A high-voltage circuit breaker fault diagnosis method based on deep learning

Abstract

The application provides a high-voltage circuit breaker fault diagnosis method based on deep learning, which comprises grouping test of the high-voltage circuit breaker, recording of data of detection parameters of the high-voltage circuit breaker under various working conditions, and classification of the recorded data according to the working conditions, wherein the various working conditions comprise a normal working condition and a plurality of fault working conditions; a reference signal is trained according to the data under the normal working condition; the reference signal is verified according to the data in the plurality of fault working conditions; and the high-voltage circuit breaker is diagnosed and recognized by using a DTW algorithm according to the reference signal. The reference signal is verified, so that local optimal solution can be effectively avoided, and the reference signal is adjusted and corrected for multiple times, which is more accurate than a traditional average or arbitrary selection mode.

Description

Technical Field

[0001] This invention belongs to the field of circuit breaker fault diagnosis technology, specifically relating to a high-voltage circuit breaker fault diagnosis method based on deep learning. Background Technology

[0002] As crucial high-voltage equipment in power systems, high-voltage circuit breakers play a vital role in control and protection, and their operational status directly impacts the safety and stability of the lines they operate on. When a short-circuit fault occurs, the high-voltage circuit breaker quickly interrupts the massive short-circuit current to protect the entire power system from damage. However, high-voltage circuit breakers are diverse in type, quantity, and structure, and are susceptible to failure due to factors such as their own quality and external operating conditions, leading to a decline in the reliability of the entire power system.

[0003] In recent years, with the development of various algorithms and artificial intelligence, fault diagnosis of high-voltage circuit breakers has become increasingly focused on real-time and predictive capabilities. One such method is high-voltage circuit breaker fault diagnosis based on prior knowledge, which compares the measured signals with prior knowledge to determine the circuit breaker's state. Among these methods, the Dynamic Time Warping (DTW) algorithm is a relatively mature diagnostic method. In the DTW algorithm, prior knowledge is the reference signal. However, existing research mostly selects a random signal from the normal state signals or averages multiple signals as the reference signal when choosing the DTW reference signal, without providing a better selection method. Summary of the Invention

[0004] In view of this, the present invention aims to provide a high-voltage circuit breaker fault diagnosis method based on deep learning to solve the problems existing in the prior art.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0006] A deep learning-based fault diagnosis method for high-voltage circuit breakers includes the following steps:

[0007] Step S1: Perform group tests on the high-voltage circuit breakers, record the test parameters of the high-voltage circuit breakers under various operating conditions, and classify the recorded data according to the operating conditions. Each operating condition includes normal operating conditions and multiple fault operating conditions.

[0008] Step S2: Train the reference signal based on data from normal operating conditions;

[0009] Step S3: Verify the reference signal based on data from multiple fault conditions;

[0010] Step S4: Diagnose and identify the high-voltage circuit breaker using the DTW algorithm based on the reference signal.

[0011] Furthermore, in step S1, the detection parameters of the high-voltage circuit breaker include one of the following: coil current, contact vibration signal, contact current, and contact travel distance. All detection parameters of the high-voltage circuit breaker are parameters that change with respect to time.

[0012] Furthermore, in step S2, the training method for the reference signal includes the following steps:

[0013] Step S201: Sample the data under normal operating conditions to obtain the sampled signal X. i = (i1, i2, ..., i m ), i = 1, 2, ..., n, where n is the total number of data under normal operating conditions, and m is the dimension of the sampled signal;

[0014] Step S202: Traverse all sample data and obtain the maximum difference Δi for each vector of the signal. k k = 1, 2, ..., m, select i k '+Δi k / 2 is used as the pivot vector element, i k To generate the maximum difference Δi k The minimum value corresponding to the time is used to obtain the reference signal X0 = (i1' + Δi1 / 2, i2' + Δi2 / 2, ..., i m '+Δi m / 2);

[0015] Step S203: Based on the circuit breaker's operating time, the reference signal X0 and the sampled signal X i The system is segmented, and feature matching is performed on each segment. The most similar sampled signal is selected from each segment, and finally, the signal source X with the most similar sampled signals is determined. s X0 serves as the new reference signal and the adjustment signal.

[0016] Furthermore, in step S203, the number of segments is more than three, and the three segments correspond to the action preparation segment, the action process segment, and the action termination segment.

[0017] Furthermore, in step S3, a clustering algorithm is used to process the sample data for each fault condition to obtain the cluster center signal and the maximum cluster distance for each fault condition sample data, and then compared with X. s The distance values ​​between the X0 and X1 signals are compared. If the distance value is less than the maximum cluster distance, the distance value is compared with the average cluster distance. If the distance value is less than the average cluster distance, X0 is used to adjust X1. s The adjustment is performed by replacing X with data from dissimilar segments X0 in sequence. s The corresponding vector elements.

[0018] Furthermore, the adjustment method includes the following steps:

[0019] Step S301: Set the number of vector elements to be replaced. The initial value is 1, and the upper limit of the number is the upper limit of the number of vector elements in the segment. The process will terminate when the upper limit is reached.

[0020] Step S302: Sort the vector elements in descending order according to the difference between them, and replace the corresponding vector elements in turn. If the distance value after replacement is greater than the average value of the cluster distance, then terminate. Otherwise, increase the number of vector elements replaced by one and repeat step S301.

[0021] Furthermore, in step S4, the diagnostic identification is implemented based on an artificial neural network. The artificial algorithm includes multiple neurons, each corresponding to the identification of different working conditions. The identification algorithm for different working conditions is the DTW algorithm. The input of the artificial neural algorithm is the data to be classified, and the output of the artificial neural algorithm is a unit vector containing only one element. The unit vector corresponds to the working condition identification result.

[0022] Furthermore, the fault diagnosis method also includes the following steps: the sample data obtained after each pair of sample identifications will be fed back into the corresponding classification sample database. When the rate of change of the sample variables in the sample database is greater than 2%, the baseline signal will be recalculated.

[0023] In summary, this invention provides a deep learning-based fault diagnosis method for high-voltage circuit breakers. The method includes grouping and testing high-voltage circuit breakers, recording the detection parameters of the circuit breakers under various operating conditions, and classifying the recorded data according to the operating conditions, including normal operating conditions and multiple fault operating conditions. A reference signal is trained based on the data from the normal operating conditions; the reference signal is verified based on data from multiple fault operating conditions; and the high-voltage circuit breaker is diagnosed and identified using the DTW algorithm based on the reference signal. This invention effectively avoids local optima by verifying the reference signal, and the multiple adjustments and corrections to the reference signal result in higher accuracy than traditional averaging or arbitrary selection methods. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A flowchart of a high-voltage circuit breaker fault diagnosis method based on deep learning provided for an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0027] As crucial high-voltage equipment in power systems, high-voltage circuit breakers play a vital role in control and protection, and their operational status directly impacts the safety and stability of the lines they operate on. When a short-circuit fault occurs, the high-voltage circuit breaker quickly interrupts the massive short-circuit current to protect the entire power system from damage. However, high-voltage circuit breakers are diverse in type, quantity, and structure, and are susceptible to failure due to factors such as their own quality and external operating conditions, leading to a decline in the reliability of the entire power system.

[0028] In recent years, with the development of various algorithms and artificial intelligence, fault diagnosis of high-voltage circuit breakers has become increasingly focused on real-time and predictive capabilities. One such method is high-voltage circuit breaker fault diagnosis based on prior knowledge, which compares the measured signals with prior knowledge to determine the circuit breaker's state. Among these methods, the Dynamic Time Warping (DTW) algorithm is a relatively mature diagnostic method. In the DTW algorithm, prior knowledge is the reference signal. However, existing research mostly selects a random signal from the normal state signals or averages multiple signals as the reference signal when choosing the DTW reference signal, without providing a better selection method.

[0029] Based on this, the present invention aims to provide a high-voltage circuit breaker fault diagnosis method based on deep learning.

[0030] The following is a detailed description of an embodiment of a deep learning-based fault diagnosis method for high-voltage circuit breakers according to the present invention.

[0031] Please see Figure 1 This embodiment provides a high-voltage circuit breaker fault diagnosis method based on deep learning, including the following steps:

[0032] Step S1: Perform group tests on the high-voltage circuit breakers, record the test parameters of the high-voltage circuit breakers under various operating conditions, and classify the recorded data according to the operating conditions. Each operating condition includes normal operating conditions and multiple fault operating conditions.

[0033] Step S2: Train the reference signal based on data from normal operating conditions;

[0034] Step S3: Verify the reference signal based on data from multiple fault conditions;

[0035] Step S4: Diagnose and identify the high-voltage circuit breaker using the DTW algorithm based on the reference signal.

[0036] This embodiment provides a deep learning-based fault diagnosis method for high-voltage circuit breakers. The method includes grouping and testing high-voltage circuit breakers, recording the detection parameters of the circuit breakers under various operating conditions, and classifying the recorded data according to the operating conditions, including normal operating conditions and multiple fault operating conditions. A reference signal is trained based on the data from the normal operating conditions; the reference signal is verified based on data from multiple fault operating conditions; and the high-voltage circuit breaker is diagnosed and identified using the DTW algorithm based on the reference signal. This invention effectively avoids local optima by verifying the reference signal, and the multiple adjustments and corrections to the reference signal result in higher accuracy than traditional averaging or arbitrary selection methods.

[0037] In an optional embodiment, in step S1, the detection parameters of the high-voltage circuit breaker include one of coil current, contact vibration signal, contact current and contact travel, and all detection parameters of the high-voltage circuit breaker are parameters of change with respect to time.

[0038] In an optional embodiment, step S2, the training method for the reference signal includes the following steps:

[0039] Step S201: Sample the data under normal operating conditions to obtain the sampled signal X. i = (i1, i2, ..., i m ), i = 1, 2, ..., n, where n is the total number of data under normal operating conditions, and m is the dimension of the sampled signal;

[0040] Step S202: Traverse all sample data and obtain the maximum difference Δi for each vector of the signal. k k = 1, 2, ..., m, select i k '+Δi k / 2 is used as the pivot vector element, i k To generate the maximum difference Δi k The minimum value corresponding to the time is used to obtain the reference signal X0 = (i1' + Δi1 / 2, i2' + Δi2 / 2, ..., i m '+Δi m / 2);

[0041] Step S203: Based on the circuit breaker's operating time, the reference signal X0 and the sampled signal X i The system is segmented, and feature matching is performed on each segment. The most similar sampled signal is selected from each segment, and finally, the signal source X with the most similar sampled signals is determined. s X0 serves as the new reference signal and the adjustment signal.

[0042] In an optional embodiment, in step S203, the number of segments is more than three, and the three segments correspond to the action preparation segment, the action process segment, and the action termination segment.

[0043] In an optional embodiment, in step S3, a clustering algorithm is used to process the sample data for each fault condition to obtain the cluster center signal and the maximum cluster distance for each fault condition sample data, and then compared with X. s The distance values ​​between the X0 and X1 signals are compared. If the distance value is less than the maximum cluster distance, the distance value is compared with the average cluster distance. If the distance value is less than the average cluster distance, X0 is used to adjust X1. s The adjustment is performed by replacing X with data from dissimilar segments X0 in sequence. s The corresponding vector elements.

[0044] In an optional embodiment, the adjustment method includes the following steps:

[0045] Step S301: Set the number of vector elements to be replaced. The initial value is 1, and the upper limit of the number is the upper limit of the number of vector elements in the segment. The process will terminate when the upper limit is reached.

[0046] Step S302: Sort the vector elements in descending order according to the difference between them, and replace the corresponding vector elements in turn. If the distance value after replacement is greater than the average value of the cluster distance, then terminate. Otherwise, increase the number of vector elements replaced by one and repeat step S301.

[0047] In an optional embodiment, in step S4, the diagnostic identification is implemented based on an artificial neural network. The artificial algorithm includes multiple neurons, each corresponding to the identification of different working conditions. The identification algorithm for different working conditions is the DTW algorithm. The input of the artificial neural algorithm is the data to be classified, and the output of the artificial neural algorithm is a unit vector containing only one element. The unit vector corresponds to the working condition identification result.

[0048] In an optional embodiment, the fault diagnosis method further includes the following steps: the sample data obtained after each pair of sample identifications is fed back into the sample database of the corresponding classification; when the rate of change of the sample variables in the sample database is greater than 2%, the baseline signal is recalculated.

[0049] Compared to existing technologies, this invention first calculates the median of the sample data using arithmetic, then segments the actual circuit breaker's operating time, and matches the most similar sample according to the operating time segment. This sample is the median value of the normal operating range, which can then be used to select a reference signal. Subsequently, this reference signal is compared with the sample centers of other fault conditions to further adjust the reference signal. The fault diagnosis method provided by this invention can effectively avoid local optima when selecting the reference signal, and the reference signal is adjusted and corrected multiple times, making it more accurate than traditional averaging or arbitrary selection methods.

[0050] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fault diagnosis method for high-voltage circuit breakers based on deep learning, characterized in that, Includes the following steps: Step S1: Perform group tests on the high-voltage circuit breakers, record the test parameters of the high-voltage circuit breakers under various operating conditions, and classify the recorded data according to the operating conditions. Each operating condition includes normal operating conditions and multiple fault operating conditions. Step S2: Train the reference signal based on data from normal operating conditions; Step S3: Verify the reference signal based on data from multiple fault conditions; Step S4: Diagnose and identify the high-voltage circuit breaker using the DTW algorithm based on the reference signal; In step S2, the training method for the reference signal includes the following steps: Step S201: Sample the data under normal operating conditions to obtain the sampled signal X. i =(i1, i2, ..., i m ), i = 1, 2, ..., n, where n is the total number of data under normal operating conditions, and m is the dimension of the sampled signal; Step S202: Traverse all sample data and obtain the maximum difference Δi for each vector of the signal. k k = 1, 2, ..., m, select i k '+Δi k / 2 is used as the pivot vector element, i k To generate the maximum difference Δi k The minimum value corresponding to the time is used to obtain the reference signal X0=(i1'+Δi1 / 2, i2'+Δi2 / 2, ..., i m '+Δi m / 2); Step S203: Based on the circuit breaker's operating time, the reference signal X0 and the sampled signal X i The system is segmented, and feature matching is performed on each segment. The most similar sampled signal is selected from each segment, and finally, the signal source X with the most similar sampled signals is determined. s X0 serves as the new reference signal and the adjustment signal. In step S203, there are three segments, which correspond to the action preparation segment, the action process segment, and the action termination segment. In step S3, a clustering algorithm is used to process the sample data for each fault condition to obtain the cluster center signal and the maximum cluster distance for each fault condition sample data, and then compared with X. s The distance value between the signal and each cluster center is calculated. If the distance value is less than the maximum cluster distance, it is compared with the average cluster distance. If the distance value is less than the average cluster distance, X0 is used to adjust X. s The adjustment is performed by replacing X with data from dissimilar segments X0 in sequence. s The corresponding vector elements; The adjustment method includes the following steps: Step S301: Set the number of vector elements to be replaced. The initial value is 1, and the upper limit of the number is the upper limit of the number of vector elements in the segment. The process will terminate when the upper limit is reached. Step S302: Sort the vector elements in descending order according to the difference between them, and replace the corresponding vector elements in turn. If the distance value after replacement is greater than the average value of the cluster distance, then terminate. Otherwise, increase the number of vector elements replaced by one and repeat step S301.

2. The high-voltage circuit breaker fault diagnosis method based on deep learning according to claim 1, characterized in that: In step S1, the detection parameters of the high-voltage circuit breaker include one of coil current, contact vibration signal, contact current and contact travel, and all detection parameters of the high-voltage circuit breaker are parameters that change with time.

3. The high-voltage circuit breaker fault diagnosis method based on deep learning according to claim 1, characterized in that: In step S4, the diagnostic identification is implemented based on an artificial neural network, which includes multiple neurons, each corresponding to the identification of different working conditions. The identification algorithm for the different working conditions is the DTW algorithm. The input of the artificial neural network is the data to be classified, and the output of the artificial neural network is a unit vector containing only one element, which corresponds to the working condition identification result.

4. A high-voltage circuit breaker fault diagnosis method based on deep learning according to any one of claims 1-3, characterized in that: The fault diagnosis method further includes the following steps: the sample data obtained after each sample identification will be fed back to the corresponding classification sample database. When the rate of change of the sample variables in the sample database is greater than 2%, the baseline signal will be recalculated.

Images