Transform feature word spectrum-based transformer oil chromatographic data anomaly identification method, device and equipment

Abstract

The invention discloses a transformer oil chromatogram data abnormity identification method based on Transform feature word spectrums, and the method comprises the steps: receiving inputted historical oil chromatogram time sequence data of a transformer, and carrying out the preprocessing through a preprocessing rule, and obtaining corresponding preprocessing data; mapping the preprocessed data according to a preset word spectrum mapping rule to obtain a corresponding feature word spectrum; encoding the feature word spectrum according to a preset feature word spectrum encoding model to generate a corresponding fault word spectrum vector; and performing anomaly recognition on the fault word spectrum vector according to a preset anomaly recognition model to obtain a corresponding anomaly recognition result. According to the transformer oil chromatography data anomaly identification method based on the Transform feature word spectrum, the fault word spectrum vector is obtained by mapping and encoding the historical oil chromatography time sequence data of the transformer, and then anomaly identification is performed on the fault word spectrum vector to obtain an accurate anomaly identification result; and the accuracy of carrying out abnormity identification on the transformer is greatly improved.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring and fault diagnosis technology, and in particular to a method, apparatus and equipment for anomaly identification of transformer oil chromatographic data based on Transformer feature word spectrum. Background Technology

[0002] Transformers in power equipment are key devices for voltage step-up or step-down conversion. Inside the transformer are coils immersed in transformer oil for insulation and heat dissipation. Current technology typically involves observing changes in the color of the transformer oil to determine if an anomaly or fault has occurred inside the transformer. However, this method relies on human experience, and different personnel may use different standards, leading to poor accuracy in anomaly identification. Therefore, existing methods for anomaly identification in transformer oil suffer from insufficient accuracy. Summary of the Invention

[0003] This invention provides a method, apparatus, and device for identifying anomalies in transformer oil chromatographic data based on Transformer feature word spectrum, aiming to solve the problem of insufficient accuracy in the existing methods for identifying anomalies in transformer oil.

[0004] In a first aspect, embodiments of the present invention provide a method for anomaly identification in transformer oil chromatographic data based on Transformer feature word spectrum, wherein the method includes: The system receives historical oil chromatography time-series data of the transformer and preprocesses it according to preprocessing rules to obtain the corresponding preprocessed data. The preprocessed data is mapped according to a preset word spectrum mapping rule to obtain the corresponding feature word spectrum; the feature word spectrum includes the oil chromatogram word vector for each time node; The feature word spectrum is encoded according to a preset feature word spectrum encoding model to generate a corresponding fault word spectrum vector; the fault word spectrum vector is used to characterize the operating status of the transformer within a continuous time period; Anomalies are identified in the fault word spectrum vectors based on a preset anomaly identification model, and the corresponding anomaly identification results are obtained.

[0005] Secondly, embodiments of the present invention also provide a transformer oil chromatographic data anomaly identification device based on Transformer feature word spectrum, wherein the device is used to perform the transformer oil chromatographic data anomaly identification method based on Transformer feature word spectrum as described in the first aspect above, and the device includes: The preprocessing unit is used to receive the input historical oil chromatography time series data of the transformer and preprocess it according to the preprocessing rules to obtain the corresponding preprocessed data; The mapping unit is used to map the preprocessed data according to the preset word spectrum mapping rules to obtain the corresponding feature word spectrum; the feature word spectrum includes the oil chromatogram word vector for each time node; The encoding unit is used to encode the feature word spectrum according to a preset feature word spectrum encoding model to generate a corresponding fault word spectrum vector; the fault word spectrum vector is used to characterize the operating status of the transformer within a continuous time period; An anomaly identification unit is used to identify anomalies in the fault word spectrum vector according to a preset anomaly identification model, and obtain the corresponding anomaly identification result.

[0006] Thirdly, embodiments of the present invention also provide a computer device, wherein the device includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When the processor executes the program stored in the memory, it implements the steps of the transformer oil chromatographic data anomaly identification method based on Transformer feature spectrum as described in the first aspect above.

[0007] Fourthly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the transformer oil chromatographic data anomaly identification method based on Transformer feature spectrum as described in the first aspect above.

[0008] This invention provides a method, apparatus, and device for anomaly identification in transformer oil chromatographic data based on Transformer feature word spectrum. The method includes: receiving input historical oil chromatographic time-series data of a transformer and preprocessing it according to preprocessing rules to obtain corresponding preprocessed data; mapping the preprocessed data according to preset word spectrum mapping rules to obtain corresponding feature word spectra; encoding the feature word spectra according to a preset feature word spectrum encoding model to generate corresponding fault word spectrum vectors; and performing anomaly identification on the fault word spectrum vectors according to a preset anomaly identification model to obtain corresponding anomaly identification results. The above-described anomaly identification method for transformer oil chromatographic data based on Transformer feature word spectrum significantly improves the accuracy of transformer anomaly identification by mapping and encoding historical oil chromatographic time-series data of the transformer to obtain fault word spectrum vectors, and then performing anomaly identification on the fault word spectrum vectors. Attached Figure Description

[0009] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart of a method for identifying anomalies in transformer oil chromatographic data based on Transformer feature word spectrum, provided in an embodiment of the present invention. Figure 2 A schematic block diagram of a transformer oil chromatographic data anomaly identification device based on Transformer feature word spectrum provided in an embodiment of the present invention; Figure 3 This is a schematic block diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0013] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0014] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0015] The embodiments of this invention provide a method for anomaly identification of transformer oil chromatographic data based on Transformer feature spectrum. This method is applied in a terminal device. The anomaly identification method of transformer oil chromatographic data based on Transformer feature spectrum is executed by application software installed in the terminal device. The terminal device is a device used to perform anomaly identification to analyze the oil chromatographic data of the transformer and obtain the anomaly identification result, such as a desktop computer, laptop computer, tablet computer or mobile phone.

[0016] like Figure 1 As shown, the method includes steps S110 to S140.

[0017] S110: Receive the input historical oil chromatography time series data of the transformer and preprocess it according to the preprocessing rules to obtain the corresponding preprocessed data.

[0018] Historical oil chromatography time-series data of transformers can be input into the terminal device. The terminal device will then preprocess the input historical oil chromatography time-series data according to the preprocessing rules to obtain the preprocessed data.

[0019] In a specific embodiment, step S110 includes the following sub-steps: excluding outliers from the historical oil chromatography time series data according to the outlier range in the preprocessing rules to obtain corresponding valid time series data; and supplementing the valid time series data with missing values ​​according to the missing value supplementation function in the preprocessing rules to obtain corresponding preprocessed data.

[0020] Historical oil chromatography time-series data contains concentration values ​​of multiple key characteristic gases at various time points. These time points are the data acquisition points. For example, if the vector data acquisition time point is set to 5 minutes, a set of concentration values ​​for multiple key characteristic gases will be collected every five minutes. These key characteristic gases may include H2, CH4, C2H2, C2H4, and C2H6. Specifically, the outlier range in the preprocessing rules can be used to determine whether the concentration values ​​of each key characteristic gas in the historical oil chromatography time-series data fall within the outlier range. If a concentration value falls within the outlier range, all concentration values ​​at the time point corresponding to that concentration value are excluded. Different key characteristic gases have their own different outlier ranges set in the preprocessing rules. For example, the outlier range for H2 is (-∞, 0) ∪ [500 mg / m³]. 3 After outlier removal from historical oil chromatography time series data, valid time series data can be obtained.

[0021] Furthermore, missing values ​​in the valid time-series data can be filled in using the missing value completion function in the preprocessing rules. This function can be based on the Lagrange mean value theorem. If the concentration value of a key characteristic gas is missing at a certain time point, the missing value completion function can be used to calculate the preceding and following concentration values ​​for that missing value, thus obtaining the corresponding completion value at the missing location. The preceding concentration value is the concentration value at the time point preceding the missing location, and the following concentration value is the concentration value at the time point following the missing location. After filling in the missing data values ​​in the valid time-series data, the preprocessed data is obtained.

[0022] S120. The preprocessed data is mapped according to the preset word spectrum mapping rules to obtain the corresponding feature word spectrum.

[0023] Furthermore, the preprocessed data is mapped according to a pre-set word spectrum mapping rule to obtain a feature word spectrum corresponding to the preprocessed data. The feature word spectrum includes oil chromatographic word vectors for each time point.

[0024] In a specific embodiment, step S120 includes the following sub-steps: extracting the corresponding multi-dimensional standard feature vector from the preprocessed data; mapping the multi-dimensional standard feature vector to a fixed-dimensional initial word vector space according to the linear embedding matrix in the word spectrum mapping rule to obtain an initial embedding vector; and superimposing residual enhancement on the initial embedding vector according to the residual enhancement function in the word spectrum mapping rule to obtain the corresponding enhancement vector as the feature word spectrum.

[0025] Specifically, the corresponding multi-dimensional standard feature vectors can be extracted from the preprocessed data. In addition to the feature vectors corresponding to the concentration values ​​of a single key feature gas, the extracted multi-dimensional standard feature vectors also include feature vectors corresponding to the correlation between the concentration values ​​of multiple key feature gases.

[0026] Furthermore, the multi-dimensional standard feature vectors obtained in the above steps are mapped to a fixed-dimensional initial word vector space according to the linear embedding matrix defined in the word spectrum mapping rules, to obtain the initial embedding vector. For example, the obtained multi-dimensional standard feature vector is X. t ={ , ,… }, by using linear embedding, multi-dimensional standard feature vectors are mapped to a fixed-dimensional initial word vector space: for ; Among them, e ti (0) W is the initial embedding vector. e ∈R d×9Let d be the embedding matrix, b be the word vector dimension, and d be the word vector dimension. e The parameters set for the initial word vector space are: the subscript t is the time parameter (corresponding to different time nodes), i is the dimension parameter of the multi-dimensional standard feature vector, i∈[1,n], and n is the total number of dimensions of the multi-dimensional standard feature vector.

[0027] To enhance the sensitivity of oil chromatography word vectors to anomalous gas changes, this invention further incorporates a residual enhancement structure during the embedding process, enabling the word vectors to simultaneously retain both original gas information and deep semantic information. This residual enhancement can be applied to the initial embedding vector using the residual enhancement function in the word spectrum mapping rules. The specific application process can be represented as follows: ; Among them, W r Let φ(•) be the residual enhancement matrix, φ(•) be the ReLU activation function, and σ(•) be the Sigmoid activation function. ti 'For superimposed residual-enhanced oil chromatography word vectors, b r These are the parameters set in the residual enhancement function. Therefore, the oil chromatography word vectors with superimposed residual enhancement are also the feature word spectrum corresponding to the preprocessed data.

[0028] The feature word spectrum is used to characterize the high-dimensional structural information of the gas composition in the oil at that time point. It not only includes concentration amplitude features, but also retains the statistical correlation and trend features between different gases and their ratios, thus forming a time series semantic unit that can be learned by the Transformer model.

[0029] In a specific embodiment, the step of extracting the corresponding multi-dimensional feature vector from the preprocessed data includes: obtaining the concentration values ​​of key feature gases at each time point; calculating the typical ratio features corresponding to the concentration values ​​at the same time point; and combining the concentration values ​​at each time point with the typical ratio features to obtain the multi-dimensional feature vector.

[0030] Specifically, the concentration values ​​corresponding to the key characteristic gases at each time point can be obtained. Further, typical ratio features corresponding to the concentration values ​​at the same time point are calculated; for example, based on the concentration values ​​corresponding to the key characteristic gases in the above steps, typical ratio features are calculated as: CH4 / H2, C2H4 / C2H6, C2H2 / C2H4, and C2H2 / CH4; then the typical ratio features include five ratio features corresponding to each time point. The five sets of typical ratio features obtained are combined with the concentration values ​​corresponding to the key characteristic gases to obtain a multi-dimensional feature vector x. t = [H2,CH4,C2H2,C2H4,C2H6,CH4 / H2,C2H4 / C2H6,C2H2 / C2H4,C2H2 / CH4] T .

[0031] In a specific embodiment, the step of combining the concentration values ​​at each time point with typical ratio features to obtain a multi-dimensional feature vector includes: combining the concentration values ​​at each time point with typical ratio features to obtain a corresponding combined feature vector; and normalizing the combined feature vector according to a preset normalization function to obtain a standard feature vector as the corresponding multi-dimensional feature vector.

[0032] Furthermore, to improve the accuracy of anomaly identification, after combining the concentration values ​​at each time point with typical ratio features to obtain a combined feature vector, the combined feature vector is normalized using a normalization function to obtain the corresponding multi-dimensional feature vector. For example, the combined feature vector X... t ={X t1 X t2 ,…X tn Z-score normalization or Min-Max normalization is used to eliminate the impact of differences in gas magnitude on model training. ; After normalization, the standard eigenvector X can be obtained. t ={ , ,… The labeled feature vector can then be used as a multi-dimensional feature vector for subsequent processing.

[0033] S130. Encode the feature word spectrum according to the preset feature word spectrum encoding model to generate the corresponding fault word spectrum vector.

[0034] Furthermore, the obtained feature word spectrum is encoded according to the feature word spectrum encoding model to generate the corresponding fault word spectrum vector. The fault word spectrum vector is used to characterize the operating state of the transformer within a continuous time period; the feature word spectrum encoding model can be set as the Transformer model.

[0035] In a specific embodiment, step S130 includes the following sub-steps: extracting a vector sequence corresponding to a continuous time period from the feature word spectrum according to the sliding time window set in the feature word spectrum coding model; adding positional encoding to the vector value of each time node in the vector sequence according to the encoding measurement in the feature word spectrum coding model to obtain the corresponding vector positioning sequence; inputting the vector positioning sequence into the encoder of the feature word spectrum coding model, and using the attention mechanism of the encoder to perform oil chromatography time pattern modeling on the encoded vector sequence to obtain the corresponding encoded vector sequence; and generating a fault word spectrum vector corresponding to the encoded vector sequence through the hidden state layer of the feature word spectrum coding model.

[0036] Specifically, a time window sequence can be constructed, such as selecting a sliding time window of length T (e.g., setting T to 60 minutes), and extracting the corresponding continuous time period vector sequence from the feature word spectrum based on the sliding time window. That is, obtaining the oil chromatography word vectors of multiple consecutive time nodes located within the sliding time window to form a vector sequence: E t ={e t-T+1 ,e t-T+2 ,… ,e t}; Furthermore, to characterize the dynamic change trend of gas content over time in oil chromatography time series, an absolute value or a corresponding position code is added to each time position k: ; e k That is, E t The element p at the time node corresponding to time position k. k If the added absolute value or positional code is used, then after adding the positional code, a vector positioning sequence corresponding to the vector sequence can be obtained.

[0037] The obtained vector localization sequence is input into the encoder of the feature word spectrum encoding model, and the attention mechanism of the encoder is used to model the oil chromatography time pattern of the encoded vector sequence: ; Then, H is the sequence of encoded vectors obtained by modeling.

[0038] Furthermore, fault word spectrum vectors corresponding to the encoded vector sequence are generated through the hidden state layer of the feature word spectrum encoding model. That is, based on the hidden state of the last layer of the Transformer model, fault word spectrum vectors for this time window are generated through pooling or sequence aggregation. ; Among them, f t The obtained fault word spectrum vector is used to represent the fault semantic features corresponding to the continuous time period and is the core input for subsequent anomaly identification and fault type discrimination.

[0039] S140. Perform anomaly identification on the fault word spectrum vector according to the preset anomaly identification model to obtain the corresponding anomaly identification result.

[0040] To accurately identify fault word spectrum vectors, a multi-layer fully connected neural network can be constructed as an anomaly detection model to identify anomalies in the obtained fault word spectrum vectors.

[0041] In a specific embodiment, step S140 includes the following sub-steps: inputting the fault word spectrum vector into the input layer of the anomaly recognition model; obtaining the node values ​​corresponding to each output node after the anomaly recognition model performs correlation analysis on the fault word spectrum vector; and obtaining the type corresponding to the output node with the largest node value as the corresponding anomaly recognition result.

[0042] The anomaly detection model includes an input layer for processing the d-dimensional fault word spectrum vector f obtained in the above steps. t The system consists of an input layer, a hidden layer (using a three-layer fully connected neural network with batch normalization and ReLU activation functions inserted between layers to enhance the network's ability to fit complex nonlinear fault modes), and an output layer (which can distinguish multiple faults simultaneously and uses a softmax activation function for output).

[0043] The structure of the hidden layer can be represented as follows: ; The softmax activation function can be expressed as: ; Where p = {p0, p1, ..., p} M}, p0 represents the normal type, and the other values ​​correspond to the abnormal types, with M being the total number of types.

[0044] The fault word spectrum vector obtained in the above steps is input into the input layer of the anomaly detection model. The number of input nodes in the input layer is equal to the number of dimensions of the fault word spectrum vector. The fault word spectrum vector is then analyzed by the hidden layer in the anomaly detection model to obtain the node value corresponding to each output node. The resulting output is p, where each type corresponds to one node value in the output.

[0045] Obtain the type corresponding to the output node with the largest node value in the output result, and use it as the anomaly identification result. If the node value corresponding to p0 in the output result is the largest, the anomaly identification result is no anomaly; if the node value corresponding to other nodes in the output result is the largest, then obtain the type corresponding to that node value as the anomaly identification result, and the anomaly identification result is that an anomaly exists.

[0046] Before using the anomaly detection model, it can be trained using a training dataset. For the training process, this application employs a loss function based on multi-class cross-entropy loss: ; Among them, y m For real tags, ω mLet ω represent the different weights for the normal class and each type of anomaly, where m is the type index, m∈[1,M]. To improve the recognition performance of a few types, ω is chosen. m as follows: ; N total N is the total number of samples in the training dataset. classes N represents the total number of types in the training dataset. m Let m be the number of samples of the m-th type.

[0047] The transformer oil chromatographic data anomaly identification method based on Transformer feature word spectrum disclosed in the above embodiments includes: receiving input historical oil chromatographic time-series data of the transformer and preprocessing it according to preprocessing rules to obtain corresponding preprocessed data; mapping the preprocessed data according to preset word spectrum mapping rules to obtain corresponding feature word spectrum; encoding the feature word spectrum according to a preset feature word spectrum encoding model to generate corresponding fault word spectrum vectors; and performing anomaly identification on the fault word spectrum vectors according to a preset anomaly identification model to obtain corresponding anomaly identification results. The above-mentioned transformer oil chromatographic data anomaly identification method based on Transformer feature word spectrum maps and encodes historical oil chromatographic time-series data of the transformer to obtain fault word spectrum vectors, and then performs anomaly identification on the fault word spectrum vectors to obtain accurate anomaly identification results, significantly improving the accuracy of transformer anomaly identification.

[0048] This invention also provides a transformer oil chromatographic data anomaly identification device based on Transformer feature spectrum. This device can be configured in a terminal device and is used to execute any embodiment of the aforementioned transformer oil chromatographic data anomaly identification method based on Transformer feature spectrum. Specifically, please refer to... Figure 2 , Figure 2 This is a schematic block diagram of a transformer oil chromatographic data anomaly identification device based on Transformer feature word spectrum provided in an embodiment of the present invention.

[0049] like Figure 2 As shown, the transformer oil chromatography data anomaly identification device 100 based on Transformer feature word spectrum includes a preprocessing unit 110, a mapping unit 120, an encoding unit 130, and an anomaly identification unit 140.

[0050] The preprocessing unit 110 is used to receive the input historical oil chromatography time series data of the transformer and preprocess it according to the preprocessing rules to obtain the corresponding preprocessed data.

[0051] The mapping unit 120 is used to map the preprocessed data according to a preset word spectrum mapping rule to obtain the corresponding feature word spectrum; the feature word spectrum includes the oil chromatogram word vector for each time node.

[0052] The encoding unit 130 is used to encode the feature word spectrum according to a preset feature word spectrum encoding model to generate a corresponding fault word spectrum vector; the fault word spectrum vector is used to characterize the operating status of the transformer within a continuous time period.

[0053] The anomaly identification unit 140 is used to identify anomalies in the fault word spectrum vector according to a preset anomaly identification model, and obtain the corresponding anomaly identification result.

[0054] The transformer oil chromatographic data anomaly identification device based on Transformer feature word spectrum provided in this embodiment of the invention applies the above-mentioned transformer oil chromatographic data anomaly identification method based on Transformer feature word spectrum. It receives input historical oil chromatographic time-series data of the transformer and preprocesses it according to preprocessing rules to obtain corresponding preprocessed data. The preprocessed data is then mapped according to preset word spectrum mapping rules to obtain corresponding feature word spectra. The feature word spectra are encoded according to a preset feature word spectrum encoding model to generate corresponding fault word spectrum vectors. Finally, anomaly identification is performed on the fault word spectrum vectors according to a preset anomaly identification model to obtain the corresponding anomaly identification result. The above-mentioned transformer oil chromatographic data anomaly identification method based on Transformer feature word spectrum maps and encodes historical oil chromatographic time-series data of the transformer to obtain fault word spectrum vectors, and then performs anomaly identification on the fault word spectrum vectors to obtain accurate anomaly identification results, significantly improving the accuracy of transformer anomaly identification.

[0055] The aforementioned transformer oil chromatographic data anomaly identification device based on Transformer feature spectrum can be implemented as a computer program, which can, for example, Figure 3 It runs on the computer device shown.

[0056] Please see Figure 3 , Figure 3 This is a schematic block diagram of a computer device provided in an embodiment of the present invention. The computer device can be a terminal device used to execute a transformer oil chromatography data anomaly identification method based on Transformer feature spectrum to analyze transformer oil chromatography data and obtain anomaly identification results.

[0057] See Figure 3The computer device 500 includes a processor 502, a memory, and a communication interface 505 connected via a communication bus 501. The memory may include a storage medium 503 and internal memory 504.

[0058] The storage medium 503 may store an operating system 5031 and a computer program 5032. When the computer program 5032 is executed, it enables the processor 502 to perform a transformer oil chromatographic data anomaly identification method based on Transformer feature spectrum. The storage medium 503 may be a volatile storage medium or a non-volatile storage medium.

[0059] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0060] The internal memory 504 provides an environment for the operation of the computer program 5032 in the storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute the transformer oil chromatographic data anomaly identification method based on Transformer feature spectrum.

[0061] This communication interface 505 is used for network communication, such as providing data transmission. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device 500 to which the present invention is applied. The specific computer device 500 may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0062] The processor 502 is used to run the computer program 5032 stored in the memory to implement the corresponding functions in the above-mentioned method for identifying anomalies in transformer oil chromatographic data based on Transformer feature word spectrum.

[0063] Those skilled in the art will understand that Figure 3 The embodiments of the computer device shown do not constitute a limitation on the specific configuration of the computer device. In other embodiments, the computer device may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. For example, in some embodiments, the computer device may include only memory and a processor. In such embodiments, the structure and function of the memory and processor are different from those shown. Figure 3 The embodiments shown are consistent and will not be described again here.

[0064] It should be understood that, in this embodiment of the invention, the processor 502 may be a Central Processing Unit (CPU), but it may also be other general-purpose processors, digital signal processors (DSPs), microcontroller units (MCUs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0065] In another embodiment of the invention, a computer-readable storage medium is provided. This computer-readable storage medium may be volatile or non-volatile. The computer-readable storage medium stores a computer program, wherein when executed by a processor, the computer program implements the steps included in the above-described method for anomaly identification of transformer oil chromatographic data based on Transformer feature spectra.

[0066] Those skilled in the art will readily understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0067] In the embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Units with the same function may be grouped into one unit. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, or it may be an electrical, mechanical, or other form of connection.

[0068] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0069] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0070] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a computer-readable storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned computer-readable storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks.

[0071] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for anomaly identification in transformer oil chromatographic data based on Transformer feature word spectrum, characterized in that, The method includes: The system receives historical oil chromatography time-series data of the transformer and preprocesses it according to preprocessing rules to obtain the corresponding preprocessed data. The preprocessed data is mapped according to a preset word spectrum mapping rule to obtain the corresponding feature word spectrum; the feature word spectrum includes the oil chromatogram word vector for each time node; The feature word spectrum is encoded according to a preset feature word spectrum encoding model to generate a corresponding fault word spectrum vector; the fault word spectrum vector is used to characterize the operating status of the transformer within a continuous time period; the feature word spectrum encoding model is the Transformer model; Anomalies are identified in the fault word spectrum vectors based on a preset anomaly identification model, and the corresponding anomaly identification results are obtained.

2. The method for anomaly identification of transformer oil chromatographic data based on Transformer feature word spectrum according to claim 1, characterized in that, The received historical oil chromatography time-series data of the transformer is preprocessed according to preprocessing rules to obtain corresponding preprocessed data, including: The historical oil chromatography time series data are excluded from outliers according to the outlier range in the preprocessing rules to obtain the corresponding valid time series data. Missing values ​​are filled into the valid time series data according to the missing value filling function in the preprocessing rules to obtain the corresponding preprocessed data.

3. The method for anomaly identification of transformer oil chromatographic data based on Transformer feature word spectrum according to claim 1, characterized in that, The step of mapping the preprocessed data according to a preset word spectrum mapping rule to obtain the corresponding feature word spectrum includes: The corresponding multi-dimensional standard feature vectors are extracted from the preprocessed data; The multi-dimensional standard feature vector is mapped to a fixed-dimensional initial word vector space based on the linear embedding matrix in the word spectrum mapping rule to obtain the initial embedding vector; The initial embedding vector is augmented with residual enhancement functions according to the word spectrum mapping rules to obtain the corresponding enhanced vector as the feature word spectrum.

4. The method for anomaly identification of transformer oil chromatographic data based on Transformer feature word spectrum according to claim 3, characterized in that, The step of extracting the corresponding multi-dimensional feature vector from the preprocessed data includes: Obtain the concentration values ​​of key characteristic gases at each time point; Calculate the typical ratio characteristics corresponding to the concentration values ​​at the same time point; The concentration values ​​at each time point are combined with typical ratio features to obtain a multi-dimensional feature vector.

5. The method for anomaly identification of transformer oil chromatographic data based on Transformer feature word spectrum according to claim 4, characterized in that, The process of combining the concentration values ​​at each time point with typical ratio features to obtain a multi-dimensional feature vector includes: The concentration values ​​at each time point are combined with typical ratio features to obtain the corresponding combined feature vector; The combined feature vectors are normalized according to a preset normalization function to obtain standard feature vectors as corresponding multi-dimensional feature vectors.

6. The method for anomaly identification of transformer oil chromatographic data based on Transformer feature word spectrum according to claim 1, characterized in that, The step of encoding the feature word spectrum according to a preset feature word spectrum encoding model to generate a corresponding fault word spectrum vector includes: According to the sliding time window set in the feature word spectrum encoding model, a vector sequence corresponding to the continuous time period of the sliding time window is extracted from the feature word spectrum; Based on the encoding measurement in the feature word spectrum encoding model, positional encoding is added to the vector value of each time node in the vector sequence to obtain the corresponding vector positioning sequence; The vector localization sequence is input into the encoder of the feature word spectrum encoding model, and the attention mechanism of the encoder is used to model the oil chromatography time pattern of the encoded vector sequence to obtain the corresponding encoded vector sequence. The fault word spectrum vector corresponding to the encoded vector sequence is generated through the hidden state layer of the feature word spectrum encoding model.

7. The method for anomaly identification of transformer oil chromatographic data based on Transformer feature word spectrum according to claim 1, characterized in that, The step of performing anomaly identification on the fault word spectrum vector according to a preset anomaly identification model to obtain the corresponding anomaly identification result includes: The fault word spectrum vector is input into the input layer of the anomaly recognition model; The anomaly identification model performs correlation analysis on the fault word spectrum vector and outputs the node values ​​corresponding to each output node. The type corresponding to the output node with the largest node value is obtained as the corresponding anomaly identification result.

8. A device for identifying anomalies in transformer oil chromatographic data based on Transformer feature word spectrum, characterized in that, The apparatus is used to perform the transformer oil chromatographic data anomaly identification method based on Transformer feature spectrum as described in any one of claims 1-7, the apparatus comprising: The preprocessing unit is used to receive the input historical oil chromatography time series data of the transformer and preprocess it according to the preprocessing rules to obtain the corresponding preprocessed data; The mapping unit is used to map the preprocessed data according to the preset word spectrum mapping rules to obtain the corresponding feature word spectrum; the feature word spectrum includes the oil chromatogram word vector for each time node; The encoding unit is used to encode the feature word spectrum according to a preset feature word spectrum encoding model to generate a corresponding fault word spectrum vector; the fault word spectrum vector is used to characterize the operating status of the transformer within a continuous time period; An anomaly identification unit is used to identify anomalies in the fault word spectrum vector according to a preset anomaly identification model, and obtain the corresponding anomaly identification result.

9. A computer device, characterized in that, The device includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When a processor executes a program stored in memory, it implements the steps of the method for identifying anomalies in transformer oil chromatographic data based on Transformer feature spectra as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for identifying anomalies in transformer oil chromatographic data based on Transformer feature word spectrum as described in any one of claims 1-7.

Images