A temporary grounding line defect diagnosis model construction method and device

Abstract

The application discloses a temporary grounding wire defect diagnosis model construction method and device, the temporary grounding wire defect diagnosis model construction method provided by the application extracts sample characteristic quantities of sample waveforms by taking historical ultrasonic detection waveforms as the sample waveforms; according to the sample characteristic quantities, the model structure of a BP neural network is initialized in combination with the number of input characteristic quantities and the number of output results; the weights and thresholds of the BP neural network are optimized through an improved CSO algorithm of a hierarchical ring topology structure, the BP neural network is trained and tested according to the sample characteristic quantities, and a trained temporary grounding wire defect detection model is obtained; when defect detection of a temporary grounding wire is performed, the defect diagnosis model can be input with corresponding ultrasonic detection waveforms, so that defect diagnosis of the temporary grounding wire defect of the defect diagnosis model can be completed, and the technical problem that a traditional manual detection method excessively depends on the experience and quality of inspectors and is prone to incorrect judgment is solved.

Description

Technical Field

[0001] This application relates to the field of temporary grounding wire defect detection technology, and in particular to a method and apparatus for constructing a temporary grounding wire defect diagnostic model. Background Technology

[0002] With the rapid development of the social economy, users' electricity demand is constantly increasing. The existing power supply and distribution network can no longer meet the increased electricity demand. Therefore, power supply departments are upgrading existing power lines or adding new power sources to meet users' needs. When upgrading lines or requiring load cut-offs, it is often necessary to cut off power to the line or the user whose power source needs to be changed. To ensure the safety of on-site workers, relevant safety measures need to be implemented at both ends of the construction work area. Specifically, temporary grounding wires are installed on both sides of the work site to ensure the safety of personnel within the work area.

[0003] The temporary grounding wire currently used by the power supply department has a cross-section of 25mm². 2 The copper temporary grounding wire is connected to the temporary grounding wire and the grounding grid at both ends using special connectors. To better protect the grounding wire, a transparent sheath is currently used as a protective sleeve for the copper temporary grounding wire. To ensure the operational status of the temporary grounding wire, the power supply department conducts regular inspections to check its operational status. If the inspection fails, it needs to be scrapped. However, in actual operation, due to differences in the work experience and skill level of the staff, it is often impossible to accurately check the status of the temporary grounding wire, which can easily lead to the omission of defective temporary grounding wires, thus leaving safety hazards. Summary of the Invention

[0004] This application provides a method and apparatus for constructing a diagnostic model for temporary grounding wire defects, which solves the technical problem that the accuracy of existing temporary grounding wire inspection methods is easily affected by differences in the work experience and skill level of the staff.

[0005] To address the aforementioned technical problems, the first aspect of this application provides a method for constructing a temporary grounding wire defect diagnosis model, comprising:

[0006] The historical ultrasonic waveform of the temporary grounding wire is obtained as a sample waveform, and the sample feature quantity of the sample waveform is extracted.

[0007] Based on the sample features, and combining the number of input features and the number of output results, initialize the model structure of the BP neural network;

[0008] The weights and thresholds of the BP neural network are optimized using the improved CSO algorithm with a hierarchical ring topology.

[0009] Based on the sample features, the BP neural network is trained and tested. When the training accuracy of the BP neural network meets the preset accuracy requirements, training is stopped and a temporary grounding wire defect diagnosis model is output. If the training accuracy of the BP neural network does not meet the preset accuracy requirements, the improved CSO algorithm with a hierarchical ring topology is used to optimize the weights and thresholds of the BP neural network.

[0010] Preferably, optimizing the weights and thresholds of the BP neural network using the improved CSO algorithm with a hierarchical ring topology specifically includes:

[0011] Based on the BP neural network, initialize the particle swarm, rule tree, and number of layers;

[0012] A preset number of particles are randomly assigned to various rings in the rule tree to form multiple populations;

[0013] According to the greedy strategy, cross operations are performed on each population in the same layer and the population is updated.

[0014] Based on the updated populations, the fitness of each particle in the population is calculated using a fitness function, and the quality evaluation value of the population is determined based on the fitness.

[0015] Based on the association between the various populations, the superiority and inferiority evaluation values ​​of the populations and their offspring populations are compared. If the offspring population is superior to the parent population based on the comparison results, the offspring population and the parent population are swapped.

[0016] If the number of iterations in the optimization process reaches a preset maximum value, the optimization stops, and the weights and thresholds of the optimized BP neural network are obtained based on the current optimal population. If the number of iterations does not reach the preset maximum value, the process returns to performing crossover operations on each population in the same layer and updating the population according to the greedy strategy.

[0017] Preferably, the extraction of sample features from the sample waveform specifically includes:

[0018] The sample waveform is denoised, and wavelet packet extraction is used to extract features from the denoised sample waveform to obtain sample features.

[0019] Preferably, the sample features specifically include: signal mean, standard deviation, slope coefficient, kurtosis coefficient, clearing index, peak value index, impulse index, and shape index.

[0020] Preferably, the step of training and testing the BP neural network based on the sample features, stopping training when the training accuracy of the BP neural network meets the preset accuracy requirement, and outputting a temporary grounding wire defect diagnosis model; if the training accuracy of the BP neural network does not meet the preset accuracy requirement, then returning to the improved CSO algorithm with a hierarchical ring topology to optimize the weights and thresholds of the BP neural network specifically includes:

[0021] Based on the sample features, the BP neural network is trained and tested, and the training accuracy of the BP neural network is calculated through a preset optimization objective function. When the training accuracy of the BP neural network meets the preset accuracy requirement, training is stopped and a temporary grounding wire defect diagnosis model is output. If the training accuracy of the BP neural network does not meet the preset accuracy requirement, the improved CSO algorithm with a hierarchical ring topology is used to optimize the weights and thresholds of the BP neural network.

[0022] The objective function is specifically:

[0023]

[0024] In the formula, f(i) represents the training accuracy of the BP neural network, N is the number of training data, and P i P represents the defect identification result output by the BP neural network for the i-th training data item. i m This represents the actual defect identification result corresponding to the i-th training data item.

[0025] Meanwhile, the second aspect of this application also provides a device for constructing a temporary grounding wire defect diagnosis model, comprising:

[0026] The sample feature acquisition unit is used to acquire the historical ultrasonic detection waveform of the temporary grounding wire as a sample waveform and extract the sample feature quantity of the sample waveform.

[0027] The neural network initialization unit is used to initialize the model structure of the BP neural network based on the sample feature quantity, the number of input feature quantities, and the number of output results;

[0028] The neural network parameter optimization unit is used to optimize the weights and thresholds of the BP neural network using an improved CSO algorithm with a hierarchical ring topology.

[0029] The defect diagnosis model training unit is used to train and test the BP neural network based on the sample feature quantity. When the training accuracy of the BP neural network meets the preset accuracy requirement, the training stops and the temporary grounding wire defect diagnosis model is output. If the training accuracy of the BP neural network does not meet the preset accuracy requirement, the improved CSO algorithm with a hierarchical ring topology is used to optimize the weights and thresholds of the BP neural network.

[0030] Preferably, the neural network parameter optimization unit is specifically used for:

[0031] Based on the BP neural network, initialize the particle swarm, rule tree, and number of layers;

[0032] A preset number of particles are randomly assigned to various rings in the rule tree to form multiple populations;

[0033] According to the greedy strategy, cross operations are performed on each population in the same layer and the population is updated.

[0034] Based on the updated populations, the fitness of each particle in the population is calculated using a fitness function, and the quality evaluation value of the population is determined based on the fitness.

[0035] Based on the association between the various populations, the superiority and inferiority evaluation values ​​of the populations and their offspring populations are compared. If the offspring population is superior to the parent population based on the comparison results, the offspring population and the parent population are swapped.

[0036] If the number of iterations in the optimization process reaches a preset maximum value, the optimization stops, and the weights and thresholds of the optimized BP neural network are obtained based on the current optimal population. If the number of iterations does not reach the preset maximum value, the process returns to performing crossover operations on each population in the same layer and updating the population according to the greedy strategy.

[0037] Preferably, the sample feature acquisition unit is specifically used for:

[0038] The historical ultrasonic detection waveform of the temporary grounding wire is obtained as a sample waveform. The sample waveform is denoised, and the feature quantity is extracted from the denoised sample waveform using wavelet packet extraction to obtain the sample feature quantity.

[0039] Preferably, the sample features specifically include: signal mean, standard deviation, slope coefficient, kurtosis coefficient, clearing index, peak value index, impulse index, and shape index.

[0040] Preferably, the defect diagnosis model training unit is specifically used for:

[0041] Based on the sample features, the BP neural network is trained and tested, and the training accuracy of the BP neural network is calculated through a preset optimization objective function. When the training accuracy of the BP neural network meets the preset accuracy requirement, training is stopped and a temporary grounding wire defect diagnosis model is output. If the training accuracy of the BP neural network does not meet the preset accuracy requirement, the improved CSO algorithm with a hierarchical ring topology is used to optimize the weights and thresholds of the BP neural network.

[0042] The objective function is specifically:

[0043]

[0044] In the formula, f(i) represents the training accuracy of the BP neural network, N is the number of training data, and P i P represents the defect identification result output by the BP neural network for the i-th training data item. i m This represents the actual defect identification result corresponding to the i-th training data item.

[0045] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0046] The temporary grounding wire defect diagnosis model construction method provided in this application uses historical ultrasonic detection waveforms as sample waveforms to extract sample feature quantities from the sample waveforms; based on the sample feature quantities, combined with the number of input feature quantities and the number of output results, the model structure of a BP neural network is initialized; the weights and thresholds of the BP neural network are optimized through an improved CSO algorithm with a hierarchical ring topology; the BP neural network is trained and tested based on the sample feature quantities to obtain a trained temporary grounding wire defect detection model. When detecting defects in temporary grounding wires, the corresponding ultrasonic detection waveform is input into this defect diagnosis model to complete the diagnosis of temporary grounding wire defects. This solves the technical problem that traditional manual inspection methods rely too heavily on the experience and quality of inspectors, leading to subjective inspection results and easy errors in judgment. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application 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 this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1This is a flowchart illustrating a method for constructing a temporary grounding wire defect diagnosis model provided in this application.

[0049] Figure 2 This is a flowchart illustrating step 103 of the method for constructing a temporary grounding wire defect diagnosis model provided in this application.

[0050] Figure 3 This is a schematic diagram of the rule tree architecture in the temporary grounding wire defect diagnosis model construction method provided in this application.

[0051] Figure 4 This is a schematic diagram of a temporary grounding wire defect diagnosis model construction device provided in this application. Detailed Implementation

[0052] This application provides a method and apparatus for constructing a diagnostic model for temporary grounding wire defects, which solves the technical problem that the accuracy of existing temporary grounding wire inspection methods is low due to the influence of differences in the work experience and skill level of the staff.

[0053] To make the inventive objectives, features, and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0054] Please see Figure 1 This application provides a method for constructing a temporary grounding wire defect diagnosis model, including:

[0055] Step 101: Obtain the historical ultrasonic test waveform of the temporary grounding wire as a sample waveform, and extract the sample feature quantity of the sample waveform.

[0056] It should be noted that the waveforms of temporary grounding wires measured in the past are collected as sample data. These sample data can be divided into two parts: training data and test data. The sample feature quantities of the sample waveforms are extracted as input quantities for constructing the temporary grounding wire defect diagnosis model.

[0057] Step 102: Based on the sample features, combined with the number of input features and the number of output results, initialize the model structure of the BP neural network.

[0058] Then, based on the sample features obtained in the previous step, and combining the number of input features and the number of output results, the model structure of the BP neural network is initialized.

[0059] Step 103: Optimize the weights and thresholds of the BP neural network using the improved CSO algorithm with a hierarchical ring topology.

[0060] It should be noted that the traditional CSO algorithm is a commonly used swarm optimization algorithm that has been applied to solve various power system optimization problems. Similar to traditional optimization algorithms, the CSO algorithm is prone to getting trapped in local optima in the later stages, preventing the attainment of the global optimum and exhibiting slow optimization speed. To address this issue, this application proposes a hierarchical ring-shaped dynamic topology strategy. An improved CSO algorithm using a hierarchical ring-shaped topology structure optimizes the weights and thresholds of the BP neural network, thereby mitigating the premature convergence problem. Simultaneously, to ensure the algorithm's operational efficiency, the accuracy of the BP neural network is further improved.

[0061] Step 104: Train and test the BP neural network based on the sample features. When the training accuracy of the BP neural network meets the preset accuracy requirements, stop training and output the temporary grounding wire defect diagnosis model. If the training accuracy of the BP neural network does not meet the preset accuracy requirements, return to the improved CSO algorithm with a hierarchical ring topology structure to optimize the weights and thresholds of the BP neural network.

[0062] Next, based on the BP neural network whose weights and thresholds were optimized in step 103, the BP neural network is trained and tested using the sample features classified in step 101. When the training accuracy of the BP neural network meets the preset accuracy requirements, training is stopped and a temporary grounding wire defect diagnosis model is output for the diagnosis of temporary grounding wire defects. If the training accuracy of the BP neural network does not meet the preset accuracy requirements, the process returns to step 103, and the weights and thresholds of the BP neural network are re-optimized. After optimization, training and testing are performed again until the training accuracy of the BP neural network meets the preset accuracy requirements.

[0063] The defect diagnosis process for temporary grounding wires can be found in the following example:

[0064] 1) Use ultrasonic testing equipment to scan the copper conductor of the temporary grounding wire to be diagnosed and obtain the waveform diagram to be tested;

[0065] 2) Preprocess the obtained detection waveform to extract waveform features. Specifically, the detection waveform is denoised, and wavelet packets are used to extract features.

[0066] 3) Input the extracted feature quantities into the trained temporary grounding wire defect diagnosis model, so as to obtain the corresponding defect diagnosis results through the calculation of the temporary grounding wire defect diagnosis model.

[0067] The temporary grounding wire defect diagnosis model construction method provided in this application uses historical ultrasonic detection waveforms as sample waveforms to extract sample feature quantities. Based on the sample feature quantities, combined with the number of input feature quantities and the number of output results, the model structure of the BP neural network is initialized. Through an improved CSO algorithm with a hierarchical ring topology, the weights and thresholds of the BP neural network are optimized. Based on the sample feature quantities, the BP neural network is trained and tested to obtain a trained temporary grounding wire defect detection model. When performing defect detection on temporary grounding wires, the corresponding ultrasonic detection waveform is input into this defect diagnosis model to complete the diagnosis of temporary grounding wire defects. This solves the technical problem that traditional manual inspection methods rely too much on the experience and quality of inspectors, resulting in subjective inspection results and easy misjudgment.

[0068] The above is a detailed description of a basic embodiment of a temporary grounding wire defect diagnosis model construction method provided by this application. The following is a detailed description of a further embodiment of a temporary grounding wire defect diagnosis model construction method provided by this application based on the above embodiment.

[0069] Please see Figure 2 Furthermore, step 103 mentioned in the previous embodiment specifically includes the following steps:

[0070] Step 1031: Initialize the particle swarm, rule tree, and number of layers based on the BP neural network.

[0071] In this embodiment, the particles represent the parameters to be optimized in the BP neural network.

[0072] Step 1032: Randomly distribute a preset number of particles into various rings in the rule tree to form multiple populations.

[0073] Among them, the generated rule tree and population relationship are as follows: Figure 3 As shown.

[0074] Step 1033: According to the greedy strategy, perform cross-operations on each population in the same layer and update the population.

[0075] Step 1034: Based on the updated populations, calculate the fitness of each particle in the population using the fitness function, and determine the population's quality evaluation value based on the fitness.

[0076] Step 1035: Based on the relationship between each population, compare the superiority and inferiority evaluation values ​​of the population and its offspring population. If the offspring population is superior to the parent population, then the offspring population and the parent population are swapped.

[0077] Step 1036: Based on the number of iterations in the optimization process, if the number of iterations reaches the preset maximum value, then stop the optimization and obtain the weights and thresholds of the optimized BP neural network based on the current optimal population. If the number of iterations does not reach the preset maximum value, then return to step 1033.

[0078] It should be noted that after the iterations from step 1033 to step 1036, better particles will be gradually adjusted to the top layer of the rule tree. If the number of iterations reaches the preset maximum value, i.e. the iteration termination condition is met, the optimization stops, and the weights and thresholds of the optimized BP neural network are obtained based on the current optimal population. If the iteration termination condition is not met, the process returns to step 1033 and enters the next cycle of optimization iteration.

[0079] Furthermore, the sample feature quantities for extracting the sample waveform in step 101 specifically include:

[0080] The sample waveform is denoised, and the wavelet packet extraction method is used to extract the feature values ​​of the denoised sample waveform to obtain the sample feature values.

[0081] Furthermore, the specific sample characteristics include: signal mean, standard deviation, slope coefficient, kurtosis coefficient, clearing index, peak value index, impulse index, and shape index.

[0082] It should be noted that, for data preprocessing, this application employs a wavelet thresholding denoising method, the main steps of which include: selecting a wavelet basis and determining the number of decomposition layers; determining the threshold rule; and determining the threshold function. After denoising the detected waveform, wavelet packet transform is used to extract the feature quantities of the detected waveform, mainly including: signal mean, standard deviation, slope coefficient, kurtosis coefficient, cleanliness index, peak value index, impulse index, and shape index. The extracted feature quantities are used as the input to the neural network.

[0083] Furthermore, step 104 mentioned in the previous embodiment specifically includes the following steps:

[0084] Based on the sample feature quantity, the BP neural network is trained and tested, and the training accuracy of the BP neural network is calculated through the preset optimization objective function. When the training accuracy of the BP neural network meets the preset accuracy requirement, the training stops and the temporary grounding wire defect diagnosis model is output. If the training accuracy of the BP neural network does not meet the preset accuracy requirement, the process returns to step 103.

[0085] The objective function is as follows:

[0086]

[0087] In the formula, f(i) represents the training accuracy of the BP neural network, N is the number of training data, and P i P represents the defect identification result output by the BP neural network for the i-th training data item. i m This represents the actual defect identification result corresponding to the i-th training data item.

[0088] Furthermore, this solution constructs a temporary grounding wire defect diagnosis model based on ultrasonic testing combined with artificial intelligence. This model enables dynamic detection of the operating status of the temporary grounding conductor, facilitating timely and accurate information about the actual condition of the grounding wire by the operating unit. Simultaneously, it addresses the shortcomings of traditional manual inspection methods, which rely excessively on the experience and skills of inspectors, leading to subjective results. This avoids errors in judgment caused by human subjectivity and improves the scientific rigor of defect diagnosis.

[0089] The above content is a detailed description of a further embodiment of the temporary grounding wire defect diagnosis model construction method provided in this application. The following is a detailed description of an embodiment of the temporary grounding wire defect diagnosis model construction device provided in this application.

[0090] Please see Figure 4 This embodiment also provides a device for constructing a temporary grounding wire defect diagnosis model, including:

[0091] The sample feature acquisition unit 201 is used to acquire the historical ultrasonic test waveform of the temporary grounding wire as a sample waveform and extract the sample feature quantity of the sample waveform.

[0092] The neural network initialization unit 202 is used to initialize the model structure of the BP neural network based on the sample feature quantity, the number of input feature quantities, and the number of output results.

[0093] The neural network parameter optimization unit 203 is used to optimize the weights and thresholds of the BP neural network using an improved CSO algorithm with a hierarchical ring topology.

[0094] The defect diagnosis model training unit 204 is used to train and test the BP neural network based on the sample feature quantity. When the training accuracy of the BP neural network meets the preset accuracy requirement, the training stops and the temporary grounding wire defect diagnosis model is output. If the training accuracy of the BP neural network does not meet the preset accuracy requirement, the improved CSO algorithm with hierarchical ring topology is used to optimize the weights and thresholds of the BP neural network.

[0095] Furthermore, the neural network parameter optimization unit 202 is specifically used for:

[0096] Initialize the particle swarm, rule tree, and number of layers based on a BP neural network;

[0097] A preset number of particles are randomly assigned to various rings in the rule tree to form multiple populations.

[0098] According to the greedy strategy, cross operations are performed on each population in the same layer and the population is updated.

[0099] Based on the updated populations, the fitness of each particle in the population is calculated using the fitness function, and the population's quality is evaluated based on the fitness.

[0100] Based on the relationships between the populations, the superiority and inferiority evaluation values ​​of the populations and their offspring populations are compared. If the offspring population is superior to the parent population, the offspring population is swapped with the parent population.

[0101] If the number of iterations in the optimization process reaches a preset maximum value, the optimization stops, and the weights and thresholds of the optimized BP neural network are obtained based on the current optimal population. If the number of iterations does not reach the preset maximum value, the process returns to performing crossover operations on each population in the same layer and updating the population according to the greedy strategy.

[0102] Furthermore, the sample feature acquisition unit is specifically used for:

[0103] Historical ultrasonic waveforms of temporary grounding wires are obtained as sample waveforms. The sample waveforms are then denoised, and wavelet packet extraction is used to extract features from the denoised sample waveforms to obtain sample features.

[0104] Furthermore, the specific sample characteristics include: signal mean, standard deviation, slope coefficient, kurtosis coefficient, clearing index, peak value index, impulse index, and shape index.

[0105] Furthermore, the defect diagnosis model training unit is specifically used for:

[0106] Based on the sample features, the BP neural network is trained and tested, and the training accuracy of the BP neural network is calculated through a preset optimization objective function. When the training accuracy of the BP neural network meets the preset accuracy requirements, training stops and a temporary grounding wire defect diagnosis model is output. If the training accuracy of the BP neural network does not meet the preset accuracy requirements, the improved CSO algorithm with a hierarchical ring topology is used to optimize the weights and thresholds of the BP neural network.

[0107] The objective function is as follows:

[0108]

[0109] In the formula, f(i) represents the training accuracy of the BP neural network, N is the number of training data, and P iP represents the defect identification result output by the BP neural network for the i-th training data item. i m This represents the actual defect identification result corresponding to the i-th training data item.

[0110] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the terminals, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0111] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0112] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0113] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0114] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0115] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0116] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 this application.

Claims

1. A method for constructing a diagnostic model for temporary grounding wire defects, characterized in that, include: The historical ultrasonic waveform of the temporary grounding wire is obtained as a sample waveform, and the sample feature quantity of the sample waveform is extracted. Based on the sample features, and combining the number of input features and the number of output results, initialize the model structure of the BP neural network; The weights and thresholds of the BP neural network are optimized using the improved CSO algorithm with a hierarchical ring topology. Based on the sample features, the BP neural network is trained and tested. When the training accuracy of the BP neural network meets the preset accuracy requirements, training is stopped and a temporary grounding wire defect diagnosis model is output. If the training accuracy of the BP neural network does not meet the preset accuracy requirements, the improved CSO algorithm with a hierarchical ring topology is used to optimize the weights and thresholds of the BP neural network. Specifically, the optimization of the weights and thresholds of the BP neural network using the improved CSO algorithm with a hierarchical ring topology includes: Based on the BP neural network, initialize the particle swarm, rule tree, and number of layers; A preset number of particles are randomly assigned to various rings in the rule tree to form multiple populations; According to the greedy strategy, cross operations are performed on each population in the same layer and the population is updated. Based on the updated populations, the fitness of each particle in the population is calculated using a fitness function, and the quality evaluation value of the population is determined based on the fitness. Based on the association between the various populations, the superiority and inferiority evaluation values ​​of the populations and their offspring populations are compared. If the offspring population is superior to the parent population based on the comparison results, the offspring population and the parent population are swapped. If the number of iterations in the optimization process reaches a preset maximum value, the optimization stops, and the weights and thresholds of the optimized BP neural network are obtained based on the current optimal population. If the number of iterations does not reach the preset maximum value, the process returns to the previous state and performs crossover operations on each population in the same layer according to a greedy strategy to update the population.

2. The method for constructing a temporary grounding wire defect diagnosis model according to claim 1, characterized in that, The extraction of sample features from the sample waveform specifically includes: The sample waveform is denoised, and wavelet packet extraction is used to extract features from the denoised sample waveform to obtain sample features.

3. The method for constructing a temporary grounding wire defect diagnosis model according to claim 2, characterized in that, The sample features specifically include: signal mean, standard deviation, slope coefficient, kurtosis coefficient, clearing index, peak value index, impulse index, and shape index.

4. The method for constructing a temporary grounding wire defect diagnosis model according to claim 1, characterized in that, The process of training and testing the BP neural network based on the sample features, stopping training when the training accuracy of the BP neural network meets the preset accuracy requirements, and outputting a temporary grounding wire defect diagnosis model; and if the training accuracy of the BP neural network does not meet the preset accuracy requirements, returning to the improved CSO algorithm with a hierarchical ring topology to optimize the weights and thresholds of the BP neural network specifically includes: Based on the sample features, the BP neural network is trained and tested, and the training accuracy of the BP neural network is calculated through a preset optimization objective function. When the training accuracy of the BP neural network meets the preset accuracy requirement, training is stopped and a temporary grounding wire defect diagnosis model is output. If the training accuracy of the BP neural network does not meet the preset accuracy requirement, the improved CSO algorithm with a hierarchical ring topology is used to optimize the weights and thresholds of the BP neural network. The objective function is specifically: In the formula, The training accuracy of the BP neural network is represented by N, where N is the number of training data points. The defect identification result output by the BP neural network for the i-th training data item. This represents the actual defect identification result corresponding to the i-th training data item.

5. A device for constructing a temporary grounding wire defect diagnosis model, characterized in that, include: The sample feature acquisition unit is used to acquire the historical ultrasonic detection waveform of the temporary grounding wire as a sample waveform and extract the sample feature quantity of the sample waveform. The neural network initialization unit is used to initialize the model structure of the BP neural network based on the sample feature quantity, the number of input feature quantities, and the number of output results. The neural network parameter optimization unit is used to optimize the weights and thresholds of the BP neural network using an improved CSO algorithm with a hierarchical ring topology. The defect diagnosis model training unit is used to train and test the BP neural network based on the sample feature quantity. When the training accuracy of the BP neural network meets the preset accuracy requirement, the training is stopped and a temporary grounding wire defect diagnosis model is output. If the training accuracy of the BP neural network does not meet the preset accuracy requirement, the improved CSO algorithm with a hierarchical ring topology is used to optimize the weights and thresholds of the BP neural network. Specifically, the neural network parameter optimization unit is used for: Based on the BP neural network, initialize the particle swarm, rule tree, and number of layers; A preset number of particles are randomly assigned to various rings in the rule tree to form multiple populations; According to the greedy strategy, cross operations are performed on each population in the same layer and the population is updated. Based on the updated populations, the fitness of each particle in the population is calculated using a fitness function, and the quality evaluation value of the population is determined based on the fitness. Based on the association between the various populations, the superiority and inferiority evaluation values ​​of the populations and their offspring populations are compared. If the offspring population is superior to the parent population based on the comparison results, the offspring population and the parent population are swapped. If the number of iterations in the optimization process reaches a preset maximum value, the optimization stops, and the weights and thresholds of the optimized BP neural network are obtained based on the current optimal population. If the number of iterations does not reach the preset maximum value, the process returns to the previous state and performs crossover operations on each population in the same layer according to a greedy strategy to update the population.

6. The temporary grounding wire defect diagnosis model construction device according to claim 5, characterized in that, The sample feature acquisition unit is specifically used for: The historical ultrasonic detection waveform of the temporary grounding wire is obtained as a sample waveform. The sample waveform is denoised, and the feature quantity is extracted from the denoised sample waveform using wavelet packet extraction to obtain the sample feature quantity.

7. The temporary grounding wire defect diagnosis model construction device according to claim 6, characterized in that, The sample features specifically include: signal mean, standard deviation, slope coefficient, kurtosis coefficient, clearing index, peak value index, impulse index, and shape index.

8. The temporary grounding wire defect diagnosis model construction device according to claim 5, characterized in that, The defect diagnosis model training unit is specifically used for: Based on the sample features, the BP neural network is trained and tested, and the training accuracy of the BP neural network is calculated through a preset optimization objective function. When the training accuracy of the BP neural network meets the preset accuracy requirement, training is stopped and a temporary grounding wire defect diagnosis model is output. If the training accuracy of the BP neural network does not meet the preset accuracy requirement, the improved CSO algorithm with a hierarchical ring topology is used to optimize the weights and thresholds of the BP neural network. The objective function is specifically: In the formula, The training accuracy of the BP neural network is represented by N, where N is the number of training data points. The defect identification result output by the BP neural network for the i-th training data item. This represents the actual defect identification result corresponding to the i-th training data item.

Images