Switch equipment fault positioning method, device, equipment, medium and product

Abstract

The embodiment of the invention discloses a switch equipment fault positioning method and device, equipment, a medium and a product, and the method comprises the steps: obtaining original vibration data collected by at least two sensors on switch equipment, extracting a time domain feature and a frequency domain feature, determining a fault type judgment result according to the time domain feature and the frequency domain feature, and determining the fault type of the switch equipment according to the fault type judgment result. Performing feature fusion on the time domain feature and the frequency domain feature to obtain a fusion feature, and determining a target fault point judgment feature according to the time domain feature, the frequency domain feature, the fusion feature and a fault type judgment result; and performing model calculation according to the target fault point judgment features to obtain a target fault point judgment result. According to the technical scheme of the embodiment of the invention, the problem that the fault point of the switch equipment cannot be accurately judged at present is solved, fault positioning can be carried out through comprehensive feature extraction and comprehensive analysis of signal features acquired by multiple sensors and through machine learning by accurately utilizing the features, the accuracy of fault positioning is improved, and the fault positioning efficiency is improved. And the operation stability of switchgear is ensured.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of vibration analysis, and in particular, to a method, device, equipment, medium, and product for fault location of switchgear. Background Art

[0002] With the improvement of industrial automation, switchgear, as a key component in the circuit control system, the stability and reliability of its performance have been increasingly emphasized. High-frequency vibration analysis, as an advanced monitoring means, has a wide range of application requirements in the switch mechanical characteristic test. At present, in important fields such as the power system, railway transportation, aerospace, and petrochemical industry, the use frequency of switchgear is extremely high. It has become an urgent need to monitor the health status of these devices in real time, prevent potential failures, and ensure the safe operation of the devices.

[0003] In the above fields, switchgear generates vibration signals during operation, and these signals contain important information about the operating state of the device. Therefore, vibration analysis technology is used to monitor the mechanical characteristics of switchgear to facilitate the real-time detection of early device failures. Summary of the Invention

[0004] The embodiments of the present invention provide a method, device, equipment, medium, and product for fault location of switchgear, which can comprehensively extract features and comprehensively analyze the features of signals collected by multiple sensors, and accurately use the features for fault location through machine learning, improve the accuracy of fault location, and ensure the operation stability of switchgear.

[0005] In a first aspect, the embodiments of the present invention provide a method for fault location of switchgear, and the method includes:

[0006] Obtain the original vibration data collected by at least two sensors arranged on the switchgear, where each sensor is arranged at a different part of the switchgear;

[0007] Perform time-domain feature extraction and frequency-domain feature extraction on the original vibration data to obtain time-domain features and frequency-domain features;

[0008] Perform feature splicing on the time-domain features and frequency-domain features to obtain target fault type judgment features;

[0009] Input the target fault type judgment features into a fault type judgment model to obtain a fault type judgment result, where the fault type judgment model includes support vector machine models corresponding to at least two fault types;

[0010] Feature fusion is performed on the time-domain features and frequency-domain features according to the preset fault location analysis weight values of each sensor to obtain fused features, where the preset fault location analysis weight values are determined according to the historical failure rates of the parts of the switchgear corresponding to each sensor;

[0011] Feature splicing is performed based on the time-domain features, frequency-domain features, fused features, and fault type judgment results to obtain target fault point judgment features;

[0012] The target fault point judgment features are input into the first fault point judgment model based on a deep neural network to obtain a first fault point judgment result;

[0013] The target fault point judgment features are input into the second fault point judgment model based on a gradient boosting decision tree to obtain a second fault point judgment result;

[0014] Weighted fusion is performed on the first fault point judgment result and the second fault point judgment result to obtain the target fault point judgment result of the switchgear.

[0015] In a second aspect, an embodiment of the present invention provides a switchgear fault location device, and the device includes:

[0016] A vibration data acquisition module, configured to acquire original vibration data collected by at least two sensors arranged on the switchgear, where each sensor is arranged at a different part of the switchgear;

[0017] A feature extraction module, configured to perform time-domain feature extraction and frequency-domain feature extraction on the original vibration data to obtain time-domain features and frequency-domain features;

[0018] A fault type judgment feature determination module, configured to splice the time-domain features and frequency-domain features to obtain target fault type judgment features;

[0019] A fault type judgment result determination module, configured to input the target fault type judgment features into a fault type judgment model to obtain a fault type judgment result, where the fault type judgment model includes support vector machine models corresponding to at least two fault types;

[0020] A fused feature determination module, configured to perform feature fusion on the time-domain features and frequency-domain features according to the preset fault location analysis weight values of each sensor to obtain fused features, where the preset fault location analysis weight values are determined according to the historical failure rates of the parts of the switchgear corresponding to each sensor;

[0021] A target fault point judgment feature determination module, configured to perform feature splicing based on the time-domain features, frequency-domain features, fused features, and fault type judgment results to obtain target fault point judgment features;

[0022] The first fault point determination result module is configured to input the target fault point determination feature into the first fault point determination model based on a deep neural network to obtain the first fault point determination result;

[0023] The second fault point determination result module is configured to input the target fault point determination feature into the second fault point determination model based on a gradient boosting decision tree to obtain the second fault point determination result;

[0024] The target fault point determination result module is configured to perform weighted fusion on the first fault point determination result and the second fault point determination result to obtain the target fault point determination result of the switchgear.

[0025] Thirdly, an embodiment of the present invention further provides a computer device, which includes:

[0026] One or more processors;

[0027] A memory for storing one or more programs;

[0028] When the above one or more programs are executed by the one or more processors, the one or more processors implement the switchgear fault location method provided in any embodiment of the present invention.

[0029] Fourthly, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the switchgear fault location method provided in any embodiment of the present invention.

[0030] Fifthly, an embodiment of the present invention further provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the switchgear fault location method provided in any embodiment of the present invention.

[0031] The embodiments in the above invention have the following advantages or beneficial effects:

[0032] In an embodiment of the present invention, by acquiring the original vibration data collected by at least two sensors arranged on the switching device, where each sensor is arranged at a different part of the switching device; performing time-domain feature extraction and frequency-domain feature extraction on the original vibration data to obtain time-domain features and frequency-domain features; splicing the time-domain features and frequency-domain features to obtain target fault type judgment features; inputting the target fault type judgment features into a fault type judgment model to obtain a fault type judgment result, where the fault type judgment model includes support vector machine models corresponding to at least two fault types; performing feature fusion on the time-domain features and frequency-domain features according to the preset fault location analysis weight values of each sensor, where the preset fault location analysis weight values are determined according to the historical failure rates of the parts of the switching device corresponding to each sensor; performing feature splicing according to the time-domain features, frequency-domain features, fusion features and fault type judgment results to obtain target fault point judgment features; inputting the target fault point judgment features into a first fault point judgment model based on a deep neural network to obtain a first fault point judgment result; inputting the target fault point judgment features into a second fault point judgment model based on a gradient boosting decision tree to obtain a second fault point judgment result; performing weighted fusion on the first fault point judgment result and the second fault point judgment result to obtain the target fault point judgment result of the switching device. The technical solution of the embodiment of the present invention solves the problem that the fault point of the switching device cannot be accurately judged at present, and can accurately perform fault location by using features through comprehensive feature extraction and comprehensive analysis of the signal features collected by multiple sensors, improving the accuracy of fault location and ensuring the operation stability of the switching device. Description of the Drawings

[0033] Figure 1 is a flowchart of a method for fault location of a switching device provided by an embodiment of the present invention;

[0034] Figure 2 is a flowchart of a method for fault location of a switching device provided by an embodiment of the present invention;

[0035] Figure 3 is a flowchart of a method for fault location of a switching device provided by an embodiment of the present invention;

[0036] Figure 4 is a schematic structural diagram of a switching device fault location device provided by an embodiment of the present invention;

[0037] Figure 5 is a schematic structural diagram of a computer device provided by an embodiment of the present invention. Detailed Embodiments

[0038] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present invention, rather than limiting the present invention. Additionally, it should be noted that for the sake of description, only the parts related to the present invention rather than all the structures are shown in the drawings.

[0039] Figure 1 The flowchart of a switchgear fault location method provided by an embodiment of the present invention is applicable to the scenario of switchgear fault location. This method can be executed by a switchgear fault location device, which can be implemented in software and / or hardware and integrated in a computer device with application development functions.

[0040] As Figure 1 shown, the switchgear fault location method of this embodiment includes the following steps:

[0041] S110. Obtain the original vibration data collected by at least two sensors arranged on the switchgear, where each sensor is arranged at a different part of the switchgear.

[0042] The sensors can be the same sensors, such as piezoelectric vibration sensors, magnetoelectric vibration sensors, or fiber optic vibration sensors, etc.

[0043] The installation positions can include switch contacts (or switch tips), drive devices (such as motors, hydraulic devices, spring mechanisms), mechanical linkage parts of the switch (such as connecting rods, hinge devices, etc.), operating mechanisms of electrical circuit breakers (such as operating levers, gearboxes of switch mechanisms), electrical connection terminals, and wiring terminals, etc.

[0044] The switch contact is one of the most critical components in an electrical switchgear. The closing or opening of the switch will cause mechanical shocks and electrical sparks, and these factors usually lead to vibrations. The vibration characteristics of the contacts can directly reflect the health status of the mechanical and electrical performance of the switch.

[0045] The drive device of the switchgear is responsible for controlling the operation of the switch. Imbalances or faults during the drive process (such as the out-of-step of the motor, pressure fluctuations in the hydraulic system, etc.) usually lead to abnormal vibrations. Monitoring the vibrations of the drive device helps to detect these problems in advance.

[0046] The mechanical linkage parts of the switchgear usually include connecting rods, hinge points, etc. These components will generate friction or impacts during the switch operation, thereby causing abnormal vibrations. Especially when the mechanical connection parts are worn or loose, the vibration signals will change.

[0047] The operating mechanism is usually composed of a series of levers, gears and other transmission components, which will generate certain vibrations during the switching operation. If there are problems inside the operating mechanism (such as gear damage, insufficient lubrication, excessive friction, etc.), it will lead to abnormal vibrations.

[0048] During the switching operation, especially under high current or high voltage conditions, the electrical connection terminals may cause local vibrations due to problems such as poor contact and looseness. The states of these components directly affect the electrical performance of the equipment.

[0049] Therefore, in this embodiment, by acquiring the vibration data, i.e., vibration signals, of parts such as the switch contacts and the driving device, and analyzing the vibration data, the fault points of the switching equipment are determined.

[0050] S120. Extract time-domain features and frequency-domain features from the original vibration data to obtain time-domain features and frequency-domain features.

[0051] The frequency-domain features can be extracted by means such as short-time Fourier transform, and the time-domain features can be extracted by relevant time-domain feature extraction algorithms, such as the root-mean-square calculation method, etc.

[0052] S130. Perform feature splicing on the time-domain features and frequency-domain features to obtain target fault type judgment features.

[0053] Feature splicing can be to splice the time-domain features after the frequency-domain features, or splice the frequency-domain features after the time-domain features to obtain the target fault type judgment features.

[0054] S140. Input the target fault type judgment features into the fault type judgment model to obtain the fault type judgment result.

[0055] Among them, the fault type judgment model includes support vector machine models corresponding to at least two fault types.

[0056] The support vector machine models adopt the one-against-many strategy, and each model corresponds to a unique target fault type. Input the target fault type judgment features into each support vector machine model, and each support vector machine model outputs the probability that the target fault type judgment features belong to the target fault type corresponding to the support vector machine model. Determine the fault type judgment result according to the highest probability among the probabilities.

[0057] S150. According to the preset fault location analysis weight values of each sensor, perform feature fusion on the time-domain features and frequency-domain features to obtain fusion features.

[0058] The time-domain features and frequency-domain features are weighted and fused according to the preset fault location analysis weight values to obtain the fused features. The preset fault location analysis weight values are determined according to the historical failure rates of the parts of the switchgear corresponding to each sensor, so that the fused features can help determine the more likely fault points based on historical data.

[0059] S160. Feature splicing is performed based on the time-domain features, frequency-domain features, fused features, and fault type judgment results to obtain the target fault point judgment features.

[0060] The time-domain features, frequency-domain features, fused features, and fault type judgment results are spliced in a preset order to obtain the target fault point judgment features. The preset order can be set according to actual needs.

[0061] When the device has multiple fault modes or different types of fault signals have obvious differences, the fault type judgment results can clearly distinguish these different states or modes, thereby improving the fault location, that is, the accuracy of the fault point judgment. Especially in the case of complex fault modes, more data noise, or multi-category fault classification, the pattern recognition of the support vector machine can help the fault point judgment model extract more effective features, making the training of the fault point judgment model more accurate. That is, step S140 gives a directional fault type, and in the machine learning algorithms of steps S170 and S180, the corresponding features can be focused on directionally.

[0062] S170. The target fault point judgment features are input into the first fault point judgment model based on a deep neural network to obtain the first fault point judgment result.

[0063] S180. The target fault point judgment features are input into the second fault point judgment model based on a gradient boosting decision tree to obtain the second fault point judgment result.

[0064] Both the first fault point judgment model and the second fault point judgment model are trained with the sample fault point judgment features processed by the feature extraction and feature fusion methods such as in steps S110 - S160. In the prediction stage, they can output the fault point judgment results according to the features learned in the training stage. The fault point judgment results can be the judgment results of the specific fault parts. The first fault point judgment model calculates through each layer and represents the possibility of belonging to each possible fault point in the form of a probability distribution, and determines the fault point corresponding to the highest possibility as the first fault point judgment result. Each decision tree in the second fault point judgment model performs node splitting and judgment based on the target fault point judgment features, and finally obtains the second fault point judgment result by weighted combination of the results of all decision trees.

[0065] S190. Perform weighted fusion on the first fault point judgment result and the second fault point judgment result to obtain the target fault point judgment result of the switchgear.

[0066] Different fault point judgment methods or data sources have their own advantages and disadvantages. Therefore, in this embodiment, the first fault point judgment result and the second fault point judgment result are comprehensively weighted and fused to reduce the error and uncertainty of single judgment, so that the target fault point judgment result can more accurately locate the fault point. The weights of different fault point judgment results can be set according to actual needs.

[0067] The technical solution of this embodiment is as follows: acquire the original vibration data collected by at least two sensors arranged on the switchgear, where each sensor is arranged at a different part of the switchgear; perform time-domain feature extraction and frequency-domain feature extraction on the original vibration data to obtain time-domain features and frequency-domain features; splice the time-domain features and frequency-domain features to obtain target fault type judgment features; input the target fault type judgment features into a fault type judgment model to obtain a fault type judgment result, where the fault type judgment model includes support vector machine models corresponding to at least two fault types; according to the preset fault location analysis weight values of each sensor, perform feature fusion on the time-domain features and frequency-domain features to obtain fusion features, where the preset fault location analysis weight values are determined according to the historical failure rates of the parts of the switchgear corresponding to each sensor; perform feature splicing according to the time-domain features, frequency-domain features, fusion features and the fault type judgment result to obtain target fault point judgment features; input the target fault point judgment features into a first fault point judgment model based on a deep neural network to obtain a first fault point judgment result; input the target fault point judgment features into a second fault point judgment model based on a gradient boosting decision tree to obtain a second fault point judgment result; perform weighted fusion on the first fault point judgment result and the second fault point judgment result to obtain the target fault point judgment result of the switchgear. The technical solution of the embodiment of the present invention solves the problem that the fault point of the switchgear cannot be accurately judged at present. Through comprehensive feature extraction and comprehensive analysis of the signal features collected by multiple sensors, accurate fault location can be achieved by machine learning using features, improving the accuracy of fault location and ensuring the operation stability of the switchgear.

[0068] Figure 2 It is a flowchart of a switchgear fault location method provided by an embodiment of the present invention. This embodiment and the switchgear fault location method in the above embodiment belong to the same inventive concept, and further describes the process of determining the fusion features. This method can be executed by a switchgear fault location device, and the device can be implemented in a software and / or hardware manner and integrated in a computer device with application development functions.

[0069] As Figure 2As shown in the figure, the method for fault location of the switching device in this embodiment includes the following steps:

[0070] S210. Obtain the original vibration data collected by at least two sensors arranged on the switching device, where each sensor is arranged at a different part of the switching device.

[0071] S220. Perform time-domain feature extraction and frequency-domain feature extraction on the original vibration data to obtain time-domain features and frequency-domain features.

[0072] In an alternative embodiment, the time-domain features include amplitude, root mean square value, and peak factor. Performing time-domain feature extraction on the original vibration data to obtain time-domain features may be to perform smoothing processing on the original vibration data to obtain smoothed vibration data; calculate the amplitude of the original vibration data according to the maximum value and minimum value of the vibration signal in the smoothed vibration data; calculate the root mean square value of the original vibration data according to the vibration signal and the number of signal sampling points in the smoothed vibration data; calculate the peak factor of the original vibration data according to the maximum value of the vibration signal and the root mean square value.

[0073] Exemplarily, the original vibration data can be smoothed by a moving average technique to obtain smoothed vibration data. Based on the smoothed vibration data, the amplitude can be calculated by the following formula:

[0074] A = max(x(t)) - min(x(t));

[0075] In the formula, A represents the amplitude; x(t) represents the time series signal of the smoothed vibration data, that is, the vibration signal; max(x(t)) represents the maximum value of the vibration signal; min(x(t)) represents the minimum value of the vibration signal;

[0076] The root mean square value can be calculated by the following formula:

[0077]

[0078] In the formula, RMS represents the root mean square value; N represents the number of signal sampling points; x i represents the i-th sampling point of the vibration signal;

[0079] The peak factor can be calculated by the following formula:

[0080]

[0081] In the formula, PF represents the peak factor; max(x(t)) represents the maximum value of the vibration signal.

[0082] In an alternative embodiment, the frequency-domain features include the main frequency component, the spectral width, and the peak frequency. Extracting the frequency-domain features from the original vibration data, the frequency-domain features obtained may be that the original vibration data is denoised to obtain denoised vibration data; the time-frequency features of the denoised vibration data are extracted by short-time Fourier transform: the main frequency component of the original vibration data is extracted according to the time-frequency features; the average frequency is calculated based on the frequency of the denoised vibration data, and the spectral width of the original vibration data is calculated according to the frequency, the average frequency, and the time-frequency features; according to the time-frequency features, the frequency point with the largest amplitude in the frequency is determined, and the frequency point with the largest amplitude is used as the peak frequency.

[0083] For the original vibration data, denoising processing is performed through a band-pass filter to obtain denoised vibration data. The frequency-domain features are calculated based on the denoised vibration data.

[0084] The calculation process of the main frequency component may be to capture the time-frequency features by short-time Fourier transform, and the calculation process of the time-frequency features is represented by the following formula:

[0085]

[0086] In the formula: X(t,f) represents the signal transformation result at time t and frequency f; x(τ) represents the time-series signal after denoising, that is, the vibration signal of the denoised vibration data; ω(τ - t) represents the Gaussian window function; e -j2πfτ represents the complex exponential basis function; t represents time; f represents frequency.

[0087] Taking the square of the modulus of the result of the short-time Fourier transform, that is, the transformed time-frequency features, the power spectral density is obtained, and the frequency corresponding to the maximum value of the power spectral density is found in each time frame to determine the main frequency component.

[0088] The spectral width calculation can be represented by the following formula:

[0089]

[0090] In the formula, W represents the spectral width, f mean represents the mean frequency of the spectrum, and f represents frequency;

[0091] Through the spectral width, the frequency expansion characteristics of the switchgear in different working states can be effectively described, thereby serving as a basis for judging the fault type and fault point. A larger spectral width may represent the working load of the equipment or the frequency expansion caused by a fault, while a smaller spectral width may represent that the equipment is in a relatively stable state.

[0092] The calculation process of the peak frequency can be represented by the following formula:

[0093]

[0094] where: f peak represents the peak frequency; represents the frequency point with the largest amplitude among the found frequencies; Re(X(t, f)) and Im(X(t, f)) respectively represent the real part and the imaginary part; |X(t, f)| represents the amplitude of the vibration signal at time t and frequency f.

[0095] S230. Perform feature splicing on the time-domain features and frequency-domain features to obtain the target fault type judgment features.

[0096] The time-domain features and frequency-domain features can be simply spliced in a front-back manner to obtain the target fault type judgment features, or the time-domain features and frequency-domain features of each sensor can be spliced to obtain the single-sensor splicing features, and then the single-sensor splicing features are spliced according to the preset sensor order to obtain the target fault type judgment features. This embodiment does not limit the splicing method.

[0097] S240. Input the target fault type judgment features into the fault type judgment model to obtain the fault type judgment result.

[0098] Among them, the fault type judgment model includes a support vector machine model corresponding to at least two fault types.

[0099] The support vector machine model will output the probability belonging to the target fault type corresponding to the support vector machine model and the probability belonging to other fault types outside the target fault type. Finally, according to the probabilities output by all support vector machine models, determine

[0100] Exemplarily, each support vector machine model, that is, each binary classifier, has a weighting coefficient w k , representing the importance of the classifier. Exemplarily, the decision function of the weighting coefficient of each classifier is as follows:

[0101]

[0102] where, w k represents the weight of the classifier corresponding to category k, and f k (x) represents the decision function of the corresponding classifier, representing the confidence in category k;

[0103] In multi-class prediction, the input sample x is evaluated through the decision function f k (x) of each classifier. Each classifier outputs a confidence value. Finally, the category is predicted through the weighted decision rule, and the classifier with the highest score is selected as the final predicted category. For example, the confidences of normal, fault A, and fault B are output, and the category with the highest score is selected as the fault type of the fault type judgment result.

[0104] The data processing process of the support vector machine realizes multi-classification through the one-vs-rest strategy, and the performance and robustness of the model are improved by means of a weighted classifier.

[0105] S250. According to the preset fault location analysis weight value, preset time-domain feature weight value, and preset frequency-domain feature weight value of each sensor, perform feature fusion on the time-domain feature and the frequency-domain feature to obtain a fused feature.

[0106] The time-domain features participating in the feature fusion can be amplitude, root mean square value, and peak factor, and the frequency-domain features can include the main frequency component and the spectral width.

[0107] Exemplarily, the calculation process of the fused feature is represented by the following formula:

[0108]

[0109] In the formula, S(t) represents the fused feature, ω m represents the weight of sensor m, that is, the preset fault location analysis weight value, α m represents the weight coefficient of the time-domain feature, that is, the preset time-domain feature weight value, β m represents the weight coefficient of the frequency-domain feature, that is, the preset frequency-domain feature weight value, A m represents the amplitude of the data collected by sensor m, RMS m represents the root mean square value of the data collected by sensor m; P m represents the peak factor of the data collected by sensor m, f peak,m represents the main frequency component of the data collected by sensor m; Δf m represents the spectral width of the data collected by sensor m. The preset fault location analysis weight value is determined according to the historical failure rate of the part of the switchgear corresponding to each sensor. The higher the historical failure rate, the higher the preset fault location analysis weight value.

[0110] By weighted fusion of the vibration signals collected by multiple sensors through time-domain features and frequency-domain features, a comprehensive fused feature is generated, which effectively reflects the working state of the equipment and provides support for subsequent fault point judgment.

[0111] S260. According to the time-domain feature, frequency-domain feature, fused feature, and fault type judgment result, perform feature splicing to obtain the target fault point judgment feature.

[0112] Splice the time-domain feature, frequency-domain feature, fused feature, and fault type judgment result in a preset order. For example, the preset order is time-domain feature - frequency-domain feature - fused feature - fault type judgment result.

[0113] S270. Input the target fault point judgment feature into the first fault point judgment model based on a deep neural network to obtain a first fault point judgment result.

[0114] S280. Input the target fault point judgment feature into the second fault point judgment model based on a gradient boosting decision tree to obtain a second fault point judgment result.

[0115] S290. Perform weighted fusion on the first fault point judgment result and the second fault point judgment result to obtain the target fault point judgment result of the switchgear.

[0116] In this embodiment, predictions are made through two different types of machine learning algorithms, and the final prediction result is obtained through weighted fusion. For example, the calculation process of the target fault point judgment result is represented by the following formula:

[0117]

[0118] In the formula, represents the first fault point judgment result, represents the second fault point judgment result.

[0119] The technical solution of this embodiment is to obtain the original vibration data collected by at least two sensors arranged on the switching device, where each sensor is arranged at a different part of the switching device; extract the time-domain features and frequency-domain features from the original vibration data to obtain time-domain features and frequency-domain features; splice the time-domain features and frequency-domain features to obtain target fault type judgment features; input the target fault type judgment features into the fault type judgment model to obtain the fault type judgment result; where the fault type judgment model includes support vector machine models corresponding to at least two fault types; according to the preset fault location analysis weight value, preset time-domain feature weight value, and preset frequency-domain feature weight value of each sensor, perform feature fusion on the time-domain features and frequency-domain features to obtain fusion features; where the preset fault location analysis weight value is determined according to the historical failure rate of the part of the switching device corresponding to each sensor; perform feature splicing according to the time-domain features, frequency-domain features, fusion features, and fault type judgment result to obtain target fault point judgment features; input the target fault point judgment features into the first fault point judgment model based on a deep neural network to obtain the first fault point judgment result; input the target fault point judgment features into the second fault point judgment model based on a gradient boosting decision tree to obtain the second fault point judgment result; perform weighted fusion on the first fault point judgment result and the second fault point judgment result to obtain the target fault point judgment result of the switching device. The technical solution of the embodiment of the present invention solves the problem that the fault point of the switching device cannot be accurately judged at present. Through comprehensive feature extraction and comprehensive analysis of the features of the signals collected by multiple sensors, accurate fault location can be achieved by machine learning using features, and the accuracy of fault location can be further improved through feature weighted fusion, ensuring the operation stability of the switching device.

[0120] Figure 3 FIG. is a flowchart of a switching device fault location method provided by an embodiment of the present invention. This embodiment and the switching device fault location method in the above embodiment belong to the same inventive concept, and further describe the process of training the fault type judgment model. This method can be executed by a switching device fault location device, and this device can be implemented in a software and / or hardware manner and integrated in a computer device with application development functions.

[0121] As Figure 3 shown, the training process of the fault type judgment model in the switching device fault location method of this embodiment includes the following steps:

[0122] S310. Determine the initial support vector machine model corresponding to at least two fault types.

[0123] The fault types at least include a normal type and at least one fault type. Regarding the classification of fault types as a multi-classification problem, select the one-versus-all strategy to perform binary classification training on the initial support vector machine model.

[0124] S320. For each initial support vector machine model, perform binary classification training on the initial support vector machine model according to the training samples, the target fault type and the non-target fault type of the initial support vector machine model, to obtain a support vector machine model corresponding to at least two fault types.

[0125] For each initial support vector machine model, use the samples of the target fault type of the initial support vector machine model as positive examples, and the samples of all other fault types, that is, the non-target fault types, as negative examples, and train to obtain support vector machine models with the same number as the number of fault types.

[0126] S330. Use the support vector machine models corresponding to at least two fault types as fault type judgment models.

[0127] In an optional implementation manner, performing binary classification training on the initial support vector machine model according to the training samples, the target fault type and the non-target fault type of the initial support vector machine model includes:

[0128] Determine the class label of each training sample of the initial support vector machine model, where the class label includes a target fault type label and a non-target fault type label; determine the initial hyperplane of the initial support vector machine model; calculate the function value of the adaptive function of the training sample according to the training sample and the class label; calculate the distance from the training sample to the hyperplane, determine the support vectors in the training samples according to the distance and a preset distance threshold, and update the support vectors through importance measurement to obtain a support vector update result; update the hyperplane parameters of the initial hyperplane according to the support vector update result to obtain a hyperplane update result; determine the loss function according to the hyperplane update result, the function value of the adaptive function, and the distance from the training sample to the hyperplane, and complete the binary classification training when the loss function converges.

[0129] Target fault type label and non-target fault type label

[0130] For class k, construct a binary classification problem, where the class labels can be +1 and -1. Mark the target fault type samples as +1 and the non-target fault types as -1. For each training sample x i , its label y i Perform conversion according to class k:

[0131]

[0132] Determine the initial hyperplane of the initial support vector machine model, and initialize the hyperplane parameters of the hyperplane to obtain an initialized hyperplane. The hyperplane parameters include a weight vector and a bias term. In each binary classification task, the support vector machine determines the hyperplane by solving the following optimization problem:

[0133]

[0134] where w k represents the weight vector of the hyperplane; b k represents the bias term of the hyperplane; ξ k,i represents the slack variable, indicating the classification error of the i-th sample; C k represents the regularization parameter.

[0135] For each sample i, the constraint condition is:

[0136]

[0137] where: represents the transformed class label;

[0138] Calculate the function value of the adaptive function of the training samples according to the training samples and class labels, and the kernel function adopts an adaptive kernel function:

[0139]

[0140] where: x and x' are two samples; σ(x) and σ(x') represent the local feature scales of the corresponding samples; adjusted according to the fluctuations in the frequency domain;

[0141] During the training process of the support vector machine model, it is required that the distances from all sample points to the hyperplane are as large as possible, while satisfying the constraint conditions of correct classification. Support vectors are those sample points closest to the hyperplane. Calculate the distance from the training samples to the hyperplane, and determine the training samples less than or equal to the preset distance threshold in the training samples as support vectors, and introduce the importance measure of the support vectors to evaluate the contribution of each support vector to the model decision boundary:

[0142]

[0143] where: ∈ represents a positive number to prevent the denominator from being zero; d i represents the distance from the support vector v i to the hyperplane; ω i represents the importance measure of the support vector;

[0144] According to the importance ω i of the support vector, dynamically update the support vector, and update the hyperplane parameters of the initial hyperplane according to the updated result of the support vector to obtain the updated result of the hyperplane.

[0145] The loss function of the support vector machine model is as follows:

[0146]

[0147] Wherein, K represents the number of categories; N k represents the number of samples in category k; i represents the sample index; x i represents the feature of the i-th sample; y i,k represents the label of sample x i in category k; w k represents the weight vector of category k; b k represents the bias term of category k; K(x i , x) represents the value of the kernel function; d i,k represents the distance from sample x i to the hyperplane of category k; ε represents a constant to prevent the denominator from being zero; λ represents the regularization parameter.

[0148] In this embodiment, multi-classification is achieved through the one-vs-rest strategy, and the performance and robustness of the model are improved by means of the dynamic support vector update mechanism. The adaptive kernel function design and the importance measure of support vectors enable the model to better adapt to the non-linearity and complexity in the data, especially suitable for processing complex data types such as vibration signals. The loss function integrates the weighted mechanism, the importance measure of support vectors, and the adaptive kernel function, and can process non-linear data in multi-classification tasks, dynamically adjust the importance of support vectors, and improve the accuracy and robustness of the model through weighted decision-making.

[0149] The technical solution of this embodiment is to determine the initial support vector machine models corresponding to at least two fault types; for each initial support vector machine model, perform binary classification training on the initial support vector machine model according to the training samples, the target fault type and the non-target fault type of the initial support vector machine model, to obtain the support vector machine models corresponding to at least two fault types; and use the support vector machine models corresponding to at least two fault types as the fault type judgment model. The technical solution of the embodiment of the present invention solves the problem that the fault point of the switchgear cannot be accurately judged at present. Through comprehensive feature extraction and comprehensive analysis of the signal features collected by multiple sensors, the fault can be accurately located by machine learning using the features, improving the accuracy of fault location, ensuring the operation stability of the switchgear, and improving the accuracy of fault type determination through the support vector machine model with the one-vs-rest strategy.

[0150] Figure 4 FIG. is a schematic structural diagram of a switchgear fault location device provided by an embodiment of the present invention. This embodiment is applicable to the scenario of switchgear fault location. The switchgear fault location device can be implemented in a software and / or hardware manner and integrated into a computer terminal device with application development functions.

[0151] Such as Figure 4As shown in the figure, the switchgear fault location device includes: a vibration data acquisition module 410, a feature extraction module 420, a fault type judgment feature determination module 430, a fault type judgment result determination module 440, a fusion feature determination module 450, a target fault point judgment feature determination module 460, a first fault point judgment result determination module 470, a second fault point judgment result determination module 480, and a target fault point judgment result determination module 490.

[0152] Among them, the vibration data acquisition module 410 is used to acquire the original vibration data collected by at least two sensors arranged on the switchgear, where each sensor is arranged at different parts of the switchgear; the feature extraction module 420 is used to perform time-domain feature extraction and frequency-domain feature extraction on the original vibration data to obtain time-domain features and frequency-domain features; the fault type judgment feature determination module 430 is used to splice the time-domain features and frequency-domain features to obtain target fault type judgment features; the fault type judgment result determination module 440 is used to input the target fault type judgment features into the fault type judgment model to obtain the fault type judgment result, where the fault type judgment model includes support vector machine models corresponding to at least two fault types; the fusion feature determination module 450 is used to perform feature fusion on the time-domain features and frequency-domain features according to the preset fault location analysis weight values of each sensor to obtain fusion features, where the preset fault location analysis weight values are determined according to the historical failure rates of the parts of the switchgear corresponding to each sensor; the target fault point judgment feature determination module 460 is used to splice the time-domain features, frequency-domain features, fusion features, and fault type judgment results to obtain target fault point judgment features; the first fault point judgment result determination module 470 is used to input the target fault point judgment features into the first fault point judgment model based on a deep neural network to obtain the first fault point judgment result; the second fault point judgment result determination module 480 is used to input the target fault point judgment features into the second fault point judgment model based on a gradient boosting decision tree to obtain the second fault point judgment result; the target fault point judgment result determination module 490 is used to perform weighted fusion on the first fault point judgment result and the second fault point judgment result to obtain the target fault point judgment result of the switchgear.

[0153] The technical solution of this embodiment obtains the original vibration data collected by at least two sensors arranged on the switching device, where each sensor is arranged at a different part of the switching device; performs time-domain feature extraction and frequency-domain feature extraction on the original vibration data to obtain time-domain features and frequency-domain features; splices the time-domain features and frequency-domain features to obtain target fault type judgment features; inputs the target fault type judgment features into a fault type judgment model to obtain a fault type judgment result, where the fault type judgment model includes support vector machine models corresponding to at least two fault types; performs feature fusion on the time-domain features and frequency-domain features according to the preset fault location analysis weight values of each sensor, where the preset fault location analysis weight values are determined according to the historical failure rates of the parts of the switching device corresponding to each sensor; splices the time-domain features, frequency-domain features, fusion features, and fault type judgment results to obtain target fault point judgment features; inputs the target fault point judgment features into a first fault point judgment model based on a deep neural network to obtain a first fault point judgment result; inputs the target fault point judgment features into a second fault point judgment model based on a gradient boosting decision tree to obtain a second fault point judgment result; performs weighted fusion on the first fault point judgment result and the second fault point judgment result to obtain the target fault point judgment result of the switching device. The technical solution of the embodiment of the present invention solves the problem that the fault point of the switching device cannot be accurately judged at present, and can accurately perform fault location by using features through comprehensive feature extraction and comprehensive analysis of the signal features collected by multiple sensors, improving the accuracy of fault location and ensuring the operation stability of the switching device.

[0154] In an alternative embodiment, the time-domain features include amplitude, root mean square value, and peak factor, and the feature extraction module 420 is specifically configured to:

[0155] Perform smoothing processing on the original vibration data to obtain smoothed vibration data; calculate the amplitude of the original vibration data according to the maximum value and minimum value of the vibration signal in the smoothed vibration data; calculate the root mean square value of the original vibration data according to the vibration signal and the number of signal sampling points in the smoothed vibration data; calculate the peak factor of the original vibration data according to the maximum value of the vibration signal and the root mean square value.

[0156] In an alternative embodiment, the frequency-domain features include main frequency components, spectral width, and peak frequency, and the feature extraction module 420 is further configured to:

[0157] Denoise the original vibration data to obtain denoised vibration data; extract the time-frequency features of the denoised vibration data through short-time Fourier transform: extract the main frequency components of the denoised vibration data according to the time-frequency features to obtain the main frequency components of the original vibration data; calculate the average frequency according to the frequency of the denoised vibration data, and calculate the spectral width of the original vibration data according to the frequency, average frequency and time-frequency features; determine the frequency point with the largest amplitude in the frequency according to the time-frequency features, and use the frequency point with the largest amplitude as the peak frequency.

[0158] In an alternative embodiment, the device further includes:

[0159] A fault type judgment model training module, configured to determine an initial support vector machine model corresponding to at least two fault types; for each initial support vector machine model, perform binary classification training on the initial support vector machine model according to the training samples, the target fault type and non-target fault type of the initial support vector machine model, to obtain a support vector machine model corresponding to at least two fault types; use the support vector machine models corresponding to at least two fault types as the fault type judgment model.

[0160] In an alternative embodiment, the fault type judgment model training module is further configured to:

[0161] Determine the class label of each training sample of the initial support vector machine model, where the class label includes a target fault type label and a non-target fault type label; determine the initial hyperplane of the initial support vector machine model; calculate the function value of the adaptive function of the training sample according to the training sample and the class label; calculate the distance from the training sample to the hyperplane, determine the support vectors in the training samples according to the distance and a preset distance threshold, and update the support vectors through importance measurement to obtain a support vector update result; update the hyperplane parameters of the initial hyperplane according to the support vector update result to obtain a hyperplane update result; determine the loss function according to the hyperplane update result, the function value of the adaptive function, and the distance from the training sample to the hyperplane, and complete the binary classification training when the loss function converges.

[0162] In an alternative embodiment, the fusion feature determination module 450 is specifically configured to:

[0163] Perform feature fusion on the time-domain features and frequency-domain features according to the preset fault location analysis weight value, preset time-domain feature weight value, and preset frequency-domain feature weight value of each sensor to obtain fusion features.

[0164] The switchgear fault location device provided by the embodiments of the present invention can execute the switchgear fault location method provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method.

[0165] Figure 5 A schematic structural diagram of a computer device provided by an embodiment of the present invention. Figure 5 A block diagram of an exemplary computer device 12 suitable for implementing the embodiments of the present invention is shown. Figure 5 The computer device 12 shown is merely an example and should not impose any limitation on the functions and usage scope of the embodiments of the present invention. The computer device 12 can be any terminal device with computing capabilities, such as intelligent controllers, servers, mobile phones and other terminal devices.

[0166] As Figure 5 shown, the computer device 12 is presented in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0167] The bus 18 represents one or more of several types of bus structures, including a memory bus or a memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus structures. By way of example, these architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

[0168] The computer device 12 typically includes a variety of computer system-readable media. These media can be any available media accessible by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0169] The system memory 28 may include computer system-readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. Merely by way of example, a storage system 34 can be used to read and write non-removable, non-volatile magnetic media ( Figure 5 not shown, commonly referred to as a "hard disk drive"). Although Figure 5Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., CD-ROM, DVD-ROM or other optical media) can be provided. In these cases, each drive can be connected to the bus 18 through one or more data medium interfaces. The system memory 28 can include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0170] A program / utility 40 having a set (at least one) of program modules 42 can be stored, for example, in the system memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include the implementation of a network environment. The program modules 42 generally perform the functions and / or methods in the embodiments described in the present invention.

[0171] The computer device 12 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and can also communicate with one or more devices that enable a user to interact with the computer device 12, and / or communicate with any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication can be carried out through the input / output (I / O) interface 22. Also, the computer device 12 can communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public network, such as the Internet) through the network adapter 20. As shown in the figure, the network adapter 20 communicates with other modules of the computer device 12 through the bus 18. It should be understood that although Figure 5 not shown, other hardware and / or software modules can be used in combination with the computer device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0172] The processing unit 16 executes various functional applications and data processing by running the programs stored in the system memory 28, for example, implementing the switch device fault location method provided by the embodiments of the present invention. The method includes:

[0173] Obtain the original vibration data collected by at least two sensors arranged on the switch device, where each sensor is arranged at a different part of the switch device;

[0174] Perform time-domain feature extraction and frequency-domain feature extraction on the original vibration data to obtain time-domain features and frequency-domain features;

[0175] Perform feature concatenation on the time-domain features and frequency-domain features to obtain the target fault type judgment features;

[0176] Input the target fault type judgment features into the fault type judgment model to obtain the fault type judgment result, where the fault type judgment model includes support vector machine models corresponding to at least two fault types;

[0177] Perform feature fusion on the time-domain features and frequency-domain features according to the preset fault location analysis weight values of each sensor to obtain the fusion features, where the preset fault location analysis weight values are determined according to the historical failure rates of the parts of the switching device corresponding to each sensor;

[0178] Perform feature concatenation on the time-domain features, frequency-domain features, fusion features and fault type judgment results to obtain the target fault point judgment features;

[0179] Input the target fault point judgment features into the first fault point judgment model based on a deep neural network to obtain the first fault point judgment result;

[0180] Input the target fault point judgment features into the second fault point judgment model based on a gradient boosting decision tree to obtain the second fault point judgment result;

[0181] Perform weighted fusion on the first fault point judgment result and the second fault point judgment result to obtain the target fault point judgment result of the switching device.

[0182] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the switching device fault location method provided in any embodiment of the present invention. The method includes:

[0183] Obtain the original vibration data collected by at least two sensors arranged on the switching device, where each sensor is arranged at a different part of the switching device;

[0184] Perform time-domain feature extraction and frequency-domain feature extraction on the original vibration data to obtain time-domain features and frequency-domain features;

[0185] Perform feature concatenation on the time-domain features and frequency-domain features to obtain the target fault type judgment features;

[0186] Input the target fault type judgment features into the fault type judgment model to obtain the fault type judgment result, where the fault type judgment model includes support vector machine models corresponding to at least two fault types;

[0187] According to the preset fault location analysis weight value of each sensor, perform feature fusion on the time-domain features and frequency-domain features to obtain fused features, where the preset fault location analysis weight value is determined according to the historical failure rate of the part of the switching device corresponding to each sensor;

[0188] Perform feature splicing based on the time-domain features, frequency-domain features, fused features, and fault type judgment results to obtain the target fault point judgment features;

[0189] Input the target fault point judgment features into the first fault point judgment model based on a deep neural network to obtain the first fault point judgment result;

[0190] Input the target fault point judgment features into the second fault point judgment model based on a gradient boosting decision tree to obtain the second fault point judgment result;

[0191] Perform weighted fusion on the first fault point judgment result and the second fault point judgment result to obtain the target fault point judgment result of the switching device.

[0192] The computer storage medium of the embodiments of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to: an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive list) of the computer-readable storage medium include: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, the computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, apparatus, or device.

[0193] The computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries the computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium can also be any computer-readable medium other than the computer-readable storage medium, and this computer-readable medium can send, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device.

[0194] The program code contained on a computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0195] The computer program code for performing the operations of the present invention can be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, Python, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0196] An embodiment of the present invention also provides a computer program product, including a computer program which, when executed by a processor, implements the switch device fault location method provided in any embodiment of the present application.

[0197] In the process of implementing the computer program product, the computer program code for performing the operations of the present invention can be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, Python, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network - including a local area network (LAN) or a wide area network (WAN) - or can be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0198] Those of ordinary skill in the art should understand that the various modules or steps of the present invention described above can be implemented using a general-purpose computing device. They can be concentrated on a single computing device or distributed across a network composed of multiple computing devices. Optionally, they can be implemented using program code executable by a computer device, so that they can be stored in a storage device and executed by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0199] Note that the above is only the preferred embodiment of the present invention and the technical principles applied. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein. Various obvious changes, re-adjustments, and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments. Without departing from the concept of the present invention, more other equivalent embodiments can be included, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for locating a switchgear fault, characterized in that: include: Acquire original vibration data collected by at least two sensors arranged on the switch device, wherein each of the sensors is arranged at a different position of the switch device; Performing time domain feature extraction and frequency domain feature extraction on the original vibration data to obtain time domain features and frequency domain features; The time domain features and the frequency domain features are combined to obtain target fault type judgment features; Inputting the target fault type judgment feature into a fault type judgment model to obtain a fault type judgment result, wherein the fault type judgment model includes a support vector machine model corresponding to at least two fault types; According to a preset fault location analysis weight value of each of the sensors, the time domain features and the frequency domain features are subjected to feature fusion to obtain fusion features, wherein the preset fault location analysis weight value is determined according to a historical failure rate of a part of the switch device corresponding to each of the sensors; Perform feature splicing according to the time domain feature, the frequency domain feature, the fusion feature and the fault type judgment result to obtain a target fault point judgment feature; Inputting the target fault point judgment feature into a first fault point judgment model based on a deep neural network to obtain a first fault point judgment result; Inputting the target fault point judgment feature into a second fault point judgment model based on a gradient enhancement decision tree to obtain a second fault point judgment result; The first fault point judgment result and the second fault point judgment result are weightedly fused to obtain a target fault point judgment result of the switch device.

2. The method according to claim 1, characterized in that: The time domain features include amplitude, root mean square value and peak factor. The time domain features are extracted from the original vibration data to obtain the time domain features, including: Smoothing the original vibration data to obtain smoothed vibration data; Calculating the amplitude of the original vibration data according to the maximum value and the minimum value of the vibration signal in the smoothed vibration data; Calculating the root mean square value of the original vibration data according to the vibration signal and the number of signal sampling points in the smoothed vibration data; The peak factor of the original vibration data is calculated according to the maximum value of the vibration signal and the root mean square value.

3. The method according to claim 1, characterized in that The frequency domain features include main frequency components, spectrum width and peak frequency. The frequency domain features are extracted from the original vibration data to obtain frequency domain features, including: Performing denoising processing on the original vibration data to obtain denoised vibration data; The time-frequency characteristics of the denoised vibration data are extracted by short-time Fourier transform: Extracting the main frequency components of the denoised vibration data according to the time-frequency characteristics to obtain the main frequency components of the original vibration data; Calculate an average frequency according to the frequency of the denoised vibration data, and calculate a frequency spectrum width of the original vibration data according to the frequency, the average frequency and the time-frequency feature; According to the time-frequency characteristics, a frequency point with the largest amplitude in the frequencies is determined, and the frequency point with the largest amplitude is taken as the peak frequency.

4. The method according to claim 1, characterized in that: The training process of the fault type judgment model includes: determining an initial support vector machine model corresponding to at least two fault types; For each of the initial support vector machine models, performing binary classification training on the initial support vector machine model according to the training samples, the target fault type and the non-target fault type of the initial support vector machine model, to obtain a support vector machine model corresponding to at least two fault types; The support vector machine model corresponding to at least two fault types is used as the fault type judgment model.

5. The method according to claim 4, characterized in that The binary classification training of the initial model of the support vector machine is performed according to the training samples, the target fault type and the non-target fault type of the initial model of the support vector machine, including: Determining a category label for each training sample of the support vector machine initial model, wherein the category label includes a target fault type label and a non-target fault type label; Determining an initial hyperplane of the support vector machine initial model; Calculate the function value of the adaptive function of the training sample according to the training sample and the category label; Calculating the distance from the training sample to the hyperplane, determining a support vector in the training sample according to the distance and a preset distance threshold, and updating the support vector by an importance metric to obtain a support vector update result; Update the hyperplane parameters of the initial hyperplane according to the support vector update result to obtain a hyperplane update result; A loss function is determined according to the hyperplane update result, the function value of the adaptive function and the distance from the training sample to the hyperplane, and the binary classification training is completed when the loss function converges.

6. The method according to claim 1, characterized in that The step of fusing the time domain features and the frequency domain features according to the preset fault location analysis weight value of each sensor to obtain the fused features includes: According to the preset fault location analysis weight value, the preset time domain feature weight value and the preset frequency domain feature weight value of each sensor, the time domain feature and the frequency domain feature are fused to obtain a fused feature.

7. A switchgear fault location device, characterized in that: include: A vibration data acquisition module, used to acquire original vibration data collected by at least two sensors arranged on the switch device, wherein each of the sensors is arranged at a different part of the switch device; A feature extraction module, used to extract time domain features and frequency domain features from the original vibration data to obtain time domain features and frequency domain features; A fault type judgment feature determination module is used to perform feature splicing on the time domain features and the frequency domain features to obtain target fault type judgment features; A fault type judgment result determination module, used for inputting the target fault type judgment feature into a fault type judgment model to obtain a fault type judgment result, wherein the fault type judgment model includes a support vector machine model corresponding to at least two fault types; A fusion feature determination module, used to perform feature fusion on the time domain feature and the frequency domain feature according to a preset fault location analysis weight value of each sensor to obtain a fusion feature, wherein the preset fault location analysis weight value is determined according to a historical failure rate of a part of the switch device corresponding to each sensor; A target fault point judgment feature determination module is used to perform feature splicing according to the time domain feature, the frequency domain feature, the fusion feature and the fault type judgment result to obtain the target fault point judgment feature; A first fault point judgment result determination module, used for inputting the target fault point judgment feature into a first fault point judgment model based on a deep neural network to obtain a first fault point judgment result; A second fault point judgment result determination module, used for inputting the target fault point judgment feature into a second fault point judgment model based on a gradient enhancement decision tree to obtain a second fault point judgment result; The target fault point judgment result determination module is used to perform weighted fusion on the first fault point judgment result and the second fault point judgment result to obtain the target fault point judgment result of the switch device.

8. A computer device, characterized in that: The computer device comprises: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the switch device fault locating method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the switching device fault location method as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for locating a fault of a switching device as claimed in any one of claims 1 to 6 is implemented.

Images