A Circuit Breaker Fault Diagnosis Method Based on Genetic Algorithm and Clustering Algorithm

Abstract

The present invention provides a circuit breaker fault diagnosis method based on genetic algorithm and clustering algorithm, which relates to the technical field of fault diagnosis. The fault diagnosis method includes the following steps: S1, collecting multi-source fault external characteristic data of the circuit breaker and performing data preprocessing; S2, constructing multiple groups of physical quantity combinations according to the multi-source fault external characteristic data of the circuit breaker obtained by preprocessing in step S1, and using the genetic algorithm to evaluate the multiple groups of physical quantity combinations; S3, constructing a multi-dimensional criterion for circuit breaker faults by using the genetic algorithm; S4, judging the circuit breaker faults by the K-means clustering method according to the multi-dimensional criterion for circuit breaker faults constructed in step S3; The genetic algorithm is used to combine and screen the multi-dimensional physical quantity information of the circuit breaker. Through the genetic algorithm, random combinations of different physical quantities are realized, the characteristics of each physical quantity are explored one by one, and the optimal circuit breaker fault criterion is screened out, solving the problem that the existing circuit breaker fault diagnosis algorithm is not universal.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and particularly to a circuit breaker fault diagnosis method based on genetic algorithm and clustering algorithm. Background Art

[0002] The high-voltage circuit breaker is an important electrical equipment, and may fail due to lightning strikes, external forces, high temperatures, improper operations, etc. during the operation of the power system. The failure of the high-voltage circuit breaker will affect the normal operation of the power system, so it is necessary to identify the fault in time and take corresponding countermeasures.

[0003] Currently, there are many available criteria for the fault diagnosis of circuit breakers, mainly including electrical signals such as voltage and current, sound signals, temperature signals, and so on. However, most judgment methods only involve one or two criteria and can only target one type or several types of faults, and do not have universality in fault diagnosis. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present invention provides a circuit breaker fault diagnosis method based on genetic algorithm and clustering algorithm.

[0005] The present invention solves the above technical problems through the following technical means: A circuit breaker fault diagnosis method based on genetic algorithm and clustering algorithm, comprising the following steps:

[0006] S1. Collect the multi-fault external characteristic data of the circuit breaker and perform data preprocessing;

[0007] S2. According to the multi-fault external characteristic data of the circuit breaker obtained by preprocessing in step S1, construct multiple groups of physical quantity combinations, and use the genetic algorithm to evaluate the multiple groups of physical quantity combinations;

[0008] S3. According to the multiple groups of physical quantity combinations constructed in step S2, use the genetic algorithm to construct multi-dimensional criteria for circuit breaker faults;

[0009] S4. According to the multi-dimensional criteria for circuit breaker faults constructed in step S3, judge the circuit breaker faults by the K-means clustering method.

[0010] Further, the multi-fault external characteristic data of the circuit breaker includes the operation data corresponding to the normal operation state and the fault state of the circuit breaker.

[0011] Further, the operation data includes current, voltage, electrical signal, temperature signal, and sound signal.

[0012] Further, the data preprocessing includes performing data transformation on the collected multi-fault external characteristic data to achieve data normalization, removing outlier samples, and integrating the data set into a database.

[0013] Further, a genetic algorithm is used to construct a multi-dimensional criterion for circuit breaker faults, which specifically includes the following steps:

[0014] S31. Based on the pre-processed multi-fault external characteristic data in step S1, randomly generate 10 groups of initial physical quantity combination vectors as the first-generation physical quantity combinations: The physical quantity combinations only contain 0 and 1, and the number of vector elements is the same as the number of physical quantities in the pre-processed multi-fault external characteristic data, where 0 indicates that the physical quantity combination does not contain the physical quantity, and 1 indicates that the physical quantity combination contains the physical quantity;

[0015] S32. Evaluate the first-generation physical quantity combinations through the fitness function in the genetic algorithm. The fitness function: The size of the difference between the sample center of the K-means clustering analysis result of the data group using this physical quantity combination and the actual classification sample center; The K-means clustering method calculates the average distance between each group of data and each classification center. The classification center is the average value vector of each sample in the known classification. Add this group of data to the group with the smallest average distance and update the classification center of this group until all data is classified. In summary, if the actual known classification center vector is For a certain physical quantity combination The fitness function F(x) is expressed as:

[0016]

[0017] where the smaller F(x) is, the better. The population iterates in the direction of decreasing F(x);

[0018] S33. After evaluating the first-generation 10 groups of physical quantities using the fitness function, select the five groups with the smallest fitness for crossover. The crossover method is: Select a cut point between two elements, and take all the elements of the parent vector 1 before the cut point and all the elements of the parent vector 2 after the cut point to form a new sub-vector;

[0019] S34. Introduce the concept of crossover rate: Let the crossover rate be ηc (0 < ηc < 1). Before crossover, randomly generate a number i between 0 and 1. If i < ηc, then perform the crossover operation. Otherwise, select any one of the two parent vectors as the next-generation sub-vector. The crossover rate is set to 80% - 90%;

[0020] S35. Add a mutation operator. Let the mutation rate be ηm (0 < ηm < 1). When performing the crossover operation, randomly generate a number j between 0 and 1 for each element of the sub-vector. If j < ηm, then perform a NOT operation on this element. The mutation rate is 0.5% - 1%;

[0021] S36. After performing the operations in steps S31 - S35, one iteration ends. Determine whether there is a combination of physical quantities in the sub - vector whose fitness meets the requirements. If there is, break out of the loop. If not, continue to execute steps S31 - S35.

[0022] Further, in step S3, dynamic programming is added to the genetic algorithm, and the optimal combination of physical quantities obtained by the genetic algorithm improved by dynamic programming is used as the multi - dimensional criterion for circuit breaker faults.

[0023] Further, adding dynamic programming to the genetic algorithm, the specific operation method is: using the number of non - zero elements in the vector as the state quantity S. Different S represents different stages. Among all the vectors generated in a limited number of iterations, start judging from S = 1 whether there is a vector whose fitness meets the requirements. If there is, stop searching. If not, continue to search in the state of S = 1 + n, where n is a positive integer.

[0024] Further, in step S4, according to the optimal combination of physical quantities obtained by the genetic algorithm improved by dynamic programming as the multi - dimensional criterion for circuit breaker faults, the K - means clustering method is used to judge the circuit breaker faults, which specifically includes the following steps:

[0025] S41. Calculate the operating data of the collected multi - fault external characteristics of the circuit breaker in the normal operating state in step S1 as the initial value of the sample center.

[0026] S42. Cluster the multi - dimensional criterion data of the circuit breaker faults collected during the operation of the circuit breaker according to the initial value of the sample center, calculate the Euclidean distance between this group of criterion data and the sample center, and classify it into the closest category.

[0027] S43. If this group of criterion data is classified into the normal operating state data, it is determined that the circuit breaker is working normally. If it is classified into the fault state data, it is determined that the circuit breaker has a fault corresponding to the fault state.

[0028] S44. If the actual operating condition of the circuit breaker is consistent with the judgment result, record this group of criterion data into this category of data and recalculate the sample center.

[0029] S45. Collect the next group of multi - dimensional criterion data of the circuit breaker and repeat steps S41 - S44.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] (1) The genetic algorithm improved by dynamic programming is used to screen multi-dimensional variable information, and the clustering method is used to realize fault diagnosis according to the breaker parameters. The method proposed by the present invention does not target a certain type or several types of faults, but is based on a variety of physical quantity information, and uses the ideas of big data and artificial intelligence to process and analyze the data, obtain the optimal fault criterion, and use this criterion to achieve good clustering analysis.

[0032] (2) The genetic algorithm is used to combine and screen the multi-dimensional physical quantity information of the breaker, and the screening criterion is whether accurate clustering analysis can be achieved. Through the genetic algorithm, the present invention realizes the random combination of different physical quantities, explores the characteristics of each physical quantity one by one and screens out the optimal breaker fault criterion, incorporates different types of signals such as sound, light, and electricity into the breaker fault diagnosis and plays a role at the same time, and can be applied to a variety of fault types, solving the problem that the existing breaker fault diagnosis algorithm is not universal.

[0033] (3) On the basis of the traditional genetic algorithm, the idea of dynamic programming is added to ensure the best clustering effect with the least physical quantities, reduce the dimension of the breaker fault criterion, ensure the rapidity of the algorithm in fault diagnosis, and reduce the burden of data acquisition.

[0034] (4) The K-means clustering method is applied in combination with the genetic algorithm. The principle of the K-means clustering algorithm is simple, the operation is convenient, the calculation speed is fast, and as the running time increases, its sample center will become more and more stable, which can ensure better and better effects in actual operation, reduce the complexity of the algorithm, and ensure the reliability of fault diagnosis.

[0035] (4) It is composed of several common simple algorithms, which is easy to program and realizes the mutual optimization effect between algorithms at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 Shows the flowchart of the breaker fault diagnosis method based on the genetic algorithm and the clustering algorithm provided by the embodiment of the present invention;

[0037] Figure 2 Shows the operation method of the crossover operator;

[0038] Figure 3 Shows the operation method of the mutation operator. DETAILED DESCRIPTION OF THE INVENTION

[0039] To enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0040] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned accompanying drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0041] The following further describes the present invention in detail in conjunction with the accompanying drawings:

[0042] The present invention proposes a circuit breaker fault diagnosis method based on genetic algorithm and clustering algorithm, aiming to solve the problems of lack of universality and single data usage in the existing circuit breaker fault diagnosis methods.

[0043] The circuit breaker fault diagnosis method based on genetic algorithm and clustering algorithm proposed by the present invention screens multi-dimensional variable information through a genetic algorithm improved by dynamic programming, and analyzes the circuit breaker parameters by using the clustering method, so as to realize the universal circuit breaker fault diagnosis applying multi-dimensional fault criteria.

[0044] The flow chart of the circuit breaker fault diagnosis method based on genetic algorithm and clustering algorithm proposed by the present invention is as Figure 1 shown, and mainly includes the following steps:

[0045] S1. Collect the multi-fault external characteristic data of the circuit breaker and perform data preprocessing;

[0046] S2. According to the multi-fault external characteristic data of the circuit breaker obtained by preprocessing in step S1, construct multiple groups of physical quantity combinations, and use the genetic algorithm to evaluate the multiple groups of physical quantity combinations;

[0047] S3. According to the multiple groups of physical quantity combinations constructed in step S2, use the genetic algorithm to construct multi-dimensional fault criteria for the circuit breaker;

[0048] S4. Based on the multi-dimensional criteria for circuit breaker faults constructed in step S3, judge the circuit breaker faults by the K-means clustering method.

[0049] For the circuit breaker fault diagnosis method based on genetic algorithm and clustering algorithm proposed by the present invention, it is necessary to collect the multi-fault external characteristic data of the circuit breaker and perform preprocessing; collect multi-dimensional physical quantity data such as current, voltage and other electrical signals, temperature signals, and sound signals of the circuit breaker and input them into the algorithm. The collected data should include the operation data corresponding to the normal operation state and various common fault states of the circuit breaker, and ensure a sufficient number of groups to ensure the accuracy of the genetic algorithm results.

[0050] For the circuit breaker fault diagnosis method based on genetic algorithm and clustering algorithm, the data preprocessing includes performing data transformation on the collected multi-fault external characteristic data to achieve data normalization, removing outlier samples and integrating the data set into a database.

[0051] For the circuit breaker fault diagnosis method based on genetic algorithm and clustering algorithm proposed by the present invention, the genetic algorithm constructs multi-dimensional criteria for circuit breaker faults, including the following steps:

[0052] S31. Based on the aforementioned preprocessed multi-fault external characteristic data, randomly generate several groups of initial physical quantity combination vectors as the first-generation physical quantity combinations. In the present invention, 10 groups of initial physical quantity combination vectors are used as the first-generation physical quantity combinations: the physical quantity combinations only contain 0 and 1, and the number of vector elements is the same as the number of physical quantities in the preprocessed multi-fault external characteristic data. Among them, 0 means that the physical quantity combination does not contain this physical quantity, that is, this physical quantity is not considered during clustering analysis, and 1 means that the physical quantity combination contains this physical quantity;

[0053] S32. Evaluate the first-generation physical quantity combinations through the fitness function in the genetic algorithm. It is necessary to define the fitness function: use the difference between the sample center of the K-means clustering analysis result of the data set using this physical quantity combination and the actual classification sample center; the K-means clustering method is to calculate the average distance between each group of data and each classification center (that is, the average value vector of each sample in the known classification), add this group of data to the group with the smallest average distance, and update the classification center of this group until all data is classified. In summary, if the actual known classification center vector is For a certain physical quantity combination The fitness function F(x) is expressed as:

[0054]

[0055] Among them, the smaller F(x) is, the better. The population should be iterated in the direction of decreasing F(x);

[0056] S33, after evaluating the first generation of 10 groups of physical quantities using the fitness function, the five groups with the smallest fitness are selected for crossover. The crossover method is as follows: Figure 2 As shown, the specific operation method is: select a point between two elements, take all elements of parent vector 1 before the tangent point and all elements of parent vector 2 after the tangent point, and combine them into a new child vector;

[0057] S34, introduce the concept of crossover rate: suppose the crossover rate is ηc (0<ηc<1), before crossover, randomly generate a number i between 0 and 1, if i<ηc, then crossover operation is performed, otherwise, any one of the two parent vectors is selected as the next generation child vector, and the crossover rate is generally set at about 80%~90%;

[0058] S35. To increase the randomness of the combination, add a mutation operator, such as Figure 3 As shown in the figure: Assume that the mutation rate is ηm (0<ηm<1). When performing the crossover operation, a number j between 0 and 1 is randomly generated for each element of the sub-vector. If j<ηm, a mutation operation is performed, that is, a non-operation is performed on the element. The mutation rate is generally between 0.5% and 1%.

[0059] S36. After completing the above operations, one iteration ends, and it is determined whether there is a combination of physical quantities in the sub-vector whose fitness (the fitness is set by the staff according to the actual situation and is not limited here) meets the requirements. If so, the loop is exited; if not, the iteration is continued (i.e., the technical execution steps S31 to S35) until the condition is met.

[0060] The circuit breaker fault diagnosis method based on genetic algorithm and clustering algorithm proposed by the present invention adds dynamic programming to the genetic algorithm to ensure that the best fault diagnosis effect is performed with the least physical quantity; the optimal physical quantity combination obtained by the genetic algorithm improved by dynamic programming is used as the multi-dimensional criterion of the circuit breaker fault, and the specific operation method is: the number of non-zero elements in the vector is used as the state quantity S, different S represents different stages, and in all vectors generated by a limited number of iterations, starting from S=1, it is judged whether there is a vector whose fitness meets the requirements, if so, the search is stopped, if not, the search is continued in the state of S=2, until a vector meeting the requirements is found in the state of S=1+n, and n is a positive integer.

[0061] The circuit breaker fault diagnosis method based on genetic algorithm and clustering algorithm proposed in the present invention, in step S4, uses the optimal physical quantity combination obtained by the genetic algorithm improved by dynamic programming as the multi-dimensional criterion of circuit breaker fault, and judges the circuit breaker fault by K-means clustering method, which specifically includes the following steps:

[0062] S41. In the calculation step S1, the operation data of the multi-fault external characteristics data of the circuit breaker collected in the normal operation state is used as the initial value of the sample center;

[0063] S42. Cluster the multi-dimensional criterion data of the circuit breaker fault collected during the operation of the circuit breaker according to the initial value of the sample center, calculate the Euclidean distance between this group of criterion data and the sample center, and classify it into the category with the closest distance;

[0064] S43. If this group of criterion data is classified into the normal operation state data, it is determined that the circuit breaker is working normally. If it is classified into the fault state data, it is determined that the circuit breaker has a fault corresponding to the fault state; corresponding countermeasures should be taken;

[0065] S44. If the actual operation condition of the circuit breaker is consistent with the judgment result, this group of criterion data is included in this type of data, and the sample center is recalculated;

[0066] S45. Collect the next group of multi-dimensional criterion data of the circuit breaker and repeat steps S41 to S44.

[0067] The above content is only to illustrate the technical idea of the present invention and cannot limit the protection scope of the present invention. Any modification made on the basis of the technical solution according to the technical idea proposed by the present invention falls within the protection scope of the claims of the present invention.

Claims

1. A circuit breaker fault diagnosis method based on genetic algorithm and clustering algorithm, characterized in that: It includes the following steps: S1. Collect the multi-fault external characteristic data of the circuit breaker and perform data preprocessing; S2. According to the multi-fault external characteristic data of the circuit breaker obtained by preprocessing in step S1, construct multiple groups of physical quantity combinations, and use the genetic algorithm to evaluate the multiple groups of physical quantity combinations; S3. According to the multiple groups of physical quantity combinations constructed in step S2, use the genetic algorithm to construct a multi-dimensional criterion for circuit breaker faults; S4. According to the multi-dimensional criterion for circuit breaker faults constructed in step S3, judge the circuit breaker faults by the K-means clustering method; Using the genetic algorithm to construct a multi-dimensional criterion for circuit breaker faults specifically includes the following steps: S31. Based on the preprocessed multi-fault external characteristic data in step S1, randomly generate 10 groups of initial physical quantity combination vectors as the first-generation physical quantity combinations: the physical quantity combinations only contain 0 and 1, and the number of vector elements is the same as the number of physical quantities in the preprocessed multi-fault external characteristic data, where 0 means that the physical quantity is not included in the physical quantity combination, and 1 means that the physical quantity is included in the physical quantity combination; S32. Evaluate the first-generation physical quantity combinations through the fitness function in the genetic algorithm. The fitness function: the size of the difference between the sample center of the K-means clustering analysis result of the data group using this physical quantity combination and the actual classification sample center; The K-means clustering method calculates the average distance between each group of data and each classification center. The classification center is the mean vector of each sample in the known classification. Add this group of data to the group with the smallest average distance and update the classification center of this group until all data is classified. In summary, if the actual known classification center vector is for a certain combination of physical quantities The fitness function F(x) is expressed as: The population iterates in the direction of decreasing F(x); S33. After evaluating the first-generation 10 groups of physical quantities using the fitness function, select the five groups with the smallest fitness for crossover. The crossover method is: select a cut point between two elements, and take all the elements of the parent vector 1 before the cut point and all the elements of the parent vector 2 after the cut point to form a new sub-vector; S34. Introduce the concept of crossover rate: Let the crossover rate be ηc (0 < ηc < 1). Before crossover, randomly generate a number i between 0 and 1. If i < ηc, then perform the crossover operation. Otherwise, select any one of the two parent vectors as the next-generation sub-vector. The crossover rate is set to 80% - 90%; S35. Add a mutation operator. Let the mutation rate be ηm (0 < ηm < 1). When performing the crossover operation, randomly generate a number j between 0 and 1 for each element of the sub-vector. If j < ηm, then perform a NOT operation on this element. The mutation rate is 0.5% - 1%; S36. After performing the operations in steps S31 - S35, one iteration ends. Judge whether there is a physical quantity combination in the sub-vector whose fitness meets the requirements. If there is, jump out of the loop. If not, continue to execute steps S31 - S35; In step S3, add dynamic programming to the genetic algorithm, and use the optimal physical quantity combination obtained by the genetic algorithm improved by dynamic programming as the multi-dimensional criterion for circuit breaker faults; Add dynamic programming to the genetic algorithm. The specific operation method is as follows: Use the number of non-zero elements in the vector as the state quantity S. Different S values represent different stages. Among all the vectors generated in a limited number of iterations, start judging from S = 1 whether there is a vector with a fitness value meeting the requirements. If there is, stop searching. If not, continue searching in the state of S = 1 + n, where n is a positive integer.

2. A circuit breaker fault diagnosis method based on a genetic algorithm and a clustering algorithm according to claim 1, characterized in that: The multi-fault external characteristic data of the circuit breaker includes the operation data corresponding to the normal operation state and the fault state of the circuit breaker.

3. A circuit breaker fault diagnosis method based on a genetic algorithm and a clustering algorithm according to claim 2, characterized in that: The operation data includes current, voltage, electrical signal, temperature signal, and sound signal.

4. A circuit breaker fault diagnosis method based on a genetic algorithm and a clustering algorithm according to claim 1, characterized in that: The data preprocessing includes performing data transformation on the collected multi-fault external characteristic data of the circuit breaker to achieve data normalization, removing outlier samples, and integrating the data set into a database.

5. A circuit breaker fault diagnosis method based on a genetic algorithm and a clustering algorithm according to claim 1, characterized in that: In step S4, use the optimal physical quantity combination obtained by the genetic algorithm improved by dynamic programming as the multi-dimensional criterion for circuit breaker faults, and judge the circuit breaker faults through the K-means clustering method. The specific steps are as follows: S41. Calculate the operation data of the collected multi-fault external characteristic data of the circuit breaker in the normal operation state in step S1 as the initial value of the sample center; S42. Cluster the multi-dimensional criterion data of the circuit breaker faults collected during the operation of the circuit breaker according to the initial value of the sample center, calculate the Euclidean distance between this group of criterion data and the sample center, and classify it into the category with the closest distance; S43. If this group of criterion data is classified into the normal operation state data, it is determined that the circuit breaker is operating normally. If it is classified into the fault state data, it is determined that the circuit breaker has a fault corresponding to the fault state; S44. If the actual operating condition of the circuit breaker is consistent with the judgment result, include this group of criterion data in this category of data and recalculate the sample center; S45. Collect the next group of multi-dimensional criterion data of the circuit breaker and repeat steps S41 to S44.

Images