Transformer fault diagnosis method, device, equipment and medium

Abstract

The invention relates to the field of power equipment fault prediction and health management, in particular to a transformer fault diagnosis method and device, equipment and a medium, and the method comprises the steps: verifying the effectiveness of each gas sensor corresponding to online monitoring gas when the transformer fault diagnosis is triggered; for any gas sensor, if the verification of the gas sensor is not passed, triggering an online calibration process to obtain a calibrated gas sensor; according to the calibrated gas sensor, gas concentration data corresponding to the online monitoring gas and measurement uncertainty corresponding to the gas concentration data are obtained; performing fault diagnosis on the transformer based on each gas concentration data and the measurement uncertainty corresponding to each gas concentration data to obtain a diagnosis result and a confidence level of the diagnosis result; and a diagnosis report is output according to the diagnosis result and the confidence level of the diagnosis result, and the credibility of transformer fault diagnosis is effectively improved by establishing quantitative correlation between data quality and a diagnosis conclusion.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power equipment fault prediction and health management, and in particular to a transformer fault diagnosis method, device, equipment and medium. BACKGROUND

[0002] Based on the analysis of the dissolved gas in the insulating oil is one of the most effective means to diagnose the internal latent fault of the transformer (such as partial discharge, overheating, arc, etc.). At present, the automatic diagnosis system based on online monitoring data has been widely used, but the reliability of the diagnosis conclusion is seriously dependent on the accuracy of the monitoring data. The inevitable drift of the online monitoring sensor will lead to the distortion of the gas concentration and the calculation of the ratio, and then may cause misdiagnosis (such as misjudging the normal state as a fault) or missed diagnosis (such as failing to discover a serious fault in time).

[0003] In the prior art, fault diagnosis and sensor calibration are two relatively independent links. The diagnosis system usually defaults that the received data is accurate, and the calibration operation is manually triggered by the operation and maintenance personnel periodically or when data anomalies are found. This disconnection brings two main problems:

[0004] 1. The effectiveness of the data before diagnosis is unknown: the diagnosis system cannot know whether the data used at present is deviated due to the sensor drift when analyzing, so that the diagnosis conclusion is established on the basis of unreliable data.

[0005] 2. The diagnosis conclusion lacks confidence evaluation: the diagnosis report usually only gives the conclusion of the fault type and severity, but does not attach the uncertainty information of the conclusion. The operation and maintenance personnel cannot judge whether the diagnosis conclusion is certain or there is a risk of large error, which affects the scientificity of the decision.

[0006] Therefore, a method capable of integrating data calibration and fault diagnosis is needed to ensure the reliability of the diagnosis conclusion and provide a clear confidence reference for the operation and maintenance decision. SUMMARY

[0007] The present application aims to provide a transformer fault diagnosis method, device, equipment and medium, which solves the problems in the prior art.

[0008] The present application is realized by the following technical solutions:

[0009] In the first aspect, the present application embodiment provides a transformer fault diagnosis method, comprising:

[0010] When triggering the transformer fault diagnosis, the effectiveness of each gas sensor corresponding to the online monitoring gas is verified;

[0011] For any gas sensor, if the verification of the gas sensor fails, triggering an online calibration process to obtain a calibrated gas sensor;

[0012] According to the calibrated gas sensor, obtaining gas concentration data corresponding to the online monitored gas and measurement uncertainty corresponding to the gas concentration data;

[0013] Based on each gas concentration data and the measurement uncertainty corresponding to each gas concentration data, performing fault diagnosis on the transformer to obtain a diagnosis result and a confidence level of the diagnosis result;

[0014] According to the diagnosis result and the confidence level of the diagnosis result, outputting a diagnosis report, wherein the diagnosis report at least includes a fault diagnosis conclusion, a confidence level of the diagnosis conclusion, and a state identifier indicating whether the data is calibrated.

[0015] Preferably, when triggering the transformer fault diagnosis, the effectiveness of each gas sensor corresponding to the online monitored gas is verified, including:

[0016] In response to a triggering instruction of the transformer fault diagnosis, obtaining a historical calibration record and a remaining uncertainty of each gas sensor;

[0017] For each gas sensor, determining a calibration duration of the gas sensor according to an interval between a latest calibration time in the historical calibration record of the gas sensor and a current time;

[0018] If the calibration duration is greater than or equal to a preset calibration duration threshold, the verification of the gas sensor fails;

[0019] If the calibration duration is less than the preset calibration duration threshold, comparing the remaining uncertainty with a preset remaining uncertainty threshold;

[0020] If the remaining uncertainty is greater than or equal to the remaining uncertainty threshold, the verification of the gas sensor fails.

[0021] Preferably, when triggering the transformer fault diagnosis, the effectiveness of each gas sensor corresponding to the online monitored gas is verified, including:

[0022] According to a drift prediction model of each gas sensor, determining a predicted drift amount of each gas sensor;

[0023] According to the predicted drift amount of each gas sensor and a preset tolerance, determining the effectiveness of the data of the online monitored gas corresponding to each gas sensor.

[0024] Preferably, for any gas sensor, if the verification of the gas sensor fails, an online calibration process is triggered to obtain a calibrated gas sensor, including:

[0025] Send a command to the online calibration device to control the online calibration device to perform calibration tests on the gas sensor using standard gas, and obtain the response output of the gas sensor at multiple standard concentration points;

[0026] Based on the correspondence between each response output and the standard concentration point, an updated calibration curve for the gas sensor is generated;

[0027] Based on the uncertainty of the updated calibration curve and the standard gas, the remaining uncertainty of the gas sensor is calculated and updated.

[0028] Preferably, the step of calculating and updating the residual uncertainty of the gas sensor based on the uncertainty of the updated calibration curve and the standard gas includes:

[0029] Based on the fitting residuals of the updated calibration curve, the standard uncertainty of the fit, which characterizes the dispersion of the gas sensor response, is calculated.

[0030] The combined standard uncertainty of the gas sensor is calculated by combining the baseline standard uncertainty of the standard gas with the fitted standard uncertainty.

[0031] The combined standard uncertainty is multiplied by a preset inclusion factor to update the residual uncertainty of the gas sensor.

[0032] Preferably, obtaining the gas concentration data of the corresponding online monitored gas and the measurement uncertainty corresponding to the gas concentration data based on the calibrated gas sensor includes:

[0033] The gas concentration data is obtained by measuring the corresponding online monitoring gas using the calibrated gas sensor.

[0034] The measurement uncertainty of the gas concentration data is calculated based on the fitting residual of the updated calibration curve of the gas sensor and the coverage of the calibration points.

[0035] Preferably, the step of performing fault diagnosis on the transformer based on each gas concentration data and the measurement uncertainty corresponding to each gas concentration data, to obtain the diagnosis result and the confidence level of the diagnosis result, includes:

[0036] Based on the gas concentration data and the corresponding measurement uncertainty, a diagnostic input dataset with uncertainty is obtained;

[0037] Based on the gas concentration data in the diagnostic input dataset, a preliminary fault diagnosis is performed to obtain the diagnostic results;

[0038] Based on the diagnostic results, the diagnostic input dataset, and the pre-trained fault diagnosis model, uncertainty propagation analysis is performed to obtain the probability distribution of the diagnostic results.

[0039] Based on the probability distribution, the confidence level is calculated to obtain the confidence level of the diagnostic result.

[0040] Secondly, embodiments of the present invention provide a transformer fault diagnosis device, comprising:

[0041] The verification module is used to verify the effectiveness of each gas sensor corresponding to the online monitoring gas when the transformer fault diagnosis is triggered.

[0042] The verification module is used to trigger an online calibration process to obtain a calibrated gas sensor if the verification of any gas sensor fails.

[0043] The measurement module is used to obtain the gas concentration data of the corresponding online monitored gas and the measurement uncertainty corresponding to the gas concentration data based on the calibrated gas sensor.

[0044] The diagnostic module is used to perform fault diagnosis on the transformer based on each gas concentration data and the measurement uncertainty corresponding to each gas concentration data, and to obtain the diagnostic results and the confidence level of the diagnostic results;

[0045] The output report module is used to output a diagnostic report based on the diagnostic results and the confidence level of the diagnostic results. The diagnostic report includes at least the fault diagnosis conclusion, the confidence level of the diagnostic conclusion, and a status indicator indicating whether the data has been calibrated.

[0046] Thirdly, embodiments of the present invention provide an electronic device, including: at least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the method of the first aspect described above.

[0047] Fourthly, embodiments of the present invention provide a storage medium storing computer program instructions, which, when executed by a processor, implement the method of the first aspect described above.

[0048] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0049] This application effectively improves the reliability of transformer fault diagnosis by establishing a quantitative correlation between data quality and diagnostic conclusions. It embeds a systematic data validity verification mechanism into the diagnostic process, automatically triggering calibration when sensor data is unreliable, thus ensuring the accuracy of basic data from the source. By inputting the measurement uncertainty obtained after calibration as a key parameter into the fault diagnosis model, it achieves a probabilistic assessment of the reliability of diagnostic conclusions, enabling maintenance personnel to distinguish between high-confidence and low-confidence diagnostic results. The final diagnostic report, which includes data status identifiers and confidence levels, provides a more complete basis for differentiated maintenance decisions, thereby enhancing the overall practicality and decision support capabilities of the fault diagnosis system. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0051] Figure 1 A flowchart illustrating the transformer fault diagnosis method provided in this application;

[0052] Figure 2 This is an example schematic diagram of the trigger transformer fault diagnosis provided in this application;

[0053] Figure 3 This is a schematic diagram of the structure of the transformer fault diagnosis device provided in this application;

[0054] Figure 4 A schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0056] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0057] It should be noted that all actions involving the acquisition of signals, information, or data in this invention are carried out in compliance with the relevant data protection laws and regulations of the locality and with authorization from the owner of the relevant device.

[0058] Example 1

[0059] Please see Figure 1 This invention provides a transformer fault diagnosis method, including:

[0060] S1. When triggering transformer fault diagnosis, verify the effectiveness of each gas sensor corresponding to the online monitoring gas;

[0061] Specifically, when the transformer fault diagnosis process is triggered, the system first verifies the validity of each gas sensor corresponding to the online monitored gas. This step assesses whether the current state of each sensor meets the data reliability requirements for diagnosis by querying its historical calibration records and combining them with preset judgment conditions. For example, the system checks whether the time interval between the sensor's last calibration and the last calibration exceeds the allowable range, or whether the deviation predicted based on the sensor drift model exceeds the tolerance threshold. If the sensor is determined to be invalid, it means that its output data may contain unacceptable errors and cannot be directly used for subsequent diagnostic analysis. This step ensures the quality of the input data from the source, laying the foundation for the reliability of the entire diagnostic process.

[0062] S2. For any gas sensor, if the verification of the gas sensor fails, an online calibration process is triggered to obtain the calibrated gas sensor.

[0063] Specifically, for gas sensors that fail verification, the system automatically triggers an online calibration process. During calibration, a standard gas is introduced into the sensor, causing it to respond at multiple known concentration points. Subsequently, based on the correspondence between the response signal and the standard concentration, a new calibration function is generated using curve fitting. This function is used to correct the sensor's original output, restoring its accurate measurement characteristics. Simultaneously, the calibration process reassesses the sensor's residual uncertainty, an indicator reflecting the overall error range remaining after calibration, synthesized from the uncertainty of the standard gas itself and the residual of the calibration curve fitting. Through this step, the sensor not only obtains an accurate measurement benchmark, but its data uncertainty is also quantified and updated, providing a traceable and reliable data source for subsequent diagnostics.

[0064] S3. Based on the calibrated gas sensor, obtain the gas concentration data of the corresponding online monitored gas and the measurement uncertainty corresponding to the gas concentration data;

[0065] Specifically, based on the calibrated gas sensors, the system performs reliable data acquisition and uncertainty quantification. Specifically, the raw signals acquired by the sensors in real time are input into a newly established calibration function to calculate precise gas concentration values. Simultaneously, combined with the updated residual uncertainty after this calibration, and considering the noise effects in real-time measurements, the measurement uncertainty corresponding to the gas concentration data is calculated using an error propagation model. The final output is the concentration value of each gas and its corresponding uncertainty value; for example, an acetylene concentration of 5.8 μL / L has an uncertainty of ±0.3 μL / L. This set of data comprehensively characterizes the measurement results and their reliability range, providing core input for subsequent probabilistic fault diagnosis.

[0066] S4. Based on the gas concentration data and the measurement uncertainty corresponding to each gas concentration data, perform fault diagnosis on the transformer to obtain the diagnosis result and the confidence level of the diagnosis result;

[0067] Specifically, the gas concentration data and their corresponding measurement uncertainties are input into the fault diagnosis model. During the execution of the diagnostic logic, the uncertainty propagation method is used to analyze the impact of the input uncertainty on the final conclusion. For example, when using the ratio method for diagnosis, the fluctuation range of the gas ratio due to the input uncertainty is calculated using error propagation theory; if a machine learning model is used, the distribution of diagnostic results under input uncertainty can be simulated through probability sampling. Based on this analysis, the system not only outputs diagnostic conclusions such as arc discharge or overheating faults, but also calculates the confidence level of the conclusion, such as 88%. This step transforms the traditional yes-or-no judgment into a probabilistic evaluation, giving the diagnostic conclusion quantifiable credibility information.

[0068] S5. Output a diagnostic report based on the diagnostic results and the confidence level of the diagnostic results. The diagnostic report shall include at least the fault diagnosis conclusion, the confidence level of the diagnostic conclusion, and a status indicator indicating whether the data has been calibrated.

[0069] Specifically, a diagnostic report is generated by combining the diagnostic results with the confidence level. The report must include at least a clear fault diagnosis conclusion, a confidence level expressed as a percentage or grade, and a data calibration status indicator. The status indicator clearly informs the user whether the data used for this diagnosis is calibrated or uncalibrated, thus assisting in operational decisions in conjunction with the confidence level. The report can be presented using a combination of structured data and visual charts, such as using color coding to intuitively reflect the level of confidence. This step integrates all the key information from the preceding processes into a standardized, complete output, completing the closed loop from data verification to decision support.

[0070] In some embodiments, the verification of the effectiveness of each gas sensor corresponding to the online monitoring gas during transformer fault diagnosis includes:

[0071] In response to the trigger command for transformer fault diagnosis, the historical calibration records and residual uncertainties of each gas sensor are acquired;

[0072] For each gas sensor, the calibration duration is determined based on the interval between the most recent calibration time and the current time in the historical calibration record of the gas sensor.

[0073] If the calibration time is greater than or equal to the preset calibration time threshold, the verification of the gas sensor fails.

[0074] If the calibration time is less than a preset calibration time threshold, then the remaining uncertainty is compared with a preset remaining uncertainty threshold.

[0075] If the remaining uncertainty is greater than or equal to the remaining uncertainty threshold, the gas sensor fails verification.

[0076] Specifically, the verification of the effectiveness of each gas sensor corresponding to the online monitored gas during transformer fault diagnosis is implemented through the following logic: In response to the diagnostic trigger command, the system first acquires the historical calibration records and residual uncertainties of each gas sensor. For each sensor, the interval between its most recent calibration time and the current time is calculated to obtain the calibration duration. If the calibration duration reaches or exceeds a preset threshold, it is considered that the sensor may have drifted beyond the limit due to long-term lack of calibration, and its verification is directly deemed unsuccessful. If the calibration duration is within the allowable range, its residual uncertainty is further compared with the preset threshold. The residual uncertainty reflects the comprehensive error range that the sensor still has after calibration. If this value exceeds the allowable upper limit, it indicates that even if the calibration has not expired, its data reliability is insufficient, and the verification is also deemed unsuccessful. By examining both time and accuracy indicators sequentially, this verification mechanism can comprehensively evaluate the actual usability of the sensors, providing reliable data access basis for subsequent diagnosis.

[0077] In some embodiments, the verification of the effectiveness of each gas sensor corresponding to the online monitoring gas during transformer fault diagnosis includes:

[0078] Based on the drift prediction model of each gas sensor, the predicted drift amount of each gas sensor is determined.

[0079] The validity of the online monitoring gas data corresponding to each gas sensor is determined based on the predicted drift amount and preset tolerance of each gas sensor.

[0080] Specifically, the data validity verification logic is further refined by introducing a sensor drift prediction model. This model, built upon historical sensor performance data, runtime, and environmental factors, outputs the predicted drift amount for each sensor in future time periods. The system compares the predicted drift amount with a preset tolerance: if the predicted drift amount does not exceed the tolerance, it indicates that although the sensor has inherent errors, they are still within an acceptable range, and its data remains valuable; if the predicted drift amount reaches or exceeds the tolerance threshold, the sensor's output data is deemed unreliable, requiring calibration intervention. This method, by quantitatively assessing the sensor's performance degradation trend, achieves a forward-looking judgment on data validity, providing a dynamic evaluation basis for ensuring the quality of diagnostic data.

[0081] In some embodiments, the verification of the effectiveness of each gas sensor corresponding to the online monitoring gas during transformer fault diagnosis includes:

[0082] Receive instructions from the mobile app sent by the remote monitoring center;

[0083] The effectiveness of each gas sensor corresponding to the online gas monitoring is verified according to the instructions of the mobile APP.

[0084] like Figure 2 As shown, the steps for triggering transformer fault diagnosis also include a timed trigger mode and a remote command mode. In the timed trigger mode, a preset time period is established. When the preset time period is reached, a signal is automatically sent to the core controller, which then initiates the calibration process. In the remote command mode, commands are received from a mobile app at the remote monitoring center. When a command to initiate calibration is received, a signal is automatically sent to the core controller, which then initiates the calibration process.

[0085] In some embodiments, if the verification of any gas sensor fails, an online calibration process is triggered to obtain a calibrated gas sensor, including:

[0086] Send a command to the online calibration device to control the online calibration device to perform calibration tests on the gas sensor using standard gas, and obtain the response output of the gas sensor at multiple standard concentration points;

[0087] Based on the correspondence between each response output and the standard concentration point, an updated calibration curve for the gas sensor is generated;

[0088] Based on the uncertainty of the updated calibration curve and the standard gas, the remaining uncertainty of the gas sensor is calculated and updated.

[0089] Specifically, when a gas sensor fails verification, the system sends a control command to the online calibration device, driving it to introduce standard gas into the sensor to be calibrated and record the sensor's response signals at different standard concentration points. Based on the correspondence between these response signals and known standard concentrations, an updated calibration curve is constructed using curve fitting methods such as regression analysis. This curve establishes a correction relationship between the sensor's original output and the actual gas concentration. Subsequently, the system comprehensively considers the uncertainty of the standard gas itself and the fitting residual of the updated calibration curve, calculating a new residual uncertainty using an uncertainty synthesis method. This residual uncertainty reflects the confidence range of the sensor's measurement results after calibration. Its update ensures that the sensor's accuracy evaluation remains consistent with its current actual performance, providing an accurate error assessment benchmark for subsequent data use.

[0090] In some embodiments, calculating and updating the residual uncertainty of the gas sensor based on the uncertainty of the updated calibration curve and the standard gas includes:

[0091] Based on the fitting residuals of the updated calibration curve, the standard uncertainty of the fit, which characterizes the dispersion of the gas sensor response, is calculated.

[0092] The combined standard uncertainty of the gas sensor is calculated by combining the baseline standard uncertainty of the standard gas with the fitted standard uncertainty.

[0093] The combined standard uncertainty is multiplied by a preset inclusion factor to update the residual uncertainty of the gas sensor.

[0094] Specifically, the system calculates the fitting standard uncertainty based on the fitting residual of the updated calibration curve. This index characterizes the dispersion of the sensor response value relative to the calibration curve, reflecting the sensor's repeatability and linearity. Subsequently, the system synthesizes the baseline standard uncertainty provided by the standard gas certificate with the aforementioned fitting standard uncertainty. The synthesized standard uncertainty is typically calculated using the sum of squares and square root method. This value comprehensively reflects the errors introduced during calibration by both the standard source and the sensor's own characteristics. Finally, the synthesized standard uncertainty is multiplied by a preset coverage factor (usually k=2, corresponding to approximately 95% confidence level) to obtain the expanded uncertainty, which is then updated as the sensor's residual uncertainty. This calculation process ensures that the residual uncertainty comprehensively and quantitatively reflects the sensor's overall measurement accuracy under current conditions, providing an updated reliability index for subsequent data use.

[0095] In some embodiments, obtaining the gas concentration data of the corresponding online monitored gas and the measurement uncertainty corresponding to the gas concentration data based on the calibrated gas sensor includes:

[0096] The gas concentration data is obtained by measuring the corresponding online monitoring gas using the calibrated gas sensor.

[0097] The measurement uncertainty of the gas concentration data is calculated based on the fitting residual of the updated calibration curve of the gas sensor and the coverage of the calibration points.

[0098] Specifically, the system uses calibrated gas sensors to measure the corresponding online monitored gases. By inputting the real-time signal output by the sensors into an updated calibration curve for conversion, accurate gas concentration data is obtained. Subsequently, based on the fitting residual of the updated calibration curve and the distribution of calibration points within the measurement range, the system calculates the measurement uncertainty of the gas concentration data. The fitting residual reflects the deviation between the calibration curve and the actual sensor response, while the coverage of calibration points affects the extrapolation reliability of the curve across different concentration ranges. By combining these two factors, the system can assess the possible error range of the current measurement results, thus providing a corresponding quantitative index for uncertainty for each set of gas concentration data, ensuring that subsequent fault diagnosis can be based on complete data quality information.

[0099] In some embodiments, the fault diagnosis of the transformer based on each gas concentration data and the measurement uncertainty corresponding to each gas concentration data, to obtain the diagnosis result and the confidence level of the diagnosis result, includes:

[0100] Based on the gas concentration data and the corresponding measurement uncertainty, a diagnostic input dataset with uncertainty is obtained;

[0101] Based on the gas concentration data in the diagnostic input dataset, a preliminary fault diagnosis is performed to obtain the diagnostic results;

[0102] Based on the diagnostic results, the diagnostic input dataset, and the pre-trained fault diagnosis model, uncertainty propagation analysis is performed to obtain the probability distribution of the diagnostic results.

[0103] Based on the probability distribution, the confidence level is calculated to obtain the confidence level of the diagnostic result.

[0104] Specifically, the system first integrates the gas concentration data and their corresponding measurement uncertainties into a diagnostic input dataset with uncertainty. This dataset comprehensively describes the concentration information of the gases to be analyzed and their respective reliability ranges. Based on the gas concentration data in this dataset, the system performs preliminary fault diagnosis, outputting preliminary diagnostic results, such as identifying potential fault types, by applying a pre-trained fault diagnosis model. Subsequently, the system combines the diagnostic results, the complete diagnostic input dataset, and the fault diagnosis model to perform uncertainty propagation analysis. It evaluates the impact of the uncertainty of the input data on the diagnostic conclusion using probabilistic methods, thereby obtaining the probability distribution of the diagnostic results. Finally, the system calculates the confidence level based on this probability distribution. By analyzing the probability weight of the target diagnostic result in the distribution, it outputs the quantitative credibility of the diagnostic result, completing the transformation from data to a reliable diagnosis.

[0105] Furthermore, the training of this fault diagnosis model is a supervised learning process, the core objective of which is to establish a mapping relationship from gas concentration data to the probability distribution of fault types. The training process begins with the data preparation stage, firstly collecting a large amount of sample data from the transformer historical operation and maintenance database. Each sample should contain: the concentration values ​​of a set of key characteristic gases (such as H2, CH4, C2H2, C2H4, C2H6), and the corresponding real fault labels (such as normal, partial discharge, low temperature overheating, high temperature overheating, arc discharge, etc.) confirmed by domain experts or post-analysis. To improve the robustness and uncertainty quantification capability of the model, data augmentation is required on the original dataset. That is, based on the statistical information of historical measurement uncertainty, random noise that conforms to its actual distribution is artificially added to each gas concentration value, thereby generating multiple sample variants with slight perturbations to simulate the uncertainty of data in the real monitoring environment.

[0106] Next, feature engineering and model building are performed, using the original gas concentrations and their calculated ratios (such as C2H2 / C2H4, CH4 / H2, C2H2 / CH4) as input features. For machine learning-based models, the output layer is a probability vector, where each dimension corresponds to the probability of a fault type occurring. The core step in model training is loss function optimization, employing a special loss function, such as evidence-based cross-entropy loss. This function not only penalizes the difference between the predicted and true types but also uses the concentration of the probability distribution output by the model (i.e., the strength of evidence) as a regularization term. In this way, the model is guided to learn: for samples with high certainty, it outputs a highly concentrated probability distribution (high confidence); for samples with ambiguous features or close to the decision boundary, it outputs a more uniform probability distribution (low confidence), thus inherently distinguishing the diagnostic difficulty of different samples.

[0107] Finally, model validation and calibration are performed, evaluating the trained model on an independent test set. Evaluation metrics are used to examine the uncertainty calibration effect, i.e., whether the confidence level of the model output matches its measured accuracy (e.g., in all samples predicted as "arc discharge" by the model with 80% confidence, approximately 80% actually are arc discharges). If undercalibration or overcalibration exists, post-processing techniques such as temperature scaling are used to fine-tune the probability distribution of the output to ensure that the confidence level provided by the finally deployed model is realistic and reliable, faithfully reflecting the uncertainty of the diagnostic conclusion.

[0108] For example, during the routine monitoring of a 500kV main transformer, the system triggers a diagnostic process for dissolved gas analysis in the insulating oil. First, data from each gas sensor is collected: hydrogen (H2) concentration is 98 μL / L, with an uncertainty of ±5 μL / L; acetylene (C2H2) concentration is 8 μL / L, with an uncertainty of ±0.5 μL / L; and ethylene (C2H4) concentration is 45 μL / L, with an uncertainty of ±3 μL / L. These data are then integrated into a diagnostic input dataset with uncertainty.

[0109] Based on gas concentration data, the system uses the three-ratio method for preliminary diagnosis. The calculated characteristic gas ratio code is 102, corresponding to the fault type of arc discharge. Subsequently, uncertainty propagation analysis is performed: considering the uncertainty of each gas concentration, 10,000 Monte Carlo simulations are conducted to obtain the distribution of the ratio codes. The simulation results show that the probability of code 102 occurring is 76%, code 101 is 15%, and code 202 is 9%.

[0110] Based on this probability distribution, the system calculates the confidence level of the diagnostic results. The confidence level for the target diagnostic result, arc discharge (corresponding to code 102), is determined to be 76%. The final diagnostic report shows: Fault type: arc discharge; Confidence level: 76%; Data status: calibrated. Maintenance personnel can use this quantitative confidence index, combined with the equipment's operating status, to develop differentiated maintenance strategies.

[0111] Example 2

[0112] Please see Figure 3 This invention provides a transformer fault diagnosis device, comprising:

[0113] The verification module 301 is used to verify the effectiveness of each gas sensor corresponding to the online monitoring gas when triggering transformer fault diagnosis.

[0114] The verification module 302 is used to trigger an online calibration process to obtain a calibrated gas sensor if the verification of any gas sensor fails.

[0115] The measurement module 303 is used to obtain the gas concentration data of the corresponding online monitored gas and the measurement uncertainty corresponding to the gas concentration data based on the calibrated gas sensor.

[0116] The diagnostic module 304 is used to perform fault diagnosis on the transformer based on each gas concentration data and the measurement uncertainty corresponding to each gas concentration data, and to obtain the diagnostic result and the confidence level of the diagnostic result;

[0117] The output report module 305 is used to output a diagnostic report based on the diagnostic results and the confidence level of the diagnostic results. The diagnostic report includes at least a fault diagnosis conclusion, the confidence level of the diagnostic conclusion, and a status indicator indicating whether the data has been calibrated.

[0118] It should be noted that each module and unit in the transformer fault diagnosis device in this embodiment corresponds one-to-one with each step in the transformer fault diagnosis method in the aforementioned embodiment. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned transformer fault diagnosis method, and will not be repeated here.

[0119] Example 3

[0120] Please see Figure 4 This embodiment provides an electronic device, including at least one processor 401 and a memory 402. Optionally, the device further includes a communication component 403. The processor 401, memory 402, and communication component 403 are connected via a bus 404.

[0121] In a specific implementation, at least one processor 401 executes computer execution instructions stored in memory 402, causing at least one processor 401 to perform the above-described method.

[0122] The specific implementation process of processor 401 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0123] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0124] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0125] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0126] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0127] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0128] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0129] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0130] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0131] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0132] In addition, 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.

[0133] If a function 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 this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a 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 of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0134] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0135] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A transformer fault diagnosis method characterized by, The method comprises the following steps: when triggering transformer fault diagnosis, verifying the effectiveness of each gas sensor corresponding to the online monitoring gas; for any gas sensor, if the verification of the gas sensor fails, triggering an online calibration process to obtain the calibrated gas sensor; obtaining the gas concentration data of the corresponding online monitoring gas and the measurement uncertainty corresponding to the gas concentration data according to the calibrated gas sensor; based on each gas concentration data and the measurement uncertainty corresponding to each gas concentration data, performing fault diagnosis on the transformer to obtain a diagnosis result and a confidence level of the diagnosis result; outputting a diagnosis report according to the diagnosis result and the confidence level of the diagnosis result, wherein the diagnosis report at least includes a fault diagnosis conclusion, a confidence level of the diagnosis conclusion, and a state identifier indicating whether the data is calibrated.

2. The method of claim 1, wherein, The method comprises the following steps: in response to a trigger instruction of transformer fault diagnosis, obtaining the historical calibration record and the residual uncertainty of each gas sensor; for each gas sensor, determining the calibration duration of the gas sensor according to the interval between the latest calibration time in the historical calibration record of the gas sensor and the current time; if the calibration duration is greater than or equal to a preset calibration duration threshold, the verification of the gas sensor fails; if the calibration duration is less than the preset calibration duration threshold, comparing the residual uncertainty with a preset residual uncertainty threshold; if the residual uncertainty is greater than or equal to the residual uncertainty threshold, the verification of the gas sensor fails.

3. The method of claim 1, wherein, The method comprises the following steps: determining the predicted drift of each gas sensor according to the drift prediction model of each gas sensor; determining the effectiveness of the data of the online monitoring gas corresponding to each gas sensor according to the predicted drift of each gas sensor and a preset tolerance.

4. The method of claim 1, wherein, The method comprises the following steps: sending an instruction to an online calibration device to control the online calibration device to calibrate and test the gas sensor using standard gas to obtain the response output of the gas sensor at multiple standard concentration points; generating an updated calibration curve of the gas sensor according to the correspondence between each response output and a standard concentration point; calculating and updating the residual uncertainty of the gas sensor based on the updated calibration curve and the uncertainty of the standard gas.

5. The method of claim 4, wherein, The method comprises the following steps: from the fitting residuals of the updated calibration curve, a fitting standard uncertainty is calculated that characterizes the response dispersion of the gas sensor combining the reference standard uncertainty of the standard gas with the fitted standard uncertainty to calculate a combined standard uncertainty of the gas sensor multiplying the combined standard uncertainty by a preset inclusion factor to update the residual uncertainty of the gas sensor.

6. The method of claim 1, wherein, The gas concentration data corresponding to the online monitoring gas and the measurement uncertainty corresponding to the gas concentration data are obtained according to the calibrated gas sensor, and the measurement uncertainty of the gas concentration data is calculated according to the fitting residual of the updated calibration curve of the gas sensor and the coverage range of the calibration points. The gas concentration data is obtained by measuring the corresponding online monitoring gas according to the calibrated gas sensor. The measurement uncertainty of the gas concentration data is calculated according to the fitting residual of the updated calibration curve of the gas sensor and the coverage range of the calibration points.

7. The method of claim 1, wherein, The transformer is diagnosed based on each gas concentration data and the measurement uncertainty corresponding to each gas concentration data, and the diagnostic result and the confidence level of the diagnostic result are obtained, including: The diagnostic input data set with uncertainty is obtained according to each gas concentration data and the corresponding measurement uncertainty. The diagnostic result is obtained by performing preliminary fault diagnosis according to the gas concentration data in the diagnostic input data set. The probability distribution of the diagnostic result is obtained by performing uncertainty propagation analysis according to the diagnostic result, the diagnostic input data set and the pre-trained fault diagnosis model. The confidence level of the diagnostic result is obtained by performing confidence level calculation according to the probability distribution.

8. A transformer fault diagnostic device characterized by comprising: Including: The verification module is configured to verify the effectiveness of each gas sensor corresponding to the online monitoring gas when triggering the transformer fault diagnosis; The verification module is configured to trigger the online calibration process to obtain the calibrated gas sensor if the verification of the gas sensor fails; The measurement module is configured to obtain the gas concentration data corresponding to the online monitoring gas and the measurement uncertainty corresponding to the gas concentration data according to the calibrated gas sensor; The diagnostic module is configured to diagnose the transformer based on each gas concentration data and the measurement uncertainty corresponding to each gas concentration data, and obtain the diagnostic result and the confidence level of the diagnostic result; The output report module is configured to output a diagnostic report according to the diagnostic result and the confidence level of the diagnostic result, and the diagnostic report at least includes a fault diagnosis conclusion, a confidence level of the diagnostic conclusion, and a state identifier indicating whether the data is calibrated.

9. An electronic device, comprising: Including: At least one processor, at least one memory and computer program instructions stored in the memory, when the computer program instructions are executed by the processor, realize the method of any one of claims 1-7.

10. A computer-readable storage medium having stored thereon computer program instructions, wherein, When the computer program instructions are executed by the processor, the method of any one of claims 1-7 is realized.

Images