SF6 switch equipment fault diagnosis method and system based on decomposition components

Abstract

The invention discloses an SF6 switch equipment fault diagnosis method and system based on decomposition components, and the method comprises the steps: firstly obtaining decomposition product data of SF6 switch equipment, and carrying out the standardization processing; and then, determining a fault type by using a pre-constructed SF6 switch equipment fault diagnosis model in combination with the standardized data. According to the model, a radial basis function is used as a kernel function, a support vector machine is used as a base classifier, and feature extraction and dimension reduction are carried out on standardized data through weighted principal component analysis. Meanwhile, parameters of the support vector machine are optimized by adopting an improved adaptive sparrow search algorithm. Through the WPCA-ASSA-SVM method, the accuracy and efficiency of SF6 switch equipment fault diagnosis are remarkably improved, a new solution and thought are provided for latent fault diagnosis based on decomposition component analysis, and development of the SF6 switch equipment fault diagnosis technology is promoted.

Description

Technical Field

[0001] The present invention belongs to the technical field of SF6 switchgear fault diagnosis, and relates to a method and system for SF6 switchgear fault diagnosis based on decomposed components. Background Art

[0002] High-voltage switchgear is an important control and protection equipment in the power system. Its safety and reliability play a decisive role in the stable operation of the power system. SF6 is currently the most excellent insulating and arc-extinguishing medium, and it occupies a dominant position in high-voltage switchgear represented by gas switchgear and GIS. Although SF6 switchgear has high safety and reliability, it may still have a variety of latent faults such as metal protrusions, free conductive particles, insulator surface contamination, suspended potential discharge, excessive contact ablation, excessive grease coating, etc. during long-term operation. These latent faults will further develop and deteriorate during the operation of the equipment. If they cannot be discovered and stopped in time, they may cause more serious safety accidents. Therefore, timely and effective status monitoring and latent fault diagnosis of SF6 switchgear are of great significance to ensure the stability of the power system.

[0003] Under the influence of local discharge, spark discharge and other factors that may occur during the operation of SF6 switchgear, the SF6 gas inside it will decompose to varying degrees and produce low-fluorine sulfides such as SF5 and SF4. When there are organic insulating materials and metal materials near the discharge area, substances such as CF4, C2F6, CO, and CO2 will also be produced. The types of SF6 decomposition products produced by internal discharge of SF6 switchgear are closely related to the materials near the discharge area, and their concentrations are inevitably directly affected by the discharge type and discharge energy. Therefore, there is an intrinsic connection between SF6 decomposition products and the types and severity of latent faults inside SF6 switchgear, and the status monitoring and latent fault diagnosis of SF6 switchgear can be achieved through decomposition component analysis. Compared with pulse current, ultra-high frequency, ultrasonic and other methods, the SF6 decomposition component analysis method has the advantages of not having to shut down the equipment, not being affected by on-site electromagnetic interference, and high sensitivity.

[0004] However, the problem of excessive grease coating has not been considered in the current research on SF6 switchgear fault diagnosis, the problem of tracing the decomposition products of SF6 has not been completely solved, there is a lack of basis for the selection of characteristic decomposition products, and the level of latent fault diagnosis of SF6 switchgear based on decomposition component analysis is difficult to improve. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention provides a method and system for SF6 switchgear fault diagnosis based on decomposition components, thereby solving the technical problems that the prior art does not consider the problem of excessive grease coating in SF6 switchgear fault diagnosis, and the SF6 decomposition products are incompletely traceable and there is no basis for selecting characteristic decomposition products, resulting in difficulty in improving the diagnosis level.

[0006] The present invention is achieved through the following technical solutions:

[0007] A method for SF6 switchgear fault diagnosis based on decomposition components comprises the following steps:

[0008] S1: Obtain decomposition product data of the SF6 switchgear, and perform standardization processing on the obtained data to obtain standardized data; specifically, the data here includes the type of decomposition products and the concentration of each decomposition product, and then the concentration of each decomposition product obtained is standardized.

[0009] S2: determining the fault type of the SF6 switchgear according to the standardized data and a pre-built SF6 switchgear fault diagnosis model;

[0010] In the process of constructing the SF6 switchgear fault diagnosis model, the radial basis function is selected as the kernel function, the support vector machine is selected as the base classifier, and the weighted principal component analysis method is used to extract features and reduce the dimension of the standardized data. The improved adaptive sparrow search algorithm is used to optimize the penalty parameters and kernel function parameters of the support vector machine; the improvement of the adaptive sparrow search algorithm includes the cat mapping initialization population, the explorer position update process and the adaptive hybrid mutation.

[0011] Preferably, the acquired data is standardized, and the standardization function is:

[0012]

[0013] Where: x * is the data after standardization, x is the data before standardization, μ is the mean of the decomposition product data obtained; σ is the standard deviation of the original data.

[0014] Preferably, the weighted principal component analysis method is used to extract features from the standardized data. During the weighting process, the weighted covariance matrix is:

[0015] S W =W T X T XW

[0016] Where: S Wis the weighted covariance data; W is an m×m diagonal weighted matrix, m is the number of features, and the diagonal elements of W are 1 / σ j Calculated, where 1 / σ j is the standard deviation of the jth feature; X is the input data set, where each column is a feature vector; T is the matrix transpose.

[0017] Preferably, the radial basis function is:

[0018] K(x i ,x j ) = exp(-γ||x i -x j || 2 ),γ>0

[0019] Among them, K(x i ,x j ) is the sample data after linear decomposition; x i and x j is the sample data; γ is the Gaussian kernel function.

[0020] Preferably, in the cat mapping initialization population, the cat mapping expression is:

[0021]

[0022] Where: a1 and b1 are arbitrary real numbers; mod1 means the decimal part of a1; w i and i is the position of the pixel in the grayscale image before transformation;

[0023] The improved updating process of the explorer position is:

[0024]

[0025] in: is the j-dimensional position information of the i-th sparrow at the t-th iteration; iter max is the maximum number of iterations; α∈[0,1] is a random number; Q is a random number that follows a normal distribution; L is a matrix of shape 1×d, in which all elements are 1; R2 is the warning value, which is a random number in the range of [0,1]; ST∈[0.5,1] is the safety threshold.

[0026] Preferably, the adaptive hybrid mutation combines Gaussian mutation and random reverse learning, specifically: Gaussian mutation is performed when the current individual fitness is less than the average fitness, and random reverse learning is performed when the current individual fitness is greater than or equal to the average fitness.

[0027] A system for SF6 switchgear fault diagnosis based on decomposition components, comprising:

[0028] A data acquisition module, wherein the data acquisition module is used to acquire the decomposition product data of the SF6 switchgear and obtain standardized data after standardizing the acquired data;

[0029] A data processing module, the data processing module is used to determine the fault type of the SF6 switchgear according to the standardized data and a pre-constructed SF6 switchgear fault diagnosis model; during the construction of the SF6 switchgear fault diagnosis model, a radial basis function is selected as a kernel function, a support vector machine is selected as a base classifier, and a weighted principal component analysis method is used to realize feature extraction and dimensionality reduction of the standardized data, and an improved adaptive sparrow search algorithm is used to realize the optimization of the penalty parameters and kernel function parameters of the support vector machine; the improvement of the adaptive sparrow search algorithm includes a cat mapping initialization population, an explorer position update process, and an adaptive hybrid mutation.

[0030] A computer device comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above method.

[0031] A computer-readable storage medium stores a computer program, wherein the computer program / instructions are executed by a processor to implement the steps of the above method.

[0032] A computer program product comprises a computer program / instruction, which implements the steps of the above method when executed by a processor.

[0033] Compared with the prior art, the present invention has the following beneficial technical effects:

[0034] The present invention discloses a method for diagnosing SF6 switchgear faults based on decomposition components. The method uses a weighted principal component analysis method to extract features and reduce the dimension of standardized data, and can more effectively screen out characteristic variables that have an important influence on fault diagnosis. These characteristic variables include the influence of excessive grease coating on SF6 decomposition products, because excessive grease coating will change the discharge conditions or produce specific decomposition products. In the process of weighted principal component analysis (WPCA), these features related to excessive grease coating will be given higher weights, so that they can be retained in the feature extraction process; in addition, by constructing a fault diagnosis model based on support vector machine (SVM) and optimizing the model using adaptive sparrow search algorithm (ASSA), the ability to identify the relationship between SF6 decomposition products and fault types is improved. The ASSA algorithm can more effectively find the optimal parameters of the SVM model through the combination of global search and local search, thereby improving the accuracy of the model in tracing the source of SF6 decomposition products; in addition, the WPCA method improves the separability of data by reducing the overlap and redundancy between features, which also helps to more accurately identify the association between SF6 decomposition products and fault types; the present invention adopts the WPCA method in the feature extraction stage, and the method selects the principal component based on the principle that the cumulative variance contribution is greater than 90%. The WPCA method can more effectively retain important information and filter interference information, which means that when selecting feature decomposition products, the WPCA method provides a clearer basis and standard. At the same time, by optimizing the SVM model parameters with the ASSA algorithm, the model's ability to identify the sensitivity of the characteristic decomposition products is further improved, thereby ensuring the effectiveness of the selected characteristic decomposition products in fault diagnosis. In summary, the present invention has made significant progress in solving the problems of the prior art not considering excessive grease coating in SF6 switchgear fault diagnosis, incomplete tracing of SF6 decomposition products, and lack of basis for the selection of characteristic decomposition products by adopting the WPCA-ASSA-SVM method. This method not only improves the accuracy and efficiency of fault diagnosis, but also provides a new way and idea for improving the level of latent fault diagnosis of SF6 switchgear based on decomposition component analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0036] Figure 1 It is a flow chart of a method for diagnosing SF6 switchgear faults based on decomposition components in the present invention;

[0037] Figure 2 It is a flow chart of a method for SF6 switchgear fault diagnosis based on decomposition component analysis in one embodiment of the present invention;

[0038] Figure 3 The figure is a schematic diagram of the structure of a system for diagnosing SF6 switchgear faults based on decomposed components in the present invention. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0040] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0042] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0043] Example 1

[0044] like Figure 1 As shown, the present invention discloses a method for SF6 switchgear fault diagnosis based on decomposition components, comprising the following steps:

[0045] S1: Obtain decomposition product data of the SF6 switchgear, and perform standardization processing on the obtained data to obtain standardized data; specifically, the data here includes the type of decomposition products and the concentration of each decomposition product, and then the concentration of each decomposition product obtained is standardized.

[0046] S2: determining the fault type of the SF6 switchgear according to the standardized data and a pre-built SF6 switchgear fault diagnosis model;

[0047] In the process of constructing the SF6 switchgear fault diagnosis model, the radial basis function is selected as the kernel function, the support vector machine is selected as the base classifier, and the weighted principal component analysis method is used to realize feature extraction and dimensionality reduction of the standardized data. The improved adaptive sparrow search algorithm is used to realize the optimization of the penalty parameters and kernel function parameters of the support vector machine; the improvement of the adaptive sparrow search algorithm includes the cat mapping initialization population, the explorer position update process and the adaptive hybrid mutation.

[0048] That is, firstly, the decomposition product data of the SF6 switchgear is collected and used as the input sample, and the input sample is standardized to obtain the standardized data; then, the weighted principal component analysis (WPCA) method is used to extract the features of the standardized data; based on the support vector machine (SVM), a preliminary fault diagnosis model is constructed; based on an improved adaptive sparrow algorithm (ASSA), the constructed fault diagnosis model is optimized, and then the construction of the SF6 switchgear fault diagnosis model based on weighted principal component analysis and adaptive sparrow search algorithm optimized support vector machine (WPCA-ASSA-SVM) proposed in the present invention is completed.

[0049] The collected data is preprocessed using the standard deviation standardization method. The conversion function of the processing method, that is, the standardization function, is:

[0050]

[0051] Where: x * is the data after standardization, x is the data before standardization, μ is the mean of the decomposition product data obtained, that is, the mean of the original data; σ is the standard deviation of the original data.

[0052] Furthermore, the weighted principal component analysis method is used to extract features from the standardized data. The weighted principal component analysis method is weighted by determining the corresponding weight according to the inverse of the standard deviation of the feature. In the weighting process, the weighted covariance matrix is ​​defined as follows:

[0053] S W =W T X T XW

[0054] Where: S W is the weighted covariance data; W is an m×m diagonal weighted matrix, m is the number of features, and the diagonal elements of W are 1 / σ j Calculated, where 1 / σ j is the standard deviation of the jth feature; X is the input data set, where each column is a feature vector; T is the matrix transpose.

[0055] Further preferably, when constructing a fault diagnosis model based on a support vector machine, the present invention selects a radial basis function (RBF) as a kernel function, because there is a complex nonlinear relationship between the SF6 switchgear fault and the decomposed components analysis (DCA) data, and RBF can nonlinearly map samples to a high-dimensional space and adaptively adjust the shape of the mapping space according to the distribution of the data, thereby better capturing the intrinsic structure of the data. The RBF expression, that is, the radial basis function is:

[0056] K(x i ,x j ) = exp(-γ||x i -x j || 2 ),γ>0

[0057] Among them, K(x i ,x j ) is the sample data after linear decomposition; x i and x j is the sample data; γ is the Gaussian kernel function.

[0058] When optimizing the support vector machine, the present invention optimizes the parameters of the SVM by improving the adaptive sparrow algorithm (ASSA). The algorithm simulates the movement of the sparrow population to optimize the parameters, and the entire population is divided into explorers, followers and guards. The explorer is responsible for foraging and providing directions for the entire sparrow population to guide the group flow; the follower will follow the explorer with the best fitness value to forage, and may compete with it for food. If the snatch is successful, it will replace the explorer to forage in a wider range; the guard is a group of sparrows randomly selected, and avoids falling into the local optimum through anti-predation strategies.

[0059] The explorer's position update formula is as follows:

[0060]

[0061] Where: is the j-dimensional position information of the i-th sparrow at the t-th iteration; itermax represents the maximum number of iterations; α∈[0,1] is a random number; Q is a random number that follows a normal distribution; L is a matrix of shape 1×d, in which all elements are 1; R2 represents the warning value, which is a random number in the range of [0,1]; ST∈[0.5,1] represents the safety threshold.

[0062] The follower's position update formula is as follows:

[0063]

[0064] Where X P Indicates the best position occupied by the producer; X worst represents the global worst position; A is a matrix of shape 1×d, each element in the matrix is ​​either 1 or -1, and satisfies A + =A T ×(A×A T ) -1 .

[0065] The position update formula of the sentinel is as follows:

[0066]

[0067] Where: represents the current global optimal position; β represents the step length and obeys the standard normal distribution; K∈[-1,1] is a random number, indicating the direction of the sparrow's movement, and is also the step length control parameter; f i 、f g and f w They represent the fitness value of the current sparrow individual, the current global best fitness and the current global worst fitness respectively; ε is a constant.

[0068] Since the initial strategy of the original SSA is a simple random method that cannot guarantee the diversity of the population, its later iterations are still prone to converge to local optimization prematurely. To solve this problem, the present invention improves the adaptive sparrow algorithm. The improvement of the adaptive sparrow search algorithm includes a cat mapping initialization population, an explorer position update process, and an adaptive hybrid mutation.

[0069] In the cat mapping initialization population, the cat mapping expression is:

[0070]

[0071] Where: a1 and b1 are arbitrary real numbers; mod1 means the decimal part of a1; w i and i is the position of the pixel in the grayscale image before transformation.

[0072] The improved updating process of the explorer position is:

[0073]

[0074] in: is the j-dimensional position information of the i-th sparrow at the t-th iteration; iter max is the maximum number of iterations; α∈[0,1] is a random number; Q is a random number that follows a normal distribution; L is a matrix of shape 1×d, in which all elements are 1; R2 is the warning value, which is a random number in the range of [0,1]; ST∈[0.5,1] is the safety threshold.

[0075] Further preferably, the adaptive hybrid mutation combines Gaussian mutation and random reverse learning, specifically: Gaussian mutation is performed when the current individual fitness is less than the average fitness, and random reverse learning is performed when the current individual fitness is greater than or equal to the average fitness.

[0076] Specifically, Gaussian variation comes from Gaussian distribution, and the formula is as follows:

[0077] mutation(x)=x(1+randn)

[0078] Where: x is the original individual position; mutation(x) is the individual position after Gaussian mutation; randn is a random number that satisfies the standard normal distribution.

[0079] The formula for random reverse learning is as follows:

[0080] x' j =k1x 1b +k2(x ub -x j )

[0081] Where: k1 and k2 are both random factors, both are random numbers between 0 and 1.

[0082] The present invention discloses a method for SF6 switchgear fault diagnosis based on decomposition components. The method is based on a weighted principal component analysis and adaptive sparrow search algorithm optimized support vector machine (WPCA-ASSA-SVM) fault diagnosis method, selects a support vector machine as a base classifier, adopts a weighted principal component analysis method to realize feature extraction and dimension reduction of original data, adopts an adaptive sparrow search algorithm to realize optimization of penalty parameters and kernel function parameters of the support vector machine, and integrates and constructs a WPCA-ASSA-SVM fault diagnosis model. When selecting the principal component according to the principle that the cumulative variance contribution is greater than 90%, the PCA method needs to retain the first 19 principal components, while the WPCA method only needs to retain the first 10 principal components. Compared with the PCA method, the WPCA method can effectively reduce the overlap of features and improve the separability of data, thereby better filtering interference information and retaining important information. The adaptive sparrow search algorithm used in the present invention is compared with the traditional genetic algorithm, gray wolf optimization algorithm and original sparrow search algorithm. When the adaptive sparrow search algorithm is used for parameter optimization, a lower fitness value can be achieved with fewer iterations, indicating that the adaptive sparrow search algorithm has a stronger optimization effect than the other three parameter optimization algorithms, has a faster convergence speed and has the ability to jump out of the local optimum, can ensure convergence to the optimal solution, and effectively improves the global optimization ability of the model. Diagnostic experiments were carried out using the fault data of SF6 switchgear running on the grid. The diagnostic experimental results show that the WPCA-ASSA-SVM fault diagnosis model performs well in multiple evaluation indicators and can realize accurate diagnosis of common latent faults of SF6 switchgear running on the grid, so that the level of latent fault diagnosis of SF6 switchgear based on decomposition component analysis is further improved.

[0083] Example 2

[0084] A method for SF6 switchgear fault diagnosis based on decomposition component analysis, comprising:

[0085] Step 1: Collect decomposition product data of SF6 switchgear and use it as input sample. After standardization of the input sample, obtain standardized data;

[0086] Step 2: Use the weighted principal component analysis method (WPCA) to extract features from the standardized data;

[0087] Step 3: Based on support vector machine (SVM), a preliminary fault diagnosis model is constructed;

[0088] Step 4: Based on an improved adaptive sparrow algorithm, the fault diagnosis model constructed in step 3 is optimized, thereby completing the construction of the SF6 switchgear fault diagnosis model based on weighted principal component analysis and adaptive sparrow search algorithm optimized support vector machine (WPCA-ASSA-SVM) proposed in the present invention;

[0089] For step 1, the present invention further improves as follows:

[0090] For high-voltage switchgear in field operation, due to differences in equipment structure, fault location, etc., the components and concentrations of gas decomposition products may still fluctuate greatly, and such fluctuations may have a significant impact on the accuracy of fault diagnosis. Therefore, in order to more comprehensively and perfectly reflect the fault information of high-voltage switchgear, this paper uses the 9 original characteristic gases of the fault sample as the basis, integrates the ratio of characteristic components, and expands and constructs a 27-dimensional fault feature set, as shown in Table 1:

[0091] Table 1 Fault feature set of high-voltage switchgear

[0092]

[0093] In the SF6 switchgear fault samples, the contents of various characteristic gas components vary significantly, ranging from thousands of ppm to only a few ppm. Considering that the differences in the order of magnitude of different fault characteristics in the samples will affect the accuracy and convergence speed of the fault diagnosis model, the standard deviation normalization method is used to preprocess the data. This method can map each feature of the sample data to a standard normal distribution with a mean of 0 and a standard deviation of 1. It is suitable for situations where the data has outliers and the minimum or maximum values ​​of the data are not fixed. The conversion function is:

[0094]

[0095] Where: μ is the mean of the original data; σ is the standard deviation of the original data.

[0096] For step 2, the present invention further improves as follows:

[0097] Principal Component Analysis (PCA) is a linear analysis method based on minimizing the reconstruction error and the mean square error. It provides a linear transformation matrix between high-dimensional and low-dimensional space, which can map high-dimensional data to low-dimensional space. It can reduce the number of features, remove redundant information, and reduce the complexity of the problem. It is widely used because of its simplicity, ease of implementation and high data interpretability.

[0098] However, PCA aims to minimize the reconstruction error and the Euclidean distance of each dimension of features, and treats each dimension of features equally, which leads to poor results when processing data of different importance. Weighted principal component analysis (WPCA) introduces weights on the basis of PCA, uses weighted distance to reduce the impact of outliers on the projection direction, highlights the features related to the principal component, and can better approximate nonlinear relationships through appropriate weight allocation. This method can effectively reduce feature overlap when processing non-uniform weight data, better retain important information, and improve data separability.

[0099] In order to evaluate the importance of each feature in the SF6 switchgear fault diagnosis feature set, the present invention determines its corresponding weight according to the inverse of the standard deviation of the feature. In the weighting process, the weighted covariance matrix is ​​defined as follows:

[0100] S W =W T X T XW

[0101] Where: X represents the input data set, where each column is a feature vector; W represents an m×m diagonal weighted matrix, where m is the number of features. The diagonal elements of W are 1 / σ j Calculated, where 1 / σ j is the standard deviation of the jth feature.

[0102] For step 3, the present invention further improves as follows:

[0103] Support vector machine (SVM) is a supervised learning algorithm with strong generalization ability. It is suitable for processing high-dimensional and nonlinear data. It is particularly effective when processing small or medium-sized data sets. Its core concept is to solve the separating hyperplane that can correctly divide the training data set and has the largest geometric spacing.

[0104] By introducing Lagrange multipliers, SVM can be expressed as the following dual problem for solving the largest separating hyperplane:

[0105]

[0106] Since the dimension may become very large after the low-dimensional space is mapped to the high-dimensional space, calculating the dot product between all samples requires a lot of computing resources. Therefore, the kernel function is introduced to solve the above dual problem, and the optimal classification function is obtained as follows:

[0107]

[0108] Where: m is the number of support vectors; K(x,x i ) represents the kernel function; x represents the input vector of the training sample; is the center of the kernel function.

[0109] Different kernel functions have different parameters and mapping characteristics, which have a significant impact on the performance and accuracy of SVM. The present invention selects radial basis function (RBF) as the kernel function because there is a complex nonlinear relationship between SF6 switchgear faults and DCA data, and RBF can nonlinearly map samples into high-dimensional space and adaptively adjust the shape of the mapping space according to the distribution of data, so as to better capture the intrinsic structure of the data. The RBF expression is:

[0110] K(x i ,x j ) = exp(-γ||x i -x j || 2 ),γ>0

[0111] For step 4, the present invention further improves as follows:

[0112] The penalty parameter C and the kernel function parameter γ have a great influence on the classification performance of SVM. The present invention optimizes the parameters of SVM by improving the adaptive sparrow algorithm (ASSA). The algorithm simulates the movement of sparrow populations to optimize parameters, and the entire population is divided into explorers, followers, and guards. Explorers are responsible for foraging and provide directions for the entire sparrow population to guide the flow of the group; followers will follow the explorer with the best fitness value to forage, and may compete with it for food. If the snatch is successful, they will replace the explorer to forage in a wider range; guards are a group of sparrows randomly selected, and avoid falling into local optimality through anti-predation strategies.

[0113] The explorer's position update formula is as follows:

[0114]

[0115] Where: is the j-dimensional position information of the i-th sparrow at the t-th iteration; iter max represents the maximum number of iterations; α∈[0,1] is a random number; Q is a random number that follows a normal distribution; L is a matrix of shape 1×d, in which all elements are 1; R2 represents the warning value, which is a random number in the range of [0,1]; ST∈[0.5,1] represents the safety threshold.

[0116] The follower's position update formula is as follows:

[0117]

[0118] Where X P Indicates the best position occupied by the producer; X worstrepresents the global worst position; A is a matrix of shape 1×d, each element in the matrix is ​​either 1 or -1, and satisfies A + =A T ×(A×A T ) -1 .

[0119] The position update formula of the sentinel is as follows:

[0120]

[0121] Where: represents the current global optimal position; β represents the step length and obeys the standard normal distribution; K∈[-1,1] is a random number, indicating the direction of the sparrow's movement, and is also the step length control parameter; f i 、f g and f w They represent the fitness value of the current sparrow individual, the current global best fitness and the current global worst fitness respectively; v is a constant.

[0122] Since the initial strategy of the original SSA is a simple random method, the diversity of the population cannot be guaranteed, and its later iterations are still prone to converge to local optimization too early. To solve this problem, the present invention improves the adaptive sparrow algorithm, and its main improvement strategies are as follows:

[0123] (1) Cat mapping initialization population:

[0124] Cat mapping, also known as Arnold mapping, is a chaotic mapping method used for repeated folding and stretching transformations in a limited area. The initial population generated by this mapping is more diverse and less likely to fall into small cycles and fixed points, which can speed up convergence and improve global search capabilities.

[0125] The cat mapping expression is:

[0126]

[0127] Where: a1 and b1 are arbitrary real numbers; mod1 means finding the decimal part of a1.

[0128] (2) Improve the explorer position update formula

[0129] Since the way the sparrows of the original sparrow search algorithm converge to the current optimal solution is to jump directly to the current optimal solution, it is easy for the algorithm to fall into the local optimum. In order to enhance the global search capability, the producer's formula is updated as follows:

[0130]

[0131] (3) Adaptive hybrid mutation

[0132] In order to further enhance the algorithm's ability to escape local optimality, the present invention combines Gaussian mutation and random reverse learning to perform adaptive hybrid mutation: Gaussian mutation is performed when the current individual fitness is less than the average fitness, and random reverse learning is performed when the current individual fitness is greater than or equal to the average fitness.

[0133] Gaussian variation comes from Gaussian distribution, and the formula is as follows:

[0134] mutation(x)=x(1+randn)

[0135] Where: x is the original individual position; mutation(x) is the individual position after Gaussian mutation; randn is a random number that satisfies the standard normal distribution.

[0136] The formula for random reverse learning is as follows:

[0137] x' j =k1x 1b +k2(x ub -x j )

[0138] Where: k1 and k2 are both random factors, both are random numbers between 0 and 1.

[0139] In addition, considering that the mutated sparrow position is not necessarily better than the original position, the present invention adopts a greedy strategy to choose whether to replace the original solution, that is, the replacement is performed only when the fitness value of the mutated solution is better than the original solution.

[0140] In the ASSA algorithm, each sparrow position represents a potential parameter solution. During the training and testing process of SVM, the goal is to improve the generalized performance of the classification model, that is, to maximize the classification accuracy of the training samples, which is reflected in this study as minimizing the fitness function. The sparrow position with the minimum fitness is better than other sparrow positions, and its value will be retained during the optimization process to obtain the optimal parameters.

[0141] In order to ensure the generalization performance, the present invention combines 5-fold cross validation and adopts the following fitness function to measure the optimality of the optimization problem considered:

[0142]

[0143] Where: k is the number of folds in the cross-validation; m is the number of each subset used for validation; Y ij is the number of correct predictions; F ijis the number of samples in the subset. This fitness function comprehensively considers the prediction accuracy of the model in different folds of cross-validation, calculates the fitness value of each fold by taking the square root of the correct prediction ratio, and averages these values. This method can better balance the performance of the model on each subset and avoid fluctuations in the fitness value due to extreme prediction accuracy on a certain subset.

[0144] Example 3

[0145] In order to further explain the scheme of the present invention, it is described by the following examples:

[0146] The present invention is a method for SF6 switchgear fault diagnosis based on decomposition component analysis, such as Figure 2 As shown, the specific implementation steps are as follows:

[0147] Step 1: Input the acquired original fault sample data as the initial fault feature, and then preprocess the data using the standard deviation standardization method. The formula for preprocessing the data using the standard deviation standardization method is defined as:

[0148]

[0149] Where: μ is the mean of the original data; σ is the standard deviation of the original data.

[0150] Step 2: Use weighted principal component analysis (WPCA) to extract features from the standardized data;

[0151] Step 3: Initialize all parameters, including the current number of iterations T and the maximum number of iterations T max , initialize the sparrow population size N, warning value ST, and number of explorers P d and the number of sentinels S d ;

[0152] Step 4: Use cat mapping to initialize the population, evaluate the fitness of each sparrow and sort them, and obtain the global optimal sparrow position with the minimum fitness. The formula for evaluating the fitness of each sparrow is:

[0153]

[0154] Where: k is the number of folds in the cross-validation; m is the number of each subset used for validation; Y ij is the number of correct predictions; F ijis the number of samples in the subset. This fitness function comprehensively considers the prediction accuracy of the model in different folds of cross-validation, calculates the fitness value of each fold by taking the square root of the correct prediction ratio, and averages these values. This method can better balance the performance of the model on each subset and avoid fluctuations in the fitness value due to extreme prediction accuracy on a certain subset.

[0155] Step 5: For each candidate sparrow, according to the cross-validation theory, use the corresponding parameters to train the support vector machine classification model (SVM);

[0156] Step 6: Update the positions of the explorer, follower, and sentinel, including:

[0157] Step 6.1: Update the explorer's position based on the formula:

[0158]

[0159] Step 6.2: Update the follower position based on the formula:

[0160]

[0161] Where: X P Indicates the best position occupied by the producer; X worst represents the global worst position; A is a matrix of shape 1×d, each element in the matrix is ​​either 1 or -1, and satisfies A + =A T ×(A×A T ) -1 .

[0162] Step 6.3: Update the formula for the sentinel's position:

[0163]

[0164] Where: represents the current global optimal position; β represents the step length and obeys the standard normal distribution; K∈[-1,1] is a random number, indicating the direction of the sparrow's movement, and is also the step length control parameter; f i 、f g and f w They represent the fitness value of the current sparrow individual, the current global best fitness and the current global worst fitness respectively; v is a constant.

[0165] Step 7: Recalculate the fitness f of each sparrow i And the average fitness f avg , and then perform adaptive hybrid mutation. The process of adaptive hybrid mutation is as follows:

[0166] If the individual fitness is less than the average fitness, Gaussian mutation is performed. The formula for Gaussian mutation is:

[0167] mutation(x)=x(1+randn)

[0168] Where: x is the original individual position; mutation(x) is the individual position after Gaussian mutation; randn is a random number that satisfies the standard normal distribution.

[0169] If the individual fitness is greater than the average fitness, random reverse learning is performed. The formula for random reverse learning is:

[0170] x' j =k1x 1b +k2(x ub -x j )

[0171] Where: k1 and k2 are both random factors, both are random numbers between 0 and 1.

[0172] Step 8: Recalculate the fitness and sort and update the global optimal sparrow position, and increase the number of iterations by one;

[0173] Step 9: Repeat steps 7 and 8 until the number of iterations reaches the set value;

[0174] Step 10: Output the optimal parameters C and γ, build an adaptive sparrow search algorithm optimized support vector machine (ASSA-SVM) fault diagnosis model with the optimal parameters, and output and analyze the fault diagnosis results.

[0175] Example 4

[0176] In addition, if Figure 3 As shown, the present invention also discloses a system for diagnosing SF6 switchgear faults based on decomposition components, comprising:

[0177] A data acquisition module, wherein the data acquisition module is used to acquire the decomposition product data of the SF6 switchgear and obtain standardized data after standardizing the acquired data;

[0178] A data processing module, the data processing module is used to determine the fault type of the SF6 switchgear according to the standardized data and a pre-constructed SF6 switchgear fault diagnosis model; during the construction of the SF6 switchgear fault diagnosis model, a radial basis function is selected as a kernel function, a support vector machine is selected as a base classifier, and a weighted principal component analysis method is used to realize feature extraction and dimensionality reduction of the standardized data, and an improved adaptive sparrow search algorithm is used to realize the optimization of the penalty parameters and kernel function parameters of the support vector machine; the improvement of the adaptive sparrow search algorithm includes a cat mapping initialization population, an explorer position update process, and an adaptive hybrid mutation.

[0179] In addition, a schematic diagram of a terminal device is provided in an embodiment of the present invention. The terminal device of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor executes the computer program, the functions of each module / unit in the above-mentioned device embodiments are implemented.

[0180] The computer program may be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to accomplish the present invention.

[0181] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0182] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0183] The memory may be used to store the computer programs and / or modules, and the processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory.

[0184] If the module / unit integrated in the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0185] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for SF6 switchgear fault diagnosis based on decomposition components, characterized in that: The following steps are involved: S1: Obtain decomposition product data of SF6 switchgear, and perform standardization processing on the obtained data to obtain standardized data; S2: determining the fault type of the SF6 switchgear according to the standardized data and a pre-built SF6 switchgear fault diagnosis model; In the process of constructing the SF6 switchgear fault diagnosis model, the radial basis function is selected as the kernel function, the support vector machine is selected as the base classifier, and the weighted principal component analysis method is used to extract features and reduce the dimension of the standardized data. The improved adaptive sparrow search algorithm is used to optimize the penalty parameters and kernel function parameters of the support vector machine; the improvement of the adaptive sparrow search algorithm includes the cat mapping initialization population, the explorer position update process and the adaptive hybrid mutation.

2. A method for SF6 switchgear fault diagnosis based on decomposition components according to claim 1, characterized in that: The obtained data is subjected to standardization processing, and the standardization function is: Where: x * is the data after standardization, x is the data before standardization, μ is the mean of the decomposition product data obtained; σ is the standard deviation of the original data.

3. The method for SF6 switchgear fault diagnosis based on decomposition components according to claim 1 is characterized in that: The weighted principal component analysis method is used to extract features from the standardized data. During the weighting process, the weighted covariance matrix is: SW=W T X T XW Where: S W is the weighted covariance data; W is an m×m diagonal weighted matrix, m is the number of features, and the diagonal elements of W are 1 / σ j Calculated, where 1 / σ j is the standard deviation of the jth feature; X is the input data set, where each column is a feature vector; T is the matrix transpose.

4. The method for SF6 switchgear fault diagnosis based on decomposition components according to claim 1 is characterized in that: The radial basis function is: K(x i ,x j )=exp(-γ||x i -x j || 2 ),γ>0 Among them, K(x i ,x j ) is the sample data after linear decomposition; x i and x j is the sample data; γ is the Gaussian kernel function.

5. The method for SF6 switchgear fault diagnosis based on decomposition components according to claim 1 is characterized in that: In the cat mapping initialization population, the cat mapping expression is: Where: a1 and b1 are arbitrary real numbers; mod1 means the decimal part of a1; w i and i is the position of the pixel in the grayscale image before transformation; The improved updating process of the explorer position is: in: is the j-dimensional position information of the i-th sparrow at the t-th iteration; iter max is the maximum number of iterations; α∈[0,1] is a random number; Q is a random number that follows a normal distribution; L is a matrix of shape 1×d, in which all elements are 1; R2 is the warning value, which is a random number in the range of [0,1]; ST∈[0.5,1] is the safety threshold.

6. The method for SF6 switchgear fault diagnosis based on decomposition components according to claim 1 is characterized in that: The adaptive hybrid mutation combines Gaussian mutation and random reverse learning, specifically: Gaussian mutation is performed when the current individual fitness is less than the average fitness, and random reverse learning is performed when the current individual fitness is greater than or equal to the average fitness.

7. A system for SF6 switchgear fault diagnosis based on decomposition components, characterized in that: include: A data acquisition module, wherein the data acquisition module is used to acquire the decomposition product data of the SF6 switchgear and obtain standardized data after standardizing the acquired data; A data processing module, the data processing module is used to determine the fault type of the SF6 switchgear according to the standardized data and a pre-constructed SF6 switchgear fault diagnosis model; during the construction of the SF6 switchgear fault diagnosis model, a radial basis function is selected as a kernel function, a support vector machine is selected as a base classifier, and a weighted principal component analysis method is used to realize feature extraction and dimensionality reduction of the standardized data, and an improved adaptive sparrow search algorithm is used to realize the optimization of the penalty parameters and kernel function parameters of the support vector machine; the improvement of the adaptive sparrow search algorithm includes a cat mapping initialization population, an explorer position update process, and an adaptive hybrid mutation.

8. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images