Lightning arrester performance detection method and device
The random forest algorithm evaluates the important level of the lightning arrester operating parameters and combines the time weight, which solves the problem that traditional detection methods rely on a single parameter, and achieves more accurate and reliable lightning arrester performance detection.
Patent Information
- Application Number
- CN202510219497.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-26
- Publication Date
- 2025-05-06
AI Technical Summary
Traditional lightning arrester performance detection methods rely on a single electrical parameter and cannot fully reflect the performance status of the lightning arrester, resulting in low detection accuracy and reliability.
The random forest algorithm is used to evaluate the important level of the lightning arrester operating parameters, and the parameter weight is determined based on the time weight, and performance detection is performed through weighted polynomial regression.
It improves the comprehensiveness and accuracy of lightning arrester performance detection, enhances prediction capabilities and robustness, and improves the accuracy and reliability of detection.
Smart Images
Figure CN119936539A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of performance detection, and more specifically, to a method and device for detecting the performance of a lightning arrester. Background Art
[0002] Lightning arresters are important protective devices used in power systems to protect electrical equipment from lightning overvoltage and operating overvoltage damage. The stability of their performance is directly related to the safe and reliable operation of the power system. Therefore, it is very important to accurately and efficiently detect the performance of lightning arresters.
[0003] In the traditional arrester performance detection method, a single electrical parameter index is usually used for simple evaluation, such as leakage current, operating voltage, etc. However, this method has many limitations. First, the operation of the arrester is affected by multiple parameters, and relying on a single parameter alone cannot fully reflect its performance status. The traditional method fails to fully consider the differences in the importance of each parameter and the impact of time factors on the parameters, and the accuracy and reliability of the arrester performance detection are low.
[0004] Therefore, an accurate and reliable method for detecting the performance of lightning arresters is needed. Summary of the invention
[0005] The purpose of the present disclosure is to provide a method and device for detecting the performance of a lightning arrester, so as to improve the accuracy and reliability of the detection of the performance of the lightning arrester.
[0006] A first aspect of an embodiment of the present disclosure provides a method for detecting performance of a lightning arrester, comprising: Obtain the importance level of each parameter in the arrester operating parameters based on the random forest algorithm, and determine the weight of each arrester operating parameter based on the time of each arrester operating parameter; Determining the weight of each parameter in the arrester operating parameters based on the importance level of each parameter in the arrester operating parameters and the weight of each arrester operating parameter; The weight of each parameter in each arrester operation parameter is used as the weight distribution of weighted polynomial regression; The performance of the arrester is tested based on weighted polynomial regression to obtain the performance test results.
[0007] A second aspect of the embodiments of the present disclosure provides a lightning arrester performance detection device, comprising: A first weight determination module is used to obtain the importance level of each parameter in the arrester operation parameters based on a random forest algorithm, and determine the weight of each arrester operation parameter based on the time of each arrester operation parameter; A second weight determination module, used to determine the weight of each parameter in the arrester operating parameters based on the importance level of each parameter in the arrester operating parameters and the weight of each arrester operating parameter; A weighted polynomial determination module is used to use the weight of each parameter in each arrester operation parameter as a weight distribution for weighted polynomial regression; The performance detection module is used to detect the performance of the arrester based on weighted polynomial regression to obtain a performance detection result.
[0008] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of the above-mentioned lightning arrester performance detection method when executing the computer program.
[0009] According to a fourth aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned lightning arrester performance detection method are implemented.
[0010] The arrester performance detection method and device provided by the embodiments of the present disclosure have the following beneficial effects: The present disclosure can accurately evaluate the importance level of each arrester operating parameter through the random forest algorithm, which helps to identify the key factors that have the greatest impact on the performance of the arrester. At the same time, combined with the time weight of each parameter, the evaluation dimension of the parameter is further refined, making the consideration of the parameter more comprehensive and detailed. After determining the importance level and time weight of each parameter, the two are combined to assign weights to each arrester operating parameter, ensuring that the key parameters are emphasized during the performance detection process, while not ignoring other parameters that may affect the performance, thereby improving the comprehensiveness and accuracy of the arrester performance detection. The present disclosure uses the weight of each parameter as the weight allocation of the weighted polynomial regression, so that the regression model can focus more on the key parameters when fitting the arrester performance data, reduce the impact of noise or minor parameters on the test results, enhance the prediction ability and robustness of the present disclosure, and thus improve the accuracy and reliability of the arrester performance detection. BRIEF DESCRIPTION OF THE DRAWINGS
[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0012] Figure 1 A schematic diagram of a flow chart of a lightning arrester performance detection method provided in an embodiment of the present disclosure; Figure 2 A structural block diagram of a lightning arrester performance detection device provided in an embodiment of the present disclosure; Figure 3 A schematic block diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION
[0013] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present disclosure. However, it should be clear to those skilled in the art that the present disclosure may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obstructing the description of the present disclosure with unnecessary details.
[0014] In order to make the purpose, technical solutions and advantages of the present disclosure more clear, specific embodiments will be described below in conjunction with the accompanying drawings.
[0015] Please refer to Figure 1 , Figure 1 A schematic diagram of a flow chart of a lightning arrester performance detection method provided in an embodiment of the present disclosure, the method comprising: S101: Obtaining the importance level of each parameter in the arrester operating parameters based on the random forest algorithm, and determining the weight of each arrester operating parameter based on the time of each arrester operating parameter.
[0016] In this embodiment, the random forest algorithm is an ensemble learning algorithm based on decision trees, which constructs multiple decision trees by randomly sampling with replacement from the original training data set, and then combines the results of these decision trees for prediction or evaluation. When used to determine the importance level of the arrester operating parameters, it will give a corresponding importance score or importance level according to the degree of influence of each parameter on the final result (such as the arrester performance evaluation result).
[0017] In a random forest, each decision tree will divide nodes according to different feature values during the construction process, so that the impurity of the data in the divided child nodes (such as Gini impurity or information entropy) is reduced. Average impurity reduction measures the importance of a feature by calculating the contribution of each feature to the impurity reduction in all decision trees. For a given feature, the average value of the impurity reduction caused by it as a dividing feature in all decision trees is calculated. This average value is the average impurity reduction score of the feature. The higher the average impurity reduction score, the greater the role of the feature in distinguishing different categories or predicting target values, that is, the more important the feature is. The importance level can be determined according to a pre-set mapping table, as shown in Table 1: Table 1 Importance level mapping table
[0018] Considering the numerous operating parameters of the arrester, such as leakage current, voltage, temperature, etc., different parameters have different degrees of influence on the performance of the arrester. Using the random forest algorithm to determine the importance level of each parameter can help clarify which parameters play a key role in the operating status and performance of the arrester and which parameters have relatively little impact. In subsequent performance testing and analysis, we can focus more on key parameters and improve the accuracy and efficiency of detection.
[0019] Arrester operating parameters refer to various data that can reflect the operating status of the arrester, such as leakage current, voltage, temperature, etc. The changes in these parameters can reflect the working performance and health status of the arrester to a certain extent.
[0020] Considering that the timeliness of the arrester operating parameters is very important, the parameters obtained at different times have different values for the current arrester performance evaluation. The parameters closer to the current time usually reflect the real-time status of the arrester, have a higher reference value for performance detection, and should be given a higher weight; while the parameters at an earlier time have a relatively low reference value, and the weight should be reduced accordingly. In this way, parameter information at different times can be used more reasonably, making the performance test results more in line with the actual situation. By setting multiple thresholds, the difference between the time when each arrester operating parameter is collected and the current time and multiple thresholds can be used as the basis for judgment, so as to set different weights for each arrester operating parameter.
[0021] In the above process, the parameters of each decision tree in the random forest determine the division of the overall importance level. Therefore, the parameters of the decision tree in the random forest can be adjusted according to the characteristics of the arrester operation data. For example, before obtaining the importance level of each parameter in the arrester operation parameters based on the random forest algorithm, it also includes: In response to the sample size of the arrester operation data being less than the first sample size, reducing the reference value of the number of decision trees by the first number of decision trees; In response to the sample size of the arrester operation data being greater than the second sample size, increasing the reference value of the number of decision trees by the second number of decision trees; In response to the discrete value of the arrester operation data being greater than the first discrete value, the reference value of the minimum sample number of the decision tree is increased by the first sample number, and the reference value of the minimum sample number of the leaf node of the decision tree is increased by the second sample number.
[0022] In this embodiment, considering that when the sample size of the arrester operation data is small, if the number of decision trees is too large, the model may be overfitted. Because a small number of samples is difficult to represent the full picture of the data, too many decision trees may over-learn some noise or special cases in the sample, and cannot be well generalized to new data. Therefore, by reducing the reference value of the number of decision trees by the first number of decision trees, the complexity of the model can be reduced, the model can be more focused on learning the real rules in the data, and the stability and generalization ability of the model can be improved. Therefore, when the sample size is less than the first sample size, the reference value of the number of decision trees is reduced.
[0023] Considering that when the sample size is sufficient, increasing the number of decision trees can allow the random forest model to better capture the complex relationships and features in the data. More decision trees can learn and analyze the data from different perspectives, integrate more information, and thus improve the accuracy and robustness of the model. Each decision tree can learn different parts of the data, and by integrating the results of the decision trees, a more comprehensive and accurate classification of parameter importance levels can be obtained. Therefore, when the sample size is greater than the second sample size, increase the number of decision trees.
[0024] Among them, the first sample number, the second sample number, the first decision tree number and the second decision tree number can be obtained based on experience and data from the test process, or can be set based on reference values for solving similar problems.
[0025] Considering that when the discrete value of the arrester operation data is large, the distribution of the data is relatively scattered, and there may be some outliers or uneven data distribution. Increasing the minimum number of samples of the decision tree can make the decision tree more cautious when dividing nodes, avoiding excessive splitting of the decision tree due to a few data points with large discrete values, thereby making the model more stable, reducing the impact of outliers on the model, and making the decision tree pay more attention to the overall distribution and main characteristics of the data, and improving the accuracy of the parameter importance level division. Increasing the minimum number of samples in the leaf nodes of the decision tree is to ensure that the data in the leaf nodes have a certain degree of representativeness and stability. When the data discrete value is large, if the number of samples in the leaf nodes is too small, the results of the leaf nodes may be too dependent on a few data points, which is prone to fluctuations and inaccuracies. Increasing the minimum number of samples of leaf nodes can make the results of leaf nodes more reliable, thereby improving the accuracy and stability of the entire random forest model for the division of the importance level of the arrester operation parameters, so that the model can better adapt to the data characteristics with large discrete values.
[0026] Therefore, when the discrete value of the arrester operation data is greater than the first discrete value, the minimum sample number of the decision tree and the minimum sample number of the leaf node can be increased.
[0027] The discreteness of the arrester operation data can be calculated by calculating the standard deviation or mean square error of the arrester operation parameter data, and the first discrete value can be obtained based on experience or data in the test process. In this embodiment, the reference value of the number of decision trees, the reference value of the minimum number of samples, and the reference value of the minimum number of samples of leaf nodes can be determined by cross-validation or corresponding values when solving similar problems.
[0028] In this embodiment, the random forest algorithm obtains and determines the importance level of each parameter in an arrester operating parameter, and the time of each arrester operating parameter determines the weight of each arrester operating parameter, which is equivalent to a total weight rather than the weight of each data therein.
[0029] In this embodiment, a comprehensive judgment may also be performed, for example, obtaining the importance level of each parameter in the arrester operating parameters based on a random forest algorithm, including: Obtaining an importance level for each parameter in the first set of arrester operating parameters based on the first number of decision trees; obtaining an importance level of each parameter in the second set of arrester operating parameters based on a second number of decision trees; obtaining an importance level of each parameter in a third set of arrester operating parameters based on a third number of decision trees; Determine the importance level of each parameter in the arrester operating parameters by performing weighted calculation on the importance level of each parameter in the first group of arrester operating parameters, the importance level of each parameter in the second group of arrester operating parameters, and the importance level of each parameter in the third group of arrester operating parameters; The sum of the first number, the second number, and the third number is the total number of decision trees in the random forest algorithm; The first group of arrester operating parameters is the arrester operating parameters whose absolute value of the difference between the acquisition time of the arrester operating parameters and the current time is greater than the first time length; The second group of arrester operating parameters are arrester operating parameters whose absolute value of the difference between the acquisition time of the arrester operating parameters and the current time is less than or equal to the first time length and greater than the second time length; The third group of arrester operating parameters are arrester operating parameters whose absolute value of the difference between the acquisition time of the arrester operating parameters and the current time is less than or equal to the second time length.
[0030] In this embodiment, it is considered that the insulation materials inside the arrester will gradually age during long-term operation. For example, in the early stage of operation, the importance of the leakage current parameter may be relatively low, because the insulation performance of the arrester is good at this time, and the leakage current is generally within the normal range and does not change much. However, as time goes by, the aging of the insulation material may cause the leakage current to gradually increase. At this time, the importance of the leakage current parameter will be significantly increased, and it will become a key parameter reflecting whether the performance of the arrester has declined.
[0031] Therefore, as time goes by, the impact of various arrester operating parameters on the arrester performance is different, so different data sets can be set to identify different parameter impacts, i.e., importance levels, through different decision trees. It can be understood that the first number of decision trees form a small random forest, and the second and third numbers are the same. The first number, the second number, and the third number can be determined based on experiments.
[0032] Similarly, the time factor is also taken into account in the above-mentioned grouping decision process, so when performing weighted calculation, the same means as the above-mentioned time-based determination of the operating parameters of each arrester can be used. The above record states that the weights of the arrester operating parameters at each time node can be determined by setting multiple thresholds, where the multiple thresholds are the first duration, the second duration, and the third duration in this embodiment. The first duration, the second duration, and the third duration can be set based on experience or basic knowledge of the industry.
[0033] The principle of weight distribution should be that the closer the data is to the current time, the greater its weight. For example, the weight of the arrester operating parameter whose absolute value of the difference between the acquisition time and the current time is greater than the first time length can be 20%, the weight of the arrester operating parameter whose absolute value of the difference between the acquisition time and the current time is less than or equal to the first time length and greater than the second time length can be 30%, and the weight of the arrester operating parameter whose absolute value of the difference between the acquisition time and the current time is less than or equal to the second time length can be 50%. When grouping the decision tree, the parameters of the decision tree can be left unset, and the default parameter values can be used to build the decision tree model.
[0034] S102: Determine the weight of each parameter in the arrester operating parameters based on the importance level of each parameter in the arrester operating parameters and the weight of each arrester operating parameter.
[0035] In this embodiment, the weight of each parameter in each arrester operating parameter is determined based on the importance level of each parameter in the arrester operating parameter and the weight of each arrester operating parameter, including: Determining the weight of each parameter in the arrester operating parameters based on the importance level of each parameter in the arrester operating parameters; The weight of each parameter in the arrester operating parameters is determined based on the weight of each parameter in the arrester operating parameters and the weight of each arrester operating parameter. The mapping table can be obtained through the aforementioned mapping table, and different weights are assigned to each importance level and normalized. For example, the weight corresponding to the parameter with an importance level of 4 is 50%, the weight corresponding to the parameter with an importance level of 3 is 35%, the weight corresponding to the parameter with an importance level of 2 is 10%, and the weight corresponding to the parameter with an importance level of 4 is 5%. Considering that there are many operating parameters of the lightning arrester and there may be multiple parameters with an importance level of 4, the sum of their corresponding weights may exceed 1 or be less than 1. At this time, normalization can be performed so that the sum of their weights is 1.
[0036] The weight of each parameter in the arrester operating parameters can be obtained through the above calculation, which can be multiplied by the weight of each arrester operating parameter and normalized to determine the weight of each parameter in the arrester operating parameters. The arrester operating parameters may include: operating voltage, leakage current, residual voltage, partial discharge, temperature or humidity, etc. One or more.
[0037] S103: using the weight of each parameter in each arrester operation parameter as the weight distribution of the weighted polynomial regression.
[0038] In this embodiment, the determined weights are applied to the weighted polynomial regression, that is, the weight of each parameter is used as the weight of the corresponding parameter in the weighted polynomial regression model. When constructing the weighted polynomial regression model, these weights will affect the contribution of each parameter to the final regression result.
[0039] For example, there is Data points , ,in is the independent variable, i.e. the arrester operating parameter, is the dependent variable, i.e. the performance status of the arrester. The purpose is to fit a The weighted polynomial regression model is The general form of a weighted polynomial of degree is , is the parameter to be estimated corresponding to the operating parameter of the first arrester, is the parameter to be estimated corresponding to the operating parameter of the second arrester, For the The parameters to be estimated corresponding to the operating parameters of the arrester are: is the intercept. In weighted polynomial regression, for each data point Assign a weight The goal is to minimize the weighted sum of squared errors, and its objective function is for: ,in are the model parameters to be estimated. Smallest , can be about Find the partial derivatives and set them to 0 to obtain a set of normal equations. By solving the normal equations, we can get The estimated value of For the The parameters to be estimated correspond to the operating parameters of the arrester. That is, the weight of each parameter in the arrester operating parameters calculated above.
[0040] S104: Detecting the performance of the arrester based on weighted polynomial regression to obtain a performance detection result.
[0041] In this embodiment, after the weighted polynomial regression is obtained, it can be used to predict the performance parameters of the arrester, such as predicting the leakage current change trend of the arrester in the future, so as to achieve the purpose of performance prediction, because the performance of the arrester is largely related to the leakage current. It should be noted that when the performance of the arrester is predicted by the leakage current, the dependent variable in the corresponding arrester operating parameters is the leakage current, and the independent variable is other operating parameters. The obtained weighted polynomial regression algorithm can be used for prediction. Taking the change of AC leakage current over time under the operating voltage as an example, the historical AC leakage current monitoring data and the corresponding time, ambient temperature, operating voltage and other information are collected as independent variables, and the AC leakage current is used as the dependent variable. By performing regression analysis on these data, a mathematical model can be established to predict the leakage current of the arrester under similar operating conditions in the future, so as to obtain how long the arrester can be used, that is, the service life. The performance test results of the arrester can be determined according to the length of time that can be used, which can be obtained through a mapping table, as shown in Table 2: Table 2 Performance test result mapping table
[0042] It can be concluded from the above that the present disclosure can accurately evaluate the importance level of each arrester operating parameter through the random forest algorithm, which helps to identify the key factors that have the greatest impact on the performance of the arrester. At the same time, combined with the time weight of each parameter, the evaluation dimension of the parameter is further refined, making the consideration of the parameter more comprehensive and detailed. After determining the importance level and time weight of each parameter, the two are combined to assign weights to each arrester operating parameter, ensuring that the key parameters are emphasized during the performance test, while other parameters that may affect the performance are not ignored, thereby improving the comprehensiveness and accuracy of the arrester performance test. The present disclosure uses the weight of each parameter as the weight distribution of the weighted polynomial regression, so that the regression model can focus more on the key parameters when fitting the arrester performance data, reduce the impact of noise or minor parameters on the test results, enhance the prediction ability and robustness of the present disclosure, and thus improve the accuracy and reliability of the arrester performance test.
[0043] In one embodiment of the present disclosure, the arrester performance detection method further includes: The number of polynomial regressions starts with a first value and increases according to a second value, and the mean square error is calculated after each increase until a third value is reached; The value corresponding to the lowest mean square error is used as the degree of weighted polynomial regression; Solve for the coefficients of a weighted polynomial regression based on the weighted least squares method.
[0044] In one embodiment of the present disclosure, the arrester performance detection method further includes: In response to at least one parameter in the importance level of each parameter in the arrester operation parameters having an importance level lower than a first level, regularizing the weighted polynomial regression based on elastic network regression to obtain a target weighted polynomial regression; The performance of the arrester is tested based on target weighted polynomial regression to obtain the performance test results.
[0045] In this embodiment, the degree of polynomial regression: In polynomial regression, the highest degree of the independent variable is the degree of polynomial regression. The reason why the degree of polynomial regression is increased from a value in a certain step size and the mean square error is calculated is that different polynomial regression degrees may have different degrees of fit to the data. If the degree is too low, the model may not be able to capture the rules in the data well, resulting in insufficient fitting; if the degree is too high, it may overfit the noise in the data, causing the model to perform poorly on new data. By gradually trying different degrees of fit and finding the degree with the lowest mean square error, the model can achieve a better balance between fitting data and generalization ability. The first value, the second value, and the third value can be determined based on experience. For example, the first value can be 1, the second value can be 1, and the third value can be 8.
[0046] The weighted least squares method is used because the various operating parameters of the arrester have different degrees of influence on its performance. By assigning weights to each parameter, the importance of each parameter in the model can be more accurately reflected, so that the solved regression coefficient can better reflect the true relationship of the data and improve the accuracy of the model.
[0047] When the importance level of at least one parameter in the arrester operation parameters is lower than the first level, it means that the contribution of this parameter to the model is relatively small, or there may be some abnormal conditions, which may easily lead to overfitting of the model. The first level can be level 3 in Table 1, that is, when parameters with importance levels of 1 or 2 appear, the weighted polynomial regression can be regularized based on elastic network regression to obtain the target weighted polynomial regression.
[0048] Elastic Net Regression by adding and The objective function after regularization is constrained by the regularization term to constrain the weighted polynomial regression. ,in, is the weighted residual sum of squares, which measures the difference between the model prediction and the true value. yes Regularization term, yes Regularization parameter, which can make some unimportant coefficients become 0 and play the role of variable selection. yes Regularization term, yes Regularization parameter, which shrinks the coefficients and prevents overfitting.
[0049] Using elastic network regression for regularization processing can constrain the parameters of the model to prevent these less important parameters from having too much influence on the model, further improve the stability and generalization ability of the model, and make the arrester performance detection results based on target weighted polynomial regression more reliable.
[0050] It can be concluded from the above that the present disclosure gradually increases the number of polynomial regressions according to a certain value, and calculates the mean square error after each increase. This strategy ensures that the optimal number of polynomial regressions can be found, balances the model's fit and generalization ability, avoids insufficient fitting due to too low a number or overfitting due to too high a number, thereby improving the accuracy and stability of lightning arrester performance detection. The present disclosure introduces weighted least squares in polynomial regression, and assigns different weights according to the importance of each operating parameter of the lightning arrester, so that the model can more accurately reflect the true relationship of the data when solving the regression coefficient, improves the accuracy of the model, and makes the performance test results closer to the actual situation. In the present disclosure, when there are parameters with a lower importance level in the operating parameters of the lightning arrester, the weighted polynomial regression is regularized by elastic network regression, which effectively constrains the influence of these parameters on the model. The regularization processing not only makes some unimportant coefficients become 0 and realizes variable selection, but also prevents overfitting by shrinking the coefficients, further enhancing the stability and generalization ability of the model, making the arrester performance detection results based on target weighted polynomial regression more accurate and reliable.
[0051] In one embodiment of the present disclosure, the parameters of elastic network regression include: lasso regression parameters and ridge regression parameters; The lightning arrester performance detection method also includes: In response to the data volatility of the arrester operating parameters satisfying the first volatility condition and no multicollinearity occurring in the target weighted polynomial regression, increasing the reference value of the lasso regression parameter based on the first lasso regression step size, and increasing the reference value of the ridge regression parameter based on the first ridge regression step size; In response to multicollinearity occurring in the target weighted polynomial regression and the data volatility of the arrester operating parameters not satisfying a first volatility condition, increasing a reference value of the ridge regression parameter based on a second ridge regression step size; In response to the data volatility of the arrester operating parameters satisfying a first volatility condition and multicollinearity occurring in the target weighted polynomial regression, increasing reference values of the lasso regression parameters based on a first lasso regression step size, and increasing reference values of the ridge regression parameters based on a third ridge regression step size; The sum of the first ridge regression step length and the second ridge regression step length is smaller than the third ridge regression step length.
[0052] In one embodiment of the present disclosure, in response to a variance inflation factor of the target weighted polynomial regression being greater than or equal to a first inflation threshold, multicollinearity occurs in the target weighted polynomial regression; In response to the variance inflation factor of the target weighted polynomial regression being less than a first inflation threshold, multicollinearity does not occur in the target weighted polynomial regression; In response to the coefficient of variation of the leakage current data in the arrester operating parameters being greater than or equal to the first variation threshold, the data volatility of the arrester operating parameters satisfies the first volatility condition; In response to the coefficient of variation of the leakage current data in the arrester operation parameters being less than a first variation threshold, the data volatility of the arrester operation parameters does not satisfy a first volatility condition.
[0053] In this embodiment, the lasso regression parameter is the Regularization parameter of the regularization term , the ridge regression parameter is the one mentioned above Regularization parameter of the regularization term .
[0054] Increase the regularization parameter appropriately and It can enhance the stability of the model. From the ridge regression part, it shrinks the coefficients towards zero. This shrinkage can reduce the large changes in coefficients caused by data fluctuations, allowing the model to maintain relatively stable coefficient estimates when facing fluctuating data. For example, when encountering outliers in leakage current caused by lightning strikes, the regularization effect of ridge regression can prevent the model from over-adjusting coefficients to fit these outliers.
[0055] Increasing the regularization parameter can also help reduce overfitting to noise. The Lasso regression part can compress some unimportant coefficients to zero, thereby screening variables and simplifying the model. In the case of large fluctuations in data and noise, this can help the model ignore those spurious relationships caused by noise. For example, if some short-term environmental factor fluctuations (such as small-scale temperature fluctuations) on the leakage current are actually noise, Lasso regression can be adjusted Compressing these corresponding coefficients makes the model pay more attention to the truly influential factors, reduces the fitting of noise, and makes the model smoother. In this way, the model can better capture the main trends and real relationships in the data and improve the accuracy of the prediction of the arrester leakage current.
[0056] Considering that large data fluctuations will bring stability problems to the model. When constructing a regression model to predict the arrester leakage current, the model will overfit the fluctuations. That is, the model will learn the noise and short-term fluctuation patterns in the data, rather than the true intrinsic relationship. For example, in lightning weather, the instantaneous peak of the leakage current may be misjudged by the model as a normal operating mode, thereby affecting the model's prediction accuracy of the leakage current under other normal conditions. When the data volatility of the arrester operating parameters meets the first volatility condition, the data may have large fluctuations or noise. At this time, adding lasso regression parameters and ridge regression parameters can make the model less sensitive to the noise in the data, improve the model's anti-interference ability and stability, prevent the model from overfitting, and better capture the real laws in the data to adapt to the fluctuating data situation to accurately detect the arrester performance.
[0057] At the same time, when multicollinearity occurs in the target weighted polynomial regression, there is a strong linear relationship between the independent variables, which will lead to unstable model parameter estimation and affect the accuracy and reliability of the model. Increasing the ridge regression parameters can reduce the impact of multicollinearity on the model by shrinking the parameters, making the model more stable and improving the generalization ability of the model, thereby more accurately detecting the performance of the arrester.
[0058] Whether the data volatility of the arrester operating parameters meets the first volatility condition can be determined by the coefficient of variation of the leakage current data in the arrester operating parameters. When the coefficient of variation is greater than or equal to the first variation threshold, it means that the data volatility of the arrester operating parameters meets the first volatility condition. Otherwise, it does not meet the first volatility condition. That is, the first volatility condition can be that the coefficient of variation of the leakage current data in the arrester operating parameters is greater than or equal to the first variation threshold, and the first variation threshold can be determined based on experience or reference parameters of related issues.
[0059] For example, calculate the standard deviation of the arrester leakage current data , ,in It is the first Data points, is the mean, and finally calculate the coefficient of variation .
[0060] Whether multicollinearity occurs in the target weighted polynomial regression can be determined based on the variance inflation factor of the target weighted polynomial regression. When the variance inflation factor of the target weighted polynomial regression is greater than or equal to the first inflation threshold, multicollinearity occurs in the target weighted polynomial regression. Otherwise, multicollinearity does not occur. Therefore, the reference values of the lasso regression parameters and the reference values of the ridge regression parameters can be appropriately increased.
[0061] Variance inflation factor is an indicator to measure the severity of multicollinearity in a multiple regression model. It reflects the multiple by which the variance of a certain independent variable is amplified due to the linear correlation between the independent variables. For example, for each independent variable , whose variance inflation factor is The calculation formula is ,in, The independent variable The coefficient of determination is obtained by regressing the dependent variable and the other independent variables as predictor variables. In simple terms, in a regression model containing multiple independent variables, each independent variable is taken as the target and fitted with other independent variables to obtain the goodness of fit. The variable used to calculate .when When , it means that there is no linear correlation between the independent variable and other independent variables, that is, there is no multicollinearity problem. When the value of is between 1 and 5, the degree of multicollinearity is considered acceptable; when When , it indicates that there is serious multicollinearity, which will have a greater negative impact on the coefficient estimation of the regression model. Therefore, the first inflation threshold can be 5. The first inflation threshold can also be set according to actual application conditions.
[0062] in, ,in, is the residual sum of squares, which represents the observed value With the predicted value The sum of squares of the differences between and reflects the part not explained by the model. is the total sum of squared deviations, which represents the observed value With the mean The sum of squares of the differences between them reflects the total variation of the observed data.
[0063] At the same time, considering that the data volatility does not meet the first volatility condition, it means that the data is relatively stable and there is no obvious noise or volatility interference. At this time, the main problem of the model is multicollinearity, rather than the need to use lasso regression to deal with outliers or unstable factors in the data. If the lasso regression parameters are adjusted at this time, it may unnecessarily change the way the model fits the data, and may introduce new problems instead of solving the existing multicollinearity problem.
[0064] This is because lasso regression is achieved by adding The regularization term, which penalizes the absolute value of the parameter, has the function of automatic feature selection, which will make the parameter estimates of some unimportant features become zero, thereby achieving the purpose of screening variables. However, in the current scenario, due to the low volatility of the data, there is no need to further screen variables through lasso regression to reduce the complexity of the model or remove the influence of noise. In addition, lasso regression may be too aggressive in setting the coefficients of some variables that still contribute to the model to zero despite the existence of collinearity, thereby losing some information and affecting the accuracy of the model.
[0065] Ridge regression can shrink parameter estimates toward zero, but will not make the parameters exactly zero. This can retain the influence of all independent variables on the dependent variable to a certain extent, while reducing the instability of parameter estimation caused by multicollinearity, making the model more stable and reliable.
[0066] Therefore, when multicollinearity occurs in the target weighted polynomial regression and the data volatility of the arrester operating parameters does not meet the first volatility condition, the reference value of the ridge regression parameter is increased based on the second ridge regression step size; and the reference value of the lasso regression parameter is not adjusted. The reference value of the ridge regression parameter and the reference value of the lasso regression parameter can be determined based on experience or the usual settings of the elastic network.
[0067] Secondly, considering that when data volatility and multicollinearity coexist, the situation is the most complex and severe, and has the greatest impact on the model. Increasing the ridge regression parameters based on the third ridge regression step length is to solve both data volatility and multicollinearity problems at the same time. It is necessary to adjust the parameters more significantly to stabilize the model and improve accuracy. Therefore, the third ridge regression step length should be greater than the sum of the first ridge regression step length and the second ridge regression step length to ensure that the model can effectively cope with this complex situation.
[0068] It can be concluded from the above that the present disclosure accurately adjusts the reference values of the lasso regression parameters and the ridge regression parameters according to the data volatility of the arrester operating parameters and whether multicollinearity occurs in the target weighted polynomial regression, ensuring that the model can maintain the optimal state when facing different data characteristics, thereby improving the accuracy of the arrester performance detection. In the present disclosure, when the data volatility of the arrester operating parameters is large, by increasing the reference values of the lasso regression parameters and the ridge regression parameters, the present disclosure is made less sensitive to the noise and outliers in the data, and the anti-interference ability and stability of the present disclosure are enhanced, which helps the model to accurately capture the performance characteristics of the arrester in a fluctuating data environment, and improves the reliability of the arrester performance detection results. In the present disclosure, when multicollinearity occurs in the target weighted polynomial regression, by increasing the reference values of the ridge regression parameters, the influence of multicollinearity on the model parameter estimation is effectively reduced, which helps to avoid the problem of inaccurate coefficient estimation caused by the linear relationship between the independent variables of the model, thereby improving the accuracy of the arrester performance detection.
[0069] In one embodiment of the present disclosure, determining the weight of each arrester operating parameter based on the time of each arrester operating parameter includes: Determine the weight corresponding to each arrester operating parameter; In response to the absolute value of the difference between the acquisition time of the arrester operating parameter and the current time being greater than the first duration, reducing the weight corresponding to the arrester operating parameter based on the first duration step; In response to the absolute value of the difference between the acquisition time of the arrester operating parameter and the current time being less than or equal to the first duration and greater than the second duration, reducing the weight corresponding to the arrester operating parameter based on the second duration step; The sum of the weights corresponding to the operating parameters of each arrester is one.
[0070] In this embodiment, considering that the operating state of the arrester changes over time, the operating parameters closer to the current time can better reflect its current real performance. However, the parameters collected at an earlier time may no longer accurately represent the current performance status due to the longer time interval. Therefore, by reducing the weight of the parameters collected at an earlier time, the performance test can be more focused on recent data, improving the accuracy of the test results.
[0071] Setting two different time thresholds and step sizes for adjusting weights between partitions is to consider the impact of time on parameter importance in a more detailed manner. Different time intervals have different impacts on the timeliness of parameters. The importance of parameters with longer time intervals decreases more significantly, so a larger step size is used to reduce the weight; while the importance of parameters with shorter time intervals decreases relatively less, so a smaller step size is used to reduce the weight. This makes the weight adjustment more in line with the actual situation and improves the accuracy of performance testing.
[0072] The first duration, the second duration, the first duration step and the second duration step may be determined based on actual conditions or based on data from an experimental process.
[0073] In one embodiment of the present disclosure, the arrester performance detection method further includes: The weights corresponding to the arrester operating parameters are determined based on the first formula.
[0074] The first formula is ,in, For the The initial weights of the parameters, Indicates that for Arrester operating parameters and their unnormalized weights. The adjustment coefficient can be determined based on experience or experiments. Expressed as The absolute value of the difference between the parameter acquisition time and the current time, Indicates the first duration, Indicates the second duration. Indicates rounding down.
[0075] Each weight can then be normalized to obtain the weight of each arrester operating parameter.
[0076] From the above, it can be concluded that the present disclosure takes into account the characteristics of the arrester performance changing over time. By reducing the weight of the operating parameters collected earlier, the recent data occupies a larger proportion in the performance test, ensuring that the test results can better reflect the current performance status of the arrester, thereby improving the accuracy of the arrester performance test.
[0077] Corresponding to the arrester performance detection method of the above embodiment, Figure 2 This is a structural block diagram of a lightning arrester performance detection device provided by an embodiment of the present disclosure. For ease of description, only the parts related to the embodiment of the present disclosure are shown. Figure 2 The arrester performance detection device 20 includes: a first weight determination module 21, a second weight determination module 22, a weighted polynomial determination module 23 and a performance detection module 24.
[0078] The first weight determination module 21 is used to obtain the importance level of each parameter in the arrester operation parameters based on the random forest algorithm, and determine the weight of each arrester operation parameter based on the time of each arrester operation parameter; A second weight determination module 22, for determining the weight of each parameter in the arrester operating parameters based on the importance level of each parameter in the arrester operating parameters and the weight of each arrester operating parameter; A weighted polynomial determination module 23, used to use the weight of each parameter in each arrester operating parameter as a weight distribution for weighted polynomial regression; The performance detection module 24 is used to detect the performance of the arrester based on weighted polynomial regression to obtain a performance detection result.
[0079] In one embodiment of the present disclosure, the arrester performance detection device 20 further includes: a regression parameter determination module; A regression parameter determination module, used for starting the number of polynomial regressions from a first value, increasing according to a second value, and calculating a mean square error after each increase until a third value is reached; The value corresponding to the lowest mean square error is used as the degree of weighted polynomial regression; Solve for the coefficients of a weighted polynomial regression based on the weighted least squares method.
[0080] In one embodiment of the present disclosure, the arrester performance detection device 20 further includes: a regularization module; A regularization module, for performing regularization processing on the weighted polynomial regression based on elastic network regression to obtain a target weighted polynomial regression in response to the importance level of at least one parameter in the importance level of each parameter in the arrester operation parameters being lower than the first level; The performance of the arrester is tested based on target weighted polynomial regression to obtain the performance test results.
[0081] In one embodiment of the present disclosure, the parameters of elastic network regression include: lasso regression parameters and ridge regression parameters; The arrester performance detection device 20 further includes: a regression parameter adjustment module; A regression parameter adjustment module, for increasing a reference value of a lasso regression parameter based on a first lasso regression step length, and increasing a reference value of a ridge regression parameter based on a first ridge regression step length, in response to the data volatility of the arrester operating parameter satisfying a first volatility condition and no multicollinearity occurring in the target weighted polynomial regression; In response to multicollinearity occurring in the target weighted polynomial regression and the data volatility of the arrester operating parameters not satisfying a first volatility condition, increasing a reference value of the ridge regression parameter based on a second ridge regression step size; In response to the data volatility of the arrester operating parameters satisfying a first volatility condition and multicollinearity occurring in the target weighted polynomial regression, increasing reference values of the lasso regression parameters based on a first lasso regression step size, and increasing reference values of the ridge regression parameters based on a third ridge regression step size; The sum of the first ridge regression step length and the second ridge regression step length is smaller than the third ridge regression step length.
[0082] In one embodiment of the present disclosure, in response to a variance inflation factor of the target weighted polynomial regression being greater than or equal to a first inflation threshold, multicollinearity occurs in the target weighted polynomial regression; In response to the variance inflation factor of the target weighted polynomial regression being less than a first inflation threshold, multicollinearity does not occur in the target weighted polynomial regression; In response to the coefficient of variation of the leakage current data in the arrester operating parameters being greater than or equal to the first variation threshold, the data volatility of the arrester operating parameters satisfies the first volatility condition; In response to the coefficient of variation of the leakage current data in the arrester operation parameters being less than a first variation threshold, the data volatility of the arrester operation parameters does not satisfy a first volatility condition.
[0083] In one embodiment of the present disclosure, the first weight determination module 21 is specifically used to determine the weight corresponding to each arrester operating parameter; In response to the absolute value of the difference between the acquisition time of the arrester operating parameter and the current time being greater than the first duration, reducing the weight corresponding to the arrester operating parameter based on the first duration step; In response to the absolute value of the difference between the acquisition time of the arrester operating parameter and the current time being less than or equal to the first duration and greater than the second duration, reducing the weight corresponding to the arrester operating parameter based on the second duration step; The sum of the weights corresponding to the operating parameters of each arrester is one.
[0084] In one embodiment of the present disclosure, it further includes: a random forest parameter determination module; a random forest parameter determination module, configured to reduce a reference value of the number of decision trees by the first number of decision trees in response to a sample size of the arrester operation data being less than a first sample size; In response to the sample size of the arrester operation data being greater than the second sample size, increasing the reference value of the number of decision trees by the second number of decision trees; In response to the discrete value of the arrester operation data being greater than the first discrete value, the reference value of the minimum sample number of the decision tree is increased by the first sample number, and the reference value of the minimum sample number of the leaf node of the decision tree is increased by the second sample number.
[0085] See also Figure 3 , Figure 3 A schematic block diagram of an electronic device provided by an embodiment of the present disclosure. Figure 3 The electronic device 300 in the embodiment shown may include: one or more processors 301, one or more input devices 302, one or more output devices 303 and one or more memories 304. The processors 301, input devices 302, output devices 303 and memories 304 communicate with each other via a communication bus 305. The memory 304 is used to store computer programs, which include program instructions. The processor 301 is used to execute the program instructions stored in the memory 304. The processor 301 is configured to call the program instructions to execute the functions of each module / unit in the above-mentioned device embodiments, such as Figure 2 The functions of modules 21 to 24 are shown.
[0086] It should be understood that in the embodiment of the present disclosure, the processor 301 may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.
[0087] The input device 302 may include a touch panel, a fingerprint collection sensor (for collecting the user's fingerprint information and fingerprint direction information), a microphone, etc., and the output device 303 may include a display (LCD, etc.), a speaker, etc.
[0088] The memory 304 may include a read-only memory and a random access memory, and provide instructions and data to the processor 301. A portion of the memory 304 may also include a non-volatile random access memory. For example, the memory 304 may also store information about the device type.
[0089] In a specific implementation, the processor 301, input device 302, and output device 303 described in the embodiments of the present disclosure can execute the implementation methods described in the first and second embodiments of the lightning arrester performance detection method provided in the embodiments of the present disclosure, and can also execute the implementation methods of the electronic device described in the embodiments of the present disclosure, which will not be repeated here.
[0090] In another embodiment of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by the processor, all or part of the processes in the above-mentioned embodiment method are implemented, and the computer program can also be completed by instructing the relevant hardware through the computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.
[0091] The computer-readable storage medium may be an internal storage unit of the electronic device of any of the aforementioned embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Furthermore, the computer-readable storage medium may also include both an internal storage unit of the electronic device and an external storage device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.
[0092] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this disclosure.
[0093] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the electronic devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0094] In the several embodiments provided in the present application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces or units, or it can be an electrical, mechanical or other form of connection.
[0095] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present disclosure.
[0096] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0097] The above are only specific embodiments of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present disclosure, and these modifications or replacements should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.
Claims
1. A lightning arrester performance detection method, characterized in that: include: Obtain the importance level of each parameter in the arrester operating parameters based on the random forest algorithm, and determine the weight of each arrester operating parameter based on the time of each arrester operating parameter; Determining the weight of each parameter in the arrester operating parameters based on the importance level of each parameter in the arrester operating parameters and the weight of each arrester operating parameter; Using the weight of each parameter in the operating parameters of each arrester as the weight distribution of weighted polynomial regression; The performance of the arrester is tested based on the weighted polynomial regression to obtain a performance test result.
2. The lightning arrester performance detection method according to claim 1, characterized in that: Also includes: The number of times of the polynomial regression starts from the first value and increases according to the second value, and the mean square error is calculated after each increase until a third value is reached; The value corresponding to the lowest mean square error is used as the degree of the weighted polynomial regression; The coefficients of the weighted polynomial regression are solved based on the weighted least squares method.
3. The lightning arrester performance detection method according to claim 1, characterized in that: Also includes: In response to the importance level of at least one parameter among the importance levels of each parameter in the arrester operation parameters being lower than the first level, regularizing the weighted polynomial regression based on elastic network regression to obtain a target weighted polynomial regression; The performance of the arrester is tested based on the target weighted polynomial regression to obtain a performance test result.
4. The arrester performance detection method according to claim 3, characterized in that: The parameters of the elastic network regression include: lasso regression parameters and ridge regression parameters; The lightning arrester performance detection method also includes: In response to the data volatility of the arrester operating parameters satisfying a first volatility condition and no multicollinearity occurs in the target weighted polynomial regression, increasing the reference value of the lasso regression parameter based on a first lasso regression step size, and increasing the reference value of the ridge regression parameter based on a first ridge regression step size; In response to multicollinearity occurring in the target weighted polynomial regression and the data volatility of the arrester operating parameter not satisfying a first volatility condition, increasing a reference value of the ridge regression parameter based on a second ridge regression step size; In response to the data volatility of the arrester operating parameters satisfying a first volatility condition and multicollinearity occurring in the target weighted polynomial regression, increasing the reference value of the lasso regression parameter based on a first lasso regression step size, and increasing the reference value of the ridge regression parameter based on a third ridge regression step size; The sum of the first ridge regression step length and the second ridge regression step length is smaller than the third ridge regression step length.
5. The arrester performance detection method according to claim 4, characterized in that: In response to a variance inflation factor of the target weighted polynomial regression being greater than or equal to a first inflation threshold, multicollinearity occurs in the target weighted polynomial regression; In response to the variance inflation factor of the target weighted polynomial regression being less than the first inflation threshold, multicollinearity does not occur in the target weighted polynomial regression; In response to the coefficient of variation of the leakage current data in the arrester operating parameter being greater than or equal to a first variation threshold, the data volatility of the arrester operating parameter satisfies a first volatility condition; In response to the coefficient of variation of the leakage current data in the arrester operating parameters being less than a first variation threshold, the data volatility of the arrester operating parameters does not satisfy a first volatility condition.
6. The lightning arrester performance detection method according to claim 1, characterized in that: Determining the weight of each arrester operating parameter based on the time of each arrester operating parameter includes: Determine the weight corresponding to each arrester operating parameter; In response to the absolute value of the difference between the acquisition time of the arrester operation parameter and the current time being greater than the first duration, reducing the weight corresponding to the arrester operation parameter based on the first duration step; In response to the absolute value of the difference between the acquisition time of the arrester operating parameter and the current time being less than or equal to the first duration and greater than the second duration, reducing the weight corresponding to the arrester operating parameter based on the second duration step; The sum of the weights corresponding to the operating parameters of the arresters is one.
7. The arrester performance detection method according to claim 1, characterized in that: Before obtaining the importance level of each parameter in the arrester operating parameters based on the random forest algorithm, the method further includes: In response to the sample size of the arrester operation data being less than the first sample size, reducing the reference value of the number of decision trees by the first number of decision trees; In response to the sample size of the arrester operation data being greater than the second sample size, increasing the reference value of the number of decision trees by the second number of decision trees; In response to the discrete value of the arrester operation data being greater than the first discrete value, the reference value of the minimum sample number of the decision tree is increased by the first sample number, and the reference value of the minimum sample number of the leaf node of the decision tree is increased by the second sample number.
8. A lightning arrester performance detection device, characterized in that: include: A first weight determination module, used for obtaining the importance level of each parameter in the arrester operation parameters based on a random forest algorithm, and determining the weight of each arrester operation parameter based on the time of each arrester operation parameter; A second weight determination module, used to determine the weight of each parameter in each arrester operating parameter based on the importance level of each parameter in the arrester operating parameter and the weight of each arrester operating parameter; A weighted polynomial determination module, used to use the weight of each parameter in the said arrester operation parameters as the weight distribution of the weighted polynomial regression; The performance detection module is used to detect the performance of the arrester based on the weighted polynomial regression to obtain a performance detection result.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.