A hybrid feature fault diagnosis method and system based on low-voltage intelligent circuit breaker

Abstract

The application discloses a kind of hybrid feature fault diagnosis method and system based on low-voltage intelligent circuit breaker, belong to low-voltage intelligent circuit breaker fault diagnosis field.Adopt fuzzy entropy, variance, energy moment, the ratio of the variance of first-order difference signal and original signal variance and the ratio of the mobility of first-order derivative and the mobility of original signal to extract fault feature, and singular value decomposition is carried out to the parameter matrix that variance, energy moment, the ratio of the variance of first-order difference signal and original signal variance and the ratio of the mobility of first-order derivative and the mobility of original signal, to avoid feature redundancy, compared with single feature detection, greatly improve the fault diagnosis accuracy.Through the feature of the vibration signal of multiple low-voltage intelligent circuit breaker is collected, overcome the problem that traditional low-voltage intelligent circuit breaker fault diagnosis research is less.Can accurately identify the fault of low-voltage intelligent circuit breaker, realize the fault diagnosis of low-voltage intelligent circuit breaker.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of low-voltage intelligent circuit breaker fault diagnosis, and relates to a hybrid feature fault diagnosis method and system based on a low-voltage intelligent circuit breaker. BACKGROUND

[0002] Circuit breakers are widely used in power systems, and their safe operation is a powerful guarantee for the stable operation of power grids and the safe use of electricity by users. As a key component for controlling and protecting low-voltage distribution networks, low-voltage intelligent circuit breakers are an indispensable part of smart grid construction, and are responsible for the important duties of closing, carrying, and breaking the normal operating current, overload current, and short-circuit current of the operating circuit. With the development of smart distribution networks in China, higher requirements are placed on the safety performance of low-voltage intelligent circuit breakers. Data shows that the main fault of current low-voltage intelligent circuit breakers is mechanical failure. Due to the special nature of its structure and position, it is not the best method to determine whether a low-voltage intelligent circuit breaker is intact by disassembling it. Regular maintenance cannot determine whether the circuit breaker is intact at all times and may degrade the performance of the circuit breaker. Therefore, developing fault detection technology for low-voltage intelligent circuit breakers is of great significance to improving the safety of low-voltage intelligent circuit breakers.

[0003] A low-voltage intelligent circuit breaker is a complex device with many types of faults. In summary, signals that can reflect the operating state of a low-voltage intelligent circuit breaker include opening and closing coil current signals, sound signals, and vibration signals. Opening and closing current signals can only be used to detect faults associated with the structure of the electromagnet, but if the fault is not associated with the electromagnet, it cannot be reflected by the opening and closing current signals. Sound signals are associated with vibration signals, and vibration generates sound, but the propagation of sound signals relies on air, which can easily lead to loss of signal information. The movement of parts of a low-voltage intelligent circuit breaker generates vibrations, and vibration signals can reflect most of the faults of a low-voltage intelligent circuit breaker. Therefore, the vibration signals of a low-voltage intelligent circuit breaker are a good carrier of state signals. However, due to the complexity of the vibration signals of a low-voltage intelligent circuit breaker, it is a major challenge to efficiently and accurately extract fault features from the vibration signals of a low-voltage intelligent circuit breaker. At present, most research on low-voltage intelligent circuit breaker fault diagnosis based on vibration signals uses entropy as an information feature. Fuzzy entropy can more accurately describe the complexity of a system, but since one-sided features can only reflect part of the characteristics of fault signals, they cannot be used for fault recognition using multiple features, and cannot more comprehensively reflect fault feature information. SUMMARY

[0004] The present application aims to solve the problem of low accuracy in fault diagnosis of low-voltage intelligent circuit breakers due to the inability to utilize multiple features for fault identification by reacting to partial features of fault signals through unilateral features in the prior art, and to provide a hybrid feature fault diagnosis method and system based on a low-voltage intelligent circuit breaker.

[0005] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0006] The hybrid feature fault diagnosis method based on a low-voltage intelligent circuit breaker proposed by the present application comprises the following steps:

[0007] The noise-reduced vibration signal is obtained, and the vibration signal is decomposed to obtain the optimal decomposition parameter of the vibration signal; the fuzzy entropy value is obtained according to the optimal decomposition parameter of the vibration signal, and a fuzzy entropy feature vector is formed;

[0008] According to the variance of the optimal decomposition parameter of the vibration signal, the energy moment, the ratio of the variance of the first-order difference signal to the variance of the original signal, and the ratio of the mobility of the first-order derivative to the mobility of the original signal, a parameter matrix is constructed;

[0009] The parameter matrix is singular value decomposed to obtain a decomposed feature vector, and a hybrid feature vector is obtained according to the decomposed feature vector and the fuzzy entropy feature vector; the hybrid feature vector is input into a support vector machine to obtain a fault type.

[0010] Preferably, the method for obtaining the noise-reduced vibration signal is as follows:

[0011] The wavelet basis function is selected as the wavelet basis of discrete wavelet transform, and the maximum decomposition layer j is determined by the ratio of the minimum frequency f min of the vibration signal to the center frequency f0of the wavelet basis function; the vibration signal x w (t) is decomposed by j layers of discrete wavelet, the threshold value of each layer of wavelet high-frequency coefficient is determined according to the principle of unbiased likelihood estimation, and the vibration signal is reconstructed according to the low-frequency coefficient of wavelet decomposition and the high-frequency coefficient after thresholding to obtain the noise-reduced vibration signal u w (t).

[0012] Preferably, the method for obtaining the optimal decomposition parameter k of the vibration signal is as follows:

[0013] The noise-reduced vibration signal u w (t) is decomposed into K finite bandwidth eigenmode functions, and the constrained variational problem is described as follows:

[0014]

[0015] Where, {u k}={u1,u2……u k} represents the K modal components obtained from the decomposition; {ω k}={ω1,ω2……ω k} represents the center frequency of each modal component. After Hilbert transformation, u k The spectrum of (t), where * represents the convolution operation, st indicates the constraint, and j is the imaginary unit. For gradient operators;

[0016] Transform the modal components and their center frequencies to the frequency domain:

[0017]

[0018]

[0019] in, They are respectively u w The Fourier transforms of λ(t) and λ(t), the Lagrange multiplier operator of λ(t), and then the iterative search algorithm of alternating directions of the multiplier operator are used to find the optimal solution to the problem. K solutions are obtained through iterative updates. The real part obtained by performing the inverse Fourier transform is the modal component u. k (t) is the optimal decomposition parameter k of the vibration signal.

[0020] Preferably, the method for obtaining fuzzy entropy values ​​and constructing fuzzy entropy feature vectors is as follows:

[0021] Given a signal sampling frequency of N, define the phase space dimension m (m≤N-2), and reconstruct the phase space X(i)=[u k (i),u k (i+1),...,u k (i+m-1)]-u0(i),i=1,2,...,N-m+1, u0(i) is the mean value,

[0022]

[0023] Where j≠i is the maximum margin between window vectors X(i) and X(j);

[0024] Fuzzy membership function The correlation between phase spaces X(i) and X(j) is as follows: And j≠i, Let n be an exponential function, n be the boundary gradient of the function, and r be the boundary width;

[0025] For each i, we can calculate its average value to obtain... The ambiguity similarity function is: The fuzzy entropy of the original vibration signal is y i = lnΦ m (r) - lnΦ m+1 (r) ; wherein, m is the reconstruction dimension; r is (0.1-0.25)S d , S d is the standard deviation of the sampling signal.

[0026] Preferably, the parameter matrix is constructed according to the variance of the optimal decomposition parameter of the vibration signal, the energy moment, the ratio of the variance of the first-order difference signal to the variance of the original signal, and the ratio of the mobility of the first-order derivative to the mobility of the original signal, and the method is as follows:

[0027]

[0028] Wherein, A i , B i , C i , D i is the variance of the i-th IMF component, the energy moment, the ratio of the variance of the first-order difference signal to the variance of the original signal, and the ratio of the mobility of the first-order derivative to the mobility of the original signal.

[0029] Preferably, the singular value decomposition is performed on the parameter matrix, and the method for obtaining the decomposed feature vector is as follows:

[0030]

[0031] Wherein, ∑ is a diagonal matrix composed of singular values, U and V T are feature vectors obtained by singular value decomposition, q=min(k,4), the diagonal matrix ∑ composed of singular values is converted into a feature vector ∑ V which can represent H, and the fuzzy entropy feature vector and the singular value vector are combined to form the final feature of the sampling signal.

[0032] Preferably, the method for obtaining the mixed feature vector according to the decomposed feature vector and the fuzzy entropy feature vector is as follows: the mixed feature vector is F=[Y ∑ V ], Y is the fuzzy entropy feature vector.

[0033] The application provides a mixed feature fault diagnosis system based on a low-voltage intelligent circuit breaker, which comprises:

[0034] A fuzzy entropy feature vector acquisition module is configured to acquire a denoised vibration signal, decompose the vibration signal to obtain optimal decomposition parameters of the vibration signal, acquire a fuzzy entropy value according to the optimal decomposition parameters of the vibration signal, and form a fuzzy entropy feature vector.

[0035] A parameter matrix construction module is configured to construct a parameter matrix according to the variance of the optimal decomposition parameter of the vibration signal, the ratio of the energy moment, the variance of the first-order differential signal and the variance of the original signal, and the ratio of the mobility of the first-order derivative and the mobility of the original signal.

[0036] A hybrid feature vector acquisition module is configured to perform singular value decomposition on the parameter matrix, acquire the decomposed feature vector, acquire a hybrid feature vector according to the decomposed feature vector and the fuzzy entropy feature vector, and input the hybrid feature vector into a support vector machine to acquire the fault type.

[0037] A computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the hybrid feature fault diagnosis method based on the low-voltage intelligent circuit breaker when executing the computer program.

[0038] A computer readable storage medium stores a computer program, and the computer program implements the steps of the hybrid feature fault diagnosis method based on the low-voltage intelligent circuit breaker when executed by a processor.

[0039] Compared with the prior art, the present application has the following beneficial effects:

[0040] The purpose of the present application is to provide a hybrid feature fault diagnosis method based on a low-voltage intelligent circuit breaker, which combines multiple feature quantities to diagnose the low-voltage circuit breaker, extracts fault features by using fuzzy entropy, variance, energy moment, the ratio of the variance of the first-order differential signal and the variance of the original signal, and the ratio of the mobility of the first-order derivative and the mobility of the original signal, avoids feature redundancy, greatly improves the fault diagnosis accuracy compared with single feature detection, collects multiple low-voltage intelligent circuit breaker vibration signal features, overcomes the problem that traditional low-voltage intelligent circuit breaker fault diagnosis research is less, can accurately identify the fault of the low-voltage intelligent circuit breaker, and realizes the fault diagnosis of the low-voltage intelligent circuit breaker.

[0041] The hybrid feature fault diagnosis system based on the low-voltage intelligent circuit breaker provided by the present application realizes fault diagnosis by dividing the system into a fuzzy entropy feature vector acquisition module, a parameter matrix construction module and a hybrid feature vector acquisition module. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0043] Figure 1 Flow chart of the mixed feature fault diagnosis method based on low-voltage intelligent circuit breaker of the present application.

[0044] Figure 2 Flow chart of the mixed feature fault diagnosis method of the present application.

[0045] Figure 3 Basic idea chart of the SVM of the present application.

[0046] Figure 4 System chart of the mixed feature fault diagnosis system based on low-voltage intelligent circuit breaker of the present application. DETAILED DESCRIPTION

[0047] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will combine the drawings in the embodiments of the present application to make a clear and complete description of the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations.

[0048] Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0049] It should be noted that: similar labels and letters represent similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0050] In the description of the embodiments of the present application, it should be noted that if the terms "upper", "lower", "horizontal", "inner" and the like indicating the orientation or position relationship are based on the orientation or position relationship shown in the drawings, or the orientation or position relationship when the product of the present application is usually placed, and are only for the convenience of describing the present application and simplifying the description, and therefore cannot be understood as limiting the present application by indicating or implying that the devices or elements referred to must have a specific orientation, construction and operation. In addition, the terms "first", "second" and the like are only used for differentiation and cannot be understood as indicating or implying relative importance.

[0051] In addition, if the term "horizontal" appears, it does not mean that the component must be absolutely horizontal, but can be slightly inclined. For example, "horizontal" only means that its direction is relatively more horizontal than "vertical", and does not mean that the structure must be completely horizontal, but can be slightly inclined.

[0052] In the description of the embodiments of the present application, it should be noted that unless otherwise explicitly specified and limited, if the terms "arrangement", "installation", "connection", "connection" appear, they should be understood in a broad sense, for example, they can be fixedly connected, or can be detachably connected, or integrally connected; can be mechanically connected, or can be electrically connected; can be directly connected, or indirectly connected through an intermediate medium; can be connected inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0053] The present application will be described in further detail below in conjunction with the drawings:

[0054] The present application proposes a hybrid feature fault diagnosis method based on a low-voltage intelligent circuit breaker, as shown in Figure 1 The method comprises the following steps:

[0055] S1, obtaining a denoised vibration signal, decomposing the vibration signal to obtain the optimal decomposition parameter of the vibration signal; obtaining the fuzzy entropy value according to the optimal decomposition parameter of the vibration signal, and constituting a fuzzy entropy feature vector;

[0056] The method for obtaining the denoised vibration signal is as follows:

[0057] Selecting a wavelet base function as a discrete wavelet change wavelet base, determining the maximum decomposition layer j through the ratio of the minimum frequency f min of the vibration signal to the center frequency f0 of the wavelet base function, and performing j-layer discrete wavelet decomposition on the vibration signal x w (t), determining the wavelet high-frequency coefficient threshold value of each layer according to the unbiased likelihood estimation principle, and reconstructing the vibration signal according to the low-frequency coefficient of the wavelet decomposition and the high-frequency coefficient after the threshold value is acted on, to obtain the denoised vibration signal u w (t).

[0058] The optimal decomposition parameter k of the vibration signal is obtained as follows:

[0059] The noise-reduced vibration signal u w The problem (t) is decomposed into K eigenmode functions with finite bandwidths, and the constrained variational problem is described as follows:

[0060]

[0061] Among them, {u k}={u1,u2……u k} represents the K modal components obtained from the decomposition; {ω k}={ω1,ω2……ω k} represents the center frequency of each modal component. After Hilbert transformation, u k The spectrum of (t), where * represents the convolution operation, st indicates the constraint, and j is the imaginary unit. For gradient operators;

[0062] Transform the modal components and their center frequencies to the frequency domain:

[0063]

[0064]

[0065] in, They are respectively u w The Fourier transforms of λ(t) and λ(t), the Lagrange multiplier operator of λ(t), and then the iterative search algorithm of alternating directions of the multiplier operator are used to find the optimal solution to the problem. K solutions are obtained through iterative updates. The real part obtained by performing the inverse Fourier transform is the modal component u. k (t) is the optimal decomposition parameter k of the vibration signal.

[0066] The method for obtaining fuzzy entropy values ​​and constructing fuzzy entropy feature vectors is as follows:

[0067] Given a signal sampling frequency of N, define the phase space dimension m (m≤N-2), and reconstruct the phase space X(i)=[u k (i),u k (i+1),...,u k (i+m-1)]-u0(i),i=1,2,...,N-m+1, u0(i) is the mean value,

[0068]

[0069] where j≠i is the maximum interval between the window vectors X(i) and X(j);

[0070] fuzzy membership function The correlation between the phase spaces X(i) and X(j) is as follows: and j≠i, is an exponential function, n is the boundary gradient of the function, and r is the boundary width;

[0071] The average value of each i is obtained as

[0072] The fuzzy degree similarity function is:

[0073] The fuzzy entropy of the original vibration signal is y i =lnΦ m (r)-lnΦ m+1 (r);where m is the reconstruction dimension; r is (0.1-0.25)S d , S d is the standard deviation of the sampling signal.

[0074] S2, according to the variance of the optimal decomposition parameter of the vibration signal, the energy moment, the ratio of the variance of the first-order difference signal to the variance of the original signal, and the ratio of the mobility of the first-order derivative to the mobility of the original signal, a parameter matrix is constructed;

[0075] According to the variance of the optimal decomposition parameter of the vibration signal, the energy moment, the ratio of the variance of the first-order difference signal to the variance of the original signal, and the ratio of the mobility of the first-order derivative to the mobility of the original signal, a parameter matrix is constructed. The method is as follows:

[0076]

[0077] where A i , B i , C i , D i are the variance, energy moment, ratio of the variance of the first-order difference signal to the variance of the original signal, and ratio of the mobility of the first-order derivative to the mobility of the original signal of the i-th IMF component, respectively.

[0078] S3, singular value decomposition is performed on the parameter matrix to obtain a decomposed feature vector, a mixed feature vector is obtained according to the decomposed feature vector and the fuzzy entropy feature vector, and the mixed feature vector is input into a support vector machine to obtain a fault type.

[0079] The method for performing singular value decomposition on the parameter matrix to obtain a decomposed feature vector is as follows:

[0080]

[0081] where ∑ is a diagonal matrix composed of singular values, U and V T are eigenvectors obtained by singular value decomposition, q = min(k, 4), and the diagonal matrix ∑ composed of singular values is converted into the eigenvector ∑ V representing H.

[0082] The method for obtaining the mixed eigenvector according to the decomposed eigenvector and the fuzzy entropy eigenvector is as follows: the mixed eigenvector is F = [Y ∑ V ], Y is the fuzzy entropy eigenvector.

[0083] As shown in FIG. Figure 2 , the specific implementation steps of the low-voltage intelligent circuit breaker mixed feature fault diagnosis method based on fuzzy entropy and singular value decomposition are as follows:

[0084] Step 1: Collect the vibration signals x w (t) of the low-voltage intelligent circuit breaker in normal state and various fault states, and establish a signal database; wherein w represents a data group, and represents the normal state and various fault states, respectively.

[0085] Step 2: Pretreat the vibration signals x w (t) collected in S1 by using the improved wavelet denoising method, that is, denoising, to obtain the denoised vibration signals u w (t), and the specific process of denoising is as follows:

[0086] Select a dbN wavelet basis function with orthogonality as the wavelet basis of discrete wavelet transform, determine the maximum decomposition layer j through the ratio of the minimum frequency f min of the useful signal to the center frequency f0 of the wavelet basis function, perform j-layer discrete wavelet decomposition on the vibration signal x w (t), determine the wavelet high-frequency coefficient threshold value of each layer according to the unbiased likelihood estimation principle, and perform wavelet reconstruction on the vibration signal according to the low-frequency coefficient of wavelet decomposition and the high-frequency coefficient after thresholding, to obtain the denoised vibration signal u w (t).

[0087] Step 3: Decompose the denoised vibration signal u w (t) in step 2 by using the optimized variational mode decomposition algorithm, and find the optimal decomposition parameters k and penalty factor a of each type of signal under the condition of ensuring the maximum fitness, wherein k is the number of intrinsic mode components (IMF) obtained by VMD decomposition;

[0088] Step 31: Initialize the whale population vector position in the WOA algorithm as [[k, a], and use this parameter to u w(t) Perform VMD decomposition to obtain k IMF components; the specific steps are as follows:

[0089] The noise-reduced signal u w The problem (t) is decomposed into K eigenmode functions with finite bandwidths. Accordingly, the constrained variational problem is described as follows:

[0090]

[0091] Among them, {u k}={u1,u2……u k} represents the K modal components obtained from the decomposition; {ω k}={ω1,ω2·····ω k} represents the center frequency of each modal component. After Hilbert transformation, u k The spectrum of (t), where * represents the convolution operation, st indicates the constraint, and j is the imaginary unit. This is the gradient operator.

[0092] Introduce the augmented Lagrangian function and transform the expression to the frequency domain:

[0093]

[0094]

[0095] in, They are respectively u w Fourier transforms of λ(t) and λ(t), where λ(t) is the Lagrange multiplier operator.

[0096] Based on the above, an alternating direction algorithm with multiplication operators is used to iteratively search for the optimal solution to the problem, and finally, K solutions are obtained through iterative updates. The real part obtained by performing the inverse Fourier transform is the modal component u. k (t).

[0097] Step 32: Calculate the k IMF components u k The kurtosis value of (t) will be the maximum value of the obtained kurtosis value. The fitness function of the WOA algorithm is used, but this is only for local optimization; the goal of global optimization is... Maximize, Maximum The corresponding k and α are the optimal parameters;

[0098] The WOA algorithm is used to optimize the parameters k and α of VMD and initialize the whale population vector position [k, a]. The specific steps are as follows:

[0099] Set the number of whales N and the number of algorithm iterations t.max Initialize the whale swarm vector position [k, a], the i-th individual position is as follows:

[0100] X i = r · (ub - lb) + lb

[0101] wherein r is a random vector between [0, 1]; X i is in the range of [ub, lb]; a is a convergence factor, linearly decreasing from 2 to 0, VMD is used to process signals according to each whale position, and the maximum value K i of the shearing degree corresponding to each whale vector individual is calculated, and its position is recorded as the global optimal position X*(t).

[0102]

[0103]

[0104]

[0105] wherein, p is a random number and is in the range of [0, 1], r is a random vector between [0, 1]; a is a convergence factor, linearly decreasing from 2 to 0, X represents the position of the whale; t is the number of iteration searches; A and C are coefficient vectors.

[0106] When searching for prey, the whale swarm spirals upward, and in the process of spiraling upward, the whale also shrinks its enclosure. Considering these two cases, and considering that the probability of selecting a shrinking enclosure and a spiral position update is 50%, the position relationship expression of the search is as follows:

[0107]

[0108]

[0109] wherein b is a constant, different b values correspond to different spiral shapes; l is a uniformly distributed random number; p is a distributed random number between [0, 1]. When searching globally, the whale outside the enclosure has randomness in searching for prey, and when |A|≥1, it indicates that the whale is outside the enclosure, at this time, random search is adopted, and |A|≤1, which indicates that the whale is inside the enclosure, the above-mentioned position updating method inside the enclosure is adopted, and the random search updating formula is as follows:

[0110]

[0111] In the formula: is the position vector of a random whale in the group.

[0112] Fitness function:

[0113] u of k IMF components k (t) kurtosis value, the maximum value of the obtained kurtosis value As the fitness function of the WOA algorithm, kurtosis (K) is a numerical statistic reflecting the vibration signal, and the specific calculation formula is as follows:

[0114]

[0115] Where, u k () is the instantaneous amplitude, is the amplitude mean, p(x) is the probability density, and σ is the standard deviation.

[0116] Step 4: Decompose the vibration signal by selecting the corresponding k and α in step 3, use the algorithm of fuzzy entropy to calculate the fuzzy entropy value of k IMF respectively, and form the fuzzy entropy feature vector Y, Y = [y1... y k ], Where y1... y k is the fuzzy entropy value of k IMF components, and the specific steps are as follows:

[0117] Determine the signal sampling frequency as N, define the phase space dimension m (m ≤ N-2), and reconstruct the phase space X(i) = [u k (i), u k (i+1),..., u k (i+m-1)]-u0(i), i = 1, 2,..., N-m+1, u0(i) is the mean,

[0118]

[0119] , and j≠i is the maximum interval between window vectors X(i) and X(j).

[0120] Introduce fuzzy membership function

[0121]

[0122] Define the correlation between X(i) and X(j):

[0123] And j≠i;

[0124] Where, is the exponential function; n is the boundary gradient of the function; r is the boundary width.

[0125] For each i, find its average value, which is

[0126] Define the fuzzy degree similarity function as

[0127] FuzzyEn(m, n, r, N) = lnΦ m (r) - lnΦ m+1 (r);

[0128] The fuzzy entropy estimation value is y i = FuzzyEn(m, n, r, N) = lnΦ m (r) - lnΦ m+1 (r);

[0129] Wherein, m is the reconstruction dimension; r is generally (0.1-0.25)S d ,S d is the standard deviation of the sampling signal. The fuzzy entropy size is closely related to the value of the parameter, and for better estimation of the conditional probability, m = n = 2 can be generally taken.

[0130] Step 5: Calculate the variance of the k IMF components u k (t) obtained in step 4, the energy moment, the ratio of the variance of the first-order difference signal to the variance of the original signal, and the ratio of the mobility of the first-order derivative to the mobility of the original signal, and construct a parameter matrix H based on the above:

[0131]

[0132] Wherein, A i , B i , C i , D i are the variance, energy moment, ratio of the variance of the first-order difference signal to the variance of the original signal, and ratio of the mobility of the first-order derivative to the mobility of the original signal of the i-th IMF component.

[0133] The variance of the signal represents the amplitude characteristics of the signal, and the formula is: A i = σ i 2

[0134] Wherein, σ i is the standard deviation of the i-th IMF component u i (t).

[0135] The energy moment of the signal is: B i = ∫t|u i (t)| 2 dt

[0136] The ratio of the variance of the first-order difference signal to the variance of the original signal reflects the mobility of the signal, and the specific formula is:

[0137] Wherein, σ iThe ratio of the variance of the first-order differential signal of the i-th IMF component to the variance of the original signal.

[0138] The ratio of the mobility of the first-order derivative of the signal to the mobility of the original signal, which indicates the similarity of each IMF component signal to a pure sine signal, is calculated according to the following formula:

[0139]

[0140] In the formula, σ i The standard deviation of the second-order difference of the i-th IMF component.

[0141] Step 6: Singular value decomposition is performed on the parameter matrix in step 5, and the decomposed singular value vector is extracted. The specific steps of singular value decomposition are as follows:

[0142]

[0143] In the formula, ∑=[diag(σ1,σ2,…σ q ) and 0] is a diagonal matrix composed of singular values, and U and V T are eigenvectors obtained by singular value decomposition, where q=min(k,4), and ∑ is converted into an eigenvector ∑ V that can represent H. The fuzzy entropy eigenvector and the singular value vector are combined to form a final feature representing the sampling signal, and the mixed eigenvector is F:

[0144] F=[Y ∑ V ]

[0145] Step 7: The eigenvector set after dimensionality reduction in step 6 is divided into training samples and test samples. The test samples are used as input vectors of a support vector machine (SVM), and the SVM is trained to obtain a corresponding SVM fault classifier. The trained support vector machine is used to diagnose the fault of the low-voltage intelligent circuit breaker, and the fault recognition of the low-voltage intelligent circuit breaker is realized.

[0146] A CNN network pruning rate automatic search method and system based on reinforcement learning

[0147] The support vector machine (SVM) is developed from linear separability, and the purpose is to obtain the optimal hyperplane in the support vector machine. It has strong classification and prediction ability, and is more suitable for small sample classification than artificial neural network. Moreover, the SVM has strong learning and generalization ability, and the basic idea is as shown in Figure 3 , wherein the origin and square point represent two different samples, and the ωx+b=0 in the middle is a class boundary, wherein ω is a hyperplane normal vector, and b is a threshold. The goal is to separate the two samples. When the separation interval is maximum, it is considered that the obtained hyperplane is optimal.

[0148] In this step, the feature vector obtained in step 6 is taken as a training and test set, the training set is used to train the model, and a classification model capable of identifying low-voltage intelligent circuit breaker faults is obtained, and the test set is used to determine the accuracy of the trained system.

[0149] The application provides a hybrid feature fault diagnosis system based on a low-voltage intelligent circuit breaker, as shown in the figure, comprising a fuzzy entropy feature vector acquisition module, a parameter matrix construction module and a hybrid feature vector acquisition module. Figure 4

[0150] The fuzzy entropy feature vector acquisition module is used to acquire the denoised vibration signal, decompose the vibration signal to obtain the optimal decomposition parameter of the vibration signal, acquire the fuzzy entropy value according to the optimal decomposition parameter of the vibration signal, and form a fuzzy entropy feature vector.

[0151] The parameter matrix construction module is used to construct a parameter matrix according to the variance of the optimal decomposition parameter of the vibration signal, the energy moment, the ratio of the variance of the first-order difference signal to the variance of the original signal, and the ratio of the mobility of the first-order derivative to the mobility of the original signal.

[0152] The hybrid feature vector acquisition module is used to singular value decompose the parameter matrix, acquire the decomposed feature vector, acquire the hybrid feature vector according to the decomposed feature vector and the fuzzy entropy feature vector, and input the hybrid feature vector into a support vector machine to acquire the fault type.

[0153] An embodiment of the application provides a terminal device, the terminal device comprising a processor, a memory, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the steps in each of the method embodiments. Alternatively, the processor executes the computer program to implement the functions of each module / unit in each of the device embodiments.

[0154] The computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the application.

[0155] The terminal device can be a desktop computer, a notebook computer, a palm computer, a cloud server and other computing devices. The terminal device can include, but is not limited to, a processor and a memory.

[0156] ​The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic, discrete hardware components, etc.

[0157] The memory can be used to store the computer programs and / or modules, and the processor realizes various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory.

[0158] The modules / units integrated in the terminal device, if realized in the form of software function units and sold or used as independent products, can be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. When the processor executes the computer program, the steps of the above-mentioned various method embodiments can be realized. The computer program includes computer program code, which can be in the form of source code, object code, executable file or some intermediate form, etc. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the computer readable medium can include or exclude contents according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electric carrier signal and telecommunication signal.

[0159] Therefore, the application provides a mixed feature fault diagnosis method based on a low-voltage intelligent circuit breaker, which combines multiple feature quantities to diagnose the low-voltage circuit breaker, adopts fuzzy entropy, variance, energy moment, the ratio of the variance of the first-order differential signal to the variance of the original signal, and the ratio of the mobility of the first-order derivative to the mobility of the original signal to extract fault features, and adopts singular value decomposition to extract features from the parameter matrix composed of the variance, the energy moment, the ratio of the variance of the first-order differential signal to the variance of the original signal, and the ratio of the mobility of the first-order derivative to the mobility of the original signal, thereby avoiding feature redundancy and greatly improving the fault diagnosis accuracy compared with single feature detection.

[0160] The above merely illustrates the preferred embodiments of the application and is not intended to limit the application. The application can be variously modified and changed for those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A hybrid feature fault diagnosis method based on low-voltage intelligent circuit breakers, characterized in that, Includes the following steps: The noise-reduced vibration signal is obtained, and the vibration signal is decomposed to obtain the optimal decomposition parameters. The fuzzy entropy value is obtained based on the optimal decomposition parameters of the vibration signal, and a fuzzy entropy feature vector is constructed. A parameter matrix is ​​constructed based on the variance, energy moment, ratio of the variance of the first-order difference signal to the variance of the original signal, and ratio of the mobility of the first derivative to the mobility of the original signal of the optimal decomposition parameters of the vibration signal. Singular value decomposition is performed on the parameter matrix to obtain the decomposed eigenvectors. A mixed eigenvector is then obtained based on the decomposed eigenvectors and the fuzzy entropy eigenvectors. This mixed eigenvector is input into a support vector machine to obtain the fault type. The optimal decomposition parameter k for the vibration signal is obtained as follows: The noise-reduced vibration signal u w The problem (t) is decomposed into K eigenmode functions with finite bandwidths, and the constrained variational problem is described as follows: Among them, {u k }={u1, u2…..u k } represents the K modal components obtained from the decomposition; }={ , } represents the center frequency of each modal component. After Hilbert transformation The spectrum, * represents the convolution operation. This refers to the constraint condition, where j is the imaginary unit. For gradient operators; Transform the modal components and their center frequencies to the frequency domain: in, They are respectively Fourier transform, The Lagrange multiplier is used, and then an alternating direction algorithm for the multiplier is employed to iteratively search for the optimal solution to the problem, resulting in K iterative updates. The real part obtained by performing an inverse Fourier transform on it is the modal component. , which is the optimal decomposition parameter k of the vibration signal; The method for performing singular value decomposition on the parameter matrix to obtain the decomposed eigenvectors is as follows: in, A diagonal matrix composed of singular values. These are the eigenvectors obtained from singular value decomposition. A diagonal matrix composed of singular values Transform into an eigenvector that can represent H The fuzzy entropy feature vector and the singular value vector are combined to form the feature of the final response sample signal.

2. The hybrid feature fault diagnosis method based on low-voltage intelligent circuit breakers according to claim 1, characterized in that, The method for obtaining the noise-reduced vibration signal is as follows: Choosing wavelet basis functions as the wavelet basis for discrete wavelet transforms, the minimum frequency f of the vibration signal is used. min The ratio of the wavelet basis function's center frequency f0 to the maximum decomposition level j determines the maximum number of decomposition layers for the vibration signal x. w (t) Perform j-level discrete wavelet decomposition, determine the high-frequency coefficient thresholds for each level of wavelet based on the unbiased likelihood estimation principle, and reconstruct the vibration signal using wavelet based on the low-frequency coefficients of the wavelet decomposition and the high-frequency coefficients after applying the thresholds, to obtain the denoised vibration signal u. w (t).

3. The hybrid feature fault diagnosis method based on low-voltage intelligent circuit breakers according to claim 1, characterized in that, The method for obtaining fuzzy entropy values ​​and constructing fuzzy entropy feature vectors is as follows: The signal sampling frequency is determined to be N, and the phase space dimension is defined. Reconstructing phase space , The mean, ; in, To reconstruct phase space and The maximum interval between; Fuzzy membership function Reconstructing phase space and The correlation between them is as follows: and , Let n be an exponential function, n be the boundary gradient of the function, and r be the boundary width; For each i, we can calculate its average value to obtain... The ambiguity similarity function is: , The fuzzy entropy of the original vibration signal is Where r is (0.1~0.25) This represents the standard deviation of the sampled signal.

4. The hybrid feature fault diagnosis method based on low-voltage intelligent circuit breakers according to claim 3, characterized in that, The method for constructing the parameter matrix based on the variance, energy moment, ratio of the variance of the first-order difference signal to the variance of the original signal, and ratio of the mobility of the first derivative to the mobility of the original signal of the optimal decomposition parameters of the vibration signal is as follows: in, These are the variance, energy moment, ratio of the variance of the first-order differential signal to the variance of the original signal, and ratio of the mobility of the first derivative to the mobility of the original signal, respectively, for the i-th IMF component.

5. The hybrid feature fault diagnosis method based on low-voltage intelligent circuit breakers according to claim 1, characterized in that, The method for obtaining the mixed feature vector based on the decomposed feature vector and the fuzzy entropy feature vector is as follows: The mixed feature vector is... , This is the fuzzy entropy feature vector.

6. A hybrid feature fault diagnosis system based on a low-voltage intelligent circuit breaker, characterized in that, The method described by any one of claims 1 to 5 includes: The fuzzy entropy feature vector acquisition module is used to acquire the noise-reduced vibration signal, decompose the vibration signal to obtain the optimal decomposition parameters of the vibration signal, and obtain the fuzzy entropy value based on the optimal decomposition parameters of the vibration signal to form a fuzzy entropy feature vector. The parameter matrix construction module is used to construct a parameter matrix based on the variance of the optimal decomposition parameters of the vibration signal, the energy moment, the ratio of the variance of the first-order difference signal to the variance of the original signal, and the ratio of the mobility of the first derivative to the mobility of the original signal. The hybrid feature vector acquisition module is used to perform singular value decomposition on the parameter matrix, obtain the decomposed feature vector, obtain a hybrid feature vector based on the decomposed feature vector and the fuzzy entropy feature vector, and input the hybrid feature vector into the support vector machine to obtain the fault type.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the hybrid feature fault diagnosis method based on low-voltage intelligent circuit breakers as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the hybrid feature fault diagnosis method based on low-voltage intelligent circuit breakers as described in any one of claims 1 to 5.

Images