Cable state evaluation method based on Bayesian optimization and XGBoost

Abstract

The invention discloses a cable state evaluation method based on Bayesian optimization and XGBoost, and the method employs a cross entropy loss function on the basis of an XGBoost algorithm, and can effectively reduce the loss. And hyper-parameters in the XGBoost algorithm are adjusted and optimized by using a Bayesian parameter optimization method, so that key features in the operation process of the cable are extracted more effectively, and finally, the state of the cable is evaluated. By using the cross entropy loss function, the complexity of the cable state can be better adapted, and the robustness of the model is improved. Meanwhile, the Bayesian hyper-parameter optimization is beneficial to finding the optimal model parameter configuration, so that the model can capture the cable state characteristics more accurately.

Description

Technical Field

[0001] The present invention belongs to the field of online monitoring of power cables, and in particular relates to a cable status evaluation method based on Bayesian optimization and XGBoost. Background Art

[0002] With the advancement of modernization, the number of urban cable lines has increased rapidly, forming an intricate cable network. Although this growth provides users with a more convenient electricity experience, due to the complex operating environment, changes in working conditions and the influence of external factors, cable lines often have defects or failures, posing a potential threat to the power system. At present, the State Grid Corporation of China is actively promoting the maintenance of power transmission and transformation equipment to enhance the monitoring and analysis of the grid operation status, so that equipment problems can be discovered and repaired in a timely manner. Therefore, the State Grid Corporation of China has put forward higher standards for cable multi-parameter monitoring and comprehensive evaluation requirements.

[0003] The power cable online monitoring system has been widely used in high-voltage cables of 110kV and above, and monitors cable temperature, ground loop current, partial discharge and other data in real time. At the same time, through daily maintenance, regular online testing and offline testing, the cable has accumulated a large amount of data, providing reliable support for the comprehensive evaluation of the cable status. Therefore, comprehensive analysis of various parameters of the cable operation status, extraction of core indicators, and construction of a multi-dimensional risk assessment model have far-reaching research and application value for the power grid.

[0004] At present, through sufficient experiments and research, it is concluded that the cable will experience aging over time, which is mainly reflected in the combined effect of external and internal factors on the insulation material inside the cable, resulting in a decrease in insulation performance, which in turn affects the operating status of the cable. In the field of cable status assessment, domestic and foreign scholars have conducted in-depth and extensive research on this phenomenon, focusing mainly on cross-linked polyethylene materials in medium and high voltage cables. These studies not only help to understand the cable aging mechanism more comprehensively, but also provide key information for the assessment and maintenance of cable status. The current cable status assessment methods are mainly divided into two categories: slicing method and mathematical modeling method. Although the slicing method is simple and intuitive, its application is limited by the actual cable operation and maintenance needs. On the other hand, although the method of establishing a mathematical model is relatively rich, it is relatively difficult to determine the key characteristic indicators due to the limited actual cable fault samples and complex influencing factors. Based on the polarization-depolarization current (PDC) detection and diagnosis method, the researchers improved the PDC method, and improved the extracted single status assessment feature quantity into multiple status assessment feature quantities with different weights, thereby improving the accuracy of cable status assessment. Later, through simulation tests on eleven typical cable defects, a comprehensive evaluation method for high-voltage cable status based on multi-state quantity characteristics and change laws was proposed, and its effectiveness was verified through examples. After that, the remaining life of 110kV XLPE cable was evaluated using the step-by-step withstand voltage method and isothermal relaxation current method, providing a basis for scientific guidance of cable maintenance and replacement.

[0005] On the other hand, since fault states are relatively rare in practical applications, only a limited amount of fault data is available, so there is an imbalance between classes in cable sample data. This challenge may cause the model to fail to learn enough when facing a faulty cable, reducing the system's ability to identify and respond to abnormal situations. There are two main methods to solve the sample imbalance problem: (1) data-level methods, which use various sampling techniques to rebalance the sample distribution of the training set to alleviate the imbalance; (2) classifier-level methods, which improve the existing classification algorithm, including the improvement of the model, training method, and loss function, so that the classifier pays more attention to minority class data. The current technology introduces class label weights in the loss function, emphasizes the influence of minority classes on model parameters, and uses cost-sensitive learning methods to alleviate data imbalance in text classification. It can be seen that although the existing technology involves the analysis and evaluation method of cable multi-state feature quantities and the problem of small sample data and imbalance between classes, it lacks an in-depth discussion of combining the two. Therefore, in the face of imbalanced cable sample data, there is an urgent need for a cable state evaluation method that can effectively combine Bayesian parameter optimization with XGBoost. Summary of the invention

[0006] In order to address the deficiencies in the prior art, the present invention aims to provide a cable state assessment method based on Bayesian optimization and XGBoost, which solves the problem of low accuracy of cable state assessment caused by imbalance between sample classes in multi-classification problems, can more effectively extract key features in the cable operation process, can better adapt to the complexity of the cable state, improve the robustness of the model, and enable the model to more accurately capture the cable state characteristics.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is:

[0008] A cable state assessment method based on Bayesian optimization and XGBoost specifically includes the following steps:

[0009] 1) performing data processing operations on the collected cable data set to obtain a processed data set;

[0010] 2) Determine the range of XGBoost model hyperparameter values;

[0011] 3) Select the cross entropy loss function as the evaluation indicator of Bayesian optimization;

[0012] 4) Select Gaussian process as the surrogate model;

[0013] 5) Select test points based on the surrogate model and evaluate the objective function at the test points;

[0014] 6) Update the parameters of the proxy model and record the best hyperparameters for each iteration;

[0015] 7) Continue iterating until a certain number of iterations is reached or the performance is no longer significantly improved, and finally the optimal hyperparameter configuration will be obtained;

[0016] 8) Substitute the optimal hyperparameter configuration into the original XGBoost model to obtain the Bayesian optimized XGBoost model;

[0017] 9) Use the Bayesian optimized XGBoost model to predict the test data, obtain the prediction results, compare the prediction results with the true labels, and calculate the evaluation indicators to comprehensively evaluate the accuracy and reliability of the model;

[0018] 10) Based on the evaluation indicators and business needs, interpret and analyze the output of the model to draw evaluation conclusions on the cable status.

[0019] The data processing of the collected cable data set D in step 1) includes missing value filling, outlier processing and data smoothing operations.

[0020] Step 2) specifically includes: selecting the objective function f as XGBoost, defining the hyperparameter space χ of the XGBoost model including the range or optional value of each hyperparameter; selecting the tree depth max_depth, learning rate learning_rate, subsampling rate subsample, and parameter gamma for controlling tree growth as optional parameters of the XGBoost model, namely:

[0021] χ={max_depth,learning_rate,subsample,gamma} (1)

[0022] And set the value range of each hyperparameter: select max_depth between 3-15, learning_rate between 0.0005-0.1, subsample between 0.5-0.98, and gamma between 0.00001-0.5.

[0023] Step 3) specifically includes: The cross entropy loss function is used to measure the difference between the predicted probability distribution of the model and the real data distribution, and is used to optimize the parameters of the model in the classification task of supervised learning, namely:

[0024]

[0025] In multi-classification problems, p represents the true category of the sample; q represents the probability distribution predicted by the model, where each element represents the probability that the model predicts that the sample belongs to the corresponding category;

[0026] Determine the evaluation metric used to measure the performance of the XGBoost model, and select the cross entropy loss function as the evaluation metric EvalMetric for Bayesian optimization, namely:

[0027]

[0028] Step 4) specifically includes: selecting a proxy model M, which establishes a probability model based on known hyperparameter configurations and corresponding performance indicator values ​​to predict the performance of unknown hyperparameter configurations;

[0029] Let Gaussian process G be the proxy model M, that is:

[0030] M=G=GP[m(x),k(x,x′)] (4)

[0031] Here, m(x) is the mean function, which defines the expectation of the random variable at x; k(x,x′) is the covariance function, which defines the covariance between the random variables at x and x′.

[0032] Step 5) specifically includes: Bayesian hyperparameters include defining the proxy model, selecting the Bayesian optimization algorithm, finding the best hyperparameter configuration, and updating the proxy model;

[0033] The Optuna hyperparameter optimization library is selected as the Bayesian optimization library, and then the iteration to find the best hyperparameter configuration begins; in each iteration, the optimization algorithm selects the next hyperparameter configuration based on the prediction of the surrogate model and evaluates its performance on the objective function.

[0034] The present invention has the following beneficial effects:

[0035] In the face of unbalanced cable sample data, the present invention adopts the cross entropy loss function based on the XGBoost algorithm, which can significantly reduce the loss. In addition, the Bayesian parameter optimization method is used to tune the hyperparameters in the XGBoost algorithm, so as to more effectively extract the key features in the cable operation process and finally realize the evaluation of the cable status. Through the use of the cross entropy loss function, it is possible to better adapt to the complexity of the cable status and improve the robustness of the model. At the same time, Bayesian hyperparameter optimization helps to find the optimal model parameter configuration, so that the model can capture the cable status characteristics more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a specific flow chart of Bayesian optimization of XGBoost in the present invention;

[0037] Figure 2 This is a diagram of the Bayesian hyperparameter tuning results of the present invention;

[0038] Figure 3 This is a comparison chart of the accuracy of different algorithms of the present invention;

[0039] Figure 4 This is a comparison chart of the recall rates of different algorithms in the present invention. DETAILED DESCRIPTION

[0040] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not intended to limit the scope of use of the present invention.

[0041] The present invention proposes a cable status assessment method based on Bayesian optimization and XGBoost, which can significantly improve the accuracy of the assessment and effectively solve the overfitting problem that traditional classification models may face. The method is implemented based on the following steps:

[0042] 1) Perform a series of data processing operations on the collected cable data set D, including missing value filling, outlier processing and data smoothing operations, to ensure the integrity and quality of the data, and obtain the processed data set D′.

[0043] Then, the cable status evaluation method based on Bayesian optimization and XGBoost mentioned in the present invention is used to evaluate the labeled data set D train Through this process, the model will learn and establish the relationship between the cable status and various features. The Bayesian optimization XGBoost process mentioned in the present invention is as follows: Figure 1 As shown. Specifically including the following steps:

[0044] 2) Determine the range of XGBoost model hyperparameter values:

[0045] Select the objective function f as XGBoost, and define the hyperparameter space χ of the XGBoost model including the range or optional value of each hyperparameter.

[0046] The present invention selects four hyperparameters, namely, tree depth (max_depth), learning rate (learning_rate), subsampling rate (subsample), and parameter (gamma) for controlling tree growth as optional parameters of the XGBoost model, namely:

[0047] χ={max_depth, learning_rate, subsample, gamma} (1).

[0048] And set the value range of each hyperparameter: max_depth is selected to vary between 3-15, learning_rate is selected to vary between 0.0005-0.1, subsample is selected to vary between 0.5-0.98, and gamma is selected to vary between 0.00001-0.5. Determining these hyperparameter spaces is the basis for the subsequent use of the Bayesian optimization algorithm.

[0049] 3) Select the cross entropy loss function as the evaluation indicator for Bayesian optimization:

[0050] The cross entropy loss function is a commonly used loss function in classification problems. It is used to measure the difference between the predicted probability distribution of the model and the true data distribution. It is usually used to optimize the parameters of the model in the classification task of supervised learning, namely:

[0051]

[0052] p represents the true probability distribution, and in multi-classification problems, p represents the true category of the sample; q represents the probability distribution predicted by the model, each element of which represents the probability that the model predicts that the sample belongs to the corresponding category.

[0053] Cross entropy has two important properties:

[0054] (1) Non-negativity, H(p,q) ≥ 0. This property makes cross entropy an effective indicator for measuring the difference between two probability distributions. (2) Convex function property. When p1 and p2 are two different probability distributions, H(p1,q)+H(p2,q) ≥ H(p1*p2,q), which ensures that the local minimum is also the global minimum, thereby simplifying the optimization process.

[0055] Determine the evaluation index to measure the performance of the XGBoost model. The present invention uses the cross entropy loss function as the evaluation index EvalMetric of Bayesian optimization, that is:

[0056]

[0057] 4) Choose Gaussian process as the surrogate model:

[0058] Select a proxy model M, which builds a probability model based on known hyperparameter configurations and corresponding performance indicator values ​​to predict the performance of unknown hyperparameter configurations;

[0059] The present invention selects Gaussian process G as the proxy model M, that is:

[0060] M=G=GP[m(x),k(x,x′)] (4)

[0061] Among them, m(x) is the mean function, which defines the expectation of the random variable at x; k(x,x′) is the covariance function, which defines the covariance between the random variables at x and x′. The covariance function is the core of the Gaussian process, which determines the spatial correlation of random variables.

[0062] 5) Select test points based on the surrogate model and evaluate the objective function at the test points:

[0063] Bayesian hyperparameters include defining the proxy model, selecting the Bayesian optimization algorithm, finding the best hyperparameter configuration, and updating the proxy model; the Bayesian optimization library usually includes multiple optimization algorithms for searching the hyperparameter space to find the best configuration.

[0064] The present invention uses the Optuna hyperparameter optimization library as the Bayesian optimization library, and then begins to iterate to find the best hyperparameter configuration. In each iteration, the optimization algorithm selects the next hyperparameter configuration based on the prediction of the proxy model and evaluates its performance on the objective function.

[0065] 6) As iterations proceed, the proxy model will be continuously updated to record the best hyperparameters for each round of iteration; this can better capture the relationship between hyperparameters and performance indicators, which helps the optimization algorithm to more intelligently select the next configuration.

[0066] 7) Continue iterating until a certain number of iterations is reached or the performance is no longer significantly improved, and the optimal hyperparameter configuration will be obtained.

[0067] 8) Substitute the optimal hyperparameter configuration into the original XGBoost model to obtain the Bayesian optimized XGBoost model f.

[0068] 9) Use the Bayesian optimized XGBoost model f to test the data D test Make a prediction, get the predicted result y′, compare the predicted result with the true label y, and calculate the evaluation index. The purpose of this evaluation stage is to understand the generalization performance of the model, that is, the performance of the model on unseen data. Through this process, the accuracy and reliability of the model can be comprehensively evaluated.

[0069] 10) Finally, based on the evaluation indicators and business needs, the output of the model is interpreted and analyzed to draw an evaluation conclusion on the cable status. This conclusion can reflect important information such as the health status and aging degree of the cable, and also provide an important reference for subsequent decision-making and maintenance.

[0070] Simulation experiment and analysis

[0071] Dataset selection

[0072] By collecting cable-related data, including cable operation time, whether the cable has experienced ground faults, cable partial discharge, cable dielectric loss, cable aging status and other parameters, and scoring according to cable evaluation standards, the cable aging degree is divided into three categories: mild aging, moderate aging and severe aging. Some data reflecting cable characteristics are shown in Table 1 below.

[0073] Table 1 Cable operation parameter data

[0074]

[0075] Analysis of Bayesian Optimization Experiment Results

[0076] Currently, in the field of hyperparameter tuning methods, the Bayesian hyperparameter tuning method is relatively excellent in terms of time cost and accuracy. It mainly establishes a proxy model (usually a Gaussian process) in the search space and updates the proxy model according to the existing observations in each iteration, so as to more intelligently select the next hyperparameter to try. This paper conducts a Bayesian hyperparameter tuning experiment on the XGBoost algorithm, setting the number of iterations to 30 rounds, and the results are as follows: Figure 2 shown.

[0077] By observing the chart, we can find that the accuracy rate keeps changing at different iteration numbers. After 30 iterations, the XGBoost classification method with Bayesian hyperparameter tuning achieves a maximum accuracy rate of 88.29%. It is worth noting that in the process of achieving this highest accuracy rate, the best number of iterations is the 13th, and the output of the best hyperparameter configuration is:

[0078]

[0079] Through Bayesian hyperparameter tuning, the model achieved significant performance improvements in a relatively short period of time, which shows that the Bayesian optimization algorithm can quickly and accurately find the optimal configuration of model parameters for the XGBoost classification algorithm.

[0080] Cable status assessment results analysis

[0081] In order to more comprehensively demonstrate the effectiveness and superiority of the improved algorithm of the present invention, the gradient boosting decision tree (GBDT), the classification algorithm CatBoost based on the gradient boosting decision tree, LightGBM, the adaptive boosting algorithm AdaBoost, the emerging model Tabnet combining neural network and decision tree, and the XGBoost algorithm before optimization were selected as the benchmark algorithms for the comparative experiment, and a comparative analysis was carried out in terms of accuracy and recall rate.

[0082] Comparative analysis of accuracy

[0083] In the cable data multi-classification problem, the accuracy of the evaluation model refers to the proportion of samples correctly classified by the model in all cable data categories to the total number of samples. The accuracy is usually expressed as a percentage, and the calculation formula is as follows:

[0084]

[0085] Among them, TP (True Positive) is a true positive example, FP (False Positive) is a false positive example, TN (True Negative) is a true negative example, and FN (False Negative) is a false negative example. This formula takes into account the contribution of each category and calculates the overall accuracy of the model in each category. The closer the accuracy value is to 1, the better the model performance is, and the closer the value is to 0, the worse the performance is. Accuracy is one of the most intuitive model performance metrics at present. It provides a summary of the model's performance on the data set and can quickly understand the performance of the model. The present invention also uses accuracy as a criterion for each classification algorithm model to evaluate the cable status. The specific performance is shown in Table 2.

[0086] Table 2 Comparison of accuracy of each classification algorithm model

[0087] Algorithm Name Model Accuracy CatBoost 76.67％ GBDT 78.33％ TabNet 83.33％ AdaBoost 73.33％ LightGBM 80.00％ XGBoost 81.67％ Method of the present invention 88.29％

[0088] Through the comparison in the above table, it can be clearly seen that the accuracy of the XGBoost algorithm after Bayesian parameter optimization has been significantly improved compared to before optimization. Although other classification algorithms CatBoost, GBDT, TabNet, AdaBoost and LightGBM have excellent performance in their respective fields, in this comparison, the method of the present invention is significantly ahead of all other algorithms with an accuracy rate of 88.29%, fully demonstrating its advantages in model prediction performance. In particular, the GBDT algorithm, although it may be faster in calculation speed, its accuracy rate of 78.33% is significantly lower than that of the method of the present invention, showing its shortcomings in prediction accuracy. In order to more intuitively show the relationship between the accuracy rates of each classification algorithm, the following plot is drawn. Figure 3 The bar graph shown.

[0089] pass Figure 3 It can be observed that there are certain differences in the accuracy of different algorithms. GBDT, CatBoost and LightGBM algorithms are ensemble learning methods based on decision trees, and they perform poorly for data with weak linear relationships. Cable status prediction is affected by many factors, among which there may be nonlinear relationships or interactions, which makes it difficult for these algorithms to accurately capture the complexity of the data. In addition, cable status prediction is affected by data quality. If there is noise data or outliers, it will affect the prediction effect of the AdaBoost algorithm. The TabNet algorithm is mainly used to process tabular data. The prediction of cable aging may involve time series data or continuity features, which may limit the performance of the TabNet algorithm on this problem.

[0090] Comparative analysis of recall rate

[0091] Recall is one of the indicators used to evaluate the performance of classification models. It measures the ratio between the number of positive samples identified by the model and the number of actual positive samples. Recall represents the proportion of samples correctly identified as positive samples among all samples that are actually positive samples. The specific calculation formula is as follows:

[0092]

[0093] Among them, TP (True Positive) is a true positive example, and FN (False Negative) is a false negative example. The higher the recall rate, the higher the proportion of correctly predicted samples among all truly positive examples, that is, the lower the false negative rate. The present invention uses the recall rate as a criterion for evaluating the cable status of each classification algorithm model, and the specific performance is shown in Table 3.

[0094] Table 3 Comparison of recall rates of various classification algorithm models

[0095] Algorithm Name Model recall CatBoost 82.61％ GBDT 79.93％ TabNet 83.33％ AdaBoost 80.94％ LightGBM 85.62％ XGBoost 82.27％ Method of the present invention 93.98％

[0096] Judging from the recall rate results, the method of the present invention performs well and has a high positive example recognition ability. Compared with the traditional XGBoost algorithm, the method of the present invention has a significant improvement in recall rate, from 82.27% to 93.98%. This improvement not only highlights the effectiveness of Bayesian optimization in hyperparameter tuning, but also shows that the optimized XGBoost has a stronger ability in identifying positive examples. Compared with other ensemble learning algorithms such as CatBoost, GBDT, TabNet, AdaBoost and LightGBM, these algorithms have been widely used and recognized in their respective fields, but when processing cable status data, their recall rates have failed to reach the height of the method of the present invention, which further proves the superior performance of the method of the present invention when processing complex data sets. In order to more intuitively reveal the differences in recall rates among different classification algorithms, such as Figure 4 The bar chart shown is used for visual display.

[0097] It can be seen that the method of the present invention shows relatively excellent results in cable status assessment. This advantage is mainly reflected in the fact that the XGBoost algorithm after Bayesian optimization is more robust when processing sample data with imbalance between classes, effectively alleviating the overfitting problem. Compared with other algorithms, this method has achieved more accurate results in cable status assessment, further confirming the effectiveness of the method proposed in the present invention in the problem of cable status assessment.

[0098] The present invention proposes a new method to solve the problem of imbalance between classes of cable sample data, that is, to optimize the XGBoost classification algorithm by using Bayesian hyperparameter optimization, so as to more accurately evaluate the cable status. This method effectively solves the overfitting problem that traditional classification models may face. Through example verification, compared with the traditional cable status feature evaluation method, the method proposed in the present invention has achieved significant improvements in accuracy and effectiveness. This advantage is mainly attributed to the use of Bayesian hyperparameters to optimize the XGBoost classification algorithm, which makes the model better adapt to the problem of class imbalance in cable sample data and improves the ability to accurately evaluate the cable aging status.

[0099] The present invention has made important progress in solving the problem of imbalance between classes in cable aging status assessment, and provides a more accurate and reliable assessment method for the field of power cable health status monitoring. It has broad application prospects in power grid operation and maintenance, and provides strong support for improving the reliability and safety of cable systems.

[0100] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A cable status assessment method based on Bayesian optimization and XGBoost, characterized in that: The specific steps include: 1) performing data processing operations on the collected cable data set to obtain a processed data set; 2) Determine the range of XGBoost model hyperparameter values; 3) Select the cross entropy loss function as the evaluation indicator of Bayesian optimization; 4) Select Gaussian process as the surrogate model; 5) Select test points based on the surrogate model and evaluate the objective function at the test points; 6) Update the parameters of the proxy model and record the best hyperparameters for each iteration; 7) Continue iterating until a certain number of iterations is reached or the performance is no longer significantly improved, and finally the optimal hyperparameter configuration will be obtained; 8) Substitute the optimal hyperparameter configuration into the original XGBoost model to obtain the Bayesian optimized XGBoost model; 9) Use the Bayesian optimized XGBoost model to predict the test data, obtain the prediction results, compare the prediction results with the true labels, and calculate the evaluation indicators to comprehensively evaluate the accuracy and reliability of the model; 10) Based on the evaluation indicators and business needs, interpret and analyze the output of the model to draw evaluation conclusions on the cable status.

2. The cable status assessment method based on Bayesian optimization and XGBoost according to claim 1, characterized in that: The data processing of the collected cable data set D in step 1) includes missing value filling, outlier processing and data smoothing operations.

3. The cable status assessment method based on Bayesian optimization and XGBoost according to claim 1, characterized in that: Step 2) specifically includes: Select the objective function f as XGBoost, define the hyperparameter space χ of the XGBoost model, including the range or optional value of each hyperparameter; select the tree depth max_depth, learning rate learning_rate, subsample rate subsample, and parameter gamma that controls tree growth as optional parameters of the XGBoost model, namely: χ={max_depth,learning_rate,subsample,gamma} (1) And set the value range of each hyperparameter: select max_depth between 3-15, learning_rate between 0.0005-0.1, subsample between 0.5-0.98, and gamma between 0.00001-0.

5.

4. The cable status assessment method based on Bayesian optimization and XGBoost according to claim 3 is characterized in that: Step 3) specifically includes: The cross entropy loss function is used to measure the difference between the predicted probability distribution of the model and the true data distribution. It is used to optimize the parameters of the model in the classification task of supervised learning, namely: In multi-classification problems, p represents the true category of the sample; q represents the probability distribution predicted by the model, where each element represents the probability that the model predicts that the sample belongs to the corresponding category; Determine the evaluation metric used to measure the performance of the XGBoost model, and select the cross entropy loss function as the evaluation metric EvalMetric for Bayesian optimization, namely:

5. The cable status assessment method based on Bayesian optimization and XGBoost according to claim 4 is characterized in that: Step 4) specifically includes: Select a proxy model M, which builds a probability model based on known hyperparameter configurations and corresponding performance indicator values ​​to predict the performance of unknown hyperparameter configurations; Let Gaussian process G be the proxy model M, that is: M=G=GP[m(x),k(x,x′)] (4) Here, m(x) is the mean function, which defines the expectation of the random variable at x; k(x,x′) is the covariance function, which defines the covariance between the random variables at x and x′.

6. The cable status assessment method based on Bayesian optimization and XGBoost according to claim 5, characterized in that: Step 5) specifically includes: Bayesian hyperparameters include defining the surrogate model, selecting the Bayesian optimization algorithm, finding the best hyperparameter configuration, and updating the surrogate model; The Optuna hyperparameter optimization library is selected as the Bayesian optimization library, and then the iteration to find the best hyperparameter configuration begins; in each iteration, the optimization algorithm selects the next hyperparameter configuration based on the prediction of the surrogate model and evaluates its performance on the objective function.

Images