Circuit breaker fault diagnosis method and device, computer equipment and storage medium

Abstract

The invention relates to a circuit breaker fault diagnosis method and device, computer equipment and a storage medium, and relates to the technical field of fault detection. The method comprises the following steps: acquiring a coil current signal and a vibration signal of a target circuit breaker; determining a key time interval and a linear slope of the coil current signal as current characteristics; the key time interval is the time corresponding to the time when the target circuit breaker starts to act to the time when the iron core touches the operating mechanism; the linear slope is determined based on current fitting values corresponding to the coil current curve equation at the moment of the first turning point and the moment of starting action of the target circuit breaker; determining a power spectrum based on an envelope line corresponding to the vibration signal, and determining an index value of the power spectrum in the key frequency interval as a vibration characteristic; and based on a plurality of pre-constructed single-class support vector machines, identifying the current characteristics and the vibration characteristics to obtain a fault diagnosis result of the target circuit breaker. By adopting the method, the fault identification accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of fault detection, and in particular to a circuit breaker fault diagnosis method, device, computer equipment and storage medium. Background Art

[0002] Circuit breakers are the most important electrical equipment in the power system, and are mainly responsible for the control and protection of other electrical equipment. At present, the fault identification of most high-voltage circuit breakers is mainly based on information such as the opening and closing coil current, travel data, vibration, and sound print signals. Among them, vibration signals and opening and closing coil current have become important indicators for online monitoring and diagnosis of circuit breakers.

[0003] In the related art, the coil current analysis technology is mainly based on the key moment point information of the typical coil current for judgment, but the actual collected coil current is not completely consistent with the typical coil current waveform, and there is noise interference, so there will be deviations in directly locating the time point. The wavelet transform in the vibration analysis technology can simultaneously characterize the time and frequency domain information of the signal, and can decompose the signal into detail parts and approximate parts for analysis, but the wavelet transform cannot accurately describe the relationship between time and frequency. Therefore, the fault identification accuracy of the high-voltage circuit breaker fault diagnosis based on the coil current or vibration signal characteristics in the related art is insufficient. Summary of the invention

[0004] Based on this, it is necessary to provide a high-voltage circuit breaker fault diagnosis method, device, computer equipment, computer-readable storage medium and computer program product that can improve the accuracy of fault identification in response to the above technical problems.

[0005] In a first aspect, the present application provides a circuit breaker fault diagnosis method. The method comprises:

[0006] Acquire coil current signal and vibration signal of target circuit breaker;

[0007] Determine the key time interval and linear slope of the coil current signal as current characteristics; the key time interval is the time corresponding to the start of the target circuit breaker from the start of the action to the iron core triggering the operating mechanism; the linear slope is determined based on the current fitting values ​​corresponding to the first turning point of the coil current curve equation and the start of the action of the target circuit breaker;

[0008] Based on the envelope corresponding to the vibration signal, a power spectrum is determined, and an index value of the power spectrum in a key frequency interval is determined as a vibration feature; the key frequency interval is determined based on a change in a spectral kurtosis coefficient value in each interval of the vibration signal;

[0009] Based on a plurality of pre-constructed single-class support vector machines, the current characteristics and the vibration characteristics are identified to obtain a fault diagnosis result of the target circuit breaker; the single-class support vector machine is trained based on the current characteristics and vibration characteristics of circuit breakers with various types of faults.

[0010] In one embodiment, determining the critical time interval and linear slope of the coil current signal as current characteristics includes:

[0011] Determine the critical time interval between the start of the action of the target circuit breaker and the iron core triggering the operating mechanism according to the coil current signal at the time when the target circuit breaker is not actuated and the time when the impact amplitude of the vibration signal appears;

[0012] Performing curve fitting on the coil current signal within the critical time interval based on an adaptive least squares method to obtain a coil current curve equation;

[0013] Based on the curvature of the coil current curve equation at each moment, the moment of the first turning point is determined, and based on the current fitting values ​​corresponding to the moment of the first turning point and the moment of the start of action of the target circuit breaker, the linear slope of the coil current signal is determined.

[0014] In one embodiment, determining the time of the first turning point based on the curvature of the coil current curve equation at each time comprises:

[0015] Determine the time of the local minimum value of the coil current curve equation in the key time interval and the time of the first local maximum value according to the sliding window strategy;

[0016] The time of determining the target local minimum is the time of the first turning point; the target local minimum is selected from the local minimums, and the time of the target local minimum is greater than the time of the first local maximum.

[0017] In one embodiment, determining the power spectrum based on the envelope corresponding to the vibration signal includes:

[0018] De-noising the vibration signal based on the peak coefficient of each window corresponding to the vibration signal to obtain a de-noised vibration signal;

[0019] Filtering the denoised vibration signal in the key frequency range through a Chebyshev filter to obtain a filtered vibration signal;

[0020] Wherein, if the minimum value of the key frequency interval is a preset frequency value, filtering is performed through low-pass filtering; if the minimum value of the key frequency interval is greater than the preset frequency value, filtering is performed through band-pass filtering; if the minimum value of the key frequency interval is less than the preset frequency value, filtering is performed through high-pass filtering;

[0021] Based on the Hilbert-Huang transform strategy, the filtered vibration signal is processed to obtain an envelope corresponding to the filtered vibration signal;

[0022] Based on the indirect autocorrelation method, a power spectrum corresponding to the envelope is determined.

[0023] In one embodiment, determining the index value of the power spectrum in the key frequency interval as a vibration feature includes:

[0024] For each frequency interval of the key frequency interval, determining a power spectrum energy value of each frequency line corresponding to the frequency interval in the power spectrum;

[0025] The average value, variance value and center of gravity value corresponding to the power spectrum energy value of each frequency line are determined as the vibration characteristics of the key frequency range.

[0026] In one embodiment, before determining the power spectrum based on the envelope corresponding to the vibration signal, the method further includes:

[0027] According to the binary classification recursive tree, the initial frequency interval of the vibration signal is split to obtain a first frequency interval and a second frequency interval;

[0028] Determine the spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval respectively, and if any one of the spectral kurtosis coefficient values ​​is greater than a spectral kurtosis threshold, determine the frequency interval corresponding to the spectral kurtosis coefficient value greater than the spectral kurtosis threshold as the initial frequency interval;

[0029] Returning to the step of splitting the initial frequency interval of the vibration signal to obtain a first frequency interval and a second frequency interval, until the spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval, respectively, are both less than the spectral kurtosis threshold;

[0030] If the spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval respectively are both less than or equal to the spectral kurtosis threshold, the initial frequency interval is determined to be a key frequency interval.

[0031] In one embodiment, the identifying the current feature and the vibration feature based on the pre-built multiple single-class support vector machines to obtain the fault diagnosis result of the target circuit breaker also includes:

[0032] The current feature and the vibration feature are used as a group of features, and are respectively input into each trained single-class support vector machine to obtain a prediction result output by each single-class support vector machine; the prediction result of the single-class support vector machine is a first prediction value or a second prediction value;

[0033] If the prediction results of each of the single-class support vector machines are the first prediction value, determining that the fault diagnosis result of the target circuit breaker is a new fault type;

[0034] If there is a single-class support vector machine whose prediction result is the second prediction value, it is determined that the fault diagnosis result of the target circuit breaker is the fault type corresponding to the single-class support vector machine whose prediction result is the second prediction value.

[0035] In a second aspect, the present application also provides a circuit breaker fault diagnosis device. The device comprises:

[0036] A signal acquisition module, used to acquire a coil current signal and a vibration signal of a target circuit breaker;

[0037] A current characteristic determination module, used to determine the key time interval and linear slope of the coil current signal as the current characteristic; the key time interval is the time corresponding to the start of the action of the target circuit breaker to the time when the iron core triggers the operating mechanism; the linear slope is determined based on the current fitting values ​​corresponding to the first turning point of the coil current curve equation and the start of the action of the target circuit breaker;

[0038] A vibration feature determination module, used to determine a power spectrum based on an envelope corresponding to the vibration signal, and determine an index value of the power spectrum in a key frequency interval as a vibration feature; the key frequency interval is determined based on a change in a spectral kurtosis coefficient value in each interval of the vibration signal;

[0039] A fault identification module is used to identify the current characteristics and the vibration characteristics based on a plurality of pre-constructed single-class support vector machines to obtain a fault diagnosis result of the target circuit breaker; the single-class support vector machines are trained based on the current characteristics and vibration characteristics of circuit breakers with various types of faults.

[0040] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method described in the first aspect are implemented.

[0041] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0042] In a fifth aspect, the present application further provides a computer program product, wherein the computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0043] The above-mentioned circuit breaker fault diagnosis method, device, computer equipment, storage medium and computer program product obtain the coil current signal and vibration signal of the target circuit breaker, and determine the time period corresponding to the start of the action of the target circuit breaker to the iron core triggering the operating mechanism as the key time interval, and determine the linear slope corresponding to the current fitting value of the coil current curve equation at the first turning point and the moment when the target circuit breaker starts to act, and use the key time interval and linear slope as the current characteristics of the target circuit breaker. Determine the power spectrum according to the envelope of the vibration signal, determine the key frequency interval according to the spectral kurtosis coefficient value of each interval of the vibration signal, and determine the index value of the power spectrum in the key frequency interval, and use the index value as the vibration feature. Predict the current signal and vibration signal through multiple pre-constructed single-class support vector machines, determine the fault diagnosis result of the target circuit breaker, and obtain the fault type of the target circuit breaker.

[0044] The coil current signal that can represent the target circuit breaker failure can be accurately obtained through the key time interval, and the linear slope can accurately obtain the signal changes generated when the target circuit breaker fails, thereby improving the accuracy of the current characteristics obtained. The accurate key frequency interval can be determined by the spectral kurtosis coefficient value, and the relationship between time and frequency can be accurately represented by the index value of the key frequency interval, thereby improving the accuracy of the vibration characteristics. The single-class support vector machine obtained by training the current characteristics and vibration characteristics of each circuit breaker can predict the fault diagnosis results of the target circuit breaker and improve the accuracy of fault identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] FIG1 is an application environment diagram of a circuit breaker fault diagnosis method in one embodiment;

[0047] FIG2 is a schematic flow chart of a circuit breaker fault diagnosis method according to an embodiment;

[0048] FIG3 is a schematic diagram of a coil current waveform in one embodiment;

[0049] FIG4 is a schematic diagram of a coil current waveform in a normal closing state in one embodiment;

[0050] FIG5 is a schematic diagram of a coil current waveform in a normal opening state in one embodiment;

[0051] FIG6 is a schematic diagram of coil current waveform fitting in one embodiment;

[0052] FIG7 is a schematic diagram of calculating the peak coefficient using a sliding window method in one embodiment;

[0053] FIG8 is a schematic diagram of a waveform of a vibration signal after denoising in one embodiment;

[0054] FIG9 is a schematic diagram of a process of determining spectral kurtosis using a binary classification recursive tree in one embodiment;

[0055] FIG10 is a schematic diagram of a vibration signal before and after filtering in one embodiment;

[0056] FIG11 is a schematic diagram of a process for constructing a PSO-OcSVM model in one embodiment;

[0057] FIG12 is a schematic diagram of a process of performing fault diagnosis based on a PSO-OcSVM model in one embodiment;

[0058] FIG13 is a schematic diagram of the recognition result of the PSO-OcSVM model in one embodiment;

[0059] FIG14 is a structural block diagram of a circuit breaker fault diagnosis device in one embodiment;

[0060] FIG. 15 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0062] The circuit breaker fault diagnosis method provided in the embodiment of the present application can be applied in the application environment shown in Figure 1. Among them, the terminal 102 communicates with the server 104 through the network, and the terminal 102 can monitor the target circuit breaker to be detected in real time, and obtain the coil current signal and vibration signal of the target circuit breaker. The terminal 102 can send the acquired coil current signal and vibration signal to the server 104, and the server 104 can determine the key time interval of the coil current signal and the linear slope of the coil current signal as the current feature. The server 104 can determine the envelope of the vibration signal, and determine the power spectrum based on the envelope, and use the index value of the power spectrum in the key frequency interval as the vibration feature. The server 104 inputs the current feature and vibration feature of the target circuit breaker into multiple single-class support vector machines, obtains the prediction results of each single-class support vector machine, and obtains the fault diagnosis result of the target circuit breaker by combining the above prediction results.

[0063] Among them, the data storage system can store data that the server 104 needs to process, such as the coil current signal and vibration signal of each circuit breaker in the past. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and circuit breakers. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Personal computers, laptops, smart phones, and tablet computers can all obtain the coil current signal and vibration signal of the circuit breaker in real time. The server 104 can be implemented as an independent server or a server cluster consisting of multiple servers.

[0064] In an exemplary embodiment, as shown in FIG2 , a circuit breaker fault diagnosis method is provided, which is described by taking the method applied to the server 104 in FIG1 as an example, and includes the following steps S202 to S208 . Among them:

[0065] Step S202, obtaining a coil current signal and a vibration signal of a target circuit breaker.

[0066] Specifically, the server can obtain the coil current signal and vibration information of the target circuit breaker from the terminal or circuit breaker in real time. The coil current signal usually refers to the current signal flowing in the control coil of the circuit breaker. The vibration signal refers to the signal corresponding to the mechanical vibration generated by the movement of the equipment components (such as springs, mechanical switches, etc.) during the operation of the circuit breaker.

[0067] Step S204: determining the key time interval and linear slope of the coil current signal as current characteristics.

[0068] Among them, the key time interval is the time corresponding to the target circuit breaker from the start of action to the iron core triggering the operating mechanism; the linear slope is determined based on the current fitting value corresponding to the first turning point of the coil current curve equation and the start of action of the target circuit breaker. The first turning point is used to indicate the time corresponding to the first time the coil current signal changes rapidly from high to low.

[0069] Specifically, the server can obtain the first time when the target circuit breaker starts to act, and the second time when the iron core in the target circuit breaker touches the operating mechanism after the target circuit breaker starts to act. The time interval between the first time and the second time is determined as the key time interval. The first time and the second time can be obtained by detecting the current action of the target circuit breaker through the detection device in the target circuit breaker, and recording the time when the action starts and the iron core touches the operating mechanism. The first time and the second time can also be obtained based on the coil current signal of the target circuit breaker, as well as the signal characteristics corresponding to the start of the action and the signal characteristics when the iron core touches the operating mechanism, to determine the corresponding time points, thereby obtaining the first time and the second time.

[0070] After the server obtains the coil current signal, it can fit the coil current signal to obtain the coil current curve equation corresponding to the target circuit breaker. The server can obtain the time corresponding to the first turning point and the time when the target circuit breaker starts to act, and determine the linear slope between the above two moments based on the current fitting values ​​corresponding to the above two moments. Finally, the key time interval and linear slope are determined as the current characteristics of the target circuit breaker.

[0071] Step S206: determining a power spectrum based on an envelope corresponding to the vibration signal, and determining an index value of the power spectrum in a key frequency interval as a vibration feature.

[0072] Among them, the critical frequency interval is determined based on the change of the spectral kurtosis coefficient value in each interval of the vibration signal; the envelope is a mathematical tool to describe the change of the signal or waveform, which is usually used to analyze and visualize the basic characteristics of the signal. The envelope is essentially a curve that depicts the change of the maximum or minimum value of the signal. The power spectrum is used to represent the power distribution of the signal at different frequencies, reflecting the distribution of the signal's energy in the frequency domain.

[0073] Specifically, the server can preprocess the acquired vibration signal, such as filtering and noise removal. The server can determine the envelope corresponding to the preprocessed vibration signal, for example, the server calculates the envelope by the Hilbert-Huang transform principle. The server can convert the envelope by the indirect autocorrelation method to obtain the corresponding power spectrum. For the power spectrum, the power spectrum energy value corresponding to each key frequency interval is determined, and through a statistical strategy, multiple index values ​​composed of the power spectrum energy value are determined, and the index value is determined as the vibration feature.

[0074] Step S208: Based on the pre-built multiple single-class support vector machines, the current characteristics and the vibration characteristics are identified to obtain the fault diagnosis result of the target circuit breaker.

[0075] Among them, the single-class support vector machine is trained based on the current characteristics and vibration characteristics of the circuit breaker of each type of fault. Each single-class support vector machine is used to predict a fault type.

[0076] Specifically, the server inputs the current characteristics and vibration characteristics of the target circuit breaker into a pre-built single-class support vector machine, and each single-class support vector machine can obtain a prediction result based on the input current characteristics and vibration characteristics. By combining the prediction results of multiple single-class support vector machines, the fault diagnosis result of the target circuit breaker can be obtained.

[0077] In the above-mentioned circuit breaker fault diagnosis method, the coil current signal and vibration signal of the target circuit breaker are obtained, and the time period corresponding to the start of the action of the target circuit breaker to the iron core triggering the operating mechanism is determined as the key time interval, and the linear slope corresponding to the current fitting value of the coil current curve equation at the first turning point and the moment when the target circuit breaker starts to act is determined, and the key time interval and the linear slope are used as the current characteristics of the target circuit breaker. The power spectrum is determined according to the envelope of the vibration signal, the key frequency interval is determined according to the spectral kurtosis coefficient value of each interval of the vibration signal, and the index value of the power spectrum in the key frequency interval is determined, and the index value is used as the vibration feature. The current signal and vibration signal are predicted by multiple pre-constructed single-class support vector machines, the fault diagnosis result of the target circuit breaker is determined, and the fault type of the target circuit breaker is obtained.

[0078] The coil current signal that can represent the target circuit breaker failure can be accurately obtained through the key time interval, and the linear slope can accurately obtain the signal changes generated when the target circuit breaker fails, thereby improving the accuracy of the current characteristics obtained. The accurate key frequency interval can be determined by the spectral kurtosis coefficient value, and the relationship between time and frequency can be accurately represented by the index value of the key frequency interval, thereby improving the accuracy of the vibration characteristics. The single-class support vector machine obtained by training the current characteristics and vibration characteristics of each circuit breaker can predict the fault diagnosis results of the target circuit breaker and improve the accuracy of fault identification.

[0079] In an exemplary embodiment, the specific implementation process of the step of “determining the critical time interval and linear slope of the coil current signal as the current characteristics” includes:

[0080] According to the coil current signal at the moment when the target circuit breaker is not actuated and the moment when the impact amplitude of the vibration signal appears, the critical time interval between the start of the action of the target circuit breaker and the iron core triggering the operating mechanism is determined; based on the adaptive least squares method, the coil current signal in the critical time interval is curve fitted to obtain the coil current curve equation; based on the curvature of the coil current curve equation at each moment, the moment of the first turning point is determined, and based on the current fitting values ​​corresponding to the moment of the first turning point and the moment when the target circuit breaker starts to act, the linear slope of the coil current signal is determined.

[0081] Specifically, first, the server can obtain the coil current signal and the vibration signal from the moment when the target circuit breaker is not actuated. The server can monitor the vibration signal, and when the vibration signal is detected to have an impact amplitude, record the corresponding moment, and use it as the moment when the target circuit breaker starts to act or the core triggers the operating mechanism. Based on this, the server can determine the moment when the coil current signal undergoes an instantaneous change based on the acquired coil current signal, as the moment when the core triggers the operating mechanism. And the moment when the vibration signal has an impact amplitude before the core triggers the operating mechanism is used as the moment when the target circuit breaker starts to act. The server determines the time interval formed by the moment when the target circuit breaker starts to act and the moment when the core triggers the operating mechanism as the key time interval.

[0082] Secondly, the server can use the adaptive least squares method to perform curve fitting on the coil current signal in the critical time interval. The least squares method mainly uses the root mean square error and the adjusted goodness of fit as evaluation indicators, and achieves the best fit of the sample value through step-by-step iteration. For example, the server can define the number of terms n of the initial polynomial as 0, that is, constant fitting, and construct a constant fitting equation based on the detected coil current value and the least squares fitting, and obtain the fitting value at each moment.

[0083] After that, the root mean square error and the adjusted goodness of fit coefficient are calculated. The number of terms n is increased, and the least squares fitting is performed again to obtain the fitting values ​​at each moment, and the root mean square error and the adjusted goodness of fit coefficient are calculated again. The server determines whether the goodness of fit coefficient obtained for the second time is greater than the preset coefficient value. If it is less than the preset coefficient value, the root mean square error obtained for the second time is updated to the root mean square error obtained for the first time, and the number of terms n is continued to be increased, and the process is returned to perform the least squares fitting again to obtain the fitting values ​​at each moment, and the steps of calculating the root mean square error and the adjusted goodness of fit coefficient are repeated until the goodness of fit coefficient is greater than the preset coefficient value.

[0084] When it is determined that the goodness of fit coefficient is greater than the preset coefficient value, it can be determined whether the current root mean square error is greater than the previous root mean square error. If it is determined that the current root mean square error is less than the previous root mean square error, the least square fitting is returned to be performed again to obtain the fitting value at each moment, and the steps of calculating the root mean square error and the adjusted goodness of fit coefficient are repeated, and the previous root mean square error is updated. If it is determined that the current root mean square error is greater than the previous root mean square error, it can be determined that the number of curve fitting items outputted is n-1, and the fitted coil current curve equation is obtained.

[0085] Finally, the server determines the curvature corresponding to each moment according to the coil current curve equation. The server determines the moment corresponding to the first turning point of the coil current value from high to low according to the curvature at different moments as the moment of the first turning point. The server can determine the linear slope between the two current fitting values ​​according to the current fitting value at the moment of the first turning point and the current fitting value at the moment when the target circuit breaker starts to operate.

[0086] In this embodiment, the critical time interval can be determined by detecting the coil current signal and vibration signal of the target circuit breaker, analyzing the time when the impact amplitude appears, and the time when the target circuit breaker does not act and starts to act. The coil current signal is curve-fitted to obtain the coil current curve equation, and the first turning point is determined by the curvature corresponding to the coil current curve equation at each moment, and finally the linear slope of the coil current signal is obtained, which can accurately obtain the critical time interval, thereby improving the accuracy of the linear slope.

[0087] In an exemplary embodiment, the specific implementation process of the step of "determining the time of the first turning point based on the curvature of the coil current curve equation at each time" includes:

[0088] The sliding window strategy is used to determine the time of the local minimum of the coil current curve equation in the key time interval and the time of the first local maximum; the time of the target local minimum is determined as the time of the first turning point.

[0089] Among them, the target local minimum is selected from the local minimum values, and the time of the target local minimum value is greater than the time of the first local maximum value.

[0090] Specifically, after determining the coil current curve equation, the server can determine the curvature of each point through a sliding window strategy, and determine multiple local minima and local maxima within the critical time interval. For example, the curvature can be a second-order derivative. When the first-order derivative corresponding to the curvature is equal to 0, and the second-order derivative corresponding to the moment is less than 0, the moment is determined to be a local maximum. When the first-order derivative corresponding to the curvature is equal to 0, and the second-order derivative corresponding to the moment is greater than 0, the moment is determined to be a local minimum. The server can determine the moment of the first local maximum, and based on the moment as a reference, determine the local minimum after the moment as the target local minimum, and determine the moment corresponding to the target local minimum as the moment of the first turning point.

[0091] In this embodiment, multiple local minima and local maxima in the key time interval are determined by calculating the curvature, and the local minimum value greater than the first local maximum value is selected as the first turning point, thereby obtaining the moment of the first turning point, which can improve the accuracy and efficiency of obtaining the first turning point.

[0092] In an exemplary embodiment, the specific implementation process of the step of “determining the power spectrum based on the envelope corresponding to the vibration signal” includes:

[0093] Based on the peak coefficient of each window corresponding to the vibration signal, the vibration signal is denoised to obtain the denoised vibration signal. The denoised vibration signal in the key frequency range is filtered by a Chebyshev filter to obtain a filtered vibration signal. Based on the Hilbert-Huang transform strategy, the filtered vibration signal is processed to obtain the envelope corresponding to the filtered vibration signal. Based on the indirect autocorrelation method, the power spectrum corresponding to the envelope is determined.

[0094] First, the server can denoise the vibration signal based on the peak coefficient of each window corresponding to the vibration signal to obtain a denoised vibration signal. Specifically, the server can calculate the peak coefficient value of each window by a non-overlapping sliding window method, and then calculate the relationship between the local minimum and the local maximum to remove invalid signals. For example, if the power frequency is 50Hz and the sampling frequency of the vibration sensor is 20KHz, the sliding window size is defined as 4000 points, and then the peak coefficient of each window is calculated.

[0095] Secondly, the server can filter the denoised vibration signal in the key frequency interval through a Chebyshev filter to obtain a filtered vibration signal. Among them, if the minimum value of the key frequency interval is the preset frequency value, it is filtered through a low-pass filter; if the minimum value of the key frequency interval is greater than the preset frequency value, it is filtered through a band-pass filter; if the minimum value of the key frequency interval is less than the preset frequency value, it is filtered through a high-pass filter. For example, if the minimum value of the key frequency interval is 0Hz, a low-pass filter is used, if it is greater than 0Hz, a band-pass filter is used, and in other cases, a high-pass filter is used.

[0096] Then, the server can process the filtered vibration signal based on the Hilbert-Huang transform strategy to obtain the envelope curve corresponding to the filtered vibration signal. The envelope curve obtained by the Hilbert-Huang transform in the embodiment of the present application.

[0097] Finally, the server can determine the power spectrum corresponding to the envelope based on the indirect autocorrelation method. Specifically, the server can perform Fourier transform on the vibration signal of the envelope to obtain F(w). The autocorrelation function R(t) is obtained by calculating the square value of F(w) and performing an inverse Fourier transform. By reconnecting the autocorrelation function R(t), R'(t) is obtained, and the window function and window size are defined. The server can correct R'(t) through the window function to obtain the corrected R"(t). After correction, Fourier transform is performed to obtain F'(w). Afterwards, the server corrects the amplitude of F'(w) to obtain the power spectrum.

[0098] In this embodiment, denoising is performed through the peak coefficient of each window, filtering is performed through Chebyshev filtering, the envelope is determined through Hilbert-Huang transform, and the power spectrum is determined based on the indirect autocorrelation method, which can improve the accuracy of the vibration signal and the accuracy of the power spectrum.

[0099] In an exemplary embodiment, the specific implementation process of the step of “determining the index value of the power spectrum in the key frequency interval as the vibration characteristic” includes:

[0100] For each frequency interval in the critical frequency interval, determine the power spectrum energy value of each frequency line corresponding to the frequency interval in the power spectrum; determine the average value, variance value and center of gravity value corresponding to the power spectrum energy value of each frequency line, which is the vibration characteristic of the critical frequency interval.

[0101] Specifically, the key frequency interval may include multiple frequency intervals. For each frequency interval, the server may determine the power spectrum energy value of each frequency line corresponding to the frequency interval based on the obtained power spectrum. The server may determine the statistical index value corresponding to the power spectrum energy value of all frequency lines, such as determining the mean value, variance value and center of gravity value, and use the statistical index value as the vibration feature corresponding to the key frequency interval.

[0102] In this embodiment, the vibration characteristics of the key frequency interval are determined by the index value corresponding to the power spectrum energy value of each frequency line in the key frequency interval, which can improve the data dimension of the vibration characteristics and thus improve the richness of the vibration characteristics.

[0103] In an exemplary embodiment, before the step of “determining the power spectrum based on the envelope corresponding to the vibration signal”, the method further includes:

[0104] According to the binary classification recursive tree, the initial frequency interval of the vibration signal is split to obtain a first frequency interval and a second frequency interval. The spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval are determined. If any spectral kurtosis coefficient value is greater than the spectral kurtosis threshold, the frequency interval corresponding to the spectral kurtosis coefficient value greater than the spectral kurtosis threshold is determined as the initial frequency interval.

[0105] Return to the step of splitting the initial frequency interval of the vibration signal to obtain a first frequency interval and a second frequency interval, until the spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval, respectively, are both less than the spectral kurtosis threshold.

[0106] If the spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval respectively are both less than or equal to the spectral kurtosis threshold, the initial frequency interval is determined to be a key frequency interval.

[0107] Specifically, the server can calculate the intermediate frequency and divide the initial frequency interval according to the vibration signal into half to obtain the first frequency interval and the second frequency interval. For example, if the initial frequency interval is 0 to 1000, the first frequency interval can be 0 to 500, and the second frequency interval can be 500 to 1000. The server can calculate the spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval according to the envelope spectrum, respectively, to obtain the first spectral kurtosis coefficient and the second spectral kurtosis coefficient, the first spectral kurtosis coefficient corresponds to the first frequency interval, and the second spectral kurtosis coefficient corresponds to the second frequency interval.

[0108] Based on this, the server can determine whether the first spectral kurtosis coefficient and the second spectral kurtosis coefficient are both greater than the spectral kurtosis threshold. If any one of the spectral kurtosis coefficients is greater than the spectral kurtosis threshold, the frequency interval corresponding to the spectral kurtosis coefficient is determined to be the initial frequency interval, and the updated initial frequency interval is segmented. The first frequency interval and the second frequency interval corresponding to the initial frequency interval are obtained again until the first spectral kurtosis coefficient and the second spectral kurtosis coefficient are both less than the spectral kurtosis threshold. Optionally, if the first spectral kurtosis coefficient and the second spectral kurtosis coefficient are both greater than the spectral kurtosis threshold, the frequency interval corresponding to the spectral kurtosis coefficient with a larger spectral kurtosis threshold can be selected as the initial frequency interval, and the initial frequency interval is segmented.

[0109] If the spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval are both less than or equal to the spectral kurtosis threshold, the initial frequency interval can be determined to be a critical frequency interval. Optionally, after the server determines the critical frequency intervals corresponding to the vibration signals corresponding to each circuit breaker under different conditions, the union of the critical frequency intervals can be used as the final critical frequency interval.

[0110] In this embodiment, the vibration signal is split multiple times through a binary classification recursive tree, and the spectral kurtosis coefficient value is used to determine whether to iterate the split, and the initial frequency interval is updated. When the initial frequency interval is not after the split, the initial frequency interval can be determined as the key frequency interval, which can improve the accuracy of determining the key frequency interval.

[0111] In an exemplary embodiment, the specific implementation process of the step of "identifying current features and vibration features based on a plurality of pre-built single-class support vector machines to obtain a fault diagnosis result of a target circuit breaker" includes:

[0112] The current characteristics and vibration characteristics are taken as a group of characteristics and respectively input into each trained single-class support vector machine to obtain the prediction results output by each single-class support vector machine; the prediction results of the single-class support vector machine are the first prediction value or the second prediction value; if the prediction results of each single-class support vector machine are the first prediction value, then the fault diagnosis result of the target circuit breaker is determined to be a new fault type; if there is a single-class support vector machine whose prediction result is the second prediction value, then the fault diagnosis result of the target circuit breaker is determined to be the fault type corresponding to the single-class support vector machine whose prediction result is the second prediction value.

[0113] Among them, the trained single-class support vector machines are all obtained through training of current characteristics and vibration characteristics of circuit breakers under different fault conditions, and each single-class support vector machine is used to predict the prediction result of a fault type.

[0114] Specifically, the server can use the current characteristics and vibration characteristics corresponding to the target circuit breaker as a set of characteristics, and input them into each trained single-class support vector machine respectively, and each trained single-class support vector machine outputs the corresponding prediction result, thereby obtaining a prediction result set. Since the prediction result of the single-class support vector machine is one of the first prediction value or the second prediction value. And the first prediction value is used to indicate that the fault type of the target circuit breaker is not a fault type that can be predicted by the single-class support vector machine, and the second prediction value is used to indicate that the fault type of the target circuit breaker is a fault type corresponding to the single-class support vector machine.

[0115] Based on this, if the prediction results of each single-class support vector machine are the first prediction value, it can be determined that the fault diagnosis result of the target circuit breaker is a new fault type that has not been seen before. If there is any single-class support vector machine whose prediction result is the second prediction value, it can be determined that the fault diagnosis result of the target circuit breaker is the fault type corresponding to the single-class support vector machine whose prediction result is the second prediction value. For example, there are three single-class support vector machines, and the first prediction value is -1, and the second prediction value is 1. If the output result of the single-class support vector machine is (-1, -1, -1), it means that the fault diagnosis result of the target circuit breaker is a new fault type. If the output result of the single-class support vector machine is (-1, 1, -1), it means that the fault diagnosis result of the target circuit breaker is the fault type corresponding to the second single-class support vector machine.

[0116] Optionally, the server can train multiple single-class support vector machines. Determine the fault type corresponding to each single-class support vector machine. At this time, there are C categories, and one of the categories i can be selected. The server can annotate the feature set (training sample), and the label value for the i-th fault type is 1, and the label value for other fault types is -1. After that, the server can standardize the feature set of the i-th category, substitute the standardized feature set into the PSO-OcSVM model framework, and determine the optimal model parameters to obtain the optimal model, and standardize the feature sets of the remaining categories and substitute them into the model for verification to determine the model effect. When the model effect meets the training completion condition, it is determined that the training of the single-class support vector machine corresponding to the i-th category is completed. The server can train the PSO-OcSVM model framework corresponding to the C categories respectively to obtain the optimal model, thereby obtaining multiple trained single-class support vector machines. Among them, PSO-OcSVM is an algorithm that combines particle swarm optimization (PSO) and single-class support vector machine (OcSVM). Among them, PSO is an optimization algorithm, and OcSVM is a machine learning method for anomaly detection.

[0117] In this embodiment, a set of prediction results can be obtained by predicting the current characteristics and vibration characteristics through multiple single-class support vector machines, and the fault diagnosis result of the target circuit breaker can be determined according to the distribution of the first prediction value and the second prediction value in the prediction results. Multiple single-class support vector machines can predict multiple existing fault types and new fault types that have never appeared, which can improve the accuracy of fault prediction.

[0118] The following describes in detail the specific implementation process of the above-mentioned circuit breaker fault diagnosis method in combination with a specific embodiment. The coil current signal contains rich circuit breaker status information. During the opening and closing process of the high-voltage circuit breaker, any movement details of the iron core will cause the current of the DC electromagnetic operation coil to change. Therefore, by monitoring the coil current during the opening and closing process, the working condition of the secondary control circuit of the high-voltage circuit breaker and the movement of the iron core can be roughly understood, providing an auxiliary criterion for maintenance personnel. In an example, the coil current signal can reflect the status of the iron core stroke, iron core jamming, coil status, iron latch and valve status connected to the iron core top rod, etc. Most of the analysis techniques for coil current are based on the typical coil current fluctuation form and divided into stages according to mechanical characteristics, and then the time and current amplitude of each stage are extracted as signal feature quantities, thereby realizing circuit breaker fault diagnosis. The typical coil current waveform is shown in Figure 3. The waveform is divided into four stages: (1) Stage 1: t0~t1 is the stage when the iron core is stationary and about to move, and the opening and closing command arrives at time t0; (2) Stage 2: t1~t2 is the stage when the iron core moves, and the iron core triggers the tripping mechanism of the operating mechanism at time t2; (3) Stage 3: t2~t3 is the stage when the iron core stops moving. If this stage lasts long enough, the current will reach the DC stable value; (4) Stage 4: t3~t4 is the stage when the coil current is cut off. After the auxiliary switch is actuated in this stage, an arc is generated between the contacts of the auxiliary switch, and the arc voltage rises rapidly, forcing the current to decrease rapidly to zero.

[0119] The existing coil current analysis technology is mainly based on the key moment point information of the typical coil current. However, the actual collected coil current is not completely consistent with the typical coil current waveform, and there is noise interference, so there will be deviations in directly locating the time point. For example, Figures 4 and 5 are the coil currents actually collected by the circuit breaker during the opening and closing process. Compared with Figure 3, it can be found that there are great differences.

[0120] The vibration signal of a high-voltage circuit breaker is a non-stationary impact signal, which includes a variety of transient vibration events, which may contain a variety of status information, such as contact wear, loose screws, transmission mechanism and many other mechanical fault signs.

[0121] In the embodiment of the present application, the curve characteristics of the coil current can be extracted. For the coil current of the circuit breaker, the main method is to extract the first turning point of the fitted coil current curve equation. The linear slope k between the time when the circuit breaker starts to operate and the time from the starting point to the time when the iron core touches the operating mechanism As a current feature, the current feature can be combined with the vibration feature to realize fault diagnosis.

[0122] First, the key interval of the coil current is extracted. According to the coil current value when the circuit breaker is not in action and the time when the impact amplitude of the vibration signal appears, the coil current signal I(t) and duration from the start of the circuit breaker to the core touching the operating mechanism are determined. .

[0123] Then, the adaptive least squares method is used to perform curve fitting on I(t). The adaptive least squares method mainly uses the root mean square error and the adjusted goodness of fit as evaluation indicators, and achieves the best fit of the sample value through step-by-step iteration. The specific steps are as follows:

[0124] (1) Define the number of terms n of the initial polynomial as 0, that is, constant fitting, and then construct a constant fitting equation based on the I(t) data to obtain the fitting value , and calculate the root mean square error and the adjusted goodness-of-fit coefficient .

[0125]

[0126] In the formula is the fitted value, is the actual value I(t), is the average value, and T represents the total amount of data.

[0127] (2) Increase the number of terms n, set n = n + 1, and then perform the least squares fit again to obtain a new fitting value , and calculate the root mean square error and the adjusted goodness-of-fit coefficient .

[0128] (3) Determine whether it is less than 0.95. If so, jump to step (2) and perform the following operations:

[0129]

[0130] if If it is greater than 0.95, go to step (4).

[0131] (4) Judgment Is the value greater than , if less than , then jump to step (2) and perform the same operation as in step (3), that is, change .if Value greater than Then the number of output curve fitting items is n-1, and the following curve equation can be obtained:

[0132]

[0133] Get the fitted coil current curve equation , calculate the curvature C(t) of the coil current curve equation at each point, then calculate the local minimum position and the first local maximum position of the curvature C(t) according to the sliding window method, and then select the local minimum value greater than the first local maximum position as the first turning point position of the coil current , and then calculate the slope k at the starting point. The calculation formulas for the curvature C(t) and the slope k are as follows:

[0134]

[0135] The adaptive least squares curve fitting method proposed in the embodiment of the present application can reduce the impact of noise on the later analysis, while retaining the change of the coil current waveform. As shown in Figure 6, the fitted curve is smoother while retaining the waveform change of the original signal. Secondly, the extracted curve features are calculated based on the attributes of the data itself, without the need for prior knowledge, which improves the robustness of the fault identification model and reduces dependence on the environment.

[0136] In an embodiment of the present application, the features of the key spectrum interval of the vibration signal can be extracted. The vibration signal of the high-voltage circuit breaker is a non-stationary impact signal, so the feature extraction of the vibration signal first refers to the fast spectral kurtosis principle to determine the key frequency intervals of different sample states, and then filters the original signal based on the Chebyshev II filter to remove invalid information, and then extracts the envelope based on the Hilbert-Huang transform, calculates the power spectrum, and finally extracts the mean, variance and centroid frequency of the power spectrum in the key frequency interval as feature vectors to construct a combined single-class support vector machine model. Specifically, it includes the following steps:

[0137] (1) Preprocess the data based on the peak coefficient value and using the sliding window method;

[0138] The non-overlapping sliding window method is used to calculate the kurtosis coefficient value (peak coefficient value) of each window, and then the relationship between the local minimum and the local maximum is calculated to remove invalid signals. Considering that the power frequency is 50Hz and the vibration sensor sampling frequency is 20KHz, the sliding window size is defined as 4000 points, and then the peak coefficient of each window is calculated. The result is shown in Figure 7. The vibration signal waveform after removing noise by this method is shown in Figure 8.

[0139] The peak factor calculation formula is as follows:

[0140]

[0141] In the formula Represents vibration signal The local peak value of Represents the number of local peaks extracted, is the signal length.

[0142] (2) Based on the fast spectral kurtosis principle, a binary classification recursive tree is constructed to identify the key frequency range of each state.

[0143] The frequency domain distribution of the collected vibration signal is calculated, and then the spectral kurtosis coefficient value of each interval after each layer division is calculated using the binary classification recursive tree. By comparing the changes in the calculated values, the key feature intervals of each state and each sample are determined. The specific algorithm flow is shown in Figure 9, where the spectral kurtosis coefficient calculation formula is as follows:

[0144]

[0145] Where is the envelope spectrum, represents the mean, and represents the absolute value.

[0146] (3) Using a Chebyshev II filter to filter the original signal according to the critical frequency range;

[0147] Through step (2), the critical frequency interval of each state sample can be determined, and then the signal is filtered using a Chebyshev II filter. If the minimum value of the critical frequency interval is 0 Hz, a low-pass filter is used; if it is greater than 0 Hz, a band-pass filter is used; in other cases, a high-pass filter is used. The filtering effect is shown in Figure 10.

[0148] (4) Extract the envelope of the filtered signal and calculate the power spectrum;

[0149] This process mainly calculates the power spectrum of the vibration signal, in which the Hilbert-Huang transform is used to extract the envelope, and then the power spectrum is calculated based on the indirect autocorrelation method. Generally speaking, the envelope obtained by the Hilbert-Huang transform can be transformed into the envelope spectrum by Fourier transform, but here the Fourier transform is suitable for stationary signals, and it is not suitable for the non-stationary vibration signal of the high-voltage circuit breaker, so the power spectrum calculation is chosen here. The envelope of the filtered signal is calculated using the principle of Hilbert-Huang transform. The Hilbert transform formula is as follows:

[0150]

[0151] In the formula is the envelope obtained by solution.

[0152] The power spectrum is calculated using the indirect autocorrelation method, which is mainly based on the theorem that the autocorrelation function and the power spectrum function are inverse Fourier transforms of each other. The specific calculation method is as follows:

[0153] ① The envelope S(t) of the vibration signal is the above , perform Fourier transform and get F(w).

[0154]

[0155] ② Calculate the square value of F(w) and perform inverse Fourier transform to obtain the autocorrelation function R(t).

[0156]

[0157] ③ Rejoin R(t) to get R'(t), and define the window function and window size width.

[0158]

[0159] In the above formula, l is the length of R(t).

[0160] In the embodiment of the present application, the window function can be specified as a triangular window function, and the function is as follows:

[0161]

[0162] In the above formula, n is the sampling point number.

[0163] ④ Use the window function to correct the R'(t) deviation and obtain R”(t).

[0164]

[0165] ⑤ Calculate the Fourier transform of R”(t) to obtain F'(w).

[0166]

[0167] ⑥ Correct the amplitude of F'(w) to obtain the power spectrum P(w).

[0168]

[0169] In the above formula, L is the length of R''(t).

[0170] (5) Extract the power spectrum mean, power spectrum variance and power spectrum centroid value of the key frequency interval.

[0171] According to the key frequency interval determined in step (2), the mean value, variance value and centroid value of the envelope spectrum in each key frequency interval are calculated. The calculation formulas for each index value are as follows:

[0172]

[0173] In the formula is the average value, is the variance value, is the value of the center of gravity frequency, is the power spectrum energy value, is the frequency value, is the number of frequency lines.

[0174] In the embodiment of the present application, by referring to the fast spectral kurtosis principle, the calculation of high and low pass filters is removed, the interval is directly intercepted to calculate the spectral kurtosis value, and only the part with large spectral kurtosis value is classified into the next layer until all the spectral kurtosis values ​​after division under a certain layer are lower than the values ​​before division. This technology does not need to manually define the number of decomposition layers, and can determine the optimal number of decomposition layers according to the data, and only performs half recursion, which shortens the running time to a certain extent.

[0175] In an exemplary embodiment, the server can build a model by combining single-class support vector machines. The current features and vibration features calculated in the above embodiment are combined to construct a feature vector. Then, single-class support vector machine models based on different states are built in turn, wherein the kernel function of the selected single-class support vector machine is a radial basis kernel function, and parameter optimization is implemented by a particle swarm algorithm. The main steps are to take a certain class as the "positive class" and the other classes as the "heterogeneous classes" in turn, and then use the "positive class" feature set as the training set to build a PSO-OcSVM model of this class, wherein a three-fold cross-validation method is used when training the model to improve the universality of the model. The algorithm framework is shown in Figure 11.

[0176] The principle of single-class support vector machine is similar to that of support vector machine. The main purpose is to find a hyperplane that separates normal samples from abnormal samples. This hyperplane is called the splitting hyperplane and is supported by support vectors. The support vectors are the normal sample points closest to the splitting hyperplane. The difference is that the single-class support vector machine determines whether it is a class, while the support vector machine determines which class it is. The main goal of a single-class support vector machine is to find abnormal or rare samples in a data set, and it does not require a large number of normal samples for training. This makes the single-class support vector machine very useful in dealing with high-dimensional data and very sparse anomaly detection problems. It also happens to be aimed at the characteristics of high-voltage circuit breakers that the amount of normal state sample data is greater than the fault samples, and the identification needs of fault categories that cannot be simulated. The application of this technology increases the ability to identify new types of high-voltage circuit breaker faults.

[0177] In one embodiment, as shown in FIG12 , a process of fault identification for multiple single-class support vector machines is disclosed. The identification results of the PSO-OcSVM model for n types of faults can be obtained. If all the identification results are -1, the fault type is determined to be a new fault type. If there is a state where the identification result is 1, the known fault type can be determined as a fault identification result. As shown in FIG13 , FIG13 shows the results of circuit breaker category identification based on normal, broken bar and undervoltage conditions. The figure mainly shows the results after dimensionality reduction of the feature vector, where the yellow area is the boundary area of ​​the model, and the points falling inside the yellow area are the correct states corresponding to the model identification, and the identification code is "1", while the points outside the area are "abnormal" states, and their identification code is "-1".

[0178] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0179] Based on the same inventive concept, the embodiment of the present application also provides a circuit breaker fault diagnosis device for implementing the circuit breaker fault diagnosis method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the circuit breaker fault diagnosis device provided below can refer to the limitations of the circuit breaker fault diagnosis method above, and will not be repeated here.

[0180] In an exemplary embodiment, as shown in FIG. 14 , a circuit breaker fault diagnosis device 1400 is provided, comprising: a signal acquisition module 1401, a current feature determination module 1402, a vibration feature determination module 1403 and a fault identification module 1404, wherein:

[0181] A signal acquisition module 1401 is used to acquire a coil current signal and a vibration signal of a target circuit breaker;

[0182] The current characteristic determination module 1402 is used to determine the key time interval and linear slope of the coil current signal as the current characteristic; the key time interval is the time corresponding to the target circuit breaker from the start of action to the iron core triggering the operating mechanism; the linear slope is determined based on the current fitting values ​​corresponding to the first turning point of the coil current curve equation and the start of action of the target circuit breaker;

[0183] The vibration characteristic determination module 1403 is used to determine the power spectrum based on the envelope corresponding to the vibration signal, and determine the index value of the power spectrum in the key frequency interval as the vibration characteristic; the key frequency interval is determined based on the change of the spectral kurtosis coefficient value of each interval of the vibration signal;

[0184] The fault identification module 1404 is used to identify the current characteristics and vibration characteristics based on multiple pre-built single-class support vector machines to obtain the fault diagnosis result of the target circuit breaker; the single-class support vector machine is trained based on the current characteristics and vibration characteristics of circuit breakers with various types of faults.

[0185] Furthermore, the current characteristic determination module 1402 is specifically used to: determine the critical time interval between the start of the target circuit breaker and the iron core triggering the operating mechanism according to the coil current signal at the time when the target circuit breaker is not actuated and the time when the impact amplitude of the vibration signal appears; perform curve fitting on the coil current signal in the critical time interval based on the adaptive least squares method to obtain the coil current curve equation; determine the time of the first turning point based on the curvature of the coil current curve equation at each time, and determine the linear slope of the coil current signal based on the current fitting values ​​corresponding to the time of the first turning point and the time when the target circuit breaker starts to act.

[0186] Furthermore, the current characteristic determination module 1402 is specifically used to: determine the time of the local minimum value of the coil current curve equation in the key time interval and the time of the first local maximum value according to the sliding window strategy; determine the time of the target local minimum value as the time of the first turning point; the target local minimum value is selected from the local minima, and the time of the target local minimum value is greater than the time of the first local maximum value.

[0187] Furthermore, the vibration characteristic determination module 1403 is specifically used to: denoise the vibration signal based on the peak coefficient of each window corresponding to the vibration signal to obtain a denoised vibration signal; filter the denoised vibration signal in the key frequency interval through a Chebyshev filter to obtain a filtered vibration signal; wherein, if the minimum value of the key frequency interval is a preset frequency value, filter it through a low-pass filter; if the minimum value of the key frequency interval is greater than the preset frequency value, filter it through a band-pass filter; if the minimum value of the key frequency interval is less than the preset frequency value, filter it through a high-pass filter; based on the Hilbert-Huang transform strategy, process the filtered vibration signal to obtain an envelope corresponding to the filtered vibration signal; based on the indirect autocorrelation method, determine the power spectrum corresponding to the envelope.

[0188] Furthermore, the vibration characteristic determination module 1403 is specifically used to: determine, for each frequency interval in the critical frequency interval, a power spectrum energy value of each frequency line corresponding to the frequency interval in the power spectrum; determine the average value, variance value and center of gravity value corresponding to the power spectrum energy value of each frequency line, which is the vibration characteristic of the critical frequency interval.

[0189] Furthermore, the device also includes a frequency interval determination module, which is used to: split the initial frequency interval of the vibration signal according to the binary classification recursive tree to obtain a first frequency interval and a second frequency interval; determine the spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval respectively, if any spectral kurtosis coefficient value is greater than the spectral kurtosis threshold, then determine the frequency interval corresponding to the spectral kurtosis coefficient value greater than the spectral kurtosis threshold as the initial frequency interval; return to execute the step of splitting the initial frequency interval of the vibration signal to obtain the first frequency interval and the second frequency interval, until the spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval are both less than the spectral kurtosis threshold; if the spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval are both less than or equal to the spectral kurtosis threshold, then determine that the initial frequency interval is a critical frequency interval.

[0190] Furthermore, the fault identification module is specifically used to: take the current characteristics and the vibration characteristics as a group of characteristics, input them into each trained single-class support vector machine respectively, and obtain the prediction results output by each single-class support vector machine; the prediction results of the single-class support vector machine are the first prediction value or the second prediction value; if the prediction results of each single-class support vector machine are the first prediction value, then determine that the fault diagnosis result of the target circuit breaker is a new fault type; if there is a single-class support vector machine whose prediction result is the second prediction value, then determine that the fault diagnosis result of the target circuit breaker is the fault type corresponding to the single-class support vector machine whose prediction result is the second prediction value.

[0191] Each module in the above-mentioned circuit breaker fault diagnosis device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0192] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG15. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store coil current signals and vibration signals of each circuit breaker. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a circuit breaker fault diagnosis method is implemented.

[0193] Those skilled in the art will understand that the structure shown in FIG. 15 is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0194] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0195] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0196] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0197] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0198] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0199] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A circuit breaker fault diagnosis method, characterized in that: The method comprises: Acquire coil current signal and vibration signal of target circuit breaker; Determine the key time interval and linear slope of the coil current signal as current characteristics; the key time interval is the time corresponding to the start of the target circuit breaker from the start of the action to the iron core triggering the operating mechanism; the linear slope is determined based on the current fitting values ​​corresponding to the first turning point of the coil current curve equation and the start of the action of the target circuit breaker; Based on the envelope corresponding to the vibration signal, a power spectrum is determined, and an index value of the power spectrum in a key frequency interval is determined as a vibration feature; the key frequency interval is determined based on a change in a spectral kurtosis coefficient value in each interval of the vibration signal; Based on a plurality of pre-constructed single-class support vector machines, the current characteristics and the vibration characteristics are identified to obtain a fault diagnosis result of the target circuit breaker; the single-class support vector machine is trained based on the current characteristics and vibration characteristics of circuit breakers with various types of faults.

2. The method according to claim 1, characterized in that Determining the critical time interval and linear slope of the coil current signal as current characteristics includes: Determine the critical time interval between the start of the action of the target circuit breaker and the iron core triggering the operating mechanism according to the coil current signal at the time when the target circuit breaker is not actuated and the time when the impact amplitude of the vibration signal appears; Performing curve fitting on the coil current signal within the critical time interval based on an adaptive least squares method to obtain a coil current curve equation; Based on the curvature of the coil current curve equation at each moment, the moment of the first turning point is determined, and based on the current fitting values ​​corresponding to the moment of the first turning point and the moment of the start of action of the target circuit breaker, the linear slope of the coil current signal is determined.

3. The method according to claim 2, characterized in that Determining the time of the first turning point based on the curvature of the coil current curve equation at each time includes: Determine the time of the local minimum value of the coil current curve equation in the key time interval and the time of the first local maximum value according to the sliding window strategy; The time of determining the target local minimum is the time of the first turning point; the target local minimum is selected from the local minimums, and the time of the target local minimum is greater than the time of the first local maximum.

4. The method according to claim 1, characterized in that: The determining of the power spectrum based on the envelope corresponding to the vibration signal comprises: De-noising the vibration signal based on the peak coefficient of each window corresponding to the vibration signal to obtain a de-noised vibration signal; Filtering the denoised vibration signal in the key frequency range through a Chebyshev filter to obtain a filtered vibration signal; Wherein, if the minimum value of the key frequency interval is a preset frequency value, filtering is performed through low-pass filtering; if the minimum value of the key frequency interval is greater than the preset frequency value, filtering is performed through band-pass filtering; if the minimum value of the key frequency interval is less than the preset frequency value, filtering is performed through high-pass filtering; Based on the Hilbert-Huang transform strategy, the filtered vibration signal is processed to obtain an envelope corresponding to the filtered vibration signal; Based on the indirect autocorrelation method, a power spectrum corresponding to the envelope is determined.

5. The method according to claim 4, characterized in that Determining the index value of the power spectrum in the key frequency interval as a vibration feature includes: For each frequency interval of the key frequency interval, determining a power spectrum energy value of each frequency line corresponding to the frequency interval in the power spectrum; The average value, variance value and center of gravity value corresponding to the power spectrum energy value of each frequency line are determined as the vibration characteristics of the key frequency range.

6. The method according to claim 4, characterized in that Before determining the power spectrum based on the envelope corresponding to the vibration signal, the method further includes: According to the binary classification recursive tree, the initial frequency interval of the vibration signal is split to obtain a first frequency interval and a second frequency interval; Determine the spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval respectively, and if any one of the spectral kurtosis coefficient values ​​is greater than a spectral kurtosis threshold, determine the frequency interval corresponding to the spectral kurtosis coefficient value greater than the spectral kurtosis threshold as the initial frequency interval; Returning to the step of splitting the initial frequency interval of the vibration signal to obtain a first frequency interval and a second frequency interval, until the spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval, respectively, are both less than the spectral kurtosis threshold; If the spectral kurtosis coefficient values ​​corresponding to the first frequency interval and the second frequency interval respectively are both less than or equal to the spectral kurtosis threshold, the initial frequency interval is determined to be a key frequency interval.

7. The method according to claim 1, characterized in that The identifying of the current feature and the vibration feature based on the pre-built multiple single-class support vector machines to obtain the fault diagnosis result of the target circuit breaker includes: The current feature and the vibration feature are used as a group of features, and are respectively input into each trained single-class support vector machine to obtain a prediction result output by each single-class support vector machine; the prediction result of the single-class support vector machine is a first prediction value or a second prediction value; If the prediction results of each of the single-class support vector machines are the first prediction value, determining that the fault diagnosis result of the target circuit breaker is a new fault type; If there is a single-class support vector machine whose prediction result is the second prediction value, it is determined that the fault diagnosis result of the target circuit breaker is the fault type corresponding to the single-class support vector machine whose prediction result is the second prediction value.

8. A circuit breaker fault diagnosis device, characterized in that: The device comprises: A signal acquisition module, used to acquire a coil current signal and a vibration signal of a target circuit breaker; A current characteristic determination module, used to determine the key time interval and linear slope of the coil current signal as the current characteristic; the key time interval is the time corresponding to the start of the action of the target circuit breaker to the time when the iron core triggers the operating mechanism; the linear slope is determined based on the current fitting values ​​corresponding to the first turning point of the coil current curve equation and the start of the action of the target circuit breaker; A vibration feature determination module, used to determine a power spectrum based on an envelope corresponding to the vibration signal, and determine an index value of the power spectrum in a key frequency interval as a vibration feature; the key frequency interval is determined based on a change in a spectral kurtosis coefficient value in each interval of the vibration signal; A fault identification module is used to identify the current characteristics and the vibration characteristics based on a plurality of pre-constructed single-class support vector machines to obtain a fault diagnosis result of the target circuit breaker; the single-class support vector machines are trained based on the current characteristics and vibration characteristics of circuit breakers with various types of faults.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images