Current transformer error solving method, system, equipment and medium

Abstract

The invention belongs to the technical field of electric power measurement, and discloses a current transformer error solving method, system, device and medium, and the method comprises the following steps: collecting the operation data of a current transformer, constructing a sample analysis data set, carrying out the component decomposition and feature extraction of the sample data set, and calculating the statistical threshold value of the sample data set; judging the condition of the current transformer, and constructing an in-phase current transformer sample; a calculation step based on a statistical magnitude threshold value: calculating the statistical magnitude of each sample in the same-phase current transformer samples, comparing the statistical magnitude with the statistical magnitude threshold value, obtaining an analysis result of the state of the current transformer, and further obtaining a related solution threshold value; a particle swarm is initialized through a multi-target weighted fitness function model and improved Logistic mapping; based on the fitness of each particle, cooperatively optimizing a particle swarm by combining a grey wolf algorithm, a particle swarm optimization algorithm and a Levy flight disturbance mechanism, and iteratively solving the position of each particle; based on a double-threshold mechanism, whether the current transformer really has a problem is judged.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power measurement, and particularly to a current transformer error solving method, system, device and medium. BACKGROUND

[0002] In modern power systems, current transformers (CTs) are key devices that ensure the reliable operation and accurate measurement of the system. The measurement accuracy of CTs plays a crucial role in the stable and efficient operation of the power system. CTs are mainly used to scale AC current to provide accurate current signals for measuring instruments, relay protection devices, etc. However, due to various factors such as manufacturing process, operating environment (such as temperature, humidity, electromagnetic interference, etc.), and device aging caused by long-term use, CTs inevitably produce errors in the form of ratio and phase differences during actual operation.

[0003] These seemingly small errors can cause a series of serious problems. In terms of power measurement, CT errors can directly lead to inaccurate energy measurement, causing economic disputes between power companies and users; in the field of relay protection, inaccurate current measurement values can cause relay protection devices to malfunction or refuse to act, threatening the safe and stable operation of the power system, and even causing large-scale power outages, causing great losses to social production and people's lives.

[0004] Traditional CT error solving methods are mainly based on analytical algorithms and empirical formulas, which have obvious limitations. On the one hand, analytical algorithms often rely on simplified model assumptions, making it difficult to accurately reflect the complex actual operating characteristics of CTs, resulting in large deviations between calculated results and actual errors; on the other hand, empirical formulas are usually derived under specific conditions and lack universality, making them less adaptable to different types and operating environments of CTs. In addition, traditional methods are generally prone to local optimal solutions, and when faced with complex error solving scenarios, they cannot find global optimal solutions, thereby severely restricting the improvement of CT error solving accuracy.

[0005] With the continuous development and intelligent upgrading of power systems, the requirement for CT measurement accuracy is increasing, and existing error solving methods have been difficult to meet actual needs. Therefore, there is an urgent need for an innovative, efficient and accurate CT error solving method to address the shortcomings of traditional methods and ensure the safe and reliable operation and accurate measurement of power systems.

[0006] Therefore, based on the above technical problems, the present application proposes a current transformer error solving method with accurate measurement accuracy, efficient operation, and the ability to ensure the stable operation of the power system. SUMMARY

[0007] The purpose of the present application is to overcome the shortcomings of the prior art, provide an innovative, efficient and accurate current transformer error solving method, to solve the shortcomings of the traditional method, and ensure the safe and reliable operation and accurate measurement of the power system.

[0008] In order to achieve the above-mentioned purpose, the present application provides a current transformer error solving method, which comprises the following steps: Step S1: Collecting the operation data of the current transformers of each measuring line at the same node of the substation within a first preset time period to construct a sample analysis data set, and combining the principal component analysis method to perform component decomposition and feature extraction on the sample analysis data set, to establish an initial steady state, and then calculate the statistical threshold value in the sample data set; Step S2: Collecting the original waveform data of the current transformers of each measuring line at the same node of the substation within a second preset time period, filtering the original waveform data based on the sliding window mean filtering algorithm to obtain filtered original waveform data, and combining the fast Fourier transform method to calculate the filtered original waveform data to obtain the real-time condition judgment of each measuring line current transformer, and then constructing the in-phase current transformer sample; Step S3: Based on the calculation step of the statistical threshold value in step S1, the statistical value of each sample in the in-phase current transformer sample is calculated and the initial steady state statistical threshold value is compared to obtain the preliminary analysis result of the current transformer state, and then the solving threshold value of the ratio error and phase error in the current transformer is set according to the accuracy level of the current transformer; Step S4: Based on the KCL principle, an initial multi-objective weighted fitness function model is established, taking the ratio error and phase error of each current transformer as the solving object and the minimum of the node current vector as the optimization target; and after obtaining the initial multi-objective weighted fitness function model, the initial multi-objective weighted fitness function model is constrained based on the solving threshold value of the ratio error and phase error of the current transformer in step S3 to obtain an optimized multi-objective weighted fitness function model; Step S5: After obtaining the optimized multi-objective weighted fitness function model, an improved Logistic mapping is introduced to initialize the particle swarm; and based on the fitness of each particle, the particle swarm is optimized in coordination by combining the grey wolf algorithm, the particle swarm optimization algorithm and the Levy flight disturbance mechanism, to iteratively update the position of each particle, until the preset stop condition is met, to obtain the final solving result of the ratio error and phase error, wherein in the iteration process, whether each particle is updated is limited based on the solving threshold value of the ratio error and phase error of the current transformer in step S3; Step S5: After obtaining the optimized multi-objective weighted fitness function model, an improved Logistic mapping is introduced to initialize the particle swarm; and based on the fitness of each particle, the particle swarm is optimized in coordination by combining the grey wolf algorithm, the particle swarm optimization algorithm and the Levy flight disturbance mechanism, to iteratively update the position of each particle, until the preset stop condition is met, to obtain the final solving result of the ratio error and phase error, wherein in the iteration process, whether each particle is updated is limited based on the solving threshold value of the ratio error and phase error of the current transformer in step S3; Step S6: Based on the double threshold mechanism, a hypothesis testing model with two judgment thresholds is constructed, the preliminary analysis result in step S3 is taken as the first discriminant object of the hypothesis testing model, and the solving result of the final difference and phase difference obtained in step S5 is taken as the second discriminant object of the hypothesis testing model. Based on the two judgment thresholds of the hypothesis testing model, the first discriminant object and the second discriminant object are compared and analyzed respectively. If the hypothesis testing model shows that both discriminant objects think that there is an over-limit problem in the current transformer at this place within a specified time period, it is determined that there is indeed an over-limit problem in the current transformer at this place.

[0009] Further, the step S1 comprises the following steps: Step S1.1: Collecting the operation data of the current transformers of the same phase of each branch under the same node of the transformer substation within a first preset time period to establish a sample analysis data set in the form of a current transformer vector, and establishing a kernel matrix composed of the sample analysis data set based on the kernel function method, wherein the operation data is the amplitude and phase value of the current transformer; Step S1.2: Transforming the kernel matrix based on the centering processing method to eliminate the mean shift in the feature space and obtain the kernel matrix after centering processing; Step S1.3: Performing feature decomposition on the kernel matrix after centering processing to obtain each eigenvalue and each eigenvector corresponding to each eigenvalue; Step S1.4: After obtaining each eigenvalue and each eigenvector corresponding to each eigenvalue, sorting each eigenvalue in descending order, combining the calculation principle of cumulative contribution rate, obtaining the first k larger eigenvalues with a cumulative contribution rate greater than or equal to 85%, and constructing a variance contribution diagonal matrix and a principal component direction matrix composed of the first k larger eigenvalues and the eigenvectors corresponding to the first k larger eigenvalues; Step S1.5: After obtaining the principal component direction matrix , calculating the principal component projection vector of each sample in the sample analysis data set; Step S1.6: After obtaining the principal component projection vector of each sample, calculating the sum of squares of the residual vectors to obtain statistic; Step S1.7: After obtaining the set of Q statistics, combining the quantile method to take the 99th quantile of the normal samples in the sample analysis data set as the optimal discriminant threshold, or combining the chi-square distribution method to determine the statistic threshold under the preset confidence level.

[0010] Further, the step S2 comprises the following steps: Step S2.1: Collecting raw waveform data of each measuring line current transformer under the same node of the transformer substation within a second preset time period through a high-precision sensor, wherein the raw waveform data is the secondary side current waveform data of the current transformer; Step S2.2: Filtering the raw current waveform data based on a sliding window mean filtering algorithm to obtain filtered waveform data; Step S2.3: Calculating the amplitude and phase values of each harmonic and the total harmonic distortion rate of each current transformer based on the filtered waveform data and combining fast Fourier transform, and then constructing the same-phase current transformer sample; Step S2.4: When the total harmonic distortion rate of a certain current transformer exceeds a preset value one or the amplitude of a certain harmonic thereof is greater than a preset value two, it is determined that the current transformer has harmonic interference.

[0011] Further, the step S3 includes the following steps: Step S3.1: After obtaining the same-phase current transformer sample, combining the step S1 statistical quantity threshold calculation step, calculating the sample statistical quantity; Step S3.2: Comparing the sample statistical quantity with the statistical quantity threshold in step S1.6, if the statistical quantity exceeds the statistical quantity threshold , it is determined that the current transformer in the transformer substation may have a problem, and the fault separation algorithm is used to locate the current transformer that may have a problem; Step S3.3: After obtaining the preliminary analysis result of the current transformer state, combining the accuracy level of the current transformer, and dividing the solving threshold of the ratio difference and the phase difference in the current transformer, wherein when the current transformer is preliminarily determined to be problem-free in step S3.2, and the accuracy level corresponding to the current transformer is 0.1S level, the ratio difference solving threshold of the current transformer is set to [-0.1%, 0.1%], and the phase difference solving threshold is set to [-5', 5']; Step S3.4: When the current transformer is preliminarily determined to have a problem in step S3.2, expanding the coverage range of the ratio difference solving threshold and the phase difference solving threshold of the current transformer according to the accuracy level corresponding to the current transformer and the regulation error limit.

[0012] Further, the step S4 includes the following steps: Step S4.1: after obtaining the samples of each current transformer, based on the KCL principle, a multi-objective weighted fitness function model is established, taking the ratio error and phase error of each current transformer as the solving object, and the sum of the primary current vectors as the optimization target, the current transformer error solving problem is converted into the minimization problem of the sum of the primary current vectors in the multi-objective weighted fitness function model, the formula of the initial multi-objective weighted fitness function model is: , wherein, is the ratio error of the CT on the i-th line in the evaluation group, is the phase error of the CT on the i-th line in the evaluation group, is the secondary side current phasor of the i-th sample.

[0013] Step S4.2: based on the initial multi-objective weighted fitness function model, the initial multi-objective weighted fitness function model is constrained by combining the solving threshold of the ratio error and phase error of the current transformer in step S3.3, and an optimized multi-objective weighted fitness function model is obtained, the formula of the optimized multi-objective weighted fitness function model is: , , wherein, is the ratio error of the CT on the i-th line in the evaluation group, is the phase error of the CT on the i-th line in the evaluation group, is the accuracy level of the CT, is the rated secondary current value of the CT, is the secondary side current phasor of the i-th sample.

[0014] Further, the step S5 includes the following steps: Step S5.1: after obtaining the optimized multi-objective weighted fitness function model, the positions of the initial particles are randomly generated in the search space, the improved Logistic mapping is introduced to diffuse the initial particle group, and the best position Pbest and the global best position Gbest of each particle are set; Step S5.2: based on the optimized multi-objective weighted fitness function model, the fitness of each initial particle is calculated, the particles ranked first, second and third in fitness are taken as α, β and δ particles, and the remaining particles are taken as ω particles, and the wolf position is calculated based on the positions of the α, β and δ particles; Step S5.3: based on the grey wolf algorithm, the updated positions of the alpha, beta and delta particles are calculated, and based on the updated positions of the alpha, beta and delta particles, the updated positions of the wolves are calculated; the position of the omega particle is updated based on the particle swarm optimization algorithm, and noise interference is applied to the omega particle by combining the Levy flight disturbance mechanism to enhance the global search ability of the omega particle; Step S5.4: if the updated positions of the particles or the wolves are within the solution threshold range of the ratio error and phase error of the current transformer, and the updated fitness of each particle is better than the best fitness corresponding to each particle, the position of each particle is updated to the current position, and the current fitness of each particle is recorded as the best fitness, wherein if the best fitness of a particle is better than the global best fitness, the current position of the particle is set as the best position Gbest; Step S5.5: based on the current fitness of each particle, the top three particles in the fitness ranking are re-set as the alpha, beta and delta particles, and steps S5.2-S5.4 are repeated to iteratively solve the ratio error and phase error of each current transformer until the preset stopping condition is met, the iteration of the particle swarm is stopped, and the global best position Gbest is output to obtain the optimal solution of the current current transformer error value, wherein the preset stopping condition is that when the descending rate of the optimized multi-objective weighted fitness function model is detected for 3 times in a row < 0.0001, or the iteration number of the particle swarm reaches 200 times.

[0015] Further, the step S6 comprises the following steps: Step S6.1: based on the double threshold mechanism, a hypothesis testing model with two judgment thresholds is constructed, and the preliminary analysis result in step S3 is taken as the first discriminant object of the hypothesis testing model, and the final solution of the ratio error and phase error obtained in step S5 is taken as the second discriminant object of the hypothesis testing model; Step S6.2: one of the judgment thresholds of the hypothesis testing model is set as: if the Mahalanobis distance in the kernel space of a sample in the current transformer sample exceeds the Mahalanobis distance threshold in the kernel space, the hypothesis testing model triggers an early warning, and the fault separation algorithm is used to locate the current transformer with problems, and then the current transformer corresponding to the data is reported to exist abnormal condition; Step S6.3: the other judgment threshold of the hypothesis testing model is set as: when it is detected that the current transformer error output by the optimized multi-objective weighted fitness function model exceeds the error threshold of the corresponding current transformer level in the regulation, the hypothesis testing model triggers an early warning, and the current transformer corresponding to the data is reported to exist abnormal condition; Step S6.4: based on the two judgment thresholds of the hypothesis testing model, the first discriminant object and the second discriminant object are compared and analyzed respectively; ​Step S6.5: If within a specified time period, the assumption test model has 80% of the results showing that both discriminators believe that the current transformer at the location has an overrange problem, then it is determined that the current transformer at the location has an overrange problem.

[0016] Based on the same inventive concept, the application also provides a current transformer error solving system, a statistical quantity threshold calculation module, a same-phase current transformer sample construction module, a current transformer state analysis module, a solving model construction module, a particle swarm optimization module, and a double threshold judgment module. The statistical quantity threshold calculation module is used to collect operation data of current transformers of each measuring line at a same node of a substation in a first preset time period, to construct a sample analysis data set, and to combine a principal component analysis method to perform component decomposition and feature extraction on the sample analysis data set, establish an initial steady state, and then calculate the statistical quantity threshold in the sample data set. The same-phase current transformer sample construction module is used to collect original waveform data of current transformers of each measuring line at a same node of a substation in a second preset time period, to perform filtering processing on the original waveform data based on a sliding window mean filtering algorithm to obtain filtered original waveform data, and to combine a fast Fourier transform method to calculate the filtered original waveform data to obtain a real-time condition judgment of each measuring line current transformer, and then construct a same-phase current transformer sample. The current transformer state analysis module is used to calculate the statistical quantity of each sample in the same-phase current transformer sample based on the calculation step of the statistical quantity threshold in step S1 of the above-mentioned current transformer error solving method, compare the statistical quantity of each sample with the statistical quantity threshold of the initial steady state to obtain a preliminary analysis result of the current transformer state, and then set the solving threshold of the ratio error and the phase error of the current transformer in combination with the accuracy level of the current transformer. The solving model construction module is used to establish an initial multi-objective weighted fitness function model taking the ratio error and the phase error of each current transformer as the solving object and the minimum of the node primary current vector as the optimization target based on the KCL principle. After obtaining the multi-objective weighted fitness function model, the solving threshold of the ratio error and the phase error of the current transformer in step S3 of the above-mentioned current transformer error solving method is combined to constrain the initial multi-objective weighted fitness function model to obtain an optimized multi-objective weighted fitness function model. The particle swarm optimization module is used to introduce an improved Logistic mapping to initialize the particle swarm after obtaining the optimized multi-objective weighted fitness function model. Based on the fitness of each particle, the particle swarm is cooperatively optimized by combining a grey wolf algorithm, a particle swarm optimization algorithm, and a Levy flight disturbance mechanism to iteratively update the position of each particle until a preset stop condition is met, and the solving result of the final ratio error and phase error is obtained. In the iteration process, whether each particle is updated is limited in combination with the solving threshold of the ratio error and the phase error of the current transformer in step S3 of the above-mentioned current transformer error solving method.The double-threshold judgment module is used for constructing a hypothesis test model with two judgment thresholds based on a double-threshold mechanism, taking the preliminary analysis result in step S3 of the above one kind of current transformer error solving method as a first discrimination object of the hypothesis test model, taking the solving result of the final difference and phase difference obtained in step S5 of the above one kind of current transformer error solving method as a second discrimination object of the hypothesis test model, and comparing and analyzing the first discrimination object and the second discrimination object respectively based on the two judgment thresholds of the hypothesis test model, wherein if the hypothesis test model has 80% of the results showing that both discrimination objects think that the current transformer at the place has an over-limit problem within a specified time length, it is determined that the current transformer at the place indeed has an over-limit problem.

[0017] Based on the same inventive concept, the application also provides a data processing device based on a current transformer error solving method, comprising a memory and a processor, the memory is used for storing a computer program, and the processor is used for executing the computer program to realize the steps of the above-mentioned data processing method based on a current transformer error solving method.

[0018] Based on the same inventive concept, the application also provides a computer readable storage medium: the computer readable storage medium stores a computer program, and the computer program is executed by a processor to realize the steps of the above-mentioned current transformer error solving method.

[0019] The application adopts the above-mentioned scheme, which has the following advantages:

[0020] 1) The calculation accuracy is significantly improved: by fusing the grey wolf algorithm and the particle swarm optimization algorithm, and introducing the Levy flight disturbance mechanism, the problem that the traditional algorithm is easy to fall into local optimum is effectively overcome, the optimal solution of CT error can be more accurately searched in the complex solution space, and the accuracy of error calculation is improved.

[0021] 2) Enhance the robustness of the algorithm: the Levy flight disturbance mechanism enables the algorithm to jump out of the local optimal region during the search process, increases the randomness and diversity of the search, improves the adaptability of the algorithm to different working conditions and data characteristics, and enhances the robustness of the algorithm.

[0022] 3) Improve the stability of the power system: accurate CT error calculation can provide reliable data support for protection, control and metering of the power system, avoid misoperation of relay protection, inaccurate power metering and other problems caused by current measurement error, and improve the stability and reliability of the power system, and ensure the safe operation of the power system. In the event of a power system fault, accurate current measurement can enable the relay protection device to act in time and accurately, and remove the faulty line to prevent the fault from expanding. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 This is a flowchart illustrating the current transformer error solving method provided by the present invention.

[0024] Figure 2 This is a schematic diagram illustrating the specific process of the current transformer error solving method provided by the present invention.

[0025] Figure 3 This is a schematic diagram of the current transformer error solving system provided by the present invention. Detailed Implementation

[0026] To facilitate understanding of the present invention, a more complete description is given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention. Example 1

[0027] See appendix Figures 1-2 As shown in the figure, this embodiment provides a method for solving the error of a current transformer, including the following steps: Step S1: Collect the operating data of current transformers on each measuring line under the same node of the substation within the first preset time period to construct a sample analysis dataset. Then, combine the principal component analysis method to perform component decomposition and feature extraction, establish an initial steady state, and then calculate the current transformers in the sample dataset. Threshold for statistics; the Q statistic test can determine whether a time series is stationary (i.e., its statistical properties do not change over time). Step S1 includes the following steps: Step S1.1: Collect the operating data of the in-phase current transformers of each branch under the same node of the substation within the first preset time period (the operating data is the amplitude and phase values ​​of the current transformers, which are obtained by multiple measurements) to establish a sample analysis dataset in vector form of the current transformers. Based on the kernel function method, establish a kernel matrix composed of the sample analysis dataset, that is, use the operating data of the in-phase current transformers related to the same node as the training dataset to form the kernel matrix: , Where n is the number of samples and m is the number of elements in a single sample collection; Calculate the kernel matrix (of the training dataset). The element in the i-th row and j-th column of the matrix is: , , in, For kernel functions, common kernel functions include the Gaussian kernel (Radial Basis Function, RBF), the p-order polynomial kernel, and the sigmoid kernel, etc., and the specific choice needs to be made based on the data; x i Let x be the vector of the i-th row of the data matrix x. j Let be the j-th column vector of the data matrix x.

[0028] The Gaussian kernel function (RBF) described above can handle complex nonlinear relationships, and its expression is as follows: , in, is a key parameter of the Gaussian kernel function.

[0029] Step S1.2: Transform the kernel matrix using a centering method to eliminate mean shift in the feature space, obtaining the centered kernel matrix. The specific expression is: , in, It is an n×n identity matrix.

[0030] Step S1.3: For the centered kernel matrix Eigenvalue decomposition is performed to obtain each eigenvalue and its corresponding eigenvector, which is equivalent to solving for each eigenvalue. With each feature vector The specific formula is as follows: .

[0031] Step S1.4: After obtaining each eigenvalue and its corresponding eigenvector, sort the eigenvalues ​​in descending order. Based on the calculation principle of cumulative contribution rate, obtain the k largest eigenvalues ​​with a cumulative contribution rate greater than or equal to 85%. And based on the first k largest eigenvalues Construct the variance contribution diagonal matrix and the first k largest eigenvalues Eigenvectors with one-to-one correspondence The principal component direction matrix formed .

[0032] Step S1.5: Obtain the principal component direction matrix Then, calculate the principal component projections for each sample in the sample analysis dataset. The specific expression is: , in, for The i-th row vector.

[0033] Step S1.6: Obtain the principal component projections for each sample. Then, by calculating the sum of squares of the residual vectors, we obtain... The statistics are as follows: , In practical calculations, the above formula can be simplified to: , in, The value of the kernel function. For a single sample The calculated Q statistic (i.e., It is the specific manifestation of the Q statistic on a single sample, obtained through a specific calculation method for each sample. The value can then be used to analyze the characteristics of each sample in the residual space. It should be noted that the above kernel function With kernel function value The application of the same function in different steps, where the kernel function... Used to represent sample x i and x j Similarity in high-dimensional feature spaces can capture nonlinear relationships between samples, from kernel function values. Used to represent the "self-benchmark" of a sample in a high-dimensional space, i.e., sample self-similarity, it can be used as a benchmark for residual calculation and measures the deviation of a sample from its principal component projection.

[0034] Step S1.7: After obtaining the set of Q statistics, combine the quantile method with the 99th percentile of the normal samples in the sample analysis dataset as the optimal discrimination threshold, or combine the chi-square distribution method to determine the preset reliability level (e.g., 95%). Statistical threshold; It should be noted that the principle of the aforementioned kernel principal component analysis (KPCA) technique is as follows: The original nonlinear characteristics of the current transformer (CT) are mapped to a high-dimensional feature space using a kernel function. Principal components are then extracted in this high-dimensional space to construct a more accurate anomaly detection boundary, making it more suitable than traditional PCA for handling nonlinear distortion problems caused by CT saturation. Secondly, the aforementioned relevant Q statistic follows a chi-square distribution, which can determine whether there is a significant difference between the actual value and the theoretical expected value, thereby improving... The reliability of the statistical threshold and the specific steps of the chi-square distribution method are readily apparent to those working in this field, and will not be elaborated upon here.

[0035] Step S2: Obtain the sample dataset After the statistical quantity threshold, the original waveform data of the current transformers of each measuring line under the same node of the transformer substation in a second preset time period is collected, the original waveform data is filtered based on a sliding window mean filtering algorithm to obtain filtered original waveform data, and the filtered original waveform data is calculated based on a fast Fourier transform method to obtain a real-time condition judgment of each measuring line current transformer, and then a same-phase current transformer sample is constructed.

[0036] Step S2.1: The original waveform data of the same-phase current transformers of each measuring line under the same node of the transformer substation in a second preset time period is collected by a high-precision sensor, wherein the original waveform data is the secondary-side current waveform data of the current transformer, and the second preset time period is after the first preset time period (the specific selection of the first preset time period and the second preset time period can be selected according to actual conditions).

[0037] Step S2.2: The original current waveform data is filtered based on a sliding window mean filtering algorithm (window length 200 ms) to obtain preprocessed (filtered) waveform data (the sliding window mean filtering algorithm and the harmonic amplitude calculation formula described above belong to common algorithm procedures, and the principle thereof is easily thought of by a person skilled in the art, and thus will not be described again); by the sliding window mean filtering algorithm, transient noise can be eliminated, and data points deviating from the mean value ± 3 times the standard deviation are linearly interpolated and corrected, which is more convenient for inputting the corrected current data into a subsequent multi-objective weighted fitness function model.

[0038] Step S2.3: Based on the filtered waveform data, the amplitudes and phase values of each harmonic of each current transformer and the total harmonic distortion rate are calculated in combination with a fast Fourier transform (FFT), and then a same-phase current transformer sample is constructed, and the calculation formula of the total harmonic distortion rate is:

[0039] Step S2.4: When the total harmonic distortion rate of a certain current transformer exceeds 5% (preset value one) or the amplitude of a certain harmonic (such as the 5th harmonic or the 7th harmonic) of the current transformer is greater than 3% (preset value two) of the fundamental wave, it is determined that the current transformer is disturbed by harmonics (under the harmonic distortion signal, the transmission characteristics of the current transformer will cause errors, and then affect the electric energy metering).

[0040] Step S3: Based on the residual space features obtained by the kernel principal component analysis method in step S1 (in combination with the initial steady state) and the same-phase current transformer sample constructed in step S2, the multi-objective weighted fitness function model is constructed. ​​​​​The steps for calculating the statistical threshold are as follows: calculate the threshold values ​​for each sample in the in-phase current transformer sample. Statistic and each sample Statistics and initial steady state Statistical thresholds are compared to obtain preliminary analysis results of the current transformer's state. Then, combined with the accuracy level of the current transformer, the calculation thresholds for the ratio difference and phase difference in the current transformer (i.e., the error limits for the ratio difference and phase difference) are set. During the normal operation of the current transformer, the ratio difference and phase difference will affect the accuracy of its power metering. Therefore, it is necessary to reduce the impact of the ratio difference and phase difference on the normal operation of the current transformer. Secondly, the specific calculation formulas for the ratio difference and phase difference are easy for those engaged in this field to understand, and will not be elaborated here.

[0041] Step S3.1: After obtaining the in-phase current transformer sample, combine it with the information from step S1. The steps for calculating the threshold of a statistic, including calculating the sample threshold. The statistics and kernel matrix are specifically for each test sample in the in-phase current transformer sample. The same kernel function method is used to calculate the kernel matrix of the sample and the training dataset. The i-th element in the kernel matrix is: , Decentralize the new kernel matrix: , Map the test samples onto the principal component directions of the high-dimensional feature space and calculate their projection: , Simplify the calculation of samples using kernel functions Statistic: .

[0042] Step S3.2: Transfer the sample Statistics and steps in S1.6 If the statistical thresholds are compared, Statistical quantity exceeds Statistical threshold This indicates that the current transformer group within the substation may have an out-of-tolerance problem. A fault separation algorithm is used to locate the current transformers potentially causing this problem. The contribution rate of the variables to the residual space is as follows: , Normalization can be used to calculate the specific percentage contribution of each element in a single measurement, as follows: .

[0043] Step S3.3: After obtaining the preliminary analysis result of the current transformer state, the solving threshold of the ratio error and the phase error in the current transformer is divided in combination with the accuracy level of the current transformer, wherein when the current transformer is preliminarily determined as no problem in step S3.2 and the accuracy level corresponding to the current transformer is 0.1S level, the solving threshold of the ratio error of the current transformer is set as [-0.1%, 0.1%] and the threshold of the phase error is [-5', 5'].

[0044] Step S3.4: When the current transformer is preliminarily determined as having a problem in step S3.2, the solving threshold of the ratio error and the phase error in the current transformer is expanded in combination with the accuracy level corresponding to the current transformer and the regulation error limit.

[0045] Step S4: After obtaining the samples of the in-phase current transformers, an initial multi-objective weighted fitness function model taking the ratio error and the phase error of each current transformer as the solving objects and the sum of the node primary current vectors and the minimum as the optimization target is established based on the KCL principle; and after obtaining the multi-objective weighted fitness function model, the initial multi-objective weighted fitness function model is constrained in combination with the solving threshold of the ratio error and the phase error of the current transformer in step S3 to obtain an optimized multi-objective weighted fitness function model.

[0046] Step S4.1: After obtaining the samples of the in-phase current transformers, an initial multi-objective weighted fitness function model taking the ratio error and the phase error of each current transformer as the solving objects and the sum of the node primary current vectors and the minimum as the optimization target is established based on the KCL principle, and the formula of the initial multi-objective weighted fitness function model is: , wherein, is the ratio error of the CT on the i-th line in the evaluation group, is the phase error of the CT on the i-th line in the evaluation group, is the secondary side current phasor of the i-th sample.

[0047] Step S4.2: After obtaining the initial multi-objective weighted fitness function model, the initial multi-objective weighted fitness function model is constrained in combination with the solving threshold of the ratio error and the phase error of the current transformer in step S3.3 to obtain an optimized multi-objective weighted fitness function model, and the formula of the above-mentioned optimized multi-objective weighted fitness function model is: , , wherein, is the ratio error of the CT on the i-th line in the evaluation group, to evaluate the phase error of the CT on the i-th line in the group, to evaluate the accuracy class of the CT, to evaluate the rated secondary current value of the CT, to evaluate the secondary-side current phasor of the i-th sample; The above s.t. is the abbreviation of "subject to", meaning "subject to" or "subject to the following constraints", which is used to separate the objective function of the optimization problem from the constraint conditions, indicating that when finding the optimal solution, the solution must meet both the optimization direction of the objective function and all the constraint conditions, so as to ensure that the optimization result is both minimized error and meets the physical limitations of the CT and the accuracy class of the CT.

[0048] Step S5: After obtaining the optimized multi-objective weighted fitness function model, an improved Logistic mapping is introduced to initialize the particle swarm; and based on the fitness of each particle, the particle swarm is cooperatively optimized by combining the grey wolf algorithm, the particle swarm optimization algorithm and the Levy flight disturbance mechanism, so as to iteratively update the position of each particle, until the preset stop condition is met, the solving result of the final ratio error and phase error is obtained, wherein, in the iteration process, the solving threshold of the current transformer ratio error and phase error in step S3 is combined to limit whether each particle is updated.

[0049] Step S5.1: After obtaining the optimized multi-objective weighted fitness function model, the position and update speed of the initial particles and other parameters are randomly generated in the search space, and then the improved Logistic mapping is introduced to diffuse the initial particle swarm (generally, the initial particle swarm is 30-60 by default, in order to balance the calculation accuracy and single calculation speed), and the best position of the i-th particle and the global best position of all particles are set. In the above initial particles, the random generation formula of the position of each particle is: , The diffusion formula of the improved Logistic mapping is: , wherein, and respectively represent the minimum and maximum values of the j-th dimension, is a random number generated in [0, 1]; is a random number following the standard normal distribution; Secondly, the speed initialization formula can be consistent with the position formula, and the speed boundary is usually 10%-20% of the position boundary.

[0050] It should be noted that an initial position is generated for the sample as an initial particle position through the improved Logistic chaotic mapping, iteration calculation is performed to ensure the diversity of the initial position of the particle, premature convergence of the particle in the optimization process is avoided, and the algorithm can effectively avoid falling into a local optimal solution and improve the global search ability, so that the algorithm is more accurate and efficient.

[0051] Step S5.2: Calculate the fitness of each initial particle based on the optimized multi-objective weighted fitness function model. Since the objective of the optimized multi-objective weighted fitness function model is the minimum value, the fitness function can take the absolute value of its reciprocal. The specific formula is: , The higher the fitness is, the better the position is. The top three particles in the fitness ranking are selected as the alpha, beta and delta particles, and the corresponding positions are , and The remaining particles are omega particles, and the wolf position is calculated based on the positions of the alpha, beta and delta particles . The specific calculation formula is: , The wolf position is determined by calculating the average of the positions of the alpha, beta and delta particles, which can guide other particles to move to more potential areas, thereby improving the search efficiency of the algorithm, avoiding the search process from falling into a local optimal solution, and helping to explore more widely in the solution space to find better solutions.

[0052] Step S5.3: Update the positions of the alpha, beta and delta particles based on the grey wolf algorithm. Specifically: First, generate random coefficients A and C: , , where a is a convergence factor (linearly decaying from 2 to 0 according to the number of iterations), and are random numbers in the range [0, 1]; Calculate the positions of the updated alpha, beta and delta particles, and update the position of the wolf based on the positions of the updated alpha, beta and delta particles : , , , , Calculate the position of the updated omega particle based on the particle swarm optimization algorithm. The specific formula is: , , wherein, is an inertia weight, and is a learning factor, and is a random number ranging in [0, 1]; and a Levy flight disturbance mechanism is combined to randomly disturb the position of the ω particle (noise interference is applied), so that the global search ability of the ω particle is enhanced, and the formula is: , wherein, is a Levy flight intensity coefficient (0.5 by default), is 1.5, is a random number subject to a standard normal distribution, is a hierarchical weight of the grey wolf algorithm.

[0053] Step S5.4: If the updated position of each particle or wolf is within the solution threshold range of the ratio difference and phase difference of the current transformer, and the fitness of the updated position of each particle is better than the best fitness corresponding to each particle, the position of each particle is updated to the current position, and the current fitness of each particle is recorded as the best fitness, wherein if the best fitness of a particle is better than the global best fitness, the current position of the particle is set as the best position Gbest.

[0054] Step S5.5: Based on the current fitness of each particle, the particles with the top three fitnesses are re-set as the α, β and δ particles, and steps S5.2-S5.4 are repeated to iteratively solve the ratio difference and phase difference of each current transformer, until a preset stop condition is met, the iteration of the particle swarm is stopped, and the global best position Gbest is output, and the optimal solution of the current current transformer error value is obtained, wherein the preset stop condition is that when the descending rate of the optimized multi-objective weighted fitness function model is < 0.001 for 3 times in succession, or the iteration number of the particle swarm reaches 200 times.

[0055] Unlike the traditional algorithm, the error calculation result will have a large deviation under complex working conditions such as harmonic interference in the power system. The grey wolf algorithm used in the embodiment can still accurately calculate the CT error under the same conditions. The introduction of the Levy flight disturbance mechanism enables the grey wolf algorithm to jump out of the local optimal region in the search process, thereby increasing the randomness and diversity of the search, improving the adaptability of the grey wolf algorithm to different working conditions and data characteristics, and improving the CT error accuracy level from 0.2 level of the traditional algorithm to 0.1 level. The accuracy of error calculation is improved, and the robustness of the algorithm is enhanced.

[0056] Step S6: Based on the double threshold mechanism, a hypothesis testing model with two judgment thresholds is constructed, the preliminary analysis result in step S3 is taken as the first discriminant object of the hypothesis testing model, the comparison analysis result of the final ratio difference and phase difference obtained in step S5 and the ratio difference and phase difference in step S3 is taken as the second discriminant object of the hypothesis testing model, and based on the two judgment thresholds of the hypothesis testing model, the first discriminant object and the second discriminant object are compared and analyzed respectively, and if the hypothesis testing model has 80% of the results showing that both discriminant objects think that there is an error problem of the current transformer at the place within a preset time period 3, it is determined that there is indeed an error problem of the current transformer at the place.

[0057] Step S6.1: Based on the double threshold mechanism, a hypothesis testing model with two judgment thresholds is constructed, and the preliminary analysis result in step S3 is taken as the first discriminant object of the hypothesis testing model, and the solving result of the final ratio difference and phase difference obtained in step S5 is taken as the second discriminant object of the hypothesis testing model.

[0058] Step S6.2: One of the judgment thresholds of the hypothesis testing model is set to: if the Mahalanobis distance of a sample in the in-phase current transformer sample exceeds the Mahalanobis distance threshold in the kernel space, the hypothesis testing model triggers a warning, and through a fault separation algorithm, the current transformer with the problem is located, and then the current transformer corresponding to the data is reported to exist an abnormal situation.

[0059] Step S6.3: The other judgment threshold of the hypothesis testing model is set to: when it is detected that the error of the current transformer output by the optimized multi-objective weighted fitness function model exceeds the error threshold of the corresponding current transformer level in the regulation, the hypothesis testing model triggers a warning, and reports that the current transformer corresponding to the data exists an abnormal situation.

[0060] Step S6.4: Based on the two judgment thresholds of the hypothesis testing model, the first discriminant object and the second discriminant object are compared and analyzed respectively.

[0061] Step S6.5: If the hypothesis testing model has 80% of the results showing that both discriminant objects think that there is an error problem of the current transformer at the place within a specified time period (which can be set according to relevant regulations or actual situation), it is determined that there is an error problem of the current transformer at the place.

[0062] In summary, the current transformer error solving method in the embodiment fuses the advantages of the Grey Wolf Optimizer (GWO) and the Particle Swarm Optimization (PSO), specifically including a Levy flight perturbation mechanism and a multi-objective weighted fitness function design; considering that the Particle Swarm Optimization (PSO) has the characteristic of fast convergence speed, but due to the sensitivity of its parameters, the Particle Swarm Optimization is prone to fall into a local optimal solution, thereby leading to deviation from the global optimal result; in view of this problem, the current transformer error solving method in the embodiment selects to combine the idea of the Grey Wolf Optimizer (GWO), in the search process, based on the role setting of different levels of “grey wolves” (such as alpha, beta, and delta wolves), thereby guiding the particle swarm to search for the global optimal solution more efficiently; secondly, accurate CT error calculation can provide reliable data support for the protection, control, and metering of the power system, avoid problems such as relay protection misoperation and inaccurate power metering caused by current measurement error, thereby improving the stability and reliability of the power system, and ensuring the safe operation of the power system; in the event of a power system fault, accurate current measurement can enable the relay protection device to act in time and accurately, and remove the faulty line to prevent the fault from expanding. Example 2

[0063] Reference Figure 3As shown, based on the same inventive concept, the application also provides a current transformer error solving system, a statistical quantity threshold calculation module, a same-phase current transformer sample construction module, a current transformer state analysis module, a solving model construction module, a particle swarm optimization module, and a double threshold judgment module. The statistical quantity threshold calculation module is used to collect operation data of current transformers of each measuring line at a same node of a substation in a first preset time period, to construct a sample analysis data set, and to combine a principal component analysis method to perform component decomposition and feature extraction on the sample analysis data set, establish an initial steady state, and then calculate the statistical quantity threshold in the sample data set. The same-phase current transformer sample construction module is used to collect original waveform data of current transformers of each measuring line at a same node of a substation in a second preset time period, to perform filtering processing on the original waveform data based on a sliding window mean filtering algorithm to obtain filtered original waveform data, and to combine a fast Fourier transform method to calculate the filtered original waveform data to obtain a real-time condition judgment of each measuring line current transformer, and then construct a same-phase current transformer sample. The current transformer state analysis module is used to calculate the statistical quantity of each sample in the same-phase current transformer sample based on the calculation step of the statistical quantity threshold in step S1 of the current transformer error solving method in the above embodiment, compare the statistical quantity of each sample with the statistical quantity threshold of the initial steady state to obtain a preliminary analysis result of the current transformer state, and then set the solving threshold of the ratio error and the phase error of the current transformer in combination with the accuracy level of the current transformer. The solving model construction module is used to establish an initial multi-objective weighted fitness function model taking the ratio error and the phase error of each current transformer as a solving object and the minimum of the node primary current vector as an optimization target based on the KCL principle. After obtaining the multi-objective weighted fitness function model, the solving threshold of the ratio error and the phase error of the current transformer in step S3 of the current transformer error solving method in the above embodiment is combined to constrain the initial multi-objective weighted fitness function model to obtain an optimized multi-objective weighted fitness function model. The particle swarm optimization module is used to introduce an improved Logistic mapping to initialize the particle swarm after obtaining the optimized multi-objective weighted fitness function model. Based on the fitness of each particle, the particle swarm is cooperatively optimized in combination with a grey wolf algorithm, a particle swarm optimization algorithm, and a Levy flight disturbance mechanism to iteratively update the position of each particle until a preset stop condition is met, and the solving result of the final ratio error and phase error is obtained. In the iteration process, whether each particle is updated is limited in combination with the solving threshold of the ratio error and the phase error of the current transformer in step S3 of the current transformer error solving method in the above embodiment.The double-threshold judgment module is configured to construct a hypothesis testing model with two judgment thresholds based on a double-threshold mechanism, take the preliminary analysis result in step S3 of the current error solving method based on a current transformer as the first discrimination object of the hypothesis testing model, take the solving result of the final ratio difference and phase difference obtained in step S5 of the current error solving method based on a current transformer as the second discrimination object of the hypothesis testing model, and perform comparative analysis on the first discrimination object and the second discrimination object based on the two judgment thresholds of the hypothesis testing model, wherein if the hypothesis testing model has 80% of the results showing that both discrimination objects think that the current transformer at the position has an over-error problem within a specified time length, it is determined that the current transformer at the position indeed has an over-error problem. Embodiment 3

[0064] Based on the same inventive concept, the present application further provides a data processing device based on a current transformer error solving method, comprising a memory and a processor, the memory is used to store a computer program, and the processor is used to execute the computer program to realize the steps of the data processing method of the current error solving method based on a current transformer in the above embodiment. Embodiment 4

[0065] Based on the same inventive concept, the present application further provides a computer readable storage medium, the computer readable storage medium stores a computer program, and the computer program is executed by a processor to realize the steps of the current error solving method in the above embodiment.

[0066] The above-mentioned embodiments are only the preferred embodiments of the present application, and do not limit the present application in any form. Any skilled person in the art can make more possible changes and decorations, or modifications to the technical solution of the present application without departing from the scope of the technical solution of the present application, using the disclosed technical content. Therefore, any equivalent changes made according to the idea of the present application without departing from the content of the technical solution of the present application shall be covered within the protection scope of the present application.

Claims

1. A method for solving the error of a current transformer, characterized in that: Includes the following steps: Step S1: Collect the operating data of current transformers on each measuring line under the same node of the substation within the first preset time period to construct a sample analysis dataset. Then, using principal component analysis (PCA), decompose and extract features from the sample analysis dataset to establish an initial steady state, and finally calculate the values ​​of the current transformers in the sample dataset. Statistical threshold; Step S2: Collect the original waveform data of the current transformers of each measuring line under the same node of the substation within the second preset time period. Based on the sliding window mean filtering algorithm, filter the original waveform data to obtain the filtered original waveform data. Combined with the fast Fourier transform method, calculate the filtered original waveform data to obtain the real-time status judgment of the current transformers of each measuring line, and then construct the in-phase current transformer sample. Step S3: Based on step S1 The steps for calculating the statistical threshold are as follows: calculate the threshold value for each sample in the in-phase current transformer sample. Statistics, and the values ​​of each sample Statistics and initial steady state Statistical thresholds are compared to obtain preliminary analysis results of the current transformer's state. Then, based on the accuracy level of the current transformer, the thresholds for solving the ratio difference and phase difference in the current transformer are set. Step S4: Based on the KCL principle, establish an initial multi-objective weighted fitness function model with the ratio difference and phase difference of each current transformer as the solution object and minimizing the primary current vector sum of the nodes as the optimization objective; after obtaining the initial multi-objective weighted fitness function model, combine the solution thresholds of the ratio difference and phase difference of the current transformers in Step S3 to constrain the initial multi-objective weighted fitness function model and obtain the optimized multi-objective weighted fitness function model; Step S5: After obtaining the optimized multi-objective weighted fitness function model, an improved Logistic mapping is introduced to initialize the particle swarm. Based on the fitness of each particle, the particle swarm is collaboratively optimized by combining the Grey Wolf algorithm, the particle swarm optimization algorithm, and the Lévy flight perturbation mechanism to iteratively update the position of each particle until the preset stopping condition is met, and the final solution results of the ratio difference and phase difference are obtained. During the iteration process, the solution threshold of the current transformer ratio difference and phase difference in step S3 is used to restrict whether each particle is updated. Step S6: Based on the dual-threshold mechanism, construct a hypothesis testing model with two judgment thresholds. Use the preliminary analysis results in Step S3 as the first discrimination object of the hypothesis testing model, and use the final solution results of the ratio difference and phase difference obtained in Step S5 as the second discrimination object of the hypothesis testing model. Based on the two judgment thresholds of the hypothesis testing model, compare and analyze the first discrimination object and the second discrimination object respectively. If, within a specified time, 80% of the results of the hypothesis testing model show that both discrimination objects believe that the current transformer has an out-of-tolerance problem, then it is determined that the current transformer does indeed have an out-of-tolerance problem.

2. The method for solving the error of a current transformer according to claim 1, characterized in that: Step S1 includes the following steps: Step S1.1: Collect the operating data of the in-phase current transformers of each branch under the same node of the substation within the first preset time period to establish a sample analysis dataset in vector form of the current transformers, and establish a kernel matrix composed of the sample analysis dataset based on the kernel function method, wherein the operating data is the amplitude and phase value of the current transformers; Step S1.2: Transform the kernel matrix based on the centering method to eliminate the mean shift in the feature space and obtain the centered kernel matrix; Step S1.3: Perform eigenvalue decomposition on the centered kernel matrix to obtain each eigenvalue and each eigenvector corresponding to each eigenvalue; Step S1.4: After obtaining each eigenvalue and its corresponding eigenvector, sort the eigenvalues ​​in descending order. Based on the calculation principle of cumulative contribution rate, obtain the k largest eigenvalues ​​with a cumulative contribution rate greater than or equal to 85%, and construct a variance contribution diagonal matrix based on the k largest eigenvalues. The principal component direction matrix consists of the eigenvectors corresponding one-to-one with the first k largest eigenvalues. ; Step S1.5: Obtain the principal component direction matrix Then, calculate the principal component projection vectors of each sample in the sample analysis dataset; Step S1.6: After obtaining the principal component projection vectors of each sample, the Q statistic is obtained by calculating the sum of squares of the residual vectors; Step S1.7: After obtaining the set of Q statistics, combine the quantile method with the 99th percentile of the normal samples in the sample analysis dataset as the optimal discrimination threshold, or combine the chi-square distribution method to determine the pre-set reliability level. Statistical threshold.

3. The method for solving the error of a current transformer according to claim 1, characterized in that: Step S2 includes the following steps: Step S2.1: Collect the original waveform data of the current transformers of the same phase of each measuring line under the same node of the substation through a high-precision sensor within the second preset time period. The original waveform data is the secondary current waveform data of the current transformer. Step S2.2: Based on the sliding window mean filtering algorithm, filter the original current waveform data to obtain the filtered waveform data; Step S2.3: Based on the filtered waveform data, combined with the fast Fourier transform, calculate the amplitude and phase values ​​of each harmonic of each current transformer, as well as the total harmonic distortion rate, and then construct the in-phase current transformer sample. Step S2.4: When the total harmonic distortion rate of a current transformer exceeds a preset value one or the amplitude of a specific harmonic exceeds a preset value two, it is determined that the current transformer has harmonic interference.

4. The method for solving the error of a current transformer according to claim 1, characterized in that: Step S3 includes the following steps: Step S3.1: After obtaining the in-phase current transformer sample, combine it with the information from step S1. The steps for calculating the threshold of a statistic, including calculating the sample... Statistic; Step S3.2: Transfer the sample Statistics and steps in S1.6 If the statistical thresholds are compared, Statistical quantity exceeds If the statistical threshold is set, it is determined that there may be a problem with the current transformer in the substation. The fault separation algorithm is then used to locate the current transformer that may have a problem. Step S3.3: After obtaining the preliminary analysis results of the current transformer status, combine the accuracy class of the current transformer to divide the solution thresholds for the ratio difference and phase difference in the current transformer. Specifically, when the current transformer is preliminarily determined to be problem-free in step S3.2 and the accuracy class corresponding to the current transformer is 0.1S, the solution threshold for the ratio difference of the current transformer is set to [-0.1%, 0.1%], and the solution threshold for the phase difference is set to [-5', 5']. Step S3.4: When it is initially determined in step S3.2 that there is a problem with the current transformer, expand the coverage of the ratio difference solution threshold and phase difference solution threshold of the current transformer according to the accuracy class and procedure error limit corresponding to the current transformer.

5. The method for solving the error of a current transformer according to claim 4, characterized in that: Step S4 includes the following steps: Step S4.1: After obtaining the samples of each in-phase current transformer, based on the KCL principle, establish a multi-objective weighted fitness function model with the ratio difference and phase difference of each current transformer as the solution object and minimizing the primary current vector sum of the nodes as the optimization objective. This transforms the current transformer error solution problem into the problem of minimizing the primary current vector sum in the multi-objective weighted fitness function model. The formula for the initial multi-objective weighted fitness function model is: ， in, To evaluate the ratio error of CT on the i-th line in the population, To evaluate the phase error of the CT on the i-th line in the population, Let be the secondary current phasor of the i-th sample; Step S4.2: Based on the initial multi-objective weighted fitness function model, and combined with the threshold values ​​for the current transformer ratio difference and phase difference obtained in Step S3.3, constraints are applied to the initial multi-objective weighted fitness function model to obtain the optimized multi-objective weighted fitness function model. The formula for the optimized multi-objective weighted fitness function model is as follows: ， ， in, To evaluate the ratio error of CT on the i-th line in the population, To evaluate the phase error of the CT on the i-th line in the population, The accuracy level of CT scans. This is the rated secondary current value of the CT. Let be the secondary current phasor of the i-th sample.

6. The method for solving the error of a current transformer according to claim 1, characterized in that: Step S5 includes the following steps: Step S5.1: After obtaining the optimized multi-objective weighted fitness function model, the positions of the initial particles are randomly generated in the search space, an improved Logistic mapping is introduced to diffuse the initial particle population, and the optimal position Pbest and the global optimal position Gbest for each particle are set. Step S5.2: Based on the optimized multi-objective weighted fitness function model, calculate the fitness of each initial particle, take the top three particles in fitness as α, β and δ particles, and the remaining particles as ω particles, and calculate the wolf position based on the positions of α, β and δ particles. Step S5.3: Based on the gray wolf algorithm, calculate the updated positions of α, β and δ particles, and based on the updated positions of α, β and δ particles, calculate the updated position of the wolf; update the position of ω particle based on particle swarm optimization algorithm, and combine the Lévy flight perturbation mechanism to apply noise interference to ω particle to enhance the global search capability of ω particle; Step S5.4: If the updated position of each particle or wolf is within the threshold range for solving the ratio difference and phase difference of the current transformer, and the fitness of each particle after the position update is better than the best fitness corresponding to each particle, then update the position of each particle to the current position and record the current fitness of each particle as the best fitness. If the best fitness of a certain particle is better than the global best fitness, then set the current position of the particle as the best position Gbest. Step S5.5: Based on the current fitness of each particle, reset the top three particles in fitness ranking as α, β, and δ particles, and repeat steps S5.2-S5.4 to iteratively solve the ratio difference and phase difference of each current transformer until a preset stopping condition is met. Then stop the iterative solution of the particle swarm and output the global best position Gbest to obtain the optimal solution of the error value of the current in-phase current transformer. The preset stopping condition is: when the rate of decrease of the optimized multi-objective weighted fitness function model is detected to be <10 for three consecutive times. -6 Or the number of iterations of the particle swarm reaches 200.

7. The method for solving the error of a current transformer according to claim 1, characterized in that: Step S6 includes the following steps: Step S6.1: Based on the dual threshold mechanism, construct a hypothesis testing model with two judgment thresholds, and use the preliminary analysis results in step S3 as the first discriminant of the hypothesis testing model, and use the final solution results of the ratio difference and phase difference obtained in step S5 as the second discriminant of the hypothesis testing model. Step S6.2: Set one of the judgment thresholds of the hypothesis testing model as follows: if a sample in the in-phase current transformer sample exceeds the core space Mahalanobis distance threshold, the hypothesis testing model triggers an early warning and locates the problematic current transformer through the fault separation algorithm, and then reports that the current transformer corresponding to the data has an abnormal situation. Step S6.3: Set another judgment threshold for the hypothesis testing model as follows: when the error of the current transformer output by the optimized multi-objective weighted fitness function model exceeds the error threshold of the corresponding current transformer level in the procedure, the hypothesis testing model triggers an early warning and reports that there is an abnormality in the current transformer corresponding to the data. Step S6.4: Based on the two judgment thresholds of the hypothesis testing model, perform comparative analysis on the first and second discriminant objects respectively; Step S6.5: If, within the specified time period, 80% of the results of the hypothesis testing model show that both judgment objects believe that the current transformer at that location has an out-of-tolerance problem, then it is determined that the current transformer at that location has an out-of-tolerance problem.

8. A current transformer error solving system, characterized in that, include: The system comprises a statistical threshold calculation module, a current transformer sample construction module, a current transformer state analysis module, a solution model construction module, a particle swarm optimization module, and a dual threshold judgment module. The statistical threshold calculation module collects operating data from current transformers on each measuring line under the same node of the substation within a first preset time period to construct a sample analysis dataset. It then uses principal component analysis (PCA) to decompose and extract features from the sample analysis dataset, establishing an initial steady state and calculating the statistical threshold in the sample dataset. The current transformer sample construction module collects raw waveform data from current transformers on each measuring line under the same node of the substation within a second preset time period. Based on a sliding window mean filtering algorithm, it filters the raw waveform data to obtain filtered raw waveform data. In addition, by combining the fast Fourier transform method, the filtered original waveform data is calculated to obtain the real-time status judgment of the current transformers of each measurement line, and then a sample of in-phase current transformers is constructed. The current transformer state analysis module is used to calculate the statistical threshold in step S1 of the current transformer error solution method according to any one of claims 1 to 7, calculate the statistical value of each sample in the in-phase current transformer sample, and compare the statistical value of each sample with the initial steady-state statistical threshold to obtain the preliminary analysis results of the current transformer state. Then, combined with the accuracy level of the current transformer, the solution thresholds for the ratio difference and phase difference in the current transformer are set. The solution model construction module is used to establish a solution model based on the KCL principle, with the ratio difference and phase difference of each current transformer as the solution object, and the node primary current vector... The initial multi-objective weighted fitness function model is optimized by minimizing the sum of quantities. After obtaining the multi-objective weighted fitness function model, the initial multi-objective weighted fitness function model is constrained by the solution thresholds for the current transformer ratio difference and phase difference in step S3 of the current transformer error solution method according to any one of claims 1 to 7, thereby obtaining an optimized multi-objective weighted fitness function model. The particle swarm optimization module is used to introduce an improved Logistic mapping to initialize the particle swarm after obtaining the optimized multi-objective weighted fitness function model. Based on the fitness of each particle, the gray wolf algorithm is combined with the optimization module. The particle swarm optimization algorithm and the Lévy flight perturbation mechanism are used to collaboratively optimize the particle swarm, iteratively updating the position of each particle until a preset stopping condition is met, thus obtaining the final solution results for the ratio difference and phase difference. During the iteration process, the solution thresholds for the ratio difference and phase difference of the current transformer in step S3 of any one of claims 1 to 7 are used to restrict whether each particle is updated. The dual-threshold judgment module is used to construct a hypothesis testing model with two judgment thresholds based on the dual-threshold mechanism, using the current transformer error solution method according to any one of claims 1 to 7. The preliminary analysis results in step S3 are used as the first discrimination object of the hypothesis testing model. The final ratio difference and phase difference calculation results obtained in step S5 of the current transformer error solution method according to any one of claims 1 to 7 are used as the second discrimination object of the hypothesis testing model. Based on the two judgment thresholds of the hypothesis testing model, the first discrimination object and the second discrimination object are compared and analyzed respectively. If, within a specified time, 80% of the results of the hypothesis testing model show that both discrimination objects believe that the current transformer has an out-of-tolerance problem, then it is determined that the current transformer does indeed have an out-of-tolerance problem.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of a current transformer error solving method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed, it implements the steps of the current transformer error solving method as described in any one of claims 1 to 7.

Images