Fault detection method and device for immersed mutual inductor

Abstract

The invention is suitable for the technical field of mutual inductors, and provides a fault detection method and device for an immersed mutual inductor, and the method comprises the steps: obtaining the real-time data of a plurality of performance indexes of the immersed mutual inductor; according to the real-time data of the plurality of performance indexes, the health degree of the immersed transformer is evaluated through a preset health degree evaluation model; according to the health degree, judging whether the immersed mutual inductor fails or not; if the immersed mutual inductor fails, extracting time sequence data of each performance index in a preset time period, and performing window division on the time sequence data of each performance index through a time window; and according to the data sudden change condition in each window, carrying out secondary confirmation on the fault. According to the invention, the fault of the immersed transformer can be detected more accurately.

Description

Technical Field

[0001] The present invention belongs to the technical field of transformers, and particularly relates to a fault detection method and device for water - immersed transformers. Background Art

[0002] As an indispensable key device in the power system, the transformer undertakes the important task of proportionally converting high voltage and large current into low voltage and small current for operations such as measurement, protection, and control. However, in the actual operating environment, the transformer may be affected by various adverse factors, and water immersion is a relatively common situation that may lead to serious consequences.

[0003] When the transformer is water - immersed, its internal insulation performance, electrical performance, and mechanical structure may all be damaged. For example, moisture may penetrate into the insulation material, reducing its insulation resistance value, resulting in a decline in insulation performance, increasing the risk of leakage, and even causing a short - circuit fault in severe cases, posing a direct threat to the safe and stable operation of the power system. At the same time, water immersion may also affect the excitation characteristics of the transformer, changing its ratio accuracy, and thus affecting the normal operation of measurement and protection devices. In addition, the presence of moisture may trigger chemical reactions such as corrosion, damaging the metal components and connectors of the transformer, and affecting its mechanical strength and stability.

[0004] Traditional transformer fault detection methods have many limitations when faced with the complex situation after water immersion. For example, methods based on single - parameter or simple threshold judgment cannot comprehensively consider the changes in multiple performance indicators of the transformer after water immersion, easily leading to misjudgment or missed judgment. In addition, traditional methods often lack the ability to deeply analyze and mine data, resulting in low accuracy. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a fault detection method and device for water - immersed transformers to more accurately detect faults of water - immersed transformers.

[0006] The first aspect of the embodiments of the present invention provides a fault detection method for water - immersed transformers, including:

[0007] Obtaining real - time data of multiple performance indicators of the water - immersed transformer;

[0008] Evaluating the health status of the water - immersed transformer through a preset health status evaluation model according to the real - time data of the multiple performance indicators;

[0009] Judging whether the water - immersed transformer is faulty according to the health status;

[0010] If the immersed current transformer fails, extract the time series data of each performance index within a preset time period, and divide the time series data of each performance index through a time window;

[0011] Perform a secondary confirmation of the fault according to the data mutation situation in each window.

[0012] Combined with the first aspect, in a possible implementation manner of the first aspect, the performing a secondary confirmation of the fault according to the data mutation situation in each window includes:

[0013] For the time series data of any one performance index, calculate the angular value between adjacent data in each window, and take the average value of each angular value in each window as the mutation value of each window;

[0014] Judge whether each window is a mutation window according to the mutation value of each window;

[0015] Determine the time points and window values of each mutation window;

[0016] According to the time points of each mutation window, select the window values of the comparison windows corresponding to each mutation window from the time series data of this performance index of multiple normal current transformers;

[0017] Perform a secondary confirmation of the fault according to the window values of each mutation window and each comparison window.

[0018] Combined with the first aspect, in a possible implementation manner of the first aspect, the determining the time points and window values of each mutation window includes:

[0019] Take the acquisition moment of the first data in the mutation window as the time point of the mutation window;

[0020] Calculate the average value of the data in the mutation window as the window value of the mutation window.

[0021] Combined with the first aspect, in a possible implementation manner of the first aspect, the performing a secondary confirmation of the fault according to the window values of each mutation window and each comparison window includes:

[0022] Generate a first window value sequence according to the window values of each mutation window;

[0023] Generate a second window value sequence corresponding to each normal current transformer according to the window values of the comparison windows of each normal current transformer;

[0024] Calculate the similarity between the first window value sequence and each second window value sequence respectively, and calculate the average value of each similarity to obtain the comprehensive similarity corresponding to this performance index;

[0025] Based on the comprehensive similarity, perform a secondary confirmation of the fault.

[0026] Combined with the first aspect, in a possible implementation manner of the first aspect, the performing a secondary confirmation of the fault based on the comprehensive similarity includes:

[0027] If the comprehensive similarity is less than a preset similarity threshold, confirm the fault of the water - immersed current transformer;

[0028] After the secondary confirmation of the fault, it further includes:

[0029] If the fault of the water - immersed current transformer is confirmed, determine the severity of the fault according to the comprehensive similarity corresponding to each performance index.

[0030] Combined with the first aspect, in a possible implementation manner of the first aspect, before obtaining the real - time data of multiple performance indexes of the water - immersed current transformer, it further includes: performing an importance analysis on the multiple performance indexes to obtain the importance weights of each performance index;

[0031] The evaluating the health of the water - immersed current transformer according to the real - time data of the multiple performance indexes through a preset health evaluation model includes:

[0032] According to Evaluate the health of the water - immersed current transformer;

[0033] where H is the health; n is the number of performance indexes; w i is the pre - calculated importance weight of the performance index; x i is the real - time data of the standardized performance index.

[0034] Combined with the first aspect, in a possible implementation manner of the first aspect, before obtaining the real - time data of multiple performance indexes of the water - immersed current transformer, it further includes: performing a correlation analysis on the multiple performance indexes to obtain the correlation between each performance index;

[0035] After obtaining the real - time data of the multiple performance indexes of the water - immersed current transformer, it further includes:

[0036] According to the correlation, pre - process the real - time data of the multiple performance indexes; where the pre - processing includes: correcting abnormal data and interpolating missing data.

[0037] Combined with the first aspect, in a possible implementation manner of the first aspect, after the secondary confirmation according to the fault, it further includes:

[0038] If it is confirmed that the water - immersed current transformer is faulty, then based on the real - time data of the multiple performance indicators and a pre - trained fault classification model, determine the fault type of the water - immersed current transformer.

[0039] Combined with the first aspect, in a possible implementation manner of the first aspect, the multiple performance indicators include at least one of the following:

[0040] Insulation resistance value, sealing integrity parameter, excitation characteristic parameter, transformation ratio, and ambient temperature.

[0041] The second aspect of the embodiments of the present invention provides a fault detection device for a water - immersed current transformer, including:

[0042] An acquisition module, configured to acquire real - time data of multiple performance indicators of the water - immersed current transformer;

[0043] An evaluation module, configured to evaluate the health status of the water - immersed current transformer according to the real - time data of the multiple performance indicators through a preset health status evaluation model; and determine whether the water - immersed current transformer is faulty according to the health status;

[0044] A confirmation module, configured to, if the water - immersed current transformer is faulty, extract the time - series data of each performance indicator within a preset time period, and perform window division on the time - series data of each performance indicator through a time window; and perform secondary confirmation of the fault according to the data mutation situation within each window.

[0045] Combined with the second aspect, in a possible implementation manner of the second aspect, the confirmation module is specifically configured to:

[0046] For the time - series data of any one performance indicator, calculate the angular value between adjacent data within each window, and use the average value of each angular value within each window as the mutation value of each window;

[0047] Judge whether each window is a mutation window according to the mutation value of each window;

[0048] Determine the time points and window values of each mutation window;

[0049] According to the time points of each mutation window, select the window values of the corresponding comparison windows from the time - series data of this performance indicator of multiple normal current transformers;

[0050] Perform secondary confirmation of the fault according to the window values of each mutation window and each comparison window.

[0051] Combined with the second aspect, in a possible implementation manner of the second aspect, the confirmation module is specifically configured to:

[0052] Take the acquisition time of the first data in the mutation window as the time point of the mutation window;

[0053] Calculate the average value of the data in the mutation window as the window value of the mutation window.

[0054] Combined with the second aspect, in a possible implementation manner of the second aspect, the confirmation module is specifically configured to:

[0055] Generate a first window value sequence according to the window values of each mutation window;

[0056] Generate a second window value sequence corresponding to each normal current transformer according to the window values of the comparison windows of each normal current transformer;

[0057] Calculate the similarity between the first window value sequence and each second window value sequence respectively, and calculate the average value of each similarity to obtain the comprehensive similarity corresponding to this performance index;

[0058] Based on the comprehensive similarity, perform a secondary confirmation of the fault.

[0059] Combined with the second aspect, in a possible implementation manner of the second aspect, the confirmation module is specifically configured to:

[0060] If the comprehensive similarity is less than the preset similarity threshold, confirm that the water - immersed current transformer is faulty;

[0061] After the secondary confirmation of the fault, the confirmation module is further configured to:

[0062] If it is confirmed that the water - immersed current transformer is faulty, determine the severity of the fault according to the comprehensive similarity corresponding to each performance index.

[0063] Combined with the second aspect, in a possible implementation manner of the second aspect, before acquiring the real - time data of multiple performance indexes of the water - immersed current transformer, the acquisition module is further configured to: perform an importance analysis on the multiple performance indexes to obtain the importance weights of each performance index;

[0064] The evaluation module is specifically configured to:

[0065] According to Evaluate the health degree of the water - immersed current transformer;

[0066] where H is the health degree; n is the number of performance indexes; w i is the pre - calculated importance weight of the performance index; x i is the real - time data of the standardized performance index.

[0067] In combination with the second aspect, in a possible implementation manner of the second aspect, before obtaining the real-time data of multiple performance indicators of the water-immersed current transformer, the obtaining module is further configured to: perform a correlation analysis on the multiple performance indicators to obtain the correlation between each performance indicator;

[0068] After obtaining the real-time data of multiple performance indicators of the water-immersed current transformer, the obtaining module is further configured to:

[0069] Preprocess the real-time data of the multiple performance indicators according to the correlation; wherein, the preprocessing includes: correcting abnormal data and interpolating missing data.

[0070] In combination with the second aspect, in a possible implementation manner of the second aspect, after the secondary confirmation of the fault, the confirmation module is further configured to:

[0071] If it is confirmed that the water-immersed current transformer fails, determine the fault type of the water-immersed current transformer according to the real-time data of the multiple performance indicators and a pre-trained fault classification model.

[0072] In combination with the second aspect, in a possible implementation manner of the second aspect, the multiple performance indicators include at least one of the following:

[0073] Insulation resistance value, seal integrity parameter, excitation characteristic parameter, turns ratio, and ambient temperature.

[0074] The beneficial effects of the embodiments of the present invention compared with the prior art are:

[0075] According to the real-time data of multiple performance indicators of the water-immersed current transformer, the embodiments of the present invention comprehensively evaluate the health status of the water-immersed current transformer by using a health assessment model to accurately detect faults of the water-immersed current transformer. By combining multiple performance indicators, the state of the current transformer can be comprehensively understood, avoiding misjudgment or missed judgment caused by traditional methods relying only on a single parameter or a simple threshold judgment, and improving the accuracy of the health assessment of the water-immersed current transformer. After initially determining that the water-immersed current transformer fails, by obtaining the time series data of each performance indicator within a preset time period and dividing the time series data of each performance indicator through a time window, and according to the data mutation situation within each window, the fault is reconfirmed, which can further improve the accuracy of fault judgment and provide effective support for the maintenance of the power system. Description of the Drawings

[0076] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0077] Figure 1 It is a schematic flowchart of the implementation process of the fault detection method for the water - immersed current transformer provided by the embodiment of the present invention;

[0078] Figure 2 It is a schematic structural diagram of the fault detection device for the water - immersed current transformer provided by the embodiment of the present invention. Detailed implementation manners

[0079] In the following description, for the purpose of illustration rather than limitation, specific details such as specific system structures and technologies are presented to thoroughly understand the embodiments of the present invention. However, those skilled in the art should clearly understand that the present invention can also be implemented in other embodiments without these specific details. In other cases, the detailed descriptions of well - known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present invention.

[0080] To illustrate the technical solutions described in the present invention, the following will be described through specific embodiments.

[0081] Traditional current transformer detection methods have many limitations when facing complex situations after water immersion. On the one hand, some methods based on single - parameter or simple threshold judgment cannot comprehensively consider the changes in multiple performance indicators of the current transformer after water immersion. For example, only by monitoring whether the insulation resistance value is lower than a certain fixed threshold to judge whether the current transformer is faulty, other parameter changes and their mutual relationships may be ignored, resulting in misjudgment or missed judgment. On the other hand, traditional methods often lack the ability to deeply analyze and mine data, and cannot effectively use a large amount of historical operation data and relevant auxiliary information to establish a more accurate health assessment model.

[0082] In view of the deficiencies of the prior art, the embodiments of the present invention provide a fault detection method for water - immersed current transformers. Refer to Figure 1 as shown, including:

[0083] Step S101, analyze a large amount of historical operation data of current transformers and establish a health assessment model.

[0084] Exemplarily, a large amount of historical operation data of instrument transformers can be collected according to the installation and operation conditions of instrument transformers in the power system. The data sources can cover instrument transformers in different substations and with different service lives. The collected performance indicators include insulation resistance value, seal integrity parameter, excitation characteristic parameter, transformation ratio, and ambient temperature, etc. The seal integrity parameter can be the pressure change at the seal, gas leakage rate, etc., and the excitation characteristic parameter can be the excitation current, excitation voltage, and their derived parameters, etc. At the same time, auxiliary information such as measurement timestamp, instrument transformer model, service life, and installation location is recorded.

[0085] After collecting these historical operation data, advanced preprocessing is required. The preprocessing here includes: outlier processing. For example, for each parameter, calculate the mean and standard deviation, and mark the parameters that exceed the mean plus twice the standard deviation as outliers; missing value processing, that is, for the missing data, use appropriate methods for imputation. Exemplarily, the correlation between different performance indicators can be analyzed (for example, there is a negative correlation between ambient temperature and insulation resistance value, that is, as the ambient temperature increases, the insulation resistance value has a downward trend), and a correlation model between performance indicators is established to provide a basis for data imputation.

[0086] In this embodiment:

[0087] For the missing insulation resistance value, if it is randomly missing, the multiple imputation method can be used. For example, select 5 samples from the dataset that have similar characteristics to the sample where the missing value is located (such as the same model, similar service life, similar ambient temperature, etc.), and establish a regression model through the insulation resistance value data of these samples to predict the missing value. If it is continuously missing, the time series imputation method can be used. According to the change trend of the insulation resistance value over time, linear interpolation is performed using the non-missing data before and after.

[0088] For the missing seal integrity parameter, if the pressure change at the seal is missing, imputation can be performed according to the correlation model of other parameters related to seal integrity (such as gas leakage rate, ambient temperature, etc.); for the missing gas leakage rate, if there is historical data record, imputation can be performed according to the average value of the historical data, and if there is no historical data, the average gas leakage rate of the same type of instrument transformer can be referred to for imputation.

[0089] For the missing excitation characteristic parameter, taking the missing excitation current as an example, if there is a correlation with other excitation characteristic parameters (such as excitation voltage, phase difference, etc.), imputation can be performed according to the correlation model; if there is no obvious correlation, the average excitation current value of the same type of instrument transformer can be referred to for imputation. For the missing parameters such as excitation voltage and the phase difference between current and voltage, similar methods can be used for processing according to the actual situation.

[0090] For the missing turns ratio, interpolation can be performed according to the standard turns ratio of transformers of the same type. If there are other relevant parameters (such as physical quantities related to the turns ratio), interpolation can also be performed according to the correlation model.

[0091] Then, based on the historical operation data after the above preprocessing, importance analysis of the performance indicators is carried out to obtain the importance weights of each performance indicator, thereby establishing a health assessment model.

[0092] Importance analysis is to determine the weight of the influence of each performance indicator on the health of the transformer according to historical operation data and experience. In this embodiment, through the analysis of a large amount of historical fault data, it is found that the insulation resistance value has a greater impact on the health of the transformer, and its weight w 绝缘电阻 = 0.3; the weight of the seal integrity parameter w 密封完整性 = 0.2; the weight of the excitation characteristic parameter w 励磁持性 = 0.2; the weight of the turns ratio w 变比 = 0.1; the weight of the ambient temperature w 环境温度 = 0.2.

[0093] The health assessment model established in this embodiment is as follows:

[0094]

[0095] Among them, H is the health; n is the number of performance indicators; w i is the pre-calculated importance weight of the performance indicator; x i is the real-time data of the performance indicator after standardization.

[0096] Step S102, obtain the real-time data of multiple performance indicators of the waterlogged transformer.

[0097] In this embodiment, the multiple performance indicators include but are not limited to the above: insulation resistance value, seal integrity parameter, excitation characteristic parameter, turns ratio, and ambient temperature.

[0098] After obtaining the real-time data of multiple performance indicators of the waterlogged transformer, the real-time data can also be preprocessed through a correlation model or other means to improve the accuracy.

[0099] Step S103, evaluate the health of the waterlogged transformer according to the real-time data of multiple performance indicators through a preset health assessment model.

[0100] In this embodiment, substitute the real-time data of multiple performance indicators into the above-established health assessment model to obtain the health of the waterlogged transformer.

[0101] Step S104, determine whether the waterlogged transformer is faulty according to the health.

[0102] Here, when the health degree of the water - immersed current transformer is less than the preset health degree threshold, it is preliminarily determined that the water - immersed current transformer has a fault.

[0103] Among them, the health degree threshold can be determined in the following way:

[0104] First, set the standard values of the performance indicators of a normal current transformer as follows: the insulation resistance value is 800 MΩ, the seal integrity parameter (taking the pressure change at the seal as an example) is - 0.02 MPa, the excitation characteristic parameter value (taking the excitation current as an example) is 0.35 A, the turns ratio is 160, and the ambient temperature is 25 °C.

[0105] Standardize the data, that is, subtract the average value detected by multiple normal current transformers from the standard value and perform standardization processing to make the data between 0 - 1 for convenient calculation.

[0106] Standardization of the insulation resistance value:

[0107] Standardization of the seal integrity parameter, taking the pressure change at the seal as an example:

[0108] Standardization of the excitation characteristic parameter, taking the excitation current as an example,

[0109] Standardization of the turns ratio:

[0110] Standardization of the ambient temperature: Set the standard ambient temperature as 25 °C, so the standardized value is x 环境温度 = 0;

[0111] According to the health degree calculation formula Calculate the health degree score H = 0.22 of the normal current transformer, then the threshold can be set to 70% of 0.22, that is, 0.154. If the health degree of the water - immersed current transformer is less than 0.154, it is preliminarily determined that the water - immersed current transformer has a fault.

[0112] Step S105, if the water - immersed current transformer has a fault, extract the time - series data of each performance indicator within a preset time period, and perform window division on the time - series data of each performance indicator through a time window.

[0113] In this embodiment, the size of the preset time period is not specifically limited. For example, it can be the time - series data of each performance indicator within 1 hour. Similarly, the division of the time window can also be set according to requirements. For example, configure the window size ω = 5.

[0114] Step S106, according to the data mutation situation within each window, perform a secondary confirmation of the fault.

[0115] The data mutation situation means that the time series data changes significantly at a certain moment, which usually indicates a fault in the mutual inductor. In this embodiment, after initially determining the fault of the water-soaked mutual inductor, mutation analysis is performed on the time series data of each performance index to reconfirm the fault and improve the certainty of the judgment.

[0116] In the embodiment of the present invention, according to the real-time data of multiple performance indexes of the water-soaked mutual inductor, the health assessment model is used to comprehensively evaluate the health of the water-soaked mutual inductor, so as to accurately detect the fault of the water-soaked mutual inductor. By combining multiple performance indexes, the state of the mutual inductor can be comprehensively understood, avoiding the misjudgment or missed judgment caused by the traditional method relying only on a single parameter or a simple threshold judgment, and improving the accuracy of the health assessment of the water-soaked mutual inductor. After initially judging the fault of the water-soaked mutual inductor, by obtaining the time series data of each performance index within a preset time period and dividing the time series data of each performance index through a time window, and according to the data mutation situation in each window, the fault is reconfirmed, which can further improve the accuracy of the fault judgment and provide effective support for the maintenance of the power system.

[0117] In some embodiments, the above step S106 can be implemented in the following manner:

[0118] For the time series data of any one performance index, calculate the angular value between adjacent data in each window, and take the average value of each angular value in each window as the mutation value of each window;

[0119] Judge whether each window is a mutation window according to the mutation value of each window;

[0120] Determine the time points and window values of each mutation window;

[0121] According to the time points of each mutation window, select the window values of the corresponding comparison windows from the time series data of this performance index of multiple normal mutual inductors;

[0122] Reconfirm the fault according to the window values of each mutation window and each comparison window.

[0123] Taking the insulation resistance value as an example, calculate the angle between all adjacent data points in the window Where For the insulation resistance value data sequence [280, 300, 320, 340, 360] in a certain window, taking the 3rd data point as an example, Then: (radian).

[0124] Calculate the average value of all angles within the calculation window As the mutation value.

[0125] Configure the mutation threshold T θ = 0.8 (radians). When it is the case, determine that the window is a mutation window. In the time series data of each performance indicator, there may be multiple mutation windows.

[0126] For the window determined to be a possible mutation window, the mutation threshold can be increased. For example, adjust it to T θ = 0.9 (radians), recalculate the mutation value and judge and confirm again.

[0127] After that, the acquisition time of the first data in the mutation window can be used as the time point of the mutation window; calculate the average value of the data in the mutation window as the window value of the mutation window. According to the window values of each mutation window, generate the first window value sequence; according to the window values of the comparison windows of each normal current transformer, generate the second window value sequence corresponding to each normal current transformer; calculate the similarity between the first window value sequence and each second window value sequence respectively, and calculate the average value of each similarity to obtain the comprehensive similarity corresponding to this performance indicator; based on the comprehensive similarity, conduct a secondary confirmation of the fault.

[0128] Exemplarily, from the mutation window of the time series data of the insulation resistance value, extract the first window value sequence as According to the time stamp of the mutation window, extract the comparison window with the same time stamp from each comparison device group (other normally operating current transformers). For example, obtain the second window value sequence of a certain comparison device group as

[0129] This embodiment does not limit the method for calculating the similarity between the data sequences of the mutation device and the comparison device group. For example, it can be through the following methods:

[0130]

[0131] Finally, calculate the average value of the similarities between all comparison device groups and the mutation device As the comprehensive similarity of the insulation resistance value. When (The typical value of the similarity threshold T ρ = 0.6), then determine that the current transformer has a fault and output the number of the faulty device.

[0132] In some other embodiments, after the secondary confirmation of the current transformer fault, it can also be:

[0133] (1) If it is confirmed that the water - immersed current transformer has a fault, then determine the severity of the fault according to the comprehensive similarity corresponding to each performance indicator.

[0134] Here, for the water - immersed current transformer determined to be faulty, the fault score can be calculated based on the number of mutation windows, the mutation values of the mutation windows, and the comprehensive similarity in the faulty device (the mutation value is positively correlated with the fault score, and the comprehensive similarity is negatively correlated with the fault score). A higher fault score value indicates a more serious fault condition of the water - immersed current transformer. This means that the mutations existing in the device not only have a larger mutation degree (larger mutation value), but also have a lower comprehensive similarity compared with the normally operating devices.

[0135] (2) Determine the fault type of the water - immersed current transformer according to the real - time data of multiple performance indicators and the pre - trained fault classification model, as specifically described below.

[0136] By performing fault classification marking on the operation data of historical faulty devices, dividing the training set and the test set, selecting the gradient - boosting tree as the initial model, and conducting model training. For each round of training, a tree is trained for each fault type, and a total of K = 5 trees are trained. For each operation data sample, the output calculated through all the trees is represented as a score vector z=(z1, z2, z3, z4, z5), where each element represents the output score of the corresponding category. Convert the score vector into a probability distribution. For each category K, the probability of being predicted as this category is

[0137] For example, for a certain operation data sample, if the score vector is (0.5, 0.3, 0.1, 0.05, 0.05), the converted probability distribution is:

[0138]

[0139] Configure the probability threshold T p = 0.6, select the fault types whose preliminary classification probability values are greater than the probability threshold, and conduct fine - classification based on the metering fault model in turn. For the mutation window of the faulty device for which fine - classification is carried out, set its timestamp as T a = 10, then obtain the operation data of the faulty device at the timestamp [T a - 3T t , T a - 1T t , establish a metering fault model to fit the data in this time period, use the metering fault model to predict the operation data at the timestamp [T a - 1T t , T a + 1T t , and calculate the deviation between the predicted value and the actual value of the operation data. Obtain the probability value of the fine - classified fault type according to the deviation type.

[0140] Comprehensive probability value calculation and fault type ranking:

[0141] Combining the probability values of the preliminary classification and the detailed classification, for each detailed classification, multiply the probability value of the detailed classification within its preliminary classification by the probability value of the detailed classification to obtain the combined probability value of the detailed classification. Sum the combined probability values of all detailed classifications within the preliminary classification to obtain the combined probability value of the preliminary classification. Then, sort the fault types in sequence according to the combined probability value of the preliminary classification and the combined probability value of the detailed classification, and output the fault type sequence of the faulty device.

[0142] Model evaluation and verification: Use the test set to evaluate the trained model, and calculate relevant performance indicators such as accuracy, recall rate, precision, and F1 score. In this embodiment, a part of the data from the existing test dataset is selected as the test set, and the test set data is input into the trained model. By comparing the output result of the model with the true labels of the test set data, the accuracy, recall rate, precision, and F1 score are calculated. In this embodiment, a test set containing 100 samples is built, among which 60 samples have normal actual health status and 40 samples have faulty actual health status. Then, the model predicts that 55 samples have normal health status, among which 50 have normal actual health status; the model predicts that 45 samples have faulty health status, among which 35 have faulty actual health status. Conduct model evaluation:

[0143] Accuracy calculation:

[0144] Recall rate calculation:

[0145] Precision calculation:

[0146] F1 score calculation:

[0147] It can be seen that the model in this embodiment already has an accuracy of 80%, which basically meets the usage requirements. Then, the parameters can be adjusted, and the model can be adjusted according to the evaluation results. To improve the accuracy, the hyperparameters of the model can be considered for adjustment, such as the learning rate, regularization parameter, etc.; if the model complexity is high, the model structure can be simplified by reducing unnecessary layers or the number of neurons; if there is overfitting in the data, methods such as data augmentation and regularization can be used for processing; if the model converges slowly during training, early stopping of training or a more suitable optimization algorithm can be considered. By continuously adjusting and optimizing the model, the performance and accuracy of the model are improved, so that the model built in this embodiment can better adapt to the fault classification of the immersed current transformer.

[0148] In this embodiment, the fault type is determined by calculating the fault score and performing automated fault diagnosis. Finally, through model evaluation and verification, relevant performance indicators are calculated and the model is adjusted to better adapt to different application scenarios and data characteristics, improving the versatility and practicality of the model.

[0149] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The order of execution of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0150] Figure 2 FIG. 7 is a schematic structural diagram of a fault detection device 20 for a water - immersed current transformer provided by an embodiment of the present invention, including:

[0151] An acquisition module 21, configured to acquire real - time data of multiple performance indicators of the water - immersed current transformer.

[0152] An evaluation module 22, configured to evaluate the health of the water - immersed current transformer according to the real - time data of multiple performance indicators through a preset health evaluation model; and determine whether the water - immersed current transformer is faulty according to the health.

[0153] A confirmation module 23, configured to, if the water - immersed current transformer is faulty, extract the time - series data of each performance indicator within a preset time period, and perform window division on the time - series data of each performance indicator through a time window; and perform secondary confirmation of the fault according to the data mutation situation within each window.

[0154] As a possible implementation manner, the confirmation module 23 is specifically configured to:

[0155] For the time - series data of any one performance indicator, calculate the angular value between adjacent data within each window, and use the average value of each angular value within each window as the mutation value of each window;

[0156] Determine whether each window is a mutation window according to the mutation value of each window;

[0157] Determine the time points and window values of each mutation window;

[0158] According to the time points of each mutation window, select the window values of the corresponding comparison windows from the time - series data of this performance indicator of multiple normal current transformers;

[0159] Perform secondary confirmation of the fault according to the window values of each mutation window and each comparison window.

[0160] As a possible implementation manner, the confirmation module 23 is specifically configured to:

[0161] Take the acquisition time of the first data in the mutation window as the time point of the mutation window;

[0162] Calculate the average value of the data in the mutation window as the window value of the mutation window.

[0163] As a possible implementation, the confirmation module 23 is specifically configured to:

[0164] Generate a first window value sequence according to the window values of each mutation window;

[0165] Generate a second window value sequence corresponding to each normal current transformer according to the window values of the comparison windows of each normal current transformer;

[0166] Calculate the similarity between the first window value sequence and each second window value sequence respectively, and calculate the average value of each similarity to obtain the comprehensive similarity corresponding to this performance index;

[0167] Based on the comprehensive similarity, perform a secondary confirmation of the fault.

[0168] As a possible implementation, the confirmation module 23 is specifically configured to:

[0169] If the comprehensive similarity is less than the preset similarity threshold, confirm the fault of the water-soaked current transformer;

[0170] After performing the secondary confirmation of the fault, the confirmation module is further configured to:

[0171] If it is confirmed that the water-soaked current transformer is faulty, determine the severity of the fault according to the comprehensive similarity corresponding to each performance index.

[0172] As a possible implementation, before obtaining the real-time data of multiple performance indexes of the water-soaked current transformer, the obtaining module 21 is further configured to: perform importance analysis on multiple performance indexes to obtain the importance weights of each performance index;

[0173] The evaluation module 22 is specifically configured to:

[0174] According to Evaluate the health of the water-soaked current transformer;

[0175] where H is the health; n is the number of performance indexes; w i is the pre-calculated importance weight of the performance index; x i is the real-time data of the standardized performance index.

[0176] As a possible implementation, before obtaining the real-time data of multiple performance indexes of the water-soaked current transformer, the obtaining module 21 is further configured to: perform correlation analysis on multiple performance indexes to obtain the correlation between each performance index;

[0177] After obtaining the real-time data of multiple performance indicators of the waterlogged instrument transformer, the obtaining module 21 is further configured to:

[0178] Preprocess the real-time data of multiple performance indicators according to the correlation; wherein, the preprocessing includes: correcting abnormal data and interpolating missing data.

[0179] As a possible implementation manner, after the secondary confirmation of the fault, the confirmation module 23 is further configured to:

[0180] If it is confirmed that the waterlogged instrument transformer is faulty, determine the fault type of the waterlogged instrument transformer according to the real-time data of multiple performance indicators and a pre-trained fault classification model.

[0181] As a possible implementation manner, the multiple performance indicators include at least one of the following:

[0182] Insulation resistance value, seal integrity parameter, excitation characteristic parameter, transformation ratio, and ambient temperature.

[0183] In the embodiments of the present invention, according to the real-time data of multiple performance indicators of the waterlogged instrument transformer, the health assessment model is used to comprehensively evaluate the health of the waterlogged instrument transformer to accurately detect the faults of the waterlogged instrument transformer. Through the combination of multiple performance indicators, the state of the instrument transformer can be comprehensively understood, avoiding misjudgment or missed judgment caused by the traditional method relying only on a single parameter or a simple threshold judgment, and improving the accuracy of the health assessment of the waterlogged instrument transformer. After initially judging the fault of the waterlogged instrument transformer, by obtaining the time series data of each performance indicator within a preset time period, and dividing the time series data of each performance indicator through a time window, and according to the data mutation situation within each window, the fault is reconfirmed, which can further improve the accuracy of fault judgment and provide effective support for the maintenance of the power system.

[0184] In the above embodiments, the descriptions of the respective embodiments have their own emphases. For parts not detailed or recorded in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0185] Those of ordinary skill in the art can realize that the templates, units, and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or by a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0186] If the module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by a computer program instructing relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described embodiments of the fault detection method for each immersion current transformer can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0187] The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A fault detection method for a submerged transformer, characterized in that: include: Obtain real-time data of multiple performance indicators of submerged transformers; According to the real-time data of the multiple performance indicators, the health of the submerged transformer is evaluated by a preset health evaluation model; According to the health degree, determining whether the submerged mutual inductor is faulty; If the submerged transformer fails, extract the time series data of each performance indicator within a preset time period, and divide the time series data of each performance indicator into windows by time windows; The fault is reconfirmed based on the data mutation in each window.

2. The fault detection method for a submerged mutual inductor according to claim 1, characterized in that: The second confirmation of the fault according to the data mutation in each window includes: For the time series data of any performance indicator, calculate the angle value between adjacent data in each window, and take the average value of each angle value in each window as the mutation value of each window; Determine whether each window is a mutation window according to the mutation value of each window; Determine the time point and window value of each mutation window; According to the time point of each mutation window, the window value of the comparison window corresponding to each mutation window is selected from the time series data of the performance indicator of multiple normal transformers; The fault is reconfirmed based on the window values ​​of each mutation window and each comparison window.

3. The fault detection method for a submerged mutual inductor according to claim 2, characterized in that: Determining the time point and window value of each mutation window includes: The moment when the first data in the mutation window is obtained is taken as the time point of the mutation window; Calculate the average value of the data in the mutation window as the window value of the mutation window.

4. The fault detection method for a submerged mutual inductor according to claim 2, characterized in that: The second confirmation of the fault according to the window values ​​of each mutation window and each comparison window includes: Generate a first window value sequence according to the window values ​​of each mutation window; Generating a second window value sequence corresponding to each normal mutual inductor according to the window value of the comparison window of each normal mutual inductor; Calculating the similarities between the first window value sequence and each second window value sequence respectively, and calculating the average value of each similarity to obtain a comprehensive similarity corresponding to the performance indicator; Based on the comprehensive similarity, the fault is reconfirmed.

5. The fault detection method for a submerged mutual inductor according to claim 4, characterized in that: The second confirmation of the fault based on the comprehensive similarity includes: If the comprehensive similarity is less than a preset similarity threshold, it is confirmed that the submerged transformer is faulty; After the fault is confirmed twice, the method further includes: If the submerged transformer is confirmed to be faulty, the severity of the fault is determined based on the comprehensive similarities corresponding to various performance indicators.

6. The fault detection method for a submerged mutual inductor according to claim 1, characterized in that: Before acquiring the real-time data of the multiple performance indicators of the submerged transformer, the method further includes: performing importance analysis on the multiple performance indicators to obtain the importance weight of each performance indicator; The step of evaluating the health of the submerged transformer according to the real-time data of the multiple performance indicators by using a preset health evaluation model includes: according to Assessing the health of the submerged transformer; Among them, H is the health level; n is the number of performance indicators; w i is the importance weight of the pre-calculated performance indicator; x i It is the real-time data of the standardized performance indicators.

7. The fault detection method for a submerged mutual inductor according to claim 1, characterized in that: Before acquiring the real-time data of the multiple performance indicators of the submerged transformer, the method further includes: performing correlation analysis on the multiple performance indicators to obtain the correlation between the various performance indicators; After obtaining the real-time data of multiple performance indicators of the submerged transformer, the method further includes: According to the correlation, the real-time data of the multiple performance indicators are preprocessed; wherein the preprocessing includes: correcting abnormal data and interpolating missing data.

8. The fault detection method for a submerged mutual inductor according to claim 1, characterized in that: After the fault is confirmed again according to the method, the method further includes: If the submerged transformer is confirmed to be faulty, the fault type of the submerged transformer is determined based on the real-time data of the multiple performance indicators and a pre-trained fault classification model.

9. The fault detection method for a submerged mutual inductor according to any one of claims 1 to 8, characterized in that: The multiple performance indicators include at least one of the following: Insulation resistance value, sealing integrity parameters, excitation characteristic parameters, transformation ratio and ambient temperature.

10. A fault detection device for a submerged transformer, characterized in that: include: An acquisition module, used to acquire real-time data of multiple performance indicators of the submerged transformer; An evaluation module, configured to evaluate the health of the submerged transformer according to the real-time data of the plurality of performance indicators through a preset health evaluation model; According to the health degree, determining whether the submerged mutual inductor is faulty; The confirmation module is used to extract the time series data of each performance indicator within a preset time period if the submerged transformer fails, and divide the time series data of each performance indicator into windows by time windows; and perform a secondary confirmation of the fault according to the data mutation in each window.

Images