High-voltage circuit breaker motor driving fault diagnosis method and system based on wavelet neural network

Abstract

The invention discloses a high-voltage circuit breaker motor driving fault diagnosis method and system based on a wavelet neural network, and relates to the technical field of high-voltage switch equipment, and the method comprises the steps: collecting the operation data of a motor driving mechanism as an initial signal of fault diagnosis; creating a wavelet neural network model of the motor driving mechanism model of the high-voltage circuit breaker, and removing high-frequency components in the initial signal by using a wavelet decomposition algorithm in the wavelet neural network model; taking the signal after the wavelet decomposition processing as input, and taking a fault feature vector in a motor driving mechanism of the high-voltage circuit breaker as output; and diagnosing gradual change faults of the high-voltage circuit breaker motor driving mechanism through the wavelet neural network model of the high-voltage circuit breaker motor driving mechanism model. According to the method, the calculation complexity of fault diagnosis by using the wavelet neural network model is reduced, and the gradual change fault of the motor driving mechanism component of the high-voltage circuit breaker is accurately diagnosed.

Description

Technical Field

[0001] The invention relates to the technical field of high-voltage switchgear, and in particular to a method and system for diagnosing motor drive faults of high-voltage circuit breaker based on a wavelet neural network. Background Art

[0002] Failure of the motor drive mechanism of a high-voltage circuit breaker will not only affect the accurate opening and closing of the moving contacts of the high-voltage circuit breaker, but may also cause the high-voltage circuit breaker to trip. The diagnosis of gradual faults in the motor drive mechanism of a high-voltage circuit breaker should accurately predict the evolution trend of component failures in the early stage of system component failure, and provide early warning and maintenance signals to operators and maintenance personnel in a timely manner. The difficulty in diagnosing gradual faults is that the fault signals of gradual faults such as bearing wear change very slowly in the early stage of degradation, and only change dramatically when close to failure. Therefore, the training data obtained in the early stage of the fault also changes slowly, and it is difficult to accurately estimate how long it will take for a gradual fault to fail.

[0003] At present, the gradual fault diagnosis methods of electromechanical systems are generally divided into two categories: analytical model-based and artificial intelligence-based diagnostic methods. Model-based fault diagnosis methods rely on obtaining an accurate system dynamic model. Among them, the recursive least squares method is often used to identify the fault characteristic parameters in the model. The disadvantage of this method is that the fault parameters do not change significantly in the initial stage of gradual faults, and the prediction accuracy is poor. Artificial intelligence-based fault diagnosis methods do not require accurate mathematical models, but require a complete database to train and verify diagnostic algorithms. Relatively speaking, data-based fault diagnosis methods have self-learning, self-organization and self-reasoning capabilities, and are more flexible to use.

[0004] To this end, this paper proposes a method for diagnosing gradual faults of high-voltage circuit breaker motor drive mechanisms based on wavelet neural networks. The wavelet multi-level decomposition algorithm is used to select the frequency components of the sensor receiving signal, and the feedback mechanism of the dynamic neural network is used to predict the evolution trend of the fault characteristic parameters. Compared with the recursive least squares method based on the analytical model, the proposed wavelet neural network-based method uses the wavelet decomposition algorithm to remove the influence of high-frequency components in the sensor measurement signal, which is convenient for estimating the change trend of fault characteristic parameters and pre-diagnosing gradual faults. The results show that this method can diagnose the mechanical and electrical gradual faults existing in the motor drive mechanism of high-voltage circuit breaker, and improve the safety and reliability of the operation of the motor drive mechanism of high-voltage circuit breaker. Summary of the invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the problem to be solved by the present invention is: how to solve the problem that in the initial stage of gradual faults in the existing methods, the fault parameters do not change obviously, the prediction accuracy is poor, and a complete database is needed to train and verify the diagnosis algorithm.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: a high-voltage circuit breaker motor drive fault diagnosis method based on wavelet neural network, comprising: collecting motor drive mechanism operation data as an initial signal for fault diagnosis; creating a wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model, and removing high-frequency components in the initial signal by using a wavelet decomposition algorithm in the wavelet neural network model; using the signal processed by wavelet decomposition as input, and using the fault feature vector in the high-voltage circuit breaker motor drive mechanism as output; and diagnosing the gradual fault of the high-voltage circuit breaker motor drive mechanism through the wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model.

[0008] As a preferred solution of the high-voltage circuit breaker motor drive fault diagnosis method of the wavelet neural network of the present invention, the initial signal includes the three-phase stator winding current of the servo motor and the speed and angular displacement of the servo motor in the motor drive mechanism.

[0009] As a preferred solution of the high-voltage circuit breaker motor drive fault diagnosis method of the wavelet neural network of the present invention, wherein: the high-voltage circuit breaker motor drive mechanism model includes a differential equation expressed as:

[0010]

[0011] Where x(k)=[i a (k)i b (k)i c (k)ω m (k)θ m (k)] T is the state vector at the kth moment, which contains the three-phase stator winding current i of the servo motor in the motor drive mechanism a (k), i b (k) and i c (k) and the speed ω of the servo motor in the motor drive mechanism m (k) and the angular displacement θ of the servo motor m (k); u(k)=[u a (k) u b (k) u c (k) T L (k)] T Represents the input vector, which contains the three-phase stator winding voltage u of the servo motor in the motor drive mechanism a (k),u b (k),uc (k) and servo load torque T L (k), the observation matrix C is the unit matrix, A and B are the state matrix and the control matrix respectively, expressed as:

[0012]

[0013]

[0014] Among them, R a is the resistance of the stator winding, L a is the inductance of the stator winding, B m is the damping coefficient, J m is the moment of inertia of the motor, T L is the load torque, k Ea (θ m ),k Eb (θ m ),k Ec (θ m ) are the three-phase back electromotive force coefficients of the motor, η a ,η b ,η c are the fault coefficients of the three-phase windings of the motor respectively.

[0015] As a preferred solution of the wavelet neural network high-voltage circuit breaker motor drive fault diagnosis method of the present invention, the wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model is expressed as:

[0016]

[0017] Among them, C is the synaptic weight matrix representing the output layer characteristics, is the wavelet function describing the hidden layer neurons, where the perceptron neuron ψ in any hidden layer k (·)(k=1,2,…,q) is expressed as:

[0018]

[0019] Among them, w a,kj is the synaptic weight of the hidden layer feedback signal x(j), w b,ki is the synaptic weight of the input signal u(i) of the input layer.

[0020] As a preferred solution of the wavelet neural network high-voltage circuit breaker motor drive fault diagnosis method of the present invention, the wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model includes a description of hidden layer neurons using memory equations, and the wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model is rewritten as:

[0021]

[0022] The state vector equation is expressed as:

[0023]

[0024] Among them, the synaptic weight vector is a (q+m)×1 dimensional vector, ξ k Defined as ξ k =[x(k)u(k)] T , is a (q+m)×1 state vector.

[0025] As a preferred solution of the wavelet neural network high-voltage circuit breaker motor drive fault diagnosis method of the present invention, the wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model also includes a real-time recursive learning process represented as:

[0026] Λ j,k+1 =Φ k (W a,k Λ j,k +U j,k+1 ),j=1,2,…,q

[0027] Among them, Λ j,k is the state vector x(k) with respect to the weight w j The partial derivative matrix of is expressed as:

[0028]

[0029] U j,k is a q×(q+m+1) matrix, with the jth row equal to the vector ξ k , the other rows are all 0, which means:

[0030]

[0031] Φ k It is a q×q diagonal matrix, and the jth diagonal element is the partial derivative of the neuron function with respect to the independent variable, expressed as:

[0032]

[0033] Let vector d(k) be the expected output of the neural network and define the error vector, expressed as:

[0034] e(k)=d(k)-y(k)=d(k)-Cx(k)

[0035] The goal of the learning process is to minimize the cost function obtained from all time squared errors, expressed as:

[0036]

[0037] Cost function E(k) for weight vector w j Differentiation, expressed as:

[0038]

[0039] If the gradient descent iterative method is used, the synaptic weight vector w j,k The adjustment amount is expressed as:

[0040]

[0041] Where η is the learning rate parameter, Δw j,k is the synaptic weight vector w j,k The amount of adjustment.

[0042] As a preferred solution of the high-voltage circuit breaker motor drive fault diagnosis method of the wavelet neural network of the present invention, the output mainly includes two parts, one of which is the state quantity of the model expressed as:

[0043] y(k)=x(k)=[i a (k)i b (k)i c (k)ω m (k)θ m (k)] T

[0044] Among them, y(k) is used to estimate the weight w of the neural network k ; The other part is the fault feature vector expressed as:

[0045]

[0046] in, is the fault feature vector, It is directly used to infer the fault of the motor drive mechanism of the high-voltage circuit breaker; the fault characteristic vector is used for fault diagnosis reasoning rules: if the position of the circuit breaker moving contact remains unchanged, but the output instruction of the motor drive mechanism changes, it is inferred that the circuit breaker moving contact is stuck; if the position of the circuit breaker moving contact deviates, but the output instruction of the motor drive mechanism is constant, it is inferred that the circuit breaker moving contact position sensor is drifting; if the impedance of the circuit breaker servo motor winding decreases, it is inferred that the circuit breaker servo motor winding parts are short-circuited; if the damping coefficient of the circuit breaker servo system increases, it is inferred that the circuit breaker motor drive mechanism is severely worn.

[0047] Another object of the present invention is to provide a system of a high-voltage circuit breaker motor drive fault diagnosis method based on a wavelet neural network, which can solve the problem of high-voltage circuit breaker motor drive fault diagnosis based on a wavelet neural network by constructing a high-voltage circuit breaker motor drive fault diagnosis system based on a wavelet neural network.

[0048] To solve the above technical problems, the present invention provides the following technical solutions: a high-voltage circuit breaker motor drive fault diagnosis system based on wavelet neural network, comprising a data acquisition module, a model creation module, an input-output module and a diagnosis module; the data acquisition module is used to collect motor drive mechanism operation data as an initial signal for fault diagnosis; the model creation module is used to create a wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model, by using a wavelet decomposition algorithm in the wavelet neural network model to remove high-frequency components in the initial signal; the input-output module is used to take the signal processed by wavelet decomposition as input, and take the fault feature vector in the high-voltage circuit breaker motor drive mechanism as output; the diagnosis module is used to diagnose the gradual fault of the high-voltage circuit breaker motor drive mechanism through the wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model.

[0049] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the high-voltage circuit breaker motor drive fault diagnosis method based on the wavelet neural network are implemented.

[0050] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the high-voltage circuit breaker motor drive fault diagnosis method based on the wavelet neural network as described above.

[0051] The beneficial effects of the present invention are as follows: the high-voltage circuit breaker motor drive fault diagnosis method of the wavelet neural network provided by the present invention can use the decoupled unscented Kalman filtering algorithm to estimate the state of synaptic weights to decouple the wavelet neural network model, reduce the computational complexity of fault diagnosis using the wavelet neural network model, and accurately diagnose gradual faults of high-voltage circuit breaker motor drive mechanism components. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0053] Figure 1 This is a flow chart of a high-voltage circuit breaker motor drive fault diagnosis method based on a wavelet neural network provided in the first embodiment of the present invention.

[0054] Figure 2 This is a wavelet neural network model framework diagram of a high-voltage circuit breaker motor drive fault diagnosis method based on a wavelet neural network provided in the first embodiment of the present invention.

[0055] Figure 3 This is a multi-frequency component wavelet neural network prediction algorithm framework diagram of the high-voltage circuit breaker motor drive fault diagnosis method based on the wavelet neural network provided by the first embodiment of the present invention.

[0056] Figure 4 This is a structural diagram of a high-voltage circuit breaker motor drive fault diagnosis system based on a wavelet neural network provided in the second embodiment of the present invention.

[0057] Figure 5 This is a diagram of fault diagnosis results of a high-voltage circuit breaker motor drive fault diagnosis method based on a wavelet neural network provided in the third embodiment of the present invention. DETAILED DESCRIPTION

[0058] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0059] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0060] Example 1, reference Figure 1 to Figure 3 , which is the first embodiment of the present invention, and provides a high-voltage circuit breaker motor drive fault diagnosis method based on a wavelet neural network, including: collecting motor drive mechanism operation data as an initial signal for fault diagnosis; creating a wavelet neural network model of a high-voltage circuit breaker motor drive mechanism model, and removing high-frequency components in the initial signal by using a wavelet decomposition algorithm in the wavelet neural network model; using the signal processed by the wavelet decomposition as input, and using the fault feature vector in the high-voltage circuit breaker motor drive mechanism as output; and diagnosing the gradual fault of the high-voltage circuit breaker motor drive mechanism through the wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model.

[0061] A fault diagnosis method for a high-voltage circuit breaker motor drive mechanism based on a wavelet neural network accurately predicts the evolution trend of component failures in the early stage of system component failure, provides early warning and maintenance signals to operators and maintenance personnel in a timely manner, and improves the safety and reliability of the operation of the high-voltage circuit breaker motor drive mechanism.

[0062] S1. Collect the motor drive mechanism operation data as the initial signal for fault diagnosis.

[0063] The operating data of the high-voltage circuit breaker motor drive mechanism under various gradual fault conditions such as motor armature winding inter-turn short circuit and transmission device wear are used to train the wavelet neural network fault diagnosis model, and then the trained wavelet neural network model is used to diagnose the gradual faults of the high-voltage circuit breaker motor drive mechanism.

[0064] S2. Create a wavelet neural network model for the high-voltage circuit breaker motor drive mechanism model, and use the wavelet decomposition algorithm in the wavelet neural network model to remove the high-frequency components in the initial signal.

[0065] S3. The signal processed by wavelet decomposition is used as input, and the fault feature vector in the motor drive mechanism of the high-voltage circuit breaker is used as output.

[0066] The following differential equations are used to establish the high-voltage circuit breaker motor drive mechanism model:

[0067]

[0068] Where x(k)=[i a (k)i b (k)i c (k)ω m (k)θ m (k)] T is the state vector at the kth moment, which contains the three-phase stator winding current i of the servo motor in the motor drive mechanism a (k), i b (k) and i c (k), and the speed ω of the servo motor in the motor drive mechanism m (k) and the angular displacement θ of the servo motor m (k); u(k)=[u a (k) u b (k) u c (k) T L (k)] T represents the input vector, which contains the three-phase stator winding voltage u of the servo motor in the motor drive mechanism a (k),u b (k),u c (k) and servo load torque T L (k); The observation matrix C is the unit matrix, A and B represent the state matrix and the control matrix, respectively, which are described as:

[0069]

[0070]

[0071] Among them, R a is the resistance of the stator winding, L a is the inductance of the stator winding, B m is the damping coefficient; J m is the moment of inertia of the motor; T L is the load torque, k Ea (θ m ),k Eb (θ m ),k Ec (θ m ) are the three-phase back electromotive force coefficients of the motor, η a ,η b ,η c are the fault coefficients of the three-phase windings of the motor respectively.

[0072] Designing a wavelet neural network model to estimate the variation characteristics of parameters in the high-voltage circuit breaker motor drive mechanism is a very critical content. The wavelet neural network model used to approximate the high-voltage circuit breaker motor drive mechanism model in the proposed method is as follows:

[0073]

[0074] Among them, C is the synaptic weight matrix representing the output layer characteristics, is the wavelet function describing the hidden layer neurons, where the perceptron neuron ψ in any hidden layer k (·)(k=1,2,…,q) is defined as Among them, w a,kj represents the synaptic weight x(j), w of the hidden layer feedback signal b,kj Represents the synaptic weight of the input signal u(i) in the input layer.

[0075] Using the memory equation to describe the hidden layer neurons, the wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model can be rewritten as:

[0076]

[0077] The wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model has the following structure: Figure 2 shown.

[0078] The state vector equation can be simplified as:

[0079]

[0080] Among them, the synaptic weight vector w j = "w a,j w b,j T(j=1,2,…,q), is a (q+m)×1-dimensional vector; ξ k Defined as ξ k =[x(k)u(k)] T , is a (q+m)×1 state vector.

[0081] Let vector d(k) be the expected output of the neural network and define the error vector as:

[0082] e(k)=d(k)-y(k)=d(k)-Cx(k)

[0083] According to the output equation y(k)=Cx(k), we can obtain e(k)=d(k)-Cx(k), where C is the synaptic weight matrix representing the characteristics of the output layer.

[0084] The goal of the learning process is to minimize the cost function obtained by all time squared error instants:

[0085]

[0086] Cost function E(k) for weight vector w j Differentiating, we get:

[0087]

[0088] Among them, Λ j,k Defined as the state vector x(k) with respect to the weight w j The partial derivative matrix of is expressed as:

[0089]

[0090] The real-time recursive learning process is expressed as:

[0091] Λ j,k+1 =Φ k (W a,k Λ j,k +U j,k+1 ),j=1,2,…,q

[0092] Among them, U j,k It is defined as a q×(q+m+1) matrix, except that the jth row is equal to the vector ξ k Except for , all other rows are 0, which is represented as:

[0093]

[0094] Φ k It is defined as a q×q diagonal matrix, expressed as:

[0095]

[0096] Right now:

[0097]

[0098] If an iterative method such as gradient descent is used, the synaptic weight vector w j,k The adjustment amount is:

[0099]

[0100] Among them, η is the learning rate parameter, Δw j,k is the synaptic weight vector w j,k The amount of adjustment.

[0101] Wavelet neural network training is essentially a state estimation problem for state equations and measurement equations. As state variables, and the input and sensor measurements As a sample for training. Its state equation and measurement equation are described as follows:

[0102]

[0103] Where h(·) is the input-output relationship function of the neural network; ε is the process noise vector with zero mean, and its covariance matrix is ​​cov(ε)=E(εε T )=Q; v is the measurement noise vector with zero mean, and its covariance matrix is ​​cov(v)=E(vv T )=R.

[0104] Given a training sample The Kalman filter is used to train the weight parameters of the wavelet neural network. In order to use the Kalman filter to predict the weight state of the wavelet neural network, it is necessary to linearize the nonlinear parts in the above state equation and measurement equation. The first-order term of the Taylar expansion is used to linearize the measurement equation, and the result is:

[0105]

[0106] Among them, H is a measurement matrix of p×s dimensions, and its linearization process is expressed as:

[0107]

[0108] A multi-frequency component wavelet neural network diagnosis algorithm framework for the fault characteristic parameter prediction of the motor drive mechanism of a high-voltage circuit breaker is designed. First, the frequency is divided into multiple frequency components, and each frequency component has an independent spectrum amplitude mode; then the weight estimation processing is performed on the divided frequency component signal, and finally the product of the above prediction weight and the wavelet neuron output is superimposed, because these components are implemented based on the wavelet orthogonal theory, which conforms to the orthogonal characteristics. The specific algorithm framework can be found in Figure 3 .

[0109] According to the attached Figure 3 It can be seen that the output of the wavelet neural network consists of two parts: one is the state quantity y(k)=x(k)=[i a (k)i b (k)i c (k)ω m (k)θ m (k)] T , and the other part is the fault feature vector y(k) is used to estimate the weight w of the neural network k . Fault feature vector It is directly used to infer the faults in the motor drive mechanism of the high-voltage circuit breaker. Its failure modes include mechanical vibration, jamming, blocking and other events, as well as electrical and electronic failure events, such as short circuit, sensor drift, device failure, etc. The fault diagnosis model based on wavelet neural network follows a recursive method and can map historical and current data to future data.

[0110] A decoupled unscented Kalman filter is used to track the state of signals of different frequency components in the model. The unscented Kalman filter algorithm ignores the coupling between the values ​​of the neural network and introduces multiple zero values ​​into the covariance matrix Pk|k, reducing the computational complexity of the filter algorithm. In addition, the unscented Kalman filter algorithm is used to process. A set of Sigma points are generated near the mean of the frequency component signal, and then the variance prediction estimate is generated by weighting. In the filtering process, this paper adopts the method of proportional correction of Sigma points to convert the n-dimensional random vector with variance Pk|k+1 Expanded to 2n+1 column vectors:

[0111]

[0112] in, is the optimal state estimate of the variable w at time k. The specific meaning of k|k-1 is that the k before the symbol “|” refers to time k, and the k-1 after the symbol “|” refers to the optimal estimate of the system state w based on all measurement data and system dynamic information up to time k-1; P k|k-1 is the optimal covariance matrix at time k; parameter is an arbitrarily small number, and in the present invention, parameter κ=0; (·) i It means taking the i-th column element of the matrix (·).

[0113] Decoupled Unscented Kalman Filter measurement update. Generate the corresponding Sigma point of the measurement equation and calculate the prediction estimate:

[0114]

[0115] Among them, Z i is the corresponding Sigma point of the measurement equation, W i Sigma point χ i The weighting factor of is obtained by the following formula:

[0116]

[0117] Calculate the Kalman gain K k :

[0118]

[0119] Among them, R k is the measurement noise covariance matrix at time k; P zz Z i and The covariance matrix between xz is i and The covariance matrix between them; K is the Kalman gain.

[0120] Decoupled unscented Kalman filter estimate update:

[0121]

[0122] Error covariance update:

[0123]

[0124] Among them, Q k is the process noise covariance matrix at time k.

[0125] S4. Diagnose the gradual fault of the high-voltage circuit breaker motor drive mechanism through the wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model.

[0126] The input and output data are processed using the decoupled unscented Kalman filter algorithm, the neural network is trained offline, and then the new measurement value is predicted online. For the decoupled unscented Kalman filter algorithm, the output data of the input signal component of the wavelet neural network is normalized data. Because the prediction of these components is based on wavelet theory, it satisfies the orthogonal characteristics. The prediction variance estimates of each component are superimposed to obtain the prediction variance of the original signal.

[0127] The fault diagnosis reasoning rule based on the fault feature vector is:

[0128] If the position of the circuit breaker moving contact remains unchanged, but the output command of the motor drive mechanism changes, it is inferred that the circuit breaker moving contact is stuck.

[0129] If there is a deviation in the position of the moving contact of the circuit breaker, and the output command of the motor drive mechanism is constant, it is inferred that the position sensor of the moving contact of the circuit breaker is drifting.

[0130] If the impedance of the circuit breaker servo motor winding becomes smaller, it is inferred that the circuit breaker servo motor winding parts are short-circuited.

[0131] If the damping coefficient of the circuit breaker servo system becomes larger, it is inferred that the circuit breaker motor drive mechanism is severely worn.

[0132] Example 2, reference Figure 4 , which is the second embodiment of the present invention, is different from the previous embodiment in that it provides a high-voltage circuit breaker motor drive fault diagnosis system based on a wavelet neural network, including: a data acquisition module 100, a model creation module 200, an input and output module 300 and a diagnosis module 400.

[0133] The data acquisition module 100 is used to collect the operation data of the motor drive mechanism as an initial signal for fault diagnosis.

[0134] The model creation module 200 is used to create a wavelet neural network model of a high-voltage circuit breaker motor drive mechanism model by using a wavelet decomposition algorithm in the wavelet neural network model to remove high-frequency components in the initial signal.

[0135] The input-output module 300 is used to take the signal processed by wavelet decomposition as input, and take the fault feature vector in the motor drive mechanism of the high-voltage circuit breaker as output.

[0136] The diagnosis module 400 is used to diagnose the gradual fault of the high-voltage circuit breaker motor drive mechanism by using the wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model.

[0137] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. 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 methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0138] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0139] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0140] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0141] Example 3, reference Figure 5 , which is the third embodiment of the present invention, and which is different from the first two embodiments in that it is used to verify and illustrate the technical effects adopted in the present invention, so as to verify the real effects of this method.

[0142] 1) Establish a wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model, whose structure is as follows Figure 2 As shown:

[0143]

[0144] The state vector equation can be simplified as Among them, the synaptic weight vector is a (q+m)×1-dimensional vector; ξ k Defined as ξ k =[x(k)u(k)] T , is a (q+m)×1 state vector.

[0145] 2) Given a training sample The Kalman filter is used to train the weight parameters of the wavelet neural network:

[0146]

[0147] Among them, H is a measurement matrix of p×s dimensions, and its linearization process is expressed as:

[0148]

[0149] 3) Select fault feature vector:

[0150]

[0151] 4) Using the proportional correction Sigma point method, the n-dimensional random vector with variance Pk|k+1 is Expanded to 2n+1 column vectors:

[0152]

[0153] in, parameter is an arbitrarily small number, and in the present invention, parameter κ=0; (·) i It means taking the i-th column element of the matrix (·).

[0154] 5) Decoupled unscented Kalman filter measurement update. Generate the corresponding Sigma point of the measurement equation and calculate the predicted estimate:

[0155]

[0156] 6) Calculate the Kalman gain K k :

[0157]

[0158] Among them, Z i is the corresponding Sigma point of the measurement equation, W i Sigma point χ i The weighting factor of is obtained by the following formula:

[0159]

[0160] 7) Decoupled unscented Kalman filter estimate update:

[0161]

[0162] 8) Error covariance update:

[0163]

[0164] If an iterative method such as gradient descent is used, the synaptic weight vector w j,k The adjustment amount is:

[0165]

[0166] The transmission mechanism wear is implanted into the normally operating motor drive mechanism system. During the experiment, the transmission mechanism wear fault diagnosis results are obtained by comparing the measurement data of the normal motor drive mechanism and the faulty motor drive mechanism. The measurement data include current, voltage, speed and displacement, among which the speed is the main fault characteristic data of the transmission mechanism wear fault. The proposed high-voltage circuit breaker motor drive mechanism fault diagnosis method based on wavelet neural network is used for fault diagnosis. The diagnosis results are as follows: Figure 5 shown. Figure 5The data from 30s to 70s are the speed data of the normal operation of the motor drive mechanism, and the data from 70s to 100s are the speed data of the motor drive mechanism after the fault is implanted. In order to verify the fault diagnosis effect of the method proposed in this implementation scheme, the normal motor drive mechanism operation data and the faulty motor drive mechanism operation data are fused together in the experiment. It can be found that the use of the proposed method to track the fault characteristic parameters can accurately determine the change trend of the fault characteristic parameters and realize the diagnosis of EMA gradual faults.

[0167] 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 preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A wavelet neural network-based high-voltage circuit breaker motor drive fault diagnosis method, characterized in that: include, Collect the operation data of the motor drive mechanism as the initial signal for fault diagnosis; A wavelet neural network model of a high-voltage circuit breaker motor drive mechanism model is created, and the high-frequency components of the initial signal are removed by using the wavelet decomposition algorithm in the wavelet neural network model; The signal processed by wavelet decomposition is used as input, and the fault feature vector in the motor drive mechanism of the high-voltage circuit breaker is used as output; The gradual fault of high-voltage circuit breaker motor drive mechanism is diagnosed through the wavelet neural network model of high-voltage circuit breaker motor drive mechanism model.

2. The high-voltage circuit breaker motor drive fault diagnosis method based on wavelet neural network as claimed in claim 1 is characterized in that: The initial signal includes the three-phase stator winding current of the servo motor and the rotation speed and angular displacement of the servo motor in the motor drive mechanism.

3. The high-voltage circuit breaker motor drive fault diagnosis method based on wavelet neural network as claimed in claim 2 is characterized in that: The high-voltage circuit breaker motor drive mechanism model includes a differential equation represented as follows: Where x(k)=[i a (k) i b (k) i c (k) ω m (k) θ m (k)] T is the state vector at the kth moment, which contains the three-phase stator winding current i of the servo motor in the motor drive mechanism a (k), i b (k) and i c (k) and the speed ω of the servo motor in the motor drive mechanism m (k) and the angular displacement θ of the servo motor m (k); u(k)=[u a (k) u b (k) u c (k) T L (k)] T represents the input vector, which contains the three-phase stator winding voltage u of the servo motor in the motor drive mechanism a (k),u b (k),u c (k) and servo load torque T L (k), the observation matrix C is the unit matrix, A and B are the state matrix and control matrix respectively, expressed as, Among them, R a is the resistance of the stator winding, L a is the inductance of the stator winding, B m is the damping coefficient, J m is the moment of inertia of the motor, T L is the load torque, k Ea (θ m ),k Eb (θ m ),k Ec (θ m ) are the three-phase back electromotive force coefficients of the motor, η a ,η b ,η c are the fault coefficients of the three-phase windings of the motor respectively.

4. The high-voltage circuit breaker motor drive fault diagnosis method based on wavelet neural network as claimed in claim 3 is characterized by: The wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model is expressed as: Among them, C is the synaptic weight matrix representing the output layer characteristics, is the wavelet function describing the hidden layer neurons, where the perceptron neuron ψ in any hidden layer k (·)(k=1,2,…,q) is expressed as, Among them, w a,kj is the synaptic weight of the hidden layer feedback signal x(j), w b,ki is the synaptic weight of the input signal u(i) of the input layer.

5. The high-voltage circuit breaker motor drive fault diagnosis method based on wavelet neural network as claimed in claim 4 is characterized in that: The wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model includes a description of hidden layer neurons using memory equations. The wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model is rewritten as: The state vector equation is expressed as, Among them, the synaptic weight vector is a (q+m)×1 dimensional vector, ξ k Defined as ξ k =[x(k) u(k)] T , is a (q+m)×1 state vector.

6. The high-voltage circuit breaker motor drive fault diagnosis method based on wavelet neural network as claimed in claim 5, characterized in that: The wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model also includes a real-time recursive learning process represented as: L j,k+1 =Φ k (W a,k L j,k +U j,k+1 ),j=1,2,…,q Among them, Λ j,k is the state vector x(k) with respect to the weight w j The partial derivative matrix of is expressed as, U j,k is a q×(q+m+1) matrix, with the jth row equal to the vector ξ k , the other rows are all 0, which means, Φ k is a q×q diagonal matrix, and the jth diagonal element is the partial derivative of the neuron function with respect to the independent variable, expressed as, Let vector d(k) be the expected output of the neural network and define the error vector, expressed as, e(k)=d(k)-y(k)=d(k)-Cx(k) The goal of the learning process is to minimize the cost function obtained from all time squared errors, expressed as, Cost function E(k) for weight vector w j Differentiation, expressed as, If the gradient descent iterative method is used, the synaptic weight vector w j,k The adjustment amount is expressed as, Among them, η is the learning rate parameter, Δw j,k is the synaptic weight vector w j,k The amount of adjustment.

7. The high-voltage circuit breaker motor drive fault diagnosis method based on wavelet neural network as claimed in claim 6 is characterized by: The output mainly includes two parts. One part is the state quantity of the model expressed as: y(k)=x(k)=[i a (k) i b (k) i c (k) ω m (k) θ m (k)] T Among them, y(k) is used to estimate the weight w of the neural network k ; The other part is the fault feature vector expressed as, in, is the fault feature vector, It is directly used to infer the fault of the motor drive mechanism of the high-voltage circuit breaker; The fault diagnosis reasoning rule for the fault feature vector is that if the position of the circuit breaker moving contact remains unchanged, but the output command of the motor drive mechanism changes, it is inferred that the circuit breaker moving contact is stuck; If the position of the moving contact of the circuit breaker deviates, but the output command of the motor drive mechanism is constant, it is inferred that the position sensor of the moving contact of the circuit breaker is drifting; If the impedance of the circuit breaker servo motor winding becomes smaller, it is inferred that the circuit breaker servo motor winding parts are short-circuited; If the damping coefficient of the circuit breaker servo system becomes larger, it is inferred that the circuit breaker motor drive mechanism is severely worn.

8. A system for high-voltage circuit breaker motor drive fault diagnosis method using the wavelet neural network according to any one of claims 1 to 7, characterized in that: It includes a data acquisition module (100), a model creation module (200), an input and output module (300) and a diagnosis module (400); The data acquisition module (100) is used to collect the operation data of the motor drive mechanism as an initial signal for fault diagnosis; The model creation module (200) is used to create a wavelet neural network model of a high-voltage circuit breaker motor drive mechanism model by using a wavelet decomposition algorithm in the wavelet neural network model to remove high-frequency components in an initial signal; The input-output module (300) is used to take the signal processed by wavelet decomposition as input, and take the fault feature vector in the motor drive mechanism of the high-voltage circuit breaker as output; The diagnosis module (400) is used to diagnose a gradual fault of a high-voltage circuit breaker motor drive mechanism by using a wavelet neural network model of the high-voltage circuit breaker motor drive mechanism model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the high-voltage circuit breaker motor drive fault diagnosis method based on the wavelet neural network according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the high-voltage circuit breaker motor drive fault diagnosis method based on a wavelet neural network according to any one of claims 1 to 7 are implemented.

Images