Electromagnetic current transformer error prediction method based on weighted extreme learning machine
Through the electromagnetic current transformer error prediction method combined with the weighted extreme learning machine and the gray wolf optimization algorithm, the safety, portability and model training speed problems in online monitoring of current transformer errors are solved, and efficient and accurate error monitoring is achieved. and prediction.
Patent Information
- Application Number
- CN202411065894.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-05
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-08-05
AI Technical Summary
The existing online monitoring technology for current transformer errors has problems such as safety, insufficient portability, insufficient data characteristics, slow model training speed and serious sample imbalance, making it difficult to effectively monitor electromagnetic current transformers in large-scale application scenarios. Error.
The electromagnetic current transformer error prediction method based on weighted limit learning machine is adopted, combined with gray wolf optimization algorithm and variational modal decomposition, time frequency domain features are extracted, and feature matrix after dimensionality reduction is constructed, and limit learning is constructed through cost-sensitive methods. machine model, quickly train and predict the error of current transformer.
It realizes fast and accurate online monitoring of current transformer errors, improves monitoring efficiency and accuracy, overcomes sample imbalance problem, and has strong engineering universality and market promotion conditions.
Smart Images

Figure CN118965294A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of current transformers, and in particular relates to an electromagnetic current transformer error prediction method based on a weighted extreme learning machine. Background Art
[0002] Electromagnetic current transformers play a vital role in power systems for measuring and monitoring large currents. However, after long-term operation, some electromagnetic current transformers may exceed the error limit, so accurate online monitoring and error evaluation are required.
[0003] my country already has relatively mature and systematic equipment manufacturing technology and error compensation methods for current transformers. At the same time, a relatively complete calibration method has been set up based on operating experience to ensure the operating stability of current transformers. Currently, the metering characteristics of current transformers are mainly evaluated through periodic calibration methods. However, actual operating data show that there is a deviation between the offline measurement error of the current transformer and the actual operating error. The periodic offline calibration method cannot fully grasp the metering performance of the current transformer under actual operating conditions. The current transformer is installed in the substation. The electromagnetic environment of the substation in the operating condition is complex, and there is a possibility of exceeding the allowable error in long-term operation. In order to accurately and timely discover the metering status of the electromagnetic current transformer in operation and obtain accurate error conditions, the online monitoring technology of electromagnetic current transformer errors has become a hot spot and difficulty in the current technical development of electric energy metering devices.
[0004] In the exploration and development of current transformer verification methods, it has experienced a development from "static verification" to "live verification" and then to "pre-calibration evaluation". At present, the most common method in the power industry is to conduct offline "static verification" of transformers under power outage conditions to grasp the operating errors of transformers in operation. In order to reduce the impact of power outages on the operation of the power grid due to the "static verification" method, scholars have proposed a transformer "live verification" method, which uses an online special standard to calibrate the errors of operating transformers under power. In order to overcome the shortcomings of the "live verification" method in terms of portability and safety, experts and scholars have tried to establish physical or mathematical models in the direction of "pre-calibration evaluation" in recent years to achieve the evaluation of the operating errors of all transformers in the network.
[0005] In the research of electromagnetic current transformer error evaluation technology, the focus is on offline analysis under simulated actual operating conditions, and analysis of the influence of multi-dimensional factors on electromagnetic current transformer errors. By simulating actual operating conditions, the error performance of electromagnetic current transformers under different conditions can be more accurately evaluated, and the influence of various factors on the error can be analyzed.
[0006] In addition, in terms of online evaluation of electromagnetic current transformer errors, research is mainly focused on electronic current transformers. This method uses traditional current transformers as reference standards to compare the relative changes in the errors of electronic current transformers with the same voltage level, phase sequence, and transformation ratio. By online monitoring of the errors of electromagnetic current transformers, their performance can be monitored in real time and potential problems can be discovered in a timely manner.
[0007] In summary, the online monitoring technology of current transformer error is of great significance to ensure the accuracy and reliability of electric energy metering. By simulating the actual operating conditions and evaluating the error of the electronic current transformer online, the monitoring and evaluation capabilities of the current transformer performance can be improved, thereby effectively preventing and solving the current transformer error problem.
[0008] Disadvantages of existing technology:
[0009] (1) At present, offline evaluation and online evaluation often require error comparison with traditional electromagnetic standard transformers. In large-scale application scenarios of transformers, both safety and portability are not as convenient as the pre-evaluation method. The pre-evaluation method requires the acquisition of some transformer states in advance, and this part of the "pre-calibrated" transformer state has a great impact on the evaluation results.
[0010] (2) Existing data-driven transformer evaluation models rarely have input features that combine time-frequency domain features. Even if these types of features are available, they are often extracted through convolutional neural networks or time series network models, thus losing the original feature interpretability.
[0011] (3) The existing transformer error evaluation model does not take into account the serious data imbalance of positive and negative samples during the model training process, that is, there are more samples under the normal state of the transformer and fewer samples with abnormal tolerance.
[0012] (4) Most of the machine learning models or deep learning models used in existing evaluation models are trained based on the principle of back propagation, which results in slow training speed.
[0013] Therefore, how to overcome the shortcomings of the existing technology is an urgent problem to be solved in the field of current transformer technology. Summary of the invention
[0014] The purpose of the present invention is to solve the deficiencies in the prior art and to provide an electromagnetic current transformer error prediction method based on a weighted extreme learning machine, specifically an electromagnetic current transformer online monitoring error evaluation method based on variational mode decomposition and a weighted extreme learning machine.
[0015] To achieve the above purpose, the technical solution adopted by the present invention is as follows:
[0016] The electromagnetic current transformer error prediction method based on weighted extreme learning machine includes the following steps:
[0017] Step (1), collecting the secondary side current signal amplitude data of the three-phase current transformer every a minute under normal operating conditions to obtain a three-phase current amplitude matrix; and collecting the angle difference and ratio difference of the three-phase current transformer at the same time;
[0018] Step (2), for the three-phase current amplitude matrix obtained in step (1), remove abnormal values in the matrix data;
[0019] Step (3), standardizing the amplitude data in the three-phase current matrix processed in step (2) to obtain a standardized three-phase current amplitude matrix;
[0020] Step (4), using the Grey Wolf Optimization Algorithm to optimize the parameters of the variational modal decomposition algorithm, and using the optimized variational modal decomposition algorithm to perform variational modal decomposition on the standardized three-phase current amplitude matrix obtained in step (3);
[0021] Step (5), extracting time domain characteristic parameters and frequency domain characteristic parameters from the modal components decomposed in step (4), and concatenating all extracted parameters to obtain a reduced-dimensional three-phase current characteristic matrix;
[0022] Step (6), using the reduced dimension three-phase current feature matrix obtained in step (5) as input, and using the angle difference and ratio difference as output, to train the extreme learning machine model to obtain a current transformer error detection model;
[0023] Step (7), using the secondary side current signal amplitude data of the three-phase current transformer to be detected to obtain the three-phase current amplitude matrix of the three-phase current transformer to be detected; then through steps (2) to (5), the corresponding three-phase current characteristic matrix is obtained, and then input into the current transformer error detection model obtained in step (6) to obtain the predicted angle difference and ratio difference of the three-phase current transformer to be detected;
[0024] Step (8), compare the prediction result of step (7) with the error threshold. If the predicted angle difference and ratio difference are both smaller than the threshold, it indicates that the current transformer is in a normal operating state at this time; otherwise, it is considered that the current transformer may be in an abnormal operating state at this time.
[0025] Furthermore, in step (1), a=15; 900 data points need to be retained for each signal after being eliminated in step (2).
[0026] Further, the specific method of step (2) is:
[0027] According to the maximum and minimum values of historical normal signal changes, abnormal values that exceed the range are removed, and non-numeric values are also eliminated.
[0028] Furthermore, the specific method of step (3) is: standardizing the data processed by step (2) using a column normalization method.
[0029] Furthermore, in step (4), the specific method of using the Grey Wolf Optimization Algorithm to optimize the parameters of the variational mode decomposition algorithm is:
[0030] Step 4.1: Initialize the parameters of the Gray Wolf Algorithm: Initialize the basic parameters of the Gray Wolf Algorithm, such as population size, decreasing coefficient, random vector, and number of iterations.
[0031] Step (4.2), gray wolf position initialization: randomly initialize the decomposition layer number k and penalty factor alpha of the variational mode decomposition algorithm as the initial position of the gray wolf algorithm;
[0032] Step (4.3), calculate the fitness of the gray wolf individuals: the fitness function takes the mean of the permutation entropy, the optimization goal is to minimize the mean of the permutation entropy, save the three wolves with the best fitness, and define the first three individuals with the smallest fitness values as α wolf, β wolf and δ wolf respectively, and their positions are defined as and The rest are uniformly defined as ωwolf;
[0033] Step (4.4), capture the prey, and capture the prey by the following formula
[0034]
[0035] In the formula, is the position of the individual after update; is the current position of the prey; t is the number of iterations of the current algorithm; is the distance between the current individual and the prey; The current location of the individual; is a synergy coefficient vector; are another synergy coefficient vector;
[0036] Step (4.5), iteratively update the gray wolf individual, and update the latest position of the gray wolf according to the following four update formulas:
[0037]
[0038] In the formula, and is the direction and step length of the movement of ω wolf toward α wolf, β wolf and δ wolf; and are the positions of the alpha wolf, beta wolf, and delta wolf; and is the distance between the current individual and α wolf, β wolf and δ wolf; recalculate the fitness and update α wolf, β wolf and δ wolf;
[0039] In step (4.5), when the number of iterations is reached, the optimal decomposition layer k and penalty factor alpha of the VMD algorithm are output.
[0040] Furthermore, the optimal decomposition level k is 2.
[0041] Further, in step (5), the time domain feature parameters include maximum value, minimum value, first-order moment, second-order origin moment and kurtosis factor; the frequency domain features include centroid frequency, mean square frequency, frequency variance and multi-scale permutation entropy;
[0042] in,
[0043] The maximum value is calculated as: max =max(x i ), where x i is the current data of the i-th sampling point;
[0044] The minimum value is calculated as: min =min(x i );
[0045] The calculation formula for the first-order moment is: Where N is the number of samples;
[0046] The calculation formula of the second-order origin moment is:
[0047] The calculation formula of the kurtosis factor is: in, is the sampling mean, σ is the sampling standard deviation;
[0048] The calculation formula of the center of gravity frequency is: P(f) is the spectrum value at frequency f;
[0049] The formula for calculating the mean square frequency is:
[0050] The frequency variance is calculated as:
[0051] The calculation formula of multi-scale permutation entropy is:
[0052] For m different numbers, there will be m! possible order patterns π; for a permutation of the number π, f(π) represents its frequency in the time series; the relative frequency is p(π) = f(π) / (N / τ-m+1).
[0053] Further, in step (5), the specific method of splicing is:
[0054] The reduced dimension three-phase current feature matrix is A-phase sample, B-phase sample, and C-phase sample from top to bottom;
[0055] The A phase sample is composed of 24-dimensional features of phases A, B, and C in sequence;
[0056] The B-phase samples are composed of 24-dimensional features of the three phases B, A, and C in sequence;
[0057] The C phase samples are formed by sequentially concatenating the 24-dimensional features of the three phases C, A, and B;
[0058] Thus, a 72*3-dimensional three-phase characteristic matrix is constructed, that is, a reduced-dimensional three-phase current characteristic matrix.
[0059] Furthermore, in the extreme learning machine model, the cost-sensitive weight matrix is embedded in the extreme learning machine objective function, and each weight factor of the newly added weight matrix in the extreme learning machine objective function is constructed in a cost-sensitive manner.
[0060] Furthermore, in step (8), if the predicted angle difference and ratio difference both exceed the corresponding threshold values, the working state of the current transformer is determined and the operation, maintenance and repair work of the relevant staff is guided.
[0061] Furthermore, the threshold range of the angle difference is ±10 points, and the threshold range of the ratio difference is ±0.2%; if the predicted angle difference or ratio difference exceeds the corresponding threshold, it is determined that the current transformer is in an out-of-tolerance state, and the staff needs to be reminded for maintenance and repair.
[0062] During the training in step (6), the data division is set in a 5-fold cross-check manner, and 4 sets of data are trained and 1 set of data is tested each time, and the final accuracy is the average accuracy of 5 times as the final result.
[0063] In step (4.3) of the present invention, the fitness function takes the mean of the permutation entropy, and the optimization goal is to minimize the mean of the permutation entropy. The smaller the mean, the more orderly and regular the data arrangement.
[0064] Since the optimal decomposition layer number k obtained by variational mode decomposition is 2, in step (5) of the present invention,
[0065] For each modal component, 5 time domain features are extracted: maximum value, minimum value, first-order moment, second-order origin moment and kurtosis factor; plus 7 frequency domain features: centroid frequency, mean square frequency, frequency variance and multi-scale permutation entropy (where the multi-scale permutation entropy m is 4, and the permutation number features on 4 scales are extracted). That is, 12 features are extracted for each modal component, and a total of 24 features for the two modal components.
[0066] Among them, for m different numbers, there will be m! possible sequence patterns π, also called permutations; for the permutation of the number π, f(π) represents its frequency in the time series; then the relative frequency is
[0067] p(π)=f(π) / (N / τ-m+1). In the present invention, m is 4.
[0068] The present invention discloses a method for predicting and evaluating the measurement error of an electromagnetic current transformer based on variational mode decomposition and extreme learning machine, and solves the following technical problems:
[0069] The present invention adopts the variational mode decomposition method to reduce the dimension of the signal and simplifies the dimension of the data set. Variational mode decomposition has a good denoising effect, can improve the smoothness of the displacement data, extract the detailed features of the data, and has a stronger signal decomposition ability than other decomposition methods. The core idea of this technology is to decompose the signal and extract features to convert high-dimensional signal data into a low-dimensional representation, thereby reducing redundant information of the data and improving the efficiency and accuracy of signal processing. The present invention avoids the problem that the sampled current transformer monitoring data belongs to a high-dimensional data set, and directly bringing it into the reliability assessment and prediction model will increase the difficulty of calculation and prolong the model running time.
[0070] The present invention selects a weighted extreme learning machine as the regression model. The extreme learning machine only needs to randomly initialize weights and biases, and does not require back propagation optimization, which effectively solves the problem of slow training speed in previous methods. In view of the serious imbalance of transformer monitoring data, the present invention solves it by embedding a cost-sensitive weight matrix in the objective function of the extreme learning machine. The details are as follows:
[0071] Given the reduced-dimensional input training set S = {(x i ,y i ),i=1,2,...,N,y i ∈{+,-}}, first traverse all samples in the training set to count the number of positive and negative samples; construct the diagonal weight factor ω of the weight matrix W according to the sample category ii , the value is the reciprocal of the total number of samples contained in the category corresponding to the sample, and the rest of the positions of W are 0;
[0072] Substitute the weight matrix into the solution formula of the output weight matrix β:
[0073]
[0074] Among them, H is the output matrix of the hidden layer, C is the regularization coefficient, N is the number of samples, L is the number of nodes in the hidden layer, and Y is the actual output value matrix;
[0075] For an unknown transformer monitoring data sample x, the transformer specific error output by the weighted extreme learning machine is substituted into the function f(x)=Hβ to be calculated.
[0076] Compared with the prior art, the present invention proposes to formulate an online evaluation process for the metering performance of current transformers based on the characteristics of the output current data of the secondary monitoring of the transformer, and studies a dimensionality reduction method for current transformer monitoring data with a high efficiency calculation rate. The error gradient of the current transformer can be accurately discovered through the model, and it has strong engineering universality and broad market promotion conditions.
[0077] Compared with the prior art, the present invention has the following beneficial effects:
[0078] 1. The present invention proposes a dimensionality reduction method for transformer secondary side monitoring data based on Grey Wolf Optimized Variational Mode Decomposition. By extracting the time-frequency domain features of each modal signal, the original 900-dimensional detection data is reduced to 72 dimensions, thereby achieving effective dimensionality reduction of the current transformer monitoring data. The GWO algorithm adopts a flexible parameter adjustment and dynamic update strategy, which can quickly converge to a better solution and has a strong global search capability; its algorithm flow is clear, the parameter adjustment is relatively intuitive, and it is easy to understand and apply. The three-phase current feature matrix is constructed by constructing the reduced-dimensional sample data based on the original monitoring data, and splicing the time-frequency domain features. The establishment of this feature matrix takes into account the differences in currents of different phases, provides a more comprehensive and accurate input for the ELM model of current angle difference ratio prediction, and more comprehensively evaluates the performance of single-phase current transformers.
[0079] 2. The present invention selects a weighted extreme learning machine as a transformer error prediction model based on three-phase current data. The weighted extreme learning machine has the characteristics of fast training speed of the traditional extreme learning machine. It only needs to randomly assign bias and input weights, calculate the output weight matrix by least squares, and construct each weight factor of the newly added weight matrix in the objective function in a cost-sensitive manner. The mean absolute error MAE (Mean Absolute Error) and mean square error MSE (Mean Square Error) are used as prediction evaluation indicators to train the extreme learning machine.
[0080] 3. Compare the weighted extreme learning machine with other prediction algorithms in the field of machine learning, including extreme learning machine (ELM), support vector machine (SVM), random forest, and extreme gradient boosting (XGBoost). The results of the comparative experiment are shown in Table 1. It can be seen that the prediction effect of the model of the present invention is the best, and it is greatly improved compared with the ELM model. The reason may be that the addition of the weight matrix better solves the problem of sample imbalance in the current transformer data set.
[0081] Table 1 Comparison of algorithm performance evaluation indicators
[0082] BRIEF DESCRIPTION OF THE DRAWINGS
[0083] Figure 1 This is the transformer performance evaluation process proposed by the present invention.
[0084] Figure 2 Obtaining the optimal parameters of VMD based on the Grey Wolf optimization algorithm.
[0085] Figure 3 It is a matrix construction combining the three-phase current characteristics. DETAILED DESCRIPTION
[0086] The present invention is further described in detail below in conjunction with embodiments.
[0087] Those skilled in the art will appreciate that the following examples are only used to illustrate the present invention and should not be considered to limit the scope of the present invention. If no specific techniques or conditions are specified in the examples, the techniques or conditions described in the literature in the art or the product specifications are used. If the manufacturer of the materials or equipment used is not specified, they are all conventional products that can be purchased.
[0088] The electromagnetic current transformer error prediction method based on weighted extreme learning machine includes the following steps:
[0089] Step (1), collecting the secondary side current signal amplitude data of the three-phase current transformer every a minute under normal operating conditions to obtain a three-phase current amplitude matrix; and collecting the angle difference and ratio difference of the three-phase current transformer at the same time;
[0090] Step (2), for the three-phase current amplitude matrix obtained in step (1), remove abnormal values in the matrix data;
[0091] Step (3), standardizing the amplitude data in the three-phase current matrix processed in step (2) to obtain a standardized three-phase current amplitude matrix;
[0092] Step (4), using the Grey Wolf Optimization Algorithm to optimize the parameters of the variational modal decomposition algorithm, and using the optimized variational modal decomposition algorithm to perform variational modal decomposition on the standardized three-phase current amplitude matrix obtained in step (3);
[0093] Step (5), extracting time domain characteristic parameters and frequency domain characteristic parameters from the modal components decomposed in step (4), and concatenating all extracted parameters to obtain a reduced-dimensional three-phase current characteristic matrix;
[0094] Step (6), using the reduced dimension three-phase current feature matrix obtained in step (5) as input, and using the angle difference and ratio difference as output, to train the extreme learning machine model to obtain a current transformer error detection model;
[0095] Step (7), using the secondary side current signal amplitude data of the three-phase current transformer to be detected to obtain the three-phase current amplitude matrix of the three-phase current transformer to be detected; then through steps (2) to (5), the corresponding three-phase current characteristic matrix is obtained, and then input into the current transformer error detection model obtained in step (6) to obtain the predicted angle difference and ratio difference of the three-phase current transformer to be detected;
[0096] Step (8), compare the prediction result of step (7) with the error threshold. If the predicted angle difference and ratio difference are both smaller than the threshold, it indicates that the current transformer is in a normal operating state at this time; otherwise, it is considered that the current transformer may be in an abnormal operating state at this time.
[0097] In step (1), a=15;
[0098] The specific method of step (2) is: according to the maximum and minimum values of the historical normal signal changes, the abnormal values that exceed the range are removed, and non-values are also removed;
[0099] After each signal is eliminated in step (2), 900 data points need to be retained.
[0100] The specific method of step (3) is: standardize the data processed by step (2) using the column normalization method.
[0101] In step (4), the specific method of using the gray wolf optimization algorithm to optimize the parameters of the variational mode decomposition algorithm is:
[0102] Step 4.1: Initialize the parameters of the Gray Wolf Algorithm: Initialize the basic parameters of the Gray Wolf Algorithm, such as population size, decreasing coefficient, random vector, and number of iterations.
[0103] Step (4.2), gray wolf position initialization: randomly initialize the decomposition layer number k and penalty factor alpha of the variational mode decomposition algorithm as the initial position of the gray wolf algorithm;
[0104] Step (4.3), calculate the fitness of the gray wolf individuals: the fitness function takes the mean of the permutation entropy, the optimization goal is to minimize the mean of the permutation entropy, save the three wolves with the best fitness, and define the first three individuals with the smallest fitness values as α wolf, β wolf and δ wolf respectively, and their positions are defined as and The rest are uniformly defined as ωwolf;
[0105] Step (4.4), capture the prey, and capture the prey by the following formula
[0106]
[0107] In the formula, is the position of the individual after update; is the current position of the prey; t is the number of iterations of the current algorithm; is the distance between the current individual and the prey; The current location of the individual; is a synergy coefficient vector; are another synergy coefficient vector;
[0108] Step (4.5), iteratively update the gray wolf individual, and update the latest position of the gray wolf according to the following four update formulas:
[0109]
[0110] In the formula, and is the direction and step length of the movement of ω wolf toward α wolf, β wolf and δ wolf; and are the positions of alpha wolf, beta wolf, and delta wolf; and is the distance between the current individual and α wolf, β wolf and δ wolf; recalculate the fitness and update α wolf, β wolf and δ wolf;
[0111] In step (4.5), when the number of iterations is reached, the optimal decomposition layer k and penalty factor alpha of the VMD algorithm are output.
[0112] The optimal decomposition level k is 2.
[0113] In step (5), the time domain feature parameters include maximum value, minimum value, first-order moment, second-order origin moment and kurtosis factor; the frequency domain features include centroid frequency, mean square frequency, frequency variance and multi-scale permutation entropy;
[0114] in,
[0115] The maximum value is calculated as: max =max(x i ), where x i is the current data of the i-th sampling point;
[0116] The minimum value is calculated as: min =min(x i );
[0117] The calculation formula for the first-order moment is: Where N is the number of samples;
[0118] The calculation formula of the second-order origin moment is:
[0119] The calculation formula of the kurtosis factor is: in, is the sampling mean, σ is the sampling standard deviation;
[0120] The calculation formula of the center of gravity frequency is: P(f) is the spectrum value at frequency f;
[0121] The formula for calculating the mean square frequency is:
[0122] The frequency variance is calculated as:
[0123] The calculation formula of multi-scale permutation entropy is:
[0124] For m different numbers, there will be m! possible order patterns π; for a permutation of the number π, f(π) represents its frequency in the time series; the relative frequency is p(π) = f(π) / (N / τ-m+1).
[0125] In step (5), the specific method of splicing is:
[0126] The reduced dimension three-phase current feature matrix is A-phase sample, B-phase sample, and C-phase sample from top to bottom;
[0127] The A phase sample is composed of 24-dimensional features of phases A, B, and C in sequence;
[0128] The B-phase samples are composed of 24-dimensional features of the three phases B, A, and C in sequence;
[0129] The C phase samples are formed by sequentially concatenating the 24-dimensional features of the three phases C, A, and B;
[0130] Thus, a 72*3-dimensional three-phase characteristic matrix is constructed, that is, a reduced-dimensional three-phase current characteristic matrix.
[0131] In the extreme learning machine model, the cost-sensitive weight matrix is embedded in the extreme learning machine objective function, and each weight factor of the newly added weight matrix in the extreme learning machine objective function is constructed in a cost-sensitive way.
[0132] In step (8), if the predicted angle difference and ratio difference both exceed the corresponding threshold values, the working state of the current transformer is determined and the operation, maintenance and repair work of the relevant staff is guided.
[0133] The threshold range of the angle difference is ±10 points, and the threshold range of the ratio difference is ±0.2%; if the predicted angle difference or ratio difference exceeds the corresponding threshold, it is determined that the current transformer is in an out-of-tolerance state, and the staff needs to be reminded for maintenance and repair.
[0134] Application Examples
[0135] The electromagnetic current transformer error prediction method based on weighted extreme learning machine includes the following steps:
[0136] (1) The performance evaluation process of the transformer of the present invention is as follows Figure 1 shown.
[0137] ① Collect the current signal amplitude data of the secondary side of the three-phase current transformer every 15 minutes under normal operating conditions.
[0138] ② Remove abnormal values that exceed the range according to the maximum and minimum values of historical normal signal changes, and also remove non-values;
[0139] After each signal is eliminated, 900 data points need to be retained.
[0140] ③ The three-phase current amplitude data matrix is standardized using the column normalization method.
[0141] ④ The Grey Wolf Optimization Algorithm is used to optimize the parameters of the variational mode decomposition algorithm to obtain the best decomposition effect, and two inherent mode components are obtained;
[0142] The obtained standardized three-phase current amplitude matrix is subjected to variational mode decomposition using the optimized variational mode decomposition algorithm;
[0143] ⑤ Calculate the time domain characteristic parameters and frequency domain characteristic parameters extracted from the modal components to obtain a 24-dimensional characteristic vector for each phase data, and concatenate all the extracted parameters to obtain a reduced-dimensional three-phase current characteristic matrix;
[0144] ⑥ Taking the reduced-dimensional three-phase current feature matrix as input and the angle difference and ratio difference as output, the extreme learning machine model is trained to obtain the current transformer error detection model;
[0145] ⑦ Using the secondary side current signal amplitude data of the three-phase current transformer to be detected, the three-phase current amplitude matrix of the three-phase current transformer to be detected is obtained; then the corresponding three-phase current characteristic matrix is obtained through the above steps, and then input into the obtained current transformer error detection model to obtain the predicted angle difference and ratio difference of the three-phase current transformer to be detected;
[0146] ⑧ Compare the angle difference and ratio difference results predicted by the model with the error threshold. If the predicted angle difference and ratio difference do not exceed the threshold, it indicates that the current transformer is in normal operation at this time; if it is greater than the threshold, it indicates that the three-phase current transformer is likely to be in abnormal operation at this time.
[0147] If the predicted angle difference or ratio difference is greater than the threshold, the working status and fault type of the current transformer are determined, and the operation, maintenance and repair work of relevant personnel are guided.
[0148] For the above process, the specific method of process ④ is as follows: a dimension reduction method for current transformer monitoring data based on VMD (variational mode decomposition).
[0149] In the Gray Wolf Optimization Algorithm, the fitness value reflects the degree of excellence of each individual in the current solution space, and the smaller or larger the value is according to the optimization target, the higher the excellence of the individual solution. In the present invention, the selection of the fitness function of the Gray Wolf Algorithm is based on the mean of the permutation entropy, which is regarded as an objective evaluation criterion for the target problem. When the permutation entropy is used as the target of the optimization algorithm, it is usually the smaller the better. Permutation entropy is an indicator used to measure the degree of disorder or randomness of a set of data. A smaller permutation entropy indicates that the arrangement between the data is more orderly and regular, while a larger permutation entropy indicates that the arrangement between the data is more disordered and random. What is optimized here is to minimize the permutation entropy mean, that is, to arrange the data in a more orderly and regular manner.
[0150] The process ④ includes the following steps
[0151] Step 4.1: Initialize the parameters of the Gray Wolf Algorithm: Initialize the basic parameters of the Gray Wolf Algorithm, such as population size, decreasing coefficient, random vector, and number of iterations.
[0152] Step (4.2), gray wolf position initialization: randomly initialize the decomposition layer number k and penalty factor alpha of the variational mode decomposition algorithm as the initial position of the gray wolf algorithm;
[0153] Step (4.3), calculate the fitness of the gray wolf individuals: the fitness function takes the mean of the permutation entropy, the optimization goal is to minimize the mean of the permutation entropy, save the three wolves with the best fitness, and define the first three individuals with the smallest fitness values as α wolf, β wolf and δ wolf respectively, and their positions are defined as and The rest are uniformly defined as ωwolf;
[0154] Step (4.4), capture the prey, and capture the prey by the following formula
[0155]
[0156] In the formula, is the position of the individual after update; is the current position of the prey; t is the number of iterations of the current algorithm; is the distance between the current individual and the prey; The current location of the individual; is a synergy coefficient vector; are another synergy coefficient vector;
[0157] Step (4.5), iteratively update the gray wolf individual, and update the latest position of the gray wolf according to the following four update formulas:
[0158]
[0159] In the formula, and is the direction and step length of the movement of ω wolf toward α wolf, β wolf and δ wolf; and are the positions of alpha wolf, beta wolf, and delta wolf; and is the distance between the current individual and α wolf, β wolf and δ wolf; recalculate the fitness and update α wolf, β wolf and δ wolf;
[0160] In step (4.5), when the number of iterations is reached, the optimal decomposition layer k and penalty factor alpha of the VMD algorithm are output.
[0161] The 900*3-dimensional three-phase current amplitude matrix data of the current transformer every 15 minutes is decomposed into two layers of sub-modal signals by the variational modal algorithm of GWO (Grey Wolf Optimization), and the time domain (maximum value x max =max(x i ), minimum value x min =min(x i ), first-order moment Second-order origin moment Kurtosis factor and frequency domain characteristics (center frequency Mean square frequency Frequency variance Multi-scale permutation entropy) was used to obtain 24 feature parameters. Finally, the 900×3 high-dimensional data set was reduced to a 24×3 low-dimensional data set, completing the data dimensionality reduction.
[0162] Before obtaining the frequency domain features, the time domain waveform of the signal is usually converted into a power spectrum through the Fourier transform of the autocorrelation function, which converts the signal from the time domain to the frequency domain to obtain its power spectrum. The formula is as follows: Where x(i) is the sampling value of the current transformer signal. After obtaining the spectrum information, the corresponding statistical characteristics can be calculated according to the statistical indicator formula.
[0163] In the frequency domain, the multi-scale permutation entropy construction method consists of two steps: first, the “coarse-grained” process is applied to the time series. For a given time series {x1, x2, …, x N}, construct a continuous coarse-grained time series by averaging consecutively increasing data points in non-overlapping windows. Multiple coarse-grained time series Each element of Calculate. Where τ represents the scaling factor, 1≤j≤N / τ. The length of each coarse-grained time series is a component of N / τ.
[0164] Next, the permutation entropy of each coarse-grained time series is calculated and then plotted as a function of the scaling factor τ. j The arrangement of the signal samples y j Derive a vector sequence of length m, v m (n) = [y n ,y n+1 ,…,y (n+m-1 )], 1≤n≤N / τ-m+1. Then, v m (n) can be arranged in ascending order: For m different numbers, there will be m! possible order patterns π, also called permutations. For a permutation of π, f(π) represents its frequency in the time series. The relative frequency is then p(π) = f(π) / (N / τ-m+1). The permutation entropy of a time series is defined as The maximum value of H(m) is log(m!), which means that the probability of all permutations is equal. The minimum value of H(m) is zero, which means that the time series is very regular. In other words, the smaller the value of H(m), the more regular the time series is. Here, the permutation entropy m is selected as 4.
[0165] The accuracy of the ratio difference and angle difference of the current transformer is often disturbed by other phase currents, which is due to the mutual correlation and influence between the phase currents in the three-phase system. The present invention proposes a matrix construction of three-phase current characteristics, and constructs sample data after dimensionality reduction based on the original monitoring data. According to the case, the A-phase sample is composed of 24-dimensional time-frequency domain features of the three phases A, B, and C, respectively, and the B-phase sample is composed of 24-dimensional time-frequency domain features of the three phases B, A, and C, respectively. The C-phase sample is composed of 24-dimensional time-frequency domain features of the three phases C, A, and B. This arrangement constructs a 72*3-dimensional three-phase feature matrix. The establishment of this feature matrix takes into account the differences in currents of different phases, provides a more comprehensive and accurate input for the ELM model of current angle difference ratio prediction, and more comprehensively evaluates the performance of single-phase current transformers.
[0166] For the above process, the specific method of process ⑥ is as follows:
[0167] Given the reduced-dimensional input training set S = {(x i ,y i ),i=1,2,...,N,y i∈{+,-}}, first traverse all samples in the training set to count the number of positive and negative samples; construct the diagonal weight factor ω of the weight matrix W according to the sample category ii , the value is the reciprocal of the total number of samples contained in the category corresponding to the sample, and the rest of the positions of W are 0.
[0168] Substitute the weight matrix into the solution formula for the output weight matrix β.
[0169]
[0170] Among them, H is the output matrix of the hidden layer, C is the regularization coefficient, N is the number of samples, L is the number of nodes in the hidden layer, and Y is the actual output value matrix;
[0171] For an unknown transformer monitoring data sample x, the transformer specific error output by the weighted extreme learning machine is substituted into the function f(x)=Hβ to be calculated.
[0172] The weighted extreme learning machine is compared with other prediction algorithms in the field of machine learning, including extreme learning machine (ELM), support vector machine (SVM), random forest, and extreme gradient boosting (XGBoost). The results of the comparative experiment are shown in Table 2. It can be seen that the prediction effect of the model is the best, and it is greatly improved compared with the ELM model. The reason may be that the addition of the weight matrix better solves the problem of sample imbalance in the current transformer data set.
[0173] Table 2 Comparison of algorithm performance evaluation indicators
[0174]
[0175] The present invention proposes to use a variational mode decomposition algorithm based on gray wolf optimization to decompose the transformer secondary side monitoring data and extract the corresponding time-frequency domain features, and send the final extracted time-frequency domain features to a weighted extreme learning machine for ratio difference angle difference prediction. The present invention adopts gray wolf optimization to select the optimal VMD decomposition parameters of the transformer secondary side monitoring current amplitude data: the number of decomposition layers and the penalty factor. Based on the optimized parameters, the original high-dimensional current amplitude data is decomposed and the features of the corresponding sub-signals in the time-frequency domain are extracted, thereby achieving dimensionality reduction of the transformer monitoring data. The use of weighted extreme learning machines in the prediction model better solves the data imbalance problem of fewer abnormal samples in the original sample data, and as a representative of single hidden layer feedforward neural networks, extreme learning machines have faster training and testing speeds than back-propagation neural networks.
[0176] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.
Claims
1. An electromagnetic current transformer error prediction method based on weighted extreme learning machine, characterized in that: The steps include: Step (1), collecting the secondary side current signal amplitude data of the three-phase current transformer every a minute under normal operating conditions to obtain a three-phase current amplitude matrix; and collecting the angle difference and ratio difference of the three-phase current transformer at the same time; Step (2), for the three-phase current amplitude matrix obtained in step (1), remove abnormal values in the matrix data; Step (3), standardizing the amplitude data in the three-phase current matrix processed in step (2) to obtain a standardized three-phase current amplitude matrix; Step (4), using the Grey Wolf Optimization Algorithm to optimize the parameters of the variational modal decomposition algorithm, and using the optimized variational modal decomposition algorithm to perform variational modal decomposition on the standardized three-phase current amplitude matrix obtained in step (3); Step (5), extracting time domain characteristic parameters and frequency domain characteristic parameters from the modal components decomposed in step (4), and concatenating all extracted parameters to obtain a reduced-dimensional three-phase current characteristic matrix; Step (6), using the reduced dimension three-phase current feature matrix obtained in step (5) as input, and using the angle difference and ratio difference as output, to train the extreme learning machine model to obtain a current transformer error detection model; Step (7), using the secondary side current signal amplitude data of the three-phase current transformer to be detected to obtain the three-phase current amplitude matrix of the three-phase current transformer to be detected; then through steps (2) to (5), the corresponding three-phase current characteristic matrix is obtained, and then input into the current transformer error detection model obtained in step (6) to obtain the predicted angle difference and ratio difference of the three-phase current transformer to be detected; Step (8), compare the prediction result of step (7) with the error threshold. If the predicted angle difference and ratio difference are both smaller than the threshold, it indicates that the current transformer is in a normal operating state at this time; otherwise, it is considered that the current transformer may be in an abnormal operating state at this time.
2. The electromagnetic current transformer error prediction method based on weighted extreme learning machine according to claim 1, characterized in that: In step (1), a=15; The specific method of step (2) is: according to the maximum and minimum values of the historical normal signal changes, the abnormal values that exceed the range are removed, and non-values are also removed; After each signal is eliminated in step (2), 900 data points need to be retained.
3. The electromagnetic current transformer error prediction method based on weighted extreme learning machine according to claim 1, characterized in that: The specific method of step (3) is: standardize the data processed by step (2) using the column normalization method.
4. The electromagnetic current transformer error prediction method based on weighted extreme learning machine according to claim 1, characterized in that: In step (4), the specific method of using the gray wolf optimization algorithm to optimize the parameters of the variational mode decomposition algorithm is: Step 4.1: Initialize the parameters of the Gray Wolf Algorithm: Initialize the basic parameters of the Gray Wolf Algorithm, such as population size, decreasing coefficient, random vector, and number of iterations. Step (4.2), gray wolf position initialization: randomly initialize the decomposition layer number k and penalty factor alpha of the variational mode decomposition algorithm as the initial position of the gray wolf algorithm; Step (4.3), calculate the fitness of the gray wolf individuals: the fitness function takes the mean of the permutation entropy, the optimization goal is to minimize the mean of the permutation entropy, save the three wolves with the best fitness, and define the first three individuals with the smallest fitness values as α wolf, β wolf and δ wolf respectively, and their positions are defined as and The rest are uniformly defined as ωwolf; Step (4.4), capture the prey, and capture the prey by the following formula In the formula, is the position of the individual after update; is the current position of the prey; t is the number of iterations of the current algorithm; is the distance between the current individual and the prey; The current location of the individual; is a synergy coefficient vector; are another synergy coefficient vector; Step (4.5), iteratively update the gray wolf individual, and update the latest position of the gray wolf according to the following four update formulas: In the formula, and is the direction and step length of the movement of ω wolf toward α wolf, β wolf and δ wolf; and are the positions of alpha wolf, beta wolf, and delta wolf; and is the distance between the current individual and α wolf, β wolf and δ wolf; recalculate the fitness and update α wolf, β wolf and δ wolf; In step (4.5), when the number of iterations is reached, the optimal decomposition layer k and penalty factor alpha of the VMD algorithm are output.
5. The electromagnetic current transformer error prediction method based on weighted extreme learning machine according to claim 4, characterized in that: The optimal decomposition level k is 2.
6. The electromagnetic current transformer error prediction method based on weighted extreme learning machine according to claim 5, characterized in that: In step (5), the time domain feature parameters include maximum value, minimum value, first-order moment, second-order origin moment and kurtosis factor; the frequency domain features include centroid frequency, mean square frequency, frequency variance and multi-scale permutation entropy; in, The maximum value is calculated as: max =max(x i ), where x i is the current data of the i-th sampling point; The minimum value is calculated as: min =min(x i ); The calculation formula for the first-order moment is: Where N is the number of samples; The calculation formula of the second-order origin moment is: The calculation formula of the kurtosis factor is: in, is the sampling mean, σ is the sampling standard deviation; The calculation formula of the center of gravity frequency is: P(f) is the spectrum value at frequency f; The formula for calculating the mean square frequency is: The frequency variance is calculated as: The calculation formula of multi-scale permutation entropy is: For m different numbers, there will be m! possible order patterns π; for a permutation of the number π, f(π) represents its frequency in the time series; the relative frequency is p(π) = f(π) / (N / τ-m+1).
7. The electromagnetic current transformer error prediction method based on weighted extreme learning machine according to claim 6, characterized in that: In step (5), the specific method of splicing is: The reduced dimension three-phase current feature matrix is A-phase sample, B-phase sample, and C-phase sample from top to bottom; The A phase sample is composed of 24-dimensional features of phases A, B, and C in sequence; The B-phase samples are composed of 24-dimensional features of the three phases B, A, and C in sequence; The C phase samples are formed by sequentially concatenating the 24-dimensional features of the three phases C, A, and B; Thus, a 72*3 dimensional three-phase characteristic matrix is constructed, that is, a reduced dimensional three-phase current characteristic matrix.
8. The electromagnetic current transformer error prediction method based on weighted extreme learning machine according to claim 1, characterized in that: In the extreme learning machine model, the cost-sensitive weight matrix is embedded in the extreme learning machine objective function, and each weight factor of the newly added weight matrix in the extreme learning machine objective function is constructed in a cost-sensitive way.
9. The electromagnetic current transformer error prediction method based on weighted extreme learning machine according to claim 1, characterized in that: In step (8), if the predicted angle difference or ratio difference exceeds the corresponding threshold value, the working state of the current transformer is determined and the operation, maintenance and repair work of the relevant staff is guided.
10. The electromagnetic current transformer error prediction method based on weighted extreme learning machine according to claim 9, characterized in that: The threshold range of the angle difference is ±10 points, and the threshold range of the ratio difference is ±0.2%; if the predicted angle difference or ratio difference exceeds the corresponding threshold, it is determined that the current transformer is in an out-of-tolerance state, and the staff needs to be reminded for maintenance and repair.
Citation Information
Patent Citations
Nonlinear error identification method and device of optical fiber current transformer, and storage medium
CN114371326A
Method and device for predicting error state of current transformer and electronic equipment
CN117195100A
Load short-term prediction method and system of hybrid prediction model
CN117973604A
