Power distribution network equipment fault identification method and device and storage medium
By collecting and processing the operating status monitoring data of distribution network equipment, using the spatial attention layer and conditioned reflection characteristics, the accurate identification of head, tail and open faults is achieved, solving the problem of low accuracy in fault identification in the existing technology, and improving the accuracy of equipment fault identification.
Patent Information
- Application Number
- CN202411674237.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-21
- Publication Date
- 2025-03-11
AI Technical Summary
The existing distribution network equipment fault identification methods are difficult to cover all possible fault types comprehensively and accurately, resulting in low accuracy in equipment fault identification.
By collecting operating status monitoring data of the distribution network, using a pre-trained open long-tail recognition model, the equipment failure types are identified as head, tail and open. Then, the direct feature set is extracted and spliced through the spatial attention layer, data augmentation is performed to obtain conditioned features, and finally the feature embedding vector is determined based on these features, and model training is performed to achieve fault recognition.
By highlighting the key features in the direct feature set and suppressing unimportant information, the model's ability to identify fault features is improved and the accuracy of equipment fault recognition is improved. At the same time, the introduction of similarity weights accurately identifies slight differences between different categories in the case of data imbalance or category overlap, further improving the accuracy of fault category identification.
Smart Images
Figure CN119598167A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of fault identification, and in particular to a method, a device and a storage medium for identifying faults of distribution network equipment. Background Art
[0002] In distribution network automation equipment, fault types often show an obvious long-tail distribution, that is, some common fault types occur more frequently, which are defined as head faults. Other rare fault types only occur under certain conditions, which are defined as tail faults. In addition, the continuous commissioning of new equipment has led to the emergence of some new fault types that have never been seen in actual operation, which are defined as open faults. The accurate identification of these unknown faults is also crucial.
[0003] Head-type faults, such as poor switch contact, equipment heating, abnormal waveform, etc., exist in large numbers in historical data and are easy to identify. Tail-type faults are relatively rare but very important, such as faults caused by natural disasters or human factors. In addition, distribution network automation equipment may also encounter some open-type faults. Existing distribution network equipment fault identification methods are difficult to fully and accurately cover all possible fault types, resulting in low accuracy in equipment fault identification. Summary of the invention
[0004] The present invention provides a distribution network equipment fault identification method, device and storage medium to solve the technical problem that the existing distribution network equipment fault identification method is difficult to comprehensively and accurately cover all possible fault types, resulting in low accuracy in equipment fault identification.
[0005] The present invention provides a method for identifying a fault in a power distribution network device, comprising:
[0006] Collecting operation status monitoring data of the distribution network, the operation status monitoring data includes equipment monitoring data and node monitoring data, the equipment monitoring data includes equipment temperature data, equipment life data and equipment performance data, and the node monitoring data includes node voltage data and node current data;
[0007] Based on a pre-trained open long-tail recognition model, the equipment fault type is identified into a head class, a tail class and an open class according to the operation status monitoring data;
[0008] Extracting a direct feature set of the running status monitoring data, and performing splicing processing on the direct feature set according to a spatial attention layer to obtain a new direct feature set; wherein the spatial attention layer is constructed according to the number of the head classes and the number of the tail classes;
[0009] Performing data enhancement processing according to the new direct feature set to obtain conditional reflection features;
[0010] Determine a feature embedding vector based on the new direct feature set and the conditional reflection feature, perform model training according to the feature embedding vector, and obtain a fault recognition model;
[0011] The data to be identified is input into the fault identification model to obtain the equipment fault identification result.
[0012] Furthermore, the direct feature set is concatenated according to the spatial attention layer to obtain a new direct feature set, including:
[0013] The spatial attention layer is used to extract the head data set and the tail data set in the direct feature set, and the fault classification feature data corresponding to the head data set and the tail data set are respectively spliced to obtain a new direct feature set.
[0014] Furthermore, performing data enhancement processing according to the new direct feature set to obtain conditional reflection features includes:
[0015] Performing data enhancement processing on the new direct features to obtain enhanced features;
[0016] Performing small batch clustering processing according to the new direct features to form cluster center results;
[0017] The enhanced features are multiplied with the cluster center results for training, and the training results are enhanced using the tail class to obtain conditional reflection features.
[0018] Further, determining a feature embedding vector based on the new direct feature set and the conditional reflection feature includes:
[0019] Use activation function to generate feature selection coefficients;
[0020] Determine the similarity weight based on the distance between the input feature and the center point of each class of the conditional reflection feature;
[0021] A feature embedding vector is determined according to the feature selection coefficient, the similarity weight and the conditional reflection feature.
[0022] Furthermore, the expression of the feature embedding vector is as follows:
[0023]
[0024] Among them, x embed is the feature embedding vector, δ is the similarity weight, x direct is the input feature, e is the feature selection coefficient, x reflection It is a conditioned reflex characteristic.
[0025] The present invention provides a distribution network equipment fault identification device, comprising:
[0026] A detection data acquisition module is used to collect the operation status monitoring data of the distribution network, wherein the operation status monitoring data includes equipment monitoring data and node monitoring data, wherein the equipment monitoring data includes equipment temperature data, equipment life data and equipment performance data, and the node monitoring data includes node voltage data and node current data;
[0027] A fault classification module, used for identifying the equipment fault type into a head class, a tail class and an open class according to the operation status monitoring data based on a pre-trained open long-tail recognition model;
[0028] A feature concatenation module, used for extracting a direct feature set of the running status monitoring data, and concatenating the direct feature set according to a spatial attention layer to obtain a new direct feature set; wherein the spatial attention layer is constructed according to the number of the head classes and the number of the tail classes;
[0029] A data enhancement module, used for performing data enhancement processing according to the new direct feature set to obtain conditional reflection features;
[0030] A model training module, used to determine a feature embedding vector based on the new direct feature set and the conditional reflection feature, and perform model training according to the feature embedding vector to obtain a fault recognition model;
[0031] The fault identification module is used to input the data to be identified into the fault identification model to obtain the equipment fault identification result.
[0032] Furthermore, the feature stitching module is also used for:
[0033] The spatial attention layer is used to extract the head data set and the tail data set in the direct feature set, and the fault classification feature data corresponding to the head data set and the tail data set are respectively spliced to obtain a new direct feature set.
[0034] Furthermore, the data enhancement module is also used for:
[0035] Performing data enhancement processing on the new direct features to obtain enhanced features;
[0036] Performing small batch clustering processing according to the new direct features to form cluster center results;
[0037] The enhanced features are multiplied with the cluster center results for training, and the training results are enhanced using the tail class to obtain conditional reflection features.
[0038] The present invention also provides a terminal device, comprising: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, the distribution network equipment fault identification method as described above is implemented.
[0039] The present invention also provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the distribution network equipment fault identification method as described above.
[0040] The embodiment of the present invention constructs an attention mechanism according to the number of head classes and the number of tail classes. The direct feature set is spliced through the attention mechanism, which can effectively highlight the key features in the direct feature set and suppress unimportant information, thereby improving the model's ability to recognize fault features, and further effectively improving the accuracy of equipment fault identification.
[0041] Furthermore, the embodiment of the present invention introduces similarity weights, which can accurately identify slight differences between different categories in the presence of data imbalance or category overlap, thereby further improving the accuracy of identifying different fault categories of the device. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Figure 1 It is a flow chart of a method for identifying a fault in a distribution network device provided by an embodiment of the present invention;
[0043] Figure 2 It is a schematic diagram of a feature processing flow provided by an embodiment of the present invention;
[0044] Figure 3 It is a structural schematic diagram of a distribution network equipment fault identification device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0045] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0046] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.
[0047] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.
[0048] See also Figure 1 The present invention provides a method for identifying a fault in a distribution network device, comprising:
[0049] S1. Collecting the operation status monitoring data of the distribution network, the operation status monitoring data includes equipment monitoring data and node monitoring data, the equipment monitoring data includes equipment temperature data, equipment life data and equipment performance data, and the node monitoring data includes node voltage data and node current data;
[0050] In an embodiment of the present invention, the operation status monitoring data of the distribution network can be collected by sensors to construct a multi-dimensional time series feature data set. When collecting data, a high-frequency sampling method is used to ensure the real-time and accuracy of the data, especially the data information during the period when the fault occurs. Among them, the operation status monitoring data may include image data and text data, and the operation status monitoring data may also include fault indicators of the equipment, including node voltage, current vector and waveform, etc.
[0051] S2, based on the pre-trained open long-tail recognition model, identify the equipment fault type into head class, tail class and open class according to the operation status monitoring data;
[0052] In the embodiment of the present invention, an open long-tail recognition model may be pre-trained, and the model may identify equipment fault types based on operating status monitoring data.
[0053] S3, extracting a direct feature set of the running status monitoring data, and concatenating the direct feature set according to the spatial attention layer to obtain a new direct feature set; wherein the spatial attention layer is constructed according to the number of head classes and the number of tail classes;
[0054] In the embodiment of the present invention, the operating status monitoring data can be used as the original data to perform multi-layer convolution operations to extract standard features, that is, the input feature set. Direct features are standard features extracted from the input data, usually from the penultimate layer of the convolutional neural network. Direct features can be updated through random gradient descent of training data and can reflect the basic information of the input data.
[0055] In an embodiment of the present invention, the spatial attention layer may be constructed in the following manner:
[0056] The number of head classes and tail classes obtained by statistical recognition are used to construct the expression of the spatial attention layer based on the number of head classes and the number of tail classes:
[0057]
[0058] Among them, N head and N tail They represent the number of head classes and tail classes respectively; MA(f) and SA(f) represent two different attention layers. The weighting of the tail class is confirmed by comparing the specific number of head and tail classes. The fewer the number of tail samples, the higher the attention weight is required.
[0059] The embodiment of the present invention performs splicing processing on the direct feature set through the attention mechanism, which can effectively highlight the key features in the direct feature set and suppress unimportant information, thereby improving the model's ability to recognize fault features.
[0060] S4, perform data enhancement processing according to the new direct feature set to obtain conditional reflection features;
[0061] See also Figure 2 In an embodiment of the present invention, a learning module can be constructed, and data enhancement processing is performed by the learning module according to a new direct feature set to obtain a conditional reflection feature.
[0062] S5. Determine a feature embedding vector based on the new direct feature set and the conditional reflection feature, perform model training according to the feature embedding vector, and obtain a fault recognition model;
[0063] S6. Input the data to be identified into the fault identification model to obtain the equipment fault identification result.
[0064] The embodiment of the present invention constructs an attention mechanism according to the number of head classes and the number of tail classes. The direct feature set is spliced through the attention mechanism, which can effectively highlight the key features in the direct feature set and suppress unimportant information, thereby improving the model's ability to recognize fault features, and further effectively improving the accuracy of equipment fault identification.
[0065] In one embodiment, step S3, concatenating the direct feature set according to the spatial attention layer to obtain a new direct feature set includes:
[0066] The spatial attention layer is used to extract the head data set and the tail data set in the direct feature set, and the fault classification feature data corresponding to the head data set and the tail data set are concatenated to obtain a new direct feature set.
[0067] In the embodiment of the present invention, by extracting the features of the head and tail data sets respectively, different fault types can be covered more comprehensively, thereby improving the accuracy of fault classification.
[0068] In one embodiment, step S4, performing data enhancement processing according to the new direct feature set to obtain conditional reflection features, includes:
[0069] S41, performing data enhancement processing on the new direct features to obtain enhanced features;
[0070] S42, performing small batch clustering processing according to the new direct features to form cluster center results;
[0071] In the embodiment of the present invention, some new direct features are extracted for small batch clustering. During the training process, the model samples samples from each category to build a small batch. These samples are used to calculate the class center and update the model memory features to form the cluster center result:
[0072]
[0073] S43, performing cumulative training on the enhanced features and the cluster center results, and using the tail class to enhance the training results to obtain conditional reflex features.
[0074] In this embodiment of the present invention, by i Multiply it by its corresponding weight o i , and aggregate these weighted class centers to calculate a new conditional reflection feature x reflection , the reflection feature is designed to enhance the direct features of the input image, especially when training data is scarce (such as in the tail class). The expression of the conditional reflection feature is as follows:
[0075]
[0076] In one embodiment, step S5, determining a feature embedding vector based on the new direct feature set and the conditional reflection feature, comprises:
[0077] S51, using activation function to generate feature selection coefficients;
[0078] In the embodiment of the present invention, an activation function may be used to analyze and select new direct features to generate feature selection coefficients.
[0079] In the embodiment of the present invention, the feature embedding vector can be determined by constructing a learning module, and the combination of direct features and conditioned reflex features is considered in the learning process. Since the head class (data-rich class) and the tail class (data-scarce class) have different feature requirements, the concept Xiangcu Anze adapts to these differences through learning. For the head class, direct features are usually sufficient, and fewer memory features can be selected; for the tail class, conditioned reflex features are particularly important.
[0080] In the embodiment of the present invention, a suitable feature combination can be selected according to the properties of the input features, and the proportion of direct features and conditioned reflex features to be used in a specific situation can be dynamically determined. Specifically, in a lightweight neural network, an activation function is used to generate a selection coefficient, and the combination of direct features and conditioned reflex features is weighted by the feature selection coefficient to form a final feature identification. The expression of the feature selection coefficient is as follows:
[0081] e = tanh(T(x direct ))
[0082] Among them, tanh represents the activation function, e represents the feature selection coefficient, and the output e of the concept selector is obtained by directly selecting the feature x. direct The T neural network is applied to determine how to combine direct features and memory features.
[0083] S52, determining a similarity weight according to the distance between the input feature and the center point of each class of the conditional reflection feature;
[0084] In this embodiment of the present invention, the similarity weight is used to measure the input feature x direct and conditioned reflex feature x reflection The center point x of each class in i Specifically, δ is calculated by finding x direct With all class centers x i The minimum Euclidean distance between
[0085]
[0086] S53: Determine a feature embedding vector according to the feature selection coefficient, the similarity weight and the conditional reflection feature.
[0087] In one embodiment, the expression of the feature embedding vector is as follows:
[0088]
[0089] Among them, x embed is the feature embedding vector, δ is the similarity weight, x direct is the input feature, e is the feature selection coefficient, xreflection is a conditional reflex feature. Represents element-wise multiplication.
[0090] In the embodiment of the present invention, the influence of direct features and conditional reflection features on the prediction of tail-type faults is comprehensively considered, thereby effectively improving the accuracy of identifying tail-type equipment faults.
[0091] In the embodiment of the present invention, adaptive similarity can not only optimize the overall accuracy of all categories, but also provide a potential solution for fairness analysis, so that the performance of the model can be improved when processing long-tail distribution data. By calculating the distance between the direct feature and the neural reflection memory center, the weight of the similarity can be dynamically adjusted: when the direct feature is close to the neuron center, the weight is large, indicating that the sample may belong to a known class. When the direct feature is far away from the feature neural center, the weight decreases, indicating that the sample may belong to an open class or an unknown class.
[0092] The implementation of the embodiments of the present invention has the following beneficial effects:
[0093] The embodiment of the present invention constructs an attention mechanism according to the number of head classes and the number of tail classes. The direct feature set is spliced through the attention mechanism, which can effectively highlight the key features in the direct feature set and suppress unimportant information, thereby improving the model's ability to recognize fault features, and further effectively improving the accuracy of equipment fault identification.
[0094] Furthermore, the embodiment of the present invention introduces similarity weights, which can accurately identify slight differences between different categories in the presence of data imbalance or category overlap, thereby further improving the accuracy of identifying different fault categories of the device.
[0095] See also Figure 3 Based on the same inventive concept as the above embodiment, the present invention provides a distribution network equipment fault identification device, comprising:
[0096] The detection data acquisition module 10 is used to collect the operation status monitoring data of the distribution network, the operation status monitoring data includes equipment monitoring data and node monitoring data, the equipment monitoring data includes equipment temperature data, equipment life data and equipment performance data, and the node monitoring data includes node voltage data and node current data;
[0097] A fault classification module 20 is used to identify the equipment fault type into a head class, a tail class and an open class according to the operation status monitoring data based on a pre-trained open long-tail recognition model;
[0098] A feature concatenation module 30 is used to extract a direct feature set of the running status monitoring data, and concatenate the direct feature set according to the spatial attention layer to obtain a new direct feature set; wherein the spatial attention layer is constructed according to the number of head classes and the number of tail classes;
[0099] A data enhancement module 40 is used to perform data enhancement processing according to the new direct feature set to obtain conditional reflection features;
[0100] A model training module 50 is used to determine a feature embedding vector based on a new direct feature set and a conditional reflection feature, and to perform model training according to the feature embedding vector to obtain a fault recognition model;
[0101] The fault identification module 60 is used to input the data to be identified into the fault identification model to obtain the equipment fault identification result.
[0102] In one embodiment, the feature stitching module is further used to:
[0103] The spatial attention layer is used to extract the head data set and the tail data set in the direct feature set, and the fault classification feature data corresponding to the head data set and the tail data set are concatenated to obtain a new direct feature set.
[0104] In one embodiment, the data enhancement module is further configured to:
[0105] Perform data enhancement processing on the new direct features to obtain enhanced features;
[0106] Perform small batch clustering based on new direct features to form cluster center results;
[0107] The enhanced features are multiplied with the cluster center results for training, and the training results are enhanced using the tail class to obtain the conditional reflection features.
[0108] In one embodiment, determining a feature embedding vector based on a new direct feature set and a conditional reflection feature includes:
[0109] Use activation function to generate feature selection coefficients;
[0110] Determine the similarity weight based on the distance between the input feature and the center point of each class of the conditional reflection feature;
[0111] The feature embedding vector is determined based on the feature selection coefficient, similarity weight and conditional reflection feature.
[0112] In one embodiment, the expression of the feature embedding vector is as follows:
[0113]
[0114] Among them, xembed is the feature embedding vector, δ is the similarity weight, x direct is the input feature, e is the feature selection coefficient, x reflection It is a conditioned reflex characteristic.
[0115] Correspondingly, an embodiment of the present invention also provides a terminal device, including: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the distribution network equipment fault identification method as described in any one of the above embodiments.
[0116] The terminal device of this embodiment includes: a processor, a memory, and a computer program and computer instructions stored in the memory and executable on the processor. When the processor executes the computer program, each step in the above embodiment 1 is implemented, for example Figure 1 Alternatively, when the processor executes the computer program, the functions of each module / unit in the above device embodiment, such as the model training module 50, are implemented.
[0117] Exemplarily, the computer program can be divided into one or more modules / units, one or more modules / units are stored in a memory and executed by a processor to complete the present invention. One or more modules / units can be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program in the terminal device. For example, the model training module 50 is used to determine the feature embedding vector based on the new direct feature set and the conditional reflection feature, and perform model training based on the feature embedding vector to obtain a fault recognition model.
[0118] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The terminal device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the schematic diagram is merely an example of a terminal device and does not constitute a limitation on the terminal device. The terminal device may include more or fewer components than shown in the diagram, or may combine certain components, or different components. For example, the terminal device may also include an input / output device, a network access device, a bus, etc.
[0119] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.
[0120] The memory can be used to store computer programs and / or modules. The processor realizes various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile terminal, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Med ia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (F l ash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.
[0121] Among them, if the module / unit integrated in the terminal device 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 present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, RandomAccess Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electric carrier signals and telecommunication signals.
[0122] Correspondingly, an embodiment of the present invention also provides a computer-readable storage medium, the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the distribution network equipment fault identification method as described in any one of the above embodiments.
[0123] The above specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.
Claims
1. A method for identifying faults in distribution network equipment, characterized in that: include: Collecting operation status monitoring data of the distribution network, the operation status monitoring data includes equipment monitoring data and node monitoring data, the equipment monitoring data includes equipment temperature data, equipment life data and equipment performance data, and the node monitoring data includes node voltage data and node current data; Based on a pre-trained open long-tail recognition model, the equipment fault type is identified into a head class, a tail class and an open class according to the operation status monitoring data; Extracting a direct feature set of the running status monitoring data, and performing splicing processing on the direct feature set according to a spatial attention layer to obtain a new direct feature set; wherein the spatial attention layer is constructed according to the number of the head classes and the number of the tail classes; Performing data enhancement processing according to the new direct feature set to obtain conditional reflection features; Determine a feature embedding vector based on the new direct feature set and the conditional reflection feature, perform model training according to the feature embedding vector, and obtain a fault recognition model; The data to be identified is input into the fault identification model to obtain the equipment fault identification result.
2. The method for identifying a fault in a distribution network device according to claim 1, wherein: The direct feature set is concatenated according to the spatial attention layer to obtain a new direct feature set, including: The spatial attention layer is used to extract the head data set and the tail data set in the direct feature set, and the fault classification feature data corresponding to the head data set and the tail data set are respectively spliced to obtain a new direct feature set.
3. The method for identifying a fault in a distribution network device according to claim 1, wherein: The data enhancement process is performed according to the new direct feature set to obtain the conditional reflection feature, including: Performing data enhancement processing on the new direct features to obtain enhanced features; Performing small batch clustering processing according to the new direct features to form cluster center results; The enhanced features are multiplied with the cluster center results for training, and the training results are enhanced using the tail class to obtain conditional reflection features.
4. The method for identifying a fault in a distribution network device according to claim 2, wherein: The determining of a feature embedding vector based on the new direct feature set and the conditional reflection feature comprises: Use activation function to generate feature selection coefficients; Determine the similarity weight based on the distance between the input feature and the center point of each class of the conditional reflection feature; A feature embedding vector is determined according to the feature selection coefficient, the similarity weight and the conditional reflection feature.
5. The method for identifying a fault in a distribution network device according to claim 4, wherein: The expression of the feature embedding vector is as follows: Among them, x embed is the feature embedding vector, δ is the similarity weight, x direct is the input feature, e is the feature selection coefficient, x reflection It is a conditioned reflex characteristic.
6. A distribution network equipment fault identification device, characterized in that: include: A detection data acquisition module is used to collect the operation status monitoring data of the distribution network, wherein the operation status monitoring data includes equipment monitoring data and node monitoring data, wherein the equipment monitoring data includes equipment temperature data, equipment life data and equipment performance data, and the node monitoring data includes node voltage data and node current data; A fault classification module, used for identifying the equipment fault type into a head class, a tail class and an open class according to the operation status monitoring data based on a pre-trained open long-tail recognition model; A feature concatenation module, used for extracting a direct feature set of the running status monitoring data, and concatenating the direct feature set according to a spatial attention layer to obtain a new direct feature set; wherein the spatial attention layer is constructed according to the number of the head classes and the number of the tail classes; A data enhancement module, used for performing data enhancement processing according to the new direct feature set to obtain conditional reflection features; A model training module, used to determine a feature embedding vector based on the new direct feature set and the conditional reflection feature, and perform model training according to the feature embedding vector to obtain a fault recognition model; The fault identification module is used to input the data to be identified into the fault identification model to obtain the equipment fault identification result.
7. The distribution network equipment fault identification device according to claim 6, characterized in that: The feature splicing module is also used for: The spatial attention layer is used to extract the head data set and the tail data set in the direct feature set, and the fault classification feature data corresponding to the head data set and the tail data set are respectively spliced to obtain a new direct feature set.
8. The distribution network equipment fault identification device according to claim 6, characterized in that: The data enhancement module is also used for: Performing data enhancement processing on the new direct features to obtain enhanced features; Performing small batch clustering processing according to the new direct features to form cluster center results; The enhanced features are multiplied with the cluster center results for training, and the training results are enhanced using the tail class to obtain conditional reflection features.
9. A terminal device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for identifying faults of distribution network equipment as described in any one of claims 1 to 5 is implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the distribution network equipment fault identification method according to any one of claims 1 to 5.