A circuit breaker fault diagnosis method and system

Abstract

This invention discloses a circuit breaker fault diagnosis method and system, relating to the field of power system safety protection technology, and solves the problem of insufficient accuracy in circuit breaker fault diagnosis. In this embodiment, SVMD is performed on the acquired mechanical vibration signal to successively extract several narrowband modal components to achieve decoupled characterization of impact and multi-frequency components; then, TCN is used to perform feature learning on the modal component set, outputting the fault category and confidence level. This scheme can reduce the interference of noise and operating condition fluctuations on diagnosis, improve the accuracy and stability of multi-fault type identification, enhance the sensitivity to early weak fault symptoms, and support maintenance priority ranking and online operation and maintenance decision-making with confidence quantification.

Description

Technical Field

[0001] This invention relates to the field of power equipment fault diagnosis technology, and in particular to a circuit breaker fault diagnosis method and system. Background Technology

[0002] As a key breaking and protection device in substations and power distribution systems, circuit breakers may exhibit typical faults such as abnormal coil circuits, mechanical jamming, insufficient spring energy storage, or mechanical degradation such as contact wear and loose transmission components.

[0003] Existing circuit breaker vibration diagnosis methods mostly rely on single time-domain or frequency-domain indicators for threshold discrimination. Mechanical vibration signals have strong impact, non-stationary, and multi-source coupling characteristics during the opening and closing process, and are significantly affected by factors such as load conditions, mechanical wear, installation differences, and environmental noise. This makes it difficult for static thresholds to be adapted to multiple operating conditions simultaneously, easily leading to false alarms and missed alarms, and making it difficult to reliably support online diagnosis and operation and maintenance decisions.

[0004] Therefore, a circuit breaker fault diagnosis method and system are needed. Summary of the Invention

[0005] To address the problem of insufficient accuracy in diagnosing circuit breaker faults in existing technologies, this invention provides a method and system for diagnosing circuit breaker faults, which can improve the accuracy of fault diagnosis. The specific technical solution is as follows: In a first aspect, embodiments of this application provide a circuit breaker fault diagnosis method, including: The mechanical vibration signal of the circuit breaker during opening and closing is acquired. Successive Variational Mode Decomposition (SVMD) is performed on this mechanical vibration signal using a variational optimization model to obtain a set of modal components. This variational optimization model includes spectral compactness constraints, residual suppression constraints, and de-overlap constraints. The spectral compactness constraint concentrates the first modal component to its corresponding center frequency; this first modal component is the modal component extracted in the current iteration. The residual suppression constraint reduces the spectral overlap between the residual signal and the first modal component. The de-overlap constraint reduces the spectral overlap between the first modal component and the second modal component; this second modal component is the modal component that has already been extracted. This set of modal components is then input into a Temporal Convolutional Network (TCN) to obtain the fault diagnosis result output by the TCN.

[0006] Preferably, the variational optimization model includes an augmented Lagrangian function, which includes constraint terms from the spectral compactness constraint, the residual suppression constraint, and the de-overlap constraint, as well as an optimization objective term and Lagrange multiplier terms. The variational optimization model performs successive variational mode decomposition on the mechanical vibration signal to obtain a set of modal components, including: for the k-th extraction in the successive variational mode decomposition, based on the variational optimization model, constructing a k-th order augmented Lagrangian function using the first residual as input; wherein, the first residual... The difference is the residual signal after the (k-1)th extraction, where k is a positive integer greater than or equal to 1. The augmented Lagrangian function of the kth order is solved by the alternating direction multiplier method to obtain the kth order modal component and the second residual. If the second residual satisfies the first convergence condition, the first k order modal components are used as the set of modal components. If the second residual does not satisfy the first convergence condition, the second residual is used as the first residual to construct the (k+1)th order augmented Lagrangian function to solve for the new second residual, until the first convergence condition is satisfied.

[0007] Preferably, the method of solving the augmented Lagrangian function of the kth order by alternating direction multipliers to obtain the kth order modal components and the second residual includes: initializing the initial values ​​of the kth order Lagrangian multipliers, center frequency, and modal components; alternately performing the following update operations until the second convergence condition is met: a. fixing the center frequency and the Lagrangian multipliers, solving for the optimal solution of the augmented Lagrangian function with respect to the kth order modal components; b. updating the center frequency based on the frequency domain energy distribution of the optimal solution; c. calculating the reconstruction error based on the optimal solution and the mechanical vibration signal; updating the Lagrangian multipliers based on the reconstruction error; and when the second convergence condition is met, determining the corresponding optimal solution as the kth order modal components, and calculating the second residual based on the kth order modal components.

[0008] Preferably, the expression for the spectral compactness constraint includes: ; in, This indicates a spectral compactness constraint. For time-differentiable operators, The square of the L2 norm, The imaginary unit, Let t be a unit impulse function, and t be a time-domain variable. The center frequency of the kth order is Let be the k-th modal component.

[0009] Preferably, the expression for the residual suppression constraint includes: ; ; in, This indicates a residual suppression constraint. For filter functions, For balancing parameters, For the second residual, It is the square of the L2 norm.

[0010] Preferably, the expression for the de-overlap constraint includes: ; ; in, This indicates the removal of overlapping constraints. is the center frequency of the extracted i-th modal component.

[0011] Preferably, the TCN includes a fused convolutional layer of causal convolution and dilated convolution, and the diagnostic result includes the fault type and the confidence level of the fault type. After inputting the set of modal components into the temporal convolutional network TCN to obtain the fault diagnosis result output by the TCN, the method further includes: if the confidence level is less than a preset threshold, outputting a diagnostic conclusion to be reviewed and recording the corresponding mechanical vibration signal sample; if the confidence level is greater than or equal to the preset threshold, outputting the fault type and the corresponding confidence level.

[0012] Secondly, embodiments of this application provide a circuit breaker fault diagnosis system, applied to the method described in the first aspect, the system comprising: The acquisition module is used to acquire the mechanical vibration signal of the circuit breaker during the opening and closing process; The decomposition module is used to perform successive variational mode decomposition on the mechanical vibration signal using a variational optimization model to obtain a set of modal components. The variational optimization model includes spectral compactness constraints, residual suppression constraints, and deoverlap constraints. The spectral compactness constraints are used to concentrate the first modal component to the corresponding center frequency. The first modal component is the modal component extracted in the current iteration. The residual suppression constraints are used to reduce the spectral overlap between the residual signal and the first modal component. The deoverlap constraints are used to reduce the spectral overlap between the first modal component and the second modal component. The second modal component is the modal component that has been extracted. The diagnostic module is used to input the set of modal components into the temporal convolutional network (TCN) to obtain the fault diagnosis results output by the TCN.

[0013] Thirdly, embodiments of this application provide a computing device, including: a memory for storing a program; and a processor for loading the program to execute the method as described in the first aspect.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device where the computer-readable storage medium is located to perform the method described in the first aspect.

[0015] Compared with existing technologies, the advantages of this invention are as follows: by performing SVMD on the acquired mechanical vibration signal, a set of several narrowband modal components is successively extracted to achieve decoupled characterization of impact and multi-frequency components; then, TCN is used to perform feature learning on these modal component sets to output diagnostic results. This scheme can reduce the interference of noise and operating condition fluctuations on diagnosis, enhance the sensitivity to early weak fault symptoms, and improve the diagnostic accuracy of circuit breaker faults. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0017] Figure 1 A schematic flowchart illustrating a circuit breaker fault diagnosis method provided in an embodiment of this application; Figure 2 This application provides a schematic diagram of the structure of a circuit breaker fault diagnosis system according to an embodiment of the present application. Figure 3 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation

[0018] 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.

[0019] It should be understood that, when used in this specification and the appended claims, 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.

[0020] 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.

[0021] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0022] Conventional circuit breaker vibration diagnosis methods often rely on single time-domain or frequency-domain indicators for threshold discrimination, such as peak value, root mean square (RMS), kurtosis, and characteristic frequency energy ratio. Mechanical vibration signals exhibit strong impact, non-stationarity, and multi-source coupling characteristics during the opening and closing process. They are also significantly affected by factors such as load conditions, mechanical wear, installation differences, and environmental noise. This makes it difficult for static thresholds to simultaneously adapt to multiple operating conditions, leading to false alarms and missed alarms. Deep learning-based diagnostic methods typically use the original vibration sequence or spectrum as network input directly, without explicitly processing the differences between "multi-band narrowband oscillations and broadband impacts" within the signal. As a result, the network is easily dominated by strong amplitudes and common patterns during the learning process, and weak fault symptoms are masked in the feature space, thereby reducing the sensitivity to early degradation and low-probability high-loss anomalies.

[0023] The objective of this application is to achieve effective decomposition and dealiasing of mechanical vibration signals under conditions of strong impact during circuit breaker opening and closing, non-stationarity, overlapping frequency bands, and significant changes in operating conditions. This reduces the interference of noise and aliasing on fault diagnosis and improves feature stability and consistency across operating conditions. Furthermore, it effectively connects the decomposed multimodal components with a deep time-series network, ensuring online feasibility while providing usable confidence quantification and verification triggering mechanisms for the output results to support alarm and maintenance decisions.

[0024] This application addresses the online fault diagnosis needs of mechanical vibration signals during the opening and closing process of circuit breakers. The core technical methods include: firstly, performing SVMD on the acquired vibration sequence to sequentially extract several narrowband modal components to achieve decoupling characterization of impact and multi-frequency components, and then organizing the modal components into time-series input samples according to channels. Next, using a TCN with causality and an extended receptive field, feature learning is performed on the multi-channel modal sequence to output fault categories and confidence levels, and alarm or verification conclusions are generated based on preset thresholds.

[0025] Please refer to details. Figure 1 , Figure 1This application provides a flowchart illustrating a circuit breaker fault diagnosis method, which is applied to a computing device. Figure 1 As shown, the method includes: Step 101: The calculation device obtains the mechanical vibration signal of the circuit breaker during the opening and closing process.

[0026] The computing device can be a terminal or a server, which collects mechanical vibration signals during the opening and closing process through vibration sensors deployed in the circuit breaker mechanism box or key structures, records the sampling frequency and action type, and forms the original vibration sequence. .

[0027] Wherein, for the original vibration sequence The computing device can perform action segment detection to determine the start time of the action. and the end time The vibration signal of the action segment was intercepted. : .

[0028] Because the energy and amplitude characteristics of the circuit breaker's operating section and background section differ by orders of magnitude, the computing device can calculate the time-domain statistical characteristics of the signal, set a reasonable threshold, and locate the start and end of the operation based on the moment when the characteristic value crosses the threshold.

[0029] Preferably, the computing device can acquire the background signal during the stationary phase when the circuit breaker is not operating, and calculate the mean of the background RMS sequence of the background signal through a sliding window. and standard deviation Then, the real-time RMS sequence of the original vibration sequence is calculated using the same sliding window; when M consecutive points in the real-time RMS sequence exceed the initial threshold, the sampling time of the first point that crosses the threshold is taken as the starting time. After the start time, when K consecutive points in the real-time RMS sequence are below the termination threshold, the sampling time of the last point that crosses the threshold is taken as the termination time. .

[0030] The initial threshold can be 3 times the mean. and 2 times the standard deviation The sum of M, the range of M is [3, 5].

[0031] The termination threshold can be 1.5 times the value. The value of K can be in the range of [3, 5].

[0032] Then, the computing device can use linear fitting or moving average methods to eliminate vibration signals in the action segment. Baseline drift; then the action segment signal Resampling or interpolation yields a discrete sequence of uniform length, followed by detrending, noise filtering, and amplitude normalization to obtain a standardized vibration sequence. To reduce the differences in length and amplitude scales among different action samples. In some other possible implementations, the computing device may first preprocess the acquired complete mechanical vibration signal, and then extract the motion segment vibration signal from the preprocessed mechanical vibration signal.

[0033] In other possible implementations, the computing device may also use other time-domain statistical characteristics such as peak value and absolute average amplitude to replace RMS in the preferred method described above, and use a threshold method to determine the start time of circuit breaker operation. and the end time Alternatively, the degree of mutation of more complex multi-dimensional feature values ​​can be determined, or it can be determined through neural networks.

[0034] Step 102: The computing device performs Successive Variational Mode Decomposition (SVMD) on the mechanical vibration signal using a variational optimization model to obtain a set of modal components.

[0035] SVMD is a signal processing method that aims to decompose a complex raw signal into a series of simple modal components; in the... During the successive extraction, SVMD can decompose the original signal into the th... It consists of two parts: first-order mode components and residual signals.

[0036] Specifically, the modal components are several narrowband components obtained after SVMD decomposition. Each component corresponds to the oscillation component of the signal near a certain center frequency and is used to characterize different physical excitations or structural response mechanisms. The residual signal is the part of the signal that remains after SVMD has successively extracted several modes and has not been interpreted by the extracted modes. It may contain broadband impulses, noise, or components that have not been sufficiently separated.

[0037] SVMD expressions include: ; ; in, The original signal, The k-th modal component is obtained from the k-th extraction. This is the residual signal after the k-th decomposition, which is the part remaining after subtracting the k-th modal component from the original signal; The remaining signal in the current residual that still needs to be decomposed includes all the modal components that have been extracted, as well as the remaining signal that has not yet been decomposed. This allows each decomposition to build upon the previous results, ensuring that the decomposition process is coherent.

[0038] Each decomposition takes the residual signal from the previous step as input and extracts new modal components until the remaining signal is obtained. It is small enough to satisfy the convergence condition.

[0039] In the modality extraction process, to ensure that the obtained modes have clear physical meaning and reduce modality aliasing, the computing device can introduce three types of constraints to form a targeted approach. Its center frequency The variational optimization model.

[0040] The variational optimization model includes spectral compactness constraints, residual suppression constraints, and de-overlap constraints. The spectral compactness constraints are used to concentrate the first mode component to the corresponding center frequency. The first mode component is the mode component extracted in the current iteration. The residual suppression constraints are used to reduce the spectral overlap between the residual signal and the first mode component. The de-overlap constraints are used to reduce the spectral overlap between the first mode component and the second mode component. The second mode component is the mode component that has been extracted.

[0041] Preferably, the computing device can first pass through Construct real values The complex-valued analytical form eliminates negative frequency components and retains only the... The presence of nearby positive frequency information ensures that subsequent frequency analysis focuses only on the actual effective frequencies, avoiding bandwidth calculation errors caused by redundant information; then multiplied by The mode is shifted to baseband; then, by taking the first-order time derivative of the baseband preference, its L2 norm energy is calculated, and the bandwidth is constrained by minimizing this energy; the expression for this spectral compactness constraint includes: ; in, This indicates a spectral compactness constraint. For time-differentiable operators, The square of the L2 norm, The imaginary unit, Let t be a unit impulse function, and t be a time-domain variable. The center frequency of the kth order is Let be the k-th modal component.

[0042] Specifically, the more concentrated the baseband signal, the slower its time-domain changes. The energy of the time derivative quantifies the drastic nature of these time-domain changes; a smaller derivative energy indicates a slower signal change, corresponding to a baseband signal frequency more concentrated around zero frequency. The narrower the bandwidth, the more concentrated it is at the center frequency.

[0043] Preferably, to reduce the residual signal in To mitigate the possibility of spectral overlap with the first mode component in the vicinity, the computing device can introduce filters. A weighted metric is applied to the residuals. The expression for this residual suppression constraint includes: ; ; in, This indicates a residual suppression constraint. For filter functions, For balancing parameters, For the second residual, It is the square of the L2 norm.

[0044] Preferably, to avoid the first mode... Mode aliasing occurs due to overlap with the extracted second mode component in the frequency domain. Therefore, a method is introduced for each extracted center frequency. Filter ,right Penalties are imposed on the energy in these frequency bands. The expression for the deoverlap constraint includes: ; ; in, This indicates the removal of overlapping constraints. is the center frequency of the extracted i-th modal component.

[0045] Based on the above three types of constraints, computing devices can construct and extract the first... Variational problem of first-order modal components: .

[0046] Then, the computing device can introduce Lagrange multipliers into the variational optimization model. Construct an augmented Lagrangian function to transform the constrained problem into an unconstrained form, and set iterative update and termination conditions; then solve the variational optimization model.

[0047] The expression for this unconstrained augmented Lagrangian function can be: .

[0048] Preferably, for the k-th extraction in successive variational mode decomposition, the computing device can construct the aforementioned k-th order augmented Lagrangian function based on the variational optimization model, using the first residual as input; solve the k-th order augmented Lagrangian function using the Alternating Direction Method of Multipliers (ADMM) to obtain the k-th order modal components and the second residual; if the second residual satisfies the first convergence condition, the first k-th order modal components are used as the set of modal components; if the second residual does not satisfy the first convergence condition, the second residual is used as the first residual to construct the (k+1)-th order augmented Lagrangian function to solve for the new second residual, until the first convergence condition is satisfied.

[0049] Wherein, the first residual is the residual signal after the (k-1)th extraction, where k is a positive integer greater than or equal to 1; when k=1, the first residual is the normalized action segment signal obtained in step 101.

[0050] The entire process of solving the augmented Lagrangian function can be divided into an outer successive decomposition iteration and an inner ADMM iteration. The ADMM iteration calculates the convergence value of the current order modal component. After the inner ADMM converges, the computing device calculates the second residual to determine whether the outer iteration has converged, that is, whether the first convergence condition has been met. If it is met, the iteration stops, and the currently converged first k order modal components are used as the modal component set, and also as the input of the subsequent TCN. If it is not met, the second residual is used as the first residual for the inner iteration until the first convergence condition is met.

[0051] Specifically, the first convergence condition may include the residual energy of the second residual being less than a preset threshold, or the second residual having no effective frequency components after frequency domain analysis, or the number of extracted modal components reaching a preset maximum value.

[0052] Preferably, the computing device can initialize the initial values ​​of the k-th order Lagrange multiplier, center frequency, and modal components; and alternately perform the following update operations until the second convergence condition is met: a. fix the center frequency and the Lagrange multiplier, and solve for the optimal solution of the augmented Lagrangian function with respect to the k-th order modal components; b. update the center frequency based on the frequency domain energy distribution of the optimal solution; c. calculate the reconstruction error based on the optimal solution and the mechanical vibration signal; update the Lagrange multiplier based on the reconstruction error; and when the second convergence condition is met, determine the corresponding optimal solution as the k-th order modal components, and calculate the second residual based on the k-th order modal components.

[0053] For each modal component, the computing device constructs its own augmented Lagrange function, initial Lagrange multipliers, initial center frequency, and initial values ​​for the modal component. With the center frequency and Lagrange multipliers fixed, the subproblem for finding the optimal solution is a strictly convex unconstrained optimization problem. The computing device can take the first-order partial derivative of the objective function and set it equal to 0; the resulting solution is the globally unique optimal solution.

[0054] The computing device can calculate the corresponding residual signal based on the optimal solution, the first k-1 determined modal components, and the normalized vibration signal of the action segment, and use the residual signal as the reconstruction error.

[0055] The expressions for updating the Lagrange multipliers include: ; , These are the Lagrange multipliers before and after the update. For the center frequency, The iteration step size, To normalize the frequency domain form of the vibration signal during the action segment, This represents the reconstruction error.

[0056] The second convergence condition includes: the relative change of modal components between two adjacent iterations being less than the first rate of change threshold; the norm square of the signal reconstruction error being less than the error threshold; and the change of the Lagrange multiplier between two adjacent iterations being less than the second rate of change threshold.

[0057] Step 103: The computing device inputs the set of modal components into a temporal convolutional network (TCN) to obtain the fault diagnosis results output by the TCN.

[0058] The computing device can analyze the modal components corresponding to the mechanical vibration signals of each circuit breaker operation. Perform uniform length truncation or resampling, and normalize the amplitude to reduce the impact of amplitude scale differences under different operating conditions on the recognition results. This applies to the same action... The road mode sequence is stacked by channel as the input tensor.

[0059] Among them, the computing device can assign fault labels to the samples based on maintenance records or test calibration results. These labels include normal states and at least one fault state. Then, a training sample set can be built, and the training set and validation set can be divided for network training and performance evaluation.

[0060] Preferably, the TCN includes a fused convolutional layer of causal convolution and dilated convolution, and the diagnostic result includes the fault type and the confidence level of the fault type. After inputting the set of modal components into the temporal convolutional network TCN to obtain the fault diagnosis result output by the TCN, the method further includes: if the confidence level is less than a preset threshold, outputting a diagnostic conclusion to be reviewed and recording the corresponding mechanical vibration signal sample; if the confidence level is greater than or equal to the preset threshold, outputting the fault type and the corresponding confidence level.

[0061] Among them, TCN, as a fault discrimination model, includes several fused convolutional layers, so that the network output depends only on the input at the current and historical moments, while covering long-term temporal correlation features such as impact, rebound and attenuation during the circuit breaker operation process.

[0062] Among them, causal convolution refers to the convolution calculation that only uses the input data at the current and historical moments and does not refer to the data at future moments, thereby meeting the causal requirements of online diagnosis; dilated convolution refers to the introduction of a fixed interval between adjacent sampling points of the convolution kernel, so that the receptive field expands rapidly with the number of layers, thereby representing long-term dependencies with fewer parameters.

[0063] For the first in TCN A dilated causal convolution of layers, with a kernel length of 1. The void coefficient is The input features are The output features are Then the convolution operation of this layer satisfies: ; in, For the first Each convolutional kernel weight matrix For bias vectors, For time indexing.

[0064] In the above formula, the time corresponding to the input data is This ensures output It only relies on the current time τ and the historical inputs before it, and does not use information from future time times, which meets the causality requirement of time series signal processing; Represents the sampling interval on the input features, when When the value is greater than 1, the convolutional kernel will skip some input points, thereby expanding the temporal receptive field without increasing the computational cost.

[0065] By setting different void ratios for different layers This allows the receptive field of each layer to expand exponentially, thereby capturing the full-cycle characteristics from short-sequence impact to long-sequence decay, and enabling better extraction of the timing fault characteristics of circuit breaker vibration signals.

[0066] Then, after extracting features through multiple layers of dilated causal convolutions, TCN can map the features into a probability vector through fully connected layers or global pooling layers. ,in Indicates belonging to the first Confidence probability of a fault class.

[0067] Then, the computing device can generate fault diagnosis results based on the probability vector and output the category. and confidence level : ; .

[0068] Among them, the computing device can be set with a preset threshold. This is used to distinguish between a reliable diagnosis and a diagnosis that needs to be verified. When Output the conclusion to be reviewed and record the mechanical vibration signal sample of the corresponding action; when Time output fault type and confidence level The system outputs diagnostic results and uses them to prioritize alarms and maintenance procedures, thereby providing decision-making support for maintenance personnel and enabling online identification and auxiliary maintenance of abnormal mechanical vibrations in circuit breakers.

[0069] In this embodiment, SVMD is performed on the acquired mechanical vibration signal to successively extract a set of narrowband modal components, thereby achieving decoupled characterization of impact and multi-frequency components. Then, TCN is used to perform feature learning on these modal component sets and output diagnostic results. This scheme can reduce the interference of noise and operating condition fluctuations on diagnosis, enhance the sensitivity to early weak fault symptoms, and improve the diagnostic accuracy of circuit breaker faults.

[0070] In this embodiment, by combining the successive variational mode decomposition (SVMD) of the mechanical vibration signal of the circuit breaker with the temporal convolutional network (TCN), the vibration sequence with strong impact, non-stationary and overlapping frequency bands is first decomposed to obtain a set of modal components to achieve decoupling of multi-frequency components and noise suppression. Then, the TCN is driven by the modal channelization input to complete the fault type identification, thereby making up for the insufficient adaptability of the working conditions caused by the reliance on a single time domain or frequency domain index and static threshold discrimination in the prior art. This can improve the stability and accuracy of the diagnostic results under different operating speeds and noise backgrounds.

[0071] The method provided in the embodiments of this application has been described above. The system provided in the embodiments of this application will be described below.

[0072] Please see Figure 2 , Figure 2This is a schematic diagram of the structure of a circuit breaker fault diagnosis system provided in an embodiment of this application, as shown below. Figure 2 As shown, the system 20 includes: The acquisition module 201 is used to acquire the mechanical vibration signal of the circuit breaker during the opening and closing process; The decomposition module 202 is used to perform successive variational mode decomposition on the mechanical vibration signal through a variational optimization model to obtain a set of modal components. The variational optimization model includes spectral compactness constraints, residual suppression constraints, and deoverlap constraints. The spectral compactness constraints are used to concentrate the first modal component to the corresponding center frequency. The first modal component is the modal component extracted in the current iteration. The residual suppression constraints are used to reduce the spectral overlap between the residual signal and the first modal component. The deoverlap constraints are used to reduce the spectral overlap between the first modal component and the second modal component. The second modal component is the modal component that has been extracted. The diagnostic module 203 is used to input the set of modal components into the temporal convolutional network (TCN) to obtain the fault diagnosis results output by the TCN.

[0073] Preferably, the variational optimization model includes an augmented Lagrangian function, which includes constraint terms from the spectral compactness constraint, the residual suppression constraint, and the deoverlap constraint, as well as an optimization objective term and Lagrange multiplier terms. The decomposition module 202 is specifically used to construct the k-th order augmented Lagrangian function based on the variational optimization model, using the first residual as input, for the k-th extraction in the successive variational mode decomposition; wherein the first residual is the residual signal after the (k-1)-th extraction. k is a positive integer greater than or equal to 1; the k-th order augmented Lagrangian function is solved by the alternating direction multiplier method to obtain the k-th order modal components and the second residual; if the second residual satisfies the first convergence condition, the first k-th order modal components are used as the set of modal components; if the second residual does not satisfy the first convergence condition, the second residual is used as the first residual to construct the (k+1)-th order augmented Lagrangian function to solve for the new second residual, until the first convergence condition is satisfied.

[0074] Preferably, the decomposition module 202 is specifically used to initialize the initial values ​​of the k-th order Lagrange multiplier, center frequency, and modal components; and alternately perform the following update operations until the second convergence condition is met: a. fix the center frequency and the Lagrange multiplier, and solve for the optimal solution of the augmented Lagrangian function with respect to the k-th order modal components; b. update the center frequency based on the frequency domain energy distribution of the optimal solution; c. calculate the reconstruction error based on the optimal solution and the mechanical vibration signal; update the Lagrange multiplier based on the reconstruction error; and when the second convergence condition is met, determine the corresponding optimal solution as the k-th order modal component, and calculate the second residual based on the k-th order modal component.

[0075] Preferably, the expression for the spectral compactness constraint includes: ; in, This indicates a spectral compactness constraint. For time-differentiable operators, The square of the L2 norm, The imaginary unit, Let t be a unit impulse function, and t be a time-domain variable. The center frequency of the kth order is Let be the k-th modal component.

[0076] Preferably, the expression for the residual suppression constraint includes: ; ; in, This indicates a residual suppression constraint. For filter functions, For balancing parameters, For the second residual, It is the square of the L2 norm.

[0077] Preferably, the expression for the de-overlap constraint includes: ; ; in, This indicates the removal of overlapping constraints. is the center frequency of the extracted i-th modal component.

[0078] Preferably, the TCN includes a fused convolutional layer of causal convolution and dilated convolution, and the diagnostic result includes the fault type and the confidence level of the fault type; the system 20 also includes an output module 204, which is used to output the diagnostic conclusion to be reviewed and record the corresponding mechanical vibration signal sample when the confidence level is less than a preset threshold; and output the fault type and the corresponding confidence level when the confidence level is greater than or equal to the preset threshold.

[0079] The circuit breaker fault diagnosis system provided in this application can be understood by referring to the relevant content in the foregoing method embodiment section, and will not be repeated here.

[0080] like Figure 3 As shown, Figure 3This is a schematic diagram of a possible logical structure of a computing device provided in an embodiment of this application. The computing device 30 includes a processor 301, a communication interface 302, a memory 303, and a bus 304. The processor 301, the communication interface 302, and the memory 303 are interconnected via the bus 304. In an embodiment of this application, the processor 301 is used to control and manage the operation of the computing device 30. For example, the processor 301 is used to execute... Figure 1 The steps in the embodiments and / or other processes used in the techniques described herein. Communication interface 302 is used to support communication by computing device 30. Memory 303 is used to store program code and data of computing device 30.

[0081] The processor 301 can be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. The processor can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a digital signal processor and a microprocessor, etc. The bus 304 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0082] In another embodiment of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including instructions that, when executed on a computer, cause the computer to perform the above-described... Figure 1 The method described in the embodiments.

[0083] Those skilled in the art will recognize that the units of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the various examples have been generally described in terms of functionality in the foregoing description. 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 the invention.

[0084] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0085] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0086] 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.

[0087] 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.

[0088] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for diagnosing circuit breaker faults, characterized in that, The method includes: Acquire the mechanical vibration signal of the circuit breaker during the opening and closing process; The mechanical vibration signal is subjected to successive variational mode decomposition using a variational optimization model to obtain a set of modal components. The variational optimization model includes spectral compactness constraints, residual suppression constraints, and de-overlap constraints. The spectral compactness constraints are used to concentrate the first modal component to the corresponding center frequency. The first modal component is the modal component extracted in the current iteration. The residual suppression constraints are used to reduce the spectral overlap between the residual signal and the first modal component. The de-overlap constraints are used to reduce the spectral overlap between the first modal component and the second modal component. The second modal component is the modal component that has been extracted. The set of modal components is input into a temporal convolutional network (TCN) to obtain the fault diagnosis results output by the TCN.

2. The method according to claim 1, characterized in that, The variational optimization model includes an augmented Lagrangian function, which includes constraint terms from the spectral compactness constraint, the residual suppression constraint, and the de-overlap constraint. The step involves performing successive variational mode decomposition on the mechanical vibration signal using a variational optimization model to obtain a set of modal components, including: For the k-th extraction in successive variational mode decomposition, based on the variational optimization model, an augmented Lagrangian function of order k is constructed with the first residual as input; wherein, the first residual is the residual signal after the (k-1)-th extraction, and k is a positive integer greater than or equal to 1; The k-th order augmented Lagrangian function is solved by alternating direction multiplier method to obtain the k-th order modal components and the second residual; If the second residual satisfies the first convergence condition, the first k modal components are taken as the set of modal components; If the second residual does not satisfy the first convergence condition, the second residual is used as the first residual to construct an augmented Lagrangian function of order k+1 to solve for the new second residual, until the first convergence condition is satisfied.

3. The method according to claim 2, characterized in that, The method of solving the k-th order augmented Lagrangian function using the alternating direction multiplier method to obtain the k-th order modal components and the second residual includes: Initialize the k-th order Lagrange multipliers, center frequency, and initial values ​​of modal components; Perform the following update operations alternately until the second convergence condition is met: a. Fix the center frequency and the Lagrange multiplier, and solve for the optimal solution of the augmented Lagrange function with respect to the k-th modal component; b. Update the center frequency based on the frequency domain energy distribution of the optimal solution; c. Calculate the reconstruction error based on the optimal solution and the mechanical vibration signal; update the Lagrange multiplier based on the reconstruction error; If the second convergence condition is met, the corresponding optimal solution is determined to be the k-th modal component, and the second residual is calculated based on the k-th modal component.

4. The method according to any one of claims 1-3, characterized in that, The expression for the spectral compactness constraint includes: ； in, This indicates a spectral compactness constraint. For time-differentiable operators, The square of the L2 norm. The imaginary unit, Let t be a unit impulse function, and t be a time-domain variable. The center frequency of the kth order is Let be the k-th modal component.

5. The method according to any one of claims 1-3, characterized in that, The expression for the residual suppression constraint includes: ； ； in, This indicates a residual suppression constraint. For filter functions, For balancing parameters, For the second residual, It is the square of the L2 norm.

6. The method according to any one of claims 1-3, characterized in that, The expression for the de-overlapping constraint includes: ； ； in, This indicates the removal of overlapping constraints. is the center frequency of the extracted i-th modal component.

7. The method according to any one of claims 1-3, characterized in that, The TCN includes a fused convolutional layer of causal convolutions and dilated convolutions, and the diagnostic result includes a fault type and a confidence level of the fault type; after inputting the set of modal components into the temporal convolutional network TCN to obtain the fault diagnosis result output by the TCN, the method further includes: If the confidence level is less than a preset threshold, output the diagnostic conclusion to be reviewed and record the corresponding mechanical vibration signal sample. If the confidence level is greater than or equal to the preset threshold, the fault type and the corresponding confidence level are output.

8. A circuit breaker fault diagnosis system, characterized in that, The system, applied to the method of any one of claims 1-7, comprises: The acquisition module is used to acquire the mechanical vibration signal of the circuit breaker during the opening and closing process; The decomposition module is used to perform successive variational mode decomposition on the mechanical vibration signal using a variational optimization model to obtain a set of modal components. The variational optimization model includes spectral compactness constraints, residual suppression constraints, and de-overlap constraints. The spectral compactness constraints are used to concentrate the first modal component towards its corresponding center frequency; the first modal component is the modal component extracted in the current iteration. The residual suppression constraints are used to reduce the spectral overlap between the residual signal and the first modal component. The de-overlap constraints are used to reduce the spectral overlap between the first modal component and the second modal component; the second modal component is the modal component that has already been extracted. The diagnostic module is used to input the set of modal components into the temporal convolutional network (TCN) to obtain the fault diagnosis results output by the TCN.

9. A computing device, characterized in that, include: Memory, used to store programs; A processor for loading the program to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the method of any one of claims 1-7.

Images