Equipment fault type determination method and device and electronic equipment

Abstract

The invention discloses an equipment fault type determination method and device and electronic equipment. The method comprises the following steps: receiving a fault type determination request; in response to the fault type determination request, obtaining target power data of the target equipment according to the equipment identifier; determining a plurality of historical data corresponding to the target power data; determining a target map feature corresponding to the target power data and historical map features corresponding to the multiple pieces of historical data; and determining a target fault type of the target device according to the target map feature and the plurality of historical map features. According to the invention, the technical problem of low accuracy of the equipment fault type determined according to the operation characteristics of the power equipment in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method and device for determining a device fault type, and an electronic device. Background Art

[0002] Determining the type of equipment failure is crucial for ensuring stable power system operation, improving maintenance efficiency, and reducing unplanned downtime. It helps operations and maintenance personnel quickly and accurately locate the problem, take effective measures to prevent the failure from escalating, and ensure the continuity and security of power supply. Currently, the primary method for determining the type of equipment failure is to collect monitoring data from power equipment and identify high-dimensional characteristics of the equipment's operation. However, this approach suffers from the technical issue of low accuracy in determining the equipment failure type.

[0003] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0004] Embodiments of the present invention provide a method, apparatus, and electronic device for determining a device fault type, to at least solve the technical problem in the prior art of low accuracy in determining a device fault type based on operating characteristics of power equipment.

[0005] According to one aspect of an embodiment of the present invention, a method for determining a device fault type is provided, comprising: receiving a fault type determination request, wherein the fault type determination request carries a device identification of a target device; in response to the fault type determination request, obtaining target power data of the target device based on the device identification; determining a plurality of historical data corresponding to the target power data, wherein the plurality of historical data respectively include historical power data and historical fault types corresponding to the historical power data, and the similarity between the historical power data and the target power data is greater than a predetermined threshold; determining a target map feature corresponding to the target power data and the target map feature of the plurality of historical data; The target map feature is determined based on the target triplet, and the target triplet is used to represent the relationship between multiple data items in the target power data and the corresponding data values. The corresponding historical map feature is determined based on the corresponding first triplet and the corresponding second triplet, and the corresponding first triplet is used to represent the relationship between multiple data items in the historical power data and the corresponding data values. The corresponding second triplet is used to represent the relationship between the historical power data and the historical fault type corresponding to the historical power data; based on the target map feature and multiple historical map features, the target fault type of the target device is determined.

[0006] Optionally, determining multiple historical data corresponding to the target power data includes: determining a target data vector corresponding to the target power data; determining multiple historical data vectors based on the target data vector, wherein the vector similarities corresponding to the multiple historical data vectors are greater than a predetermined similarity threshold, and the corresponding vector similarity is the similarity between the corresponding historical data vector and the target data vector; determining multiple historical data corresponding to the target power data based on the multiple historical data vectors.

[0007] Optionally, determining the target map features corresponding to the target power data and the historical map features corresponding to the multiple historical data respectively includes: calling a map feature determination model, wherein the map feature determination model is provided with target model parameters, and the target model parameters are determined based on initial model parameters and sample data, and the sample data includes sample power data and sample map features corresponding to the sample power data; inputting the target power data and the corresponding historical data into the first module of the map feature determination model respectively to obtain a target triplet corresponding to the target power data and a historical triplet corresponding to the corresponding historical data, and inputting the target triplet and the corresponding historical triplet into the second module of the map feature determination model respectively to obtain the target map feature and the corresponding historical map feature, wherein the corresponding historical triplet includes the corresponding first triplet and the corresponding second triplet.

[0008] Optionally, before calling the spectrum feature determination model, it also includes: determining an initial feature determination model, wherein the initial feature determination model is provided with initial model parameters; determining an initial spectrum feature based on the initial feature determination model and sample power data; determining a characteristic error value between the initial spectrum feature and the sample spectrum feature; when the characteristic error value is less than an error threshold, determining the initial model parameter as a target model parameter; and determining the spectrum feature determination model based on the target model parameters and the initial feature determination model.

[0009] Optionally, based on the multiple historical data vectors, multiple historical data corresponding to the target power data are determined, including: determining target distances between the target power data and multiple target cluster centers, wherein the multiple target cluster centers are determined based on multiple initial cluster centers, and the multiple initial cluster centers are determined based on the power data corresponding to the multiple historical data vectors; determining multiple historical fault types corresponding to the target power data based on multiple target distances, wherein the target distances corresponding to the multiple historical fault types are less than a predetermined distance threshold; determining the historical power data corresponding to the multiple historical fault types; and determining multiple historical data corresponding to the target power data based on the multiple historical fault types and the historical power types corresponding to the multiple historical fault types.

[0010] Optionally, the target fault type of the target device is determined based on the target map feature and multiple historical map features, including: determining the similarity values ​​corresponding to the target map features of the multiple historical map features respectively; determining the type indexes of the historical fault types corresponding to the multiple historical map features respectively based on the multiple similarity values; and determining the target fault type corresponding to the target device based on the multiple type indices.

[0011] Optionally, after determining the target fault type of the target device based on the target map features and multiple historical map features, it also includes: determining a re-inspection instruction based on the target fault type of the target device, wherein the re-inspection instruction carries a type identifier corresponding to the re-inspection item type, and the re-inspection item type is determined based on the target fault type of the target device; sending the re-inspection instruction to the re-inspection device; updating the target fault type of the target device based on the received re-inspection data to obtain the updated fault type corresponding to the target device.

[0012] According to one aspect of an embodiment of the present invention, a device for determining a device fault type is provided, comprising: a receiving module for receiving a fault type determination request, wherein the fault type determination request carries a device identification of a target device; a responding module for obtaining target power data of the target device based on the device identification in response to the fault type determination request; a first determining module for determining a plurality of historical data corresponding to the target power data, wherein the plurality of historical data respectively include historical power data and historical fault types corresponding to the historical power data, and a similarity between the historical power data and the target power data is greater than a predetermined threshold; a second determining module for determining a target power data corresponding to the target power data The target map feature and the historical map features corresponding to the multiple historical data respectively, wherein the target map feature is determined based on the target triplet, and the target triplet is used to represent the relationship between multiple data items in the target power data and the corresponding data values ​​respectively, and the corresponding historical map feature is determined based on the corresponding first triplet and the corresponding second triplet, and the corresponding first triplet is used to represent the relationship between multiple data items in the historical power data and the corresponding data values ​​respectively, and the corresponding second triplet is used to represent the relationship between the historical power data and the historical fault type corresponding to the historical power data; a third determination module is used to determine the target fault type of the target device based on the target map feature and multiple historical map features.

[0013] According to one aspect of an embodiment of the present invention, an electronic device is provided, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement any of the above-mentioned methods for determining a device fault type.

[0014] According to one aspect of an embodiment of the present invention, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device can execute any of the above-mentioned methods for determining a device fault type.

[0015] In an embodiment of the present invention, a method is adopted in which a fault type determination request is received; in response to the fault type determination request, target power data of a target device is obtained based on a device identifier; a plurality of historical data corresponding to the target power data is determined; a target spectrum feature corresponding to the target power data and a plurality of historical data respectively corresponding to the historical spectrum features are determined; and a target fault type of a target device is determined based on the target spectrum feature and the plurality of historical spectrum features. By determining the target spectrum feature corresponding to the target power data and the historical spectrum feature corresponding to the plurality of historical data respectively, the purpose of determining the target fault type of a target device based on the target spectrum feature and the plurality of historical spectrum features is achieved. , and the historical graph features corresponding to multiple historical data whose similarity with the target power data is greater than a predetermined threshold, can determine the multi-dimensional feature information between the target power data type and the data value, the multi-dimensional feature information between the historical power data type and the data value, and the correlation relationship between the historical power data and the corresponding historical fault type. By comparing the target graph features with multiple historical graph features and analyzing the similarity of the multi-dimensional feature information corresponding to the target power data and the historical power data, the fault type corresponding to the target power data can be determined, thereby improving the accuracy of determining the equipment fault type, and further solving the technical problem in the prior art of low accuracy of determining the equipment fault type based on the operation characteristics of the power equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0017] Figure 1 is a flow chart of a method for determining a device fault type according to an embodiment of the present invention;

[0018] Figure 2 This is a schematic diagram of cloud-edge collaboration provided by an optional embodiment of the present invention;

[0019] Figure 3 is a flow chart of a method for determining a device fault type provided by an optional embodiment of the present invention;

[0020] Figure 4 It is a structural block diagram of a device for determining a device fault type according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0022] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] Example 1

[0024] According to an embodiment of the present invention, an embodiment of a method for determining a device fault type is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0025] Figure 1 FIG. 1 is a flow chart of a method for determining a device fault type according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0026] Step S102: receiving a fault type determination request, wherein the fault type determination request carries a device identifier of a target device.

[0027] In step S102 provided in the present application, a fault type determination request is received.

[0028] Among them, a fault type determination request is involved. The fault type determination request refers to a request to evaluate the current state of the target device to determine whether there is a fault and the specific fault type.

[0029] Among them, the target equipment is involved, and the target equipment refers to the power equipment that needs to be evaluated for the fault type.

[0030] Among them, device identification is involved. Device identification refers to the label or serial number used to uniquely identify the target device.

[0031] In this step, a fault type determination request is received, which is the starting step of determining the fault type of the target device. The fault type determination request carries the unique identification information of the target power device, clearly indicating the power device that needs to be diagnosed.

[0032] Step S104 : In response to the fault type determination request, target power data of the target device is acquired according to the device identification.

[0033] In step S104 provided in this application, target power data of the target device is acquired.

[0034] Among them, target power data is involved. Target power data refers to power data collected by monitoring equipment that needs to be analyzed to determine the equipment status and fault type of the target equipment, such as the voltage, current, temperature, partial discharge signal of the target equipment, and other real-time measurement values ​​obtained from the power equipment.

[0035] In this step, after receiving the fault type determination request, target power data of the target device is acquired.

[0036] Step S106 , determining multiple historical data corresponding to the target power data, wherein the multiple historical data respectively include historical power data and historical fault types corresponding to the historical power data, and the similarity between the historical power data and the target power data is greater than a predetermined threshold.

[0037] In step S106 provided in the present application, a plurality of historical data corresponding to the target power data are determined.

[0038] Among them, historical data is involved. Historical data refers to the power equipment operation data of the target equipment collected and recorded in the historical time period, including power data and fault types.

[0039] Among them, historical power data is involved. Historical power data refers to the power data of the target equipment under different fault types in the historical time period.

[0040] Among them, historical fault types are involved. Historical fault types refer to the specific types of faults that occurred in the target equipment during the historical period, such as overheating, partial discharge, overload, circuit breaker, etc.

[0041] Among them, a predetermined threshold is involved. The predetermined threshold refers to a standard set in advance to determine whether the similarity between the target power data and the historical power data is high enough so that the fault type can be learned or inferred from the historical data.

[0042] This step compares the currently collected target power data with stored historical data to identify historical data whose similarity with the target data exceeds a predetermined threshold. This historical data not only contains the operating data of the power equipment at different points in time, but also includes the fault type diagnosis results of the equipment at that time. By comparing the target data with historical data, the fault types recorded in the historical data can be used to predict or assist in determining the type of fault that the power equipment may currently encounter, thereby improving the accuracy and reliability of the current fault type diagnosis.

[0043] Step S108, determine the target map features corresponding to the target power data and the historical map features corresponding to the multiple historical data, wherein the target map features are determined based on the target triples, and the target triples are used to represent the relationship between the multiple data items in the target power data and the corresponding data values, and the corresponding historical map features are determined based on the corresponding first triples and the corresponding second triples, and the corresponding first triples are used to represent the relationship between the multiple data items in the historical power data and the corresponding data values, and the corresponding second triples are used to represent the relationship between the historical power data and the historical fault types corresponding to the historical power data.

[0044] In step S108 provided in the present application, target graph features corresponding to the target power data and historical graph features corresponding to the plurality of historical data are determined.

[0045] Among them, target graph features are involved. Target graph features refer to the features extracted after the target power data is converted into a knowledge graph, which is used to describe the multi-dimensional information of the current state of the power equipment, such as the equipment's operating parameters, environmental conditions, and potential signs of failure.

[0046] Among them, historical graph features are involved. Historical graph features refer to features extracted from the knowledge graph built based on historical power data.

[0047] Among them, the target triplet is involved, and the target triplet refers to a triplet that describes the association relationship between each data item and its numerical value in the target power data.

[0048] Among them, the first triplet is involved, and the first triplet refers to a triplet used to represent the relationship between multiple data items and corresponding data values ​​in historical power data.

[0049] Among them, the second triplet is involved, and the second triplet refers to the triplet representing the relationship between historical power data and historical fault types.

[0050] In this step, the target power data is first converted into a series of target triplets, which describe the relationship between data items (such as voltage and current) and specific values. Target graph features are then extracted from these relationships. Simultaneously, for historical power data, two sets of triplets are constructed: the first triplet describes the relationship between historical data items and values, while the second triplet associates historical data with known historical fault types, thereby forming corresponding historical graph features.

[0051] Through this step, the target graph features corresponding to the target power data and the historical graph features corresponding to multiple historical data are determined. By using the form of knowledge graph, we can more comprehensively understand the operating status of the target equipment and the relationship between the fault type and the operating status, so as to more accurately judge the fault type and improve the accuracy of fault location.

[0052] Step S110 : determining a target fault type of a target device based on a target graph feature and a plurality of historical graph features.

[0053] In step S110 provided in the present application, a target fault type of a target device is determined.

[0054] Among them, the target fault type is involved, and the target fault type refers to the device fault type of the target device in the current operating state.

[0055] Through this step, by comparing the target map features and multiple historical map features and analyzing the similarity of the multi-dimensional feature information corresponding to the target power data and the historical power data, the fault type corresponding to the target power data can be determined, thereby improving the accuracy of determining the equipment fault type.

[0056] Through the above steps S102-S110, a fault type determination request can be received; in response to the fault type determination request, target power data of the target device is obtained according to the device identification; multiple historical data corresponding to the target power data are determined; target map features corresponding to the target power data and historical map features corresponding to the multiple historical data are determined; a target fault type of the target device is determined based on the target map features and the multiple historical map features, and the purpose of determining the target fault type of the target device based on the target map features and the multiple historical map features is achieved by determining the target map features corresponding to the target power data and the historical map features corresponding to the multiple historical data. spectral features, and historical spectrum features corresponding to multiple historical data whose similarity with the target power data is greater than a predetermined threshold, can determine the multi-dimensional feature information between the target power data type and the data value, the multi-dimensional feature information between the historical power data type and the data value, and the correlation between the historical power data and the corresponding historical fault type. By comparing the target spectrum features with multiple historical spectrum features and analyzing the similarity of the multi-dimensional feature information corresponding to the target power data and the historical power data, the fault type corresponding to the target power data can be determined, thereby improving the accuracy of determining the equipment fault type, and further solving the technical problem in the prior art of low accuracy of determining the equipment fault type based on the operation characteristics of the power equipment.

[0057] As an optional embodiment, determining multiple historical data corresponding to target power data includes: determining a target data vector corresponding to the target power data; determining multiple historical data vectors based on the target data vector, wherein the vector similarities corresponding to the multiple historical data vectors are greater than a predetermined similarity threshold, and the corresponding vector similarities are the similarities between the corresponding historical data vectors and the target data vector; determining multiple historical data corresponding to the target power data based on the multiple historical data vectors.

[0058] In this embodiment, a specific step of determining a plurality of history data corresponding to target power data is described.

[0059] Among them, the target data vector is involved. The target data vector refers to the vector obtained by processing the target power data and converting the text information therein into numerical information.

[0060] Among them, historical data vectors are involved. Historical data vectors refer to vectors obtained by converting text information of power equipment monitoring data within a historical time period into numerical information.

[0061] Among them, vector similarity is involved. Vector similarity refers to an indicator that measures the similarity between the target data vector and the historical data vector.

[0062] Among them, a predetermined similarity threshold is involved. The predetermined similarity threshold refers to a numerical standard set in advance and is used to screen out a set of historical data vectors that are highly similar to the target data vector.

[0063] In this step, the collected real-time monitoring data on power equipment is first converted into a numerical vector to obtain a target data vector. Next, a vector similarity calculation method is used to filter out multiple historical data vectors from the historical dataset that have a high similarity with the target data vector (i.e., exceeding a predetermined similarity threshold). After finding multiple historical data vectors that meet the similarity requirements, a reverse lookup can be performed to obtain the corresponding historical monitoring data. This data contains detailed records of past equipment conditions that occurred similarly, including specific fault types and power data.

[0064] Through this step, the target power data is converted into a target data vector, the similarity between the target data vector and other data vectors is compared, and multiple historical data vectors are determined to improve the accuracy of the determined historical data vectors. Then, multiple historical data with high similarity to the target power data are determined to improve the accuracy of the determined historical data, and the similarity between the historical map features and the target map features is improved to more accurately determine the equipment fault type corresponding to the target power data.

[0065] As an optional embodiment, determining the target spectrum features corresponding to the target power data and the historical spectrum features corresponding to multiple historical data respectively includes: calling a spectrum feature determination model, wherein the spectrum feature determination model is provided with target model parameters, and the target model parameters are determined based on the initial model parameters and sample data, and the sample data includes sample power data and sample spectrum features corresponding to the sample power data; inputting the target power data and the corresponding historical data into the first module of the spectrum feature determination model respectively to obtain the target triplet corresponding to the target power data and the historical triplet corresponding to the corresponding historical data, and inputting the target triplet and the corresponding historical triplet into the second module of the spectrum feature determination model respectively to obtain the target spectrum features and the corresponding historical spectrum features, wherein the corresponding historical triplet includes the corresponding first triplet and the corresponding second triplet.

[0066] In this embodiment, specific steps of determining target graph features corresponding to target power data and historical graph features corresponding to a plurality of historical data are described.

[0067] Among them, a spectrum feature determination model is involved, which refers to a model used to convert monitoring data of power equipment into spectrum features.

[0068] Among them, the target model parameters are involved. The target model parameters refer to a parameter set in the graph feature determination model, which is used to optimize the model's processing of target device monitoring data and extraction of graph features.

[0069] Among them, the initial model parameters are involved. The initial model parameters refer to a set of parameters set before the training graph feature determination model begins. They provide basic guidance on how the model processes input data.

[0070] In this step, first, the spectrum feature determination model that has been optimized and trained is retrieved. The spectrum feature determination model is provided with target model parameters. The target model parameters are obtained based on the initial model parameters and a large amount of sample data training, ensuring that the model can effectively identify key information in the power data. Next, the monitoring data (including the current target power data and historical data) is input into the spectrum feature determination model. The target triples corresponding to the target power data and the historical triples corresponding to the corresponding historical data are determined by the first module of the spectrum feature determination model. These triples capture the entities in the monitoring data and their mutual relationships. Then, the target triples corresponding to the target power data and the historical triples corresponding to the corresponding historical data are further input into the second module of the model, and highly generalized spectrum features are extracted therefrom to obtain target spectrum features and corresponding historical spectrum features. Through this step, the spectrum feature determination model is used to convert the monitoring data into spectrum features, and the operating mode and fault signs of the power equipment are understood more deeply, thereby significantly improving the accuracy and efficiency of fault diagnosis.

[0071] As an optional embodiment, before calling the spectrum feature determination model, it also includes: determining an initial feature determination model, wherein the initial feature determination model is provided with initial model parameters; determining the initial spectrum feature based on the initial feature determination model and sample power data; determining a characteristic error value between the initial spectrum feature and the sample spectrum feature; when the characteristic error value is less than an error threshold, determining the initial model parameter as the target model parameter; and determining the spectrum feature determination model based on the target model parameters and the initial feature determination model.

[0072] In this embodiment, the specific steps of determining the graph feature determination model are described.

[0073] Among them, the initial feature determination model is involved. The initial feature determination model refers to a preliminary model framework used to extract knowledge graph features from the monitoring data of power equipment.

[0074] Among them, the initial graph features are involved, and the initial graph features refer to the features extracted from the sample power data through the initial model parameters.

[0075] Among them, the feature error value is involved, which refers to a quantitative indicator that measures the difference between the initial map features and the sample map features.

[0076] This involves an error threshold, which is a pre-set threshold used to determine whether the feature error value is low enough to be acceptable. If the feature error value is lower than the error threshold, it indicates that the model parameters have achieved a satisfactory performance level.

[0077] In this step, an initial feature determination model (AFDM) is first defined. This model is a basic model architecture with initial model parameters. Next, using sample power data as input, the AFDM extracts initial spectral features. These features are then compared with known sample spectral features to determine feature error values ​​to quantify the model's performance. If the feature error value falls below a preset error threshold, this indicates good model performance, and the current initial model parameters are determined as the target model parameters. Finally, the final spectral feature determination model is created using the target model parameters and the AFDM architecture.

[0078] Through this step, the model parameters are iteratively optimized until the characteristic error value meets the preset threshold requirement. The optimized model parameters can better capture the key information in the power data, reduce the misjudgment of fault types caused by inaccurate feature extraction, and significantly improve the accuracy of the model in extracting graph features.

[0079] As an optional embodiment, based on multiple historical data vectors, multiple historical data corresponding to the target power data are determined, including: determining the target distances between the target power data and multiple target cluster centers, wherein the multiple target cluster centers are determined based on multiple initial cluster centers, and the multiple initial cluster centers are determined based on the power data corresponding to multiple historical data vectors; determining multiple historical fault types corresponding to the target power data based on multiple target distances, wherein the target distances corresponding to the multiple historical fault types are less than a predetermined distance threshold; determining the historical power data corresponding to the multiple historical fault types; and determining the multiple historical data corresponding to the target power data based on the multiple historical fault types and the historical power types corresponding to the multiple historical fault types.

[0080] In this embodiment, a specific step of determining a plurality of history data corresponding to target power data is described.

[0081] Among them, the target cluster center is involved. The target cluster center refers to multiple cluster centers obtained by clustering the power data corresponding to multiple historical data vectors multiple times until the cluster center no longer changes significantly.

[0082] Among them, the target distance is involved. The target distance refers to the distance between the target power data vector and multiple target cluster centers. It is used to measure the similarity between the target data and the power equipment status that has appeared in history.

[0083] Among them, the initial clustering center is involved, and the initial clustering center refers to a plurality of clustering centers obtained after the power data corresponding to a plurality of historical data vectors are initially clustered.

[0084] Here, a predetermined distance threshold is involved. The predetermined distance threshold refers to a pre-set value used to determine whether the target distance is small enough to indicate that the target power data is closely related to a certain historical fault type.

[0085] In this step, first, target cluster centers are constructed based on historical data. Each target cluster center represents a pattern of one or a group of equipment operating states. Next, the target distance between the target power data vector and all target cluster centers is determined to determine the correlation between the target data and the historical data. If the distance between the target power data and a cluster center is less than a predetermined distance threshold, then it is considered that the target data may be related to the fault type represented by the cluster. Finally, a specific historical fault type is selected from the associated cluster, and the corresponding historical power data is obtained, thereby forming multiple historical data sets related to the target power data for subsequent fault type diagnosis and analysis.

[0086] Through this step, we can make full use of the information in the historical data, quickly analyze a large amount of historical data, and find the fault type most relevant to the target data by comparing it with the current data. While accelerating the fault diagnosis process, we can accurately map the current data to the historical fault type and improve the accuracy of fault classification.

[0087] As an optional embodiment, the target fault type of the target device is determined based on the target map feature and multiple historical map features, including: determining the similarity values ​​corresponding to the target map features of the multiple historical map features respectively; determining the type index of the historical fault types corresponding to the multiple historical map features respectively based on the multiple similarity values; and determining the target fault type corresponding to the target device based on the multiple type indices.

[0088] In this embodiment, specific steps of determining a target fault type corresponding to a target device are described.

[0089] Among them, the similarity value is involved. The similarity value refers to a numerical value that measures the degree of matching between two different graph features to evaluate the similarity between the target graph features and the historical graph features.

[0090] This involves the type index, which quantifies the contribution or relevance of historical fault types to the target device's fault type based on the similarity between historical graph features and target graph features. A higher type index indicates a greater relevance of the historical fault type to the target device's current performance.

[0091] In this step, first, the similarity value between the target graph feature and each historical graph feature is determined by comparing the triple relationship and attribute values ​​constructed between the two in the knowledge graph. Next, a type index is assigned to each historical fault type based on the determined similarity value. The type index reflects the correlation between the historical fault type and the current state of the target device, that is, the historical fault type corresponding to the historical graph feature with a higher similarity value will obtain a higher type index. Finally, the type indexes of all historical fault types are comprehensively considered, and the highest one or the one that meets certain conditions is selected as the target fault type. Through this step, the similarity between historical data and current data is quantified, and the corresponding type index is assigned, so that the device state can be identified more precisely and the accuracy of the determined target fault type can be improved.

[0092] As an optional embodiment, after determining the target fault type of the target device based on the target map features and multiple historical map features, it also includes: determining a re-inspection instruction based on the target fault type of the target device, wherein the re-inspection instruction carries a type identifier corresponding to the re-inspection item type, and the re-inspection item type is determined based on the target fault type of the target device; sending the re-inspection instruction to the re-inspection device; updating the target fault type of the target device based on the received re-inspection data to obtain the updated fault type corresponding to the target device.

[0093] In this embodiment, specific steps of determining the update failure type corresponding to the target device are described.

[0094] Among them, re-inspection instructions are involved. Re-inspection instructions refer to instructions that require more detailed inspection or testing after the fault type is preliminarily determined to confirm the fault type or eliminate misjudgment.

[0095] Among them, type identification is involved. Type identification refers to a unique code or label used to distinguish different re-inspection items, so that the re-inspection equipment can identify and perform corresponding re-inspection operations.

[0096] Among them, it involves updating the fault type. The updated fault type refers to the latest fault type of the target device determined after receiving and analyzing the re-inspection data.

[0097] In this step, the necessary re-inspection items are determined based on the target fault type of the initial diagnosis. A re-inspection instruction, including the re-inspection item type identifier, is then generated and sent to the corresponding re-inspection device. After the re-inspection device executes the instruction, it transmits the re-inspection data back to the system, which then updates the fault type determination and obtains a more accurate diagnosis of the fault type, thus updating the fault type. This addition of a re-inspection step effectively reduces uncertainty in the initial diagnosis, ensuring a more accurate and reliable determination of the fault type.

[0098] Based on the above embodiment and optional embodiment, an optional implementation manner is provided, which is described in detail below.

[0099] In related technologies, when performing status checks on power equipment, high-dimensional features are directly generated from the equipment's monitoring data. These features are then fed into a training model to determine the equipment's status, or fault type. This method of determining fault type lacks feature analysis at the monitoring data level, which can lead to inaccurate fault types being determined.

[0100] In light of this, an optional embodiment of the present invention provides a device status detection method that uses a large language model to retrieve and identify historical data with one or more fault types similar to the monitored data. This method then compares features of this historical data with the monitored data to predict the corresponding fault type of the device. After predicting the device's fault type, the power equipment is re-inspected using both manual inspection equipment and robotic live inspection equipment. The results of these two re-inspections are used to comprehensively determine the device's fault type, thereby improving the accuracy of fault type determination.

[0101] The following is a detailed introduction to the device status detection method provided by an optional embodiment of the present invention.

[0102] S1. Receive a fault type determination request.

[0103] An optional embodiment of the present invention provides a method for detecting the status of substation equipment. The substation where the target device is located includes a large number of substation equipment, each of which performs its own function and works together to ensure the normal operation of the substation. To ensure the long-term and safe operation of the substation, not only daily inspections are required, but also timely inspections and understanding of the faults and conditions when equipment at a specific point in the substation experiences problems.

[0104] In order to perform multi-dimensional detection of substations, an optional implementation method of the present invention adopts a cloud-edge collaborative approach to set up the cloud, edge and end sides. The cloud-edge collaborative computing platform is an intelligent computing architecture developed to meet the new challenges brought about by the development of the Internet of Things (IoT). To manage digital cloud computing, the platform can sink computing power from the centralized cloud to the edge nodes close to the data source, and optimize data processing capabilities through a distributed computing model. The cloud-edge collaborative computing model not only extends the capabilities of cloud native, but also enables data processing, business applications and artificial intelligence (AI) models to be executed at the edge close to the data source by deploying edge nodes, solving the problems of real-time response, data privacy, and maintenance convenience encountered by the Internet of Things when it is implemented. The cloud-edge collaborative computing platform has significantly improved data processing efficiency and system response speed in actual applications in multiple industries, realized diversified intelligent applications, and contributed to the digital transformation of various industries. The cloud can be responsible for advanced analysis and global optimization.

[0105] Figure 2 This is a schematic diagram of cloud-edge collaboration provided by an optional embodiment of the present invention, such as Figure 2As shown, the device side includes robotic live-line detection equipment, manual detection equipment, and online monitoring equipment. The robotic live-line detection equipment includes a multifunctional inspection robot, which can perform inspections within substations. The multifunctional inspection robot can be equipped with a smart mobile terminal, an infrared thermal imaging detection terminal, a voiceprint imaging detection terminal, and a partial discharge detection terminal. The infrared thermal imaging detection terminal, voiceprint imaging detection terminal, and partial discharge detection terminal communicate with the multifunctional inspection robot via the smart mobile terminal. The smart mobile terminal can be installed in the multifunctional inspection robot and is responsible for controlling the multifunctional inspection robot at the substation site to perform various inspection operations, controlling the robot's movements within the substation, and implementing services such as robotic arm posture control and inspection data transmission. The smart mobile terminal provides cloud-edge bidirectional collaboration, data exchange, and command forwarding capabilities. The smart mobile terminal utilizes a hybrid networking approach using Wireless LAN Authentication and Privacy Infrastructure (WAPI) and / or 5G-APN (5G-APN), enabling autonomous network access and dual-network hot-swap capabilities. When using a 5G-APN for ground-based scenarios, the smart mobile terminal automatically activates its built-in high-performance security chip (TF encryption card) and establishes a broadband IoT connection with the edge. The smart mobile terminal communicates with the multifunctional inspection robot using Ethernet wired communication. The multifunctional inspection robot, serving as a field vehicle for substation operations, uses its robotic arm to autonomously grasp various live-line inspection instruments and conduct on-site live-line inspections. The multifunctional inspection robot includes a controller for controlling its movements. The controller connects to the smart mobile terminal via wired Ethernet and receives real-time control policies from the cloud, forwarded by the smart mobile terminal using the Transmission Control Protocol (TCP). It drives the four wheels, robotic arm, and other components to perform related operations and uploads real-time status information, inspection data, and maps (including partial discharge, infrared, and voiceprint imaging). The multifunctional inspection robot consists of a torso and four limbs. The limb part includes four mechanical legs that touch the ground and are used to move the robot body within the substation. The torso part is used to carry equipment such as a mechanical arm to realize the multiple functions of the multifunctional inspection robot. According to one embodiment of the present invention, the torso part includes a controller, which serves as a motion control module of the multifunctional inspection robot. The torso part also includes a camera for collecting image information, as well as various types of sensors such as distance sensors, which are used to sense the surrounding environment, the distance between the edge of the torso part and the target obstacle and other information. The robot body can be realized as a wheeled robot, a tracked robot and a legged robot. The optional implementation manner of the present invention does not limit the specific action mode of the robot body.

[0106] The terminal side also includes infrared thermal imaging detection terminals, voiceprint imaging detection terminals, and partial discharge detection terminals. Partial discharge monitoring terminals can include ultrasonic partial discharge detection terminals and transient partial discharge detection terminals. Multiple live detection terminals utilize a unified sensor fusion design to form a detection terminal that is easily mounted on a robotic arm. This terminal can be mounted on a robotic arm to perform controlled live detection operations. The infrared thermal imaging detection terminal connects to a smart mobile terminal via wireless LAN technology (Wi-Fi), using the Hypertext Transfer Protocol (HTTP). It receives detection commands in real time and automatically transmits data maps and status information. The voiceprint imaging detection terminal connects to a smart mobile terminal via Bluetooth, using the Bluetooth Serial Port Protocol (BTSPP). It receives detection commands in real time and automatically transmits voiceprint imaging maps and status information. The partial discharge monitoring terminal can also connect to a smart mobile terminal via Bluetooth, using the BTSPP protocol. It receives detection commands in real time and transmits data packets in hexadecimal format. After forwarding, the cloud uses an analysis algorithm to parse the data into measurement data and map files. Infrared thermal imaging detection terminals are suitable for measuring the temperature of power equipment in substations and outputting temperature data. Voiceprint imaging detection terminals are suitable for monitoring power equipment in substations, obtaining images of superimposed partial discharge and ultrasonic partial discharge (PD), as well as ultrasonic average values, peak values, and spectra. Ultrasonic partial discharge detection terminals are suitable for detecting ultrasonic average values, peak values, and spectra of power equipment. Ultrasonic partial discharge detection terminals are suitable for detecting ultra-high frequency (UHF) average values, peak values, and spectra of power equipment. Manual detection equipment includes smart mobile terminals, infrared thermal imaging detection terminals, partial discharge detection terminals, and visual ultrasonic detection terminals. Manual detection equipment can be carried by personnel conducting substation inspections, using handheld devices or other means to conduct inspections of power equipment. Visual ultrasonic detection terminals enable visual inspection of power equipment, capturing ultrasonic images of power equipment for equipment status analysis based on these and other data. Online monitoring equipment includes digital metering instruments, infrared temperature sensors, high-frequency pulse current sensors, and partial discharge monitoring sensors. Digital metering instruments include various instruments for real-time monitoring of substation equipment, such as those for voltage and current. Infrared temperature sensors are suitable for monitoring the temperature of power equipment and the environment. High-frequency pulse current sensors are suitable for capturing high-frequency current pulse signals generated by partial discharge. By detecting high-frequency current pulse signals generated within power equipment, such as switchgear, the presence of partial discharge can be determined. Partial discharge detection sensors are suitable for detecting partial discharge in power equipment. These include ultrasonic partial discharge sensors and ultra-high frequency partial discharge sensors, each suitable for detecting ultrasonic average, peak, and spectrum information of power equipment. Ultra-high frequency partial discharge detection terminals are suitable for detecting ultra-high frequency average, peak, and spectrum information of power equipment.

[0107] The edge includes access node devices. These can be implemented as secure wireless access nodes for the power information intranet. These devices are broadband-narrowband converged communication devices, specifically designed to connect to robotic live-line detection equipment and manual inspection equipment via broadband IoT, and to online monitoring equipment via narrowband IoT. The access node devices also communicate with the cloud via a network. This network can be a comprehensive data network built for data communication purposes. One or more access node devices can be deployed on the edge.

[0108] The cloud can be implemented as one or more cloud servers. It manages the operations of end-side devices, integrates business data, and performs modal fitting and feature mining based on business data, thereby analyzing device status. Online monitoring devices are suitable for collecting monitoring data and uploading it to access node devices via narrowband IoT. After receiving the online monitoring data, the access node device uploads it to the business system deployed in the cloud server via the integrated data network.

[0109] The cloud, edge, and device sides together form a cloud-edge-device collaborative device status detection business system. The system adopts a cloud-edge collaborative strategy, enabling autonomous transmission of detection data, dispatching detection tasks, and re-detection of suspected faults.

[0110] S2. In response to the fault type determination request, obtain target power data of the target device according to the device identification.

[0111] Online monitoring equipment can continuously monitor various types of power equipment deployed in locations such as substations. Online monitoring equipment includes digital metering instruments, infrared temperature sensors, high-frequency pulse current sensors, and partial discharge monitoring sensors. Partial discharge detection sensors include ultrasonic partial discharge sensors and ultra-high frequency partial discharge sensors. Monitoring data includes voltage and current data collected by digital metering instruments; surface temperature data and ambient temperature data of power equipment collected by infrared temperature sensors; high-frequency current pulse signals collected by high-frequency pulse current sensors; ultrasonic partial discharge data collected by ultrasonic partial discharge sensors, including ultrasonic average value, peak value, and spectrum information; and ultra-high frequency partial discharge data collected by ultra-high frequency partial discharge sensors, including ultra-high frequency average value, peak value, and spectrum information. Each different type of monitoring data can be stored in a separate monitoring data file. The formats of different monitoring data files can vary. Target power data for the target device is obtained based on the device identifier.

[0112] Figure 3 This is a flow chart of a method for determining a device fault type provided by an optional embodiment of the present invention. Figure 3 As shown, S3 to S5 are the specific process of the method.

[0113] S3. Determine a plurality of historical data corresponding to the target power data.

[0114] In response to receiving monitoring data collected by an online monitoring device on the electric power device, the monitoring data is input into a large language model to determine a first diagnostic feature data set of the electric power device.

[0115] The large language model is pre-connected to the data and historical data is input into the large language model. The historical data includes multiple historical monitoring data points for each power device (similar to the historical power data described above), as well as the corresponding fault and maintenance records (similar to the historical fault types described above) for each item of historical monitoring data. Fault records record the fault type corresponding to the historical monitoring data, while maintenance records record the specific repair method used when the fault occurred. The large language model reads the historical data and then processes the historical data of different fault types, including document conversion. This process aims to complete various conversion tasks, such as removing redundant content and segmenting long documents into shorter sections to accommodate the context window size of the large language model. The large language model then applies the converted historical data documents to a text embedding model, converting the unstructured text into an embedded representation. This text-based embedded representation enables semantic search to find the most similar text segments. The historical data embedding vectors (similar to the historical data vectors described above) generated from the converted embedded representations are then stored for retrieval using the embedded vectors.

[0116] The monitoring data of the preset time period (the same as the target power data mentioned above) is input into the large language model. The large language model obtains the monitoring data file through the file location of each monitoring data file and reads the monitoring data of the preset time period; then processes the different types of monitoring data read, including document conversion. Subsequently, the large language model converts the converted monitoring data document into an embedded representation through a text embedding model; then the monitoring data embedding vector generated by the conversion into the embedded representation (the same as the target data vector mentioned above) is used as the first query content and input into the retriever to query multiple historical data documents that are most similar to the first query content.

[0117] Subsequently, multiple historical data documents and monitoring data documents are clustered according to a clustering algorithm, and each cluster obtained by clustering represents a fault type of the power equipment; then, one or more fault types closest to the monitoring data document are determined. Then, the first historical data document corresponding to one or more closest fault types is obtained as a first diagnostic feature data set. When determining the one or more fault types closest to the monitoring data document based on the clustering results, the distance between the monitoring data document and each cluster can be calculated based on the clustering results, and the fault type corresponding to one or more clusters with a distance less than a preset value is determined as the one or more fault types closest to the monitoring data document; or the fault type corresponding to a preset number of clusters with the smallest distance is selected as the one or more fault types closest to the monitoring data document.

[0118] S4. Determine target graph features corresponding to the target power data and historical graph features corresponding to the plurality of historical data.

[0119] The first diagnostic feature data set and the monitoring data are structurally transformed to generate knowledge graph features of the first diagnostic feature data set and the monitoring data.

[0120] When generating knowledge graph features for historical data documents in the first diagnostic feature data set, for the first historical data document of each fault type in the first diagnostic feature data set, a triple of the knowledge graph corresponding to the first historical data document is generated based on the data type included (such as the data type included in the monitoring data and historical monitoring data) and the data value corresponding to each data type (the same as the above-mentioned historical triple), and a knowledge graph of the first diagnostic feature data set is generated based on the triple of the knowledge graph corresponding to the first historical data document. Subsequently, feature extraction of multiple dimensions is performed based on the knowledge graph to generate knowledge graph features (the same as the above-mentioned historical graph features) for each fault type in the diagnostic feature data set. Feature extraction of multiple dimensions includes one-hop, two-hop, and three-hop reasoning on the knowledge graph to extract low-dimensional and high-dimensional features.

[0121] The process of generating the knowledge graph of monitoring data is also the same, including: generating triples of the knowledge graph corresponding to the monitoring data (the same as the target triples mentioned above) for the data types included in the power equipment recorded in the monitoring data document corresponding to the monitoring data (such as the data types included in the monitoring data and historical monitoring data) and the data values ​​corresponding to each data type, and generating the knowledge graph of the monitoring data based on the triples of the knowledge graph corresponding to the monitoring data. Subsequently, feature extraction of multiple dimensions is performed based on the knowledge graph of the monitoring data to generate knowledge graph features of the monitoring data (the same as the target graph features mentioned above). Feature extraction of multiple dimensions includes one-hop, two-hop and three-hop reasoning on the knowledge graph to extract low-dimensional and high-dimensional features.

[0122] When generating knowledge graph features based on diagnostic feature datasets and monitoring data, time domain alignment is also performed. Data values ​​of different data types collected at the same time within a preset time period or a time period included in historical data are extracted to generate time domain aligned data values ​​and knowledge graph features.

[0123] S5. Determine the target fault type of the target device based on the target graph feature and multiple historical graph features.

[0124] The first diagnostic feature data set and the knowledge graph features of the monitoring data are input into the fault type diagnosis model to obtain a first predicted fault type.

[0125] The fault type diagnosis model can be trained using a neural network model. The fault type diagnosis model receives knowledge graph features of one or more fault types and knowledge graph features of monitoring data, makes judgments on the monitoring data, and outputs the probability value corresponding to each fault type of the power equipment.

[0126] The fault type diagnosis model calculates the similarity between the knowledge graph features of the monitoring data and the knowledge graph features of each fault type, and determines the probability value corresponding to each fault type of the power equipment based on the similarity. When calculating the similarity, vectorized comparison can be used, and the similarity score between the knowledge graph features of the monitoring data and the knowledge graph features of each fault type can be calculated using cosine similarity, Manhattan distance, etc. Subsequently, the similarity score is normalized and the result is converted into a probability distribution to ensure that the sum of the probabilities of all fault types is 1. Subsequently, the fault type with the highest probability value is used as the first predicted fault type (the same as the target fault type mentioned above).

[0127] Subsequently, a re-inspection task is generated based on the output of the fault type diagnosis model and sent to both manual inspection equipment and robotic live-line inspection equipment. After analyzing the monitoring data of the power equipment, if the equipment status is determined to be abnormal, a re-inspection task is generated based on the fault type. Re-inspection tasks include manual inspection tasks and robotic linkage tasks. The cloud server sends the re-inspection tasks to the access node devices, which then send the manual inspection tasks to the manual inspection equipment and the robotic linkage tasks to the robotic live-line inspection equipment.

[0128] Each re-inspection task includes one or more inspection items determined based on the predicted fault type. When determining the inspection items based on the predicted fault type, all possible fault types output by the fault type diagnosis model are determined, and the determination is made based on the historical monitoring data corresponding to each possible fault type and the monitoring data when no fault occurs. If the monitoring data of this type deviates significantly from the normal value of the monitoring data when a fault occurs (higher or lower than 20% of the normal value), then the monitoring data of this type is determined to be the inspection item that needs to be re-inspected. Subsequently, a re-inspection task is generated based on one or more inspection items.

[0129] When a manual inspection task is assigned to a manual inspection device, the intelligent mobile terminal receives the task. The inspector responsible for the manual inspection task then uses manual inspection equipment, including one or more of infrared thermal imaging, partial discharge, and visual ultrasonic inspection terminals, to conduct a manual re-inspection of the power equipment based on the inspection items included in the task. The intelligent mobile terminal then sends the re-inspection data to the cloud server via the access node device.

[0130] When a robot linkage task is issued to the robot live-line detection device, the intelligent mobile terminal in the robot live-line detection device receives the task. The multifunctional inspection robot then uses the robot detection devices, including one or more of the following: infrared thermal imaging detection terminal, voiceprint imaging detection terminal, and partial discharge detection terminal, to conduct a robot re-inspection of the power equipment based on the inspection items included in the robot linkage task. The intelligent mobile terminal in the robot live-line detection device then transmits the re-inspection data to the cloud server via the access node device.

[0131] Subsequently, in response to receiving the re-inspection data, the fault type of the power equipment is determined based on the re-inspection data. After receiving the re-inspection data, the large language model performs document conversion on the re-inspection data. The converted re-inspection data document is then passed through a text embedding model to convert the unstructured text into an embedded representation. The data embedding vector corresponding to the re-inspection data generated from the conversion into the embedded representation is then input into a search engine to search for multiple historical data documents that are most similar to the second query. Subsequently, the multiple historical data documents and monitoring data documents are clustered using a clustering algorithm to determine one or more fault types that are most similar to the re-inspection data documents. One or more second historical data documents corresponding to the most similar fault types are then obtained as a second diagnostic feature dataset. The second diagnostic feature dataset and the monitoring data are structured and converted to generate knowledge graph features for the second diagnostic feature dataset and the monitoring data. The knowledge graph features for the second diagnostic feature dataset and the monitoring data are then input into a fault type diagnosis model. The fault type is determined based on the second predicted fault type output by the fault type diagnosis model and the first preset fault type.

[0132] If the second predicted fault type is the same as the first predicted fault type, the same predicted fault type is determined as the final fault type. If the second predicted fault type is different from the first predicted fault type, the predicted fault type with the highest probability value between the first predicted fault type and the second predicted fault type is determined as the final fault type.

[0133] After determining the fault type, the monitoring data and re-inspection data are labeled according to the fault type. As historical monitoring data, the corresponding fault type is stored in the fault record to update the training data of the large language model.

[0134] After determining the fault type, maintenance information is generated based on the maintenance method corresponding to the fault type in the historical data, and the maintenance information is pushed so that the operation and maintenance personnel can handle the fault of this type according to the maintenance information.

[0135] Through the above optional implementation, at least the following beneficial effects can be achieved:

[0136] (1) By comparing the target graph features with multiple historical graph features, the similarity of the multi-dimensional feature information corresponding to the target power data and the historical power data is analyzed to determine the fault type corresponding to the target power data, thereby improving the accuracy of determining the equipment fault type;

[0137] (2) Convert the target power data into a target data vector, compare the similarity between the target data vector and other data vectors, determine multiple historical data vectors, improve the accuracy of the determined historical data vectors, and then determine multiple historical data with high similarity to the target power data, improve the accuracy of the determined historical data, and improve the similarity between the historical map features and the target map features, so as to more accurately determine the equipment fault type corresponding to the target power data.

[0138] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the present invention is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0139] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0140] Example 2

[0141] According to an embodiment of the present invention, there is also provided a device for implementing the above-mentioned method for determining the type of equipment failure. Figure 4 FIG. 1 is a structural block diagram of a device for determining a device fault type according to an embodiment of the present invention. Figure 4 As shown, the apparatus includes: a receiving module 402, a responding module 404, a first determining module 406, a second determining module 408 and a third determining module 410. The apparatus will be described in detail below.

[0142] The receiving module 402 is used to receive a fault type determination request, wherein the fault type determination request carries a device identification of the target device; the responding module 404 is connected to the above-mentioned receiving module 402, and is used to respond to the fault type determination request and obtain the target power data of the target device according to the device identification; the first determining module 406 is connected to the above-mentioned responding module 404, and is used to determine a plurality of historical data corresponding to the target power data, wherein the plurality of historical data respectively include historical power data and historical fault types corresponding to the historical power data, and the similarity between the historical power data and the target power data is greater than a predetermined threshold; the second determining module 408 is connected to the above-mentioned first determining module 406, and is used to determine whether the target power data corresponds to the target power data. The corresponding target map feature and the historical map features corresponding to the multiple historical data respectively, wherein the target map feature is determined based on the target triplet, the target triplet is used to represent the relationship between the multiple data items in the target power data and the corresponding data values, the corresponding historical map feature is determined based on the corresponding first triplet and the corresponding second triplet, the corresponding first triplet is used to represent the relationship between the multiple data items in the historical power data and the corresponding data values, and the corresponding second triplet is used to represent the relationship between the historical power data and the historical fault type corresponding to the historical power data; a third determination module 410, connected to the above-mentioned second determination module 408, is used to determine the target fault type of the target device based on the target map feature and the multiple historical map features.

[0143] It should be noted here that the above-mentioned receiving module 402, response module 404, first determination module 406, second determination module 408 and third determination module 410 correspond to steps S102 to S110 in the method for implementing the equipment fault type determination method. The instances and application scenarios implemented by multiple modules and corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiment 1.

[0144] Example 3

[0145] According to another aspect of an embodiment of the present invention, an electronic device is provided, including: a processor; and a memory for storing processor-executable instructions, wherein the processor is configured to execute the instructions to implement any of the above-mentioned methods for determining a device fault type.

[0146] Example 4

[0147] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device can execute any of the above-mentioned methods for determining a device fault type.

[0148] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0149] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0150] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0151] The units described as separate components may or may not be physically separate, and 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 units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0152] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0153] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, 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. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc. Various media that can store program codes.

[0154] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for determining a device fault type, characterized in that: include: receiving a fault type determination request, wherein the fault type determination request carries a device identifier of a target device; In response to the fault type determination request, acquiring target power data of the target device according to the device identifier; Determining a plurality of historical data corresponding to the target power data, wherein the plurality of historical data respectively include historical power data and historical fault types corresponding to the historical power data, and a similarity between the historical power data and the target power data is greater than a predetermined threshold; Determine a target graph feature corresponding to the target power data and historical graph features corresponding to the multiple historical data, wherein the target graph feature is determined based on a target triplet, the target triplet is used to represent the relationship between multiple data items in the target power data and corresponding data values, and the corresponding historical graph feature is determined based on a corresponding first triplet and a corresponding second triplet, the corresponding first triplet is used to represent the relationship between multiple data items in the historical power data and corresponding data values, and the corresponding second triplet is used to represent the relationship between the historical power data and the historical fault type corresponding to the historical power data; A target fault type of the target device is determined based on the target graph feature and multiple historical graph features.

2. The method according to claim 1, characterized in that The determining of a plurality of historical data corresponding to the target power data includes: Determining a target data vector corresponding to the target power data; Determining a plurality of historical data vectors based on the target data vector, wherein vector similarities corresponding to the plurality of historical data vectors are greater than a predetermined similarity threshold, and the corresponding vector similarities are similarities between the corresponding historical data vectors and the target data vector; A plurality of historical data corresponding to the target power data is determined according to the plurality of historical data vectors.

3. The method according to claim 1, characterized in that The determining of the target graph features corresponding to the target power data and the historical graph features corresponding to the plurality of historical data includes: Retrieving a graph feature determination model, wherein the graph feature determination model is provided with target model parameters, the target model parameters are determined based on initial model parameters and sample data, the sample data including sample power data and sample graph features corresponding to the sample power data; The target power data and the corresponding historical data are respectively input into the first module of the graph feature determination model to obtain a target triplet corresponding to the target power data and a historical triplet corresponding to the corresponding historical data, and the target triplet and the corresponding historical triplet are respectively input into the second module of the graph feature determination model to obtain the target graph feature and the corresponding historical graph feature, wherein the corresponding historical triplet includes the corresponding first triplet and the corresponding second triplet.

4. The method according to claim 3, characterized in that Before retrieving the atlas feature determination model, the method further includes: determining an initial feature determination model, wherein the initial feature determination model is provided with initial model parameters; Determine the model and sample power data based on the initial features to determine the initial map features; Determining a characteristic error value between the initial atlas characteristic and the sample atlas characteristic; When the characteristic error value is less than an error threshold, determining the initial model parameters as target model parameters; The atlas feature determination model is determined based on the target model parameters and the initial feature determination model.

5. The method according to claim 2, characterized in that The determining, based on the plurality of historical data vectors, a plurality of historical data corresponding to the target power data includes: Determining target distances between the target power data and a plurality of target cluster centers, wherein the plurality of target cluster centers are determined based on a plurality of initial cluster centers, and the plurality of initial cluster centers are determined based on the power data corresponding to the plurality of historical data vectors; Determining, based on a plurality of target distances, a plurality of historical fault types corresponding to the target power data, wherein the target distances respectively corresponding to the plurality of historical fault types are less than a predetermined distance threshold; Determining historical power data corresponding to the multiple historical fault types respectively; A plurality of historical data corresponding to the target power data is determined according to the plurality of historical fault types and the historical power types respectively corresponding to the plurality of historical fault types.

6. The method according to claim 1, characterized in that The determining the target fault type of the target device based on the target graph feature and the plurality of historical graph features includes: Determining similarity values ​​corresponding to the plurality of historical graph features and the target graph feature respectively; Determining, based on the multiple similarity values, type indexes of the historical fault types corresponding to the multiple historical graph features respectively; A target fault type corresponding to the target device is determined according to a plurality of type indexes.

7. The method according to any one of claims 1 to 6, characterized in that After determining the target fault type of the target device based on the target graph feature and the plurality of historical graph features, the method further includes: Determining a re-inspection instruction based on the target fault type of the target device, wherein the re-inspection instruction carries a type identifier corresponding to a re-inspection item type, and the re-inspection item type is determined based on the target fault type of the target device; sending the re-inspection instruction to the re-inspection device; The target fault type of the target device is updated according to the received re-inspection data, and an updated fault type corresponding to the target device is obtained.

8. A device for determining a device fault type, characterized in that: include: A receiving module, configured to receive a fault type determination request, wherein the fault type determination request carries a device identifier of a target device; a response module, configured to obtain target power data of the target device according to the device identifier in response to the fault type determination request; a first determining module, configured to determine a plurality of historical data corresponding to the target power data, wherein the plurality of historical data respectively include historical power data and historical fault types corresponding to the historical power data, and a similarity between the historical power data and the target power data is greater than a predetermined threshold; a second determination module, configured to determine a target map feature corresponding to the target power data and historical map features corresponding to the plurality of historical data, wherein the target map feature is determined based on a target triplet, the target triplet is used to represent the relationship between a plurality of data items in the target power data and corresponding data values, the corresponding historical map feature is determined based on a corresponding first triplet and a corresponding second triplet, the corresponding first triplet is used to represent the relationship between a plurality of data items in the historical power data and corresponding data values, and the corresponding second triplet is used to represent the relationship between the historical power data and the historical fault type corresponding to the historical power data; The third determination module is used to determine the target fault type of the target device based on the target graph feature and multiple historical graph features.

9. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the method for determining a device fault type according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the method for determining a device fault type according to any one of claims 1 to 7.

Images