A method, system, device and medium for judging a transformer fault type

Abstract

The application discloses a transformer fault type judgment method, system, device and medium, when a vibration sub-signal data set is received, a multi-scale operation is performed on the vibration sub-signal data set to generate vibration sub-signal sample data, the vibration sub-signal sample data is input into a preset initial random forest diagnosis model for training to generate a corresponding target random forest diagnosis model, in response to a received fault judgment request, a corresponding transformer to be judged is selected and vibration sub-signal data is acquired, a multi-scale operation is performed on the vibration sub-signal data to output multi-scale grid entropy values, the multi-scale grid entropy values are input into the target random forest diagnosis model to generate target fault label category data, the target fault label category data is compared with preset actual fault label category data, and the fault type of the transformer to be judged is determined according to a comparison result; the advantages of the multi-scale grid entropy and the random forest are combined to realize accurate transformer online fault detection.

Description

Technical Field

[0001] This invention relates to the field of transformer fault detection technology, and in particular to a method, system, device and medium for determining transformer fault types. Background Technology

[0002] Transformers, as one of the most important pieces of equipment in a power system, operate under various complex electrical and mechanical conditions. Therefore, effective detection of potential transformer faults is of significant engineering value for maintaining the efficient operation of the power grid.

[0003] Currently, a new transformer fault detection method overcomes the shortcomings of traditional offline diagnostic methods. It is convenient, safe, and reliable, meeting power supply requirements, and has become one of the main detection methods for power transformer faults. Generally, online transformer detection methods can be categorized into two types: dissolved gas analysis (DGA) and vibration analysis. DGA primarily relies on the detection of dissolved gas components in transformer oil, a diagnostic technique for faults caused by changes in insulating oil characteristics. However, the diagnostic effectiveness of DGA is significantly affected by the concentration of dissolved gases, leading to certain limitations in its application. Therefore, inspectors typically use vibration signals as an important indicator for assessing the health status of transformers, and vibration-based transformer fault diagnosis has been widely applied.

[0004] When a fault occurs inside a transformer, its vibration signal often fluctuates over a wide range of scales. Conventional time-domain and frequency-domain analysis methods are difficult to capture the sensitive fault characteristics of local and global faults at multiple time scales. Fault detection methods based on signal processing, such as wavelet analysis and empirical mode decomposition, are computationally too complex and difficult for on-site monitoring personnel to understand and analyze. At the same time, the analysis results are easily affected by the frequency of environmental noise, which often reduces the performance of the detection model in the detection task. As a result, it is impossible to accurately determine the type of transformer fault. Summary of the Invention

[0005] This invention provides a method, system, device, and medium for determining transformer fault types. It addresses the technical problem that when a fault occurs inside a transformer, its vibration signal often fluctuates across a wide range of scales; conventional time-domain and frequency-domain analysis methods struggle to capture sensitive fault characteristics at multiple time scales; and signal processing-based fault detection methods, such as wavelet analysis and empirical mode decomposition, are computationally too complex and difficult for on-site monitoring personnel to understand and analyze. Furthermore, the analysis results are easily affected by environmental noise frequencies, often reducing the performance of the detection model in detection tasks. These factors contribute to the inability to accurately determine the type of transformer fault.

[0006] The first aspect of this invention provides a method for determining transformer fault types, comprising:

[0007] When a vibration sub-signal dataset is received, multi-scale operations are performed on the vibration sub-signal dataset to generate corresponding vibration sub-signal sample data.

[0008] The vibration sub-signal sample data is used as input to train a preset initial random forest diagnostic model to generate a corresponding target random forest diagnostic model;

[0009] In response to a received fault diagnosis request, the transformer to be diagnosed corresponding to the fault diagnosis request is selected and the resonator signal data is acquired;

[0010] Perform multi-scale operations on the vibrator signal data and output the corresponding multi-scale grid entropy value;

[0011] The multi-scale grid entropy value is input into the target random forest diagnostic model to generate corresponding target fault label category data;

[0012] The target fault label category data is compared with the preset actual fault label category data, and the fault type of the transformer to be judged is determined based on the comparison result.

[0013] Optionally, the step of using the vibration sub-signal sample data as input to train a preset initial random forest diagnostic model to generate a corresponding target random forest diagnostic model includes:

[0014] The vibration sub-signal sample data is input into the initial random forest diagnostic model to generate corresponding training fault label category data;

[0015] Calculate the training loss value between the training fault label category data and the associated standard fault label category data;

[0016] Compare the training loss value with a preset training standard threshold;

[0017] If the training loss value is less than or equal to the training standard threshold, then training is stopped and a target random forest diagnostic model is generated.

[0018] Optionally, it also includes:

[0019] If the training loss value is greater than the training standard threshold, the parameters of the initial random forest diagnostic model are adjusted according to the preset gradient, and the process jumps to the step of inputting the vibration sub-signal sample data into the initial random forest diagnostic model to generate the corresponding training fault label category data.

[0020] Optionally, the step of comparing the target fault label category data with preset actual fault label category data and determining the fault type of the transformer to be judged based on the comparison result includes:

[0021] Compare the target fault label category data with the preset actual fault label category data;

[0022] If the target fault label category data is consistent with the actual fault label category data, then the fault type associated with the actual fault label category data is determined for the transformer to be judged;

[0023] If the target fault label category data is inconsistent with the actual fault label category data, it is determined that the transformer to be judged has not experienced the fault type associated with the actual fault label category data, and the process jumps to the step of using the multi-scale grid entropy value to input the target random forest diagnostic model to generate the corresponding target fault label category data.

[0024] A second aspect of the present invention provides a transformer fault type determination system, comprising:

[0025] The vibration sub-signal sample data acquisition module is used to perform multi-scale operations on the vibration sub-signal dataset when a vibration sub-signal dataset is received, and generate corresponding vibration sub-signal sample data.

[0026] The target random forest diagnostic model generation module is used to train a preset initial random forest diagnostic model by inputting the vibration sub-signal sample data, and generate a corresponding target random forest diagnostic model.

[0027] The response module is used to respond to the received fault judgment request, select the transformer to be judged corresponding to the fault judgment request, and obtain the resonator signal data;

[0028] The multi-scale grid entropy value acquisition module is used to perform multi-scale operations on the vibrating sub-signal data and output the corresponding multi-scale grid entropy value.

[0029] The target fault label category data acquisition module is used to input the multi-scale grid entropy value into the target random forest diagnostic model to generate corresponding target fault label category data.

[0030] The comparison module is used to compare the target fault label category data with the preset actual fault label category data, and determine the fault type of the transformer to be judged based on the comparison result.

[0031] Optionally, the target random forest diagnostic model generation module includes:

[0032] The training label category data acquisition submodule is used to input the multi-scale grid entropy value into the initial random forest diagnostic model to generate the corresponding training label category data.

[0033] The training loss value acquisition submodule is used to calculate the training loss value between the training label category data and the associated standard label category data.

[0034] The comparison submodule is used to compare the training loss value with a preset training standard threshold.

[0035] The first data processing submodule is used to stop training and generate a target random forest diagnostic model if the training loss value is less than or equal to the training standard threshold.

[0036] Optionally, the target random forest diagnostic model generation module further includes:

[0037] The second data processing submodule is used to adjust the parameters of the initial random forest diagnostic model according to a preset gradient if the training loss value is greater than the training standard threshold, and then jump to the step of inputting the vibration sub-signal sample data into the initial random forest diagnostic model to generate corresponding training fault label category data.

[0038] Optionally, the comparison module includes:

[0039] The fault label category data comparison submodule is used to compare the target fault label category data with the preset actual fault label category data;

[0040] The first determination submodule is used to determine the fault type associated with the actual fault label category data of the transformer to be determined if the target fault label category data is consistent with the actual fault label category data.

[0041] The second determination submodule is used to determine that the transformer to be judged has not experienced the fault type associated with the actual fault label category data if the target fault label category data is inconsistent with the actual fault label category data, and then jump to execute the step of using the multi-scale grid entropy value to input the target random forest diagnostic model to generate the corresponding target fault label category data.

[0042] A third aspect of the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the transformer fault type determination method as described in any of the preceding claims.

[0043] The fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed, it implements the method for determining the type of transformer fault as described in any of the preceding claims.

[0044] As can be seen from the above technical solutions, the present invention has the following advantages:

[0045] Upon receiving a vibration sub-signal dataset, multi-scale operations are performed on the dataset to generate corresponding vibration sub-signal sample data. This sample data is then used to train a pre-defined initial random forest diagnostic model, generating a corresponding target random forest diagnostic model. In response to a received fault diagnosis request, the transformer to be diagnosed is selected, and vibration sub-signal data is acquired. Multi-scale operations are performed on this data, outputting corresponding multi-scale grid entropy values. These entropy values ​​are then input into the target random forest diagnostic model to generate corresponding target fault label category data. This target fault label category data is compared with pre-defined actual fault label category data, and the fault to be diagnosed is determined based on the comparison results. The text discusses various aspects of transformer fault detection, including the challenges of capturing sensitive local and global fault features across multiple time scales when internal transformer faults occur. It also addresses the computational complexity of signal processing-based fault detection methods like wavelet analysis and empirical mode decomposition, which are difficult for on-site monitoring personnel to understand and analyze. Furthermore, the analysis results are susceptible to environmental noise, often reducing the performance of the detection model. This leads to the inability to accurately determine the type of transformer fault. The text emphasizes the need for intelligent fault detection, utilizing multi-scale grid entropy to extract deep, highly expressive, and separable fault features. Random forests are easy to implement, have fast training speeds, and can efficiently identify multiple fault states. Therefore, combining the advantages of multi-scale grid entropy and random forests enables accurate online transformer fault detection. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart illustrating the steps of a method for determining transformer fault types according to Embodiment 1 of the present invention.

[0048] Figure 2 This is a flowchart illustrating the steps of a method for determining transformer fault types according to Embodiment 2 of the present invention.

[0049] Figure 3 This is a schematic diagram of the multi-scale grid entropy grid division provided in Embodiment 2 of the present invention;

[0050] Figure 4This is a structural block diagram of a transformer fault type determination system provided in Embodiment 3 of the present invention. Detailed Implementation

[0051] This invention provides a method, system, device, and medium for determining transformer fault types. It addresses the problem that when a transformer experiences internal faults, its vibration signals often fluctuate across a wide range of scales; conventional time-domain and frequency-domain analysis methods struggle to capture sensitive fault characteristics at multiple time scales; and signal processing-based fault detection methods, such as wavelet analysis and empirical mode decomposition, are computationally too complex and difficult for on-site monitoring personnel to understand and analyze. Furthermore, the analysis results are easily affected by environmental noise frequencies, often reducing the performance of the detection model in detection tasks. These factors contribute to the inability to accurately determine the type of transformer fault.

[0052] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0053] Please see Figure 1 , Figure 1 The flowchart illustrates the steps of a method for determining transformer fault types according to Embodiment 1 of the present invention.

[0054] This invention provides a method for determining transformer fault types, comprising:

[0055] Step 101: When the vibrating sub-signal dataset is received, perform multi-scale operations on the vibrating sub-signal dataset to generate corresponding vibrating sub-signal sample data.

[0056] A vibrating sub-signal dataset refers to a dataset composed of vibrating sub-signals from different common fault types of transformers in a substation, which are then integrated to form a single vibrating sub-signal dataset.

[0057] Multi-scale operation refers to processing the vibrator signal dataset using multi-scale grid entropy.

[0058] Vibrator signal sample data refers to the sample data used as input for training a preset initial random forest diagnostic model.

[0059] In this embodiment of the invention, when a vibration sub-signal dataset composed of vibration signals of different common fault types of transformers in a substation is received, the vibration sub-signal dataset is processed by multi-scale grid entropy to generate corresponding vibration sub-signal sample data.

[0060] Step 102: Use the vibrational sub-signal sample data to train the preset initial random forest diagnostic model and generate the corresponding target random forest diagnostic model.

[0061] The initial random forest diagnostic model refers to a diagnostic module built based on the random forest strategy. Random forest is a classifier that uses multiple trees to train and predict samples.

[0062] The target random forest diagnostic model refers to the target random forest diagnostic model generated by training the initial random forest diagnostic model.

[0063] In this embodiment of the invention, the initial random forest diagnostic model is trained based on the obtained vibrating sub-signal sample data to generate the corresponding target random forest diagnostic model.

[0064] Step 103: In response to the received fault judgment request, select the transformer to be judged corresponding to the fault judgment request and obtain the resonator signal data.

[0065] A fault diagnosis request refers to a fault diagnosis request sent by maintenance personnel to determine the type of fault that has occurred in a transformer within a substation.

[0066] Vibration sub-signal data refers to the sub-signal data obtained by preprocessing the vibration signals of the transformer under different states.

[0067] In this embodiment of the invention, in response to a fault judgment request sent by maintenance personnel regarding the type of fault occurring in a transformer within a substation, the fault judgment request is read, the corresponding transformer to be judged is determined, and the vibration signals of the transformer to be judged under different states are obtained.

[0068] Step 104: Perform multi-scale operations on the vibrator signal data and output the corresponding multi-scale grid entropy values.

[0069] Multi-scale grid entropy refers to the entropy value generated by performing multi-scale operations on the vibrator signal data using multi-scale grid entropy.

[0070] In this embodiment of the invention, the vibrator signal data is subjected to multi-scale operations by multi-scale grid entropy, and the corresponding multi-scale grid entropy value is output.

[0071] Step 105: Input the multi-scale grid entropy value into the target random forest diagnostic model to generate the corresponding target fault label category data.

[0072] Target fault label category data refers to the fault label category data obtained by analyzing the data obtained by inputting multi-scale grid entropy values ​​into the target random forest diagnostic model.

[0073] In this embodiment of the invention, multi-scale grid entropy values ​​are input into the target random forest diagnostic model, and the corresponding fault label category data is output.

[0074] Step 106: Compare the target fault label category data with the preset actual fault label category data, and determine the fault type of the transformer to be judged based on the comparison results.

[0075] Actual fault label category data refers to the actual fault type that occurred in the transformer.

[0076] In this embodiment of the invention, the fault label category data output by the target random forest diagnostic model is compared with the actual fault type of the transformer to determine whether they are consistent. If they are consistent, the fault type associated with the actual fault label category data of the transformer to be judged is determined. If they are inconsistent, the fault type associated with the actual fault label category data of the transformer to be judged is determined and the process jumps to the step of using multi-scale grid entropy value input to the target random forest diagnostic model to generate the corresponding target fault label category data.

[0077] In this embodiment of the invention, when a vibrating sub-signal dataset is received, multi-scale operations are performed on the vibrating sub-signal dataset to generate corresponding vibrating sub-signal sample data. The vibrating sub-signal sample data is then used to train a preset initial random forest diagnostic model to generate a corresponding target random forest diagnostic model. In response to a received fault diagnosis request, the transformer to be diagnosed corresponding to the fault diagnosis request is selected, and vibrating sub-signal data is acquired. Multi-scale operations are performed on the vibrating sub-signal data to output corresponding multi-scale grid entropy values. These multi-scale grid entropy values ​​are then input into the target random forest diagnostic model to generate corresponding target fault label category data. The target fault label category data is compared with preset actual fault label category data, and the fault label category is determined based on the comparison results. The problem lies in accurately determining the type of transformer fault. It addresses the issue that when a fault occurs inside a transformer, its vibration signals often fluctuate across a wide range of scales. Conventional time-domain and frequency-domain analysis methods struggle to capture sensitive local and global fault features across multiple time scales. Signal processing-based fault detection methods, such as wavelet analysis and empirical mode decomposition, are computationally too complex and difficult for on-site monitoring personnel to understand and analyze. Furthermore, the analysis results are easily affected by environmental noise frequencies, often reducing the performance of the detection model. This leads to the technical problem of inaccurately determining the type of transformer fault. To achieve intelligent fault detection, multi-scale grid entropy can extract deep, highly expressive, and separable fault features. Random forests are easy to implement, have fast training speeds, and can efficiently identify multiple fault states. Therefore, combining the advantages of multi-scale grid entropy and random forests can achieve accurate online transformer fault detection.

[0078] Please see Figure 2 , Figure 2 The flowchart illustrates the steps of a method for determining transformer fault types according to Embodiment 2 of the present invention.

[0079] This invention provides a method for determining transformer fault types, comprising:

[0080] Step 201: When the vibrating sub-signal dataset is received, perform multi-scale operations on the vibrating sub-signal dataset to generate corresponding vibrating sub-signal sample data.

[0081] In this embodiment of the invention, the specific implementation process of step 201 is similar to that of step 101, and will not be repeated here.

[0082] Step 202: Use the vibrational sub-signal sample data to train the preset initial random forest diagnostic model and generate the corresponding target random forest diagnostic model.

[0083] In this embodiment of the invention, the specific implementation process of step 202 is similar to that of step 102, and will not be repeated here.

[0084] Step 203: In response to the received fault judgment request, select the transformer to be judged corresponding to the fault judgment request and obtain the resonator signal data.

[0085] In this embodiment of the invention, the specific implementation process of step 203 is similar to that of step 103, and will not be repeated here.

[0086] In one example of the present invention, the acquired vibration sub-signal data refers to acquiring the vibration signals of the transformer to be judged under different fault types, such as inter-turn short circuit, core loosening, and winding deformation; and labeling the corresponding signals as {F1, F2, F3} in sequence, and then dividing the vibration signals of different fault types into sub-signals with a length of 1024 points without overlap.

[0087] Step 204: Perform multi-scale operations on the vibrator signal data and output the corresponding multi-scale grid entropy values.

[0088] In this embodiment of the invention, multi-scale grid entropy is used to analyze all types of vibrator signal data to obtain multi-scale grid entropy values ​​that can highly express fault characteristics.

[0089] In one example of the present invention, such as Figure 3 As shown, the vibration sub-signal data is normalized; the normalized signal is then processed at multiple scales to obtain sub-signals at multiple scales, given the required scale; a two-dimensional planar plot is generated for each sub-signal; given an n, the two-dimensional planar plot is divided into an n*n grid; the number of points in each grid is counted, and the grid entropy of each sub-signal is calculated. Finally, the obtained grid entropies are concatenated to form the final multi-scale grid entropy value.

[0090] The specific steps for calculating the entropy of a multi-scale grid are as follows:

[0091] Suppose we have a time series x = {x(i), i = 1, 2, ..., N} of length N. Normalize the series:

[0092]

[0093] In the formula, min(x) represents the minimum amplitude in the signal, max(x) represents the maximum amplitude in the signal, and x represents a time series.

[0094] The corresponding subsequence x is obtained by averaging the s data points within the non-overlapping window. s . subsequencex s Each element is defined as:

[0095]

[0096] in, denoted as floor function, s represents the total number of data points within the non-overlapping window, j represents the j-th point of sequence x, i represents the i-th point of sequence x, x(i) represents the normalized time series, and N represents the length of the time series.

[0097] For each subsequence x s Construct a point in a Cartesian coordinate system using adjacent elements, whose coordinates are (x, y, y). s (k),x s (k+1)), where k represents x s The k-th point in the middle. This forms a two-dimensional square plane.

[0098] Divide the two-dimensional plane into n×n equal-sized square grids using n horizontal and vertical lines. Calculate the proportion p(l) of the number of points in each grid relative to the total number of points in the plane.

[0099]

[0100] Where p(l) represents the proportion of points in each grid to the total number of points in the plane, l represents the l-th grid, and N l This represents the number of points in the l-th grid.

[0101] According to the Shannon entropy definition, the grid entropy GE at the s-scale is... s Its formula is:

[0102]

[0103] In the formula, GE s Represents the grid entropy value for a single level;

[0104] Given a time scale s max Finally, the multi-scale grid entropy values ​​are obtained:

[0105] GE(x,n,s max )={GE s},s=1,2,…,s max

[0106] In the formula, GE(x,n,s max ) represents the multi-level, multi-scale grid entropy value.

[0107] It is worth mentioning that the parameter settings for multi-scale grid entropy are shown in Table 1. To better understand the grid partitioning, a schematic diagram of the multi-scale grid entropy grid partitioning is shown below. Figure 4 As shown.

[0108] Table 1 Multiscale grid entropy parameter settings

[0109] name Value Number of grid lines n 10 <![CDATA[Maximum time scale s max > 20

[0110] Step 205: Input the multi-scale grid entropy value into the target random forest diagnostic model to generate the corresponding target fault label category data.

[0111] Furthermore, step 205 may include the following sub-steps:

[0112] S11. Input the vibrator signal sample data into the initial random forest diagnostic model to generate the corresponding training fault label category data.

[0113] Fault label category data refers to the label category data generated by inputting vibrator signal sample data into the initial random forest diagnostic model, which is used to determine whether the initial random forest diagnostic model has been trained successfully.

[0114] In this embodiment of the invention, the generated vibrator signal sample data is input into the initial random forest diagnostic model for model training, and corresponding training fault label category data is generated.

[0115] S12. Calculate the training loss value between the training fault label category data and the associated standard fault label category data.

[0116] Standard fault label category data refers to the standard fault label category data corresponding to the vibrator signal sample data.

[0117] The training loss value refers to the loss value between the training fault label category data and the standard fault label category data.

[0118] In this embodiment of the invention, the training loss value between the training fault label category data and the associated standard fault label category data is calculated.

[0119] S13. Compare the training loss value with the preset training standard threshold.

[0120] The training standard threshold refers to the threshold used to determine whether the initial random forest diagnostic model has been successfully trained.

[0121] In this embodiment of the invention, the training loss value is compared with a preset training standard threshold.

[0122] S14. If the training loss value is less than or equal to the training standard threshold, stop training and generate the target random forest diagnostic model.

[0123] In this embodiment of the invention, if the training loss value is less than or equal to the training standard threshold, training is stopped and a target random forest diagnostic model is generated.

[0124] Furthermore, step 205 may also include the following sub-steps.

[0125] S15. If the training loss value is greater than the training standard threshold, adjust the parameters of the initial random forest diagnostic model according to the preset gradient, and jump to the step of inputting the vibrating sub-signal sample data into the initial random forest diagnostic model to generate the corresponding training fault label category data.

[0126] In this embodiment of the invention, if the training loss value is greater than the training standard threshold, the parameters of the initial random forest diagnostic model are adjusted according to the preset gradient, and the process jumps to the step of inputting the vibrating sub-signal sample data into the initial random forest diagnostic model to generate the corresponding training fault label category data.

[0127] In one example of the present invention, when inputting the vibrating sub-signal sample data into the initial random forest diagnostic model for model training, the vibrating sub-signal sample data is divided into a training set and a validation set according to a certain ratio; wherein, 70% of the samples are used as the training set and the remaining 30% of the samples are used as the test set.

[0128] A target random forest diagnostic model was constructed using the training set, and the model was trained. The random forest contained 300 base estimators and had a maximum depth of 12.

[0129] It is worth mentioning that the parameters in the random forest are determined: the number of base estimators and the maximum depth. A target random forest diagnostic model is generated by training using the Gini index. The gain Z of the target random forest diagnostic model is used as the evaluation criterion for model training. The expressions for the Gini index and the gain Z are:

[0130]

[0131]

[0132] In the formula, Z represents the gain of the target random forest diagnostic model, c represents the fault category, p(i|t) represents the probability that sample i belongs to the t-th category, Gini(parent) is the Gini index metric given the parent node, M is the total number of records on the parent node, and M(v j ) for child node v j The number of associated records, Gini(v j ) represents node v j The Gini index.

[0133] In another example of this invention, samples from the test set are input into a trained target random forest diagnostic model to verify its diagnostic performance. The label categories provided by the random forest's decision are compared with the actual labels of the transformer (if the actual fault label of the transformer is F1, and the label output by the target random forest diagnostic model is also F1, it indicates accurate detection; if the target random forest diagnostic model outputs other fault labels, it indicates detection is successful), and the detection accuracy is calculated to complete the transformer fault detection.

[0134] Calculate the accuracy using the following formula:

[0135]

[0136] Where, N 正 N represents the number of correctly predicted samples. 总 It represents the total number of samples.

[0137] The diagnostic results are shown in Table 2. The method proposed in this invention achieved an accuracy rate of 98.25%, which is superior to commonly used fault characteristic characterization methods based on RMS values ​​and kurtosis. The comparative results demonstrate the advanced nature of the transformer fault diagnosis method provided by this invention.

[0138] Table 2 Comparison of Diagnostic Results

[0139] Fault Characterization Method Accuracy (ACC) Valid Value (RMS) 92.52％ Kurtosis 90.76％ Multiscale grid entropy 98.25％

[0140] Step 206: Compare the target fault label category data with the preset actual fault label category data.

[0141] Actual fault label category data refers to the actual fault category labels of the transformer.

[0142] In this embodiment of the invention, the target fault label category data is compared with the preset actual fault label category data.

[0143] Step 207: If the target fault label category data is consistent with the actual fault label category data, then determine the fault type associated with the actual fault label category data of the transformer to be judged.

[0144] In this embodiment of the invention, if the target fault label category data is consistent with the actual fault label category data, it is determined that the target fault label category data output by the target random forest diagnostic model is accurate, and the fault type associated with the actual fault label category data of the transformer to be determined is determined.

[0145] Step 208: If the target fault label category data is inconsistent with the actual fault label category data, then determine that the transformer to be judged has not experienced a fault type associated with the actual fault label category data and jump to the step of using multi-scale grid entropy value input to the target random forest diagnostic model to generate the corresponding target fault label category data.

[0146] In this embodiment of the invention, if the target fault label category data is inconsistent with the actual fault label category data, it is determined that the target fault label category data output by the target random forest diagnostic model is inaccurate, and the process jumps to the step of using multi-scale grid entropy values ​​to input into the target random forest diagnostic model to generate the corresponding target fault label category data.

[0147] In this embodiment of the invention, when a vibrating sub-signal dataset is received, multi-scale operations are performed on the vibrating sub-signal dataset to generate corresponding vibrating sub-signal sample data. The vibrating sub-signal sample data is then used to train a preset initial random forest diagnostic model to generate a corresponding target random forest diagnostic model. In response to a received fault diagnosis request, the transformer to be diagnosed corresponding to the fault diagnosis request is selected, and vibrating sub-signal data is acquired. Multi-scale operations are performed on the vibrating sub-signal data to output corresponding multi-scale grid entropy values. These multi-scale grid entropy values ​​are then input into the target random forest diagnostic model to generate corresponding target fault label category data. The target fault label category data is compared with preset actual fault label category data, and the fault label category is determined based on the comparison results. The problem lies in accurately determining the type of transformer fault. It addresses the issue that when a fault occurs inside a transformer, its vibration signals often fluctuate across a wide range of scales. Conventional time-domain and frequency-domain analysis methods struggle to capture sensitive local and global fault features across multiple time scales. Signal processing-based fault detection methods, such as wavelet analysis and empirical mode decomposition, are computationally too complex and difficult for on-site monitoring personnel to understand and analyze. Furthermore, the analysis results are easily affected by environmental noise frequencies, often reducing the performance of the detection model. This leads to the technical problem of inaccurately determining the type of transformer fault. To achieve intelligent fault detection, multi-scale grid entropy can extract deep, highly expressive, and separable fault features. Random forests are easy to implement, have fast training speeds, and can efficiently identify multiple fault states. Therefore, combining the advantages of multi-scale grid entropy and random forests can achieve accurate online transformer fault detection.

[0148] Please see Figure 4 , Figure 4 This is a structural block diagram of a transformer fault type determination system provided in Embodiment 3 of the present invention.

[0149] This invention provides a transformer fault type determination system, comprising:

[0150] The vibration sub-signal sample data acquisition module 301 is used to perform multi-scale operations on the vibration sub-signal dataset when a vibration sub-signal dataset is received, and generate corresponding vibration sub-signal sample data.

[0151] The target random forest diagnostic model generation module 302 is used to train the preset initial random forest diagnostic model by inputting vibration sub-signal sample data, and generate the corresponding target random forest diagnostic model.

[0152] The response module 303 is used to respond to the received fault judgment request, select the transformer to be judged corresponding to the fault judgment request, and obtain the resonator signal data;

[0153] The multi-scale grid entropy acquisition module 304 is used to perform multi-scale operations on the vibrator signal data and output the corresponding multi-scale grid entropy value.

[0154] The target fault label category data acquisition module 305 is used to input multi-scale grid entropy values ​​into the target random forest diagnostic model to generate corresponding target fault label category data.

[0155] The comparison module 306 is used to compare the target fault label category data with the preset actual fault label category data, and determine the fault type of the transformer to be judged based on the comparison results.

[0156] Furthermore, the target random forest diagnostic model generation module 302 includes:

[0157] The training label category data acquisition submodule is used to input the multi-scale grid entropy value into the initial random forest diagnostic model and generate the corresponding training label category data.

[0158] The training loss value acquisition submodule is used to calculate the training loss value between the training label category data and the associated standard label category data;

[0159] The comparison submodule is used to compare the training loss value with the preset training standard threshold.

[0160] The first data processing submodule is used to stop training and generate a target random forest diagnostic model if the training loss value is less than or equal to the training standard threshold.

[0161] Furthermore, the target random forest diagnostic model generation module 302 also includes:

[0162] The second data processing submodule is used to adjust the parameters of the initial random forest diagnostic model according to the preset gradient if the training loss value is greater than the training standard threshold, and then jump to the step of inputting the vibration sub-signal sample data into the initial random forest diagnostic model to generate the corresponding training fault label category data.

[0163] Furthermore, the comparison module 306 includes:

[0164] The fault label category data comparison submodule is used to compare the target fault label category data with the preset actual fault label category data;

[0165] The first determination submodule is used to determine the fault type associated with the actual fault label category data of the transformer to be judged if the target fault label category data is consistent with the actual fault label category data.

[0166] The second determination submodule is used to determine the fault type of the transformer to be judged that is not associated with the actual fault label category data if the target fault label category data is inconsistent with the actual fault label category data, and then jump to the step of using multi-scale grid entropy value input to target random forest diagnostic model to generate corresponding target fault label category data.

[0167] In this embodiment of the invention, when a vibrating sub-signal dataset is received, multi-scale operations are performed on the vibrating sub-signal dataset to generate corresponding vibrating sub-signal sample data. The vibrating sub-signal sample data is then used to train a preset initial random forest diagnostic model to generate a corresponding target random forest diagnostic model. In response to a received fault diagnosis request, the transformer to be diagnosed corresponding to the fault diagnosis request is selected, and vibrating sub-signal data is acquired. Multi-scale operations are performed on the vibrating sub-signal data to output corresponding multi-scale grid entropy values. These multi-scale grid entropy values ​​are then input into the target random forest diagnostic model to generate corresponding target fault label category data. The target fault label category data is compared with preset actual fault label category data, and the fault label category is determined based on the comparison results. The problem lies in accurately determining the type of transformer fault. It addresses the issue that when a fault occurs inside a transformer, its vibration signals often fluctuate across a wide range of scales. Conventional time-domain and frequency-domain analysis methods struggle to capture sensitive local and global fault features across multiple time scales. Signal processing-based fault detection methods, such as wavelet analysis and empirical mode decomposition, are computationally too complex and difficult for on-site monitoring personnel to understand and analyze. Furthermore, the analysis results are easily affected by environmental noise frequencies, often reducing the performance of the detection model. This leads to the technical problem of inaccurately determining the type of transformer fault. To achieve intelligent fault detection, multi-scale grid entropy can extract deep, highly expressive, and separable fault features. Random forests are easy to implement, have fast training speeds, and can efficiently identify multiple fault states. Therefore, combining the advantages of multi-scale grid entropy and random forests can achieve accurate online transformer fault detection.

[0168] An electronic device according to an embodiment of the present invention includes: a memory and a processor, wherein the memory stores a computer program; when the computer program is executed by the processor, the processor performs a method for determining the type of transformer fault as described in any of the above embodiments.

[0169] The memory can be an electronic memory such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM. The memory has storage space for program code used to perform any of the method steps described above. For example, the storage space for program code may include individual program codes for implementing the various steps in the methods described above. This program code can be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact discs (CDs), memory cards, or floppy disks. The program code may be compressed, for example, in a suitable form. When run by a computing processing device, this code causes the computing processing device to perform the various steps in the methods described above.

[0170] This invention provides a computer-readable storage medium storing a computer program thereon, which, when executed, implements a method for determining transformer fault types as described in any embodiment of this invention.

[0171] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0172] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0173] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0174] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0175] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0176] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions 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.

Claims

1. A method for determining transformer fault types, characterized in that, include: When a vibration sub-signal dataset is received, multi-scale operations are performed on the vibration sub-signal dataset to generate corresponding vibration sub-signal sample data. The vibration sub-signal sample data is used as input to train a preset initial random forest diagnostic model to generate a corresponding target random forest diagnostic model; In response to a received fault diagnosis request, the transformer to be diagnosed corresponding to the fault diagnosis request is selected and the resonator signal data is acquired; Performing multi-scale operations on the vibration sub-signal data and outputting the corresponding multi-scale grid entropy value includes: normalizing the vibration sub-signal data, performing multi-scale processing on the normalized signal, obtaining sub-signals at multiple scales given the required scale, generating a two-dimensional planar graph for each sub-signal, dividing the two-dimensional planar graph into an n*n grid given an n, counting the number of points in each grid, calculating the grid entropy of each sub-signal, and splicing the obtained grid entropies together to form the multi-scale grid entropy value. The multi-scale grid entropy value is input into the target random forest diagnostic model to generate corresponding target fault label category data; The target fault label category data is compared with the preset actual fault label category data, and the fault type of the transformer to be judged is determined based on the comparison result.

2. The method for determining transformer fault types according to claim 1, characterized in that, The step of using the vibration sub-signal sample data as input to train a preset initial random forest diagnostic model to generate a corresponding target random forest diagnostic model includes: The vibration sub-signal sample data is input into the initial random forest diagnostic model to generate corresponding training fault label category data; Calculate the training loss value between the training fault label category data and the associated standard fault label category data; Compare the training loss value with a preset training standard threshold; If the training loss value is less than or equal to the training standard threshold, then training is stopped and a target random forest diagnostic model is generated.

3. The method for determining transformer fault types according to claim 2, characterized in that, Also includes: If the training loss value is greater than the training standard threshold, the parameters of the initial random forest diagnostic model are adjusted according to the preset gradient, and the process jumps to the step of inputting the vibration sub-signal sample data into the initial random forest diagnostic model to generate the corresponding training fault label category data.

4. The method for determining transformer fault types according to claim 1, characterized in that, The step of comparing the target fault label category data with the preset actual fault label category data, and determining the fault type of the transformer to be judged based on the comparison result, includes: Compare the target fault label category data with the preset actual fault label category data; If the target fault label category data is consistent with the actual fault label category data, then the fault type associated with the actual fault label category data is determined for the transformer to be judged; If the target fault label category data is inconsistent with the actual fault label category data, it is determined that the transformer to be judged has not experienced the fault type associated with the actual fault label category data, and the process jumps to the step of using the multi-scale grid entropy value to input the target random forest diagnostic model to generate the corresponding target fault label category data.

5. A system for determining transformer fault types, characterized in that, include: The vibration sub-signal sample data acquisition module is used to perform multi-scale operations on the vibration sub-signal dataset when a vibration sub-signal dataset is received, and generate corresponding vibration sub-signal sample data. The target random forest diagnostic model generation module is used to train a preset initial random forest diagnostic model by inputting the vibration sub-signal sample data, and generate a corresponding target random forest diagnostic model. The response module is used to respond to the received fault judgment request, select the transformer to be judged corresponding to the fault judgment request, and obtain the resonator signal data; The multi-scale grid entropy value acquisition module is used to perform multi-scale operations on the vibration sub-signal data and output the corresponding multi-scale grid entropy value. The module includes: normalizing the vibration sub-signal data, performing multi-scale processing on the normalized signal, obtaining sub-signals at multiple scales given a required scale, generating a two-dimensional planar graph for each sub-signal, dividing the two-dimensional planar graph into an n*n grid given an n, counting the number of points in each grid, calculating the grid entropy of each sub-signal, and splicing the obtained grid entropies together to form the multi-scale grid entropy value. The target fault label category data acquisition module is used to input the multi-scale grid entropy value into the target random forest diagnostic model to generate corresponding target fault label category data. The comparison module is used to compare the target fault label category data with the preset actual fault label category data, and determine the fault type of the transformer to be judged based on the comparison result.

6. The transformer fault type determination system according to claim 5, characterized in that, The target random forest diagnostic model generation module includes: The training label category data acquisition submodule is used to input the multi-scale grid entropy value into the initial random forest diagnostic model to generate the corresponding training label category data. The training loss value acquisition submodule is used to calculate the training loss value between the training label category data and the associated standard label category data. The comparison submodule is used to compare the training loss value with a preset training standard threshold. The first data processing submodule is used to stop training and generate a target random forest diagnostic model if the training loss value is less than or equal to the training standard threshold.

7. The transformer fault type determination system according to claim 6, characterized in that, The target random forest diagnostic model generation module also includes: The second data processing submodule is used to adjust the parameters of the initial random forest diagnostic model according to a preset gradient if the training loss value is greater than the training standard threshold, and then jump to the step of inputting the vibration sub-signal sample data into the initial random forest diagnostic model to generate corresponding training fault label category data.

8. The transformer fault type determination system according to claim 5, characterized in that, The comparison module includes: The fault label category data comparison submodule is used to compare the target fault label category data with the preset actual fault label category data; The first determination submodule is used to determine the fault type associated with the actual fault label category data of the transformer to be determined if the target fault label category data is consistent with the actual fault label category data. The second determination submodule is used to determine that the transformer to be judged has not experienced the fault type associated with the actual fault label category data if the target fault label category data is inconsistent with the actual fault label category data, and then jump to execute the step of using the multi-scale grid entropy value to input the target random forest diagnostic model to generate the corresponding target fault label category data.

9. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor causes the processor to perform the steps of the transformer fault type determination method as described in any one of claims 1-4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the method for determining the type of transformer fault as described in any one of claims 1-4.

Images