Transformer fault rapid diagnosis method and system

Abstract

The invention discloses a transformer fault rapid diagnosis method and system, and relates to the technical field of fault detection. The method comprises the following steps: setting a transformer standard operation state matrix of a first transformer control mode; the operation state attribute set of the transformer standard operation state matrix is sent to a sensor array control unit, and a transformer monitoring operation state matrix is received; calculating a running state deviation vector matrix; carrying out integrated model initialization based on the first transformer control mode, and configuring a fault type prediction model; processing the operation state deviation vector matrix through a fault type prediction model, and outputting a fault type list; and performing electrical preventive test detection according to the fault type list. The technical problem that in the prior art, the transformer fault detection efficiency is low is solved, the fault type is recognized in real time through the fault type prediction model, electrical preventive test detection is rapidly carried out, and the technical effect of improving the fault detection efficiency is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of fault detection, and in particular to a method and system for rapid diagnosis of transformer faults. Background Art

[0002] Transformers are vital equipment in power systems, responsible for voltage conversion and stable supply. However, transformers are easily affected by various factors during long-term operation, causing them to fail. Traditional transformer fault detection systems can usually only perform preliminary detection of whether there are abnormalities in the equipment, and cannot directly determine the specific type of fault or the specific location of the faulty component. In order to accurately locate the problem, it is often necessary to rely on complex electrical preventive test and detection processes, which is not only time-consuming, but may also lead to extended downtime and affect the normal operation of the power system. With the increase in electricity demand, traditional fault detection methods have gradually exposed their shortcomings, especially in terms of the real-time and accuracy of fault diagnosis. Existing technologies generally have problems such as low detection efficiency and inaccurate fault location, and cannot achieve rapid prediction and preventive treatment of fault types. Summary of the invention

[0003] The present application provides a method and system for rapid diagnosis of transformer faults, which solves the technical problem of low transformer fault detection efficiency in the prior art.

[0004] In view of the above problems, the present application provides a method and system for rapid diagnosis of transformer faults.

[0005] In a first aspect of the present application, a method for rapid diagnosis of transformer faults is provided, the method comprising: A transformer standard operating state matrix of a first transformer control mode is set; an operating state attribute set of the transformer standard operating state matrix is ​​sent to a sensor array control unit, and a transformer monitoring operating state matrix is ​​received; an operating state deviation vector matrix of the transformer standard operating state matrix and the transformer monitoring operating state matrix is ​​calculated; an integrated model is initialized based on the first transformer control mode, and a fault type prediction model is configured; the operating state deviation vector matrix is ​​processed by the fault type prediction model, and a fault type list is output; and electrical preventive test detection is performed according to the fault type list.

[0006] A second aspect of the present application provides a transformer fault rapid diagnosis system, the system comprising: State matrix setting module: sets the transformer standard operation state matrix of the first transformer control mode; sensor control module: sends the operation state attribute set of the transformer standard operation state matrix to the sensor array control unit, and receives the transformer monitoring operation state matrix; deviation calculation module: calculates the operation state deviation vector matrix of the transformer standard operation state matrix and the transformer monitoring operation state matrix; model configuration module: performs integrated model initialization based on the first transformer control mode, and configures the fault type prediction model; fault prediction module: processes the operation state deviation vector matrix through the fault type prediction model, and outputs a fault type list; test detection module: performs electrical preventive test detection according to the fault type list.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, the transformer standard operating state matrix of the first transformer control mode is set. Next, the operating state attribute set of the transformer standard operating state matrix is ​​sent to the sensor array control unit to receive the transformer monitoring operating state matrix. Then, the operating state deviation vector matrix of the transformer standard operating state matrix and the transformer monitoring operating state matrix is ​​calculated. Next, the integrated model is initialized based on the first transformer control mode, and the fault type prediction model is configured; then the operating state deviation vector matrix is ​​processed by the fault type prediction model, and a fault type list is output. Finally, electrical preventive test detection is performed according to the fault type list. The technical problem of low transformer fault detection efficiency in the prior art is solved. Through the fault type prediction model, the fault type is identified in real time and the electrical preventive test detection is quickly performed, thereby achieving the technical effect of improving the fault detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 A schematic flow chart of a method for rapid transformer fault diagnosis provided in an embodiment of the present application.

[0010] Figure 2 A schematic diagram of the structure of a transformer fault rapid diagnosis system provided in an embodiment of the present application.

[0011] Explanation of the reference numerals: state matrix setting module 11 , sensor control module 12 , deviation calculation module 13 , model configuration module 14 , fault prediction module 15 , test detection module 16 . DETAILED DESCRIPTION

[0012] The present application solves the technical problem of low transformer fault detection efficiency in the prior art by providing a transformer fault rapid diagnosis method and system.

[0013] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0014] It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or are inherent to these processes, methods, products or devices.

[0015] Embodiment 1, as Figure 1 As shown, the present application provides a method for rapid diagnosis of transformer faults, wherein the method comprises: The transformer standard operating state matrix of the first transformer control mode is set.

[0016] Setting the transformer standard operating state matrix of the first transformer control mode means defining a standard operating state matrix based on various parameters of the transformer under normal and stable operating conditions. This matrix contains various performance indicators of the transformer under various typical working conditions, such as temperature, voltage, current, load, power factor, vibration and other data. The transformer standard operating state matrix is ​​used as a reference benchmark for fault diagnosis.

[0017] Furthermore, setting a transformer standard operation state matrix of the first transformer control mode includes: Perform fault sample association analysis on the first transformer control mode to obtain the operating status attribute set; perform healthy sample analysis on the operating status attribute set according to the first transformer control mode to obtain a standard interval set of operating status attributes; and construct the transformer standard operating status matrix according to the standard interval set of operating status attributes.

[0018] Specifically, the fault samples of the first control mode of the transformer are analyzed for correlation, including analyzing historical fault data or experimental data to identify the relationship between the fault and the transformer operating state; by analyzing different fault samples, key operating state attributes related to the occurrence of the fault are extracted, such as temperature, current, voltage, vibration, etc. Analysis is performed based on the normal health state data of the transformer; by studying the performance of the transformer during normal and stable operation, the standard intervals (normal ranges) of various operating state attributes are determined, such as the normal current interval, temperature interval, etc. These standard intervals represent the working characteristics of the transformer when there is no fault.

[0019] After obtaining the set of operating state attributes and the set of standard intervals, the transformer standard operating state matrix is ​​constructed based on these data. The transformer standard operating state matrix combines various operating state attributes and their standard intervals to form a multi-dimensional matrix structure, where each dimension represents an operating state attribute, and the interval within each dimension represents the range of variation of the attribute under normal conditions.

[0020] In summary, the transformer standard operating state matrix that sets the first transformer control mode establishes an operating state matrix based on the standard interval through the analysis of fault samples and the induction of healthy samples, which provides reliable data support for the rapid diagnosis of transformer faults.

[0021] Furthermore, performing fault sample association analysis on the first transformer control mode to obtain the operating state attribute set includes: A first fault sample of the first transformer control mode is obtained, wherein the first fault sample includes a first abnormal operating state attribute set whose initial operating state attribute deviation is greater than or equal to an initial operating state attribute deviation threshold; until an Mth fault sample of the first transformer control mode is obtained, wherein the Mth fault sample includes an Mth abnormal operating state attribute set whose initial operating state attribute deviation is greater than or equal to the initial operating state attribute deviation threshold, M≥50000, and M is an integer; counting the first abnormal operating state attribute set until the ratio of the first abnormal operating state attribute occurrence frequency of the Mth abnormal operating state attribute set to M is set as the first abnormal operating state attribute sensitivity; until counting the first abnormal operating state attribute set until the ratio of the Qth abnormal operating state attribute occurrence frequency of the Mth abnormal operating state attribute set to M is set as the Qth abnormal operating state attribute sensitivity, where Q represents the total number of abnormal operating state attributes; extracting the first abnormal operating state attribute sensitivity until the abnormal operating state attribute whose sensitivity of the Qth abnormal operating state attribute is greater than or equal to the sensitivity threshold is set as the operating state attribute set.

[0022] Specifically, a first fault sample of the first transformer control mode is obtained, the first fault sample includes a set of abnormal operating state attributes whose deviation in the initial operating state is greater than or equal to the set initial operating state attribute deviation threshold. Optionally, the operating state of the transformer is monitored, and by comparing with the benchmark data in the normal state, those operating state attributes with obvious deviations are found, and these abnormal attributes are classified as the first fault sample. Continue to collect fault samples until the Mth fault sample is obtained, each fault sample also includes a set of abnormal operating state attributes whose deviation in the initial operating state is greater than or equal to the initial operating state attribute deviation threshold. Assume that M≥50000, indicating that the number of fault samples is at least 50,000, and M is an integer greater than or equal to 50,000. All collected fault samples are analyzed, the frequency of each abnormal operating state attribute in the fault sample is counted, and the ratio to the total number of samples M is calculated. This ratio represents the frequency of occurrence of the abnormal operating state attribute, that is, the sensitivity. In this way, the sensitivity of the first abnormal operating state attribute is calculated, that is, the ratio of the attribute in all fault samples; similarly, the sensitivity of the Qth abnormal operating state attribute is calculated, where Q represents the total number of abnormal operating state attributes. After obtaining the sensitivity of all abnormal operating state attributes, those abnormal operating state attributes whose sensitivity is greater than or equal to the set sensitivity threshold are screened out. These attributes are considered to be key abnormal state attributes that are highly correlated with faults and appear frequently, so they are extracted as the final set of operating state attributes.

[0023] Furthermore, a healthy sample analysis is performed on the operating state attribute set according to the first transformer control mode to obtain an operating state attribute standard interval set, including: Taking the first transformer control mode as a constraint, collecting a first healthy sample, wherein the first healthy sample has a first operating state attribute feature value set of the operating state attribute set; Until the Nth healthy sample is collected with the first transformer control mode as a constraint, wherein the Nth healthy sample has the Nth operating state attribute feature value set of the operating state attribute set, N≥50000, and N is an integer; Based on the first running status attribute feature value set to the Nth running status attribute feature value set, concentrated intervals of the running status attribute feature values ​​of the running status attribute sets are counted and set as the running status attribute standard interval set.

[0024] Under the constraint of the first transformer control mode, the first healthy sample is first collected. The first healthy sample contains a first operating state attribute characteristic value set corresponding to the operating state attribute set, that is, records the actual values ​​of various attributes of the transformer under normal operating conditions. The first healthy sample is used to provide the state data of the transformer under fault-free conditions, providing a basis for subsequent analysis of the standard interval. Under the same first transformer control mode constraint, subsequent healthy samples are collected until the Nth healthy sample is obtained. Each healthy sample contains a corresponding operating state attribute characteristic value set in the operating state attribute set, N≥50000, indicating that at least 50000 healthy samples are required (N is an integer greater than or equal to 50000). The data of these healthy samples can reflect the normal operating state of the transformer under different conditions.

[0025] Based on the set of all collected operating state attribute feature values, statistical analysis is performed to determine the concentration interval of each operating state attribute, that is, to count the distribution range of each feature value. This interval represents the fluctuation range of the attribute under normal operation. By performing cluster analysis or distribution statistics on the feature values ​​of each attribute in these healthy samples, the standard interval of each attribute is calculated. These intervals are the standard interval set, which represents the reasonable range of each attribute under normal working conditions.

[0026] Further, extracting the first abnormal operation state attribute sensitivity until the abnormal operation state attribute whose sensitivity of the Qth abnormal operation state attribute is greater than or equal to the sensitivity threshold, setting it as the operation state attribute set, further includes: Extract the abnormal operating state attributes whose sensitivity of the first abnormal operating state attribute is greater than or equal to the sensitivity threshold until the sensitivity of the Qth abnormal operating state attribute is greater than or equal to the sensitivity threshold, and set them as a set of candidate operating state attributes; perform pairwise Pearson correlation coefficient analysis on the set of candidate operating state attributes to obtain a set of Pearson correlation coefficients; cluster the set of candidate operating state attributes according to the set of Pearson correlation coefficients to obtain multiple clusters of candidate operating state attributes; extract one candidate operating state attribute from each of the multiple clusters of candidate operating state attributes, and add them into the set of operating state attributes.

[0027] From all abnormal operating state attributes, those abnormal operating state attributes whose sensitivity is greater than or equal to the sensitivity threshold are screened out. These attributes are set as the set of candidate operating state attributes. These attributes have high sensitivity when a fault occurs and can effectively reflect the fault signs of the transformer. Pearson correlation coefficient analysis is performed on all attributes in the set of candidate operating state attributes. The Pearson correlation coefficient is used to measure the degree of linear correlation between attributes, ranging from -1 to 1, indicating a positive or negative correlation between attributes. Through this analysis method, it is possible to identify whether there is a strong correlation between the candidate attributes, thereby avoiding the selection of redundant attributes in subsequent processing.

[0028] Based on the Pearson correlation coefficient set, the selected operating state attributes are clustered; through cluster analysis of the correlation coefficient, highly correlated attributes are classified into one category to reduce the impact of highly correlated attributes on fault diagnosis. The selected operating state attributes in each cluster are screened, and the most representative attribute is extracted as the representative attribute of the cluster, and added to the final operating state attribute set. In this way, by selecting representative attributes from multiple clusters, the final operating state attribute set contains both key fault information and avoids redundancy and duplication between attributes.

[0029] The operating state attribute set of the transformer standard operating state matrix is ​​sent to the sensor array control unit, and the transformer monitoring operating state matrix is ​​received.

[0030] The operating state attribute set of the transformer standard operating state matrix is ​​sent to the sensor array control unit. The sensor array control unit is responsible for data interaction with each sensor of the transformer and controls the monitoring task of the sensor according to these attribute values ​​to ensure that the sensor array can collect data according to the predetermined standards. The sensor array control unit monitors the operating state of the transformer in real time by cooperating with the sensor and generates the transformer monitoring operating state matrix based on the real-time collected data. The transformer monitoring operating state matrix contains various attribute data of the transformer in actual operation, reflecting the actual working state of the current transformer.

[0031] An operating state deviation vector matrix of the transformer standard operating state matrix and the transformer monitoring operating state matrix is ​​calculated.

[0032] The transformer standard operating state matrix is ​​set through the above steps, and contains the standard values ​​of various performance indicators (such as current, voltage, temperature, load, etc.) of the transformer under normal operating conditions; the transformer monitoring operating state matrix is ​​the actual operating data collected in real time by the sensor array, reflecting the actual performance of the transformer at the current moment. Both the standard operating state matrix and the monitoring operating state matrix contain the same set of operating state attributes, which represent various operating indicators of the transformer, such as current, voltage, temperature, etc. The operating state deviation vector is the difference between the standard operating state matrix and the monitoring operating state matrix. Specifically, for each operating state attribute (such as current, voltage, temperature, etc.), the deviation can be calculated in the following way: deviation value = monitoring value - standard value; for each operating state attribute, calculate its deviation between the standard state and the monitoring state. The deviation value can be positive or negative, representing the difference between the monitoring state and the standard state. If the deviation value is positive, it means that the monitoring value is greater than the standard value, otherwise it is negative, indicating that the monitoring value is less than the standard value. For all operating status attributes, the deviation value of each attribute is obtained through the above calculation. All deviation values ​​constitute a vector, which represents the deviation of all attributes. The deviation vectors of each attribute are combined into a matrix to obtain a complete operating status deviation vector matrix. Each row of this matrix represents the deviation value of each attribute at a sampling time, and each column corresponds to a specific operating status attribute.

[0033] The integrated model is initialized based on the first transformer control mode, and the fault type prediction model is configured.

[0034] The first transformer control mode is a standard mode set based on the transformer's design parameters, operating conditions, and control strategy. The first transformer control mode represents the transformer's operating behavior under normal conditions and provides a reference framework for fault detection. The initialization of the control mode includes determining which operating state attributes (such as current, voltage, temperature, etc.) are key factors affecting transformer performance and setting standard ranges for these attributes. The integrated model is initialized using the preset transformer control mode, and the fault type prediction model is set to prepare for accurate prediction of fault types.

[0035] Furthermore, the integrated model is initialized based on the first transformer control mode, and the fault type prediction model is configured, including: Obtain a fault type prediction unit model; count historical work logs of the fault type prediction unit model in a preset time zone in the past; count prediction accuracy based on the historical work logs; when the prediction accuracy is less than or equal to a prediction accuracy threshold, collect at least 5 fault type prediction unit models for parallel integration configuration, use the intersection function to fully connect the output layer of the integration model, and obtain the fault type prediction model.

[0036] The fault type prediction unit model is a preliminarily trained model responsible for predicting the fault type of the transformer's operating status. The fault type prediction unit model is usually based on an ensemble learning method, which can process multi-dimensional input data and give possible fault type prediction results. The fault type prediction unit model uses a combination of multiple basic learning algorithms, such as decision trees, support vector machines (SVMs), neural networks, etc., to perform prediction tasks.

[0037] The working data of the model is collected by monitoring and recording the historical work logs of the fault type prediction unit model. These logs contain the input data, processing time, prediction results, and comparison of the actual fault types of the model. According to the data in the historical work logs, the prediction accuracy of the model is statistically analyzed, including the difference between the fault type output by the comparison model and the actual fault type, and the accuracy is calculated. The prediction accuracy = the number of correct predictions / the total number of predictions; if the prediction accuracy is less than or equal to the set prediction accuracy threshold (the threshold can be set according to actual needs, such as 90%), the model needs to be optimized. In this case, at least 5 fault type prediction unit models are collected and configured for parallel integration. Each prediction unit model provides multiple prediction results for the integrated model by running in parallel.

[0038] The intersection function is used to connect the output layers of multiple parallel models and combine the prediction results of all models. The intersection function usually processes the outputs of multiple models based on majority voting, weighted average or other integration methods. The intersection function outputs the final fault type prediction based on the results of multiple prediction units; the integrated model after the intersection function is fully connected becomes the final fault type prediction model.

[0039] Furthermore, the steps of constructing the fault type prediction unit model include: Collecting a state deviation vector record matrix of the first transformer control mode operation, wherein the operation state deviation vector record matrix has a detected fault type identification set; The detected fault type identification set is used as supervisory data, and the state deviation vector record matrix is ​​used as input data to train a classifier to generate the fault type prediction unit model.

[0040] The state deviation vector record matrix is ​​generated by monitoring the deviation data of the transformer under different operating conditions. Each row represents the deviation value of the operating state attribute at a specific moment or time period, and each column corresponds to the deviation of different operating state attributes (such as current, voltage, temperature, etc.). The state deviation vector record matrix reflects the difference between the current operating state of the transformer and the standard operating state; the state deviation vector record matrix also includes a set of detected fault type identifications corresponding to each data sample, that is, whether the transformer has a fault at each moment or time period, and the fault type (such as overload, short circuit, overtemperature, etc.).

[0041] The detected fault type identification set is used as supervision data, providing the true label (fault type) corresponding to each sample. These labels are used to guide the learning process of the classifier, allowing the classifier to identify the correct fault type from the input data. The data in the state deviation vector record matrix is ​​used as input to the classifier for training. The deviation data in each row is used as the feature input of the model, indicating the difference between the operating state of the transformer at that time point and the standard state. These deviation values ​​are the basis for the classifier to learn the performance characteristics of the transformer under different fault conditions. The classifier will discover the potential patterns of different fault types through these input data.

[0042] The classifier is trained using appropriate classification algorithms (such as decision trees, support vector machines, neural networks, etc.) using input data and corresponding supervision labels (fault type identification sets); during the training process, the classifier will iteratively learn based on the input data and corresponding fault type labels in the state deviation vector record matrix to minimize the prediction error and optimize the model parameters. After training, the classifier will generate a fault type prediction unit model that can predict the fault type based on the new operating status data.

[0043] The operating state deviation vector matrix is ​​processed by the fault type prediction model to output a fault type list.

[0044] The deviation vector is processed using the trained fault type prediction model to predict and generate a list of possible fault types, clarifying possible fault problems of the transformer, such as overload fault, insulation fault, and overtemperature.

[0045] Conduct electrical preventive tests based on the fault type list.

[0046] The fault type list is generated by the fault type prediction model, which includes the possible fault types of the transformer at different time points or states. Each fault type represents a specific problem that may exist in the transformer, such as overload, overtemperature, electrical short circuit, etc. According to the main fault types identified in the fault type list, select the corresponding electrical preventive test to further confirm the fault type and ensure timely maintenance or replacement measures to avoid equipment downtime or serious failures.

[0047] In summary, the embodiments of the present application have at least the following technical effects: First, the transformer standard operating state matrix of the first transformer control mode is set. Next, the operating state attribute set of the transformer standard operating state matrix is ​​sent to the sensor array control unit to receive the transformer monitoring operating state matrix. Then, the operating state deviation vector matrix of the transformer standard operating state matrix and the transformer monitoring operating state matrix is ​​calculated. Next, the integrated model is initialized based on the first transformer control mode, and the fault type prediction model is configured; then the operating state deviation vector matrix is ​​processed by the fault type prediction model, and a fault type list is output. Finally, electrical preventive test detection is performed according to the fault type list. The technical problem of low transformer fault detection efficiency in the prior art is solved. Through the fault type prediction model, the fault type is identified in real time and the electrical preventive test detection is quickly performed, thereby achieving the technical effect of improving the fault detection efficiency.

[0048] Embodiment 2, based on the same inventive concept as a method for rapid diagnosis of transformer faults in the above embodiment, Figure 2 As shown, the present application provides a transformer fault rapid diagnosis system, wherein the system comprises: State matrix setting module 11: sets the transformer standard operating state matrix of the first transformer control mode; sensor control module 12: sends the operating state attribute set of the transformer standard operating state matrix to the sensor array control unit, and receives the transformer monitoring operating state matrix; deviation calculation module 13: calculates the operating state deviation vector matrix of the transformer standard operating state matrix and the transformer monitoring operating state matrix; model configuration module 14: performs integrated model initialization based on the first transformer control mode, and configures the fault type prediction model; fault prediction module 15: processes the operating state deviation vector matrix through the fault type prediction model, and outputs a fault type list; test detection module 16: performs electrical preventive test detection according to the fault type list.

[0049] Furthermore, the state matrix setting module 11 is used to perform the following method: Perform fault sample association analysis on the first transformer control mode to obtain the operating status attribute set; perform healthy sample analysis on the operating status attribute set according to the first transformer control mode to obtain a standard interval set of operating status attributes; and construct the transformer standard operating status matrix according to the standard interval set of operating status attributes.

[0050] Furthermore, the state matrix setting module 11 is used to perform the following method: A first fault sample of the first transformer control mode is obtained, wherein the first fault sample includes a first abnormal operating state attribute set whose initial operating state attribute deviation is greater than or equal to an initial operating state attribute deviation threshold; until an Mth fault sample of the first transformer control mode is obtained, wherein the Mth fault sample includes an Mth abnormal operating state attribute set whose initial operating state attribute deviation is greater than or equal to the initial operating state attribute deviation threshold, M≥50000, and M is an integer; counting the first abnormal operating state attribute set until the ratio of the first abnormal operating state attribute occurrence frequency of the Mth abnormal operating state attribute set to M is set as the first abnormal operating state attribute sensitivity; until counting the first abnormal operating state attribute set until the ratio of the Qth abnormal operating state attribute occurrence frequency of the Mth abnormal operating state attribute set to M is set as the Qth abnormal operating state attribute sensitivity, where Q represents the total number of abnormal operating state attributes; extracting the first abnormal operating state attribute sensitivity until the abnormal operating state attribute whose sensitivity of the Qth abnormal operating state attribute is greater than or equal to the sensitivity threshold is set as the operating state attribute set.

[0051] Furthermore, the state matrix setting module 11 is used to perform the following method: Taking the first transformer control mode as a constraint, a first healthy sample is collected, wherein the first healthy sample has a first operating state attribute feature value set of the operating state attribute set; until the first transformer control mode is used as a constraint, an Nth healthy sample is collected, wherein the Nth healthy sample has an Nth operating state attribute feature value set of the operating state attribute set, N≥50000, and N is an integer; based on the first operating state attribute feature value set until the Nth operating state attribute feature value set, a concentrated interval of the operating state attribute feature values ​​of the operating state attribute set is counted, and is set as the operating state attribute standard interval set.

[0052] Furthermore, the state matrix setting module 11 is used to perform the following method: Extract the abnormal operating state attributes whose sensitivity of the first abnormal operating state attribute is greater than or equal to the sensitivity threshold until the sensitivity of the Qth abnormal operating state attribute is greater than or equal to the sensitivity threshold, and set them as a set of candidate operating state attributes; perform pairwise Pearson correlation coefficient analysis on the set of candidate operating state attributes to obtain a set of Pearson correlation coefficients; cluster the set of candidate operating state attributes according to the set of Pearson correlation coefficients to obtain multiple clusters of candidate operating state attributes; extract one candidate operating state attribute from each of the multiple clusters of candidate operating state attributes, and add them into the set of operating state attributes.

[0053] Furthermore, the model configuration module 14 is used to execute the following method: Obtain a fault type prediction unit model; count historical work logs of the fault type prediction unit model in a preset time zone in the past; count prediction accuracy based on the historical work logs; when the prediction accuracy is less than or equal to a prediction accuracy threshold, collect at least 5 fault type prediction unit models for parallel integration configuration, use the intersection function to fully connect the output layer of the integration model, and obtain the fault type prediction model.

[0054] Furthermore, the model configuration module 14 is used to execute the following method: The state deviation vector record matrix of the first transformer control mode operation is collected, wherein the operation state deviation vector record matrix has a detected fault type identification set; the detected fault type identification set is used as supervision data and the state deviation vector record matrix is ​​used as input data to train a classifier to generate the fault type prediction unit model.

[0055] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0056] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0057] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for rapid diagnosis of transformer faults, characterized in that: The method comprises: Setting a transformer standard operating state matrix of a first transformer control mode; Sending the operating state attribute set of the transformer standard operating state matrix to the sensor array control unit, and receiving the transformer monitoring operating state matrix; Calculating an operating state deviation vector matrix of the transformer standard operating state matrix and the transformer monitoring operating state matrix; Initialize the integrated model based on the first transformer control mode and configure the fault type prediction model; Processing the operating state deviation vector matrix through the fault type prediction model and outputting a fault type list; Conduct electrical preventive tests based on the fault type list.

2. The method according to claim 1, characterized in that The transformer standard operation state matrix of the first transformer control mode is set, including: Performing fault sample association analysis on the first transformer control mode to obtain the operating state attribute set; Performing a healthy sample analysis on the operating state attribute set according to the first transformer control mode to obtain an operating state attribute standard interval set; The transformer standard operating state matrix is ​​constructed according to the operating state attribute standard interval set.

3. The method according to claim 2, characterized in that Performing fault sample association analysis on the first transformer control mode to obtain the operating state attribute set includes: Obtaining a first fault sample of the first transformer control mode, wherein the first fault sample includes a first abnormal operating state attribute set whose initial operating state attribute deviation is greater than or equal to an initial operating state attribute deviation threshold; until an Mth fault sample of the first transformer control mode is obtained, wherein the Mth fault sample includes an Mth abnormal operating state attribute set whose initial operating state attribute deviation is greater than or equal to an initial operating state attribute deviation threshold, M≥50000, and M is an integer; Counting the ratio of the occurrence frequency of the first abnormal operation state attribute of the first abnormal operation state attribute set until the Mth abnormal operation state attribute set to M, and setting it as the first abnormal operation state attribute sensitivity; Until the first abnormal operation state attribute set is counted, the ratio of the occurrence frequency of the Qth abnormal operation state attribute of the Mth abnormal operation state attribute set to M is set as the Qth abnormal operation state attribute sensitivity, where Q represents the total number of abnormal operation state attributes; The abnormal operating state attributes whose sensitivity of the first abnormal operating state attribute is greater than or equal to the sensitivity threshold are extracted and set as the operating state attribute set.

4. The method according to claim 2, characterized in that Performing a healthy sample analysis on the operating state attribute set according to the first transformer control mode to obtain an operating state attribute standard interval set includes: Taking the first transformer control mode as a constraint, collecting a first healthy sample, wherein the first healthy sample has a first operating state attribute feature value set of the operating state attribute set; Until the Nth healthy sample is collected with the first transformer control mode as a constraint, wherein the Nth healthy sample has the Nth operating state attribute feature value set of the operating state attribute set, N≥50000, and N is an integer; Based on the first running status attribute feature value set to the Nth running status attribute feature value set, concentrated intervals of the running status attribute feature values ​​of the running status attribute sets are counted and set as the running status attribute standard interval set.

5. The method according to claim 3, characterized in that Extracting the first abnormal operation state attribute sensitivity until the abnormal operation state attribute whose sensitivity of the Qth abnormal operation state attribute is greater than or equal to the sensitivity threshold, setting the abnormal operation state attribute set as the operation state attribute set, further comprising: Extracting the abnormal operating state attributes whose sensitivity of the first abnormal operating state attribute is greater than or equal to the sensitivity threshold until the Qth abnormal operating state attribute sensitivity is greater than or equal to the sensitivity threshold, and setting them as a candidate operating state attribute set; Performing pairwise Pearson correlation coefficient analysis on the candidate running state attribute sets to obtain a Pearson correlation coefficient set; Clustering the set of candidate operating state attributes according to the set of Pearson correlation coefficients to obtain multiple clusters of candidate operating state attributes; A candidate running state attribute is extracted from each of the multiple clusters of candidate running state attributes, and added into the running state attribute set.

6. The method according to claim 1, characterized in that Initialize the integrated model based on the first transformer control mode and configure the fault type prediction model, including: Obtaining a fault type prediction unit model; Counting historical work logs of the fault type prediction unit model in a preset time zone in the past; Counting the prediction accuracy based on the historical work log; When the prediction accuracy is less than or equal to the prediction accuracy threshold, at least 5 fault type prediction unit models are collected for parallel integration configuration, and the output layer of the integrated model is fully connected using the intersection function to obtain the fault type prediction model.

7. The method according to claim 6, characterized in that The steps of constructing the fault type prediction unit model include: Collecting a state deviation vector record matrix of the first transformer control mode operation, wherein the operation state deviation vector record matrix has a detected fault type identification set; The detected fault type identification set is used as supervisory data, and the state deviation vector record matrix is ​​used as input data to train a classifier to generate the fault type prediction unit model.

8. A transformer fault rapid diagnosis system, characterized in that: A method for quickly diagnosing transformer faults according to any one of claims 1 to 7 is implemented, the system comprising: State matrix setting module: setting the transformer standard operation state matrix of the first transformer control mode; The sensor control module sends the operating state attribute set of the transformer standard operating state matrix to the sensor array control unit, and receives the transformer monitoring operating state matrix; Deviation calculation module: calculates the operating state deviation vector matrix of the transformer standard operating state matrix and the transformer monitoring operating state matrix; Model configuration module: initialize the integrated model based on the first transformer control mode and configure the fault type prediction model; Fault prediction module: processes the operating state deviation vector matrix through the fault type prediction model and outputs a fault type list; Test detection module: perform electrical preventive test detection according to the fault type list.

Images