Transformer unknown fault diagnosis method based on anova and igwo-svdd, medium and system

Abstract

This invention discloses a method, medium, and system for diagnosing unknown transformer faults based on ANOVA and IGWO-SVDD, relating to the field of transformer fault diagnosis technology. The method includes acquiring a sample set of transformer faults, wherein each sample in the sample set contains multiple different types of dissolved gases in the oil; calculating the content ratio of different dissolved gases in the oil in each sample as data features of multiple different feature types for each sample; using ANOVA to reduce the dimensionality of the data features of each sample to obtain the optimal feature subset; constructing an improved Grey Wolf Optimization Algorithm (IGWO-SVDD) fault diagnosis model based on support vector data description; and inputting the optimal feature subset into the IGWO-SVDD fault diagnosis model to obtain transformer fault diagnosis results for different fault types. This invention can improve the accuracy of fault type identification in transformer fault diagnosis tasks.

Description

Technical Field

[0001] This invention relates to the field of transformer fault diagnosis technology, and more specifically to a method, medium, and system for diagnosing unknown transformer faults based on ANOVA and IGWO-SVDD. Background Technology

[0002] Transformers are a crucial component of power systems, and their operational stability directly impacts grid security. A transformer failure can result in significant economic losses. Therefore, routine maintenance requires timely detection of potential faults and improved repair precision.

[0003] Currently, in the process of diagnosing transformer faults, the identification of unmarked transformer faults is usually treated as an anomaly detection problem, and a single classification algorithm is used to solve the unknown faults of the transformer.

[0004] However, existing single-classification algorithms have low classification accuracy when solving unknown transformer faults and cannot effectively identify different types of transformer faults.

[0005] Therefore, how to improve the accuracy of fault type identification in transformer fault diagnosis is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides a method, medium and system for diagnosing unknown faults in transformers based on ANOVA and IGWO-SVDD, which is used to improve the accuracy of fault type identification in transformer fault diagnosis tasks.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] First, this invention discloses a method for diagnosing unknown transformer faults based on ANOVA and IGWO-SVDD, comprising the following steps:

[0009] A sample set of multiple transformer fault types is obtained, wherein each fault sample in the sample set contains multiple dissolved gases in oil;

[0010] Calculate the content ratio of dissolved gases in different oils in each fault sample, and use it as a data feature of multiple different feature types in each fault sample;

[0011] ANOVA was used to reduce the dimensionality of the data features in each fault sample to obtain the optimal feature subset.

[0012] Construct an improved Grey Wolf Optimization Algorithm (IGWO-SVDD) fault diagnosis model based on support vector data description;

[0013] By inputting the optimal feature subset into the IGWO-SVDD fault diagnosis model, transformer fault diagnosis results for different fault types are obtained.

[0014] Preferably, the dissolved gases in the oil include H2, CH4, C2H4, C2H2, and C2H6.

[0015] Preferably, ANOVA is used to reduce the dimensionality of the data features in each fault sample, specifically including the following steps:

[0016] All data features of the same transformer fault type and the same feature type in the samples are grouped into the same dataset;

[0017] Calculate the between-group sum of squared errors and the within-group sum of squared errors for datasets with the same feature type;

[0018] The statistics of the feature type are calculated based on the sum of squared errors between groups and within groups of datasets with the same feature type.

[0019] Arrange the statistics of all feature types in descending order, and select the data features of the first preset number of feature types as the data features of each sample after dimensionality reduction.

[0020] Preferably, the formula for calculating the sum of squared errors between groups of the same feature type of dataset is:

[0021]

[0022] in, The sum of squared errors between groups represents the datasets with the same feature type; k represents the number of transformer fault types; m i This represents the number of data features in the dataset of the i-th type of transformer fault under the same feature type. This represents the mean of the data features in the dataset for the j-th type of transformer fault, which shares the same feature type. This represents the mean of the data features in a dataset containing all transformer fault types under the same feature type;

[0023] The formula for calculating the sum of squared errors within groups of datasets with the same feature type is as follows:

[0024]

[0025] in, x represents the sum of squared errors within groups of datasets with the same feature type; ij This represents the j-th data feature in the dataset of the i-th transformer fault type under the same feature type.

[0026] Preferably, the statistics of the feature type are calculated using the following formula:

[0027]

[0028] Where F represents the statistic of the same feature type, and k-1 represents The degrees of freedom, nk represents The degrees of freedom, where n represents the total number of samples.

[0029] Preferably, the construction of an improved Grey Wolf Optimization Algorithm (IGWO-SVDD) fault diagnosis model based on support vector data description includes the following steps:

[0030] The initial population of the Grey Wolf algorithm is optimized using a reverse algorithm.

[0031] The individual positions of the individuals after optimization of the initial population of the Grey Wolf Algorithm are updated using a differential algorithm, resulting in the improved Grey Wolf Algorithm IGWO.

[0032] Based on the improved Grey Wolf Optimization Algorithm (IGWO), the Support Vector Data Description (SVDD) model is optimized to obtain the IGWO-SVDD fault diagnosis model.

[0033] Preferably, the initial population optimization of the gray wolf algorithm is performed using a reverse algorithm, specifically including the following steps:

[0034] The initial population for the Gray Wolf algorithm is generated using the circle chaotic mapping, as shown in the following expression:

[0035]

[0036] In the formula, x i+1 x i These are the particle positions in the (i+1)th and ith iterations, respectively;

[0037] The initial population is initialized using a reverse learning algorithm to generate a reverse population.

[0038] The reverse population is ;

[0039] In the formula, OX id Indicates a reverse population; X maxd X represents the maximum value in the population; mind X represents the minimum value in the population; id Indicates a positive population;

[0040] Compare the current population X with the reverse population OX. id The population is merged to obtain a new population X'={X∪OX}. id}

[0041] Preferably, the position of individuals in the initial population optimized by the gray wolf algorithm is updated using a differential algorithm, specifically including the following steps:

[0042] The mutation operation is performed on the updated individual positions, and the specific expression is as follows:

[0043]

[0044] In the formula, v i (t) represents the solution for the position of the mutated individual, t represents the iteration number, and x r1 (t), x r2 (t), x r3 (t) represents 3 distinct individuals in the current population, and w represents the mutation factor;

[0045] The crossover operation is performed based on the position of the mutated individuals, as shown in the following expression:

[0046]

[0047] In the formula, u i (t) represents the solution for the individual position after crossover; rand represents a random function; CR represents the crossover factor; x i (t) is the optimal individual position vector in the current population;

[0048] Select operation:

[0049] A greedy strategy is used to compare the parent and child individuals to select a new individual. The specific expression is as follows:

[0050]

[0051] x i (t+1) represents the individual position solution after the selection operation; f(u i (t) represents the fitness function value obtained by substituting the crossover individual position into the fitness function f; f(x) i (t) represents the fitness function value obtained by substituting the optimal individual position vector in the current population into the fitness function f.

[0052] In a second aspect, the present invention discloses a computer-readable storage medium storing computer program instructions; when the computer program instructions are executed by a processor, they implement the transformer unknown fault diagnosis method described in the first aspect above.

[0053] Thirdly, the present invention also discloses a transformer unknown fault diagnosis system, comprising: a computer-readable storage medium as described in the second aspect above.

[0054] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a method, medium and system for diagnosing unknown faults in transformers based on ANOVA and IGWO-SVDD, which has the following beneficial effects. Attached Figure Description

[0055] 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 embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0056] Figure 1 This is a schematic diagram of the overall process of the transformer unknown fault diagnosis method provided by the present invention.

[0057] Figure 2 This is a schematic diagram of the process for constructing the IGWO-SVDD fault diagnosis model provided by the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Example 1

[0060] This invention discloses a method for diagnosing unknown transformer faults based on ANOVA and IGWO-SVDD, such as... Figure 1 As shown, it includes the following steps:

[0061] A sample set of multiple transformer fault types, wherein each fault sample in the sample set contains multiple different types of dissolved gases in oil.

[0062] Calculate the content ratio of dissolved gases in different oils in each fault sample, and use it as a data feature of multiple different feature types for each sample;

[0063] ANOVA was used to reduce the dimensionality of the data features of each sample to obtain the optimal feature subset.

[0064] Construct an improved Grey Wolf Optimization Algorithm (IGWO-SVDD) fault diagnosis model based on support vector data description;

[0065] By inputting the optimal feature subset into the IGWO-SVDD fault diagnosis model, transformer fault diagnosis results for different fault types are obtained.

[0066] Each step will be explained further below.

[0067] First, when collecting data on various transformer fault types using different sensors, the dissolved gases in the oil corresponding to each fault type mainly include five gases: H2, CH4, C2H4, C2H2, and C2H6. In this step, different fault types are used as sample sets. In this step, transformer fault types include medium-low temperature overheating, high temperature overheating, partial discharge, low-energy discharge, and high-energy discharge. In addition, data on normal transformer conditions are considered a special known fault type. The purpose of this invention is to determine other unknown transformer fault types based on one known fault type.

[0068] Secondly, the content ratio of dissolved gases in different oils in each fault sample is calculated as data features of multiple different feature types.

[0069] Then, ANOVA was used to reduce the dimensionality of the data features of each sample to obtain the optimal feature subset.

[0070] In transformer fault characteristic variance analysis, factors refer to the set of all transformer fault types, levels refer to each type of fault, and observations refer to the specific characteristic values ​​of each fault sample. The specific steps for characteristic variance analysis of dissolved gases in transformer oil are as follows:

[0071] 1) Make a hypothesis and test whether the means of the k levels of the factor are equal. Propose the following hypothesis:

[0072] Not fully equal

[0073] In the formula Let Hi be the mean of the i-th level, corresponding to the mean of the fault characteristics under the i-th type of transformer fault. If we reject the null hypothesis H0, it means that the independent variable has a significant impact on the dependent variable, indicating that this type of fault has a significant impact on the dissolved gas characteristics in transformer oil.

[0074] 2) Construct the test statistic. Calculate the mean of each sample, the sum of squared errors, and the F-statistic. The specific steps are as follows:

[0075] Calculate the mean of each sample, including the sample mean of the i-th fault type and the overall mean of all observations, using the following formulas:

[0076]

[0077]

[0078] In the formula, m represents the total sample size; m i x is the sample size for the i-th type of fault; ij This is the j-th observation of the i-th fault type; The sample mean of the j-th fault type; This is the overall mean of all observations.

[0079] Calculate the sum of squared errors for different samples, including calculating... Value and value.

[0080] in, Indicates the mean of each group The formula for calculating the sum of squared errors from the overall mean (the sum of squared errors between groups for datasets with the same feature type) is:

[0081]

[0082] In the formula, The sum of squared errors between groups represents the datasets with the same feature type; k represents the number of transformer fault types; m i This represents the number of data features in the dataset of the i-th type of transformer fault under the same feature type (the sample size of the i-th type of fault). This represents the mean of the data features in the dataset of the j-th type of transformer fault under the same feature type (the sample mean of the j-th type of fault). This represents the mean of the data features in a dataset of all transformer fault types under the same feature type (the overall mean of all observations).

[0083] The sum of squared errors between individual sample data and their group mean (the sum of squared errors between groups of datasets with the same feature type) is calculated using the following formula:

[0084]

[0085] In the formula, x represents the sum of squared errors within groups of datasets with the same feature type; ij This represents the j-th data feature (the j-th observation of the i-th fault type) in the dataset of the i-th transformer fault type under the same feature type.

[0086] Calculate the statistic E to eliminate the influence of the number of observations on the magnitude of the sum of squared errors. SSA and E SSE It needs to be averaged out.

[0087] Using k-1 degrees of freedom, The degrees of freedom used are nk, where n represents the total sample size. The influence of the observed values ​​on the magnitude of the sum of squared errors is shown. The formulas for the mean square and the statistical test are as follows:

[0088]

[0089]

[0090]

[0091] In the formula, F represents the statistic of the same characteristic type, and k-1 represents... The degrees of freedom, nk represents The degrees of freedom are given by F, where n represents the total sample size. A larger F value indicates a more significant impact of this type of fault on the dissolved gas characteristics in transformer oil. The statistic is compared with the critical value F at a given significance level α to make a decision.

[0092] When making a decision, based on the given significance level α, find the critical value F corresponding to the denominator degrees of freedom df1 = k−1 and the numerator degrees of freedom df2 = n−k in the distribution table of the test statistic F. α (k−1, n−k). If F> F α If the null hypothesis H0 is true, then we reject the null hypothesis H0; otherwise, we do not reject the null hypothesis H0.

[0093] Based on a given significance level α, corresponding to different degrees of freedom, the statistical value F of each candidate ratio feature is calculated. Features are then selected according to the magnitude of the value to form an initial optimal feature subset, which is used as input to the IGWO-SVDD fault diagnosis model.

[0094] like Figure 2 In this embodiment, an improved Grey Wolf Optimization Algorithm (IGWO-SVDD) fault diagnosis model based on support vector data description is constructed, including the following steps:

[0095] The initial population of the Grey Wolf algorithm is optimized using a reverse algorithm; specifically...

[0096] First, the initial population for the Gray Wolf algorithm is generated using the circle chaotic mapping, as shown in the following expression:

[0097]

[0098] In the formula, x i+1 x i These are the particle positions in the (i+1)th and ith iterations, respectively;

[0099] The initial population is initialized using a reverse learning algorithm to generate a reverse population.

[0100] The reverse population is ;

[0101] In the formula, OX id Indicates a reverse population; X maxd X represents the maximum value in the population; mind X represents the minimum value in the population; id Represents a positive population; positive population X id The expression is , where y id This represents the coefficient of inertia.

[0102] Compare the current population X with the obtained reverse population OX. id The population is merged to obtain a new population X'={X∪OX}. id}

[0103] The individual positions of the individuals after optimization of the initial population in the Gray Wolf Algorithm are updated using a differential algorithm, resulting in the improved Gray Wolf Optimization Algorithm IGWO. The specific steps are as follows:

[0104] The mutation operation is performed on the updated individual positions, and the specific expression is as follows:

[0105]

[0106] In the formula, v i (t) represents the solution for the position of the mutated individual, t represents the iteration number, and x r1 (t), x r2 (t), x r3 (t) represents 3 distinct individuals in the current population, and w represents the mutation factor;

[0107] The crossover operation is performed based on the position of the mutated individuals, as shown in the following expression:

[0108]

[0109] In the formula, u i (t) represents the solution for the individual position after crossover; rand represents a random function; CR represents the crossover factor; x i (t) is the optimal individual position vector in the current population;

[0110] Select operation:

[0111] A greedy strategy is used to compare the parent and child individuals to select a new individual. The specific expression is as follows:

[0112]

[0113] x i(t+1) represents the individual position solution after the selection operation; f(u i (t) represents the fitness function value obtained by substituting the crossover individual position into the fitness function f; f(x) i (t) represents the fitness function value obtained by substituting the optimal individual position vector in the current population into the fitness function f.

[0114] Based on the improved Grey Wolf Optimization Algorithm (IGWO), the Support Vector Data Description (SVDD) model is optimized to obtain the IGWO-SVDD fault diagnosis model.

[0115] In the above steps, the G-means function is used as the objective function to optimize the relevant parameters of the support vector data description SVDD, and the fitness function of the Grey Wolf algorithm is calculated. The specific expression is shown below:

[0116]

[0117] In the formula, TP represents the number of items correctly classified in the positive category, FN represents the number of items incorrectly classified in the positive category, TN represents the number of items correctly classified in the negative category, and FP represents the number of items incorrectly classified in the negative category.

[0118] The formula for calculating the fitness function f is shown below:

[0119]

[0120] In the formula, C represents the cross-validation fold number.

[0121] In the SVDD model, assuming the hypersphere has a radius of R and a center, the specific expression of the SVDD model is as follows:

[0122]

[0123] In the formula, α i , i Called the Lagrange operator, x s These are called support vectors, and z represents the collected feature parameters. Represents the kernel function, and x i x j These are the i-th training sample and the j-th training sample, respectively.

[0124] In this embodiment, a Gaussian kernel function is used, and its expression is:

[0125]

[0126] In the formula, σ is the kernel function parameter.

[0127] Support Vector Data Description (SVDD) is a single-class classification method that requires relatively little training data and can detect outliers. SVDD maps the training data to a high-dimensional space and then constructs a hypersphere, ensuring that as many training samples as possible are contained within it. Data within the hypersphere is considered training data, while data outside the hypersphere is classified as other types. After constructing the hypersphere, a decision function is used to determine whether test data belongs to the same class as the training data. If the decision function is 1, the distance from the data point to the center of the hypersphere is less than the hypersphere's radius, indicating that the data belongs to the same class as the training data; otherwise, it belongs to another class.

[0128] The optimization problem of SVDD can be described as follows:

[0129]

[0130] In the formula, C represents the penalty factor. This is the relaxation term introduced.

[0131] In the IGWO-SVDD fault diagnosis model, the relevant parameters describing SVDD using support vector data are optimized. The specific steps are as follows:

[0132] Step 1: Initialize the SVDD penalty factor C and kernel function parameter σ represented by the particle swarm in the improved gray wolf algorithm;

[0133] Step 2: Train SVDD using the penalty factor C to be optimized and the kernel function parameter σ represented by each particle, and calculate the fitness value of each particle;

[0134] Step 3: For each particle, compare the current fitness value under the penalty factor C and kernel function parameter σ with the best historical fitness value obtained in the previous step. If the current fitness value is better than the best historical fitness value, then update the particle's historical best penalty factor C and kernel function parameter σ.

[0135] Step 4: Compare the historical best fitness value of each particle with the fitness value under the best penalty factor C and kernel function parameter σ searched by the particle swarm of the improved gray wolf algorithm. The penalty factor C and kernel function parameter σ represented by the particle with the best fitness value are taken as the best penalty factor C and kernel function parameter σ of the current particle swarm.

[0136] Step 5 updates the particle velocity, penalty factor C, and kernel function parameter σ;

[0137] Step 6: Determine whether the particle meets the mutation conditions. If it does, mutate the penalty factor C and kernel function parameter σ represented by the particle.

[0138] Step 7: If the termination condition is met, output the final parameters to be optimized and train the SVDD model using the optimal parameters. Otherwise, go to Step 5.

[0139] Example analysis and calculation:

[0140] Data on low-temperature overheating, high-temperature overheating, partial discharge, low-energy discharge, high-energy discharge, and normal conditions were acquired. One of these conditions was designated as a known fault, while the remaining five were considered unknown. ANOVA was used to calculate statistics, and the data were then sorted and filtered. The sample set was divided into training and test sets in a 4:1 ratio. Tables 1 to 6 show the results for different types of transformer faults when the known fault type was used and the others were unknown. The specific results are shown in Tables 1-6. As can be seen from the tables, IGWO-SVDD can effectively identify different types of transformer faults, and its classification accuracy is significantly better than other methods.

[0141] Table 1: Normal is known, the rest are unknown.

[0142] Comparison Algorithm Known state Unknown state G-means(%) SVDD 69.5 73.2 72.1 PSO-SVDD 75.5 78.98 75.23 GWO-SVDD 81.23 84.32 83.11 IGWO-SVDD 89.34 91.23 90.55

[0143] Table 2 lists known cases of low-temperature overheating, and the rest are unknown.

[0144] Comparison Algorithm Known state Unknown state G-means(%) SVDD 71.24 76.34 74.55 PSO-SVDD 76.58 78.36 77.45 GWO-SVDD 83.45 87.32 85.15 IGWO-SVDD 90.45 93.23 92.61

[0145] Table 3 shows that high-temperature overheating is known, while the rest are unknown.

[0146] Comparison Algorithm Known state Unknown state G-means(%) SVDD 72.66 78.56 75.33 PSO-SVDD 76.34 79.56 77.54 GWO-SVDD 84.56 87.56 86.45 IGWO-SVDD 91.24 93.55 92.67

[0147] Table 4 shows that partial discharge is known, while the rest are unknown.

[0148] Comparison Algorithm Known state Unknown state G-means(%) SVDD 74.55 76.43 75.44 PSO-SVDD 78.23 80.32 79.23 GWO-SVDD 86.34 88.23 87.55 IGWO-SVDD 91.23 94.23 93.45

[0149] Table 5 shows that low-energy discharge is known, while the rest are unknown.

[0150] Comparison Algorithm Known state Unknown state G-means(%) SVDD 77.23 79.24 78.24 PSO-SVDD 75.12 78.23 77.25 GWO-SVDD 80.23 83.24 82.34 IGWO-SVDD 91.32 94.55 92.57

[0151] Table 6 shows that high-energy discharges are known, while the rest are unknown.

[0152] Comparison Algorithm Known state Unknown state G-means(%) SVDD 80.23 83.22 82.56 PSO-SVDD 83.12 86.23 84.56 GWO-SVDD 84.45 87.56 86.23 IGWO-SVDD 91.24 96.32 93.23

[0153] Example 2

[0154] This embodiment discloses a computer-readable storage medium storing computer program instructions; when the computer program instructions are executed by a processor, they implement the arbitrary transformer unknown fault diagnosis method described in Embodiment 1.

[0155] Example 3

[0156] This invention discloses a transformer unknown fault diagnosis system, which includes the computer-readable storage medium described in Embodiment 2.

[0157] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0158] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for diagnosing unknown transformer faults based on ANOVA and IGWO-SVDD, characterized in that, Includes the following steps: A sample set of multiple transformer fault types is obtained, wherein each fault sample in the sample set contains multiple dissolved gases in oil; Calculate the content ratio of dissolved gases in different oils in each fault sample, and use it as a data feature of multiple different feature types in each fault sample; ANOVA was used to reduce the dimensionality of the data features in each fault sample to obtain the optimal feature subset. A fault diagnosis model for the improved Grey Wolf Optimization Algorithm (IGWO-SVDD) based on support vector data description is constructed. Specifically, this includes: initial population optimization of the Grey Wolf Algorithm using a reverse algorithm, which specifically includes the following steps: The initial population for the Gray Wolf algorithm is generated using the circle chaotic mapping, as shown in the following expression: ； In the formula, x i+1 x i These are the particle positions in the (i+1)th and ith iterations, respectively; The initial population is initialized using a reverse learning algorithm to generate a reverse population. The reverse population is ; In the formula, OX id Indicates a reverse population; X maxd X represents the maximum value in the population; mind X represents the minimum value in the population; id Indicates a positive population; Compare the current population X with the reverse population OX. id The population is merged to obtain a new population X'={X∪OX}. id }; The improved Grey Wolf Optimization Algorithm (IGWO) is obtained by updating the individual positions of the initial population after optimization using a differential algorithm. The specific steps for updating the individual positions of the initial population after optimization using a differential algorithm are as follows: The mutation operation is performed on the updated individual positions, and the specific expression is as follows: ； In the formula, v i (t) represents the solution for the position of the mutated individual, t represents the iteration number, and x r1 (t), x r2 (t), x r3 (t) represents 3 distinct individuals in the current population, and w represents the mutation factor; The crossover operation is performed based on the position of the mutated individuals, as shown in the following expression: ； In the formula, u i (t) represents the solution for the individual position after crossover; rand represents a random function; CR represents the crossover factor; x i (t) is the optimal individual position vector in the current population; Select operation: A greedy strategy is used to compare the parent and child individuals to select a new individual. The specific expression is as follows: ； x i (t+1) represents the individual position solution after the selection operation; f(u i (t) represents the fitness function value obtained by substituting the crossover individual position into the fitness function f; f(x) i (t) represents the fitness function value obtained by substituting the optimal individual position vector in the current population into the fitness function f; Based on the improved Grey Wolf Optimization Algorithm (IGWO), the Support Vector Data Description (SVDD) model is optimized to obtain the IGWO-SVDD fault diagnosis model. By inputting the optimal feature subset into the IGWO-SVDD fault diagnosis model, transformer fault diagnosis results for different fault types are obtained.

2. The transformer unknown fault diagnosis method based on ANOVA and IGWO-SVDD according to claim 1, characterized in that, The dissolved gases in the oil include H2, CH4, C2H4, C2H2, and C2H6.

3. The transformer unknown fault diagnosis method based on ANOVA and IGWO-SVDD according to claim 1, characterized in that, The dimensionality reduction of the data features in each fault sample is performed using analysis of variance (ANOVA), specifically including the following steps: All data features of the same transformer fault type and the same feature type in the samples are grouped into the same dataset; Calculate the between-group sum of squared errors and the within-group sum of squared errors for datasets with the same feature type; The statistics of the feature type are calculated based on the sum of squared errors between groups and within groups of datasets with the same feature type. Arrange the statistics of all feature types in descending order, and select the data features of the first preset number of feature types as the data features of each sample after dimensionality reduction.

4. The transformer unknown fault diagnosis method based on ANOVA and IGWO-SVDD according to claim 3, characterized in that, The formula for calculating the sum of squared errors between groups for datasets with the same feature type is: ； in, The sum of squared errors between groups represents the datasets with the same feature type; k represents the number of transformer fault types; m i This represents the number of data features in the dataset of the i-th type of transformer fault under the same feature type. This represents the mean of the data features in the dataset for the j-th type of transformer fault, which shares the same feature type. This represents the mean of the data features in a dataset containing all transformer fault types under the same feature type; The formula for calculating the sum of squared errors within groups of datasets with the same feature type is as follows: ； in, x represents the sum of squared errors within groups of datasets with the same feature type; ij This represents the j-th data feature in the dataset of the i-th transformer fault type under the same feature type.

5. The transformer unknown fault diagnosis method based on ANOVA and IGWO-SVDD according to claim 4, characterized in that, The statistics for the feature type are calculated using the following formula: ； Where F represents the statistic of the same feature type, and k-1 represents The degrees of freedom, nk represents The degrees of freedom, where n represents the total number of samples.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by a processor, they implement the transformer unknown fault diagnosis method as described in any one of claims 1 to 5.

7. A transformer unknown fault diagnosis system, characterized in that, include: The computer-readable storage medium as described in claim 6.

Images