Circuit breaker fault detection optimization method and system based on SVD-XGBoost

Abstract

The invention discloses a circuit breaker fault detection optimization method and system based on SVD-XGBoost. The method comprises the steps of obtaining an original vibration signal generated when a circuit breaker is in an operation state; performing data enhancement on the original vibration signal based on an NLMS algorithm so as to adaptively filter noise, and obtaining a time domain signal after data enhancement; constructing an SVD-XGBoost model based on combination of SVD (singular value decomposition) and an XGBoost algorithm, converting an enhanced time domain signal into a frequency domain signal through fast Fourier transform, performing data dimension reduction on the frequency domain signal by adopting singular value decomposition, and extracting target features; according to the method, the extracted target features are input to the trained SVD-XGBoost model, the optimized network is adopted to carry out fault classification on the circuit breaker vibration signals, a circuit breaker fault identification result is obtained, SVD dimension reduction is introduced to convert high-dimensional features into low-dimensional features, the number of the features is reduced, the calculation efficiency is improved, and the fault classification precision and effect are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical equipment maintenance, and in particular to a circuit breaker fault detection optimization method and system based on SVD-XGBoost. Background Art

[0002] High-voltage circuit breakers are essential components in the main electrical wiring of power systems, and play a key role in maintaining the stability, reliability and economy of power systems. They have two main functions: one is to separate and close normal currents to cope with changes in grid load and ensure the normal operation of the circuit; the other is to cut off fault currents to isolate fault areas and protect the safe operation of power systems. When high-voltage circuit breakers lose reliability or mechanical parts are damaged, two main problems will arise. On the one hand, it may cause instability in the power system or power equipment. On the other hand, it may cause the relay protection to fail to trigger accurately and lose the ability to respond quickly. This will trigger the operation of the upper level protection, expand the scope of the fault, lose selectivity, and in severe cases may cause the entire system to be disconnected. Early fault diagnosis of high-voltage circuit breakers mainly uses offline detection methods, including regular maintenance and preventive maintenance. Offline detection requires the measurement of various parameters of high-voltage circuit breakers, such as the stroke, movement time, movement speed and surface resistance of the electromagnet contacts, the current resistance of the closing and opening coil circuit, and the sound, vibration signals generated during operation, and the torsional deformation of mechanical parts. However, offline detection requires power outages of the power system, which causes unnecessary economic losses and has a certain impact on people's lives. Therefore, a new maintenance method is urgently needed to replace offline detection methods.

[0003] At present, the research on fault diagnosis of mechanical components of high-voltage circuit breakers involves a variety of detection signals, including coil current and voltage signals, contact travel and speed signals, energy storage motor current signals, mechanical vibration and sound signals, etc. Traditional signal processing methods conduct time domain and frequency domain analysis on the original signal and extract time-frequency domain information. Although some results have been achieved, there are certain shortcomings. At the same time, the processing process of traditional methods is relatively complicated and the time complexity is high, which increases the relevant application cost and industrialization difficulty. In summary, when dealing with the high-dimensional problem of circuit breaker sound data, it is easy to lose information and the accuracy of frequency domain analysis is limited by the frequency resolution. The processing process is relatively complicated and the time complexity is high, which increases the application cost and industrialization difficulty of related technologies. Summary of the invention

[0004] The purpose of the present invention is to provide a circuit breaker fault detection optimization method, device and equipment based on SVD-XGBoost. Aiming at the high-dimensional problem of circuit breaker sound data, it is proposed to use NLMS technology to perform data enhancement on the original vibration signal, and then use fast Fourier transform to convert the enhanced signal into a time-frequency spectrum graph, use singular value decomposition to complete data dimensionality reduction on the time-frequency spectrum graph, and select the integrated algorithm SVD-XGBoost as the classification algorithm to realize fault diagnosis.

[0005] The present invention provides a circuit breaker fault detection optimization method based on SVD-XGBoost, comprising:

[0006] Acquire the original vibration signal generated when the circuit breaker is in operation;

[0007] Based on the NLMS algorithm, data enhancement is performed on the original vibration signal to adaptively filter out noise, thereby obtaining a time domain signal after data enhancement;

[0008] An SVD-XGBoost model is constructed based on the combination of SVD singular value decomposition and XGBoost algorithm. The enhanced time domain signal is converted into a frequency domain signal through fast Fourier transform. The frequency domain signal is subjected to data dimension reduction by singular value decomposition and the target features are extracted.

[0009] The extracted target features are input into the trained SVD-XGBoost model, and the optimized network is used to classify the circuit breaker vibration signal to obtain the circuit breaker fault identification result.

[0010] Preferably, the data enhancement of the original vibration signal based on the NLMS algorithm to adaptively filter out noise to obtain a time domain signal after data enhancement includes:

[0011] The original vibration signal x(n) is represented as a time series as input, where n represents the nth sample point;

[0012] Initialize the NLMS algorithm parameters, including the weight vector and step size factor;

[0013] Perform preliminary noise suppression on each sample point n and estimate the signal

[0014] y(n)=w T (n)*x(n);

[0015] The residual noise component is estimated by calculating the error between the original vibration signal x(n) and the signal y(n) output by the preliminary noise suppression:

[0016] e(n)=x(n)-y(n);

[0017] Update according to the adaptive weight formula:

[0018]

[0019] Calculate the output enhancement signal:

[0020]

[0021] Iterate and update until all sample points are traversed or the error converges to the preset error threshold;

[0022] Where e(n) is the error signal, μ is the step size factor, and w T (n) is the transpose of the weight vector, w(n) is the weight vector, w(n+1) is the updated weight vector, ||x(n)|| 2 is the square of the modulus of the input signal vector, ∈ is a small positive number.

[0023] Preferably, converting the enhanced time domain signal into a frequency domain signal by fast Fourier transform comprises:

[0024] Input signal after NLMS enhancement Perform a discrete Fourier transform:

[0025]

[0026] Among them, X(k) represents the frequency domain signal, the kth frequency component; N represents the total number of sampling points of the signal; e represents a natural constant; and i represents an imaginary unit.

[0027] Preferably, the adopting singular value decomposition to perform data dimension reduction on the frequency domain signal and extract target features comprises:

[0028] Constructing the target matrix A based on the frequency domain signal after discrete Fourier transformation can perform SVD singular value decomposition:

[0029] A=U∑V T ;

[0030] Among them, A is the original matrix, representing the frequency domain signal data; U is the left singular matrix, an orthogonal matrix; Σ is a diagonal matrix, and the diagonal elements are singular values; V is the right singular matrix, an orthogonal matrix, and the transpose is VT;

[0031] Select the first m largest singular values ​​and the corresponding right singular vectors, and project the high-dimensional data into a low-dimensional space consisting of m key features for dimensionality reduction in the following way:

[0032]

[0033] Where Σm represents the diagonal matrix of the first m selected singular values; The columns of the right singular matrix corresponding to the first m singular values, z is the target feature after dimensionality reduction.

[0034] Preferably, the extracted target features are input into the trained SVD-XGBoost model, and the optimized network is used to classify the circuit breaker vibration signal to obtain the circuit breaker fault identification result, which includes:

[0035] The target features obtained through SVD dimensionality reduction are used as the input of the XGBoost model;

[0036] Construct the objective function based on the loss function and regularization term:

[0037]

[0038] Among them, Obj(θ) is the objective function, is the loss function, which is used to measure the difference between the prediction and the actual result; Ω(fk) is the regularization term that controls the complexity of the model, N represents the number of samples, and K is the number of decision trees included in the model;

[0039] Training classification: Build a decision tree group through a weighted voting mechanism and calculate the gain value of each branch to determine the optimal splitting strategy:

[0040]

[0041] Among them, G j represents the gradient sum of all samples belonging to the jth node, I j represents the sample set of the jth node, H j represents the sum of the second-order derivatives of all samples belonging to the jth node, g i is the first-order gradient of the ith sample, h i is the second-order gradient of the i-th sample;

[0042] The score of the split point is calculated based on the gradient sum and the second-order derivative sum, so as to adjust and select the optimal split point:

[0043]

[0044] Among them, score represents the score of the optimized split point, λ is the regularization parameter, and H all represents the gradient and second-order derivative on the entire data set, and γ is the penalty parameter for node partitioning.

[0045] Preferably, the obtaining of the original vibration signal generated when the circuit breaker is in operation includes:

[0046] By adjusting the degree of jamming of the iron core on the circuit breaker to simulate different fault conditions, or by setting different degrees of resistance to simulate different degrees of fault, data samples of different mechanical fault types of the circuit breaker are obtained, wherein the mechanical fault type of the circuit breaker includes different degrees of fault signals simulated by loose screws on the base of the circuit breaker, jamming of the iron core and insufficient lubrication.

[0047] The present invention also provides a circuit breaker fault detection optimization system based on SVD-XGBoost, comprising:

[0048] A data acquisition module, used to obtain the original vibration signal generated when the circuit breaker is in operation;

[0049] A data enhancement module, used for performing data enhancement on the original vibration signal based on the NLMS algorithm to adaptively filter out noise, thereby obtaining a time domain signal after data enhancement;

[0050] SVD-XGBoost model building module, used to build an SVD-XGBoost model based on SVD singular value decomposition and XGBoost algorithm, convert the enhanced time domain signal into a frequency domain signal through fast Fourier transform, use singular value decomposition to reduce the data dimension of the frequency domain signal and extract target features;

[0051] The fault identification module is used to input the extracted target features into the trained SVD-XGBoost model, use the optimized network to classify the circuit breaker vibration signal, and obtain the circuit breaker fault identification result.

[0052] Preferably, the fault identification module includes:

[0053] The target establishment unit is used to use the target features obtained through SVD dimensionality reduction as the input of the XGBoost model and construct the target function based on the loss function and regularization term:

[0054]

[0055] Among them, Obj(θ) is the objective function, is the loss function, which is used to measure the difference between the prediction and the actual result; Ω(fk) is the regularization term that controls the complexity of the model, N represents the number of samples, and K is the number of decision trees included in the model;

[0056] Training classification units is used to build decision tree groups through a weighted voting mechanism, and the gain value of each branch is calculated to determine the optimal splitting strategy:

[0057]

[0058] Among them, G jrepresents the gradient sum of all samples belonging to the jth node, I j represents the sample set of the jth node, H j represents the sum of the second-order derivatives of all samples belonging to the jth node, g i is the first-order gradient of the ith sample, h i is the second-order gradient of the i-th sample;

[0059] The scoring optimization unit is used to calculate the score of the split point based on the gradient sum and the second-order derivative sum, so as to adjust and select the optimized split point:

[0060]

[0061] Among them, score represents the score of the optimized split point, λ is the regularization parameter, and H all represents the gradient and second-order derivative on the entire data set, and γ is the penalty parameter for node partitioning.

[0062] The present invention also provides an electronic device, comprising:

[0063] A memory, the memory being used to store a processing program;

[0064] A processor, wherein when executing the processing program, the processor implements the circuit breaker fault detection optimization method based on SVD-XGBoost as described in the embodiment of the present invention.

[0065] The present invention also provides a readable storage medium, characterized in that a processing program is stored on the readable storage medium, and when the processing program is executed by a processor, the circuit breaker fault detection optimization method based on SVD-XGBoost as described in an embodiment of the present invention is implemented.

[0066] With respect to the prior art, the present invention has the following beneficial effects:

[0067] The present invention proposes a circuit breaker fault detection optimization method based on SVD-XGBoost. Firstly, the NLMS algorithm is used to perform data enhancement on the original vibration signal. Then, the enhanced signal is converted into a time-frequency spectrum using the fast Fourier transform. Then, the singular value decomposition is used to complete the data dimension reduction of the time-frequency spectrum. The integrated algorithm SVD-XGBoost is selected as the classification algorithm to finally realize fault diagnosis. Finally, the effectiveness and practicality of the SVD-XGBoost model denoising method are verified through experimental research and comparative analysis.

[0068] The present invention integrates multiple technologies such as NLMS data processing, SVD dimension reduction and XGBOOST fault diagnosis to form a comprehensive circuit breaker vibration signal processing system. The innovation of the present invention is based on SVD dimension reduction and XGBoost classification and recognition. Traditional fault classification methods usually directly use the original high-dimensional features for modeling and classification, which will face the problems of high data dimension and high computational complexity. The SVD-XGBoost model adopted by the present invention converts high-dimensional features into low-dimensional features by introducing SVD dimension reduction technology, reduces the number of features, improves computational efficiency, and can retain the main data features, reducing the risk of information loss.

[0069] The SVD-XGBoost model used in the present invention takes the low-dimensional features reduced by SVD as input, and then uses the XGBoost model for fault classification and identification. XGBoost is a powerful gradient boosting tree model with excellent nonlinear modeling and high-accuracy classification capabilities. By combining SVD and XGBoost, deep learning and nonlinear modeling of the reduced-dimensional features are achieved, improving the accuracy and effect of fault classification.

[0070] The SVD-XGBoost model used in the present invention can also be applied to multi-class fault classification problems, that is, to simultaneously identify and classify multiple different types of faults. Multi-class fault classification problems are challenging, but the XGBoost model has good performance when dealing with multi-class problems and can be optimized by adjusting parameters. SVD-XGBoost combines the advantages of SVD dimensionality reduction and XGBoost models to achieve efficient classification and identification of multi-class faults.

[0071] The SVD-XGBoost model adopted in the present invention is suitable for fault diagnosis and classification tasks in real-time environments. By using the efficient XGBoost model and the feature representation of SVD dimensionality reduction, it can process and analyze the fault data stream in real time, and quickly and accurately identify and classify faults, which is of great significance for real-time fault diagnosis and prediction, and measures can be taken in time to avoid potential power system failures and damages. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 A schematic diagram of the steps of a circuit breaker fault detection optimization method based on SVD-XGBoost in one embodiment of the present invention;

[0073] Figure 2 It is a principle block diagram of SVD-XGBoost in one embodiment of the present invention;

[0074] Figure 3 This is a normal state spectrum diagram of the original vibration signal in one embodiment of the present invention;

[0075] Figure 4 This is a spectrum diagram of the stuck state of the original vibration signal in one embodiment of the present invention;

[0076] Figure 5 This is a spectrum diagram of the screw loosening state of the original vibration signal in one embodiment of the present invention;

[0077] Figure 6 It is a spectrum diagram of the lubrication shortage state of the original vibration signal in one embodiment of the present invention;

[0078] Figure 7 This is an example diagram of a confusion matrix used in an embodiment of the present invention to quantitatively analyze diagnosis results. DETAILED DESCRIPTION

[0079] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0080] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0081] It should be noted that the concepts such as "once" and "second" mentioned in the present application are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0082] It should be noted that the modifications of "one" and "plurality" mentioned in the disclosure of this application are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0083] Embodiment 1

[0084] like Figure 1 As shown, the present invention also provides a circuit breaker fault detection optimization method based on SVD-XGBoost, comprising:

[0085] S1: Acquire the original vibration signal generated when the circuit breaker is in operation. The original vibration signal is obtained by simulating different fault conditions by adjusting the degree of jamming of the iron core on the circuit breaker, or simulating different fault degrees by setting different degrees of resistance, so as to obtain data samples of different mechanical fault types of the circuit breaker. The mechanical fault types of the circuit breaker include different degrees of fault signals simulated by loose screws on the base of the circuit breaker, jamming of the iron core and insufficient lubrication.

[0086] In this embodiment, a vibration signal acquisition system is built using a PCB Piezotronics accelerometer 356A04 piezoelectric acceleration sensor and a NI9234 data acquisition card produced by National Instruments (NI). The sampling rate of the circuit breaker vibration signal is 25.6kHz. When recording vibration signals under four mechanical states, the starting point is the same as the acquisition signal time length. Among them, for the loose base screw, the tightness of the base screw can be adjusted to simulate the jamming problem, the jamming problem can be simulated by adjusting the jamming degree of the iron core to simulate different fault conditions, and the problem of insufficient lubrication can be simulated by setting different degrees of resistance to simulate data samples of different fault degrees. Four different state information of the circuit breaker operation is collected according to Table 1, and the state type is marked.

[0087] Table 1. Simulated mechanical failure types

[0088]

[0089] In this embodiment, before obtaining the original vibration signal generated when the circuit breaker is in operation in step S1, the steps include: constructing a vibration signal acquisition system with a circuit breaker, setting a vibration sensor on the circuit breaker, starting a working operation mode to simulate different mechanical fault types of the circuit breaker; activating a signal acquisition device based on the vibration signal acquisition system to capture the analog vibration signal of the circuit breaker in the mechanical state, converting the captured analog vibration signal into a digital signal, collecting and recording it according to a preset sampling rate, and transmitting the collected analog vibration signal to a fault diagnosis system to perform a circuit breaker fault detection operation; wherein the vibration signal acquisition system includes a circuit breaker, a signal acquisition device and a fault diagnosis system connected in sequence, and the signal acquisition device is powered by an external power supply to ensure the normal operation of the acquisition work; the simulated vibration signal includes loose screws on the base of the circuit breaker, stuck iron core, insufficient lubrication and related fault signals of different degrees simulated by adjusting the corresponding mechanical parts on the circuit breaker. In order to effectively detect circuit breaker faults, this embodiment designs and builds a high-precision vibration signal acquisition system, combined with the following algorithm optimization, to improve the accuracy and efficiency of fault identification. The specific acquisition and identification process is as follows: Step 1, signal acquisition device power supply: First, the signal acquisition device is powered by an external power supply to ensure the stability of the power supply voltage and current to ensure the normal operation of the acquisition device. Step 2, configuration of the vibration sensor: stick the vibration sensor on the top of the arc extinguishing chamber of the circuit breaker. Step 3, vibration signal acquisition: activate the signal acquisition device and start collecting vibration data. The acquisition device converts the analog vibration signal into a digital signal through the built-in analog-to-digital converter (ADC) and records it at a predetermined sampling rate such as 25.6kHz. Step 4, data input fault diagnosis system: the collected vibration signal data is transmitted to the fault diagnosis system by wire. Step 5, fault data analysis and identification: the data input to the fault diagnosis system first passes through the preprocessing module to filter, reduce noise and enhance the signal to highlight the fault characteristics, and then the fault diagnosis algorithm proposed in this embodiment is used for state identification. Step 6, display of identification results: the analysis results are displayed through the user interface, and the user can directly observe the status information of the circuit breaker. Among them, the signal acquisition device: the sensor reads the ADC data through the parallel interface, with a sampling frequency of up to 1MHz, and the collected data is downsampled to 32kHz through the FPGA module to make it more suitable for subsequent signal processing and analysis. When setting up the acquisition, the coordinate zero point of the sampling signal is set to when the circuit breaker is about to act to ensure that the data collected each time remains consistent. Vibration signals are captured under four mechanical conditions, including loose base screws, core jamming, and insufficient lubrication. Different degrees of faults are simulated by adjusting the conditions of related mechanical components.

[0090] S2: Based on the NLMS algorithm, data enhancement is performed on the original vibration signal to adaptively filter out noise, thereby obtaining a time domain signal after data enhancement;

[0091] S3: Based on the SVD singular value decomposition and XGBoost algorithm, an SVD-XGBoost model is constructed. The enhanced time domain signal is converted into a frequency domain signal through fast Fourier transform. The frequency domain signal is subjected to data dimension reduction by singular value decomposition and the target features are extracted.

[0092] S4: The extracted target features are input into the trained SVD-XGBoost model, and the optimized network is used to classify the circuit breaker vibration signal to obtain the circuit breaker fault identification result.

[0093] The SVD-XGBoost used in this embodiment is a method that combines singular value decomposition (SVD) and XGBoost algorithms, which is mainly used to solve the problem of feature selection and dimensionality reduction of high-dimensional and sparse data. First, it is necessary to collect relevant data of vibration signals. The collected vibration signal data is preprocessed and sorted by NLMS data to construct a matrix. During this period, steps such as noise removal, filtering, and downsampling are included to obtain clean and suitable data for analysis. The processed vibration signal data is constructed into a matrix. Each column represents a vibration signal sample point, and each row represents a specific time point and a measurement of the vibration signal. Ensure that the size of the data matrix is ​​appropriate and can be processed by SVD. Apply SVD to the constructed vibration signal data matrix. As needed, the most important singular values ​​and corresponding singular vectors in the SVD decomposition can be retained to achieve dimensionality reduction processing. Use the data after SVD dimensionality reduction for XGBoost classification and recognition: prepare dimensionality reduction data and labels, divide training sets and test sets, process features, train XGBoost models, verify performance, tune parameters, and input the new data into the model for classification after dimensionality reduction. The NLMS-SVD and XGBoost fault classification and identification methods are suitable for large-scale, high-dimensional, complex and real-time fault classification problems. The embodiment of the present invention adopts NLMS data processing, which can adaptively filter out noise, reduce the complexity of signal processing, and reduce the computational burden. This helps to improve the efficiency and performance of the system. At the same time, the advantages of SVD dimensionality reduction and XGBoost model are combined to improve the accuracy and efficiency of fault classification. SVD reduces the dimension of the data set and retains the most critical information by extracting the main components of the data. This is particularly effective for large data sets. The data after dimensionality reduction is more concentrated and clean, which helps to reveal the key features and patterns in the signal, and provides more accurate input for subsequent classification. It effectively combines data preprocessing, dimensionality reduction and classification decisions. Each step provides more optimized input for the subsequent process, enhancing the effect of the entire process.

[0094] Circuit breakers are vital equipment in power systems, and their failures may cause serious power outages. Traditional fault detection methods usually rely on preset thresholds or empirical rules, which are difficult to deal with complex and changeable fault characteristics. Therefore, this embodiment proposes a circuit breaker fault detection method that combines advanced signal processing technology and machine learning algorithms to improve detection accuracy and efficiency. It is mainly divided into four steps: signal enhancement, signal conversion, feature extraction, and fault classification. The detailed process is as follows: Figure 1 As shown in the figure, the recursive update is the estimated output, the nature of the output: it represents the signal value after noise suppression processing; the time step refers to the discrete time point in the time series; the correspondence between time and sample: in digital signal processing, time is discretized, and each sample point n corresponds to a discrete time step.

[0095] Step 1: Signal enhancement - Normalized Least Mean Square (NLMS) algorithm

[0096] (1) The original vibration signal x(n) is represented as a time series as input, where n represents the nth sample point;

[0097] (2) Initialization: Initialize the NLMS algorithm parameters, including the weight vector and step size factor μ. The weight vector w(0) is initialized to the zero vector, and the experimental setting is used to balance the convergence speed and stability.

[0098] (3) Recursive update, i.e. noise suppression and signal enhancement:

[0099] Perform preliminary noise suppression on each sample point n:

[0100] a. Signal estimation is preliminary noise suppression: The signal estimation step achieves preliminary noise suppression

[0101] y(n)=w T (n)*x(n);

[0102] Among them, y(n) is the estimated output, w T (n) is the transpose of the weight vector.

[0103] b. Error calculation, i.e. residual noise evaluation: Error calculation evaluates the remaining noise component. The error calculation is performed based on the original vibration signal x(n) and the signal y(n) output by the preliminary noise suppression to evaluate the residual noise component:

[0104] e(n)=x(n)-y(n);

[0105] Where e(n) is the error signal.

[0106] c. Adaptive weight update, i.e. signal enhancement: The weight update process continuously optimizes the filter and enhances the useful signal. Here, the NLMS adaptive filter is used and updated according to the adaptive weight formula:

[0107]

[0108] Among them, w(n) is the weight vector, w(n+1) is the updated weight vector, ||x(n)|| 2 is the square of the modulus of the input signal vector, ∈ is a small positive number.

[0109] (4) Output enhanced signal: Calculate the output enhanced signal:

[0110]

[0111] (5) Termination conditions:

[0112] The algorithm terminates when any of the following conditions is met:

[0113] a. After processing all sample points: n reaches the last index of the signal sequence.

[0114] b. Error convergence: |e(n)|<ε_threshold, where ε_threshold is a predetermined error threshold.

[0115] (6) Iterative Optimization:

[0116] Repeat steps (3)-(5) until the termination condition is met.

[0117] Step 2: Signal conversion - Fast Fourier transform (FFT)

[0118] (1) Input signal after NLMS enhancement

[0119] (2) Perform discrete Fourier transform:

[0120]

[0121] Among them, X(k) represents the frequency domain signal, the kth frequency component; N represents the total number of sampling points of the signal; e represents the natural constant, which is approximately 2.71828; i represents the imaginary unit, satisfying i 2 =-1.

[0122] Step 3: Feature Extraction - Singular Value Decomposition (SVD)

[0123] (1) Matrix construction: Reorganize X(k) into matrix A.

[0124] (2) The target matrix A is constructed based on the frequency domain signal after discrete Fourier transform and SVD singular value decomposition can be performed:

[0125] A=U∑V T ;

[0126] Among them, A is the original matrix, representing the frequency domain signal data; U is the left singular matrix, an orthogonal matrix; Σ is a diagonal matrix, and the diagonal elements are singular values; V is the right singular matrix, an orthogonal matrix, and the transpose is V T ;

[0127] (3) Dimensionality reduction:

[0128] By applying SVD decomposition to the time-frequency spectrum, three matrices U, Σ, and V can be obtained. Similar to the case of the spectrum, the singular values ​​are sorted according to their sizes, and the column vectors or row vectors corresponding to the first k singular values ​​are selected to achieve dimensionality reduction and extract the main feature information. These features comprehensively reflect the key characteristics of the circuit breaker vibration signal, including the main frequency components, energy distribution, time-frequency patterns, etc.

[0129] Specifically, the first m largest singular values ​​and the corresponding right singular vectors are selected, and the high-dimensional data is projected into a low-dimensional space consisting of m key features for dimensionality reduction in the following way:

[0130]

[0131] Where Σm represents the diagonal matrix of the first m selected singular values; The columns of the right singular matrix corresponding to the first m singular values, z is the target feature after dimensionality reduction, which will be used as input for the subsequent fault classification model.

[0132] Step 4. Fault Classification — XGBoost

[0133] (1) Input features: The target feature z obtained by SVD dimensionality reduction is used as the input of the XGBoost model;

[0134] (2) Define the objective function: Input the feature information after dimensionality reduction and construct the objective function, that is, construct the objective function based on the loss function and regularization term:

[0135]

[0136] Among them, Obj(θ) is the objective function, is the loss function, which is used to measure the difference between the prediction and the actual result; Ω(fk) is the regularization term that controls the complexity of the model, N represents the number of samples, and K is the number of decision trees included in the model;

[0137] (3) Training classification: Build a decision tree group through a weighted voting mechanism and calculate the gain value of each branch to determine the optimal splitting strategy:

[0138]

[0139] Among them, G j represents the gradient sum of all samples belonging to the jth node, I j represents the sample set of the jth node, H j represents the sum of the second-order derivatives of all samples belonging to the jth node, g i is the first-order gradient of the ith sample, h i is the second-order gradient of the i-th sample;

[0140] The score of the split point is calculated based on the gradient sum and the second-order derivative sum, so as to adjust and select the optimal split point:

[0141]

[0142] Among them, score represents the score of the optimized split point, λ is the regularization parameter, and H all represents the gradient and second-order derivative on the entire data set, and γ is the penalty parameter for node partitioning.

[0143] The key innovation of the embodiment of the present invention is to integrate multiple technologies such as NLMS data processing, SVD dimension reduction and XGBOOST fault diagnosis to form a comprehensive circuit breaker vibration signal processing system. SVD dimension reduction and XGBoost classification and recognition are the innovations of the present invention. Traditional fault classification methods usually directly use the original high-dimensional features for modeling and classification, which will face the problems of high data dimension and large computational complexity. The SVD-XGBoost method converts high-dimensional features into low-dimensional features by introducing SVD dimension reduction technology, reduces the number of features, improves computational efficiency, and can retain the main data features, reducing the risk of information loss. The SVD-XGBoost model uses the low-dimensional features of SVD dimension reduction as input, and then uses the XGBoost model for fault classification and identification. XGBoost is a powerful gradient boosting tree model with excellent nonlinear modeling and high-accuracy classification capabilities. By combining SVD and XGBoost, deep learning and nonlinear modeling of the reduced-dimensional features are realized, which improves the accuracy and effect of fault classification. The SVD-XGBoost model can also be applied to multi-class fault classification problems, that is, to simultaneously identify and classify multiple different types of faults. The multi-class fault classification problem is challenging, but the XGBoost model has good performance in dealing with multi-class problems and can be optimized by adjusting parameters. SVD-XGBoost combines the advantages of SVD dimensionality reduction and XGBoost model to achieve efficient classification and identification of multi-class faults. The SVD-XGBoost model is suitable for fault diagnosis and classification tasks in real-time environments. By using the efficient XGBoost model and SVD dimensionality reduction feature representation, it can process and analyze fault data streams in real time, and quickly and accurately identify and classify faults, which is of great significance for real-time fault diagnosis and prediction, and timely measures can be taken to avoid potential power system failures and damage.

[0144] After the above steps are performed for on-site collection and application, the algorithm verification and analysis of this embodiment is as follows:

[0145] (1) The collected vibration signal may contain various noises. The normalized least mean square (NLMS) algorithm is used to enhance the signal, reduce noise interference, and highlight fault characteristics.

[0146] (2) Then, the signal after data enhancement is subjected to fast Fourier transform. By applying the enhanced data to frequency domain analysis, it can help improve the fault diagnosis and monitoring capabilities of vibration signals. In addition, time-frequency analysis can show the effect of signal enhancement, allowing users to observe the changes in frequency trajectory. The four states after transformation are as follows: Figure 3-6 shown.

[0147] (3) Perform singular value decomposition (SVD) on the frequency domain signal and select the eigenvector corresponding to the most representative singular value.

[0148] (4) In order to verify the generalization feature ability of the model, 20% of each type of sample is randomly selected as a test set to verify the effectiveness of the model, and the remaining 80% is used as a training set to train the model. In addition, in order to avoid network overfitting, all types of samples and corresponding labels are randomly sorted as a data set and input into the XGBoost algorithm in a shuffled order to ensure the effectiveness of learning. The parameters of the XGBoost model are: the maximum tree degree is 20; the learning rate is selected as 0.4; the number of categories is 4, and the shrinking step size is 0.8.

[0149] To further understand the detailed classification of each state type, output the confusion matrix, divide each element by the sum of its row, and then round it to 4 decimal places to normalize the confusion matrix. The result is as follows Figure 7 shown.

[0150] from Figure 7 It can be found that the embodiment of the present invention has a high accuracy in fault diagnosis of circuit breakers. The diagnostic accuracy of normal state and loose screw state reaches 100%, the accuracy of stuck state reaches 98.81%, and the accuracy of insufficient lubrication reaches 99.05%, which can effectively distinguish easily confused fault types.

[0151] Embodiment 2

[0152] Based on the same concept, the present invention provides a circuit breaker fault detection optimization system based on SVD-XGBoost, comprising:

[0153] A data acquisition module, used to obtain the original vibration signal generated when the circuit breaker is in operation;

[0154] A data enhancement module, used for performing data enhancement on the original vibration signal based on the NLMS algorithm to adaptively filter out noise, thereby obtaining a time domain signal after data enhancement;

[0155] SVD-XGBoost model building module, used to build an SVD-XGBoost model based on SVD singular value decomposition and XGBoost algorithm, convert the enhanced time domain signal into a frequency domain signal through fast Fourier transform, use singular value decomposition to reduce the data dimension of the frequency domain signal and extract target features;

[0156] The fault identification module is used to input the extracted target features into the trained SVD-XGBoost model, use the optimized network to classify the circuit breaker vibration signal, and obtain the circuit breaker fault identification result.

[0157] Specifically, the fault identification module includes:

[0158] The target establishment unit is used to use the target features obtained through SVD dimensionality reduction as the input of the XGBoost model and construct the target function based on the loss function and regularization term:

[0159]

[0160] Among them, Obj(θ) is the objective function, is the loss function, which is used to measure the difference between the prediction and the actual result; Ω(fk) is the regularization term that controls the complexity of the model, N represents the number of samples, and K is the number of decision trees included in the model;

[0161] Training classification units is used to build decision tree groups through a weighted voting mechanism, and the gain value of each branch is calculated to determine the optimal splitting strategy:

[0162]

[0163] Among them, G j represents the gradient sum of all samples belonging to the jth node, I j represents the sample set of the jth node, H j represents the sum of the second-order derivatives of all samples belonging to the jth node, g i is the first-order gradient of the ith sample, h i is the second-order gradient of the i-th sample;

[0164] The scoring optimization unit is used to calculate the score of the split point based on the gradient sum and the second-order derivative sum, so as to adjust and select the optimized split point:

[0165]

[0166] Among them, score represents the score of the optimized split point, λ is the regularization parameter, and H all represents the gradient and second-order derivative on the entire data set, and γ is the penalty parameter for node partitioning.

[0167] It should be noted that the division of the modules in the embodiment of the device / system is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or physically separated. Moreover, these modules can be implemented in the form of software calling through processing elements; or in the form of hardware; or some units can be implemented in the form of software calling through processing elements, and some units can be implemented in the form of hardware.

[0168] The implementation principle of the above modules has been described in the above embodiments, so it will not be repeated here.

[0169] Embodiment 3

[0170] Based on the same concept, in some embodiments of the present application, an electronic device is also provided. The electronic device includes a memory and a processor, wherein the memory is used to store a processing program, and the processor executes the processing program according to instructions. When the processor executes the processing program, the circuit breaker fault detection optimization method based on SVD-XGBoost in the aforementioned embodiment is implemented.

[0171] In some embodiments of the present application, a readable storage medium is further provided, which may be a non-volatile readable storage medium or a volatile readable storage medium. The readable storage medium stores instructions, and when the instructions are executed on a computer, an electronic device including the readable storage medium executes the aforementioned circuit breaker fault detection optimization method based on SVD-XGBoost.

[0172] It is understandable that, for the aforementioned circuit breaker fault detection optimization method based on SVD-XGBoost, if it is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or part of the contribution to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-only memory, ROM), random access memory (Random access memory, RAM), disk or optical disk and other media that can store program code.

[0173] Computer readable storage media may include data signals propagated in baseband or as part of a carrier wave, wherein readable program codes are carried. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device. The program codes contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination thereof.

[0174] The program code involved in executing the technical solution disclosed in this application can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Python, C++, etc., and also conventional procedural programming languages ​​such as C language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0175] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A circuit breaker fault detection optimization method based on SVD-XGBoost, characterized in that: include: Acquire the original vibration signal generated when the circuit breaker is in operation; Based on the NLMS algorithm, data enhancement is performed on the original vibration signal to adaptively filter out noise, thereby obtaining a time domain signal after data enhancement; An SVD-XGBoost model is constructed based on the combination of SVD singular value decomposition and XGBoost algorithm. The enhanced time domain signal is converted into a frequency domain signal through fast Fourier transform. The frequency domain signal is subjected to data dimension reduction by singular value decomposition and the target features are extracted. The extracted target features are input into the trained SVD-XGBoost model, and the optimized network is used to classify the circuit breaker vibration signal to obtain the circuit breaker fault identification result.

2. The circuit breaker fault detection optimization method based on SVD-XGBoost according to claim 1 is characterized in that: The data enhancement of the original vibration signal based on the NLMS algorithm to adaptively filter out noise to obtain a time domain signal after data enhancement includes: The original vibration signal x(n) is represented as a time series as input, where n represents the nth sample point; Initialize the NLMS algorithm parameters, including the weight vector and step factor; Perform preliminary noise suppression on each sample point n and estimate the signal y(n)=w T (n)*x(n): The residual noise component is estimated by calculating the error between the original vibration signal x(n) and the signal y(n) output by the preliminary noise suppression: e(n)=x(n)-y(n); Update according to the adaptive weight formula: Calculate the output enhancement signal: Iterate and update until all sample points are traversed or the error converges to the preset error threshold; Where e(n) is the error signal, μ is the step size factor, and w T (n) is the transpose of the weight vector, w(n) is the weight vector, w(n+1) is the updated weight vector, ||x(n)|| 2 is the square of the modulus of the input signal vector, ∈ is a small positive number.

3. The circuit breaker fault detection optimization method based on SVD-XGBoost according to claim 1, characterized in that: The method of converting the enhanced time domain signal into a frequency domain signal by fast Fourier transform comprises: Input signal after NLMS enhancement Perform a discrete Fourier transform: Among them, X(k) represents the frequency domain signal, the kth frequency component; N represents the total number of sampling points of the signal; e represents a natural constant; and i represents an imaginary unit.

4. The circuit breaker fault detection optimization method based on SVD-XGBoost according to claim 1, characterized in that: The adopting singular value decomposition to perform data dimension reduction on the frequency domain signal and extract target features comprises: Constructing the target matrix A based on the frequency domain signal after discrete Fourier transformation can perform SVD singular value decomposition: A=U∑V T ; Among them, A is the original matrix, representing the frequency domain signal data; U is the left singular matrix, an orthogonal matrix; Σ is a diagonal matrix, and the diagonal elements are singular values; V is the right singular matrix, an orthogonal matrix, and the transpose is VT; Select the first m largest singular values ​​and the corresponding right singular vectors, and project the high-dimensional data into a low-dimensional space consisting of m key features for dimensionality reduction in the following way: Where Σm represents the diagonal matrix of the first m selected singular values; The columns of the right singular matrix corresponding to the first m singular values, z is the target feature after dimensionality reduction.

5. The circuit breaker fault detection optimization method based on SVD-XGBoost according to claim 1, characterized in that: The extracted target features are input into the trained SVD-XGBoost model, and the optimized network is used to classify the circuit breaker vibration signal, and the circuit breaker fault identification results obtained include: The target features obtained through SVD dimensionality reduction are used as the input of the XGBoost model; Construct the objective function based on the loss function and regularization term: Among them, Obj(θ) is the objective function, is the loss function, which is used to measure the difference between the prediction and the actual result; Ω(fk) is the regularization term that controls the complexity of the model, N represents the number of samples, and K is the number of decision trees included in the model; Training classification: Build a decision tree group through a weighted voting mechanism and calculate the gain value of each branch to determine the optimal splitting strategy: Among them, G j represents the gradient sum of all samples belonging to the jth node, I j represents the sample set of the jth node, H j represents the sum of the second-order derivatives of all samples belonging to the jth node, g i is the first-order gradient of the ith sample, h i is the second-order gradient of the i-th sample; The score of the split point is calculated based on the gradient sum and the second-order derivative sum, so as to adjust and select the optimal split point: Among them, score represents the score of the optimized split point, λ is the regularization parameter, and H all represents the gradient and second-order derivative on the entire data set, and γ is the penalty parameter for node partitioning.

6. The circuit breaker fault detection optimization method based on SVD-XGBoost according to claim 1, characterized in that: The obtaining of the original vibration signal generated when the circuit breaker is in operation comprises: By adjusting the degree of jamming of the iron core on the circuit breaker to simulate different fault conditions, or by setting different degrees of resistance to simulate different degrees of fault, data samples of different mechanical fault types of the circuit breaker are obtained, wherein the mechanical fault type of the circuit breaker includes different degrees of fault signals simulated by loose screws on the base of the circuit breaker, jamming of the iron core and insufficient lubrication.

7. A circuit breaker fault detection optimization system based on SVD-XGBoost, characterized in that: include: A data acquisition module, used to obtain the original vibration signal generated when the circuit breaker is in operation; A data enhancement module, used for performing data enhancement on the original vibration signal based on the NLMS algorithm to adaptively filter out noise, thereby obtaining a time domain signal after data enhancement; SVD-XGBoost model building module, used to build an SVD-XGBoost model based on SVD singular value decomposition and XGBoost algorithm, convert the enhanced time domain signal into a frequency domain signal through fast Fourier transform, use singular value decomposition to reduce the data dimension of the frequency domain signal and extract target features; The fault identification module is used to input the extracted target features into the trained SVD-XGBoost model, use the optimized network to classify the circuit breaker vibration signal, and obtain the circuit breaker fault identification result.

8. The circuit breaker fault detection optimization system based on SVD-XGBoost according to claim 7, characterized in that: The fault identification module comprises: The target establishment unit is used to use the target features obtained through SVD dimensionality reduction as the input of the XGBoost model and construct the target function based on the loss function and regularization term: Among them, Obj(θ) is the objective function, is the loss function, which is used to measure the difference between the prediction and the actual result; Ω(fk) is the regularization term that controls the complexity of the model, N represents the number of samples, and K is the number of decision trees included in the model; Training classification units is used to build decision tree groups through a weighted voting mechanism, and the gain value of each branch is calculated to determine the optimal splitting strategy: Among them, G j represents the gradient sum of all samples belonging to the jth node, I j represents the sample set of the jth node, H j represents the sum of the second-order derivatives of all samples belonging to the jth node, g i is the first-order gradient of the ith sample, h i is the second-order gradient of the i-th sample; The scoring optimization unit is used to calculate the score of the split point based on the gradient sum and the second-order derivative sum, so as to adjust and select the optimized split point: Among them, score represents the score of the optimized split point, λ is the regularization parameter, and H all represents the gradient and second-order derivative on the entire data set, and γ is the penalty parameter for node partitioning.

9. An electronic device, characterized in that: include: A memory, the memory being used to store a processing program; A processor, wherein when executing the processing program, the processor implements the circuit breaker fault detection optimization method based on SVD-XGBoost as described in any one of claims 1 to 6.

10. A readable storage medium, characterized in that: The readable storage medium stores a processing program, and when the processing program is executed by the processor, the circuit breaker fault detection optimization method based on SVD-XGBoost according to any one of claims 1 to 6 is implemented.

Images