.

[0086] Multiple target fault paths are obtained from multiple fault paths, and the starting node in the multiple target fault paths is identified;

[0087] Output module 3 is used to output the starting node of multiple target fault paths and generate fault warnings to remind maintenance personnel to carry out timely repairs.

[0088] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0089] This application also discloses an electronic device. (See reference...) Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application. The electronic device 300 may include: at least one processor 301, at least one network interface 304, a user interface 303, a memory 305, and at least one communication bus 302.

[0090] The communication bus 302 is used to enable communication between these components.

[0091] The user interface 303 may include a display screen and a camera. Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0092] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0093] The processor 301 may include one or more processing cores. The processor 301 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 305, and by calling data stored in memory 305. Optionally, the processor 301 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 301 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 301 and may be implemented as a separate chip.

[0094] The memory 305 may include random access memory (RAM) or read-only memory. Optionally, the memory 305 may include a non-transitory computer-readable storage medium. The memory 305 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 305 may also be at least one storage device located remotely from the aforementioned processor 301. (Refer to...) Figure 3 The memory 305, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for an AI-based method for detecting fault areas in power distribution cabinets.

[0095] exist Figure 3In the illustrated electronic device 300, the user interface 303 is mainly used to provide an input interface for the user and acquire user input data; while the processor 301 can be used to call an application program stored in the memory 305 for detecting fault areas in a power distribution cabinet based on AI. When executed by one or more processors 301, the electronic device 300 performs one or more of the methods described in the above embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simplicity, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0096] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0097] In the various embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between apparatuses or units may be electrical or other forms.

[0098] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0099] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0100] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0101] The above description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Other embodiments of this disclosure will be readily apparent to those skilled in the art upon consideration of the specification and the disclosure of practical truths.

[0102] This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A method for detecting fault areas in a power distribution cabinet based on AI, characterized in that, The method, applied in a power distribution cabinet monitoring system, includes: Acquire multimodal monitoring data of the target power distribution cabinet, extract operating condition features from the multimodal monitoring data, and generate an operating condition vector of the target power distribution cabinet; The operating condition vector is input into a preset typical operating condition vector library for operating condition decomposition to obtain multiple sub-operating condition vectors. Multiple sub-operating condition vectors are input into a preset power distribution cabinet fault detection model to obtain fault information corresponding to each of the multiple sub-operating condition vectors. The fault information includes faulty components, fault types, and fault confidence levels. Multiple fault information entries are input into a preset fault reasoning knowledge graph to obtain multiple fault paths. Each fault path includes multiple fault nodes, and each fault node corresponds to a fault information entry. Multiple target fault paths are obtained by filtering from the multiple fault paths, and the starting node in the multiple target fault paths is identified; Output the starting node of multiple target fault paths and generate a fault warning to remind maintenance personnel to carry out timely repairs.

2. The method according to claim 1, characterized in that, The preset typical working condition vector library is constructed from multiple typical working condition vectors. Each typical working condition vector corresponds to a working condition matching degree threshold determined by the number of working condition dimensions and the working condition type, specifically as follows: The multiple typical operating condition vectors are divided into multiple operating condition type groups according to multiple preset operating condition types, including load type, environment type and equipment status type; Obtain historical working condition data of multiple typical working condition vectors in the first working condition type group, wherein the first working condition type group is any one of the multiple working condition type groups; Calculate the pairwise similarity between historical working condition data of various typical working condition vectors in the first working condition type group to obtain the historical working condition data similarity set of the first working condition type group. Calculate the mean similarity of the historical working condition data similarity set, and use the mean similarity as the matching degree benchmark threshold for any typical working condition vector in the first working condition type group; Based on the number of working condition dimensions of the first typical working condition vector, the matching degree benchmark threshold is adjusted using a preset working condition matching degree threshold adjustment formula to obtain the working condition matching degree threshold of the first typical working condition vector. The first typical working condition vector is any one of the multiple typical working condition vectors in the first working condition type group.

3. The method according to claim 2, characterized in that, The step of inputting the operating condition vector into a preset typical operating condition vector table for operating condition decomposition to obtain multiple sub-operating condition vectors is as follows: The matching degree between the operating condition vector and the multiple typical operating condition vectors is calculated to obtain the operating condition matching degree between the operating condition vector and the multiple typical operating condition vectors. The operating condition matching degree between the second typical operating condition vector and the operating condition vector is compared with the operating condition matching degree threshold corresponding to the second typical operating condition vector, wherein the second typical operating condition vector is any one of the multiple typical operating condition vectors; If the working condition matching degree between the second typical working condition vector and the operating working condition vector is greater than or equal to the working condition matching degree threshold corresponding to the second typical working condition vector, then based on the multiple feature dimensions of the second typical working condition vector, working condition features consistent with the multiple features of the second typical working condition vector are extracted from the operating working condition vector and reconstructed into a sub-operating working condition vector.

4. The method according to claim 1, characterized in that, The preset fault reasoning knowledge graph consists of multiple fault nodes, causal association edges between the multiple fault nodes, and a conditional probability table on each causal association edge. Each fault node represents a fault mode of an electrical component in the target power distribution cabinet. Each causal association edge represents the fault propagation causal relationship between the two fault nodes. The conditional probability table stores the conditional probability information of the fault nodes at both ends of the causal association edge.

5. The method according to claim 4, characterized in that, The step of inputting multiple fault information entries into a preset fault reasoning knowledge graph to obtain multiple fault paths is as follows: Based on the fault type and fault component of multiple fault information, the multiple fault information is mapped to the preset fault reasoning knowledge graph to obtain multiple mapped fault nodes; Taking multiple mapped fault nodes as the starting point for tracing the source, and based on the causal association edges between each fault node in the preset fault reasoning knowledge graph and the conditional probability table on each causal association edge, the fault source is traced along the causal association edges to obtain multiple fault propagation links between multiple mapped fault nodes. The multiple fault propagation links are output to obtain multiple fault paths.

6. The method according to claim 1, characterized in that, The process of selecting multiple target fault paths from the multiple fault paths specifically involves: The multiple fault nodes in the first fault path are divided into multiple measured fault nodes and multiple hypothetical fault nodes, and the first fault path is any one of the multiple fault paths. Calculate the total number of fault nodes in the first fault path and the total number of measured fault nodes; Calculate the mean fault confidence of multiple measured fault nodes in the first fault path; The reliability score of the first fault path is determined based on the total number of fault nodes, the total number of measured fault nodes, and the average fault confidence. If the reliability score of the first fault path is greater than or equal to a preset reliability score threshold, then the first fault path is determined to be the target fault path.

7. An AI-based power distribution cabinet fault area detection system, characterized in that, The system is a power distribution cabinet monitoring system, which includes an acquisition module, a processing module, and an output module, wherein: The acquisition module is used to acquire multimodal monitoring data of the target power distribution cabinet, extract operating condition features from the multimodal monitoring data, and generate an operating condition vector of the target power distribution cabinet. The processing module is used to input the operating condition vector into a preset typical operating condition vector library for operating condition decomposition to obtain multiple sub-operating condition vectors. Multiple sub-operating condition vectors are input into a preset power distribution cabinet fault detection model to obtain fault information corresponding to each of the multiple sub-operating condition vectors. The fault information includes faulty components, fault types, and fault confidence levels. Multiple fault information entries are input into a preset fault reasoning knowledge graph to obtain multiple fault paths. Each fault path includes multiple fault nodes, and each fault node corresponds to a fault information entry. Multiple target fault paths are obtained by filtering from the multiple fault paths, and the starting node in the multiple target fault paths is identified; The output module is used to output the starting node of multiple target fault paths and generate fault warnings to remind maintenance personnel to carry out timely repairs.

8. An electronic device, characterized in that, The device includes a processor (301), a memory (305), a user interface (303), and a network interface (304). The memory (305) is used to store instructions. The user interface (303) and the network interface (304) are used to communicate with other devices. The processor (301) is used to execute the instructions stored in the memory (305) to cause the electronic device (300) to perform the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1 to 6.

Images