High-altitude electrode discharge voltage prediction method and system, device and medium

Abstract

A high-altitude electrode discharge voltage prediction method and system, a device and a medium, relating to the technical field of electrode discharge. The method comprises: acquiring a measured dataset of a high-altitude true electrode, and performing normalization processing and data partitioning on the measured dataset in a preset selection ratio to generate an initial training set and a test set; on the basis of the degree of influence between a feature variable and a target variable in the initial training set, constructing a target training set; carrying out model construction on the basis of a support vector regression method and an ant lion optimization algorithm in combination with the target training set, to generate an intermediate intelligent model; using the test set to test the intermediate intelligent model, and determining a target intelligent model; and inputting the measured dataset into the target intelligent model for discharge voltage calculation, to generate discharge voltage prediction data corresponding to the high-altitude true electrode. By means of the ant lion optimization intelligent algorithm and the target intelligent model, the accuracy of prediction results can be improved.

Description

A method, system, device and medium for predicting high-altitude electrode discharge voltage

[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on November 20, 2023, with application number 202311555856.0 and invention name “A method, system, device and medium for predicting high-altitude electrode discharge voltage”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present invention relates to the field of electrode discharge technology, and in particular to a method, system, equipment and medium for predicting high-altitude electrode discharge voltage. Background Art

[0003] At present, high-altitude areas are accelerating the construction of new power systems. Ball-plate gaps with electrode diameters of 0.6 to 1.6 m are widely used in the design of external insulation in valve halls of high-altitude converter stations. As the altitude increases, the air density decreases, resulting in a decrease in the external insulation performance of power equipment, which affects the external insulation discharge voltage of the valve hall equipment.

[0004] Currently, research on predicting discharge voltage at real-world high-altitude electrodes focuses primarily on two aspects: physical models of discharge mechanisms and empirical formulas for the statistical characteristics of discharge. Existing methods for predicting discharge voltage at high altitudes rely on empirical formulas and statistical models, but these methods fail to account for complex nonlinear relationships and the impact of high-altitude environments on prediction results, resulting in low prediction accuracy.

[0005] Summary of the Invention

[0006] The present invention provides a method, system, device and medium for predicting the discharge voltage of electrodes at high altitudes, which solves the technical problem that existing methods for predicting the discharge voltage of electrodes at high altitudes rely on empirical formulas and statistical models, but these methods fail to take into account complex nonlinear relationships and the impact of the high-altitude environment on the prediction results, resulting in low prediction accuracy.

[0007] The present invention provides a method for predicting discharge voltage of electrodes at high altitudes, comprising:

[0008] Obtaining an actual measurement data set of a real-type high-altitude electrode, normalizing and partitioning the actual measurement data set according to a preset selection ratio to generate an initial training set and a test set;

[0009] Constructing a target training set according to the degree of influence between the characteristic variables in the initial training set and the target variable;

[0010] Building a model based on the support vector regression method and the ant lion optimization algorithm in combination with the target training set to generate an intermediate intelligent model;

[0011] Testing the intermediate intelligent model using the test set to determine a target intelligent model;

[0012] The actual measurement data set is input into the target intelligent model to calculate the discharge voltage, and the discharge voltage prediction data corresponding to the high-altitude real electrode is generated.

[0013] Optionally, the step of normalizing and dividing the actual measurement data set according to a preset selection ratio to generate an initial training set and a test set includes:

[0014] Performing maximum and minimum normalization processing on the actual measurement data set to generate a sample parameter set;

[0015] The sample parameter set is divided according to a preset selection ratio to generate an initial training set and a test set.

[0016] Optionally, the step of constructing a target training set according to the degree of influence between the characteristic variables and the target variable in the initial training set includes:

[0017] Dividing the parameter set of air pressure, temperature, relative humidity and voltage rise rate in the initial training set into characteristic variables;

[0018] Dividing the parameter set where the U50 discharge voltage in the initial training set is located into a target variable;

[0019] The Pearson correlation coefficient is used to calculate the correlation between the characteristic variables and the target variables, and generate the correlation coefficient corresponding to the characteristic variables;

[0020] The characteristic variable corresponding to the maximum value of the correlation coefficient is used as the target training set.

[0021] Optionally, the step of constructing a model based on the support vector regression method and the ant lion optimization algorithm in combination with the target training set to generate an intermediate intelligent model includes:

[0022] The support vector regression method and Gaussian kernel function are used to build the model and generate the initial intelligent model;

[0023] An ant lion optimization algorithm is used to optimize the initial penalty coefficient and the initial kernel function parameters in the initial intelligent model to generate a target penalty coefficient and a target kernel function parameter;

[0024] The target penalty coefficient and the target kernel function parameter are used to update the initial intelligent model to generate an intermediate intelligent model.

[0025] Optionally, the step of using the support vector regression method and the Gaussian kernel function to construct the model and generate the initial intelligent model includes:

[0026] Support vector regression method is used to formalize and determine the model constraints and model objective function;

[0027] The model constraints are:

[0028] In the formula, ω represents the weight; n represents the dimension of vector y; b represents the bias value; f(x i ) represents x i The hyperplane of the mapping; x i represents the characteristic variable; y i represents the target variable; C is the initial penalty coefficient; ε (ε>0) is the maximum allowable regression error; l ε is an insensitive loss function;

[0029] The objective function of the model is:

[0030] Where f(x) represents the model objective function; α i ' and α represent Lagrange multipliers; k represents Gaussian kernel function; a represents intercept; x represents characteristic parameter; x i represents the characteristic variable;

[0031] The model constraints, the model objective function and the Gaussian kernel function are used to construct the model to generate an initial intelligent model;

[0032] The Gaussian kernel function is: k(x i ,x j )=exp(-γ||x i -x j || 2 );

[0033] Where k represents the Gaussian kernel function; x i ,x j Represents any two sets of feature parameters; γ represents the initial kernel function parameter.

[0034] Optionally, the step of optimizing the initial penalty coefficient and initial kernel function parameters in the initial intelligent model using the ant lion optimization algorithm to generate target penalty coefficient and target kernel function parameters includes:

[0035] An ant lion optimization algorithm is used to initialize the initial penalty coefficient and the initial kernel function parameters in the initial intelligent model to generate an ant lion set;

[0036] Calculating the fitness value of each ant and each ant lion in the ant lion set, generating a fitness set and counting the number of iterations;

[0037] The ant lion corresponding to the maximum fitness value in the fitness set is used as the initial elite ant lion;

[0038] A roulette wheel strategy is used to implement the random walk of the ants to update the positions of the ants and generate multiple ant positions;

[0039] Calculating the fitness corresponding to all the ant positions to generate an ant fitness set;

[0040] constructing iteration data using the number of iterations and the ant fitness set;

[0041] When the iterative data meets the stopping condition, the initial elite ant lion at the current moment is corrected and updated using a preset correction factor to generate a target elite ant lion;

[0042] The initial penalty coefficient and initial kernel function parameters corresponding to the target elite ant lion are used as the target penalty coefficient and target kernel function parameters.

[0043] Optionally, the step of testing the intermediate intelligent model using the test set to determine the target intelligent model includes:

[0044] Inputting the test set into the intermediate intelligent model to calculate the discharge voltage and generate the discharge voltage;

[0045] performing an error calculation using a difference between the discharge voltage and a true measurement value in the test set to generate an error rate;

[0046] When the error rate meets a preset error threshold, the intermediate intelligent model is used as the target intelligent model;

[0047] When the error rate does not meet the preset error threshold, the process jumps to executing the step of constructing a model based on the support vector regression method and the ant lion optimization algorithm in combination with the target training set to generate an intermediate intelligent model.

[0048] The present invention also provides a high-altitude electrode discharge voltage prediction system, comprising:

[0049] An initial training set and test set generation module is used to obtain an actual measurement data set of a real-type high-altitude electrode, normalize and divide the actual measurement data set according to a preset selection ratio, and generate an initial training set and a test set;

[0050] A target training set construction module is used to construct a target training set according to the degree of influence between the characteristic variables and the target variables in the initial training set;

[0051] An intermediate intelligent model generation module is used to construct a model based on the support vector regression method and the ant lion optimization algorithm in combination with the target training set to generate an intermediate intelligent model;

[0052] a target intelligent model determination module, configured to test the intermediate intelligent model using the test set to determine a target intelligent model;

[0053] The discharge voltage prediction data generation module is used to input the actual measurement data set into the target intelligent model to calculate the discharge voltage and generate the discharge voltage prediction data corresponding to the high-altitude true electrode.

[0054] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of implementing any of the above-mentioned high-altitude electrode discharge voltage prediction methods.

[0055] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed, any of the above-mentioned methods for predicting discharge voltage of an electrode at high altitude is implemented.

[0056] It can be seen from the above technical solutions that the present invention has the following advantages:

[0057] The present invention obtains an actual measurement data set of high-altitude true-type electrodes, normalizes and divides the actual measurement data set according to a preset selection ratio, and generates an initial training set and a test set. Based on the degree of influence between the characteristic variables and the target variables in the initial training set, a target training set is constructed. A model is constructed based on the support vector regression method and the ant lion optimization algorithm in combination with the target training set to generate an intermediate intelligent model. The intermediate intelligent model is tested using the test set to determine the target intelligent model. The actual measurement data set is input into the target intelligent model to calculate the discharge voltage, and discharge voltage prediction data corresponding to the high-altitude true-type electrodes is generated. This solves the technical problem that the existing high-altitude electrode discharge voltage prediction methods rely on empirical formulas and statistical models, but these methods cannot take into account the complex nonlinear relationships and the influence of the high-altitude environment on the prediction results, resulting in low accuracy of the prediction results. Based on the ant lion optimization intelligent algorithm and the target intelligent model, a reasonable reference can be provided for the insulation design of the true-type hardware of the converter valve in high-altitude areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0059] FIG1 is a flowchart of a method for predicting discharge voltage of an electrode at high altitude provided by a first embodiment of the present invention;

[0060] FIG2 is a flowchart of the steps of a method for predicting discharge voltage of an electrode at high altitude provided by a second embodiment of the present invention;

[0061] FIG3 is a heat map of correlation analysis of sample sets provided by Example 2 of the present invention;

[0062] FIG4 is a flowchart of a method for predicting discharge voltage of an electrode at high altitude provided by a second embodiment of the present invention;

[0063] FIG5 is a structural block diagram of a high-altitude electrode discharge voltage prediction system provided by a third embodiment of the present invention. DETAILED DESCRIPTION

[0064] Embodiments of the present invention provide a high-altitude electrode discharge voltage prediction method, system, device, and medium to address the technical problem that existing high-altitude electrode discharge voltage prediction methods rely on empirical formulas and statistical models, but these methods fail to consider complex nonlinear relationships and the impact of high-altitude environments on prediction results, resulting in low prediction accuracy.

[0065] In order to make the purpose, features, and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0066] Please refer to FIG1 , which is a flowchart of a method for predicting discharge voltage of an electrode at high altitude provided in a first embodiment of the present invention.

[0067] A method for predicting discharge voltage of an electrode at high altitude is provided in Example 1 of the present invention, comprising:

[0068] Step 101: Obtain an actual measurement data set of a real-type high-altitude electrode, perform normalization processing and data partitioning on the actual measurement data set according to a preset selection ratio, and generate an initial training set and a test set.

[0069] In this embodiment of the present invention, actual and real-time measurement data of discharges at high-altitude, real-world electrodes with different gap types are collected to construct an actual measurement dataset. This dataset is then subjected to maximum and minimum normalization to generate a sample parameter set. The sample parameter set is then divided according to a preset selection ratio to generate initial training and test sets.

[0070] Step 102: Construct a target training set based on the degree of influence between the feature variables in the initial training set and the target variable.

[0071] In this embodiment of the present invention, the parameter set containing air pressure, temperature, relative humidity, and voltage rise rate in the initial training set is classified as a feature variable. The parameter set containing the U50 discharge voltage in the initial training set is classified as a target variable. The Pearson correlation coefficient is used to calculate the correlation between the feature variables and the target variable, generating a correlation coefficient corresponding to the feature variable. The feature variable corresponding to the maximum correlation coefficient is used as the target training set.

[0072] Step 103: construct a model based on the support vector regression method and the ant lion optimization algorithm in combination with the target training set to generate an intermediate intelligent model.

[0073] In this embodiment of the present invention, a support vector regression method and a Gaussian kernel function are used to construct a model and generate an initial intelligent model. The ant lion optimization algorithm is then used to optimize the initial penalty coefficient and initial kernel function parameters in the initial intelligent model to generate a target penalty coefficient and target kernel function parameters. The target penalty coefficient and target kernel function parameters are then used to update the initial intelligent model to generate an intermediate intelligent model.

[0074] Step 104: Use the test set to test the intermediate intelligent model to determine the target intelligent model.

[0075] In this embodiment of the present invention, a test set is input into an intermediate intelligent model to calculate the discharge voltage and generate the discharge voltage. The difference between the discharge voltage and the actual measured value in the test set is used to calculate the error and generate the error rate. When the error rate meets a preset error threshold, the intermediate intelligent model is used as the target intelligent model. If the error rate does not meet the preset error threshold, the execution jumps to the step of constructing a model based on the support vector regression method and the ant lion optimization algorithm in conjunction with the target training set to generate the intermediate intelligent model.

[0076] Step 105: Input the actual measurement data set into the target intelligent model to calculate the discharge voltage and generate discharge voltage prediction data corresponding to the high-altitude real electrode.

[0077] In an embodiment of the present invention, the real-time measurement data in the actual measurement data set is input into the target intelligent model to calculate the discharge voltage, thereby obtaining the discharge voltage prediction data corresponding to the high-altitude true electrode.

[0078] In an embodiment of the present invention, an actual measurement data set of a high-altitude true-type electrode is obtained, and the actual measurement data set is normalized and divided according to a preset selection ratio to generate an initial training set and a test set. Based on the degree of influence between the characteristic variables and the target variables in the initial training set, a target training set is constructed. A model is constructed based on the support vector regression method and the ant lion optimization algorithm in combination with the target training set to generate an intermediate intelligent model. The intermediate intelligent model is tested using the test set to determine the target intelligent model. The actual measurement data set is input into the target intelligent model to calculate the discharge voltage, and discharge voltage prediction data corresponding to the high-altitude true-type electrode is generated. This solves the technical problem that the existing high-altitude electrode discharge voltage prediction methods rely on empirical formulas and statistical models, but these methods cannot take into account the complex nonlinear relationships and the influence of the high-altitude environment on the prediction results, resulting in low accuracy of the prediction results. Based on the ant lion optimization intelligent algorithm and the target intelligent model, a reasonable reference can be provided for the insulation design of the true-type hardware of the converter valve in high-altitude areas.

[0079] Please refer to FIG. 2 , which is a flowchart of a method for predicting discharge voltage of an electrode at high altitude provided in a second embodiment of the present invention.

[0080] Another method for predicting discharge voltage of an electrode at high altitude provided by Example 2 of the present invention includes:

[0081] Step 201: Obtain an actual measurement data set of a real-type high-altitude electrode, perform normalization processing and data partitioning on the actual measurement data set according to a preset selection ratio, and generate an initial training set and a test set.

[0082] Furthermore, step 201 may include the following sub-steps S11-S12:

[0083] S11. Perform maximum and minimum normalization processing on the actual measurement data set to generate a sample parameter set.

[0084] S12. Divide the sample parameter set according to the preset selection ratio to generate the initial training set and test set.

[0085] In the embodiment of the present invention, actual measurement data of discharge of different gap types of high-altitude real electrodes are collected, and the air pressure x1, temperature x2, relative humidity x3, voltage rise rate x4, ball electrode diameter x5, gap length x6 and U50 discharge voltage y are normalized to obtain the sample parameter set X regu =[x1,x2,x3,x4,x5,x6,y].

[0086] Preferably, the process of normalizing the sample parameter set and analyzing its correlation includes:

[0087] Use maximum and minimum normalization to normalize the sample data to [-1, 1], the formula is:

[0088] Where X is the original measured data set; X b and X a are the minimum and maximum values ​​in the data respectively; X regu The normalized data set is the sample parameter set.

[0089] As shown in Figure 3, the preset selection ratio is 50%. 50% of the sample parameter set is randomly selected as the initial training set to ensure the stability of the model training, and the remaining 50% of the data set is used as the test set. In the 50% sample parameter set, the air pressure (column vector x1[x 11 , x 12 ,...,x 1n ] T , assuming that there are 2n groups of total sample parameters), temperature (column vector x2[x 21 , x 22 ,...,x 2n ] T ), relative humidity (column vector x3[x 31 , x 32 ,...,x 3n ] T ), voltage rise rate (column vector x4[x 41 , x 42 ,...,x 4n ] T ) and U50 discharge voltage (column vector y[y 1， y2,...,y n ] T ) The number of column vector elements of the corresponding parameter set is the same. The parameter set of U50 discharge voltage is an n*1 column vector, and the parameter set of air pressure, temperature, relative humidity and voltage rise rate is an n*4 matrix. Therefore, the dimension of the parameter set of air pressure, temperature, relative humidity and voltage rise rate is 4 times that of the parameter set corresponding to U50 discharge voltage, but the row vector (x 11 , x 21 , x 31 , x 41 , y1) correspond in sequence, and are counted as the measured data volume of one discharge test, with a total of n groups.

[0090] Step 202: Construct a target training set based on the degree of influence between the feature variables in the initial training set and the target variable.

[0091] Furthermore, step 202 may include the following sub-steps S21-S24:

[0092] S21. Divide the parameter set of air pressure, temperature, relative humidity and voltage rise rate in the initial training set into feature variables.

[0093] S22. Divide the parameter set containing the U50 discharge voltage in the initial training set into the target variable.

[0094] S23. Use the Pearson correlation coefficient to calculate the correlation between the characteristic variables and the target variables, and generate the correlation coefficient corresponding to the characteristic variables.

[0095] S24. The characteristic variable corresponding to the maximum value of the correlation coefficient is used as the target training set.

[0096] In this embodiment of the present invention, 50% of the sample parameter set is selected as the initial training set, and the parameter set containing the normalized air pressure, temperature, relative humidity, and voltage rise rate is divided into feature variables. The parameter set containing the U50 discharge voltage is divided into the target variable.

[0097] The Pearson Correlation Coefficient can be used to analyze the correlation between any two sets of vectors. The calculation formula is as follows:

[0098] Where ρ(A,B) represents the correlation between vector A and vector B; A represents the sample parameter set X regu Any column vector of , B represents the sample parameter set X regu Any column vector of μ A represents the mean of vector A, μ B represents the mean of vector B, A i Represents the i-th value of column vector A, B i Denotes the i-th value of column vector B, and E represents the covariance. Sample parameter set vectors are grouped two by two. The correlation between the sample set vectors is calculated, and the degree of influence of the characteristic variable on the discharge voltage is determined based on the correlation coefficient. The sample parameter set here includes characteristic variables and target variables. By analyzing the correlation between the sample parameter sets in pairs, the degree of influence of the characteristic variables on the discharge voltage can be determined. The influence of each characteristic variable on the discharge voltage is determined, and based on the degree of influence, the characteristic variables with the greatest impact on the discharge voltage are selected as the input for model training. Specifically, the characteristic variable corresponding to the maximum correlation coefficient is used as the target training set.

[0099] Step 203: Use support vector regression method and Gaussian kernel function to build a model and generate an initial intelligent model.

[0100] Furthermore, step 203 may include the following sub-steps S31-S32:

[0101] S31. Use support vector regression method to formalize and determine the model constraints and model objective function.

[0102] S32. Use model constraints, model objective function and Gaussian kernel function to build the model and generate an initial intelligent model.

[0103] In the embodiment of the present invention, the model constraints and the model objective function are determined by formalizing the support vector regression method. The formalized SVR (Support Vector Regression) support vector regression method, that is, the model constraints are:

[0104] In the formula, ω represents the weight; n represents the dimension of vector y; b represents the bias value; f(x i ) represents x i The hyperplane of the mapping; x i represents the characteristic variable; y i represents the target variable; C is the initial penalty coefficient; ε (ε>0) is the maximum allowable regression error; l ε is an insensitive loss function.

[0105] Where, l ε (z) represents the insensitive loss function corresponding to the true error of the sample set; z represents the true error of the sample set; f|z| represents the absolute value of the true error of the sample set. Mapping the original data to a high-dimensional plane, the model objective function can be finally obtained:

[0106] Where f(x) represents the model objective function corresponding to the discharge voltage U(x); α i ' and α represent Lagrange multipliers; k represents Gaussian kernel function; a represents intercept; x represents characteristic parameter; x i Represents the characteristic variable. Since the RBF kernel function, also known as the Gaussian kernel function, can realize nonlinear mapping, the number of parameters and computational difficulty are relatively small, the RBF kernel function is selected for the establishment of the SVR model to construct the initial intelligent model:

[0107] The Gaussian kernel function is: k(x i ,x j )=exp(-γ||x i -x j || 2 );

[0108] Where k represents the Gaussian kernel function; x i ,x j Represents any two sets of feature parameters; γ represents the initial kernel function parameter.

[0109] Step 204: Use the ant lion optimization algorithm to optimize the initial penalty coefficient and initial kernel function parameters in the initial intelligent model to generate target penalty coefficient and target kernel function parameters.

[0110] Furthermore, step 204 may include the following sub-steps S41-S48:

[0111] S41. Use the ant lion optimization algorithm to initialize the initial penalty coefficient and the initial kernel function parameters in the initial intelligent model to generate an ant lion set.

[0112] S42. Calculate the fitness value of each ant and each ant lion in the ant lion set, generate a fitness set and count the number of iterations.

[0113] S43. The antlion with the maximum fitness value in the fitness set is selected as the initial elite antlion.

[0114] S44. Use the roulette strategy to implement the random walk of ants to update the positions of ants and generate multiple ant positions.

[0115] S45. Calculate the fitness corresponding to all ant positions and generate an ant fitness set.

[0116] S46. Use the number of iterations and the ant fitness set to construct iteration data.

[0117] S47. When the iterative data meets the stopping condition, the initial elite ant lion at the current moment is corrected and updated using a preset correction factor to generate a target elite ant lion.

[0118] S48. Using the initial penalty coefficient and initial kernel function parameters corresponding to the target elite ant lion as the target penalty coefficient and target kernel function parameters.

[0119] In an embodiment of the present invention, it can be seen from the above-mentioned initial intelligent model that its penalty coefficient C and kernel function parameter γ jointly determine the SVR regression performance. The present invention selects the ALO algorithm, namely the ant lion optimization algorithm, to optimize the penalty coefficient C and kernel function parameter γ, and introduces a correction factor to improve the accuracy and robustness of the model prediction.

[0120] Use Python to create an ALO antlion optimization algorithm object, limit the search range to [-1,-1,-1,-1], the variable value range to [1,1,1,1], set the number of antlions and ants to be the same and both to 80, and set the maximum number of iterations to 30.

[0121] Calculate the fitness value of each ant and each antlion in the antlion set, generate a fitness set, and count the number of iterations. The antlion with the maximum fitness in the fitness set is used as the initial elite antlion. A roulette wheel strategy is used to implement the ant's random walk to update the ant's position and generate multiple ant positions. Specifically, the roulette wheel strategy is used to select which antlion will prey on a particular ant. Each ant can only be preyed on by one antlion, and the antlion with higher fitness has a greater probability of capturing the ant. In addition, once an ant falls into a trap set by an antlion, the antlion will throw sand to the edge of the trap to prevent the ant from escaping. At this time, the range of the ant's random walk will be drastically reduced. This phenomenon is simulated by the following equation:

[0122] Where c t represents the minimum value of the random walk of all variables at the tth iteration; d t represents the maximum value of the random walk of all variables at the tth iteration; I is the scaling factor; T is the maximum number of iterations; and v is a number that changes as the number of iterations increases. When the fitness value of an ant is lower than that of an antlion, it is considered that the antlion has captured it. At this time, the antlion will update its position based on the position of the ant, that is, the position of the initial elite antlion is updated to the position corresponding to the maximum fitness in the ant fitness set, thus generating an intermediate elite antlion:

[0123] Where, It is represented as the position of the j-th ant lion in the t-th selection generation; is the position of the i-th ant in the t-th generation; f is the fitness function.

[0124] When the number of iterations reaches the maximum number of iterations or the fitness of the ant lion is lower than the fitness of the ant, that is, when the iteration data meets the stopping condition, the iteration stops, and the preset correction factor is used to correct and update the elite ant lion at the current moment to generate the target elite ant lion Al elite The initial penalty coefficient and initial kernel function parameters corresponding to the target elite ant lion are used as the target penalty coefficient and target kernel function parameters, as shown in the following formula.

[0125] Step 205: Update the initial intelligent model using the target penalty coefficient and the target kernel function parameters to generate an intermediate intelligent model.

[0126] In an embodiment of the present invention, the initial intelligent model is updated by optimizing the initial penalty coefficient and the initial kernel function parameters using the ant lion optimization algorithm, and the ALO-SVR model training is completed to obtain an intermediate intelligent model.

[0127] Step 206: Use the test set to test the intermediate intelligent model to determine the target intelligent model.

[0128] Furthermore, step 206 may include the following sub-steps S51-S54:

[0129] S51. Input the test set into the intermediate intelligent model to calculate the discharge voltage and generate the discharge voltage.

[0130] S52: Calculate the error using the difference between the discharge voltage and the actual measurement value in the test set to generate an error rate.

[0131] S53. When the error rate meets the preset error threshold, the intermediate intelligent model is used as the target intelligent model.

[0132] S54. When the error rate does not meet the preset error threshold, jump to the step of building a model based on the support vector regression method and the ant lion optimization algorithm combined with the target training set to generate an intermediate intelligent model.

[0133] The preset error threshold refers to a critical value that the error rate cannot exceed, which is set in advance based on actual needs.

[0134] In this embodiment of the present invention, the remaining 50% of the dataset is used as a test set and substituted into the trained ALO-SVR intelligent model, or intermediate intelligent model, to calculate the discharge voltage. This is then compared with the actual measured value, and the error rate of the intermediate intelligent model is calculated to verify the effectiveness of the intelligent prediction model. If the error rate is less than a preset error threshold, the intermediate intelligent model is used as the target intelligent model. Otherwise, the process jumps to the step of generating the intermediate intelligent model by constructing a model based on the support vector regression method and the ant lion optimization algorithm in conjunction with the target training set.

[0135] Step 207: Input the actual measurement data set into the target intelligent model to calculate the discharge voltage and generate discharge voltage prediction data corresponding to the high-altitude real electrode.

[0136] In the embodiment of the present invention, the specific implementation process of step 207 is similar to that of step 105 and will not be repeated here.

[0137] In an embodiment of the present invention, as shown in Figure 4, measured variables such as air pressure, temperature, relative humidity, gap type, and discharge voltage are input. Correlation analysis and data set partitioning are then performed, i.e., the measured data set is normalized and partitioned to generate an initial training set and a test set. The influence between the characteristic variables in the initial training set and the target variable is then determined to construct a target training set. The initial intelligent model is trained using the target training set. The specific training process is as follows: initializing parameters, randomly generating ant and ant lion positions, calculating the fitness values ​​of each ant and ant lion in the ant lion set, and generating a fitness set. The ant lion with the maximum fitness in the fitness set is selected as the initial elite ant lion. The ant lion with the maximum fitness in the fitness set is selected as the initial elite ant lion. The positions of the initial elite ant lions are obtained, the ant lion positions are generated, and the number of iterations is counted. When the number of iterations meets a preset iteration threshold, the current intermediate elite ant lion is corrected and updated using a preset correction factor to generate a target elite ant lion. The initial penalty coefficient and initial kernel function parameters corresponding to the target elite ant lion are used as the target penalty coefficient and target kernel function parameters. When the number of iterations does not meet the preset iteration threshold, a roulette wheel strategy is used to update the ant position corresponding to the ant lion position, generating multiple ant positions. When the ant fitness corresponding to an ant position is lower than the corresponding ant lion fitness, the ant lion position is updated to the position corresponding to the maximum fitness in the ant fitness set, generating an intermediate elite ant lion. The intermediate elite ant lion is used as the initial elite ant lion, and the execution jumps to the steps of obtaining the initial elite ant lion position, generating the ant lion position, and counting the number of iterations. When the ant fitness corresponding to the ant position is better than (i.e., greater than) the corresponding ant lion fitness, the current intermediate elite ant lion is corrected and updated using a preset correction factor to generate a target elite ant lion. The initial penalty coefficient and initial kernel function parameters corresponding to the target elite ant lion are used as the target penalty coefficient and target kernel function parameters. The test set is input into the intermediate intelligent model to calculate the discharge voltage. The discharge voltage is generated as the output of the test set prediction. The prediction result is compared with the actual measured value, and the error rate of the prediction model is calculated to verify the effectiveness of the intelligent prediction model.

[0138] Please refer to FIG5 , which is a structural block diagram of a high-altitude electrode discharge voltage prediction system provided by a third embodiment of the present invention.

[0139] A high-altitude electrode discharge voltage prediction system provided in Example 3 of the present invention includes:

[0140] The initial training set and test set generation module 501 is used to obtain the actual measurement data set of the high-altitude real electrode, normalize and divide the actual measurement data set according to a preset selection ratio, and generate the initial training set and test set.

[0141] The target training set construction module 502 is used to construct a target training set according to the degree of influence between the characteristic variables in the initial training set and the target variables.

[0142] The intermediate intelligent model generation module 503 is used to construct a model based on the support vector regression method and the ant lion optimization algorithm in combination with the target training set to generate an intermediate intelligent model.

[0143] The target intelligent model determination module 504 is used to test the intermediate intelligent model using a test set to determine the target intelligent model.

[0144] The discharge voltage prediction data generation module 505 is used to input the actual measurement data set into the target intelligent model to calculate the discharge voltage and generate discharge voltage prediction data corresponding to the high-altitude real electrode.

[0145] Optionally, the initial training set and test set generation module 501 includes:

[0146] The sample parameter set generation module is used to perform maximum and minimum normalization processing on the actual measurement data set to generate a sample parameter set.

[0147] The initial training set and test set generation submodule is used to divide the sample parameter set according to the preset selection ratio to generate the initial training set and test set.

[0148] Optionally, the target training set construction module 502 includes:

[0149] The feature variable division module is used to divide the parameter set of air pressure, temperature, relative humidity and voltage rise rate in the initial training set into feature variables.

[0150] The target variable division module is used to divide the parameter set containing the U50 discharge voltage in the initial training set into target variables.

[0151] The correlation coefficient generation module is used to calculate the correlation between the feature variables and the target variables using the Pearson correlation coefficient, and generate the correlation coefficient corresponding to the feature variables.

[0152] The target training set construction submodule is used to take the characteristic variable corresponding to the maximum value of the correlation coefficient as the target training set.

[0153] Optionally, the intermediate intelligent model generation module 503 includes:

[0154] The initial intelligent model generation module is used to construct the model using the support vector regression method and Gaussian kernel function to generate the initial intelligent model.

[0155] The target penalty coefficient and target kernel function parameter generation module is used to optimize the initial penalty coefficient and initial kernel function parameters in the initial intelligent model using the ant lion optimization algorithm to generate the target penalty coefficient and target kernel function parameters.

[0156] The intermediate intelligent model generation submodule is used to update the initial intelligent model using the target penalty coefficient and the target kernel function parameters to generate an intermediate intelligent model.

[0157] Optionally, the initial intelligent model generation module may perform the following steps:

[0158] Support vector regression method is used to formalize and determine the model constraints and model objective function;

[0159] The model constraints are:

[0160] In the formula, ω represents the weight; n represents the dimension of vector y; b represents the bias value; f(x i ) represents x i The hyperplane of the mapping; x i represents the characteristic variable; y i represents the target variable; C is the initial penalty coefficient; ε (ε>0) is the maximum allowable regression error; l ε is an insensitive loss function;

[0161] The model objective function is:

[0162] Where f(x) represents the model objective function; α i ' and α represent Lagrange multipliers; k represents Gaussian kernel function; a represents intercept; x represents characteristic parameter; x i represents the characteristic variable;

[0163] Model constraints, model objective functions and Gaussian kernel functions are used to construct the model and generate an initial intelligent model;

[0164] The Gaussian kernel function is: k(x i ,x j )=exp(-γ||x i -x j || 2 );

[0165] Where k represents the Gaussian kernel function; x i ,x j Represents any two sets of feature parameters; γ represents the initial kernel function parameter.

[0166] Optionally, the target penalty coefficient and target kernel function parameter generation module may perform the following steps:

[0167] The ant lion optimization algorithm is used to initialize the initial penalty coefficient and initial kernel function parameters in the initial intelligent model to generate an ant lion set.

[0168] Calculate the fitness value of each ant and each ant lion in the ant lion set, generate a fitness set and count the number of iterations;

[0169] The ant lion corresponding to the maximum fitness in the fitness set is used as the initial elite ant lion;

[0170] The roulette wheel strategy is used to implement the random walk of ants to update the positions of ants and generate multiple ant positions;

[0171] Calculate the fitness corresponding to all ant positions and generate an ant fitness set;

[0172] The iteration number and ant fitness set are used to construct iteration data;

[0173] When the iterative data meets the stopping condition, the preset correction factor is used to correct and update the initial elite ant lion at the current moment to generate the target elite ant lion;

[0174] The initial penalty coefficient and initial kernel function parameters corresponding to the target elite ant lion are used as the target penalty coefficient and target kernel function parameters.

[0175] Optionally, the target intelligent model determination module 504 includes:

[0176] The discharge voltage generation module is used to input the test set into the intermediate intelligent model to calculate the discharge voltage and generate the discharge voltage.

[0177] The error rate generation module is used to calculate the error using the difference between the discharge voltage and the actual measurement value in the test set to generate the error rate.

[0178] The target intelligent model determination submodule is used to use the intermediate intelligent model as the target intelligent model when the error rate meets the preset error threshold.

[0179] The jump execution module is used to jump to the step of building a model based on the support vector regression method and the ant lion optimization algorithm combined with the target training set to generate an intermediate intelligent model when the error rate does not meet the preset error threshold.

[0180] An embodiment of the present invention further provides an electronic device, comprising: a memory and a processor, wherein the memory stores a computer program; when the computer program is executed by the processor, the processor executes the high-altitude electrode discharge voltage prediction method as described in any of the above embodiments.

[0181] The memory may be an electronic memory such as a flash memory, an EEPROM (Electrically Erasable Programmable Read-Only Memory), an EPROM, a hard disk, or a ROM. The memory has storage space for program codes for executing any of the method steps in the above method. For example, the storage space for program codes may include individual program codes for implementing the various steps in the above method. These program codes may be read from or written to one or more computer program products. These computer program products include program code carriers such as a hard disk, a compact disk (CD), a memory card, or a floppy disk. The program code may be compressed, for example, in an appropriate form. When these codes are executed by a computing and processing device, the computing and processing device executes the various steps in the high-altitude electrode discharge voltage prediction method described above.

[0182] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for predicting the high-altitude electrode discharge voltage according to any of the above embodiments is implemented.

[0183] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0184] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0185] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0186] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0187] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0188] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for predicting the discharge voltage of high - altitude electrodes, characterized in that, it includes: Obtain the actual measurement data set of high - altitude full - scale electrodes, perform normalization processing and data partitioning on the actual measurement data set according to a preset selection ratio, and generate an initial training set and a test set; Construct a target training set according to the influence degree between the feature variables and the target variable in the initial training set; Based on the support vector regression method and the ant lion optimization algorithm, combine the target training set to construct a model and generate an intermediate intelligent model; Use the test set to test the intermediate intelligent model to determine the target intelligent model; Input the actual measurement data set into the target intelligent model for discharge voltage calculation, and generate the discharge voltage prediction data corresponding to the high - altitude full - scale electrodes.

2. The method for predicting the discharge voltage of high - altitude electrodes according to claim 1, characterized in that, The step of performing normalization processing and data partitioning on the actual measurement data set according to a preset selection ratio to generate an initial training set and a test set includes: Perform maximum - minimum normalization processing on the actual measurement data set to generate a sample parameter set; Divide the sample parameter set according to the preset selection ratio to generate an initial training set and a test set.

3. The method for predicting the discharge voltage of high - altitude electrodes according to claim 1, characterized in that, The step of constructing a target training set according to the influence degree between the feature variables and the target variable in the initial training set includes: Divide the parameter set where the air pressure, temperature, relative humidity, and voltage rise rate are located in the initial training set into feature variables; Divide the parameter set where the U50 discharge voltage is located in the initial training set into target variables; Use the Pearson correlation coefficient to calculate the correlation between the feature variables and the target variable respectively, and generate the correlation coefficient corresponding to the feature variable; Take the feature variable corresponding to the maximum value of the correlation coefficient as the target training set.

4. The method for predicting the discharge voltage of high - altitude electrodes according to claim 3, characterized in that, The step of constructing a model based on the support vector regression method and the ant lion optimization algorithm in combination with the target training set to generate an intermediate intelligent model includes: Use the support vector regression method and the Gaussian kernel function to construct a model to generate an initial intelligent model; model; Use the ant lion optimization algorithm to optimize the initial penalty coefficient and the initial kernel function parameter in the initial intelligent model to generate a target penalty coefficient and a target kernel function parameter; Use the target penalty coefficient and the target kernel function parameter to update the initial intelligent model to generate an intermediate intelligent model.

5. The method for predicting the discharge voltage of high - altitude electrodes according to claim 4, characterized in that, The step of using the support vector regression method and the Gaussian kernel function to construct a model to generate an initial intelligent model includes: Formalize the support vector regression method to determine the model constraint conditions and the model objective function; The model constraint conditions are as follows: where ω represents the weight; n represents the dimension of the vector y; b represents the bias value; f(x i ) represents the hyperplane of the mapping of x i ; x i represents the feature variable; y i represents the target variable; C is the initial penalty coefficient; ε (ε > 0) is the maximum allowable error of regression; l ε is the insensitive loss function; The objective function of the model is as follows: where f(x) represents the model objective function; α i ' and α represent Lagrange multipliers; k represents the Gaussian kernel function; a represents the intercept; x represents the characteristic parameter; x i represents the characteristic variable; Use the model constraint conditions, the model objective function, and the Gaussian kernel function to construct a model to generate an initial intelligent model; The Gaussian kernel function is: k(x i , x j ) = exp(-γ||x i - x j || 2 ); where k represents the Gaussian kernel function; x i , x j represent any two sets of characteristic parameters; γ represents the initial kernel function parameter.

6. The high-altitude electrode discharge voltage prediction method according to claim 4, characterized in that, the step of using the ant lion optimization algorithm to optimize the initial penalty coefficient and the initial kernel function parameter in the initial intelligent model to generate a target penalty coefficient and a target kernel function parameter includes: using the ant lion optimization algorithm to perform parameter initialization on the initial penalty coefficient and the initial kernel function parameter in the initial intelligent model to generate an ant lion set; calculating the fitness values of each ant and each ant lion in the ant lion set to generate a fitness set and counting the number of iterations; taking the ant lion corresponding to the maximum fitness value in the fitness set as the initial elite ant lion; using the roulette wheel strategy to realize the random walk of the ants to update the positions of the ants to generate multiple ant positions; calculating the fitness corresponding to all the ant positions to generate an ant fitness set; using the number of iterations and the ant fitness set to construct iterative data; when the iterative data meets the stop condition, using a preset correction factor to correct and update the initial elite ant lion at the current moment to generate a target elite ant lion; taking the initial penalty coefficient and the initial kernel function parameter corresponding to the target elite ant lion as the target penalty coefficient and the target kernel function parameter.

7. The high-altitude electrode discharge voltage prediction method according to claim 1, characterized in that, the step of using the test set to test the intermediate intelligent model to determine the target intelligent model includes: inputting the test set into the intermediate intelligent model to calculate the discharge voltage to generate the discharge voltage; using the difference between the discharge voltage and the true measurement value in the test set to calculate the error to generate an error rate; when the error rate meets the preset error threshold, taking the intermediate intelligent model as the target intelligent model; when the error rate does not meet the preset error threshold, jumping to execute the step of constructing an intermediate intelligent model by combining the target training set based on the support vector regression method and the ant lion optimization algorithm.

8. A high-altitude electrode discharge voltage prediction system, characterized in that, comprising: an initial training set and test set generation module, configured to obtain the actual measurement data set of the high-altitude full-scale electrode, perform normalization processing and data division on the actual measurement data set according to a preset selection ratio to generate an initial training set and a test set; a target training set construction module, configured to construct a target training set according to the influence degree between the feature variables and the target variables in the initial training set; an intermediate intelligent model generation module, configured to construct a model by combining the support vector regression method and the ant lion optimization algorithm with the target training set to generate an intermediate intelligent model; a target intelligent model determination module, configured to use the test set to test the intermediate intelligent model to determine the target intelligent model; a discharge voltage prediction data generation module, configured to input the actual measurement data set into the target intelligent model to calculate the discharge voltage to generate the discharge voltage prediction data corresponding to the high-altitude full-scale electrode.

9. An electronic device, characterized in that, It includes a memory and a processor. A computer program is stored in the memory. When the computer program is executed by the processor, the processor is caused to execute the steps of the high-altitude electrode discharge voltage prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, on which a computer program is stored, characterized in that when the computer program is executed, it implements the high-altitude electrode discharge voltage prediction method according to any one of claims 1 to 7.

Images