Transformer fault identification method and system based on modal information fusion

Abstract

The invention relates to the technical field of transformer fault recognition, in particular to a transformer fault recognition method and system based on modal information fusion, and the method comprises the steps: obtaining a target transformer, carrying out the mechanical modal collection of the target transformer, obtaining a vibration signal time domain graph, carrying out the thermal modal collection of the target transformer, obtaining an infrared image, analyzing the infrared image, and obtaining a vibration signal time domain graph; acquiring a temperature difference value set, and if it is confirmed that a temperature difference value greater than a preset temperature threshold value exists in the temperature difference value set, carrying out chemical mode acquisition on the target transformer, acquiring a gas concentration text, carrying out feature extraction on the vibration signal time-domain graph, the infrared image and the gas concentration text, and carrying out feature extraction on a vibration signal feature matrix; feature vectors in the infrared image feature matrix and the gas concentration feature matrix are matched, optimal multi-modal features are determined, and transformer fault recognition based on modal information fusion is completed based on the optimal multi-modal features. According to the invention, the accuracy and reliability of transformer fault identification can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of transformer fault identification, and in particular to a transformer fault identification method, system, electronic device and computer-readable storage medium based on modal information fusion. Background Art

[0002] Modal information fusion is a technology that combines data from different physical and information sources to improve the accuracy and comprehensiveness of fault identification. A transformer is an electrical device used to transform AC voltage levels. It works on the principle of electromagnetic induction and can convert AC power of one voltage level into AC power of another voltage level without changing the frequency.

[0003] Traditional transformer fault identification methods are often based on the analysis of single-mode information, such as relying only on vibration signals, infrared images or gas concentrations. However, these methods have certain limitations in practical applications, because transformer faults often manifest as comprehensive anomalies of multiple modal information. In addition, traditional methods are mainly based on the experience of operators and maintenance personnel. These empirical rules may not be sufficient to cope with complex and changeable fault modes, resulting in a high false alarm rate. Therefore, how to improve the accuracy and reliability of transformer fault identification has become an urgent problem to be solved in the current power industry. Summary of the invention

[0004] The present invention provides a transformer fault identification method based on modal information fusion and a computer-readable storage medium, the main purpose of which is to improve the accuracy and reliability of transformer fault identification and effectively save time and human resources.

[0005] To achieve the above object, the present invention provides a transformer fault identification method based on modal information fusion, comprising:

[0006] Receive a transformer fault identification instruction, obtain a target transformer, perform mechanical modal acquisition on the target transformer according to the transformer fault identification instruction, and obtain a vibration signal time domain diagram;

[0007] Calibrate a pre-built thermal imager, and use the calibrated thermal imager to collect thermal modal data of a target transformer to obtain an infrared image, wherein a detection frequency is preset in the thermal imager;

[0008] Analyze the infrared image to obtain temperature data, visualize the temperature data according to a preset detection frequency, and obtain a temperature line graph, wherein the horizontal axis of the temperature line graph is time and the vertical axis of the temperature line graph is temperature value;

[0009] Get the temperature difference set based on the temperature line graph;

[0010] Determine whether there is a temperature difference greater than a preset temperature threshold in the temperature difference value set;

[0011] If it is confirmed that there is a temperature difference greater than a preset temperature threshold in the temperature difference set, chemical mode acquisition is performed on the target transformer to obtain the gas concentration, and the gas concentration text is obtained according to the gas concentration;

[0012] Extract features from the vibration signal time domain graph, infrared image and gas concentration text, and use pre-built matrices to represent the extracted features, thus obtaining vibration signal feature matrix, infrared image feature matrix and gas concentration feature matrix;

[0013] Matching pre-constructed feature vectors in the vibration signal feature matrix, the infrared image feature matrix, and the gas concentration feature matrix to obtain a multi-modal feature set;

[0014] Identify the optimal multimodal features based on the multimodal feature set;

[0015] Transformer fault identification based on modal information fusion is completed based on the optimal multimodal features.

[0016] Optionally, analyzing the infrared image to obtain temperature data includes:

[0017] Using a pre-built Gaussian filter to remove noise from the infrared image to obtain a denoised image, dividing the denoised image according to a preset equidistant division to obtain a plurality of division areas;

[0018] Extract one partition area from the multiple partition areas in sequence, and perform the following operations on the extracted partition area:

[0019] Pixel values ​​are extracted from the extracted divided areas to obtain the total pixel value, and the average temperature of the total pixel value is calculated using the pre-built average temperature formula, where the average temperature formula is as follows:

[0020]

[0021] Among them, T average，m represents the average temperature of the mth partition area, x mn Represents the pixel value of the nth pixel in the mth region, N m represents the total number of pixels in the mth partition area, m represents the index of the partition area, and n represents the pixel index;

[0022] The average temperature is summarized to obtain the temperature data corresponding to multiple divided areas.

[0023] Optionally, acquiring a temperature difference value set based on a temperature line graph includes:

[0024] Extract temperature values ​​from the temperature line chart in sequence, and perform the following operations on the extracted temperature values:

[0025] Based on the extracted temperature value, the current temperature value is obtained, the next temperature value adjacent to the extracted temperature value is extracted to obtain the post-temperature value, the temperature difference between the current temperature value and the post-temperature value is calculated, and the temperature difference is imported into a pre-constructed blank array to obtain a filled array, and the post-temperature value is used as the temperature value, and the step of extracting temperature values ​​from the temperature line chart in sequence is returned until it is confirmed that the temperature values ​​in the temperature line chart have been extracted, and the temperature differences in the filled array are summarized to obtain a temperature difference set.

[0026] Optionally, calculating the temperature difference between the current temperature value and the subsequent temperature value includes:

[0027] The temperature difference between the current temperature value and the post-set temperature value is calculated using the pre-built temperature difference formula, where the temperature difference formula is as follows:

[0028]

[0029] Where T represents the temperature difference, τ represents the time constant, and T i Indicates the current temperature value, T i+1 represents the post-temperature value, exp represents the exponential function with the natural constant e as the base, t i represents the i-th time point, t current Indicates the current time.

[0030] Optionally, performing chemical mode acquisition on the target transformer to obtain gas concentration, and obtaining gas concentration text according to the gas concentration includes:

[0031] Using a pre-built oil sample collector to collect oil samples from a target transformer to obtain oil sample samples, wherein a collection time is preset;

[0032] Extracting gas from the oil sample to obtain gas concentration, analyzing the gas concentration to obtain multiple gas concentrations, and storing the multiple gas concentrations in preset to-be-filled texts to obtain multiple initial texts, wherein the multiple gas concentrations include: hydrogen concentration, methane concentration, ethane concentration and ethylene concentration;

[0033] The following operation is performed on each of the multiple initial texts:

[0034] Read the target data in the initial text, and perform format conversion operation on the target data to obtain string data;

[0035] Summarize the string data to obtain a string data set, and store the string data set in a pre-constructed blank text to obtain a filled text;

[0036] Bubble sort is performed on the string data set in the fill text to obtain the sorted text, and the sorted text is confirmed as the gas concentration text.

[0037] Optionally, the feature extraction is performed on the vibration signal time domain graph, the infrared image and the gas concentration text, and the extracted features are represented by a pre-constructed matrix to obtain a vibration signal feature matrix, an infrared image feature matrix and a gas concentration feature matrix, including:

[0038] The pre-built bidirectional gated neural network is used to extract image features from the vibration signal time domain image to obtain a vibration signal feature matrix, wherein the vibration signal feature matrix represents:

[0039] G z T ={z1 T , z2 T , …, z L T}

[0040] Among them, G z represents the vibration signal feature matrix, z1 represents the first vibration signal feature vector, z2 represents the second vibration signal feature vector, and z L represents the Lth vibration signal eigenvector;

[0041] The image features of the infrared image are extracted using a bidirectional gated neural network to obtain an infrared image feature matrix, wherein the infrared image feature matrix represents:

[0042] G h T ={h1 T , h2 T ,…,h L T}

[0043] Among them, G h represents the infrared image feature matrix, h1 represents the first infrared image feature vector, h2 represents the second infrared image feature vector, and h L represents the Lth infrared image feature vector;

[0044] The bidirectional gated neural network is used to extract text features from the gas concentration text to obtain a gas concentration feature matrix, where the gas concentration feature matrix represents:

[0045] G q T ={q1 T ,q2 T , …, q L T}

[0046] Among them, Gg represents the gas concentration characteristic matrix, q1 represents the first gas concentration characteristic vector, q1 represents the second gas concentration characteristic vector, and q1 represents the Lth gas concentration characteristic vector.

[0047] Optionally, the pre-constructed feature vectors in the vibration signal feature matrix, the infrared image feature matrix and the gas concentration feature matrix are matched to obtain a multimodal feature set, including:

[0048] Each feature vector in the vibration signal feature matrix, the infrared image feature matrix and the gas concentration feature matrix is ​​paired to obtain a multimodal feature set, wherein the multimodal feature set is expressed as:

[0049] M={(z1 T , h1 T ,q1 T ),…,(z z T ,h h T ,q g T ),…,(z L T ,h L T ,q L T )}

[0050] {1 <z<L,1<h<L,1<q<L}

[0051] Among them, M represents the multimodal feature set, z z T represents the zth vibration signal eigenvector, h h T represents the hth infrared image feature vector, q q T represents the qth gas concentration eigenvector.

[0052] Optionally, the determining the optimal multimodal feature based on the multimodal feature set includes:

[0053] Extracting a target multimodal feature from the multimodal feature set, performing fault diagnosis on the target multimodal feature to obtain a first diagnosis score, extracting the next multimodal feature adjacent to the extracted target multimodal feature to obtain an adjacent multimodal feature, performing fault diagnosis on the adjacent multimodal feature to obtain a second diagnosis score;

[0054] Determine the magnitude of the first diagnostic score and the second diagnostic score;

[0055] If the first diagnostic score is greater than the second diagnostic score, the target multimodal feature corresponding to the first diagnostic score is removed from the multimodal feature set, the retained multimodal features are summarized to obtain an updated multimodal feature set, the updated multimodal feature set is the multimodal feature set, the second diagnostic score is the first diagnostic score, and the step of extracting the next multimodal feature adjacent to the extracted target multimodal feature is returned until the multimodal feature set is an empty set, and the retained target multimodal feature is confirmed as the optimal multimodal feature;

[0056] If the first diagnostic score is less than or equal to the second diagnostic score, the adjacent multimodal features corresponding to the second diagnostic score are eliminated from the multimodal feature set, the retained multimodal features are summarized to obtain the latest multimodal feature set, the latest multimodal feature set is used as the multimodal feature set, and the step of extracting the next multimodal feature adjacent to the extracted target multimodal feature is returned until the multimodal feature set is an empty set, and the retained target multimodal feature is confirmed as the optimal multimodal feature.

[0057] Optionally, the performing fault diagnosis on the target multimodal feature to obtain a first diagnosis score includes:

[0058] The pre-built multi-layer perceptron model is used to perform fault diagnosis on the multi-modal features to obtain the first diagnostic score, where the multi-layer perceptron model is as follows:

[0059] Score(z i T ,h i T ,q i T )=MLP(concat(z i T ,h i T ,q i T ))

[0060] Among them, Score(z i T ,h i T ,q i T ) represents the fault diagnosis score, MLP represents the multi-layer perceptron model, concat(z i T ,h i T ,q i T ) represents multimodal features, z i T represents the eigenvector of the i-th vibration signal, h iT represents the i-th infrared image feature vector, q i T represents the i-th gas concentration feature vector.

[0061] To achieve the above object, the present invention also provides a transformer fault identification system based on modal information fusion, comprising:

[0062] A modal data acquisition module is used to receive a transformer fault identification instruction, obtain a target transformer, perform mechanical modal acquisition on the target transformer according to the transformer fault identification instruction, obtain a vibration signal time domain diagram, calibrate a pre-built thermal imager, and use the calibrated thermal imager to perform thermal modal acquisition on the target transformer to obtain an infrared image, wherein a detection frequency is preset in the thermal imager, analyze the infrared image, obtain temperature data, visualize the temperature data according to the preset detection frequency, obtain a temperature line graph, wherein the horizontal axis of the temperature line graph is time and the vertical axis of the temperature line graph is temperature value, and obtain a temperature difference value set based on the temperature line graph;

[0063] A text generation module is used to determine whether there is a temperature difference greater than a preset temperature threshold in the temperature difference value set. If it is confirmed that there is a temperature difference greater than the preset temperature threshold in the temperature difference value set, chemical mode acquisition is performed on the target transformer to obtain the gas concentration, and the gas concentration text is obtained according to the gas concentration;

[0064] A feature extraction module is used to extract features from the vibration signal time domain diagram, infrared image and gas concentration text, and use a pre-constructed matrix to represent the extracted features, thereby obtaining a vibration signal feature matrix, an infrared image feature matrix and a gas concentration feature matrix, and matching pre-constructed feature vectors in the vibration signal feature matrix, the infrared image feature matrix and the gas concentration feature matrix to obtain a multi-modal feature set;

[0065] The fault diagnosis module is used to identify the optimal multimodal features based on the multimodal feature set, and complete transformer fault identification based on modal information fusion based on the optimal multimodal features.

[0066] In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:

[0067] A memory storing at least one instruction;

[0068] The processor executes the instructions stored in the memory to implement the transformer fault identification method based on modal information fusion as described above.

[0069] In order to solve the above problem, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned transformer fault identification method based on modal information fusion.

[0070] The present invention is to solve the problem described in the background technology. The present invention receives a transformer fault identification instruction, obtains a target transformer, performs mechanical modal acquisition on the target transformer according to the transformer fault identification instruction, and obtains a vibration signal time domain diagram. The transformer fault identification instruction of the present invention is the starting point of the process, ensuring that the system can accurately respond to fault identification requirements and lock the specific transformer that needs to be detected. Through mechanical modal acquisition, the vibration information of the transformer during operation can be obtained, the pre-built thermal imager is calibrated, and the calibrated thermal imager is used to perform thermal modal acquisition on the target transformer to obtain an infrared image, wherein the detection frequency is preset in the thermal imager. The present invention ensures the accuracy of the collected data through the calibration of the thermal imager, and the thermal modal acquisition can The temperature distribution on the transformer surface is obtained to provide strong support for fault location. The infrared image is analyzed to obtain temperature data. The temperature data is visualized according to the preset detection frequency to obtain a temperature line graph, wherein the horizontal axis of the temperature line graph is time and the vertical axis of the temperature line graph is temperature value. The present invention removes infrared image noise through a Gaussian filter to ensure image quality. At the same time, the image is divided into equal intervals to achieve refined regional processing. The average temperature formula is used to calculate the accurate average temperature of each area to ensure that the temperature data is accurate and reliable. Finally, the temperature data of each area is summarized to fully reflect the temperature distribution of the transformer, providing accurate data support for fault identification. The temperature difference set is obtained based on the temperature line graph. The present invention systematically calculates the temperature line graph. The difference between each adjacent temperature point in the temperature difference value set is used to generate a comprehensive temperature difference value set, which provides reliable data support for determining the temperature change amplitude and identifying anomalies, and determines whether there is a temperature difference greater than a preset temperature threshold in the temperature difference value set. The present invention can promptly detect the abnormal temperature rise by comparing with the preset temperature threshold. If it is confirmed that there is a temperature difference greater than the preset temperature threshold in the temperature difference value set, the target transformer is subjected to chemical mode acquisition to obtain the gas concentration, and the gas concentration text is obtained according to the gas concentration. Through systematic oil sample collection and gas concentration analysis, a standardized text containing multiple key gas concentrations is generated to ensure the accuracy and availability of the data. At the same time, the generation of standardized text is conducive to information extraction and vibration signal The time domain diagram, infrared image and gas concentration text are used for feature extraction, and the extracted features are represented by a pre-constructed matrix to obtain a vibration signal feature matrix, an infrared image feature matrix and a gas concentration feature matrix. The present invention uses a bidirectional gated neural network for feature extraction to extract key information from each modal data. The form of the feature matrix is ​​convenient for subsequent data processing and pattern recognition. The pre-constructed feature vectors in the vibration signal feature matrix, the infrared image feature matrix and the gas concentration feature matrix are matched to obtain a multi-modal feature set. The present invention realizes information fusion between different modal data through matching of feature vectors. The multi-modal feature set integrates a variety of fault indication information, and the optimal multi-modal feature is confirmed based on the multi-modal feature set.The method of the present invention determines the optimal multimodal features through an iterative comparison and elimination process, which can gradually screen out the features that contribute most to fault diagnosis, improve the accuracy and efficiency of fault diagnosis, and reduce the interference of redundant features on the diagnosis results, providing a more streamlined and effective feature set for subsequent transformer fault identification. Based on the optimal multimodal features, the transformer fault identification based on modal information fusion is completed. The present invention uses the optimal multimodal features for fault identification, which can comprehensively consider various operating status information of the transformer and realize accurate and comprehensive fault detection and diagnosis. Therefore, the present invention can improve the accuracy and reliability of transformer fault identification and effectively save time and human resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 A schematic diagram of a flow chart of a transformer fault identification method based on modal information fusion provided by an embodiment of the present invention;

[0072] Figure 2 A functional module diagram of a transformer fault identification system based on modal information fusion provided by an embodiment of the present invention;

[0073] Figure 3 A schematic diagram of the structure of an electronic device for implementing the transformer fault identification method based on modal information fusion provided by an embodiment of the present invention.

[0074] Description of reference numerals:

[0075] 1. Electronic device; 10. Processor; 11. Memory; 12. Bus.

[0076] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0077] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0078] The embodiment of the present application provides a transformer fault identification method based on modal information fusion. The execution subject of the transformer fault identification method based on modal information fusion includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the transformer fault identification method based on modal information fusion can be executed by software or hardware installed in a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0079] Reference Figure 1FIG. 1 is a flow chart of a transformer fault identification method based on modal information fusion provided by an embodiment of the present invention. In this embodiment, the transformer fault identification method based on modal information fusion includes:

[0080] S1. Receive a transformer fault identification instruction, obtain a target transformer, perform mechanical modal acquisition on the target transformer according to the transformer fault identification instruction, and obtain a vibration signal time domain diagram.

[0081] It should be explained that the transformer fault identification instruction is generally initiated by the automation equipment, and is used to start the fault detection and analysis of the specific transformer. The step of obtaining the target transformer is: quickly locate and select the target device according to the identification provided by the transformer fault identification instruction. The step of collecting the mechanical modal of the target transformer according to the transformer fault identification instruction and obtaining the vibration signal time domain diagram is: the vibration signal generated by the vibration of the internal mechanical parts of the target transformer will propagate in the air to form a sound wave, and the sound wave is captured by an acoustic sensor to obtain a vibration signal set, and each vibration signal in the collected vibration signal set is converted into an electrical signal, and digitized to obtain vibration data, and a vibration signal time domain diagram is drawn according to the vibration data. The vibration signal time domain diagram is a chart that shows the change of the vibration signal of the equipment over time.

[0082] It should be explained that the step of using an acoustic sensor to capture sound waves and obtain a vibration signal set is: using an acoustic sensor and a preset detection frequency to obtain an acoustic signal set, normalizing each acoustic signal in the acoustic signal set, and obtaining a normalized acoustic signal set;

[0083] Extract a normalized acoustic signal from the normalized acoustic signal set in turn, and perform the following operations on the extracted normalized acoustic signal:

[0084] Determining whether the normalized acoustic signal is within a preset acoustic signal interval;

[0085] If the normalized acoustic signal is not in the preset acoustic signal interval, the normalized acoustic signal is removed from the normalized acoustic signal set to obtain a standard acoustic signal set, the standard acoustic signal set is used as the normalized acoustic signal set, and the process of extracting one normalized acoustic signal from the normalized acoustic signal set is returned to the step of extracting one normalized acoustic signal from the normalized acoustic signal set in sequence;

[0086] If the normalized acoustic signal is within the preset acoustic signal interval, the normalized acoustic signal is retained to obtain an initial acoustic signal, and the process returns to the step of sequentially extracting a normalized acoustic signal from the normalized acoustic signal set;

[0087] The initial acoustic signals are aggregated to obtain an initial acoustic signal set, and the initial acoustic signal set is confirmed as a vibration signal set.

[0088] It is understandable that the acoustic sensor is a device used to collect acoustic wave signals generated by the operation of the target transformer in the environment. The acoustic signal set refers to the set of all acoustic wave signals collected by the acoustic sensor. Normalization refers to the operation of adjusting the amplitude of the acoustic signal to within the standard range. The normalized acoustic signal set refers to the set obtained after normalizing each acoustic signal in the acoustic signal set. The standard acoustic signal set refers to the set obtained after removing signals that do not conform to the preset acoustic signal interval from the normalized acoustic signal set. The initial acoustic signal refers to a normalized acoustic signal that conforms to the preset acoustic signal interval.

[0089] Exemplarily, the automation equipment sends a fault identification instruction, instructing to perform vibration abnormality detection on the high-voltage transformer numbered T-1234. The instruction contains the location information of the transformer, such as transformer No. 3 in area A of the substation. The detection method is mechanical modal acquisition, and it is required to immediately generate a time domain diagram of the vibration signal.

[0090] S2. Calibrate a pre-built thermal imager, and use the calibrated thermal imager to collect thermal modal data of a target transformer to obtain an infrared image, wherein a detection frequency is preset in the thermal imager.

[0091] It should be explained that a thermal imager refers to a device that can detect the temperature distribution on the surface of a target transformer and convert it into a visible image. The steps of calibrating the pre-built thermal imager are: adjusting the measurement parameters of the thermal imager by using a known standard temperature reference source so that it can accurately reflect the actual temperature of the object, wherein the measurement parameters include: temperature range, resolution and response time. Thermal modal acquisition refers to an operation of measuring and analyzing the temperature distribution of a target transformer through a thermal imager. An infrared image refers to an image captured and generated by a thermal imager, which shows the temperature distribution on the surface of the target transformer. Each pixel in the infrared image represents a temperature value, and different colors or color levels are usually used to represent different temperature ranges.

[0092] S3. Analyze the infrared image to obtain temperature data, visualize the temperature data according to a preset detection frequency, and obtain a temperature line graph, wherein the horizontal axis of the temperature line graph is time and the vertical axis of the temperature line graph is temperature value.

[0093] In detail, analyzing the infrared image to obtain temperature data includes:

[0094] Using a pre-built Gaussian filter to remove noise from the infrared image to obtain a denoised image, dividing the denoised image according to a preset equidistant division to obtain a plurality of division areas;

[0095] Extract one partition area from the multiple partition areas in sequence, and perform the following operations on the extracted partition area:

[0096] Pixel values ​​are extracted from the extracted divided areas to obtain the total pixel value, and the average temperature of the total pixel value is calculated using the pre-built average temperature formula, where the average temperature formula is as follows:

[0097]

[0098] Among them, T average，m represents the average temperature of the mth partition area, x mn Represents the pixel value of the nth pixel in the mth region, N m represents the total number of pixels in the mth partition area, m represents the index of the partition area, and n represents the pixel index;

[0099] The average temperature is summarized to obtain the temperature data corresponding to multiple divided areas.

[0100] It should be explained that the Gaussian filter is an image processor that uses Gaussian filtering technology to perform denoising on an image. The embodiment of the present invention uses a Gaussian filter to denoise the infrared image, and the Gaussian filtering technology is a prior art and will not be described in detail here. The denoised image refers to an image that has been processed by a Gaussian filter and has reduced the effect of noise on the infrared image. The equidistant division described in the embodiment of the present invention refers to dividing the denoised image into multiple regions of equal size according to a preset rule. For example, the preset equidistant division is 20 × 20 pixels.

[0101] It is understandable that multiple divided regions refer to a group of independent regions obtained by equally dividing the denoised image. Pixel value extraction refers to the operation of obtaining the pixel values ​​of all pixels in each divided region. The total pixel value refers to the sum of the pixel values ​​of all pixels in a certain divided region. Pixel indexing refers to the numbering of each pixel in the process of infrared image processing so that each pixel can be uniquely identified and accessed in subsequent processing.

[0102] S4. Obtain a temperature difference value set based on the temperature line graph.

[0103] In detail, the step of obtaining a temperature difference value set based on a temperature line graph includes:

[0104] Extract temperature values ​​from the temperature line chart in sequence, and perform the following operations on the extracted temperature values:

[0105] Based on the extracted temperature value, the current temperature value is obtained, the next temperature value adjacent to the extracted temperature value is extracted to obtain the post-temperature value, the temperature difference between the current temperature value and the post-temperature value is calculated, and the temperature difference is imported into a pre-constructed blank array to obtain a filled array, and the post-temperature value is used as the temperature value, and the step of extracting temperature values ​​from the temperature line chart in sequence is returned until it is confirmed that the temperature values ​​in the temperature line chart have been extracted, and the temperature differences in the filled array are summarized to obtain a temperature difference set.

[0106] It should be explained that a temperature line chart is a data chart used to visualize temperature changes over time. By connecting the temperature values ​​in the time series with a line, the trend and fluctuation of temperature changes can be intuitively displayed. The current temperature value refers to the temperature value being extracted from the temperature line chart. The subsequent temperature value refers to the next temperature value adjacent to the current temperature value.

[0107] It is understandable that the temperature difference value described in the embodiment of the present invention refers to the difference between the current temperature value and the post-set temperature value. The temperature difference value set refers to the set of temperature difference values ​​calculated between every two adjacent temperature values ​​in the temperature line graph. A blank array refers to an array without any data. A filled array refers to an array obtained by adding the temperature difference value to the blank array.

[0108] For example, the temperature values ​​from first to last in the temperature line graph are [20, 25, 30, 28], and the blank array is [None, None, None]. A temperature value 20 is extracted from the temperature line graph, and the temperature value 20 is the current temperature value. The next temperature value 25 of the temperature value 20 is extracted from the extracted temperature line graph, and the temperature value 25 is the subsequent temperature value. The temperature difference between the current temperature value 20 and the subsequent temperature value 25 is calculated, and the temperature difference is 5. The temperature difference is 5. Import the temperature difference into the blank array, and the filled array is [5, None, None]. Set the temperature value 25 as the current temperature value, continue to extract the next temperature value 30 as the post-temperature value, calculate the temperature difference between the current temperature value 25 and the post-temperature value 30, and get a temperature difference of 5. Import the temperature difference of 5 into the filling array, and get a filling array of [5, 5, None]. Set the post-temperature value 30 as the current temperature value, continue to extract the next temperature value 28 as the post-temperature value, calculate the temperature difference between the current temperature value 30 and the post-temperature value 28, and get a temperature difference of 2. Import the temperature difference of 2 into the filling array, and get a filling array of [5, 5, 2].

[0109] In detail, the calculation of the temperature difference between the current temperature value and the subsequent temperature value includes:

[0110] The temperature difference between the current temperature value and the post-set temperature value is calculated using the pre-built temperature difference formula, where the temperature difference formula is as follows:

[0111]

[0112] Where T represents the temperature difference, τ represents the time constant, and T i Indicates the current temperature value, T i+1 represents the post-temperature value, exp represents the exponential function with the natural constant e as the base, t i represents the i-th time point, t current Indicates the current time.

[0113] It should be explained that the time constant refers to a positive constant used to control the decay speed of the time weight factor. The time weight factor refers to a factor expressed by an exponential function of the base e of the natural logarithm. The current time refers to the current time value.

[0114] S5. Determine whether there is a temperature difference greater than a preset temperature threshold in the temperature difference set. If it is confirmed that there is a temperature difference greater than the preset temperature threshold in the temperature difference set, perform chemical mode acquisition on the target transformer to obtain gas concentration, and obtain gas concentration text according to the gas concentration.

[0115] In detail, the chemical mode acquisition of the target transformer is performed to obtain the gas concentration, and the gas concentration text is obtained according to the gas concentration, including:

[0116] Using a pre-built oil sample collector to collect oil samples from a target transformer to obtain oil sample samples, wherein a collection time is preset;

[0117] Extracting gas from the oil sample to obtain gas concentration, analyzing the gas concentration to obtain multiple gas concentrations, and storing the multiple gas concentrations in preset to-be-filled texts to obtain multiple initial texts, wherein the multiple gas concentrations include: hydrogen concentration, methane concentration, ethane concentration and ethylene concentration;

[0118] The following operation is performed on each of the multiple initial texts:

[0119] Read the target data in the initial text, and perform format conversion on the target data to obtain string data;

[0120] Summarize the string data to obtain a string data set, and store the string data set in a pre-constructed blank text to obtain a filled text;

[0121] Bubble sort is performed on the string data set in the fill text to obtain the sorted text, and the sorted text is confirmed as the gas concentration text.

[0122] It should be explained that the oil sample collector refers to a device for collecting transformer oil samples. The oil sample collection described in the embodiment of the present invention refers to the operation of collecting oil samples from the target transformer using the oil sample collector. The oil sample refers to the oil sample collected from the transformer by the oil sample collector. The step of extracting gas from the oil sample is: extracting gas from the oil sample using a vacuum pumping device and a preset vacuum pressure. The step of analyzing the gas concentration is: analyzing the gas concentration using a gas chromatograph. The gas chromatograph gas analysis in the present invention is a prior art and will not be repeated here.

[0123] It is understandable that the target data refers to the gas concentration data in the initial text. The initial text is the text used to store the gas concentration. The format conversion operation refers to the operation of performing a format conversion operation on the read target data to make the target data into string data. The string data set refers to a data set consisting of a series of string data. Blank text refers to text without any data. The filled text refers to the text obtained after adding string data to the blank text. Bubble sort refers to a simple sorting algorithm that repeatedly traverses the list to be sorted, compares adjacent string data in turn, and exchanges the positions of adjacent string data when necessary until the entire list is completely sorted. The sorted text refers to the text containing the sorted string data set.

[0124] S6. Extract features from the vibration signal time domain graph, infrared image and gas concentration text, and use a pre-constructed matrix to represent the extracted features, thereby obtaining a vibration signal feature matrix, an infrared image feature matrix and a gas concentration feature matrix.

[0125] In detail, the feature extraction of the vibration signal time domain graph, the infrared image and the gas concentration text is performed, and the extracted features are represented by a pre-constructed matrix to obtain a vibration signal feature matrix, an infrared image feature matrix and a gas concentration feature matrix, including:

[0126] The pre-built bidirectional gated neural network is used to extract image features from the vibration signal time domain image to obtain a vibration signal feature matrix, wherein the vibration signal feature matrix represents:

[0127] G z T ={z1 T , z2 T , …, z L T}

[0128] Among them, G z represents the vibration signal feature matrix, z1 represents the first vibration signal feature vector, z2 represents the second vibration signal feature vector, and z L represents the Lth vibration signal eigenvector;

[0129] The image features of the infrared image are extracted using a bidirectional gated neural network to obtain an infrared image feature matrix, wherein the infrared image feature matrix represents:

[0130] G h T ={h1 T , h2 T ,…,h L T}

[0131] Among them, G h represents the infrared image feature matrix, h1 represents the first infrared image feature vector, h2 represents the second infrared image feature vector, and h L represents the Lth infrared image feature vector;

[0132] The bidirectional gated neural network is used to extract text features from the gas concentration text to obtain a gas concentration feature matrix, where the gas concentration feature matrix represents:

[0133] G q T ={q1 T ,q2 T , …, q L T}

[0134] Among them, G g represents the gas concentration characteristic matrix, q1 represents the first gas concentration characteristic vector, q1 represents the second gas concentration characteristic vector, and q1 represents the Lth gas concentration characteristic vector.

[0135] It should be explained that the bidirectional gated neural network is a neural network structure based on a gated recurrent unit, which can capture contextual information in sequence data. The bidirectional gated neural network described in the embodiment of the present invention is a prior art and will not be described here. Image feature extraction refers to the operation of extracting local and global features of an image using structures such as convolutional layers and pooling layers, and then vectorizing the features through a fully connected layer to obtain the final features.

[0136] It can be understood that the vibration signal feature matrix refers to the matrix composed of feature vectors extracted from the vibration signal time domain diagram. The infrared image feature matrix refers to the matrix composed of feature vectors extracted from the infrared image. Text feature extraction refers to extracting useful feature information from the gas concentration text. The gas concentration feature matrix refers to the matrix composed of feature vectors extracted from the gas concentration text.

[0137] S7. Match the pre-constructed feature vectors in the vibration signal feature matrix, the infrared image feature matrix, and the gas concentration feature matrix to obtain a multimodal feature set.

[0138] In detail, the pre-constructed feature vectors in the vibration signal feature matrix, the infrared image feature matrix and the gas concentration feature matrix are matched to obtain a multimodal feature set, including:

[0139] Each feature vector in the vibration signal feature matrix, the infrared image feature matrix, and the gas concentration feature matrix is ​​paired to obtain a multimodal feature set, wherein the multimodal feature set is expressed as:

[0140] M={(z1 T , h1 T ,q1 T ),…,(z z T ,h h T ,q g T ),…,(z L T ,h L T ,q L T )}

[0141] {1 <z<L,1<h<L,1<q<L}

[0142] Among them, M represents the multimodal feature set, z z T represents the zth vibration signal eigenvector, h h T represents the hth infrared image feature vector, q g T represents the qth gas concentration eigenvector.

[0143] It should be understood that the step of pairing each feature vector in the vibration signal feature matrix, the infrared image feature matrix and the gas concentration feature matrix is: pairing the vibration signal feature vector, the infrared image feature vector and the gas concentration feature vector at the same time step to form a multimodal feature vector. The multimodal feature set refers to a set composed of feature vectors extracted from the vibration signal feature matrix, the infrared image feature matrix and the gas concentration feature matrix.

[0144] S8. Identify the optimal multimodal feature based on the multimodal feature set.

[0145] In detail, the determining of the optimal multimodal feature based on the multimodal feature set includes:

[0146] Extracting a target multimodal feature from the multimodal feature set, performing fault diagnosis on the target multimodal feature to obtain a first diagnosis score, extracting the next multimodal feature adjacent to the extracted target multimodal feature to obtain an adjacent multimodal feature, performing fault diagnosis on the adjacent multimodal feature to obtain a second diagnosis score;

[0147] Determine the magnitude of the first diagnostic score and the second diagnostic score;

[0148] If the first diagnostic score is greater than the second diagnostic score, the target multimodal feature corresponding to the first diagnostic score is removed from the multimodal feature set, the retained multimodal features are summarized to obtain an updated multimodal feature set, the updated multimodal feature set is the multimodal feature set, the second diagnostic score is the first diagnostic score, and the step of extracting the next multimodal feature adjacent to the extracted target multimodal feature is returned until the multimodal feature set is an empty set, and the retained target multimodal feature is confirmed as the optimal multimodal feature;

[0149] If the first diagnostic score is less than or equal to the second diagnostic score, the adjacent multimodal features corresponding to the second diagnostic score are eliminated from the multimodal feature set, the retained multimodal features are summarized to obtain the latest multimodal feature set, the latest multimodal feature set is used as the multimodal feature set, and the step of extracting the next multimodal feature adjacent to the extracted target multimodal feature is returned until the multimodal feature set is an empty set, and the retained target multimodal feature is confirmed as the optimal multimodal feature.

[0150] It should be explained that the target multimodal feature refers to a specific multimodal feature vector selected for fault diagnosis in the multimodal feature set. The adjacent multimodal feature refers to the next multimodal feature vector adjacent to the target multimodal feature in the multimodal feature set. The updated multimodal feature set refers to the set of multimodal feature vectors remaining after the target multimodal feature is removed. The latest multimodal feature set refers to the set of multimodal feature vectors remaining after the adjacent modal features are removed.

[0151] For example, the multimodal feature dataset M = {(z1 T , h1 T ,q1 T ), (z2 T , h2 T ,q2 T ), (z3 T , h3 T ,q3 T )}, select the first multimodal feature vector (z1 T , h1 T ,q1 T) is the target multimodal feature, and fault diagnosis is performed on the target multimodal feature. The first diagnosis score is 0.85, and the next multimodal feature vector (z2 T , h2 T ,q2 T ) is an adjacent multimodal feature. Fault diagnosis is performed on the adjacent multimodal features, and the second diagnosis score is 0.92. The first diagnosis score is less than the second diagnosis score. The adjacent multimodal features corresponding to the second diagnosis score are removed from the multimodal feature set, and the latest multimodal feature set is obtained as M = {(z1 T , h1 T ,q1 T ), (z3 T , h3 T ,q3 T )}, with multimodal features (z1 T , h1 T ,q1 T ) is the target modal feature, and the next multimodal feature vector adjacent to the target modal feature is extracted from the updated multimodal feature set (z3 T , h3 T ,q3 T ) is an adjacent multimodal feature. Fault diagnosis is performed on the adjacent multimodal features, and the second diagnostic score is 0.78. The first diagnostic score is greater than the second diagnostic score. Then the target multimodal feature corresponding to the first diagnostic score is removed from the multimodal feature set, and the updated multimodal feature set is M = {(z3 T , h3 T ,q3 T )}, extract the only remaining multimodal feature from the updated multimodal feature set (z3 T , h3 T ,q3 T ), the multimodal feature set is an empty set, the iteration ends, and the retained target multimodal feature (z3 T , h3 T ,q3 T ) is confirmed as the optimal multimodal feature. The ultimate goal is to extract the smallest modal feature among all multimodal data. Compared with the traditional technology of calculating all the scores and then finding the minimum data, this algorithm significantly saves computing energy.

[0152] Further, the performing fault diagnosis on the target multimodal feature to obtain a first diagnosis score includes:

[0153] The pre-built multi-layer perceptron model is used to perform fault diagnosis on the multi-modal features to obtain the first diagnostic score, where the multi-layer perceptron model is as follows:

[0154] Score(z iT ,h i T ,q i T )=MLP(concat(z i T ,h i T ,q i T ))

[0155] Among them, Score(z i T ,h i T ,q i T ) represents the first diagnostic score, MLP represents the multi-layer perceptron model, concat(z i T ,h i T ,q i T ) represents multimodal features, z i T represents the eigenvector of the i-th vibration signal, h i T represents the i-th infrared image feature vector, q i T represents the i-th gas concentration feature vector.

[0156] It should be explained that the multilayer perceptron model is a feedforward neural network that learns the feature representation of the input data and finally outputs a predicted value model.

[0157] S9. Complete transformer fault identification based on modal information fusion based on the optimal multimodal features.

[0158] It should be explained that the optimal multimodal feature described in the embodiment of the present invention refers to the feature vector with the most representativeness and diagnostic capability screened out from the multimodal feature set through fault diagnosis scoring.

[0159] Exemplarily, the multimodal feature set is M = {(z1 T , h1 T ,q1 T ), (z2 T , h2 T ,q2 T ), (z3 T , h3 T ,q3 T )}, multimodal feature A(z1 T , h1 T ,q1 T) has a fault diagnosis score of 0.85, and the multimodal feature B(z2 T , h2 T ,q2 T ) is 0.92, and the multimodal feature C(z3 T , h3 T q3 T ) has a fault diagnosis score of 0.78, so the most representative one is the multimodal feature C(z3) with the lowest score. T , h3 T ,q3 T ).

[0160] The present invention is to solve the problem described in the background technology. The present invention receives a transformer fault identification instruction, obtains a target transformer, performs mechanical modal acquisition on the target transformer according to the transformer fault identification instruction, and obtains a vibration signal time domain diagram. The transformer fault identification instruction of the present invention is the starting point of the process, ensuring that the system can accurately respond to fault identification requirements and lock the specific transformer that needs to be detected. Through mechanical modal acquisition, the vibration information of the transformer during operation can be obtained, the pre-built thermal imager is calibrated, and the calibrated thermal imager is used to perform thermal modal acquisition on the target transformer to obtain an infrared image, wherein the detection frequency is preset in the thermal imager. The present invention ensures the accuracy of the collected data through the calibration of the thermal imager, and the thermal modal acquisition can The temperature distribution on the transformer surface is obtained to provide strong support for fault location. The infrared image is analyzed to obtain temperature data. The temperature data is visualized according to the preset detection frequency to obtain a temperature line graph, wherein the horizontal axis of the temperature line graph is time and the vertical axis of the temperature line graph is temperature value. The present invention removes infrared image noise through a Gaussian filter to ensure image quality. At the same time, the image is divided into equal intervals to achieve refined regional processing. The average temperature formula is used to calculate the accurate average temperature of each area to ensure that the temperature data is accurate and reliable. Finally, the temperature data of each area is summarized to fully reflect the temperature distribution of the transformer, providing accurate data support for fault identification. The temperature difference set is obtained based on the temperature line graph. The present invention systematically calculates the temperature line graph. The difference between each adjacent temperature point in the temperature difference value set is used to generate a comprehensive temperature difference value set, which provides reliable data support for determining the temperature change amplitude and identifying anomalies, and determines whether there is a temperature difference greater than a preset temperature threshold in the temperature difference value set. The present invention can promptly detect the abnormal temperature rise by comparing with the preset temperature threshold. If it is confirmed that there is a temperature difference greater than the preset temperature threshold in the temperature difference value set, the target transformer is subjected to chemical mode acquisition to obtain the gas concentration, and the gas concentration text is obtained according to the gas concentration. Through systematic oil sample collection and gas concentration analysis, a standardized text containing multiple key gas concentrations is generated to ensure the accuracy and availability of the data. At the same time, the generation of standardized text is conducive to information extraction and vibration signal The time domain diagram, infrared image and gas concentration text are used for feature extraction, and the extracted features are represented by a pre-constructed matrix to obtain a vibration signal feature matrix, an infrared image feature matrix and a gas concentration feature matrix. The present invention uses a bidirectional gated neural network for feature extraction to extract key information from each modal data. The form of the feature matrix is ​​convenient for subsequent data processing and pattern recognition. The pre-constructed feature vectors in the vibration signal feature matrix, the infrared image feature matrix and the gas concentration feature matrix are matched to obtain a multi-modal feature set. The present invention realizes information fusion between different modal data through matching of feature vectors. The multi-modal feature set integrates a variety of fault indication information, and the optimal multi-modal feature is confirmed based on the multi-modal feature set.The method of the present invention determines the optimal multimodal features through an iterative comparison and elimination process, which can gradually screen out the features that contribute most to fault diagnosis, improve the accuracy and efficiency of fault diagnosis, and reduce the interference of redundant features on the diagnosis results, providing a more streamlined and effective feature set for subsequent transformer fault identification. Based on the optimal multimodal features, the transformer fault identification based on modal information fusion is completed. The present invention uses the optimal multimodal features for fault identification, which can comprehensively consider various operating status information of the transformer and realize accurate and comprehensive fault detection and diagnosis. Therefore, the present invention can improve the accuracy and reliability of transformer fault identification and effectively save time and human resources.

[0161] like Figure 2 , which is a functional module diagram of a transformer fault identification system based on modal information fusion provided by an embodiment of the present invention.

[0162] The transformer fault identification system 100 based on modal information fusion of the present invention can be installed in an electronic device. According to the functions to be implemented, the transformer fault identification system 100 based on modal information fusion can include a modal data acquisition module 101, a text generation module 102, a feature extraction module 103 and a fault diagnosis module 104. The module of the present invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device;

[0163] The modal data acquisition module 101 is used to receive a transformer fault identification instruction, obtain a target transformer, perform mechanical modal acquisition on the target transformer according to the transformer fault identification instruction, obtain a vibration signal time domain diagram, calibrate a pre-built thermal imager, and use the calibrated thermal imager to perform thermal modal acquisition on the target transformer to obtain an infrared image, wherein a detection frequency is preset in the thermal imager, analyze the infrared image, obtain temperature data, visualize the temperature data according to the preset detection frequency, obtain a temperature line graph, wherein the horizontal axis of the temperature line graph is time and the vertical axis of the temperature line graph is temperature value, and obtain a temperature difference value set based on the temperature line graph;

[0164] The text generation module 102 is used to determine whether there is a temperature difference greater than a preset temperature threshold in the temperature difference value set, and if it is confirmed that there is a temperature difference greater than the preset temperature threshold in the temperature difference value set, chemical mode acquisition is performed on the target transformer to obtain the gas concentration, and the gas concentration text is obtained according to the gas concentration;

[0165] The feature extraction module 103 is used to extract features from the vibration signal time domain diagram, the infrared image and the gas concentration text, and use a pre-constructed matrix to represent the extracted features, thereby obtaining a vibration signal feature matrix, an infrared image feature matrix and a gas concentration feature matrix, and matching pre-constructed feature vectors in the vibration signal feature matrix, the infrared image feature matrix and the gas concentration feature matrix to obtain a multimodal feature set;

[0166] The fault diagnosis module 104 is used to identify the optimal multimodal feature based on the multimodal feature set, and complete transformer fault identification based on modal information fusion based on the optimal multimodal feature.

[0167] In detail, each module in the transformer fault identification system 100 based on modal information fusion in the embodiment of the present invention is used in the same manner as above. Figure 1 The transformer fault identification method based on modal information fusion described in the present invention has the same technical means and can produce the same technical effects, so it will not be repeated here.

[0168] like Figure 3 , is a schematic diagram of the structure of an electronic device for implementing a transformer fault identification method based on modal information fusion provided by an embodiment of the present invention.

[0169] The electronic device 1 may include a processor 10, a memory 11 and a bus 12, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as a transformer fault identification method program based on modal information fusion.

[0170] The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. The memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, such as a mobile hard disk of the electronic device 1. The memory 11 may also be an external storage device of the electronic device 1 in other embodiments, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 11 also includes an internal storage unit of the electronic device 1 and an external storage device. The memory 11 can not only be used to store application software and various types of data installed in the electronic device 1, such as the code of the transformer fault identification method program based on modal information fusion, but also can be used to temporarily store data that has been output or is to be output.

[0171] The processor 10 may be composed of an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips, etc. The processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect various components of the entire electronic device, and executes or executes programs or modules stored in the memory 11 (such as a transformer fault identification method program based on modal information fusion, etc.), and calls data stored in the memory 11 to execute various functions of the electronic device 1 and process data.

[0172] The bus 12 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus 12 may be divided into an address bus, a data bus, a control bus, etc. The bus 12 is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.

[0173] Figure 3 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 3 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0174] For example, although not shown, the electronic device 1 may also include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include any components such as one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, etc. The electronic device 1 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0175] Furthermore, the electronic device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0176] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device. The display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

[0177] The transformer fault identification method program based on modal information fusion stored in the memory 11 of the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve:

[0178] Receive a transformer fault identification instruction, obtain a target transformer, perform mechanical modal acquisition on the target transformer according to the transformer fault identification instruction, and obtain a vibration signal time domain diagram;

[0179] Calibrate a pre-built thermal imager, and use the calibrated thermal imager to collect thermal modal data of a target transformer to obtain an infrared image, wherein a detection frequency is preset in the thermal imager;

[0180] Analyze the infrared image to obtain temperature data, visualize the temperature data according to a preset detection frequency, and obtain a temperature line graph, wherein the horizontal axis of the temperature line graph is time and the vertical axis of the temperature line graph is temperature value;

[0181] Get the temperature difference set based on the temperature line graph;

[0182] Determine whether there is a temperature difference greater than a preset temperature threshold in the temperature difference value set;

[0183] If it is confirmed that there is a temperature difference greater than a preset temperature threshold in the temperature difference set, chemical mode acquisition is performed on the target transformer to obtain the gas concentration, and the gas concentration text is obtained according to the gas concentration;

[0184] Extract features from the vibration signal time domain graph, infrared image and gas concentration text, and use pre-built matrices to represent the extracted features, thus obtaining vibration signal feature matrix, infrared image feature matrix and gas concentration feature matrix;

[0185] Matching pre-constructed feature vectors in the vibration signal feature matrix, the infrared image feature matrix, and the gas concentration feature matrix to obtain a multi-modal feature set;

[0186] Identify the optimal multimodal features based on the multimodal feature set;

[0187] Transformer fault identification based on modal information fusion is completed based on the optimal multimodal features.

[0188] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to Figures 1 to 3 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0189] Furthermore, if the module / unit integrated in the electronic device 1 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. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0190] The present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor of an electronic device, the computer program can implement:

[0191] Receive a transformer fault identification instruction, obtain a target transformer, perform mechanical modal acquisition on the target transformer according to the transformer fault identification instruction, and obtain a vibration signal time domain diagram;

[0192] Calibrate a pre-built thermal imager, and use the calibrated thermal imager to collect thermal modal data of a target transformer to obtain an infrared image, wherein a detection frequency is preset in the thermal imager;

[0193] Analyze the infrared image to obtain temperature data, visualize the temperature data according to a preset detection frequency, and obtain a temperature line graph, wherein the horizontal axis of the temperature line graph is time and the vertical axis of the temperature line graph is temperature value;

[0194] Get the temperature difference set based on the temperature line graph;

[0195] Determine whether there is a temperature difference greater than a preset temperature threshold in the temperature difference value set;

[0196] If it is confirmed that there is a temperature difference greater than a preset temperature threshold in the temperature difference set, chemical mode acquisition is performed on the target transformer to obtain the gas concentration, and the gas concentration text is obtained according to the gas concentration;

[0197] Extract features from the vibration signal time domain graph, infrared image and gas concentration text, and use pre-built matrices to represent the extracted features, thus obtaining vibration signal feature matrix, infrared image feature matrix and gas concentration feature matrix;

[0198] Matching pre-constructed feature vectors in the vibration signal feature matrix, the infrared image feature matrix, and the gas concentration feature matrix to obtain a multi-modal feature set;

[0199] Identify the optimal multimodal features based on the multimodal feature set;

[0200] Transformer fault identification based on modal information fusion is completed based on the optimal multimodal features.

[0201] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, systems and methods can be implemented in other ways. For example, the system embodiments described above are only illustrative, and actual implementation may have other division methods.

[0202] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0203] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0204] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0205] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A transformer fault identification method based on modal information fusion, characterized in that: The method comprises: Receive a transformer fault identification instruction, obtain a target transformer, perform mechanical modal acquisition on the target transformer according to the transformer fault identification instruction, and obtain a vibration signal time domain diagram; Calibrate a pre-built thermal imager, and use the calibrated thermal imager to collect thermal modal data of a target transformer to obtain an infrared image, wherein a detection frequency is preset in the thermal imager; Analyze the infrared image to obtain temperature data, visualize the temperature data according to a preset detection frequency, and obtain a temperature line graph, wherein the horizontal axis of the temperature line graph is time and the vertical axis of the temperature line graph is temperature value; Get the temperature difference set based on the temperature line graph; Determine whether there is a temperature difference greater than a preset temperature threshold in the temperature difference value set; If it is confirmed that there is a temperature difference greater than a preset temperature threshold in the temperature difference set, chemical mode acquisition is performed on the target transformer to obtain the gas concentration, and the gas concentration text is obtained according to the gas concentration; Extract features from the vibration signal time domain graph, infrared image and gas concentration text, and use pre-built matrices to represent the extracted features, thus obtaining vibration signal feature matrix, infrared image feature matrix and gas concentration feature matrix; Matching pre-constructed feature vectors in the vibration signal feature matrix, the infrared image feature matrix, and the gas concentration feature matrix to obtain a multi-modal feature set; Identify the optimal multimodal features based on the multimodal feature set; Transformer fault identification based on modal information fusion is completed based on the optimal multimodal features.

2. The transformer fault identification method based on modal information fusion according to claim 1 is characterized in that: The step of analyzing the infrared image to obtain temperature data includes: Using a pre-built Gaussian filter to remove noise from the infrared image to obtain a denoised image, dividing the denoised image according to a preset equidistant division to obtain a plurality of division areas; Extract one partition area from the multiple partition areas in sequence, and perform the following operations on the extracted partition area: Pixel values ​​are extracted from the extracted divided areas to obtain the total pixel value, and the average temperature of the total pixel value is calculated using the pre-built average temperature formula, where the average temperature formula is as follows: Among them, T average,m represents the average temperature of the mth partition area, x mn Represents the pixel value of the nth pixel in the mth region, N m represents the total number of pixels in the mth partition area, m represents the index of the partition area, and n represents the pixel index; The average temperature is summarized to obtain the temperature data corresponding to multiple divided areas.

3. The transformer fault identification method based on modal information fusion according to claim 2 is characterized in that: The step of obtaining a temperature difference value set based on a temperature line graph includes: Extract temperature values ​​from the temperature line chart in sequence, and perform the following operations on the extracted temperature values: Based on the extracted temperature value, the current temperature value is obtained, the next temperature value adjacent to the extracted temperature value is extracted to obtain the post-temperature value, the temperature difference between the current temperature value and the post-temperature value is calculated, and the temperature difference is imported into a pre-constructed blank array to obtain a filled array, and the post-temperature value is used as the temperature value, and the step of extracting temperature values ​​from the temperature line chart in sequence is returned until it is confirmed that the temperature values ​​in the temperature line chart have been extracted, and the temperature differences in the filled array are summarized to obtain a temperature difference set.

4. The transformer fault identification method based on modal information fusion according to claim 3 is characterized in that: The step of calculating the temperature difference between the current temperature value and the subsequent temperature value includes: The temperature difference between the current temperature value and the post-set temperature value is calculated using the pre-built temperature difference formula, where the temperature difference formula is as follows: Where T represents the temperature difference, τ represents the time constant, and T i Indicates the current temperature value, T i+1 represents the post-temperature value, exp represents the exponential function with the natural constant e as the base, t i represents the i-th time point, t current Indicates the current time.

5. The transformer fault identification method based on modal information fusion according to claim 4 is characterized in that: The chemical mode acquisition of the target transformer is performed to obtain the gas concentration, and the gas concentration text is obtained according to the gas concentration, including: Using a pre-built oil sample collector to collect oil samples from a target transformer to obtain oil sample samples, wherein a collection time is preset; Extracting gas from the oil sample to obtain gas concentration, analyzing the gas concentration to obtain multiple gas concentrations, and storing the multiple gas concentrations in preset to-be-filled texts to obtain multiple initial texts, wherein the multiple gas concentrations include: hydrogen concentration, methane concentration, ethane concentration and ethylene concentration; The following operation is performed on each of the multiple initial texts: Read the target data in the initial text, and perform format conversion operation on the target data to obtain string data; Summarize the string data to obtain a string data set, and store the string data set in a pre-constructed blank text to obtain a filled text; Bubble sort is performed on the string data set in the fill text to obtain the sorted text, and the sorted text is confirmed as the gas concentration text.

6. The transformer fault identification method based on modal information fusion according to claim 5 is characterized in that: The feature extraction is performed on the vibration signal time domain diagram, the infrared image and the gas concentration text, and the extracted features are represented by using a pre-constructed matrix to obtain a vibration signal feature matrix, an infrared image feature matrix and a gas concentration feature matrix, including: The pre-built bidirectional gated neural network is used to extract image features from the vibration signal time domain image to obtain a vibration signal feature matrix, wherein the vibration signal feature matrix represents: G z T ={z1 T ,z2 T ,…,z L T } Among them, G z represents the vibration signal feature matrix, z1 represents the first vibration signal feature vector, z2 represents the second vibration signal feature vector, and z L represents the Lth vibration signal eigenvector; The image features of the infrared image are extracted using a bidirectional gated neural network to obtain an infrared image feature matrix, wherein the infrared image feature matrix represents: G h T ={h1 T ,h2 T ,…,h L T } Among them, G h represents the infrared image feature matrix, h1 represents the first infrared image feature vector, h2 represents the second infrared image feature vector, and h L represents the Lth infrared image feature vector; The bidirectional gated neural network is used to extract text features from the gas concentration text to obtain a gas concentration feature matrix, where the gas concentration feature matrix represents: G q T ={q1 T ,q2 T ,…,q L T } Among them, G q represents the gas concentration characteristic matrix, q1 represents the first gas concentration characteristic vector, q1 represents the second gas concentration characteristic vector, and q1 represents the Lth gas concentration characteristic vector.

7. The transformer fault identification method based on modal information fusion according to claim 6 is characterized in that: The pre-constructed feature vectors in the vibration signal feature matrix, the infrared image feature matrix and the gas concentration feature matrix are matched to obtain a multi-modal feature set, including: Each feature vector in the vibration signal feature matrix, the infrared image feature matrix, and the gas concentration feature matrix is ​​paired to obtain a multimodal feature set, wherein the multimodal feature set is expressed as: M={(z1 T ,h1 T ,q1 T ),…,(z z T ,h h T ,q q T ),…,(z L T ,h L T ,q L T )} {1 <z<L,1<h<L,1<q<L} Among them, M represents the multimodal feature set, z z T represents the zth vibration signal eigenvector, h h T represents the hth infrared image feature vector, q q T represents the qth gas concentration eigenvector.

8. The transformer fault identification method based on modal information fusion according to claim 7 is characterized in that: The determining of the optimal multimodal feature based on the multimodal feature set includes: Extracting a target multimodal feature from the multimodal feature set, performing fault diagnosis on the target multimodal feature to obtain a first diagnosis score, extracting the next multimodal feature adjacent to the extracted target multimodal feature to obtain an adjacent multimodal feature, performing fault diagnosis on the adjacent multimodal feature to obtain a second diagnosis score; Determine the magnitude of the first diagnostic score and the second diagnostic score; If the first diagnostic score is greater than the second diagnostic score, the target multimodal feature corresponding to the first diagnostic score is removed from the multimodal feature set, the retained multimodal features are summarized to obtain an updated multimodal feature set, the updated multimodal feature set is the multimodal feature set, the second diagnostic score is the first diagnostic score, and the step of extracting the next multimodal feature adjacent to the extracted target multimodal feature is returned until the multimodal feature set is an empty set, and the retained target multimodal feature is confirmed as the optimal multimodal feature; If the first diagnostic score is less than or equal to the second diagnostic score, the adjacent multimodal features corresponding to the second diagnostic score are eliminated from the multimodal feature set, the retained multimodal features are summarized to obtain the latest multimodal feature set, the latest multimodal feature set is used as the multimodal feature set, and the step of extracting the next multimodal feature adjacent to the extracted target multimodal feature is returned until the multimodal feature set is an empty set, and the retained target multimodal feature is confirmed as the optimal multimodal feature.

9. The transformer fault identification method based on modal information fusion according to claim 8, characterized in that: The performing fault diagnosis on the target multimodal feature to obtain a first diagnosis score includes: The pre-built multi-layer perceptron model is used to perform fault diagnosis on the multi-modal features to obtain a first diagnostic score, where the multi-layer perceptron model is as follows: Score(z i T ,h i T ,q i T )=MLP(concat(z i T ,h i T ,q i T )) Among them, Score(z i T ,h i T ,q i T ) represents the fault diagnosis score, MLP represents the multi-layer perceptron model, concat(z i T ,h i T ,q i T ) represents multimodal features, z i T represents the eigenvector of the i-th vibration signal, h i T represents the i-th infrared image feature vector, q i T represents the i-th gas concentration feature vector.

10. A transformer fault identification system based on modal information fusion, characterized in that: The system comprises: A modal data acquisition module is used to receive a transformer fault identification instruction, obtain a target transformer, perform mechanical modal acquisition on the target transformer according to the transformer fault identification instruction, obtain a vibration signal time domain diagram, calibrate a pre-built thermal imager, and use the calibrated thermal imager to perform thermal modal acquisition on the target transformer to obtain an infrared image, wherein a detection frequency is preset in the thermal imager, analyze the infrared image, obtain temperature data, visualize the temperature data according to the preset detection frequency, obtain a temperature line graph, wherein the horizontal axis of the temperature line graph is time and the vertical axis of the temperature line graph is temperature value, and obtain a temperature difference value set based on the temperature line graph; A text generation module is used to determine whether there is a temperature difference greater than a preset temperature threshold in the temperature difference value set. If it is confirmed that there is a temperature difference greater than the preset temperature threshold in the temperature difference value set, chemical mode acquisition is performed on the target transformer to obtain the gas concentration, and the gas concentration text is obtained according to the gas concentration; A feature extraction module is used to extract features from the vibration signal time domain diagram, infrared image and gas concentration text, and use a pre-constructed matrix to represent the extracted features, thereby obtaining a vibration signal feature matrix, an infrared image feature matrix and a gas concentration feature matrix, and matching pre-constructed feature vectors in the vibration signal feature matrix, the infrared image feature matrix and the gas concentration feature matrix to obtain a multi-modal feature set; The fault diagnosis module is used to identify the optimal multimodal features based on the multimodal feature set, and complete transformer fault identification based on modal information fusion based on the optimal multimodal features.

Images