Dry-type reactor fault detection method and system based on CPO-LSSVM

Abstract

The invention discloses a dry-type reactor fault detection method and system based on CPO-LSSVM. The method comprises the following steps: S1, obtaining multi-source data of a reactor; s2, removing abnormal data in the multi-source data of the electric reactor; and S3, inputting the complemented multi-source data in the step S2 into a pre-constructed dry-type reactor fault detection model so as to realize the diagnosis of the reactor fault, wherein the dry type reactor fault detection model is constructed through a least square method support vector machine (LSSVM), and parameters of the least square method support vector machine are optimized through a crown porcupine optimizer (CPO) to obtain the dry type reactor fault detection model. The method has the advantages of accurate and reliable fault detection and the like.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of dry-type reactors, and in particular to a dry-type reactor fault detection method and system based on CPO-LSSVM. Background Art

[0002] As an indispensable key equipment in the power system, the stable operation of dry-type reactors is crucial to the safety and efficiency of the power grid. However, in practical applications, dry-type reactors often fail due to various factors, such as turn-to-turn short circuits, insulation aging, increased vibration and noise, etc. These failures not only affect the service life of the reactor, but may also cause significant damage to the entire power system.

[0003] With the increasing demand for electricity and the continuous expansion of the scale of power grids, the importance of dry-type reactors in power systems has become increasingly prominent. Therefore, the research on dry-type reactor fault diagnosis is particularly urgent. At present, although there are some methods for dry-type reactor fault diagnosis, such as detection coil method and space magnetic field detection method, these methods still have problems such as low sensitivity and poor anti-interference ability in practical applications, making it difficult to accurately and timely detect early faults. Summary of the invention

[0004] In view of the technical problems existing in the prior art, the present invention provides a dry-type reactor fault detection method and system based on CPO-LSSVM with accurate and rapid fault detection.

[0005] In order to solve the above technical problems, the technical solution proposed by the present invention is:

[0006] A dry-type reactor fault detection method based on CPO-LSSVM comprises the following steps:

[0007] S1. Obtain multi-source data of reactor;

[0008] S2. Eliminate abnormal data from multi-source data of the reactor;

[0009] S3. Input the multi-source data completed in step S2 into a pre-built dry-type reactor fault detection model to realize the diagnosis of the reactor fault; wherein the dry-type reactor fault detection model is constructed by a least squares support vector machine LSSVM, and the parameters of the least squares support vector machine are optimized by the crown porcupine optimizer CPO.

[0010] Preferably, in step S1 , the multi-source data includes reactor sound signals, temperature, current and voltage data.

[0011] Preferably, in step S2, the acquired abnormal reactor data is eliminated by outlier detection, and the average value of the two digits before and after the eliminated data is used for replacement and completion by continuous interpolation.

[0012] Preferably, in step S3, the specific process of constructing the dry-type reactor fault detection model is:

[0013] S31. Construct objective function:

[0014]

[0015] Where: w is the weight vector, b is the bias term, e is the slack variable, γ is the penalty factor, and the objective function is set to minimize the square loss function of the normal data and abnormal data of the reactor;

[0016] S32, introducing Lagrange multipliers: by introducing Lagrange multipliers αi, the above optimization problem is transformed into an unconstrained optimization problem, and its dual problem is solved;

[0017] According to the KKT condition, we take the derivative of w, b, e and αi at the saddle point and set them to 0 to obtain a linear system. Solving this linear system yields αi and b.

[0018] S33, training the LSSVM model using the training set data, and obtaining the model parameters by solving the above linear system;

[0019] S34. Use CPO optimization algorithm to optimize the parameters of LSSVM to improve the accuracy of the model;

[0020] S35. Use the test set data to evaluate the performance of the model.

[0021] Preferably, in step S34, the parameters of LSSVM include a penalty factor γ and kernel function parameters.

[0022] Preferably, the specific process of step S34 is:

[0023] S341, initialization phase: randomly generate a population, and the dimension of each individual represents a parameter of LSSVM;

[0024] S342, Periodic population reduction: During the optimization process, the population size is periodically reduced according to the loop variable T to speed up convergence and maintain population diversity;

[0025] S343, Exploration phase: Simulating crested porcupine responses to predators;

[0026] S344, Exploitation phase: When a predator approaches, the crested porcupine adopts a defensive strategy;

[0027] S345. Implement CPO for optimization: Alternate exploration and mining phases, balancing the relationship between exploring new areas and exploiting known good solutions to find the optimal LSSVM parameters.

[0028] Preferably, the specific process of step S341 is:

[0029] The dimensions of the crested porcupine population were randomly initialized within the search space for each problem using the formula:

[0030]

[0031] Where N' represents the number of crested porcupine individuals, the population size N', and the individual represents the multi-source data of the reactor after data processing, is the i-th candidate solution in the search space, and are the lower and upper limits of the search range respectively; is a random initialization vector between 0 and 1. The initialization vector can be expressed as follows:

[0032]

[0033] Where X i，j represents the jth position of the ith solution, N' is the number of candidate solutions, and d represents the dimension of the given problem.

[0034] Preferably, in step S342, the mathematical model for periodically reducing the population size is:

[0035]

[0036] Where T is a variable that determines the number of loops, t is the maximum number of function evaluations, % represents the remainder or modulo operator, and N min is the minimum number of individuals in the newly generated population.

[0037] The present invention also discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described above are executed.

[0038] The present invention further discloses a dry-type reactor fault detection system based on CPO-LSSVM, comprising a memory and a processor connected to each other, wherein a computer program is stored in the memory, and when the computer program is run by the processor, the steps of the method described above are executed.

[0039] Compared with the prior art, the advantages of the present invention are:

[0040] The present invention discloses a dry-type reactor fault detection method based on CPO (crown porcupine optimizer)-LSSVM (least squares support vector machine). The method collects multi-source data, such as reactor sound signals, temperature, current, voltage and other data, and processes the data by using outlier detection to filter out abnormal data; then, the data is completed by using continuous interpolation method; then, a dry-type reactor fault detection model based on CPO (crown porcupine optimizer)-LSSVM (least squares support vector machine) is constructed to realize fault identification of dry-type reactor; the above fault detection is accurate and fast, highly reliable and highly sensitive, which can not only effectively improve the operating reliability of the reactor, reduce the economic losses and safety hazards caused by faults, but also provide strong guarantee for the safe and stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 The figure is a flow chart of a dry-type reactor fault detection method in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0043] like Figure 1 As shown, an embodiment of the present invention provides a dry-type reactor fault detection method based on CPO (Crown Porcupine Optimizer)-LSSVM (Least Squares Support Vector Machine), which specifically includes the steps of:

[0044] S1. Acquire multi-source data of the reactor, where the multi-source data includes sound signals, temperature, current, voltage and other data of the reactor;

[0045] S2, using outliers and continuous interpolation method to remove and complete multi-source data of reactors;

[0046] S3. Input the multi-source data completed in step S2 into a pre-built dry-type reactor fault detection model to realize the diagnosis of the reactor fault; wherein the dry-type reactor fault detection model is constructed by a least squares support vector machine (LSSVM), and the parameters of the least squares support vector machine (such as penalty factor γ, kernel function parameters, etc.) are optimized by a crown porcupine optimizer (CPO).

[0047] Specifically, in step S1, when acquiring multi-source data of the reactor, multi-channel acquisition is adopted through a detection device applied on site, and each measuring point is dispersed as much as possible to ensure the comprehensive reliability of the multi-source data of the reactor.

[0048] Specifically, in step S2, outlier detection is used to remove the acquired abnormal reactor data, and the continuous interpolation rule uses the average value of the two digits before and after the removed data for replacement and completion.

[0049] Specifically, in step S3, a dry-type reactor fault detection model based on CPO (Crown Porcupine Optimizer)-LSSVM (Least Squares Support Vector Machine) is constructed; wherein LSSVM (Least Squares Support Vector Machine) is a variant of the support vector machine (SVM) algorithm, which performs classification prediction by minimizing the regularized square loss function. The specific steps of constructing the dry-type reactor fault detection model are as follows:

[0050] S31. Constructing the objective function

[0051]

[0052] Among them: w is the weight vector, b is the bias term, e is the slack variable, γ is the penalty factor, yi is the multi-source data set collected by the reactor as the sample label, the objective function is set to the square loss function of the normal data and abnormal data of the reactor as the target, and the sample labels are mainly normal samples and fault samples, belonging to the category yi respectively;

[0053] S32. Introducing Lagrange multipliers: By introducing Lagrange multipliers αi, the above optimization problem is transformed into an unconstrained optimization problem, and its dual problem is solved.

[0054] The Lagrange function is: L(w,b,e;αi)=J(w,e)-i=1∑Nαi{yi(wTφ(xi)+b)-1+ei}(2)

[0055] Solution according to KKT conditions: According to the Karush-Kuhn-Tucker (KKT) conditions, we differentiate w, b, e and αi at the saddle point and set them to 0, and obtain a linear system. Solving this linear system can obtain αi and b.

[0056] S33, using the training set data (the training data set and the test data set are allocated according to the acquisition data set 8:2) to train the LSSVM model, and obtain the model parameters by solving the above linear system;

[0057] S34. Use the CPO optimization algorithm to optimize the parameters of LSSVM (such as penalty factor γ, kernel function parameters, etc.) to improve the accuracy of the model;

[0058] S35. Use the test set data to evaluate the performance of the model, including indicators such as accuracy, recall, and F1 score.

[0059] S36. Obtain a fault identification result.

[0060] In step S34, the specific steps of optimizing the parameters of LSSVM (such as penalty factor γ, kernel function parameters, etc.) using the CPO optimization algorithm are as follows:

[0061] S341, initialization phase:

[0062] The dimensions of the crested porcupine population were randomly initialized within the search space for each problem using the formula:

[0063]

[0064] Where N' represents the number of crested porcupine individuals (population size N', individuals represent multi-source data of reactors after data processing), is the i-th candidate solution in the search space, and are the lower and upper limits of the search range respectively; is a random initialization vector between 0 and 1. The initialization vector can be expressed as follows:

[0065]

[0066] Where X i，j represents the jth position of the ith solution, N' is the number of candidate solutions, and d represents the dimension of the given problem.

[0067] S342, Periodic population reduction technology

[0068] The strategy simulates that not all crested porcupines will activate the defense mechanism, but only those that are threatened. Therefore, in this strategy, some crested porcupines are tried to be taken out of the population during the optimization process to speed up the convergence, and put them back into the population to improve diversity and avoid falling into local minima; the loop determines the number of times the process is performed during the optimization process based on the loop variable T. The mathematical model of periodic reduction of scale is as follows:

[0069]

[0070] Where T is a variable that determines the number of loops, t is the maximum number of function evaluations, % represents the remainder or modulo operator, and N min is the minimum number of individuals in the newly generated population, so the population size cannot be less than N min .

[0071] S343, Exploration Stage

[0072] According to the crested porcupine's defensive behavior, when the predator is far away, the targeting strategy:

[0073] The visual and sound strategies allow crested porcupines to scare predators away, unlike the scent and physical attack strategies, where predators come very close. Through these defense mechanisms, the response of crested porcupines to predator release after implementing defensive behaviors was simulated to explore the search space to reach promising areas, which may involve near-optimal solutions.

[0074] When the crested porcupine becomes aware of the predator, the predator has two options, either move towards it or away from it. In the first option, the distance between the predator and the crested porcupine decreases as the predator moves towards the crested porcupine. This option encourages exploration of the area between the predator and the crested porcupine in an attempt to speed up the convergence rate. In contrast, in the second option, the distance between the predator and the crested porcupine is maximized as the predator chooses to move away. This strategy encourages exploration of distant areas to identify unvisited areas, which may involve the desired solution. A normal distribution is used to generate random values ​​to mathematically simulate these options.

[0075] If these random values ​​are less than 1 or greater than -1, they are encouraged to get closer to the crested porcupine. Otherwise, the predator will move away from the crested porcupine. In general, the mathematical simulation of this behavior is as follows:

[0076]

[0077] in, is the best solution for function evaluation t, is a vector generated between the current crested porcupine and a randomly selected crested porcupine from the population to represent the position of the predator at iteration t, τ1 is a random number based on a normal distribution, and τ2 is a random value in the interval [0, 1]; generate The mathematical formula is as follows:

[0078]

[0079] Where r is a random number between [1, N].

[0080] S344, mining stage

[0081] According to the defensive behavior of the crested porcupine, when a predator approaches, it has two defense strategies, namely, scent and physical attack strategies. These strategies refer to the exploitation of promising areas, which are dedicated to the exploitation search (local). The search space is used and the optimal solution is approached based on the scent and physical attack strategies. In this strategy, the crested porcupine secretes a foul smell that spreads in its surrounding area to deter predators from approaching it. To mathematically simulate this behavior, the following formula is proposed:

[0082]

[0083] where r1 and r2 are two random integers between [1, N], and τ3 is a random value generated between 0 and 1. Where r3 is a random number between [1, N], and δ is a parameter used to control the search direction, is the position of the i-th individual at iteration t, γ t is the defense factor defined using the equation; is the best solution obtained, representing CP, is the position of the i-th individual at iteration t, representing the predator at that position, α is the convergence speed factor, and τ4 is a random value within the interval [0, 1], is the average force of CP affecting the i-th predator; γ t is a random value within the interval [0, 1], is the odor diffusion factor, using the equation:

[0084]

[0085] where represents the objective function value of the i-th individual at iteration t, γ t is a small value to avoid division by zero, is a carrier including a number randomly generated between 0 and 1, rand is a variable including a number randomly generated between 0 and 1, N is the population size, t is the number of the current iteration, and t max is the maximum number of iterations.

[0086] S345. Implement CPO for optimization

[0087] Exploration phase. Each phase has a defense strategy for the crested porcupine.

[0088] The exploration phase includes visual and sound defense strategies. The exploration phase includes odor and physical attack defense strategies. In the exploration phase, when τ6 < τ7, a trade-off is made between the line-of-sight defense strategy and the sound defense strategy, where τ6 and τ7 are two values randomly generated between 0 and 1. In the exploitation phase, when τ8 < Tf, the odor defense strategy is executed; otherwise, the physical attack defense strategy is executed. Tf is a predetermined constant value between 0 and 1 used to make a trade-off between local exploitation (the third defense mechanism) and global defense. A trade-off is made between exploitation (the third defense mechanism) and global exploitation (the fourth defense mechanism). If this factor takes too small a value, it may lead to stagnation at a local minimum, and if it takes too large a value, it may lead to too slow a convergence speed.

[0089]

[0090] where, τ6, τ7, τ8, τ9, and τ 10 are numerical values randomly generated between 0 and 1. In each generation, Evaluation is performed using an objective function that needs to be minimized or maximized to achieve the desired outcome.

[0091] The present invention discloses a dry-type reactor fault detection method based on CPO (crown porcupine optimizer)-LSSVM (least squares support vector machine). The method collects multi-source data, such as reactor sound signals, temperature, current, voltage and other data, and processes the data by using outlier detection to filter out abnormal data; then, the data is completed by using continuous interpolation method; then, a dry-type reactor fault detection model based on CPO (crown porcupine optimizer)-LSSVM (least squares support vector machine) is constructed to realize fault identification of dry-type reactor; the above fault detection is accurate and fast, highly reliable and highly sensitive, which can not only effectively improve the operating reliability of the reactor, reduce the economic losses and safety hazards caused by faults, but also provide strong guarantee for the safe and stable operation of the power system.

[0092] The present invention also discloses a computer-readable storage medium, on which a computer program is stored, and the computer program executes the steps of the above method when it is run by a processor. The present invention further discloses a dry-type reactor fault detection system based on CPO-LSSVM, including a memory and a processor connected to each other, the memory storing a computer program, and the computer program executes the steps of the above method when it is run by the processor. The medium and system of the present invention correspond to the above method and also have the advantages described in the above method.

[0093] The present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiment when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable storage media include: any entity or device that can carry computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. The memory is used to store computer programs and / or modules, and the processor implements various functions by running or executing computer programs and / or modules stored in the memory, and calling data stored in the memory. The memory may include high-speed random access memory and may also include non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0094] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A dry-type reactor fault detection method based on CPO-LSSVM, characterized in that: Includes steps: S1. Obtain multi-source data of reactors; S2. Eliminate abnormal data from multi-source data of the reactor; S3. Input the multi-source data completed in step S2 into a pre-built dry-type reactor fault detection model to realize the diagnosis of the reactor fault; wherein the dry-type reactor fault detection model is constructed by a least squares support vector machine LSSVM, and the parameters of the least squares support vector machine are optimized by the crown porcupine optimizer CPO.

2. The dry-type reactor fault detection method based on CPO-LSSVM according to claim 1 is characterized in that: In step S1 , the multi-source data includes reactor sound signals, temperature, current and voltage data.

3. The dry-type reactor fault detection method based on CPO-LSSVM according to claim 1 is characterized in that: In step S2, the acquired abnormal reactor data is eliminated by outlier detection, and the average value of the two digits before and after the eliminated data is used for replacement and completion by continuous interpolation.

4. The dry-type reactor fault detection method based on CPO-LSSVM according to claim 1, 2 or 3, characterized in that: In step S3, the specific process of constructing the dry-type reactor fault detection model is as follows: S31. Construct objective function: Where: w is the weight vector, b is the bias term, e is the slack variable, γ is the penalty factor, and the objective function is set to minimize the square loss function of the normal data and abnormal data of the reactor; S32, introducing Lagrange multipliers: by introducing Lagrange multipliers αi, the above optimization problem is transformed into an unconstrained optimization problem, and its dual problem is solved; According to the KKT condition, we take the derivative of w, b, e and αi at the saddle point and set them to 0 to obtain a linear system. Solving this linear system yields αi and b. S33, training the LSSVM model using the training set data, and obtaining the model parameters by solving the above linear system; S34. Use CPO optimization algorithm to optimize the parameters of LSSVM to improve the accuracy of the model; S35. Use the test set data to evaluate the performance of the model.

5. The dry-type reactor fault detection method based on CPO-LSSVM according to claim 4 is characterized in that: In step S34, the parameters of LSSVM include penalty factor γ and kernel function parameters.

6. The dry-type reactor fault detection method based on CPO-LSSVM according to claim 4 is characterized in that: The specific process of step S34 is: S341, initialization phase: randomly generate a population, and the dimension of each individual represents a parameter of LSSVM; S342, Periodic population reduction: During the optimization process, the population size is periodically reduced according to the loop variable T to speed up convergence and maintain population diversity; S343, Exploration phase: Simulating crested porcupine responses to predators; S344, Exploitation phase: When a predator approaches, the crested porcupine adopts a defensive strategy; S345. Implement CPO for optimization: Alternate exploration and mining phases, balancing the relationship between exploring new areas and exploiting known good solutions to find the optimal LSSVM parameters.

7. The dry-type reactor fault detection method based on CPO-LSSVM according to claim 6 is characterized in that: The specific process of step S341 is: The dimensions of the crested porcupine population were randomly initialized within the search space for each problem using the formula: Where N' represents the number of crested porcupine individuals, the population size N', and the individual represents the multi-source data of the reactor after data processing, is the i-th candidate solution in the search space, and are the lower and upper limits of the search range respectively; is a random initialization vector between 0 and 1. The initialization vector is represented as follows: Where X i，j represents the jth position of the ith solution, N' is the number of candidate solutions, and d represents the dimension of the given problem.

8. The dry-type reactor fault detection method based on CPO-LSSVM according to claim 7, characterized in that: In step S342, the mathematical model for periodically reducing the population size is: Where T is a variable that determines the number of loops, t is the maximum number of function evaluations, % represents the remainder or modulo operator, and N min is the minimum number of individuals in the newly generated population.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computer program performs the steps of the method according to any one of claims 1 to 8.

10. A dry-type reactor fault detection system based on CPO-LSSVM, comprising a memory and a processor connected to each other, wherein a computer program is stored in the memory, characterized in that: When the computer program is executed by a processor, the computer program performs the steps of the method according to any one of claims 1 to 8.

Images