Distribution network grounding grid corrosion fault identification method and device based on PSO-LS-SVM algorithm

Abstract

The invention discloses a distribution network grounding grid corrosion fault identification method and device based on a PSO-LS-SVM algorithm, and is used for solving the technical problem that the existing grounding grid corrosion fault identification method generally depends on the traditional resistance measurement or simple signal processing technology, so that the corrosion fault identification precision is low. The method comprises the steps of obtaining node potential signal data of a to-be-measured grounding grid; preprocessing the node potential signal data of the grounding grid to be detected by adopting a wavelet packet transformation method to generate a target fault feature vector; and inputting the target fault feature vector into an improved PSO-LS-SVM classification diagnosis model for fault identification, and generating a target grounding grid corrosion fault identification result.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular, to a method and device for identifying corrosion faults of a distribution network grounding grid based on a PSO-LS-SVM algorithm. Background Art

[0002] With the continuous development of power systems, the distribution network, as an important part of the power network, and its grounding grid, as an important component of distribution network equipment, the reliability of the grounding grid operation is of great significance for the safety of surrounding equipment and operating personnel. Most of the grounding grid systems of power stations and distribution networks built in the initial stage use steel materials with slightly more economical cost as grounding materials, rather than copper materials with better corrosion resistance.

[0003] Therefore, after years of operation of the grounding grid, the electrical performance of the corroded steel conductors has significantly decreased, resulting in problems such as poor connection and reduced current dissipation performance, and it is impossible to ensure the safety protection of in-station equipment under abnormal conditions such as lightning and overvoltage. For this problem, an effective method is urgently needed to achieve timely detection and treatment of grounding grid faults and avoid accidents.

[0004] Existing methods for identifying corrosion faults of grounding grids usually rely on traditional resistance measurements or simple signal processing techniques. However, with the continuous expansion of the scale of the distribution network, the potential data of the grounding grid shows a trend of large scale and complexity, and these methods cannot be processed in real time in a complex electromagnetic environment, resulting in low efficiency in identifying corrosion faults. Summary of the Invention

[0005] The present invention provides a method and device for identifying corrosion faults of a distribution network grounding grid based on a PSO-LS-SVM algorithm, which are used to solve the technical problem that existing methods for identifying corrosion faults of grounding grids usually rely on traditional resistance measurements or simple signal processing techniques, resulting in low accuracy in identifying corrosion faults.

[0006] A method for identifying corrosion faults of a distribution network grounding grid based on a PSO-LS-SVM algorithm provided by the first aspect of the present invention includes:

[0007] Obtaining the potential signal data of the grounding grid nodes to be measured;

[0008] Preprocessing the potential signal data of the grounding grid nodes to be measured by using the wavelet packet transform method to generate a target fault feature vector;

[0009] Inputting the target fault feature vector into an improved PSO-LS-SVM classification and diagnosis model for fault identification to generate a target grounding grid corrosion fault identification result.

[0010] Optionally, preprocessing the potential signal data of the grounding grid node to be measured by using the wavelet packet transform method to generate a target fault feature vector, including:

[0011] Decompose the potential signal data of the grounding grid node to be measured to determine a plurality of high and low frequency band signals;

[0012] Calculate the band energy corresponding to each of the high and low frequency band signals according to the wavelet packet coefficients corresponding to each of the high and low frequency band signals;

[0013] Generate a target fault feature vector according to the band energy corresponding to each of the high and low frequency band signals.

[0014] Optionally, the model training process of the improved PSO-LS-SVM classification and diagnosis model is specifically as follows:

[0015] Obtain the training potential signal data of the grounding grid node;

[0016] Preprocess the training potential signal data of the grounding grid node by using the wavelet packet transform method to generate a training fault feature vector;

[0017] Based on the particle swarm algorithm and the preset parameter optimization interval, use the training fault feature vector to train the model of the initial PSO-LS-SVM classification and diagnosis model to determine the trained improved PSO-LS-SVM classification and diagnosis model.

[0018] Optionally, based on the particle swarm algorithm and the preset parameter optimization interval, using the training fault feature vector to train the model of the initial PSO-LS-SVM classification and diagnosis model to determine the trained improved PSO-LS-SVM classification and diagnosis model, including:

[0019] Generate a particle swarm within the preset parameter optimization interval and initialize the model parameter combinations corresponding to each particle in the particle swarm;

[0020] Input the training fault feature vector into the PSO-LS-SVM classification and diagnosis model associated with the model parameter combination corresponding to each particle for fault identification, and output the training grounding grid corrosion fault identification results corresponding to each particle;

[0021] Calculate the fitness value corresponding to each particle according to the training grounding grid corrosion fault identification results corresponding to each particle;

[0022] Determine the global optimal model parameter combination based on the model parameter combinations and fitness values corresponding to each particle;

[0023] If the fitness value corresponding to the global optimal model parameter combination is less than the preset fitness threshold or the number of iterations reaches the preset iteration threshold, determine the global optimal model parameter combination as the target model parameter combination, and obtain the trained improved PSO-LS-SVM classification and diagnosis model;

[0024] Otherwise, update the model parameter combinations corresponding to each particle based on the global optimal model parameter combination, and perform the steps of inputting the training fault feature vectors into the PSO-LS-SVM classification and diagnosis models associated with the updated model parameter combinations corresponding to each particle for fault identification, and outputting the training grounding grid corrosion fault identification results corresponding to each particle.

[0025] Optionally, the improved PSO-LS-SVM classification and diagnosis model is specifically:

[0026] ;

[0027] Wherein, is the target grounding grid corrosion fault identification result; is the Lagrange multiplier; is the output of x k ; is the kernel function, , C is the kernel function parameter; b is the bias term; x is the target fault feature vector corresponding to the measured grounding grid node potential signal data; is the feature vector of the k-th reference grounding grid node potential signal data for kernel function calculation.

[0028] Optionally, the calculation process of the band energy is specifically:

[0029] ;

[0030] Wherein, is the band energy of the high and low frequency band signal j of the m-th layer; is the key quantity for calculating ; is the reconstruction coefficient of the wavelet packet decomposition of the high and low frequency band signal j of the m-th layer; N is the number of wavelet packet coefficients included in the current high and low frequency band.

[0031] A device for identifying grounding grid corrosion faults in a distribution network based on the PSO-LS-SVM algorithm provided in the second aspect of the present invention includes:

[0032] An acquisition module for acquiring measured grounding grid node potential signal data;

[0033] A preprocessing module for preprocessing the measured grounding grid node potential signal data by using the wavelet packet transform method to generate a target fault feature vector;

[0034] A fault identification module, configured to input the target fault feature vector into an improved PSO-LS-SVM classification and diagnosis model for fault identification, and generate a target grounding grid corrosion fault identification result.

[0035] A computer device provided in the third aspect of the present invention includes a memory and a processor. A computer program is stored in the memory. When the computer program is executed by the processor, the processor is caused to execute the steps of the method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm as described in any one of the above.

[0036] A computer-readable storage medium provided in the fourth aspect of the present invention has a computer program stored thereon. When the computer program is executed, the steps of the method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm as described in any one of the above are implemented.

[0037] A computer program product provided in the fifth aspect of the present invention. The computer program product includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer is caused to execute the steps of the method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm as described in any one of the above.

[0038] It can be seen from the above technical solutions that the present invention has the following advantages:

[0039] The above technical solution of the present invention provides a method for identifying corrosion faults in a distribution network grounding grid based on the PSO-LS-SVM algorithm. First, obtain the node potential signal data of the grounding grid to be measured; then, use the wavelet packet transform method to preprocess the node potential signal data of the grounding grid to be measured to generate a target fault feature vector; finally, input the target fault feature vector into an improved PSO-LS-SVM classification and diagnosis model for fault identification to generate a target grounding grid corrosion fault identification result. Based on the above solution, the process of using the wavelet packet transform method to preprocess the obtained node potential signal data of the grounding grid to be measured to generate a target fault feature vector, and then using the improved PSO-LS-SVM classification and diagnosis model to perform fault identification according to the target fault feature vector to generate a target grounding grid corrosion fault identification result. The present invention uses an improved PSO-LS-SVM classification and diagnosis model that combines the particle swarm optimization (PSO) and least squares support vector machine (LS-SVM) algorithms, and can process in real time when facing the potential data of a large-scale grounding grid, thereby improving the identification efficiency of corrosion faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0041] Figure 1 It is a flowchart of the steps of a method for identifying corrosion faults in a distribution network grounding grid based on the PSO-LS-SVM algorithm provided by Embodiment 1 of the present invention;

[0042] Figure 2 It is a schematic diagram of the three-layer tree structure of the wavelet packet transform provided by Embodiment 1 of the present invention;

[0043] Figure 3 It is a schematic diagram of the principle of the support vector machine provided by Embodiment 1 of the present invention;

[0044] Figure 4 It is a flowchart of the steps of the model training process of the improved PSO-LS-SVM classification and diagnosis model provided by Embodiment 2 of the present invention;

[0045] Figure 5 It is a schematic flowchart of the parameter optimization process of the improved PSO-LS-SVM classification and diagnosis model provided by Embodiment 2 of the present invention;

[0046] Figure 6 Schematic diagram of the simulated grounding grid provided in the second embodiment of the present invention and schematic diagram of the experimental site

[0047] Figure 7 Schematic diagram of the connection of the experimental device provided in the second embodiment of the present invention

[0048] Figure 8 Schematic diagram of the excavation detection of the fault location provided in the second embodiment of the present invention

[0049] Figure 9 Structural block diagram of a distribution network grounding grid corrosion fault identification device based on the PSO-LS-SVM algorithm provided in the third embodiment of the present invention Detailed implementation manners

[0050] The embodiment of the present invention provides a distribution network grounding grid corrosion fault identification method and device based on the PSO-LS-SVM algorithm, which are used to solve the technical problem that the existing grounding grid corrosion fault identification methods usually rely on traditional resistance measurement or simple signal processing techniques, resulting in low identification accuracy of corrosion faults

[0051] In order to make the invention purpose, features, and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the embodiments described below are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention

[0052] Please refer to Figure 1 , Figure 1 Flowchart of the steps of a distribution network grounding grid corrosion fault identification method based on the PSO-LS-SVM algorithm provided in the first embodiment of the present invention

[0053] A distribution network grounding grid corrosion fault identification method based on the PSO-LS-SVM algorithm provided by the present invention includes

[0054] Step 101: Obtain the node potential signal data of the grounding grid to be measured

[0055] It should be noted that in order to accurately obtain the potential signal data of the distribution network grounding grid, the present invention uses a high-precision potential acquisition device FLUKE8845A to measure the node potential with a downlead. For the node potential without a downlead, the interpolation method is used for calculation to ensure the integrity and accuracy of the data

[0056] In this embodiment, the node potential signal data of the grounding grid to be measured is obtained

[0057] Step 102: Preprocess the node potential signal data of the grounding grid to be measured by using the wavelet packet transform method to generate a target fault feature vector.

[0058] It should be noted that since the advent of wavelet transform, due to its excellent performance in signal processing, it has been widely used. Wavelet transform is of great significance in the field of signal processing. It realizes multi-resolution analysis through the localization of time domain and frequency domain. However, due to the limitations of the algorithm, wavelet transform can only decompose the low-frequency part of the signal and cannot perform more refined processing and approximation on the high-frequency part.

[0059] Specifically, Step 102 may include the following sub-steps S21 - S23:

[0060] Step S21: Decompose the node potential signal data of the grounding grid to be measured to determine multiple high- and low-frequency band signals;

[0061] Step S22: Calculate the band energy corresponding to each high- and low-frequency band signal according to the wavelet packet coefficients corresponding to each high- and low-frequency band signal;

[0062] Step S23: Generate a target fault feature vector according to the band energy corresponding to each high- and low-frequency band signal.

[0063] The high- and low-frequency band signals include high-frequency band signals and low-frequency band signals.

[0064] It should be noted that, please refer to Figure 2 , the wavelet packet transform is an improvement of the standard wavelet transform. It can not only decompose low-frequency signals but also perform more refined decomposition on high-frequency signals. Through the decomposition and reconstruction of signals, the wavelet packet transform can map signals with different characteristics to different frequency bands, thereby realizing the refined analysis of signals. Taking the three-layer wavelet packet transform as an example, its basic process is as Figure 2 shown. S represents the original signal, H represents the low-frequency part of the signal, G represents the high-frequency part of the signal, and the numbers represent the number of layers of signal decomposition.

[0065] Furthermore, the decomposition of the wavelet packet transform can be expressed by the following formula:

[0066] ;

[0067] Among them, is the wavelet packet decomposition coefficient of the high- and low-frequency band signal j at the 2nth layer; is the wavelet packet decomposition coefficient (wavelet packet coefficient) of the high- and low-frequency band signal j at the (2n + 1)th layer; is the wavelet packet reconstruction coefficient of the high- and low-frequency band signal j + 1 at the nth layer; and They are the low-pass filter coefficients and high-pass filter coefficients of the wavelet packet decomposition respectively; k is the summation index; l is the displacement parameter.

[0068] Furthermore, the essence of the wavelet packet transform decomposition is to decompose the signal into different frequency bands by using conjugate orthogonal filters, and use different frequency bands to characterize different characteristics of the signal. The reconstruction process of the wavelet packet transform is through and solving , that is, reconstructing the wavelet packet coefficients corresponding to the high and low frequency band signals to obtain the corresponding wavelet packet decomposition reconstruction coefficients, and its recurrence formula is as follows:

[0069] ;

[0070] where, is the low-pass filter coefficient of the wavelet packet reconstruction; is the high-pass filter coefficient of the wavelet packet reconstruction.

[0071] Furthermore, through the reconstruction coefficients of the wavelet packet decomposition, the fluctuation coefficients of each corresponding frequency band can be obtained, and the energy of the signals in each frequency band can be calculated. In the present invention, the frequency band energy is used to characterize the fault characteristics. Therefore, in the analysis, the frequency bands closely related to the fault are selected and their energy values are calculated. The frequency band energy calculation formula is as follows:

[0072] ;

[0073] where, is the frequency band energy of the high and low frequency band signal j of the m-th layer; is the key quantity used to calculate ; is the reconstruction coefficient of the wavelet packet decomposition of the high and low frequency band signal j of the m-th layer; N is the number of wavelet packet coefficients included in the current high and low frequency bands.

[0074] Furthermore, use all the frequency band energy values to construct the target fault feature vector:

[0075] ;

[0076] where, E is the target fault feature vector; is the frequency band energy of the high and low frequency band signal 2j - 1 of the m-th layer.

[0077] In this embodiment, the original potential signal is decomposed and reconstructed by the wavelet packet transform, the energy signals in different frequency bands are extracted, the fault feature vector is constructed, and it is used as the input signal for fault identification.

[0078] Step 103: Input the target fault feature vector into the improved PSO-LS-SVM classification and diagnosis model for fault identification, and generate the target grounding grid corrosion fault identification result.

[0079] It should be noted that after extracting the corrosion fault feature data, the present invention constructs an improved PSO-LS-SVM classification and diagnosis model for classifying the feature data to achieve the identification of grounding grid corrosion faults.

[0080] Furthermore, please refer to Figure 3 , the support vector machine (SVM) is one of the commonly used methods in fault classification, with good robustness and generalization ability. Its essence is to find an optimal hyperplane to divide the input data into two categories and maintain the maximum distance between the categories. The basic principle of SVM is as Figure 3 shown.

[0081] The least squares support vector machine (LS-SVM) is an improvement of SVM. It replaces the inequality constraint in the standard SVM algorithm with an equality constraint, transforming the quadratic programming problem into a system of linear equations, thereby reducing the difficulty of solution. Compared with the standard SVM algorithm, LS-SVM has better advantages in dealing with small samples and nonlinear problems.

[0082] Assume that the sample signal to be processed is , where is the input, is the output of

[0083] ;

[0084] where, is the transpose of the weight vector. In the support vector machine (SVM), is the normal vector of the classification hyperplane; b is the bias term (also called the intercept) in the classification hyperplane equation. It is used to adjust the position of the classification hyperplane; is the feature vector obtained by transforming the input sample x k through a certain mapping function . In the support vector machine, is used to map the input data from the original space to a high-dimensional feature space in order to find a linearly separable hyperplane in the high-dimensional space.

[0085] Starting from the loss function, LS-SVM uses the second norm optimization formula. At the same time, in order to solve the Figure 2 in and the possible outliers between, the slack variable is added, and The following objective function and constraints can be obtained:

[0086] ;

[0087] Among them, is the weight vector or normal vector. In the Support Vector Machine (SVM), is the normal vector of the classification hyperplane, which is used to define the direction of the classification hyperplane. Its role in the formula: is the regularization term, which is used to control the complexity of the model. Minimizing can maximize the margin of the classification hyperplane, thereby improving the generalization ability of the model; is the penalty parameter, which is a constant greater than zero and is used to constrain the complexity and generalization ability of the model; N is the number of samples or the total number of data points.

[0088] The LS-SVM formula can be simplified through the Lagrange function:

[0089] ;

[0090] Among them, is the Lagrange function; is the objective function, , which is the optimization objective of LS-SVM, aiming to find a classification hyperplane to make the complexity of the model (controlled by ) and the classification error (controlled by ) as small as possible; is the Lagrange multiplier, and the optimization objective is transformed into the solution of a single parameter .

[0091] By taking the partial derivatives of , b, and and setting them to zero:

[0092] ;

[0093] Finally, through simplification, the LS-SVM classification decision equation (improved PSO-LS-SVM classification diagnosis model) can be obtained:

[0094] ;

[0095] Among them, is the recognition result of the target grounding grid corrosion fault; is the Lagrange multiplier; is the output of the input sample x k ; is the kernel function, , C is the kernel function parameter; b is the bias term; x is the target fault feature vector corresponding to the potential signal data of the ground grid node to be measured; is the feature vector of the k-th reference ground grid node potential signal data for kernel function calculation, that is, the training feature vector corresponding to the historical sample (training ground grid node potential signal data) used during model training. The similarity between the sample x to be measured (ground grid node potential signal data to be measured) and the training sample x i (training ground grid node potential signal data) is calculated through the kernel function, thereby affecting the classification decision.

[0096] Furthermore, As the kernel function, it can map the input in the low-dimensional space to the high-dimensional space and find the optimal classification hyperplane in the high-dimensional space. Commonly used kernel functions include linear kernel function, polynomial kernel function, radial basis function (RBF), and sigmoid kernel function, etc. Considering the computational complexity and non-linear approximation ability, this patent selects the radial basis function.

[0097] In this embodiment, the target fault feature vector is input into the improved PSO-LS-SVM classification and diagnosis model for fault identification, and the target ground grid corrosion fault identification result is generated.

[0098] As a comparison of technical effects, it can be referred to in combination with the prior art. With the continuous development of the power system, the distribution network, as an important part of the power network, the corrosion problem of its ground grid has gradually become a key factor affecting power safety and equipment stability. Existing methods for identifying ground grid corrosion faults usually rely on traditional resistance measurement or simple signal processing techniques. These methods have poor adaptability in complex electromagnetic environments, resulting in low identification accuracy of corrosion faults and prone to misdiagnosis or missed diagnosis. Moreover, with the continuous expansion of the distribution network scale, the potential data of the ground grid shows a trend of large scale and complexity. Existing methods cannot meet the requirements of real-time processing and high-precision diagnosis in the face of these massive data in terms of computational efficiency.

[0099] Therefore, how to improve the accuracy of ground grid corrosion fault identification, enhance the computational efficiency of large-scale data processing, and overcome the influence of stray electromagnetic interference on fault identification accuracy has become an urgent technical problem to be solved. In view of the above problems, the present invention proposes a method for identifying ground grid corrosion faults in a distribution network based on the PSO-LS-SVM algorithm, aiming to combine the particle swarm optimization (PSO) and least squares support vector machine (LS-SVM) algorithms to efficiently and accurately identify ground grid corrosion faults and solve problems such as low identification accuracy and poor computational efficiency in the prior art.

[0100] In an embodiment of the present invention, the present invention provides a method for identifying corrosion faults in a distribution network grounding grid based on the PSO-LS-SVM algorithm. First, obtain the node potential signal data of the grounding grid to be measured; then, use the wavelet packet transform method to preprocess the node potential signal data of the grounding grid to be measured to generate a target fault feature vector; finally, input the target fault feature vector into an improved PSO-LS-SVM classification and diagnosis model for fault identification to generate a target grounding grid corrosion fault identification result; based on the above solution, the process of using the wavelet packet transform method to preprocess the obtained node potential signal data of the grounding grid to be measured to generate a target fault feature vector, and then using the improved PSO-LS-SVM classification and diagnosis model to perform fault identification according to the target fault feature vector to generate a target grounding grid corrosion fault identification result. The present invention uses an improved PSO-LS-SVM classification and diagnosis model that combines the particle swarm optimization (PSO) and least squares support vector machine (LS-SVM) algorithms. When facing the potential data of a large-scale grounding grid, it can be processed in real time, thereby improving the identification efficiency of corrosion faults.

[0101] For better illustration, refer to Figure 4 , which shows the step flowchart of the model training process of the improved PSO-LS-SVM classification and diagnosis model provided in the second embodiment of the present invention. This process may include the following steps:

[0102] Step 401, obtain the training grounding grid node potential signal data.

[0103] The training grounding grid node potential signal data is the potential signal data for model training.

[0104] In this embodiment, obtain the training grounding grid node potential signal data.

[0105] Step 402, use the wavelet packet transform method to preprocess the training grounding grid node potential signal data to generate a training fault feature vector.

[0106] In this embodiment, use the wavelet packet transform method to preprocess the training grounding grid node potential signal data to generate a training fault feature vector.

[0107] Step 403, based on the particle swarm algorithm and the preset parameter optimization interval, use the training fault feature vector to train the model of the initial PSO-LS-SVM classification and diagnosis model to determine the trained improved PSO-LS-SVM classification and diagnosis model.

[0108] Further, step 403 may include the following sub-steps S431 - S4313:

[0109] Step S431, within the preset parameter optimization range, generate a particle swarm and initialize the model parameter combinations corresponding to each particle in the particle swarm;

[0110] Step S432, input the training fault feature vectors into the PSO - LS - SVM classification and diagnosis models associated with the model parameter combinations corresponding to each particle for fault identification, and output the training grounding grid corrosion fault identification results corresponding to each particle;

[0111] Step S433, calculate the fitness value corresponding to each particle according to the training grounding grid corrosion fault identification results corresponding to each particle;

[0112] Step S434, determine the global optimal model parameter combination based on the model parameter combinations and fitness values corresponding to each particle;

[0113] Step S435, if the fitness value corresponding to the global optimal model parameter combination is less than the preset fitness threshold or the number of iterations reaches the preset iteration number threshold, determine the global optimal model parameter combination as the target model parameter combination, and obtain the trained improved PSO - LS - SVM classification and diagnosis model;

[0114] Step S436, otherwise, update the model parameter combinations corresponding to each particle based on the global optimal model parameter combination, and execute the step of inputting the training fault feature vectors into the PSO - LS - SVM classification and diagnosis models associated with the updated model parameter combinations corresponding to each particle for fault identification, and output the training grounding grid corrosion fault identification results corresponding to each particle.

[0115] It should be noted that based on the above derivation, it can be seen that the actual classification effect of the LS - SVM classification decision equation is mainly determined by the penalty parameter and the kernel function parameter C. The traditional parameter selection method mostly uses the cross - validation method, which relies on experience and the effect is average. To improve the accuracy of fault diagnosis, there has been an SVM model optimized based on the artificial bee colony algorithm (ABC, Artificial Bee Colony). However, when optimizing the parameters, this algorithm is prone to falling into the local optimal solution. There is also an LS - SVM model optimized based on the genetic algorithm (GA, Genetic Algorithm) for predicting the corrosion rate of the grounding grid, but its randomness and blindness are relatively strong and the efficiency is low.

[0116] Furthermore, the present invention uses the Particle Swarm Optimization (PSO) algorithm to optimize the parameters of the LS-SVM classification equation. The Particle Swarm Optimization algorithm has the advantages of fast search speed, simple structure, and easy engineering implementation. The PSO algorithm has the characteristics of robustness and parallelism, and has strong global search ability, and this search ability does not depend on a specific solution model. Therefore, the PSO algorithm is used to automatically optimize the parameters.

[0117] Furthermore, please refer to Figure 5 , the parameter optimization process of the improved PSO-LS-SVM classification diagnosis model can be simply divided into six steps:

[0118] (1) The particles in the particle swarm represent the parameter combination of LS-SVM, and the parameter optimization interval is initialized (this parameter optimization interval can be set by yourself according to needs);

[0119] (2) Set the initial particle swarm size to N;

[0120] (3) Randomly initialize the velocity and position of each particle;

[0121] (4) Calculate the fitness function of the particle;

[0122] (5) According to the fitness function, calculate and update the individual particle's optimal position and the global optimal position;

[0123] (6) If the optimal solution is found within the number of iterations, output the LS-SVM parameter optimization result; otherwise, initialize the velocity and position of some particles with probability P. At the same time, perform a linear search in the negative gradient direction with probability according to the gradient information to determine the moving step size of the particles. Then return to step (4).

[0124] Specifically, within the preset parameter optimization interval, generate a particle swarm, and initialize the model parameter combination (position) corresponding to each particle in the particle swarm; input the training fault feature vectors into the PSO-LS-SVM classification diagnosis model associated with the initialized model parameter combination corresponding to each particle for fault identification, and output the training grounding grid corrosion fault identification results corresponding to each particle; calculate the fitness value corresponding to each particle according to the training grounding grid corrosion fault identification results corresponding to each particle; based on the initialized model parameter combination and fitness value corresponding to each particle, determine the individual historical optimal model parameter combination and individual optimal fitness value corresponding to each particle, where, if it is the first iteration currently, record the initialized model parameter combination and its fitness value of each particle as the individual historical optimal model parameter combination and individual optimal fitness value of each particle respectively; if it is not the first iteration currently, it is necessary to compare the fitness value corresponding to each current particle with the individual optimal fitness value to determine the individual historical optimal model parameter combination and individual optimal fitness value of each particle.

[0125] Furthermore, traverse the individual optimal fitness values of all particles, set the model parameter combination corresponding to the minimum individual optimal fitness value as the global optimal model parameter combination, and record its fitness value as the global optimal fitness value. Based on each individual's historical optimal model parameter combination and the global optimal model parameter combination, update the initialized model parameter combinations corresponding to multiple particles, determine the updated model parameter combinations corresponding to each particle (the existing update process can be referred to), and count the update times in real time. Specifically, initialize the velocities and positions (model parameter combinations) of some particles with probability P (the specific value can be set according to needs), so as to obtain the updated model parameter combinations corresponding to the particles. At the same time, perform a linear search in the negative gradient direction with probability according to the gradient information to determine the moving step size of the particles.

[0126] Furthermore, input the training fault feature vectors into the PSO-LS-SVM models corresponding to the updated model parameter combinations of each particle, and output the updated training grounding grid corrosion fault recognition results. Calculate the updated fitness values of each particle according to the new results. For each particle, if its updated fitness value < the historical individual optimal fitness value of this particle: update the updated model parameter combination of this particle to the new individual historical optimal model parameter combination; update the updated fitness value to the new individual optimal fitness value.

[0127] Furthermore, traverse the new individual optimal fitness values of all particles, set the model parameter combination corresponding to the minimum individual optimal fitness value as the new global optimal model parameter combination, and record its fitness value as the new global optimal fitness value.

[0128] Furthermore, if any of the following conditions is met, stop the iteration: the new global optimal fitness value < the preset fitness threshold, or the update times reach the preset iteration times threshold. Otherwise, return to step S432 to continue the iteration until the fitness value corresponding to the global optimal model parameter combination is less than the preset fitness threshold or the update times reach the preset iteration times threshold. Take the global optimal model parameter combination determined by the fitness value corresponding to the global optimal model parameter combination being less than the preset fitness threshold or the update times reaching the preset iteration times threshold as the target model parameter combination, and use the target model parameter combination to update the model parameters of the initial PSO-LS-SVM classification and diagnosis model to determine the trained improved PSO-LS-SVM classification and diagnosis model.

[0129] Furthermore, to verify the effectiveness of the model, a scaled-down simulated grounding grid was constructed in the laboratory. The grid side length of the simulated grounding grid is 1 meter, and the branch resistance is about several tens of milliohms, which is consistent with the structure of the actual grounding grid. The simulated grounding grid is made by welding the same flat steel material as the actual grounding grid. By clamping two flat steels, the corrosion of the grounding grid is simulated, increasing the branch resistance to about 6 times that of the normal branch, thereby generating corrosion fault samples of different degrees.

[0130] In the experiment, diagonal nodes 1 and 32 of the simulated grounding grid were selected as the current injection points, and a constant current of 6A was injected to generate the current field required for the experiment. To accurately obtain the potential data, a high-precision potential acquisition device FLUKE8845A was used to measure the node potential with a down-lead, while for the node potential without a down-lead, the interpolation method was used for calculation to ensure the integrity and accuracy of the data.

[0131] The collected potential data was preprocessed and decomposed in terms of waveform. First, the wavelet packet analysis technique was used to decompose the signal at multiple scales, and the energy in a specific frequency band was extracted as the feature vector of the training data. Subsequently, the feature vector was input into the trained improved PSO-LS-SVM fault recognition model to achieve the classification and recognition of grounding grid corrosion faults.

[0132] The schematic diagram of the construction of the simulated grounding grid and the connection of the experimental device is as Figure 6 shown. During the experiment, by adjusting the frequency and direction of the injected current, potential data in different states was collected to provide diverse training samples for the model and ensure its generalization ability in the actual engineering environment.

[0133] Furthermore, to verify the actual effect of the present invention, a 330kV high-voltage distribution network that has been in operation for more than 15 years was selected for on-site verification. According to the actual situation of the distribution network, a group of underground leads located in the diagonal direction on the ground surface were selected as the current injection and extraction positions, and multi-strand copper core cables with a diameter of were used to inject excitation currents of different frequencies into the grounding grid under test. The connection method of the experimental device is as Figure 7 shown.

[0134] Further, during the experiment, a mobile electromagnetic induction signal acquisition device was used to collect the excitation electromagnetic induction intensity signals of the grounding grid. During the signal acquisition process, by observing the changes in the electromagnetic induction intensity values, the topological distribution of the grounding grid conductors could be judged. The moving direction of the acquisition device was adjusted to move along the strong signal distribution direction, that is, the acquisition device was always kept directly above the grounding grid conductors. After the acquisition was completed, the wavelet packet transform was used to process the electromagnetic induction data to form the eigenvalue vector of the corrosion fault. The eigenvalues were input into the fault recognition model (improved PSO-LS-SVM classification diagnosis model) for analysis and processing, and finally the analysis results were obtained.

[0135] Further, through the analysis results of the data from this test, it was found that there was a corrosion alarm in the tested area. During the power outage of the distribution network, manual excavation detection was carried out at this location, and it was found that the surface of the grounding grid conductors at this location was severely rusted, and obvious thinning phenomena could be observed at the local conductor positions, such as Figure 8 shown. The experimental results prove that the present invention can effectively identify the corrosion faults of the grounding grid.

[0136] In the embodiment of the present invention, by combining the particle swarm optimization (PSO) and least squares support vector machine (LS-SVM) algorithms, an improved PSO-LS-SVM classification diagnosis model capable of efficiently and accurately identifying the corrosion faults of the grounding grid is proposed to solve the problems of low recognition accuracy and poor calculation efficiency in the prior art.

[0137] Please refer to Figure 9 , Figure 9 which is the structural block diagram of a distribution network grounding grid corrosion fault recognition device provided in Embodiment 3 of the present invention.

[0138] A distribution network grounding grid corrosion fault recognition device provided by the present invention includes:

[0139] An acquisition module 901, configured to acquire the node potential signal data of the grounding grid to be measured;

[0140] A preprocessing module 902, configured to preprocess the node potential signal data of the grounding grid to be measured by using the wavelet packet transform method to generate a target fault feature vector;

[0141] A fault recognition module 903, configured to input the target fault feature vector into the improved PSO-LS-SVM classification diagnosis model for fault recognition to generate a target grounding grid corrosion fault recognition result.

[0142] Further, the preprocessing module 902 is specifically configured to:

[0143] Decompose the node potential signal data of the grounding grid to be measured to determine multiple high- and low-frequency band signals;

[0144] Calculate the band energy corresponding to each high- and low-frequency band signal according to the wavelet packet coefficients corresponding to each high- and low-frequency band signal;

[0145] Generate a target fault feature vector according to the band energy corresponding to each high- and low-frequency band signal.

[0146] In an optional device embodiment, it further includes:

[0147] The first module is used to obtain the training grounding grid node potential signal data;

[0148] The second module is used to preprocess the training grounding grid node potential signal data by using the wavelet packet transform method to generate a training fault feature vector;

[0149] The third module is used to train the model of the initial PSO-LS-SVM classification and diagnosis model by using the training fault feature vector based on the particle swarm algorithm and the preset parameter optimization interval to determine the trained improved PSO-LS-SVM classification and diagnosis model.

[0150] Further, the third module is specifically used for:

[0151] Generate a particle swarm within the preset parameter optimization interval and initialize the model parameter combinations corresponding to each particle in the particle swarm;

[0152] Input the training fault feature vector into the PSO-LS-SVM classification and diagnosis model associated with the model parameter combination corresponding to each particle for fault identification, and output the training grounding grid corrosion fault identification results corresponding to each particle;

[0153] Calculate the fitness value corresponding to each particle according to the training grounding grid corrosion fault identification results corresponding to each particle;

[0154] Determine the global optimal model parameter combination based on the model parameter combination and fitness value corresponding to each particle;

[0155] If the fitness value corresponding to the global optimal model parameter combination is less than the preset fitness threshold or the number of iterations reaches the preset iteration number threshold, determine the global optimal model parameter combination as the target model parameter combination to obtain the trained improved PSO-LS-SVM classification and diagnosis model;

[0156] Otherwise, update the model parameter combinations corresponding to each particle based on the globally optimal model parameter combination, and perform the steps of inputting the training fault feature vectors into the PSO-LS-SVM classification and diagnosis models associated with the updated model parameter combinations corresponding to each particle for fault identification, and outputting the training grounding grid corrosion fault identification results corresponding to each particle.

[0157] Furthermore, the improved PSO-LS-SVM classification and diagnosis model is specifically as follows:

[0158] ;

[0159] where is the target grounding grid corrosion fault identification result; is the Lagrange multiplier; is the output of x k ; is the kernel function, , C is the kernel function parameter; b is the bias term; x is the target fault feature vector corresponding to the measured grounding grid node potential signal data; is the feature vector of the k-th reference grounding grid node potential signal data for kernel function calculation.

[0160] Furthermore, the calculation process of the band energy is specifically as follows:

[0161] ;

[0162] where is the band energy of the high and low frequency band signal j of the m-th layer; is the key quantity for calculating ; is the reconstruction coefficient of the wavelet packet decomposition of the high and low frequency band signal j of the m-th layer; N is the number of wavelet packet coefficients included in the current high and low frequency band.

[0163] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described devices and modules can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0164] The embodiment of the present invention also provides a computer device, including a memory and a processor, and a computer program is stored in the memory; when the computer program is executed by the processor, the processor is caused to execute the steps of the method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm as described in any of the foregoing embodiments.

[0165] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program / instructions are stored. When the computer program / instructions are executed by a processor, the steps of the method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm in any of the above embodiments are implemented.

[0166] An embodiment of the present invention also provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the steps of the method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm in any of the above embodiments are implemented.

[0167] In several embodiments provided by the present application, it should be understood that the disclosed device and method can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of the device or unit can be in an electrical, mechanical or other form.

[0168] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0169] As mentioned above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying corrosion faults in a distribution network grounding grid based on the PSO-LS-SVM algorithm, characterized in that, Including: Obtain the node potential signal data of the grounding grid to be measured; Preprocess the node potential signal data of the grounding grid to be measured by using the wavelet packet transform method to generate a target fault feature vector; Input the target fault feature vector into an improved PSO-LS-SVM classification and diagnosis model for fault identification to generate a target grounding grid corrosion fault identification result.

2. The method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm according to claim 1, characterized in that The step of preprocessing the node potential signal data of the grounding grid to be measured by using the wavelet packet transform method to generate a target fault feature vector includes: Decompose the node potential signal data of the grounding grid to be measured to determine a plurality of high and low frequency band signals; Calculate the band energy corresponding to each of the high and low frequency band signals according to the wavelet packet coefficients corresponding to each of the high and low frequency band signals; Generate a target fault feature vector according to the band energy corresponding to each of the high and low frequency band signals.

3. The method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm according to claim 1, characterized in that, The model training process of the improved PSO-LS-SVM classification and diagnosis model is specifically as follows: Obtain the training grounding grid node potential signal data; Preprocess the training grounding grid node potential signal data by using the wavelet packet transform method to generate a training fault feature vector; Based on the particle swarm algorithm and a preset parameter optimization interval, use the training fault feature vector to train the model of the initial PSO-LS-SVM classification and diagnosis model to determine the trained improved PSO-LS-SVM classification and diagnosis model.

4. The method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm according to claim 3, wherein, The step of based on the particle swarm algorithm and a preset parameter optimization interval, using the training fault feature vector to train the model of the initial PSO-LS-SVM classification and diagnosis model to determine the trained improved PSO-LS-SVM classification and diagnosis model includes: Generate a particle swarm within the preset parameter optimization interval and initialize the model parameter combinations corresponding to each particle in the particle swarm; Input the training fault feature vector into the PSO-LS-SVM classification and diagnosis models associated with the model parameter combinations corresponding to each particle for fault identification, and output the training grounding grid corrosion fault identification results corresponding to each particle; Calculate the fitness value corresponding to each particle according to the training grounding grid corrosion fault identification results corresponding to each particle; Determine the global optimal model parameter combination based on the model parameter combinations and fitness values corresponding to each particle; If the fitness value corresponding to the global optimal model parameter combination is less than the preset fitness threshold or the number of iterations reaches the preset iteration number threshold, determine the global optimal model parameter combination as the target model parameter combination to obtain the trained improved PSO-LS-SVM classification and diagnosis model; Otherwise, update the model parameter combinations corresponding to each particle based on the global optimal model parameter combination, and execute the step of inputting the training fault feature vector into the PSO-LS-SVM classification and diagnosis models associated with the updated model parameter combinations corresponding to each particle for fault identification, and output the training grounding grid corrosion fault identification results corresponding to each particle.

5. The method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm according to claim 1, wherein The improved PSO-LS-SVM classification and diagnosis model is specifically: ； Among them, is the recognition result of the corrosion fault of the target grounding grid; is the Lagrange multiplier; is the output of x k ; is the kernel function, , where C is the kernel function parameter; b is the bias term; x is the target fault feature vector corresponding to the potential signal data of the grounding grid nodes to be measured; is the feature vector of the k-th reference grounding grid node potential signal data for kernel function calculation.

6. The method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm according to claim 2, characterized in that, The calculation process of the band energy is specifically: ； Among them, is the band energy of the high and low frequency band signal j of the m-th layer; is used to calculate the key quantity; is the reconstruction coefficient of the wavelet packet decomposition of the high and low frequency band signal j of the m-th layer; N is the number of wavelet packet coefficients included in the current high and low frequency band.

7. A device for identifying corrosion faults of a distribution network grounding grid based on the PSO-LS-SVM algorithm, characterized in that, Including: An acquisition module for acquiring the potential signal data of the grounding grid nodes to be measured; A preprocessing module for preprocessing the potential signal data of the grounding grid nodes to be measured by using the wavelet packet transform method to generate a target fault feature vector; A fault identification module for inputting the target fault feature vector into an improved PSO-LS-SVM classification and diagnosis model for fault identification to generate a target grounding grid corrosion fault identification result.

8. A computer device, characterized in that, It includes a memory and a processor. A computer program is stored in the memory. When the computer program is executed by the processor, the processor is caused to execute the steps of the method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm according to any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer is caused to execute the method for identifying the corrosion fault of the distribution network grounding grid based on the PSO-LS-SVM algorithm according to any one of claims 1-6.

Images