A disconnector fault diagnosis method and system

Abstract

The application belongs to the field of mechanical fault diagnosis of disconnectors, and provides a disconnector fault diagnosis method and system. The method comprises obtaining real-time vibration signals of the disconnector, and sequentially performing singular value filtering noise reduction and hybrid modal decomposition processing on the vibration signals to extract feature vectors; and performing fault diagnosis on the disconnector according to the feature vectors and a deep weighted fusion model based on an SVM classifier; wherein the construction process of the deep weighted fusion model based on the SVM classifier is as follows: obtaining an initial weak SVM classifier according to the feature vectors of the disconnector vibration signals and a linear support vector machine with a Gaussian kernel as an initial kernel function; and iteratively optimizing the initial weak SVM classifier by means of a deep fusion weighting algorithm and optimal allocation of the weights of the disconnector vibration signal samples to obtain an SVM classifier satisfying a preset condition and serving as the deep weighted fusion model based on the SVM classifier.

Description

Technical Field

[0001] This invention belongs to the field of mechanical fault diagnosis of disconnecting switches, and particularly relates to a method and system for diagnosing faults in disconnecting switches. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] A failure of a disconnecting switch can jeopardize the safe and stable operation of the power system. Mechanical faults account for a large proportion of these failures, thus necessitating effective sensing and fault diagnosis of the disconnecting switch's mechanical condition. Accurate and reliable fault diagnosis methods are required for rapid and accurate identification of the disconnecting switch's status information and for effective fault diagnosis.

[0004] During the operation of disconnecting switches, the movement of a series of mechanical components, such as the disconnector, during opening and closing generates a large amount of vibration signals on the surface of the operating mechanism housing. Currently, analyzing the disconnecting switch status based on vibration signals is an important means of fault diagnosis. Using artificial intelligence learning methods for fault classification has become mainstream, with existing methods including neural network learning, deep coding learning, and support vector machines (SVM). However, the inventors discovered that SVM lacks reliable generalization performance when faced with highly random samples, resulting in less than ideal classification performance for disconnecting switch faults. Summary of the Invention

[0005] In order to solve the technical problems existing in the background art, the present invention provides a method and system for diagnosing disconnecting switches, which can improve the fault diagnosis effect of disconnecting switches.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] The first aspect of the present invention provides a method for diagnosing faults in disconnecting switches, comprising:

[0008] The real-time vibration signal of the disconnector switch is acquired, and singular value filtering and mixed mode decomposition are performed on it in sequence to extract the feature vector.

[0009] Based on the feature vector and the deep weighted fusion model based on SVM classifier, fault diagnosis is performed on the disconnecting switch;

[0010] The construction process of the deep weighted fusion model based on the SVM classifier is as follows:

[0011] Based on the feature vector of the vibration signal of the disconnector switch and the linear support vector machine with Gaussian kernel as the initial kernel function, an initial weak SVM classifier is obtained;

[0012] The initial weak SVM classifier is iteratively optimized by using a deep fusion weighting algorithm and optimizing the weight allocation of vibration signal samples from disconnecting switches. An SVM classifier that meets the preset conditions is then obtained and used as a deep weighted fusion model based on the SVM classifier.

[0013] As one implementation method, the fault diagnosis results of the disconnecting switch include normal operating conditions, loose screw conditions, and stuck connecting rod conditions.

[0014] As one implementation method, in the process of optimizing the weight allocation of vibration signal samples from disconnecting switches, a loss interval is used to measure the weight allocation of each classification sample by the current deep fusion weighting algorithm.

[0015] As one implementation method, a loss interval greater than zero indicates that the base classifier correctly classifies more samples than it misclassifies.

[0016] As one implementation method, in the process of optimizing the allocation of weights for vibration signal samples from disconnecting switches, the samples are divided into the following four categories after the weight update strategy is adjusted:

[0017] a) Samples that were correctly classified by the previous strong classifier but were misclassified by the current strong classifier;

[0018] b) Samples that were correctly classified by the previous strong classifier and are also correctly classified by the current strong classifier;

[0019] c) Samples that were misclassified by the previous strong classifier and are also misclassified by the current strong classifier;

[0020] d represents samples that were misclassified by the previous strong classifier but were correctly classified by the current strong classifier.

[0021] A second aspect of the present invention provides a fault diagnosis system for disconnecting switches, comprising:

[0022] The feature extraction module is used to acquire the real-time vibration signal of the disconnector switch, and then perform singular value filtering and noise reduction and mixed mode decomposition on it in sequence to extract the feature vector;

[0023] The fault diagnosis module is used to diagnose faults in the disconnector based on the feature vector and the deep weighted fusion model based on the SVM classifier.

[0024] The construction process of the deep weighted fusion model based on the SVM classifier is as follows:

[0025] Based on the feature vector of the vibration signal of the disconnector switch and the linear support vector machine with Gaussian kernel as the initial kernel function, an initial weak SVM classifier is obtained;

[0026] The initial weak SVM classifier is iteratively optimized by using a deep fusion weighting algorithm and optimizing the weight allocation of vibration signal samples from disconnecting switches. An SVM classifier that meets the preset conditions is then obtained and used as a deep weighted fusion model based on the SVM classifier.

[0027] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the disconnector switch fault diagnosis method described above.

[0028] A fourth aspect of the present invention provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the disconnector switch fault diagnosis method described above.

[0029] Compared with the prior art, the beneficial effects of the present invention are:

[0030] This invention iteratively optimizes the initial weak SVM classifier by using a deep fusion weighting algorithm and optimizing the weight allocation of vibration signal samples from disconnecting switches. By increasing the weight of weak classifiers with low classification error rates, they play a greater role in the voting process; conversely, by decreasing the weight of weak classifiers with high classification error rates, they play a smaller role in the voting process. This improves the robustness of the SVM classifier and thus enhances the accuracy of disconnecting switch fault classification.

[0031] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0032] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0033] Figure 1 This is a flowchart of the disconnector switch fault diagnosis method according to an embodiment of the present invention;

[0034] Figure 2 This is a time-domain diagram of the vibration signal of the disconnecting switch according to an embodiment of the present invention;

[0035] Figure 3 This is the recognition result of the DFW-SVM classifier in an embodiment of the present invention. Detailed Implementation

[0036] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0037] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0038] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0039] Example 1

[0040] Reference Figure 1 This embodiment provides a method for diagnosing faults in disconnecting switches, which includes:

[0041] S101: Acquire the real-time vibration signal of the disconnector switch, and sequentially perform singular value filtering and noise reduction and mixed mode decomposition processing on it to extract feature vectors;

[0042] S102: Based on the feature vector and the deep weighted fusion model based on the SVM classifier, perform fault diagnosis on the disconnect switch;

[0043] The construction process of the deep weighted fusion model based on the SVM classifier is as follows:

[0044] Based on the feature vector of the vibration signal of the disconnector switch and the linear support vector machine with Gaussian kernel as the initial kernel function, an initial weak SVM classifier is obtained;

[0045] The initial weak SVM classifier is iteratively optimized by using a deep fusion weighting algorithm and optimizing the weight allocation of vibration signal samples from disconnecting switches. An SVM classifier that meets the preset conditions is then obtained and used as a deep weighted fusion model based on the SVM classifier.

[0046] Among them, the deep weighted fusion model based on SVM classifier is particularly suitable for small-sample fault classification. The probability of mechanical failure in disconnect switches during actual operation is not high, so the number of fault samples obtained through vibration signals is small. This model performs well in this case. Furthermore, disconnect switches are usually not equipped with separate relay protection devices. Using this model as a diagnostic algorithm combined with the mechanical vibration signals of the disconnect switch, a unique protection device for the disconnect switch can be constructed, filling the gap in non-invasive intelligent diagnosis of disconnect switches.

[0047] In practical implementation, the construction process of the deep weighted fusion model based on the SVM classifier is as follows:

[0048] Step 1: Input the feature vector of the obtained disconnector switch vibration signal into a linear support vector machine with an initial kernel function of Gaussian kernel and a penalty factor C of 9 to obtain an initial weak classifier;

[0049] Suppose the input training sample set T = {(x1,y1),(x2,y2),…,(x...} N ,y N )}, where x i ∈R n Let y be the sample feature vector. i ∈{+1,-1} is the corresponding label, i=1,2,…N;

[0050] Output using a separating hyperplane and a classification decision function:

[0051] (1) Select the penalty parameter C>0, construct and solve the convex quadratic programming problem.

[0052]

[0053] Where, α i α j Represents the Lagrange multipliers;

[0054] Find the optimal solution to the above quadratic programming problem.

[0055] (2) Calculate the optimal plane vector ω * and b *

[0056] Choose α * a certain component Meet the conditions calculate

[0057]

[0058] (3) Solving for the separating hyperplane

[0059] ω * ·x+b * =0 (3)

[0060] The selected classification decision function is The kernel function chosen is the Gaussian kernel function. Where υ is the velocity parameter, which is usually taken as 1.

[0061] Step 2: Iteratively optimize the SVM classifier using the Deep Fusion Weighted (DFW) algorithm to obtain a strong classifier.

[0062] The DFW algorithm is a deep learning algorithm designed to improve the classification ability of weak classifiers. It increases the weights of samples misclassified by the previous weak classifier and decreases the weights of correctly classified samples. This way, incorrectly classified data receives more attention from subsequent weak classifiers due to their increased weight. In strengthening weak classifiers into strong classifiers, DFW employs a weighted majority voting method. Specifically, it increases the weights of weak classifiers with lower classification error rates, giving them a greater role in the voting process, while decreasing the weights of weak classifiers with higher classification error rates, giving them a smaller role. This weighted iteration is repeated to obtain the reinforcement-learned strong classifier, DFW-SVM.

[0063] The DFW-SVM algorithm process is as follows:

[0064] a. Given a training set: S = {(x1,y1),(x2,y2),…,(x...} i ,y i ),…,(x m ,y m )}, where x i It is an instance sample: y i It is a class sample, and y i ∈Y={-1,+1}, where m represents the number of training samples.

[0065] b. Initialize the weight distribution of the training data:

[0066]

[0067] c. Using a weighted distribution D m The basic classifier G is obtained by learning from the training dataset. m (x), then calculate G m (x) Classification error rate on the training sample set:

[0068]

[0069]

[0070] d. Calculate G m (x) Internal classification error parameter β m :

[0071]

[0072] Where e m To iterate the classification error rate, update the weight distribution of the training dataset:

[0073]

[0074]

[0075] Z m It is a normalization factor, which can make G m+1 Transform it into a probability distribution.

[0076] e. Construct a linear combination of the base classifiers SVM to obtain the final classifier:

[0077]

[0078]

[0079] Here, sign(f(x)) represents the sign function.

[0080] Step 3: Propose a self-updating strategy to enhance the weights of the DFW algorithm and optimize the weight allocation of the DFW algorithm.

[0081] (1) The loss function of the DFW algorithm can be approximated as a function that maximizes the non-standardized mean of the intervals and minimizes the variance of the non-standardized interval distribution:

[0082]

[0083] Where, σ 2 ξ represents the variance of the non-standardized sample loss interval; ξ represents the mean of the non-standardized sample loss interval.

[0084] The loss function is used to model the samples using standardized intervals, and its statistical definition is as follows:

[0085]

[0086] Where T represents the total number of non-standardized sample loss intervals; G t (x i Let $\mathbf{t}$ represent the base classifier corresponding to the loss margin of the $t$-th unstandardized sample. The loss margin reflects the classification of each sample by the base classifier, and a margin greater than zero indicates that the base classifier correctly classifies more samples than it misclassifies. The DFW algorithm can be viewed as adjusting the loss margin of samples to shift the sample margins towards a direction greater than zero, thereby continuously reducing the sample error rate. As the DFW algorithm iterates, it continuously increases the number of samples with negative margin increments, so that the current classifier pays more attention to samples misclassified in the previous round, causing the training sample margins to continuously move positively. Therefore, using the loss margin can measure the current algorithm's weight allocation for each class of samples. For classification cases where the error rate is already high, calculating the loss margin can better allocate the weights of each classifier.

[0087] (2) The weight self-updating strategy is a specific implementation algorithm for optimizing the weight allocation of the DFW algorithm based on the loss interval theory. In this algorithm, the updated strategy of the sample weights is divided into the following four categories after adjustment:

[0088] a) Samples that were correctly classified by the previous strong classifier but were misclassified by the current strong classifier.

[0089] b) Samples that were correctly classified by the previous strong classifier and are also correctly classified by the current strong classifier.

[0090] c) Samples that were misclassified by the previous strong classifier and are also misclassified by the current strong classifier.

[0091] d represents samples that were misclassified by the previous strong classifier but were correctly classified by the current strong classifier.

[0092] According to fractal theory, the interval in case a) is larger than that in case c), but the interval in case a) will gradually decrease. To suppress the negative trend of the interval, the sample weights in case a) should increase by a larger margin than those in case c). Similarly, the interval in case 2) should be larger than that in case 4). To ensure the continued increase of the interval in case 2), the sample weights in case 4) should increase by a larger margin than those in case 2). The sample weight update strategy of the enhanced weight self-update algorithm will be given below.

[0093] Similar to step a) of the DFW algorithm, given a training sample set, initialize the sample weights.

[0094] Similar to step c) of the DFW algorithm, after training the basic classifier, the weights of the basic classifier are assigned based on the error rate.

[0095] The self-updating strategy of the enhanced weight self-updating algorithm is as follows:

[0096] remember H t (x i ) represents the strong classifier in round t.

[0097] like The sample weights are then updated as follows:

[0098] W t+1 (x i ) = W t (x i )β t exp{(1-ξ 2 )G t (x i )-3(-1+σ 2 )H t (x i )y i} / margin i (14)

[0099] like but:

[0100] W t+1 (x i ) = W t (x i )β t exp{(1+ξ 2 )G t (x i )-(1+σ 2 )H t (x i )y i} / margin i (15)

[0101] like but:

[0102] W t+1 (x i ) = W t (x i )β t exp{(1-ξ 2 )G t (x i )-3(-1+σ2)H t (x i )y i} / margin i (16)

[0103] like but:

[0104] W t+1 (x i ) = W t (x i )β t exp{(1+ξ 2 )G t (x i )+(1+σ 2 )H t (x i )y i} / margin i (17)

[0105] After optimizing the weight allocation using the weight-reinforcing self-update algorithm, the final strong classifier is obtained:

[0106]

[0107] The following is a specific example. A total of 90 sets of mechanical vibration data from disconnecting switches were collected through experiments and simulations: 30 sets under normal operating conditions, 30 sets under loose screw conditions, and 30 sets under connecting rod jamming conditions. The DFW-SVM algorithm was used to identify the fault states based on the collected vibration data. The obtained time-domain diagram of the original vibration signal of the disconnecting switch under normal operating conditions is shown below. Figure 2 As shown.

[0108] To improve the accuracy and effectiveness of fault identification, the energy entropy and commonly used components such as the root mean square value and variance of the IMF envelope in the time-frequency domain were input into the DFW-SVM classifier as feature vectors. A total of 90 sets of feature vectors were generated, as shown in Table 1. Sets 1-30 represent the normal state of the disconnector switch, sets 31-60 represent the loose state of the auxiliary switch screws, and sets 61-90 represent the stuck state of the connecting rod. The category labels for the normal, loose, and stuck states of the disconnector switch were denoted as 1, 2, and 3, respectively. From the 90 sets of feature data, 60 sets (20 sets of normal disconnector switch data, 20 sets of loose disconnector switch data, and 20 sets of stuck disconnector switch data) were selected for training, and the remaining 30 sets were used for testing.

[0109] Table 1 Test Set Samples

[0110] Disconnect switch status type serial number Label Normal state 1-30 1 Auxiliary switch screw loose state 31-60 2 Linkage jamming 61-90 3

[0111] Depend on Figure 3 It can be seen that all 10 sets of connecting rod jamming fault signals were correctly classified, while one misclassification occurred in each of the 10 sets of loose screw fault signals and the 10 sets of normal operating condition signals. Overall, the diagnostic accuracy rate reached over 93%, achieving a satisfactory diagnostic result.

[0112] To verify that the proposed DFW-SVM classifier has better generalization performance and reliability, the same set of data was input into a linear support vector machine (SVM) and a DFW-SVM classifier with optimized weight ratios through a weight self-update strategy for state recognition. The above experiments were repeated ten times, and the state recognition accuracy of each classifier is statistically shown in Table 2.

[0113] Table 2 Comparison of recognition accuracy for each category status

[0114] algorithm Accuracy (%) SVM 84.5 DFW-SVM 93.4

[0115] The above results demonstrate that the DFW-SVM-based fault diagnosis method for disconnecting switches can effectively identify and process the original vibration signals, thereby enabling intelligent identification of the operating status of the disconnecting switches. Compared with other algorithms, the model exhibits superior and stable performance, higher accuracy, and greater reliability. This algorithm provides a basis for fault diagnosis of disconnecting switches and serves as a reference for maintenance personnel.

[0116] Example 2

[0117] This embodiment provides a fault diagnosis system for disconnecting switches, which includes the following modules:

[0118] The feature extraction module is used to acquire the real-time vibration signal of the disconnector switch, and then perform singular value filtering and noise reduction and mixed mode decomposition on it in sequence to extract the feature vector;

[0119] The fault diagnosis module is used to diagnose faults in the disconnector based on the feature vector and the deep weighted fusion model based on the SVM classifier.

[0120] The construction process of the deep weighted fusion model based on the SVM classifier is as follows:

[0121] Based on the feature vector of the vibration signal of the disconnector switch and the linear support vector machine with Gaussian kernel as the initial kernel function, an initial weak SVM classifier is obtained;

[0122] The initial weak SVM classifier is iteratively optimized by using a deep fusion weighting algorithm and optimizing the weight allocation of vibration signal samples from disconnecting switches. An SVM classifier that meets the preset conditions is then obtained and used as a deep weighted fusion model based on the SVM classifier.

[0123] It should be noted that each module in this embodiment corresponds one-to-one with each step in Embodiment 1, and their specific implementation processes are the same, so they will not be repeated here.

[0124] Example 3

[0125] This embodiment provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps in the disconnector switch fault diagnosis method described above.

[0126] Example 4

[0127] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the disconnector switch fault diagnosis method described above.

[0128] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as 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.

[0129] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for diagnosing faults in disconnecting switches, characterized in that, include: The real-time vibration signal of the disconnector switch is acquired, and singular value filtering and mixed mode decomposition are performed on it in sequence to extract the feature vector. Based on the feature vector and the deep weighted fusion model based on SVM classifier, fault diagnosis is performed on the disconnecting switch; The construction process of the deep weighted fusion model based on the SVM classifier is as follows: Based on the feature vector of the vibration signal of the disconnector switch and the linear support vector machine with Gaussian kernel as the initial kernel function, an initial weak SVM classifier is obtained; The initial weak SVM classifier is iteratively optimized by using a deep fusion weighting algorithm and optimizing the weight allocation of vibration signal samples from disconnecting switches. A strong SVM classifier that meets the preset conditions is then obtained and used as a deep weighted fusion model based on the SVM classifier. Among them, the fault diagnosis results of the disconnecting switch include normal operating conditions, loose screw conditions, and stuck connecting rod conditions; In the process of optimizing the allocation of weights for vibration signal samples from disconnecting switches, the samples are divided into the following four categories after the weight update strategy is adjusted: a) Samples that were correctly classified by the previous strong classifier but were misclassified by the current strong classifier; b) Samples that were correctly classified by the previous strong classifier and are also correctly classified by the current strong classifier; c) Samples that were misclassified by the previous strong classifier and are also misclassified by the current strong classifier; d) Samples that were misclassified by the previous strong classifier but were correctly classified by the current strong classifier; The loss function of the deep fusion weighted algorithm is approximated as a function that maximizes the non-standardized mean of the margins and minimizes the variance of the non-standardized margin distribution: ； in, The variance of the non-standardized sample loss interval is represented. This represents the mean of the loss interval for non-standardized samples; The loss function is used to model the samples using standardized intervals, and its statistical definition is as follows: ； Where T represents the total number of non-standardized sample loss intervals; Let represent the base classifier corresponding to the loss margin of the t-th unstandardized sample; the loss margin reflects how each sample is classified by the base classifier. A value greater than zero indicates that the number of samples correctly classified by the base classifier is greater than the number of samples misclassified. The Deep Fusion Weighted Algorithm (DFW) adjusts the interval of the sample loss, shifting the sample interval towards a value greater than 0, thereby continuously reducing the sample error rate. As the DFW algorithm iterates, it continuously increases the number of samples with negative interval increments, causing the current classifier to pay more attention to samples misclassified in the previous round, resulting in a continuous positive shift in the training sample intervals. This utilizes the loss interval... This measures the weight allocation of the current algorithm for each class of samples.

2. The fault diagnosis method for disconnecting switches as described in claim 1, characterized in that, In optimizing the weighting of vibration signal samples from disconnecting switches, the loss interval is used to measure the weighting of each classification sample by the current deep fusion weighting algorithm.

3. The fault diagnosis method for disconnecting switches as described in claim 2, characterized in that, A loss margin greater than zero indicates that the base classifier correctly classifies more samples than it misclassifies.

4. A fault diagnosis system for disconnecting switches, characterized in that, include: The feature extraction module is used to acquire the real-time vibration signal of the disconnector switch, and then perform singular value filtering and noise reduction and mixed mode decomposition on it in sequence to extract the feature vector; The fault diagnosis module is used to diagnose faults in the disconnector based on the feature vector and the deep weighted fusion model based on the SVM classifier. The construction process of the deep weighted fusion model based on the SVM classifier is as follows: Based on the feature vector of the vibration signal of the disconnector switch and the linear support vector machine with Gaussian kernel as the initial kernel function, an initial weak SVM classifier is obtained; The initial weak SVM classifier is iteratively optimized by using a deep fusion weighting algorithm and optimizing the weight allocation of vibration signal samples from disconnecting switches. An SVM classifier that meets the preset conditions is then obtained and used as a deep weighted fusion model based on the SVM classifier. Among them, the fault diagnosis results of the disconnecting switch include normal operating conditions, loose screw conditions, and stuck connecting rod conditions; In the process of optimizing the allocation of weights for vibration signal samples from disconnecting switches, the samples are divided into the following four categories after the weight update strategy is adjusted: a) Samples that were correctly classified by the previous strong classifier but were misclassified by the current strong classifier; b) Samples that were correctly classified by the previous strong classifier and are also correctly classified by the current strong classifier; c) Samples that were misclassified by the previous strong classifier and are also misclassified by the current strong classifier; d) Samples that were misclassified by the previous strong classifier but were correctly classified by the current strong classifier; The loss function of the deep fusion weighted algorithm is approximated as a function that maximizes the non-standardized mean of the margins and minimizes the variance of the non-standardized margin distribution: ； in, The variance of the non-standardized sample loss interval is represented. This represents the mean of the loss interval for non-standardized samples; The loss function is used to model the samples using standardized intervals, and its statistical definition is as follows: ； Where T represents the total number of non-standardized sample loss intervals; Let represent the base classifier corresponding to the loss margin of the t-th unstandardized sample; the loss margin reflects how each sample is classified by the base classifier. A value greater than zero indicates that the number of samples correctly classified by the base classifier is greater than the number of samples misclassified. The Deep Fusion Weighted Algorithm (DFW) adjusts the interval of the sample loss, shifting the sample interval towards a value greater than 0, thereby continuously reducing the sample error rate. As the DFW algorithm iterates, it continuously increases the number of samples with negative interval increments, causing the current classifier to pay more attention to samples misclassified in the previous round, resulting in a continuous positive shift in the training sample intervals. This utilizes the loss interval... This measures the weight allocation of the current algorithm for each class of samples.

5. The disconnector switch fault diagnosis system as described in claim 4, characterized in that, In optimizing the weighting of vibration signal samples from disconnecting switches, the loss interval is used to measure the weighting of each classification sample by the current deep fusion weighting algorithm.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the disconnector switch fault diagnosis method as described in any one of claims 1-3.

7. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the disconnector switch fault diagnosis method as described in any one of claims 1-3.

Images