A mechanical fault diagnosis method for high-voltage circuit breakers

Abstract

The present invention relates to the field of high-voltage electrical equipment, and particularly to a method for diagnosing mechanical faults of a high-voltage circuit breaker, including: Step 1: Fusing the vibration signal and the coil current of the circuit breaker to form a fused characteristic parameter; the coil current is the coil current when the high-voltage circuit breaker operates; the vibration signal is the vibration signal measured on the base when the high-voltage circuit breaker operates; Step 2: Constituting a training data set with the fused characteristic parameters of the circuit breaker in different states; Step 3: Constructing a fault diagnosis model and training the fault diagnosis model with the training data set; Step 4: Using the trained fault diagnosis model to diagnose the faults of the circuit breaker. The present invention improves the diagnosis accuracy rate.

Description

Technical Field

[0001] The present invention relates to the field of high-voltage electrical equipment, and particularly to a method for diagnosing mechanical faults of a high-voltage circuit breaker. Background Art

[0002] A high-voltage circuit breaker is an important control and protection device in the power system, and its reliability is directly related to the safety of the power grid. Therefore, it is necessary to conduct long-term tracking and maintenance on it; the method of regular disassembly and maintenance is time-consuming and laborious, and may cause damage or shortened life of the high-voltage circuit breaker due to improper maintenance. Therefore, it is very necessary to develop an on-line monitoring method for the circuit breaker.

[0003] At present, the main monitoring objects of the circuit breaker are vibration signals and coil currents. Different monitoring objects respectively reflect different state information. At present, there is no feature extraction and fault diagnosis method based on fused information. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for diagnosing mechanical faults of a high-voltage circuit breaker to improve the diagnostic accuracy.

[0005] To solve the above technical problems, the technical solution of the present invention is: A method for diagnosing mechanical faults of a high-voltage circuit breaker, comprising:

[0006] Step 1: Fuse the vibration signal and coil current of the circuit breaker to form a fused feature parameter; the coil current is the coil current when the high-voltage circuit breaker operates; the vibration signal is the vibration signal measured on the base when the high-voltage circuit breaker operates;

[0007] Step 2: Construct a training data set from the fused feature parameters of the circuit breaker in different states;

[0008] Step 3: Construct a fault diagnosis model and train the fault diagnosis model with the training data set;

[0009] Step 4: Use the trained fault diagnosis model to diagnose the faults of the circuit breaker.

[0010] According to the above solution, step 1 is specifically:

[0011] Step 1.1: Extract the coil current feature parameter according to the coil current;

[0012] Step 1.2: Extract the vibration signal feature parameter according to the vibration signal;

[0013] Step 1.3: Combine the coil current feature parameter and the vibration signal feature parameter into a one-dimensional vector to obtain the fused feature parameter.

[0014] According to the above solution, in step 1.1, the coil current feature parameter is the time of each extreme point of the coil current.

[0015] According to the above solution, in step 1.2, the steps for extracting the vibration signal characteristic parameters are as follows:

[0016] Step 1.21: Use variational mode decomposition to decompose the vibration signal into k intrinsic mode components;

[0017] Step 1.22: Perform time-domain segmentation on the decomposed intrinsic mode matrix, calculate the maximum singular value of each sub-matrix, and combine the maximum singular values of each sub-matrix to obtain the vibration signal characteristic parameters.

[0018] According to the above solution, the specific content of step 1.21 is as follows:

[0019] Use variational mode decomposition to process the original vibration signal. Select the decomposition numbers from 1 to 15 respectively, stack and reconstruct the obtained intrinsic mode components to obtain the reconstructed vibration signal, and calculate the normalized distance d between the original signal and the reconstructed signal. The expression is as follows:

[0020]

[0021] In the formula, i represents the i-th data, n represents the number of data, r represents the reconstructed signal, r i represents the i-th reconstructed signal, x represents the original signal, x i represents the i-th original signal;

[0022] Calculate the modal repetition coefficient c of the signal. The expression is as follows:

[0023]

[0024] In the formula, k represents the decomposition number, ω is the central frequency of the intrinsic mode component, ω i represents the central frequency of the i-th intrinsic mode component, and ε is a very small number, taking 0.01;

[0025] Scale the normalized distance and modal repetition coefficient curves to the same range, and take the upper envelope of the two curves. Use the k value at the minimum point of the envelope as the decomposition number of variational mode decomposition.

[0026] According to the above solution, the specific content of step 1.22 is as follows:

[0027] Z is the intrinsic mode component matrix, which is composed of k one-dimensional intrinsic mode components:

[0028]

[0029] In the formula, z k1 …z kn represents the k-th mode component, z k1 represents the first data in the k-th mode component, zkn Denote the nth data in the kth modal component as z 11 …z 1n Denote the 1st modal component as z 11 Denote the 1st data in the 1st modal component as z 1n Denote the nth data in the 1st modal component;

[0030] Perform time-domain segmentation on Z to obtain the submatrix zs i :

[0031] Z = [zs1, zs2,... zs 32 (4)

[0032] Perform singular value decomposition on the submatrix zs i to obtain the orthogonal matrices U and V, and the diagonal matrix Σ. Let λ be the diagonal elements of the diagonal matrix, and v i be the maximum value among the diagonal elements.

[0033] zs i = UΣV T (5)

[0034] Σ = diag(λ1, λ2,..., λ r ) (6)

[0035] v i = max(Σ) (7)

[0036] where V T is the conjugate transpose of the orthogonal matrix V, λ r is the rth diagonal element of the diagonal matrix Σ, and v i is the maximum value among the diagonal elements.

[0037] Combine all the maximum singular values v i to form a vector v, which is the vibration signal characteristic parameter.

[0038] According to the above scheme, step 2 is specifically as follows: Collect and obtain the fusion characteristic parameters of the circuit breaker in different states, record the state of the circuit breaker, mark it with numbers, and correspond the two to form a training data set.

[0039] According to the above scheme, in step 3, the fault diagnosis model uses a deep neural network model.

[0040] The present invention has the following beneficial effects:

[0041] 1. Combine the vibration signal with the coil current to form a fusion feature parameter including the characteristic parameters of the coil current and the characteristic parameters of the vibration signal. The fusion feature parameter is the mechanical state characteristic parameter, which can extract more fault types. Based on the extracted fusion feature parameter, fault diagnosis is carried out, expanding the fault diagnosis range and reliability of the circuit breaker and improving the diagnosis accuracy rate.

[0042] 2. Propose a method to determine the number of decompositions of the vibration signal by combining the modal repetition coefficient and the normalized distance, which is more comprehensive and has better decomposition effect compared with the existing methods. Description of the Drawings

[0043] Figure 1 It is a schematic flow chart of an embodiment of the present invention;

[0044] Figure 2 It is the time-domain diagram of the coil current in this embodiment;

[0045] Figure 3 It is the envelope diagram of the curve of the normalized distance and the modal repetition coefficient in this embodiment;

[0046] Figure 4 It is the relationship diagram between the vibration signal and the maximum singular value in this embodiment. Detailed Embodiment

[0047] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0048] Please refer to Figures 1 to 4 , the present invention is a mechanical fault diagnosis method for a high-voltage circuit breaker, which includes:

[0049] Step 1: Combine the vibration signal and the coil current of the circuit breaker to form a fusion feature parameter; specifically:

[0050] Step 1.1: Extract the characteristic parameters of the coil current according to the coil current; the coil current is the coil current when the high-voltage circuit breaker operates, and the characteristic parameters of the coil current are the time of each extreme point of the coil current. Figure 1 is the coil current waveform. By taking the extreme points and the maximum value, four time parameters t1, t2, t3, and t4 in Figure 1 are obtained.

[0051] Step 1.2: Extract the characteristic parameters of the vibration signal according to the vibration signal; the vibration signal is the vibration signal measured on the base when the high-voltage circuit breaker operates, and the vibration signal is the vibration acceleration signal; the steps to extract the characteristic parameters of the vibration signal are:

[0052] Step 1.21: Decompose the vibration signal into k intrinsic mode components by using variational mode decomposition; specifically:

[0053] The original vibration signal is processed using variational mode decomposition. The number of decompositions is selected as 1 - 15 respectively. The obtained intrinsic mode components are superimposed and reconstructed to obtain the reconstructed vibration signal. The normalized distance d between the original signal and the reconstructed signal is calculated, and the expression is as follows:

[0054]

[0055] In the formula, i represents the i-th data, n represents the number of data, r represents the reconstructed signal, r i represents the i-th reconstructed signal, x represents the original signal, x i represents the i-th original signal;

[0056] The modal repetition coefficient c of the signal is calculated, and the expression is as follows:

[0057]

[0058] In the formula, k represents the number of decompositions, ω is the central frequency of the intrinsic mode component, ω i represents the central frequency of the i-th intrinsic mode component, and ε is a very small number, taking 0.01;

[0059] The curves of the normalized distance and the modal repetition coefficient are scaled to the same range, and the upper envelope of the two curves is taken. The k value at the minimum point of the envelope is used as the number of variational mode decompositions; the result is shown in the appendix Figure 2 , the two curves intersect at k = 6. Take the first half of the normalized distance curve (k = 2 - 5), take the second half of the modal repetition coefficient curve (k = 6 - 15), and then take the k value at the minimum of the combined curve (k = 7) as the final number of decompositions.

[0060] Step 1.22: Perform time-domain segmentation on the obtained intrinsic mode matrix, calculate the maximum singular value of each sub-matrix, and combine the maximum singular values of each sub-matrix to obtain the vibration signal characteristic parameters; specifically:

[0061] Z is the intrinsic mode component matrix, which consists of k one-dimensional intrinsic mode components:

[0062]

[0063] In the formula, z k1 …z kn represents the k-th mode component, z k1 represents the first data in the k-th mode component, z kn represents the n-th data in the k-th mode component, z 11 …z 1n represents the first mode component, z 11 represents the first data in the first mode component,1n Denote the nth data in the first modal component;

[0064] Perform time-domain segmentation on Z to obtain the sub-matrix zs i :

[0065] Z = [zs1, zs2,... zs 32 (4)

[0066] For the sub-matrix zs i Perform singular value decomposition to obtain the orthogonal matrices U and V, and the diagonal matrix Σ. λ is the diagonal element of the diagonal matrix, and v i is the maximum value among the diagonal elements.

[0067] zs i = UΣV T (5)

[0068] Σ = diag(λ1, λ2,..., λ r ) (6)

[0069] v i = max(Σ) (7)

[0070] where V T is the conjugate transpose of the orthogonal matrix V, λ r is the rth diagonal element of the diagonal matrix Σ, and v i is the maximum value among the diagonal elements.

[0071] Combine all the maximum singular values v i to form a vector v, which is the vibration signal characteristic parameter.

[0072] In this embodiment, the original vibration signal is 1×2560. After modal decomposition, it becomes 6×2560. Here, the signal is segmented into 32 sub-matrices, and after segmentation, it becomes 6×80×32. Calculate the maximum singular value of each sub-matrix (6×80), refer to Appendix Figure 3 .

[0073] Step 1.3: Combine the coil current characteristic parameter and the vibration signal characteristic parameter to form a one-dimensional vector to obtain the fusion characteristic parameter.

[0074] Step 2: Construct a training data set from the fusion characteristic parameters of the circuit breaker in different states: Collect the fusion feature vectors in different states of the circuit breaker, and record the state of the circuit breaker. Use numbers such as 0, 1, 2, etc. for marking, and make a training data set by corresponding the two;

[0075] Step 3: Construct a fault diagnosis model and train the fault diagnosis model with a training data set; the fault diagnosis model can adopt machine learning or a deep neural network model. When machine learning is selected, a support vector machine can be used as the fault diagnosis model.

[0076] Step 4: Use the trained fault diagnosis model to conduct breaker fault diagnosis.

[0077] Parts not involved in the present invention are the same as or implemented by the prior art.

[0078] The above content is a further detailed description of the present invention in combination with specific implementation manners. It cannot be determined that the specific implementation of the present invention is only limited to these descriptions. For those of ordinary skill in the technical field to which the present invention pertains, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should all be regarded as belonging to the protection scope of the present invention.

Claims

1. A mechanical fault diagnosis method for a high-voltage circuit breaker, characterized in that: include Step 1: Fusing the vibration signal and coil current of the circuit breaker to form a fusion characteristic parameter; the coil current is the coil current when the high-voltage circuit breaker is in operation; the vibration signal is the vibration signal measured on the base when the high-voltage circuit breaker is in operation; specifically: Step 1.1: Extract the coil current characteristic parameter according to the coil current, where the coil current characteristic parameter is the time of each extreme value point of the coil current; Step 1.2: Extract the vibration signal characteristic parameters according to the vibration signal. The steps are as follows: Step 1.21: Decompose the vibration signal into k eigenmode components using variational mode decomposition; specifically: The original vibration signal is processed by variational mode decomposition. The number of decompositions is selected as 1-15. The intrinsic mode components obtained by decomposition are superimposed and reconstructed to obtain the reconstructed vibration signal. The normalized distance d between the original signal and the reconstructed signal is calculated. The expression is as follows: Wherein, i represents the i-th data, n represents the number of data, r represents the reconstructed signal, and r i represents the i-th reconstructed signal, x represents the original signal, and x i represents the i-th original signal; Calculate the modal repetition coefficient c of the signal, the expression is as follows: where k represents the number of decompositions, ω is the central frequency of the intrinsic mode component, ω i represents the central frequency of the i-th intrinsic mode component, ε is a very small number, taking 0.01; Scale the normalized distance and modal repetition coefficient curves to the same range, take the upper envelope of the two curves, and use the k value of the minimum point of the envelope as the number of variational mode decompositions Step 1.22: Perform time domain segmentation on the decomposed intrinsic mode matrix, calculate the maximum singular value of each sub-matrix, and combine the maximum singular values ​​of each sub-matrix to obtain the vibration signal characteristic parameters; Step 1.3: Combining the coil current characteristic parameters and the vibration signal characteristic parameters into a one-dimensional vector to obtain the fusion characteristic parameters; Step 2: The fusion feature parameters of the circuit breakers in different states form a training data set; Step 3: Build a fault diagnosis model and train the fault diagnosis model using a training data set; Step 4: Use the trained fault diagnosis model to perform circuit breaker fault diagnosis.

2. The mechanical fault diagnosis method of the high-voltage circuit breaker according to claim 1, wherein: The step 1.22 is specifically as follows: Z is the eigenmode component matrix, which consists of k one-dimensional eigenmode components composition: where z k1 z kn represents the k-th modal component, z k1 represents the first data in the k-th modal component, z kn represents the n-th data in the k-th modal component, z 11 z 1n represents the first modal component, z 11 represents the first data in the first modal component, z 1n represents the n-th data in the first modal component; Perform time-domain segmentation on Z to obtain the sub-matrix zs i : Z = [zs1, zs2,... zs 32 (4) For sub-matrix zs i Perform singular value decomposition to obtain orthogonal matrices U and V, and diagonal matrix Σ. Let λ be the diagonal element of the diagonal matrix, and v i be the maximum value among the diagonal elements; zs i = UΣV T (5) Σ = diag(λ1, λ2,..., λ r ) (6) v i = max(Σ) (7) where V T is the conjugate transpose of the orthogonal matrix V, λ r is the r-th diagonal element of the diagonal matrix Σ, and v i is the maximum value among the diagonal elements; Combine all the maximum singular values v i to form a vector v, which is the characteristic parameter of the vibration signal.

3. The mechanical fault diagnosis method of the high-voltage circuit breaker according to claim 1, characterized in that: Step 2 is specifically as follows: collect and obtain the fusion feature parameters of the circuit breaker in different states, record the state of the circuit breaker, mark it with numbers, and correspond the two to form a training data set.

4. The mechanical fault diagnosis method of the high-voltage circuit breaker according to claim 1, characterized in that: In step 3, the fault diagnosis model adopts a deep neural network model.

Images