A switch cabinet equipment fault prediction method based on a self-organizing fuzzy cerebellar model neural network learning

Abstract

The application provides a switch cabinet equipment fault prediction method based on a self-organizing fuzzy cerebellar model neural network learning, comprising the following steps: S1: performing atlas analysis on local discharge detection data from a switch cabinet to obtain feature data of the atlas, storing the feature data in a two-dimensional matrix, forming M*N feature input, and then removing noise of original data and reducing dimensions of input data by using a self-correlation covariance method; S2: establishing a self-organizing cerebellar neural network model with a three-layer mapping relationship; S3: designing an optimized learning rate of the model by using a Lyapunov function; S4: taking training data as model input, and training the fuzzy cerebellar neural network model; and S5: performing step S1 operation on current local discharge detection data of the switch cabinet, taking obtained data as feature input, inputting the feature input into the model, and predicting a switch cabinet equipment fault; the above method can master a health state of the switch cabinet in real time, and promotes healthy development of the electric power industry.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment detection, and in particular to a switch cabinet equipment fault prediction method based on self-organizing fuzzy cerebellum model neural network learning. Background Art

[0002] Switchgear is a common type of power transmission and distribution equipment, providing control and protection during the power transmission and distribution process. It contains disconnectors, circuit breakers, and related protective devices. When a power system fault occurs, the disconnectors can disconnect the interconnected equipment, protecting both the connected power equipment and the safety of power operators. This demonstrates the crucial role switchgear plays in power systems.

[0003] Switchgear equipment is crucial to the stable operation of the distribution network. Predicting switchgear equipment failures in advance can better assess the safety risks of equipment status and distribution system operation, and formulate reasonable condition-based maintenance strategies in advance.

[0004] To this end, the present invention proposes a switch cabinet equipment fault prediction method based on self-organizing fuzzy cerebellum model neural network learning. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to propose a switch cabinet equipment fault prediction method based on self-organizing fuzzy cerebellum model neural network learning.

[0006] In order to solve the above problems, the present invention adopts the following technical solutions:

[0007] A switchgear equipment fault prediction method based on self-organizing fuzzy cerebellar model neural network learning, characterized by comprising the following steps:

[0008] S1: Spectral analysis is performed on the partial discharge detection data from the switchgear. The characteristic data of the spectrum is stored in a two-dimensional matrix to form an M×N feature input. Then, the autocorrelation covariance method is used to remove the noise of the original data and reduce the dimension of the input data.

[0009] S2: Establish a self-organizing cerebellar neural network model with three-layer mapping relationships;

[0010] S3: Use Lyapunov function to design the optimal learning rate of the model to make the model system stable;

[0011] S4: Using the training data as model input, the fuzzy brain neural network model is trained until the model converges;

[0012] S5: Execute step S1 on the current switchgear partial discharge detection data, use the obtained data as feature input, and input it into the model to predict switchgear equipment failure.

[0013] Furthermore, S1 specifically includes the following steps:

[0014] S1.1: Perform spectrum analysis on the partial discharge detection data of the switchgear to obtain the frequency and voltage relationship data;

[0015] S1.2: The data of the spectrum is stored in an M×N two-dimensional matrix, where the horizontal axis is M and the vertical axis is N. The values ​​stored in the matrix represent the amplitude of the discharge;

[0016] S1.3: Define x to be in [1, M], y to be in [1, N], Δm and Δn to be the dimensions of the two-dimensional matrix after covariance transformation, and f(x, y) to be the matrix element value of the two-dimensional matrix with the horizontal coordinate x and the vertical coordinate y. is the mean of f(x,y); the autocorrelation covariance is used to operate between the two-dimensional pixels (x,y) and (x+Δm,y+Δn) in the atlas of size M×N:

[0017]

[0018] Where γ is the normalized autocorrelation coefficient, and A(0,0) represents the value of A(Δm,Δn) when Δm=0 and Δn=0.

[0019] Furthermore, the autocorrelation covariance is calculated as follows:

[0020]

[0021] Furthermore, S2 specifically includes the following steps:

[0022] S2.1: A fuzzy cerebellar neural network model is proposed. First, according to step S1, the M×N atlas data is calculated by autocorrelation covariance to obtain the characteristic information data of Δm×Δn, and Δm×Δn is reorganized into X=[x1 x2…x n ] T , where n = Δm × Δn;

[0023] S2.2: The fuzzy cerebellar neural network consists of an input layer, an auxiliary memory layer, a receptive field layer, a weight memory layer, and an output layer;

[0024] In the input layer: the input data of dimension n is represented as X = [x1 x2…x n ] T ,In the auxiliary memory layer, the Gaussian function is selected as the activation function as follows:

[0025]

[0026]

[0027]

[0028]

[0029] In the formula, the superscripts c and p represent the data of the current moment and the previous moment respectively, u is the number of layers of the input layer, and m is the number of layers of the input layer. ik is the input weight value of the i-th input k-th receiving domain, the symbol exp represents the exponential operation of e, ε ik is the value of the input data after weight adjustment, is the value of the activation function;

[0030] The fuzzy rule operation is established as follows:

[0031]

[0032] Where θ ik is a function of the given fuzzy membership;

[0033] At the receptive field layer, the calculation formula for the kth receptive field is as follows:

[0034]

[0035] in, represents the output of the activation function for the i-th input k-th receiving field, b k Represents the cumulative multiplication of n activation functions;

[0036] In the memory layer, the weight values ​​between the receiving field and the output layer are stored in the weight memory layer and expressed as:

[0037]

[0038] Among them, w represents the weight value between the receiving domain and the output layer, The dimension of the Euclidean space is u;

[0039] In the output layer, the output of the fuzzy cerebellar neural network model is represented as:

[0040]

[0041]

[0042] where ∈ k is the output of the receiving domain, y0 is the output of the model, y0=1 represents that the switchgear equipment is normal, and y0=0 represents that the switchgear equipment is faulty;

[0043] S2.3: Formulate the self-organizing layer-adding rules for the receiving domain layer and define the following parameters ∪ ik :

[0044]

[0045] Where ||·||2 represents the 2-norm. Continuing to define the following parameters: (Formula 12 and Formula 13)

[0046] For k=1,2,…,p, let the objective function ∪ ik Take the variable value k at the minimum value, and then assign this variable value k to The expression is as follows:

[0047]

[0048] For i=1,2,…,n, let the objective function The variable i when taking the maximum value is expressed as follows:

[0049]

[0050] Among them, K g is the set layer increase threshold;

[0051] S2.4: Formulate the self-organized layer reduction rule for the receiving domain layer and define the following parameter Z k :

[0052]

[0053] Among them, v k＝ w k b k ;

[0054] For k = 1, 2, ..., p, the objective function Z k Take the variable value k at the minimum value, and then assign this variable value k to The expression is as follows:

[0055]

[0056] in, K c is the set layer reduction threshold.

[0057] Furthermore, S3 specifically includes the following steps:

[0058] S3.1: The objective function is defined as:

[0059]

[0060] Where, e0=T0-y0, T0 is the reference output of the neural network, T0=1 represents that the switchgear equipment is normal, and T0=0 represents that the switchgear equipment is faulty, y0 is the output of the model, and e0 represents the output error of the neural network;

[0061] S3.2: Establish the Lyapunov function as follows:

[0062]

[0063] Where N is the number of iterations, e0(N)=T0-y0(N);

[0064] Define the increment of the Lyapunov function as:

[0065]

[0066] According to Taylor's formula expansion, Δe0(N) can be expressed as:

[0067]

[0068]

[0069]

[0070] The symbols Indicates approximately equal to, symbol represents the differential, Δw k and and They are weight values ​​w k and and Increment at each step;

[0071] After substitution, it is expressed as:

[0072]

[0073]

[0074]

[0075] Therefore, ΔV(N) is expressed as:

[0076]

[0077]

[0078]

[0079] Among them, η k and η ik The learning rate;

[0080] f k =y0(1-y0)b k

[0081]

[0082]

[0083] Select the parameters of each variable to meet the following conditions to make the inequality hold true, so as to ensure that the model converges accurately during the training process;

[0084]

[0085]

[0086]

[0087] Furthermore, S4 is specifically:

[0088] Obtain historically collected switchgear partial discharge detection data and fault detection results;

[0089] Step S1 is performed on the historically collected switchgear partial discharge detection data, and the obtained data is input as feature input and fault detection results into the model to train the model until the model converges; wherein the fault detection result input is the detection result at the next moment.

[0090] Based on the above scheme, the present invention also provides a computer-readable storage medium, which stores at least one instruction, at least one program, code set or instruction set. The at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the above-mentioned switch cabinet equipment fault prediction method based on self-organizing fuzzy cerebellar model neural network learning.

[0091] Beneficial effects:

[0092] The present invention provides a switch cabinet equipment fault prediction method based on self-organizing fuzzy cerebellum model neural network learning, which can grasp the health status of the switch cabinet in real time and promote the healthy development of the power industry.

[0093] The present invention first performs a spectral analysis on switchgear partial discharge detection data from a distribution network site. The spectral data is stored in a two-dimensional matrix, where the position in the matrix represents the phase of the discharge, and the value represents the amplitude of the discharge. An autocorrelation covariance method is used to remove noise from the raw data and reduce the dimensionality of the input data. A self-organizing fuzzy cerebellar neural network model with a three-layer mapping relationship is then established. Finally, the Lyapunov function is used to design the optimal learning rate for this model. This method exhibits good generalization and rapid learning speed, ensuring accurate and real-time switchgear fault prediction and is suitable for further application. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0095] Figure 1 It is a brief schematic diagram of the present invention. DETAILED DESCRIPTION

[0096] The present invention will be described in further detail below with reference to the accompanying drawings and examples. It is particularly noted that the following examples are intended only to illustrate the present invention and are not intended to limit the scope of the present invention. Similarly, the following examples are only some embodiments of the present invention and are not intended to be exhaustive. All other embodiments obtained by those of ordinary skill in the art without creative effort are intended to fall within the scope of protection of the present invention.

[0097] like Figure 1 As shown, this embodiment provides a switch cabinet equipment fault prediction method based on self-organizing fuzzy cerebellar model neural network learning, including the following steps:

[0098] S1: Spectral analysis is performed on the switchgear partial discharge detection data from the distribution network site. The characteristic data of the spectrum is stored in a two-dimensional matrix, forming an M×N feature input. Then, the autocorrelation covariance method is used to remove the noise of the original data and reduce the dimension of the input data. Specifically:

[0099] S1.1: Perform spectrum analysis on the partial discharge detection data of the switchgear to obtain the frequency and voltage relationship data;

[0100] S1.2: The data of the spectrum is stored in an M×N two-dimensional matrix, where the horizontal axis is M and the vertical axis is N. The values ​​stored in the matrix represent the amplitude of the discharge;

[0101] S1.3: Define x to be in [1, M], y to be in [1, N], Δm and Δn to be the dimensions of the two-dimensional matrix after covariance transformation, and f(x, y) to be the matrix element value of the two-dimensional matrix with the horizontal coordinate x and the vertical coordinate y. It represents the mean of f(x,y).

[0102] Next, the autocorrelation covariance used is calculated as follows:

[0103]

[0104] Finally, the autocorrelation covariance is used to calculate the two-dimensional pixels (x, y) and (x+Δm, y+Δn) in the atlas of size M×N:

[0105]

[0106] Where γ is the normalized autocorrelation coefficient, the above formula transforms the M×N two-dimensional matrix into an array with rows of Δm×Δn and columns of 1, a process referred to as reorganization. A(0,0) represents the value of A(Δm,Δn) when Δm=0 and Δn=0.

[0107] S2: Establish a self-organizing cerebellar neural network model with a three-layer mapping relationship, specifically including:

[0108] S2.1: A fuzzy cerebellar neural network model is proposed. First, according to step S1, the M×N atlas data is calculated by autocorrelation covariance to obtain the characteristic information data of Δm×Δn, and Δm×Δn is reorganized into X=[x1 x2…x n ] T , where n = Δm × Δn;

[0109] S2.2: The fuzzy cerebellar neural network consists of an input layer, an auxiliary memory layer, a receptive field layer, a weight memory layer, and an output layer;

[0110] In the input layer: the input data of dimension n is represented as X = [x1 x2…x n ] T ,In the auxiliary memory layer, the Gaussian function is selected as the activation function as follows:

[0111]

[0112]

[0113]

[0114]

[0115] In the formula, the superscripts c and p represent the data of the current moment and the previous moment respectively, u is the number of layers of the input layer, and m is the number of layers of the input layer. ik is the input weight value of the i-th input k-th receiving domain, the symbol exp represents the exponential operation of e, ε ik is the value of the input data after weight adjustment, is the value of the activation function;

[0116] The fuzzy rule operation is established as follows:

[0117]

[0118] Where θ ik is a function of the given fuzzy membership;

[0119] At the receptive field layer, the calculation formula for the kth receptive field is as follows:

[0120]

[0121] in, represents the output of the activation function for the i-th input k-th receiving field, b k Represents the cumulative multiplication of n activation functions;

[0122] In the memory layer, the weight values ​​between the receiving field and the output layer are stored in the weight memory layer and expressed as:

[0123]

[0124] Among them, w represents the weight value between the receiving domain and the output layer, The dimension of the Euclidean space is u; in the output layer, the output of the fuzzy cerebellar neural network model is expressed as:

[0125]

[0126]

[0127] where ∈ k is the output of the receiving domain, y0 is the output of the model, y0=1 represents that the switchgear equipment is normal, and y0=0 represents that the switchgear equipment is faulty;

[0128] S2.3: Formulate the self-organizing layer-adding rules for the receiving domain layer and define the following parameters ∪ ik :

[0129]

[0130] Where ||·||2 represents the 2-norm. Continuing to define the following parameters: (Formula 12 and Formula 13)

[0131] For k=1,2,…,p, let the objective function ∪ ik Take the variable value k at the minimum value, and then assign this variable value k to The expression is as follows:

[0132]

[0133] For i=1,2,…,n, let the objective function The variable i when taking the maximum value is expressed as follows:

[0134]

[0135] Among them, K g is the set layer increase threshold;

[0136] S2.4: Formulate the self-organized layer reduction rule for the receiving domain layer and define the following parameter Z k :

[0137]

[0138] Among them, v k＝ w k b k ;

[0139] For k = 1, 2, ..., p, the objective function Z k Take the variable value k at the minimum value, and then assign this variable value k to The expression is as follows:

[0140]

[0141] in, K c is the set layer reduction threshold.

[0142] S3: Use the Lyapunov function to design the optimal learning rate of the model to stabilize the model system; specifically, the following steps are included:

[0143] S3.1: The objective function is defined as:

[0144]

[0145] Where, e0=T0-y0, T0 is the reference output of the neural network, T0=1 represents that the switchgear equipment is normal, and T0=0 represents that the switchgear equipment is faulty, y0 is the output of the model, and e0 represents the output error of the neural network;

[0146] S3.2: Establish the Lyapunov function as follows:

[0147]

[0148] Where N is the number of iterations, e0(N)=T0-y0(N);

[0149] Define the increment of the Lyapunov function as:

[0150]

[0151] According to Taylor's formula expansion, Δe0(N) can be expressed as:

[0152]

[0153]

[0154]

[0155] The symbols Indicates approximately equal to, symbol represents the differential, Δw k and and They are weight values ​​w k and and Increment at each step;

[0156] After substitution, it is expressed as:

[0157]

[0158]

[0159]

[0160] in:

[0161]

[0162] in:

[0163]

[0164] in:

[0165]

[0166] in:

[0167]

[0168] in:

[0169]

[0170] in:

[0171]

[0172] Therefore, ΔV(N) can be further expressed as:

[0173]

[0174]

[0175]

[0176] Among them, η k and η ik The learning rate;

[0177] f k =y0(1-y0)b k

[0178]

[0179]

[0180] Finally, we designed the following model learning rate to make ΔV(N) < 0, ensuring the convergence of the tracking error e0(N); that is, the parameters of each variable are selected to meet the following conditions to make the inequality hold, so as to ensure that the model accurately converges during the training process;

[0181]

[0182]

[0183]

[0184] In the above inequality, T0 is a constant, and the others are variables, η k and η ik is the learning rate and iterates.

[0185] S4: Use the training data as model input to train the fuzzy brain neural network model until the model converges; specifically:

[0186] Obtain historically collected switchgear partial discharge detection data and fault detection results; perform step S1 on the historically collected switchgear partial discharge detection data, input the obtained data as feature input and the fault detection results into the model, and train the model until the model converges; wherein the fault detection result input is the detection result at the next moment. That is, when the input is the M×N feature input at moment n, the fault detection result input is the detection result at moment n+1.

[0187] S5: Execute step S1 on the current switchgear partial discharge detection data, use the obtained data as feature input, and input it into the model to predict switchgear equipment failure.

[0188] In addition, the functional units in various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0189] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or 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 and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0190] The above descriptions are only some embodiments of the present invention and do not limit the scope of protection of the present invention. Any equivalent device or equivalent process transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A switchgear equipment fault prediction method based on self-organizing fuzzy cerebellum model neural network learning, characterized in that: The steps include: S1: Perform a spectrum analysis on the partial discharge detection data from the switchgear, obtain the feature data of the spectrum and store it in a two-dimensional matrix to form an M×N feature input. Then, the autocorrelation covariance method is used to remove the noise of the original data and reduce the dimension of the input data. The specific steps include the following: S1.1: Perform spectrum analysis on the partial discharge detection data of the switchgear to obtain the frequency and voltage relationship data; S1.2: The data of the spectrum is stored in an M×N two-dimensional matrix, where the horizontal axis is M and the vertical axis is N. The values ​​stored in the matrix represent the amplitude of the discharge; S1.3: Define x to be in [1, M], y to be in [1, N], Δm and Δn to be the dimensions of the two-dimensional matrix after covariance transformation, and f(x, y) to be the matrix element value of the two-dimensional matrix with the horizontal coordinate x and the vertical coordinate y. is the mean of f(x, y); the autocorrelation covariance is used to calculate the two-dimensional pixels (x, y) and (x+Δm, y+Δn) in the atlas of size M×N: Where γ is the normalized autocorrelation coefficient, A(0,0) represents the value of A(Δm, Δn) when Δm=0 and Δn=0; S2: Establish a self-organizing cerebellar neural network model with a three-layer mapping relationship; specifically, the following steps are included: S2.1: A fuzzy cerebellar neural network model is proposed. First, according to step S1, the M×N atlas data is calculated by autocorrelation covariance to obtain the characteristic information data of Δm×Δn, and Δm×Δn is reorganized into X=[x1 x2…x n ] T , where n = Δm × Δn; S2.2: The fuzzy cerebellar neural network consists of an input layer, an auxiliary memory layer, a receptive field layer, a weight memory layer, and an output layer; In the input layer: the input data of dimension n is represented as X = x1 x2…x n ] T , In the auxiliary memory layer, the Gaussian function is selected as the activation function as follows: In the formula, the superscripts c and p represent the data of the current moment and the previous moment respectively, u is the number of layers of the input layer, and m is the number of layers of the input layer. ik is the input weight value of the i-th input k-th receiving domain, the symbol exp represents the exponential operation of e, ε ik is the value of the input data after weight adjustment, is the value of the activation function; The fuzzy rule operation is established as follows: Where θ ik is a function of the given fuzzy membership; At the receptive field layer, the calculation formula for the kth receptive field is as follows: in, represents the output of the activation function for the i-th input k-th receiving field, b k Represents the cumulative multiplication of n activation functions; In the memory layer, the weight values ​​between the receiving field and the output layer are stored in the weight memory layer and expressed as: Among them, w represents the weight value between the receiving domain and the output layer, The dimension of the Euclidean space is u; In the output layer, the output of the fuzzy cerebellar neural network model is represented as: where ∈ k is the output of the receiving domain, y0 is the output of the model, y0=1 represents that the switchgear equipment is normal, and y0=0 represents that the switchgear equipment is faulty; S2.3: Formulate the self-organizing layer-adding rules for the receiving domain layer and define the following parameters ∪ ik : Where ||·||2 represents the 2-norm; For k = 1, 2, ..., p, the objective function ∪ ik Take the variable value k at the minimum value, and then assign this variable value k to The expression is as follows: For i=1, 2, ..., n, let the objective function The variable i when taking the maximum value is expressed as follows: Among them, K g is the set layer increase threshold; S2.4: Formulate the self-organized layer reduction rule for the receiving domain layer and define the following parameter Z k : Among them, v k＝ w k b k ; For k = 1, 2, ..., p, the objective function Z k Take the variable value k at the minimum value, and then assign this variable value k to The expression is as follows: in, K c is the set layer reduction threshold; S3: Use Lyapunov function to design the optimal learning rate of the model to make the model system stable; S4: Using the training data as model input, the fuzzy brain neural network model is trained until the model converges; S5: Execute step S1 on the current switchgear partial discharge detection data, use the obtained data as feature input, and input it into the model to predict switchgear equipment failure.

2. The switch cabinet equipment fault prediction method based on self-organizing fuzzy cerebellar model neural network learning according to claim 1 is characterized in that: The autocorrelation covariance used is calculated as follows:

3. The switchgear equipment fault prediction method based on self-organizing fuzzy cerebellar model neural network learning according to claim 1 is characterized in that: S3 specifically includes the following steps: S3.1: The objective function is defined as: Where, e0=T0-y0, T0 is the reference output of the neural network, T0=1 represents that the switchgear equipment is normal, and T0=0 represents that the switchgear equipment is faulty, y0 is the output of the model, and e0 represents the output error of the neural network; S3.2: Establish the Lyapunov function as follows: Where N is the number of iterations, e0(N)=T0-y0(N); Define the increment of the Lyapunov function as: According to Taylor's formula expansion, Δe0(N) is expressed as: Among them, the symbol represents the differential, Δw k and and They are weight values ​​w k and and Increment at each step; After substitution, it is expressed as: Therefore, ΔV(N) is expressed as: Among them, η k and η ik The learning rate; f k =y0(1-y0)b k Select the parameters of each variable to meet the following conditions to make the inequality hold true, so as to ensure that the model converges accurately during the training process; 4. The switch cabinet equipment fault prediction method based on self-organizing fuzzy cerebellar model neural network learning according to claim 1 is characterized in that: S4 is specifically: Obtain historically collected switchgear partial discharge detection data and fault detection results; Step S1 is performed on the historically collected switchgear partial discharge detection data, and the obtained data is input as feature input and fault detection results into the model to train the model until the model converges; wherein the fault detection result input is the detection result at the next moment.

5. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor to implement the switch cabinet equipment fault prediction method based on self-organizing fuzzy cerebellar model neural network learning as described in one of claims 1 to 4.

Images