Gas insulated switch fault diagnosis method based on multi-source heterogeneous field self-adaption

Abstract

The application discloses a gas insulated switch fault diagnosis method based on multi-source heterogeneous field self-adaption, establishes a gas insulated switch experimental data set and a gas insulated switch field data set; multi-source dimension alignment is performed on the gas insulated switch experimental data set, so that the feature dimensions of each sample are kept consistent; the dimension-aligned gas insulated switch experimental data set and the gas insulated switch field data set are taken as a source domain and a target domain respectively, heterogeneous field self-adaption calculation is performed, a mapping matrix of each data set sample is obtained, and thus the fault samples from experiments and fields are mapped to a common subspace; a classifier is used to model the features of the common subspace, a gas insulated switch fault diagnosis model is built, and then the fault diagnosis of the gas insulated switch is realized. The application fully integrates the data features of the experimental fault samples and the field fault samples of the gas insulated switch, so as to solve the problem of the scarcity of the field fault samples, and thus the needs of the field fault diagnosis of the gas insulated switch are met.

Description

Technical Field

[0001] This invention belongs to the field of power equipment fault diagnosis technology, specifically relating to a fault diagnosis method for gas-insulated switches based on multi-source heterogeneous domain adaptive method. Background Technology

[0002] Gas-insulated circuit breakers are important devices in power systems, performing control and protection functions. Their condition affects the reliable operation of the power system, making rapid and accurate diagnosis of gas-insulated circuit breaker fault conditions crucial.

[0003] Sulfur hexafluoride (SF6) is widely used as the insulating medium for gas-insulated circuit breakers (GCBs) due to its stable chemical properties and excellent arc-quenching performance. When a fault exists inside a GCB, SF6 gas reacts chemically with trace amounts of water, oxygen, and solid insulating materials within the equipment under electrical or thermal influences, producing decomposition products such as SO2 and SOF2. The type and content of these decomposition products are closely related to the type and intensity of the fault. Therefore, SF6 decomposition products can serve as a basis for diagnosing faults in gas-insulated circuit breakers.

[0004] Currently, fault diagnosis methods for gas-insulated switches based on sulfur hexafluoride (SF6) decomposition products generate a large number of highly accurate fault samples through fault experiments on gas-insulated switches. These samples are then combined with mathematical statistics and machine learning methods to construct a diagnostic model, demonstrating high accuracy. However, instruments used for on-site detection of SF6 decomposition products have limited capabilities, detecting far fewer chemical types than experimental instruments. In such cases, the diagnostic model built using experimental samples is no longer applicable to SF6 decomposition products detected on-site, because supervised models require the dimensionality of the test samples to match that of the training model.

[0005] If a fault diagnosis model is built using field fault samples generated during actual operation, the number of field fault samples is relatively small, while most supervised models require sufficient fault samples to achieve adequate diagnostic accuracy. If experimental samples are used to supplement the field fault samples, the decomposition products of sulfur hexafluoride under different faults and the fact that field fault samples are more affected by complex operating environments compared to experimental samples will lead to problems with different data dimensions and data distributions when constructing a fault diagnosis model for gas-insulated switches by combining experimental samples and fault samples generated during actual operation. Summary of the Invention

[0006] The technical problem this invention aims to solve is to address the shortcomings of the existing technology by providing an adaptive gas-insulated switchgear fault diagnosis method based on multi-source heterogeneous domains. This method utilizes SF6 decomposition products to achieve fault diagnosis of gas-insulated switches operating in the field. This method effectively integrates experimental fault samples and field fault samples, solving the technical problem of inconsistent fault sample dimensions and distribution. It has the advantages of being fast and accurate, and meets the needs of online diagnosis.

[0007] The present invention adopts the following technical solution:

[0008] A fault diagnosis method for gas-insulated switches based on multi-source heterogeneous domain adaptation includes the following steps:

[0009] S1. Establish experimental datasets to construct the source domain H of the gas-insulated switch fault diagnosis model. s The target domain feature X of the gas-insulated switch fault diagnosis model is constructed by establishing a field dataset. t ;

[0010] S2, the source domain H obtained in step S1 s After preprocessing, we obtain a dimension of d. s Source domain features X s ;

[0011] S3. Set the number of iterations to t = 1, and the maximum number of iterations to t = T; Apply the source domain features X obtained in step S2... s The target domain features X obtained in step S1 t Perform heterogeneous domain adaptive computation to obtain the feature mapping matrix P of the source domain and the target domain. s and P t Through the feature mapping matrix P s and P t The features of the source and target domains are mapped to a common subspace of dimension d, and the mapped source and target domains are X′, respectively. s and X′ t ;

[0012] S4. Use a classifier to process the X′ obtained in step S3. s 、X′ t and the labels Y corresponding to the source and target domains s Y t Construct a fault diagnosis model for gas-insulated switches; calculate X′. s and X′ t Data distribution differences MMD, calculation of gas-insulated switch fault diagnosis model for X′ s and X′ t The classification loss CL;

[0013] S5. The number of iterations t = T. The fault diagnosis model corresponding to the minimum data distribution difference MMD and classification loss CL obtained in step S4 is taken as the optimal fault diagnosis model for gas-insulated switches.

[0014] S6. Use the gas-insulated switch fault diagnosis model obtained in step S5 to realize the fault diagnosis of the gas-insulated switch.

[0015] Specifically, in step S1, the experimental dataset includes the types and contents of SF6 decomposition products collected under typical fault experiments of gas-insulated switches; the field dataset includes the types and contents of SF6 decomposition products collected when faults occur in actual operation of gas-insulated switches.

[0016] Specifically, step S2 is as follows:

[0017] Divide the source domain fault samples, and let the aligned source domain X be... s The dimension is d s If each dimension of the sample is treated as a subsample, then The mapped subsamples have the following labels: By measuring the similarity between subsamples and within each subsample, the mapping matrix from each subsample to the common subspace is obtained. pass Implement source domain H s Dimensional alignment is performed to form the preprocessed source domain feature X. s .

[0018] Specifically, in step S3, the heterogeneous domain adaptive computation is as follows:

[0019] Let the dimension of the common subspace be d, and initialize the source domain X. s and target domain X t The sample weights are α and β; by establishing a generalized eigenvalue solving problem, heterogeneous domain adaptive computation is performed to obtain the feature mapping matrix P of the source domain and the target domain. s and P t The features of the source and target domains are mapped to a common subspace of dimension d using the mapping matrix, resulting in the mapped source and target domains, i.e., the mapped source domain X′. s =X s ×P s Target domain X′ t =X t ×P t Establish and solve the quadratic programming problem to obtain the updated sample weights α′ and β′.

[0020] Furthermore, we establish the following optimization problem: solving for the mapping matrix P between the source and target domains. s and Pt :

[0021]

[0022] Where mMMD(·) is the marginal distribution after mapping the source and target domains, and cMMD(·) is the conditional distribution after mapping the source and target domains.

[0023] Furthermore, the quadratic programming problem is as follows:

[0024]

[0025] in, For source domain X s The i-th sample; For the target domain X t The j-th sample; K ts For K st The transpose of .

[0026] Specifically, in step S4, the construction of the gas-insulated switch fault diagnosis model is as follows:

[0027] Merge X′ s 、X′ t As training samples, merge Y s Y t As training labels; set classifier parameters, including training sample weights, kernel function, and penalty parameters; where the sample weights are the weights α′ and β′ obtained in step S3; train the classifier to obtain the fault diagnosis model of the gas-insulated switch.

[0028] Specifically, in step S4, X′ is calculated. s and X′ t The data distribution differences in MMD are as follows:

[0029]

[0030] Calculate the fault diagnosis model of gas-insulated switch for X′ s and X′ t The classification loss CL is as follows:

[0031]

[0032] Among them, L[Y train f(X) train [)] is the classification loss function of the classifier, X train =[X′ s ;X′ t ], Y train =[Ys ;Y t ], where τ is the regularization parameter. Let be the mapping matrix of the source domain. Let be the feature matrix composed of source domain fault samples, and ns be the number of samples in the source domain; The mapping matrix for the target domain. Let n be the feature matrix composed of fault samples in the target domain. t C represents the number of samples in the target domain; C represents the total number of fault categories. These represent the number of samples corresponding to each type of label in the source and target domains, respectively. These are the fault sample matrices for the source and target domains under each label.

[0033] Specifically, step S6 is as follows:

[0034] Sample X of SF6 decomposition products to be diagnosed u Mapped to a common subspace, the gas-insulated switch fault diagnosis model obtained in step S5 is used to analyze the mapped sample X′ to be diagnosed. u Classification is performed to enable fault diagnosis of gas-insulated switches.

[0035] Secondly, embodiments of the present invention provide a gas-insulated switch fault diagnosis system based on multi-source heterogeneous domain adaptation, characterized in that it includes:

[0036] The data module establishes the source domain H of the experimental dataset for constructing a fault diagnosis model for gas-insulated switches. s The target domain feature X of the gas-insulated switch fault diagnosis model is constructed by establishing a field dataset. t ;

[0037] The preprocessing module processes the source domain H obtained from the data module. s After preprocessing, we obtain a dimension of d. s Source domain features X s ;

[0038] The mapping module sets the number of iterations to t=1 and the maximum number of iterations to t=T; it processes the source domain features X obtained from the preprocessing module. s The target domain features X obtained from the data module t Perform heterogeneous domain adaptive computation to obtain the feature mapping matrix P of the source domain and the target domain. s and P t Through the feature mapping matrix P s and P t The features of the source and target domains are mapped to a common subspace of dimension d, and the mapped source and target domains are X′, respectively. s and X′ t ;

[0039] The classification module uses a classifier to process the X′ obtained from the mapping module. s 、X′ t A fault diagnosis model for a gas-insulated switch is constructed using the labels Ys and Yt corresponding to the source and target domains; X′ is calculated. s and X′ t Data distribution differences MMD, calculation of gas-insulated switch fault diagnosis model for X′ s and X′ t The classification loss CL;

[0040] Select the module, iterate t=T times, and take the fault diagnosis model that minimizes the data distribution difference MMD and classification loss CL obtained by the classification module as the optimal fault diagnosis model for gas-insulated switches.

[0041] The diagnostic module utilizes the fault diagnosis model of the gas-insulated switch obtained from the selection module to achieve fault diagnosis of the gas-insulated switch.

[0042] Compared with the prior art, the present invention has at least the following beneficial effects:

[0043] A multi-source heterogeneous domain adaptive fault diagnosis method for gas-insulated switches (GES) is developed based on the decomposition characteristics of SF6 under GES faults and combined with a multi-source heterogeneous domain adaptive computation method to build a fault diagnosis model suitable for field-operated GES. First, a dataset for building the GES fault model is collected. The types and contents of SF6 decomposition products during GES faults in field operation are collected, along with the types and contents of SF6 decomposition products under different fault experiments, such as partial discharge, spark discharge, and arc discharge experiments. Then, the experimental datasets are dimensionally aligned to map fault samples in the experimental datasets to the same dimensional space. Heterogeneous domain adaptive computation is performed on the experimental and field datasets to solve the mapping matrix between them, mapping the two datasets to a common subspace, thus solving the problems of different fault sample dimensions and sample distributions between the datasets. A classifier is used to learn and model the mapped experimental and field fault samples, using the data distribution and classification loss of the fault samples as optimization objectives to obtain the optimal GES fault diagnosis model, achieving fault diagnosis of GES.

[0044] Furthermore, the SF6 decomposition products of gas-insulated circuit breakers are related to the fault type and fault intensity. For example, under discharge faults, SO2, SOF2, and SO2F2 are mainly produced; under partial discharge, the SO2F2 / SOF2 ratio is higher than under spark discharge, while the SO2F2 content is lower under arc discharge. The products formed by SF6 decomposition are also related to the fault temperature; for example, CO2, SO2, and H2S are produced at approximately 120℃, 300℃, and 360℃, respectively. Therefore, SF6 decomposition products can be used as a basis for judging gas-insulated circuit breaker faults and for establishing gas-insulated circuit breaker fault diagnosis models.

[0045] Furthermore, fault samples from gas-insulated switches operating in the field are collected to more closely reflect the actual operating conditions of gas-insulated switches; however, the number of field fault samples is limited. Therefore, by collecting the types and contents of SF6 decomposition products under different fault experiments, the training sample for establishing a fault diagnosis model for gas-insulated switches is expanded.

[0046] Furthermore, dimensional alignment is performed on the experimental fault dataset because the types of SF6 decomposition products vary under different fault conditions and detection conditions, resulting in different sample dimensions, which cannot be directly used as input for building a fault diagnosis model. Dimensional alignment considers the similarity between and within sub-samples of different dimensions to establish a generalized feature problem. The resulting mapping matrix for each sub-sample ensures that samples belonging to the same fault type are closer together, while samples belonging to different fault types are farther apart, facilitating subsequent training of the classifier.

[0047] Furthermore, the field fault dataset exhibits a significant difference in sample distribution compared to the experimental fault dataset due to the complex environment of actual operation. Additionally, the instruments used in the field to detect SF6 decomposition products have limited accuracy, resulting in fewer types of SF6 decomposition products compared to those obtained from the fault experiments. In other words, the sample dimensions differ between the two datasets. Therefore, this invention uses the experimental dataset, which generates a large number of fault labels, as the source domain, and the field dataset, with fewer fault labels, as the target domain. It fully integrates the features of each dataset through heterogeneous domain adaptive computation. During heterogeneous domain adaptive computation, a Laplace matrix is ​​introduced to ensure that the mapped source and target domains retain as many sample features as possible from the original fault data while minimizing domain offset. Moreover, considering the influence of gas pressure, adsorbent, and external environment on SF6 decomposition products, sample weights are assigned to the source and target domains when solving the mapping matrix. These weights are updated by solving a quadratic programming problem, thereby evaluating the confidence level of each fault sample and its corresponding fault label and improving the accuracy of the established gas-insulated switchgear fault diagnosis model.

[0048] Furthermore, the maximum mean difference (MMD) of the mapped fault samples and the classification loss (CL) of the classifier for the fault samples are used as optimization objectives. Cross-iteration calculations are employed to find the optimal fault sample mapping matrix, sample weights, and gas-insulated switch fault diagnosis model. The maximum mean difference (MMD) is a classic method for measuring the distance between data distributions, comprising marginal and conditional distributions. A smaller MMD indicates closer proximity between samples of the same fault type in the experimental and fault datasets, and vice versa. The classification loss (CL) is used to evaluate the effectiveness of the classifier in training the gas-insulated switch fault diagnosis model. The loss function is related to the classifier; for example, when the classifier is a support vector machine or softmax, the classification loss is calculated using the Hinge loss function and the cross-entropy function, respectively. A smaller CL indicates higher accuracy in the gas-insulated switch fault diagnosis model established by the classifier.

[0049] It is understandable that the beneficial effects of the second aspect mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0050] In summary, the present invention has the following advantages:

[0051] (1) By using methods such as dimension alignment and heterogeneous domain adaptive computing, the features of SF6 decomposition products from multiple sources are effectively integrated, thus avoiding the impact of scarce field fault samples on the establishment of gas-insulated switch fault models.

[0052] (2) Using SF6 decomposition products as the basis for judging the fault status of gas-insulated switches does not require disassembling the equipment or shutting down the power, and can meet the needs of online diagnosis.

[0053] (3) When establishing the fault diagnosis model for gas-insulated switches, issues such as sample weight, sample distribution, and classification loss were considered, which improved the accuracy of the fault diagnosis model.

[0054] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating the entire process of the method of the present invention;

[0056] Figure 2 This is a flowchart for dimensional alignment of the experimental dataset (source domain);

[0057] Figure 3 This is a flowchart for heterogeneous domain adaptive computation on experimental datasets (source domain) and field datasets (target domain);

[0058] Figure 4 A flowchart for building a fault diagnosis model for gas-insulated switches;

[0059] Figure 5 This is a flowchart for diagnosing faults in gas-insulated switches. Detailed Implementation

[0060] 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, not all, of the embodiments of the present invention. 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.

[0061] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0062] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0063] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this invention generally indicates that the preceding and following objects have an "or" relationship.

[0064] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0065] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0066] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0067] This invention provides a multi-source heterogeneous domain adaptive fault diagnosis method for gas-insulated switches. It collects the types and contents of SF6 decomposition products under various fault experiments of gas-insulated switches to establish an experimental dataset. It also collects the types and contents of SF6 decomposition products generated when gas-insulated switches experience faults during actual operation to establish a field dataset. The experimental dataset is aligned across multiple dimensions to ensure consistency in the feature dimensions of each sample. Domain adaptive calculation is performed on the experimental and field datasets to obtain a mapping matrix for each dataset. This mapping matrix maps the samples of each dataset to a common subspace. A classifier is used to model the features of the common subspace, building a fault diagnosis model for gas-insulated switches, thereby achieving fault diagnosis. This invention fully integrates the data features of experimental and field fault samples of gas-insulated switches to address the problem of scarce field fault samples, thus meeting the needs of field fault diagnosis for gas-insulated switches.

[0068] Please see Figure 1 This invention discloses a fault diagnosis method for gas-insulated switches based on multi-source heterogeneous domain adaptation, comprising the following steps:

[0069] S1. Establish experimental datasets and field datasets, and construct the source domain H of the gas-insulated switch fault diagnosis model respectively. s and target domain X t ;

[0070] The experimental dataset includes the types and contents of SF6 decomposition products collected under typical fault experiments of gas-insulated switches, such as partial discharge experiments, spark discharge experiments, arc discharge experiments, and overheating experiments; the field dataset includes the types and contents of SF6 decomposition products collected when faults occur in actual operation of gas-insulated switches.

[0071] S2, Source domain preprocessing, i.e., preprocessing the source domain H s Dimension alignment is performed to ensure that the feature dimensions of the SF6 decomposition products generated by each fault experiment are consistent. The preprocessed source domain is denoted as X. s The source domain X after dimension alignment s The dimension is d s ;

[0072] For source domain Hs Dimension alignment is performed because the types of decomposition products generated by SF6 under different faults vary, resulting in different dimensions of the fault samples. Specifically: First, the source domain fault samples are divided, and the aligned source domain is defined as X. s The dimension is d s Each dimension of the sample is treated as a subsample. By measuring the similarity between and within subsamples, the mapping matrix from each subsample to the common subspace is obtained. Thus through Implement source domain H s Dimensional alignment.

[0073] The principle is explained as follows:

[0074]

[0075] in, For subsamples Similarity matrix; For subsamples Similarity matrix; For subsamples and Similarity matrices between them.

[0076] Please see Figure 2 The specific steps for dimensional alignment of the source domain are as follows:

[0077] S201. Let X be the source domain after dimension alignment. s The dimension is d s If we divide the source domain samples and treat each dimension of the sample as a subsample, then... The corresponding fault sample label is Compare each subsample with the subsample with the largest dimension. Perform mapping calculations, where N is the number of dimensions contained in the source domain samples;

[0078] S202. Set the loop count k to 1;

[0079] S203, Solving for subsamples and Similarity matrices between like The i-th sample and If the fault labels of the j-th samples are the same, then otherwise

[0080] S204, Solving for subsamples Similarity matrix like If the fault labels of the i-th sample and the j-th sample are the same, then otherwise Constructing a similarity matrix diagonal matrix for Where N n For subsamples Number of samples; construct similarity matrix diagonal matrix for Where N N For subsamples The number of samples;

[0081] S205, Solving for subsamples Similarity matrix like If the fault labels of the i-th sample and the j-th sample are the same, then otherwise Constructing a similarity matrix diagonal matrix for Where NN represents the subsample Number of samples; construct similarity matrix diagonal matrix for Where N n For subsamples The number of samples;

[0082] S206, Evaluation Subsample In-sample and subsample Similarity, i.e. Evaluation Subsample In-sample and subsample Similarity, i.e.

[0083] S207, Subsample and The mapping matrices are respectively The problem is solved by establishing a generalized eigenvalue (as shown in equation (1)). Then the mapped subsamples Convert to in for The front d n Okay, d n With subsamples The dimensions are consistent; Λ is the largest d s The eigenvectors corresponding to each eigenvalue;

[0084]

[0085] S208. Determine if the number of iterations has reached the number of dimensions N of the source domain samples. If so, merge the sub-samples after mapping transformation. Forming the preprocessed source domain X s Otherwise, repeat step S203.

[0086] S3. Set the number of iterations to t = 1 and the maximum number of iterations to t = T; for the source domain features X s and target domain features X t Perform heterogeneous domain adaptive computation to obtain the feature mapping matrix P of the source domain and the target domain. s and P t The features of the source and target domains are mapped to a common subspace of dimension d using a mapping matrix. The mapped source and target domains are X′ and X′, respectively. s and X′ t ;

[0087] Please see Figure 3 Source Domain X s and target domain X t The corresponding fault label is Y s and Y t By analyzing the source domain X s and target domain X t X′ is obtained by performing heterogeneous domain adaptive computation. s and X′ t The specific steps are as follows:

[0088] S301. Let the dimension of the common subspace be d, and initialize the source domain X. s and target domain X t The sample weights are α and β;

[0089] S302. Adaptive computation of heterogeneous domain is performed by establishing a generalized eigenvalue solving problem (as shown in equation (2)) and solving the feature mapping matrix P. s and P t ;

[0090]

[0091] Where γ is the trade-off parameter for adaptive computation in heterogeneous domains; Λ is the eigenvector corresponding to the d largest eigenvalues;

[0092] Matrix V bs V bt V ws and V wt Equation (3) is used to ensure that, in the mapped source and target domain samples, samples with the same label are closer together, while samples with different labels are farther apart. Where L... bs and L wsL is the Laplace matrix of the source domain intrinsic features and penalized features; bt and L wt is the Laplace matrix of the inherent features and penalized features of the target domain.

[0093]

[0094] Matrix G ss G tt G st G is obtained through equation (4). ts For G st The transpose of . Where δ is the correction parameter;

[0095] ns is the number of samples in the source domain. This represents the number of samples corresponding to each type of label in the source domain. For various types of samples in the source domain The diagonal matrix generated by the unit vectors; J tt Similarly;

[0096]

[0097] S303, Solving equation (2) yields the feature mapping matrix P of the source and target domains. s and P t ;

[0098] The mapping matrix P between the source domain and the target domain s and P t :

[0099]

[0100] Where mMMD(·) is the marginal distribution after mapping the source and target domains, and cMMD(·) is the conditional distribution after mapping the source and target domains.

[0101] S304. Based on the mapping matrix, map the features of the source domain and the target domain to a common subspace of dimension d to obtain the mapped source domain and target domain, i.e., the mapped source domain X′. s =X s ×P s Target domain X′ t =X t ×P t ;

[0102] S305. Establish and solve the quadratic programming problem as shown in equation (5) to obtain the updated sample weights α′ and β′.

[0103]

[0104] in, For source domain X s The i-th sample; For the target domain X t The j-th sample; K ts For K st The transpose of .

[0105] S4. Use a classifier to classify X′ s 、X′ t and the labels Y corresponding to the source and target domains s Y t Build a fault diagnosis model for gas-insulated switches and calculate X′. s and X′ t Data distribution difference MMD, calculate fault diagnosis model for X′ s and X′ t The classification loss CL; if the number of iterations t = T, then proceed to step S5; otherwise, if t = t + 1, repeat step S3.

[0106] Different classification models can be selected for the classifier, such as support vector machine, random forest, softmax, etc.

[0107] Please see Figure 4 Taking support vector machine as an example, the specific steps for building a fault diagnosis model for gas-insulated switches are as follows:

[0108] S401, Merge X′ s 、X′ t As training samples, merge Y s Y t As training labels;

[0109] S402. Set the classifier parameters, including training sample weights, kernel function, penalty parameters, etc.; where the sample weights are the weights α′ and β′ obtained in step S305.

[0110] S403. Classifier training yields a fault diagnosis model for gas-insulated switches.

[0111] The calculation methods for MMD and CL are as follows:

[0112]

[0113]

[0114] Among them, L[Y train f(X) train [)] is the classification loss function of the classifier, X train =[X′s ;X′ t ], Y train =[Y s ;Y t ], namely the training samples and training labels mentioned in step S401; τ is the regularization parameter.

[0115] The data distribution difference (MMD) consists of marginal distribution difference (mMMD) and conditional distribution difference (cMMD), used to evaluate the domain shift effect between the source and target domains; the classification loss (CL) is used to evaluate the classifier's performance on the source domain X′. s and target domain X′ t The modeling effect.

[0116] S5. The fault diagnosis model corresponding to the minimum MMD and CL is taken as the final fault diagnosis model for gas-insulated switches. The fault samples used in establishing this model are source domain and target domain samples mapped to a common subspace, determined by the source domain mapping matrix P. s Mapping matrix P of the target domain t Therefore, the source domain fault sample is X′. s =X s ×P s The target domain fault sample is X′ t =X t ×P t ;

[0117] The smaller the MMD, the more original sample features of the dataset can be retained in the mapped source and target domains, and the closer the fault samples with the same fault label are. The smaller the CL, the better the modeling effect of the classifier. Therefore, the fault diagnosis model with the smallest MMD and CL is regarded as the optimal fault diagnosis model for gas-insulated switches.

[0118] S6. Sample X, the decomposition product of SF6 to be diagnosed. u Mapped to a common subspace, the gas-insulated switch fault diagnosis model is used to analyze the mapped sample X′ to be diagnosed. u Classification is performed to enable fault diagnosis of gas-insulated switches.

[0119] Please see Figure 5 The specific steps for diagnosing faults in gas-insulated switches are as follows:

[0120] S601, Input the SF6 decomposition product sample X to be diagnosed. u ;

[0121] S602. Map the sample to be diagnosed to a common subspace, i.e., X′. u =X u ×P t ;

[0122] S603, Using the gas-insulated switch fault diagnosis model to analyze X′ u Classification;

[0123] S604. Obtain the fault diagnosis results for the sample to be diagnosed.

[0124] In another embodiment of the present invention, a gas-insulated switch fault diagnosis system based on multi-source heterogeneous domain adaptation is provided. This system can be used to implement the above-mentioned gas-insulated switch fault diagnosis method based on multi-source heterogeneous domain adaptation. Specifically, the gas-insulated switch fault diagnosis system based on multi-source heterogeneous domain adaptation includes a data module, a preprocessing module, a mapping module, a classification module, a selection module, and a diagnosis module.

[0125] The data module establishes the source domain H of the experimental dataset to construct a fault diagnosis model for gas-insulated switches. s The target domain feature X of the gas-insulated switch fault diagnosis model is constructed by establishing a field dataset. t ;

[0126] The preprocessing module processes the source domain H obtained from the data module. s After preprocessing, we obtain a dimension of d. s Source domain features X s ;

[0127] The mapping module sets the number of iterations to t=1 and the maximum number of iterations to t=T; it processes the source domain features X obtained from the preprocessing module. s The target domain features X obtained from the data module t Perform heterogeneous domain adaptive computation to obtain the feature mapping matrix P of the source domain and the target domain. s and P t Through the feature mapping matrix P s and P t The features of the source and target domains are mapped to a common subspace of dimension d, and the mapped source and target domains are X′, respectively. s and X′ t ;

[0128] The classification module uses a classifier to process the X′ obtained from the mapping module. s 、X′ t and the labels Y corresponding to the source and target domains s Y t Construct a fault diagnosis model for gas-insulated switches; calculate X′. s and X′ t Data distribution differences MMD, calculation of gas-insulated switch fault diagnosis model for X′ s and X′ t The classification loss CL;

[0129] Select the module, iterate t=T times, and take the fault diagnosis model that minimizes the data distribution difference MMD and classification loss CL obtained by the classification module as the optimal fault diagnosis model for gas-insulated switches.

[0130] The diagnostic module utilizes the fault diagnosis model of the gas-insulated switch obtained from the selection module to achieve fault diagnosis of the gas-insulated switch.

[0131] In another embodiment of the present invention, a terminal device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used for the operation of a gas-insulated switch fault diagnosis method based on multi-source heterogeneous domain adaptation, including:

[0132] The source domain H of the experimental dataset is used to construct a fault diagnosis model for gas-insulated switches. s The target domain feature X of the gas-insulated switch fault diagnosis model is constructed by establishing a field dataset. t ; For source domain H s After preprocessing, we obtain a dimension of d. s Source domain features X s ;

[0133] Set the number of iterations to t = 1 and the maximum number of iterations to t = T; for the source domain features X s and target domain features X t Perform heterogeneous domain adaptive computation to obtain the feature mapping matrix P of the source domain and the target domain. s and P t Through the feature mapping matrix P s and P t The features of the source and target domains are mapped to a common subspace of dimension d, and the mapped source and target domains are X′, respectively. s and X′t ; Use a classifier to classify X′ s 、X′ t and the labels Y corresponding to the source and target domains s Y t Construct a fault diagnosis model for gas-insulated switches; calculate X′. s and X′ t Data distribution differences MMD, calculation of gas-insulated switch fault diagnosis model for X′ s and X′ t The classification loss CL is calculated; the number of iterations t = T is used to determine the optimal fault diagnosis model for gas-insulated switches, which is the fault diagnosis model corresponding to the minimum data distribution difference MMD and classification loss CL; the fault diagnosis of gas-insulated switches is achieved using the gas-insulated switch fault diagnosis model.

[0134] In another embodiment of the present invention, a storage medium is also provided, specifically a computer-readable storage medium (memory). This computer-readable storage medium is a memory device in a terminal device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and extended storage media supported by the terminal device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device.

[0135] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the gas-insulated switch fault diagnosis method based on multi-source heterogeneous domain adaptation in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by the processor to perform the following steps:

[0136] Set the number of iterations to t = 1 and the maximum number of iterations to t = T; for the source domain features Xs and the target domain features X... t Perform heterogeneous domain adaptive computation to obtain the feature mapping matrix P of the source domain and the target domain. s and P t Through the feature mapping matrix P s and P t The features of the source and target domains are mapped to a common subspace of dimension d, and the mapped source and target domains are X′, respectively. s and X′ t; Use a classifier to classify X′ s 、X′ t and the labels Y corresponding to the source and target domains s Y t Construct a fault diagnosis model for gas-insulated switches; calculate X′. s and X′ t Data distribution differences MMD, calculation of gas-insulated switch fault diagnosis model for X′ s and X′ t The classification loss CL is calculated; the number of iterations t = T is used to determine the optimal fault diagnosis model for gas-insulated switches, which is the fault diagnosis model corresponding to the minimum data distribution difference MMD and classification loss CL; the fault diagnosis of gas-insulated switches is achieved using the gas-insulated switch fault diagnosis model.

[0137] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. 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 claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0138] The invention will be explained below using field fault data from circuit breakers as an example. Field fault descriptions are summarized, and typical insulation faults in gas-insulated switches—partial discharge, arc discharge, and local overheating—are taken as diagnostic targets. SF6 decomposition products generated under fault experiments and field faults are collected as datasets and inputs to the fault diagnosis model. After processing, the experimental dataset and the field dataset contain 280 sets of experimental fault samples and 76 sets of field fault samples, respectively. The dimensions covered by each dataset and the types of SF6 decomposition products are shown in Table 1.

[0139] Table 1 Overview of the Fault Dataset

[0140]

[0141]

[0142] As shown in Table 1, due to different experimental conditions, the dimensionality of the SF6 decomposition products under various faults monitored in the experimental datasets is not entirely the same, but the SF6 decomposition products obtained from each group of experiments are within the same feature space. There are also differences in data distribution and dimensionality between the experimental datasets and the field datasets.

[0143] The field dataset was divided into target domain X at an 8:2 ratio. t and the sample to be diagnosed X u . Reference Figure 1 The iteration count is set to 50. Assuming the dimension of the dimension-aligned experimental dataset is 6, we first perform dimension alignment on the experimental dataset. Next, assuming the dimension of the common subspace is 3, after heterogeneous domain adaptive computation, the experimental dataset X... s and the on-site dataset X t Mapped to X′ respectively s =X s ×P s and X′ t =X t ×P t Support vector machines were used to analyze the experimental dataset X′ in the common subspace. s and the on-site dataset X′ t Fault classification modeling is performed, and the data distribution difference and classification loss under each iteration are calculated according to equations (6) and (7). After the iteration is completed, the fault diagnosis model with the smallest data distribution difference and classification loss is taken as the final fault diagnosis model of the gas-insulated switch. At this time, the original fault sample (as shown in equation (8)) is transformed into equation (9).

[0144]

[0145]

[0146] Experimental dataset X s and the on-site dataset X t The mapping matrix P s and P t As shown in equations (10) and (11) respectively.

[0147]

[0148]

[0149] Then X′ u =X u ×P t The final gas-insulated switch fault diagnosis model was used to analyze X′. u The classification accuracy of the samples to be diagnosed was 88.24% in the test.

[0150] In summary, the present invention provides an adaptive gas-insulated switch fault diagnosis method based on multi-source heterogeneous domain, which can achieve fault diagnosis of gas-insulated switches in operation in the field with good accuracy.

[0151] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments 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. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0152] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0153] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0154] In the embodiments provided by this invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For instance, the division of modules or 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 through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0155] 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.

[0156] 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.

[0157] If the integrated module / 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, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed 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 electrical carrier signals and telecommunication signals.

[0158] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0159] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0160] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0161] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A fault diagnosis method for gas-insulated switches based on multi-source heterogeneous domain adaptation, characterized in that, Includes the following steps: S1. Establish experimental dataset to construct the source domain of the gas-insulated switch fault diagnosis model. Establishing field datasets to construct target domain features for gas-insulated switch fault diagnosis models The experimental dataset includes the types and contents of SF6 decomposition products collected under typical fault experiments of gas-insulated switches; the field dataset includes the types and contents of SF6 decomposition products collected when faults occur in actual operation of gas-insulated switches. S2. The source domain obtained in step S1 Preprocessing is performed to obtain the dimension as Source domain features ; S3. Set the number of iterations to t=1 and the maximum number of iterations to t=T; process the source domain features obtained in step S2. and the target domain features obtained in step S1 Perform heterogeneous domain adaptive computation to obtain the feature mapping matrices of the source and target domains. and Through the feature mapping matrix and Map the features of the source and target domains to a dimension of... The common subspace, the mapped source domain and target domain are respectively and ; S4. Use a classifier to process the results obtained in step S3. , and the tags corresponding to the source and target domains , Build a fault diagnosis model for gas-insulated switches; calculate... and Data distribution differences MMD, calculation of gas-insulated switch fault diagnosis model and The classification loss CL; S5. The number of iterations t=T is used to determine the fault diagnosis model that minimizes the data distribution difference MMD and classification loss CL obtained in step S4. S6. Use the gas-insulated switch fault diagnosis model obtained in step S5 to realize the fault diagnosis of the gas-insulated switch.

2. The gas-insulated switch fault diagnosis method based on multi-source heterogeneous domain adaptation according to claim 1, characterized in that, Step S2 is as follows: Divide the source domain fault samples, and set the aligned source domain. The dimension is If each dimension of the sample is treated as a subsample, then , For the mapped subsamples, the corresponding fault sample label vector is: , Assuming faulty sample labels, a mapping matrix is ​​obtained by measuring the similarity between and within subsamples, and mapping each subsample to a common subspace. ,pass Realize source domain Dimensional alignment is used to form preprocessed source domain features. .

3. The gas-insulated switch fault diagnosis method based on multi-source heterogeneous domain adaptation according to claim 1, characterized in that, In step S3, the heterogeneous domain adaptive computation specifically involves: Let the dimension of the common subspace be . Initialize the source domain and target domain The sample weights are and By establishing a generalized eigenvalue solving problem, adaptive computation is performed in heterogeneous domains to obtain the feature mapping matrices of the source and target domains. and Based on the mapping matrix, the features of the source and target domains are mapped to a dimension of... From the common subspace, we obtain the mapped source domain and target domain, i.e., the mapped source domain. Target domain Establish and solve a quadratic programming problem to obtain the updated sample weights. and .

4. The gas-insulated switch fault diagnosis method based on multi-source heterogeneous domain adaptation according to claim 3, characterized in that, Establish the following optimization problem: solve for the mapping matrix between the source and target domains. and : in, The edge distribution after mapping the source and target domains. This represents the conditional distribution after mapping the source and target domains.

5. The gas-insulated switch fault diagnosis method based on multi-source heterogeneous domain adaptation according to claim 3, characterized in that, The quadratic programming problem is as follows: in, , For source domain The i-th sample; , For the target domain The j-th sample; Let be the mapping matrix of the source domain; The mapping matrix for the target domain; .

6. The gas-insulated switch fault diagnosis method based on multi-source heterogeneous domain adaptation according to claim 1, characterized in that, In step S4, the specific steps for building the fault diagnosis model for gas-insulated switches are as follows: merge , As training samples, merge , As training labels; Set the classifier parameters, including training sample weights, kernel function, and penalty parameters; where the sample weights are the weights obtained in step S3. and The classifier was trained to obtain a fault diagnosis model for the gas-insulated switch.

7. The gas-insulated switch fault diagnosis method based on multi-source heterogeneous domain adaptation according to claim 1, characterized in that, In step S4, calculate and The data distribution differences in MMD are as follows: Calculation of gas-insulated switch fault diagnosis model and The classification loss CL is as follows: in, Let be the classification loss function of the classifier. ; , ; , For regularization parameters, Let be the mapping matrix of the source domain. The feature matrix is ​​composed of source domain fault samples. The number of samples in the source domain; The mapping matrix for the target domain. The feature matrix consists of fault samples in the target domain. The number of samples in the target domain; This represents the total number of fault categories. , These represent the number of samples corresponding to each type of label in the source and target domains, respectively. , These are the fault sample matrices for the source and target domains under each label.

8. The gas-insulated switch fault diagnosis method based on multi-source heterogeneous domain adaptation according to claim 1, characterized in that, Step S6 is as follows: Samples of SF6 decomposition products to be diagnosed Mapped to a common subspace, the gas-insulated switch fault diagnosis model obtained in step S5 is used to analyze the mapped samples to be diagnosed. Classification is performed to achieve fault diagnosis of gas-insulated switches.

9. A fault diagnosis system for gas-insulated switches based on multi-source heterogeneous domain adaptation, characterized in that, include: The data module establishes the source domain for constructing a fault diagnosis model for gas-insulated switches by building an experimental dataset. Establishing field datasets to construct target domain features for gas-insulated switch fault diagnosis models The experimental dataset includes the types and contents of SF6 decomposition products collected under typical fault experiments of gas-insulated switches; the field dataset includes the types and contents of SF6 decomposition products collected when faults occur in actual operation of gas-insulated switches. The preprocessing module processes the source domain obtained from the data module. Preprocessing is performed to obtain the dimension as Source domain features ; The mapping module sets the number of loop iterations to t=1 and the maximum number of loop iterations to t=T. Source domain features obtained from the preprocessing module Target domain features obtained from the data module Perform heterogeneous domain adaptive computation to obtain the feature mapping matrices of the source and target domains. and Through the feature mapping matrix and Map the features of the source and target domains to a dimension of... The common subspace, the mapped source domain and target domain are respectively and ; The classification module uses a classifier to process the data obtained from the mapping module. , and the tags corresponding to the source and target domains , Build a fault diagnosis model for gas-insulated switches; calculate... and Data distribution differences MMD, calculation of gas-insulated switch fault diagnosis model and The classification loss CL; Select the module, iterate t=T times, and take the fault diagnosis model that minimizes the data distribution difference MMD and classification loss CL obtained by the classification module as the optimal fault diagnosis model for gas-insulated switches. The diagnostic module utilizes the fault diagnosis model of the gas-insulated switch obtained from the selection module to achieve fault diagnosis of the gas-insulated switch.

Images