Cable insulation state evaluation method and device, electronic equipment and storage medium

Abstract

The invention provides a cable insulation state evaluation method and device, electronic equipment and a storage medium, and belongs to the technical field of high-voltage cable detection, and the method comprises the steps: obtaining a cable characteristic parameter and a corresponding cable insulation state; arranging and combining the cable characteristic parameters to obtain a plurality of characteristic subsets, and pre-training the plurality of characteristic subsets based on a native lightweight gradient elevator model to obtain first cable characteristic parameters; constructing a cable insulation state evaluation model, optimizing hyper-parameters of the cable insulation state evaluation model based on a Bayesian optimization algorithm, and training the cable insulation state evaluation model based on the first cable characteristic parameters and the corresponding first cable insulation state to obtain a completely trained cable insulation state evaluation model; and evaluating the characteristic parameters of the to-be-evaluated cable based on the completely trained cable insulation state evaluation model. According to the invention, the evaluation precision of the cable insulation state is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of high-voltage cable detection, and particularly to a method, device, electronic device and storage medium for evaluating the insulation state of a cable. Background Art

[0002] Crosslinked polyethylene (XLPE) cables have become important power transmission equipment in the power system due to their excellent mechanical strength, heat resistance and insulation characteristics. However, due to the long laying line and complex operating environment of XLPE cables, with the increase of the operating years, the insulation of the cables will gradually age and the remaining life will gradually decrease. At present, many high-voltage cables in the power system have been in operation for many years, and many cables have reached the end of their aging life. Therefore, carrying out the evaluation of the cable insulation state is of great significance for ensuring power supply safety.

[0003] In the field of cable insulation state evaluation, traditional methods are divided into two types. One is to measure the cable insulation parameters, for example, obtaining a single characteristic parameter of an aging cable in the laboratory and using this characteristic parameter to evaluate the cable. The other is to collect multiple cable aging characteristic parameters and establish a cable insulation aging state evaluation system using various mathematical tools. For the first method, Patent CN114325260A discloses an insulation state evaluation method for high-voltage cables based on thermal resistance change. This method obtains the thermal resistivity of the target high-voltage cable before and after a preset aging time by establishing an equivalent cable thermal circuit model and a heat balance equation, and evaluates the insulation state of the target high-voltage cable according to the change degree of the thermal resistivity before and after the preset aging time. For the second method, Patent CN115932468A discloses an XLPE cable operation state evaluation method based on macroscopic parameters. This method detects three macroscopic parameters of the cable: insulation resistance, dielectric loss tangent, and partial discharge amount, combines and weights the three parameters, and calculates the overall aging parameter to evaluate the cable aging degree. In recent years, methods based on artificial intelligence and machine learning have also been widely applied in cable insulation state evaluation. Patent CN117171642A discloses a distribution network cable insulation state evaluation method based on a long short-term memory network. An insulation state evaluation model for distribution network cables based on a long short-term memory network is trained through weighted fusion characteristic parameters and the historical insulation state of the corresponding distribution network cables, so as to evaluate the insulation state of the cables. Patent CN113869772A discloses a cable insulation state evaluation method based on grid search optimization support vector machine. This method tunes the optimal penalty factor and Gaussian kernel parameter of the support vector machine through grid search, and then inputs the training set into the support vector machine to train the support vector machine to complete the evaluation of the cable insulation state.

[0004] While the aforementioned methods can effectively analyze the aging status of cable insulation, they also have certain technical drawbacks. For example, laboratory measurements of cable insulation parameters often assess the insulation status of a cable based on changes in a single feature or a specific category of features. This method fails to reflect the overall aging status of the cable and cannot quantitatively assess the performance. While mathematical tools can quantitatively determine the insulation status of a cable, the weighting of different characteristic parameters is highly subjective, making the resulting insulation status objective. Long short-term memory networks are a type of recurrent neural network in deep learning. Deep learning has extensive applications in image, audio, and natural language processing, but in the field of cable insulation assessment, datasets are often tabular. Numerous studies have shown that deep learning models often perform less effectively than machine learning algorithms based on tree models when processing tabular data. Support vector machines are currently the most commonly used linear classification algorithm in machine learning, but their performance is limited for nonlinear classification problems and requires dimensionality increase. Therefore, improving the accuracy and efficiency of cable insulation assessment remains a pressing technical challenge. Summary of the Invention

[0005] In view of this, it is necessary to provide a cable insulation status assessment method, device, electronic device and storage medium to solve the technical problems of low cable insulation status assessment accuracy and evaluation efficiency in the prior art.

[0006] In order to solve the above technical problems, in a first aspect, the present invention provides a method for evaluating the insulation status of a cable, comprising:

[0007] Acquiring cable operation data, wherein the cable operation data includes cable characteristic parameters and corresponding cable insulation status;

[0008] Arrange and combine the cable characteristic parameters to obtain multiple feature subsets, and pre-train the multiple feature subsets based on a native lightweight gradient boosting machine model to obtain first cable characteristic parameters;

[0009] Constructing a cable insulation state assessment model, optimizing hyperparameters of the cable insulation state assessment model based on a Bayesian optimization algorithm, and training the cable insulation state assessment model based on the first cable characteristic parameter and its corresponding first cable insulation state to obtain a fully trained cable insulation state assessment model;

[0010] The characteristic parameters of the cable to be evaluated are evaluated based on the fully trained cable insulation status evaluation model.

[0011] In a possible implementation, the cable characteristic parameters include cable operating environment parameters, cable operating years, cable partial discharge amount, and cable visual state conditions;

[0012] The cable insulation state includes good state, general state, signs of aging, relatively serious aging, and complete aging.

[0013] In a possible implementation, pre-training the multiple feature subsets based on the native lightweight gradient boosting machine model to obtain first feature parameters includes:

[0014] Pre-training the multiple feature subsets based on the native lightweight gradient boosting machine model to obtain the K-fold cross-validation accuracy rate of each first feature subset;

[0015] Determining the feature subset with the highest K-fold cross-validation accuracy rate as the first feature parameter.

[0016] In a possible implementation, pre-training the multiple feature subsets based on the native lightweight gradient boosting machine model includes:

[0017] Initializing the predicted values of each feature subset;

[0018] Iteratively calculating the probability of each feature subset in each category, and determining the predicted value of each feature subset based on the probability;

[0019] Determining the negative gradient of each feature subset based on the predicted value and the true value of each feature subset, where the negative gradient is the target that the new tree needs to fit in each iteration process;

[0020] Determining the leaf node values after splitting of the leaf nodes based on the negative gradient of each feature subset, and optimizing the native lightweight gradient boosting machine model based on the leaf node values.

[0021] In a possible implementation, the target that the new tree needs to fit is: where y i,k is the true probability of the feature parameter x ij p k,m-1 (x ij ) is the predicted probability of the feature parameter x ij , i is the number of samples, j is the number of features of each sample, m is the number of iterations, and k is the number of categories;

[0022] Optimizing the native lightweight gradient boosting machine model based on the leaf node values includes: where F k(X) is the optimized lightweight gradient boosting machine model, η is the learning rate of the model, I is the index of the sample set of a certain leaf node, h is the number of the h-th leaf node on the decision tree, H is the maximum number of leaf nodes set on the decision tree, and γ h,k,m is the leaf node value after leaf node splitting, m is the number of iterations, and M is the maximum number of iterations set.

[0023] In a possible implementation, the hyperparameters of the cable insulation state evaluation model include the maximum tree depth, the number of leaf nodes, the number of iterations, the data sampling ratio, the learning rate, the coefficient of the L1 regularization term, and the coefficient of the L2 regularization term.

[0024] In a possible implementation, optimizing the hyperparameters of the cable insulation state evaluation model based on the Bayesian optimization algorithm includes:

[0025] Defining the objective function of Bayesian optimization, where the input of the Bayesian optimization objective function is the hyperparameters of the cable insulation state evaluation model, and the output of the Bayesian optimization objective function is the accuracy of the K-fold cross-validation of the lightweight gradient boosting machine with the selected features;

[0026] Performing Gaussian process processing on the objective function to obtain the posterior probability of the objective function;

[0027] According to the mean and covariance of the posterior probability, using the acquisition function to update the parameters of the objective function to obtain the optimal hyperparameters.

[0028] On the other hand, the present invention also provides a cable insulation state evaluation device, including:

[0029] A data acquisition module, configured to acquire cable operation data, where the cable operation data includes cable characteristic parameters and their corresponding cable insulation states;

[0030] A data pre-training module, configured to perform permutation and combination on the cable characteristic parameters to obtain a plurality of feature subsets, and pre-train the plurality of feature subsets based on the original lightweight gradient boosting machine model to obtain the first cable characteristic parameters;

[0031] A model training module, configured to build a cable insulation state evaluation model, optimize the hyperparameters of the cable insulation state evaluation model based on the Bayesian optimization algorithm, and train the cable insulation state evaluation model based on the first cable characteristic parameters and their corresponding first cable insulation states to obtain a trained and complete cable insulation state evaluation model;

[0032] A model evaluation module, configured to evaluate the cable characteristic parameters to be evaluated based on the trained and complete cable insulation state evaluation model.

[0033] In a second aspect, the present invention further provides an electronic device, including a memory and a processor, wherein,

[0034] the memory is used for storing a program;

[0035] the processor is coupled to the memory and is used for executing the program stored in the memory to implement the steps in the cable insulation state evaluation method in any of the above implementation manners.

[0036] In a third aspect, the present invention further provides a computer-readable storage medium for storing computer-readable programs or instructions, and when the programs or instructions are executed by a processor, the steps in the cable insulation state evaluation method in any of the above implementation manners can be implemented.

[0037] The beneficial effects of the present invention are as follows:

[0038] By introducing the lightweight gradient boosting machine algorithm based on gradient boosting trees into the field of cable insulation evaluation, the present invention performs permutations and combinations on cable characteristic parameters to obtain multiple feature subsets, then pre-trains the multiple feature subsets based on the native lightweight gradient boosting machine model to complete feature screening, and then uses the Bayesian optimization algorithm to optimize the hyperparameters of the cable insulation state evaluation model, and trains the cable insulation state evaluation model through the first characteristic parameter to obtain a trained complete cable insulation state evaluation model. By using the first characteristic parameter for model training, the redundancy between different features is reduced, the training effect of the model is improved, and the Bayesian optimization algorithm is used to optimize the hyperparameters of the cable insulation state evaluation model, further improving the evaluation accuracy and evaluation efficiency of the model. Description of the Drawings

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention, and those skilled in the art can obtain other drawings without creative efforts based on these drawings.

[0040] Figure 1 It is a schematic flowchart of an embodiment of the cable insulation state evaluation method provided by the present invention;

[0041] Figure 2 It is a schematic flowchart of a method for determining the first cable characteristic parameter provided by an embodiment of the present invention;

[0042] Figure 3 It is a schematic flowchart of a method for optimizing the hyperparameters of the cable insulation state evaluation model based on the Bayesian optimization algorithm provided by an embodiment of the present invention;

[0043] Figure 4 A schematic flowchart of pre-training the multiple feature subsets based on a native lightweight gradient boosting machine model provided by an embodiment of the present invention;

[0044] Figure 5 A schematic structural diagram of an embodiment of a cable insulation status evaluation device provided by the present invention;

[0045] Figure 6 A schematic structural diagram of an embodiment of an electronic device provided by the present invention. Detailed implementation manners

[0046] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present invention.

[0047] The "first" description involved in the embodiments of the present invention is only for descriptive purposes, and cannot be construed as indicating or implying its relative importance or implicitly indicating the quantity of the indicated technical features. Therefore, the technical features defined with "first" may explicitly or implicitly include at least one of such features.

[0048] Referring to "embodiment" herein means that a specific feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the present invention. The phrase appears in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein may be combined with other embodiments.

[0049] The present invention provides a cable insulation status evaluation method, device, electronic device, and storage medium, which will be described separately below.

[0050] Figure 1 A schematic flowchart of an embodiment of the cable insulation status evaluation method provided by the present invention, as Figure 1 shown, the cable insulation status evaluation method includes:

[0051] S101: Collect cable operation data, where the cable operation data includes cable characteristic parameters and their corresponding cable insulation status;

[0052] Among them, the cable characteristic parameters include cable operating environment parameters, cable operating years, cable partial discharge amount, and cable visual state conditions; the cable insulation state includes good state, general state, signs of aging, relatively serious aging, and complete aging.

[0053] Specifically, collect cable operation data, and divide the cable operation data into a training set and a test set. Assume the data set D = {(x ij , C i ),} (i = 1, 2, 3,..., n1, j = 0, 1, 2, 3..., n2). Among them, n1 is the number of samples in the data set, n2 is the number of features of each sample in the data set, C i represents the sample target class label, and x ij is n2 feature parameters.

[0054] S102: Perform permutations and combinations on the cable characteristic parameters to obtain multiple feature subsets, and pre-train the multiple feature subsets based on the original lightweight gradient boosting machine model to obtain the first cable characteristic parameters;

[0055] It can be understood that by collecting the historical operation data of the cable, using the characteristic parameters as the input, and the historical insulation state of the corresponding distribution network cable as the output, the insulation of the cable can be effectively evaluated through machine learning algorithms. However, the collected characteristic parameters will inevitably have redundancy. There is a so-called curse of dimensionality problem in machine learning. It shows that the more features in the training data set are not necessarily better. Too many redundant features or features irrelevant to the classification effect often lead to a decrease in the training effect.

[0056] Therefore, the present invention selects the method of feature combination plus pre-training to accurately and efficiently complete feature screening. By introducing the lightweight gradient boosting machine algorithm based on the gradient boosting tree into the field of cable insulation evaluation, using combined features, and traversing all feature subsets to complete feature screening, the redundancy between different features is reduced.

[0057] Specifically, perform permutations and combinations on the collected cable characteristic parameters. For example, if n cable characteristic parameters are collected in the data set, then 2^ n -1 kinds of feature subsets can be constructed. Then pre-train the multiple feature subsets based on the original lightweight gradient boosting machine model to obtain the first cable characteristic parameters;

[0058] S103: Construct a cable insulation state evaluation model, optimize the hyperparameters of the cable insulation state evaluation model based on the Bayesian optimization algorithm, and train the cable insulation state evaluation model based on the first cable characteristic parameters and their corresponding first cable insulation states to obtain a well-trained cable insulation state evaluation model;

[0059] It can be understood that by using the selected feature combination, i.e., the first cable feature parameter as the input and the first cable insulation state corresponding to the first cable feature parameter as the output, the constructed cable insulation state evaluation model is trained. Specifically, a data set is constructed based on the feature combination, 70% of the data set is used as the training set, and the hyperparameters optimized by Bayesian optimization are used to start the formal training of the model. After the training is completed, the remaining 30% of the data is used as the test set to analyze indicators such as accuracy to evaluate the training effect of the model, and a well-trained cable insulation state evaluation model is obtained.

[0060] S104: Evaluate the cable feature parameters to be evaluated based on the well-trained cable insulation state evaluation model.

[0061] In the present invention, by introducing the lightweight gradient boosting machine algorithm based on the gradient boosting tree into the field of cable insulation evaluation, the cable feature parameters are permuted and combined to obtain multiple feature subsets, and then the multiple feature subsets are pre-trained based on the original lightweight gradient boosting machine model to complete the feature screening. Then, the Bayesian optimization algorithm is used to optimize the hyperparameters of the cable insulation state evaluation model, and the cable insulation state evaluation model is trained by using the first feature parameter to obtain a well-trained cable insulation state evaluation model. By using the first feature parameter for model training, the redundancy between different features is reduced, the training effect of the model is improved, and the Bayesian optimization algorithm is used to optimize the hyperparameters of the cable insulation state evaluation model, further improving the evaluation accuracy and evaluation efficiency of the model.

[0062] In some embodiments of the present invention, before permuting and combining the cable feature parameters, it further includes: preprocessing the cable operation data.

[0063] It can be understood that different types of numerical data have different unit dimensions, such as ℃ (Celsius) of ambient temperature and pc (picocoulomb) of partial discharge. Therefore, Min-Max normalization can be used to reduce the impact on the diagnostic accuracy caused by the dimensional difference between different feature parameters:

[0064]

[0065] In the formula, Y is the original feature parameter value, Y max is the maximum value observed in the feature column, and Y min is the minimum value observed in the feature column.

[0066] For ordered categorical data, such as when the input features are ranked as low, medium, high, etc., the following formula is used for preprocessing:

[0067]

[0068] Where x is the total number of N scores, and i is the order of the scores. For example, for an excellent - good - poor rating system, the poor rating can be converted to 0.17, the good rating can be converted to 0.5, and the excellent rating can be converted to 0.83.

[0069] For unordered categorical data, such as whether a cable has failed during operation is marked in some datasets, it can be converted into 01, 10 according to one - hot encoding, and then trained to further improve the evaluation accuracy of the model.

[0070] In some embodiments of the present invention, as Figure 2 shown, step S102 includes:

[0071] S201: Pre - train multiple feature subsets based on the native lightweight gradient boosting machine model to obtain the K - fold cross - validation accuracy of each feature subset;

[0072] The basic idea of K - fold cross - validation is to divide the dataset into several parts, and then take turns using one part as the test set and the other parts as the training set. After multiple trainings and tests, the average result is finally taken.

[0073] S202: Determine the feature subset with the highest K - fold cross - validation accuracy as the first feature parameter.

[0074] It can be understood that by pre - training each feature subset through the native lightweight gradient boosting machine model, the cross - validation accuracy of different feature subsets can be output. Then, by analyzing the K - fold cross - validation accuracy of different feature subsets, the feature subset corresponding to the highest accuracy is found, and the feature quantity corresponding to this feature subset is used as the input for the final model training. Feature screening can be accurately and efficiently completed, further improving the evaluation accuracy of the cable insulation status evaluation model.

[0075] First, it should be noted that the gradient boosting tree algorithm is widely used in classification and regression tasks. It improves the prediction ability of the model by constructing a series of weak learners (decision trees). The core idea of the gradient boosting tree algorithm is to combine multiple weak learners into a strong learner. Each weak learner attempts to correct the errors of the previous learner, thereby gradually improving the performance of the overall model. The lightweight gradient boosting machine is a histogram - based gradient boosting algorithm. Based on the gradient unilateral sampling and feature bundling technology, it can accelerate the training process of traditional gradient boosting and reduce memory occupancy while achieving almost the same accuracy.

[0076] The lightweight gradient boosting machine can be summarized into the following four steps during training:

[0077] 1. Discretize continuous features into histograms to reduce the amount of calculation.

[0078] 2. Build a decision tree using a histogram-based algorithm and traverse the histogram intervals each time a split point is selected.

[0079] 3. Improve the accuracy of the model by gradually adding new decision trees to correct the errors of the previous tree.

[0080] 4. Parallelize the features and data to further accelerate the training process.

[0081] In some embodiments of the present invention, as Figure 3 shown, the pre-training of multiple feature subsets by the native lightweight gradient boosting machine model in step S201 includes:

[0082] S301: Initialize the predicted values of each feature subset;

[0083] For each feature subset, that is, initialize the predicted values of the feature parameters, including: F k,0 (x ij ) = 0.

[0084] S302: Iteratively calculate the probabilities of each feature subset for each category, and determine the predicted values of each feature subset based on the probabilities;

[0085] Iteratively calculate the predicted probabilities of the feature subset, that is, the feature parameters for each category: Among them, the number of iterations is m, and the total number of categories in a specific classification problem is K.

[0086] S303: Determine the negative gradient of each feature subset based on the predicted values and the true values of each feature subset, where the negative gradient is the target that the new tree needs to fit in each iteration process;

[0087] Calculate the difference between the predicted value and the true value to obtain the negative gradient of the feature parameter x ij : The negative gradient is the target that the new tree needs to fit in each iteration, y i,k is the true probability of the feature parameter x ij , p k,m-1 (x ij ) is the predicted probability of the feature parameter x ij , i is the number of samples, j is the number of features of each sample, m is the number of iterations, and k is the number of categories.

[0088] S304: Determine the leaf node values after the leaf nodes are split based on the negative gradients of each feature subset, and optimize the native lightweight gradient boosting machine model based on the leaf node values.

[0089] Among them, the leaf node values after the leaf nodes are split are: After the leaf node splits, the leaf node values determine how samples are partitioned into different leaf nodes, directly affecting the performance of the tree model, R h,k,m is the sample set on the leaf node, and h is the number of leaf nodes on the decision tree. Then, based on the leaf node values, the native lightweight gradient boosting machine model is optimized as follows: Obtain the final model:

[0090] F k (X) is the optimized lightweight gradient boosting machine model, η is the learning rate of the model, used to control the contribution of each tree to the final model, I is the index of the sample set of a certain leaf node, used to represent the attribution of the samples after splitting. h is the number of leaf nodes on the decision tree, and H is the maximum number of leaf nodes set on the decision tree, used to control the complexity of the tree: γ h,k,m is the leaf node value after the leaf node splits, m is the number of iterations, and M is the maximum number of iterations set, that is, the total number of decision trees constructed.

[0091] First of all, it should be noted that the Bayesian optimization algorithm is an algorithm for global optimization, especially suitable for black-box functions with high computational cost and no gradient information, and is often used for hyperparameter tuning in machine learning. The core idea of the Bayesian optimization algorithm is to use a surrogate model to approximate the objective function and optimize the acquisition function to select the next evaluation point. The Bayesian optimization algorithm predicts the behavior of the objective function by constructing a surrogate model and uses the acquisition function to guide the sampling process, so that the global optimal solution can be quickly approximated with a small number of samples. The main steps of the Bayesian optimization algorithm can be summarized as follows:

[0092] 1. Select the initial observation points to evaluate the objective function to complete the initialization.

[0093] 2. Use surrogate models such as Gaussian processes to model the objective function.

[0094] 3. Define and optimize the acquisition function. Using the output of the surrogate model, the acquisition function combines the predicted mean and uncertainty to calculate and select the next evaluation point.

[0095] 4. Calculate the value of the objective function at this point and record it, and update the surrogate model with the new data points.

[0096] 5. Repeat the above steps until the stopping condition is met.

[0097] Since there are 26 hyperparameters in the lightweight gradient boosting machine algorithm that are only used to control the model learning process, but not all hyperparameters have a significant impact on the classification performance of the final model. Therefore, the present invention selects 7 of them, covering aspects such as tree structure, learning rate, regularization, etc., ensuring sufficient optimization space for the model while avoiding a decrease in optimization efficiency due to too high a parameter dimension. In some embodiments of the present invention, the hyperparameters of the cable insulation state evaluation model include the maximum tree depth, the number of leaf nodes, the number of iterations, the data sampling ratio, the learning rate, the L1 regularization term coefficient, and the L2 regularization term coefficient. As Figure 4 shown, optimizing the hyperparameters of the cable insulation state evaluation model based on the Bayesian optimization algorithm in step S103 includes:

[0098] S401: Define the objective function of Bayesian optimization;

[0099] It can be understood that the maximum tree depth, the number of leaf nodes, the number of iterations, the data sampling ratio, the learning rate, the L1 regularization term coefficient, and the L2 regularization term coefficient are used as the input xi of the Bayesian optimization objective function, and the accuracy of the K-fold cross-validation of the feature-selection lightweight gradient boosting machine is used as the output yi of the objective function. The search ranges of the maximum tree depth, the number of leaf nodes, the number of iterations, the data sampling ratio, the learning rate, the L1 regularization term coefficient, and the L2 regularization term coefficient are [3, 30], [10, 100], [100, 1000], [0.8, 1], [0.01, 0.3], [0, 1], [0, 1].

[0100] S402: Perform Gaussian process processing on the objective function to obtain the posterior probability of the objective function;

[0101] Use models such as Gaussian process to model the objective function. Among them, in Gaussian process, it is assumed that the joint distribution of these data points is a multivariate normal distribution: y ∼ N(m, K), where m is the mean vector and K is the covariance matrix. After establishing prior knowledge using the probability model, the prior knowledge is updated by observing the output of the objective function at different input points to form a posterior distribution. Based on the posterior distribution, the next sampling point is selected.

[0102] S403: According to the mean and covariance of the posterior probability, use the acquisition function to update the parameters of the objective function to obtain the optimal hyperparameters.

[0103] Among them, common acquisition functions include expected improvement, probability improvement, and upper confidence bound. Taking expected improvement as an example, its definition is: α EI (x) = E[max(0, f(x) - f(x + ))], where f(x+) is the current optimal value and E represents expectation. By optimizing the acquisition function α(x), the next evaluation point x next: x next = argmaxα(x), and finally evaluate the objective function f(x next ), and record the new data point (x next , f(x next )), then update the Gaussian process model, and repeat the above steps 50 times. Find the optimal generation of K-fold cross-validation during 50 Bayesian optimization iterations. The values of the maximum tree depth, the number of leaf nodes, the number of iterations, the data sampling ratio, the learning rate, the coefficient of the L1 regularization term, and the coefficient of the L2 regularization term are the required results.

[0104] To better implement the cable insulation state evaluation method in the embodiments of the present invention, correspondingly, based on the cable insulation state evaluation method, as Figure 5 shown, the embodiments of the present invention also provide a cable insulation state evaluation device. The cable insulation state evaluation device 500 includes:

[0105] A data acquisition module 501, configured to acquire cable operation data, where the cable operation data includes cable characteristic parameters and their corresponding cable insulation states;

[0106] A data pre-training module 502, configured to perform permutations and combinations on the cable characteristic parameters to obtain multiple feature subsets, and pre-train the multiple feature subsets based on the native lightweight gradient boosting machine model to obtain the first cable characteristic parameters;

[0107] A model training module 503, configured to construct a cable insulation state evaluation model, optimize the hyperparameters of the cable insulation state evaluation model based on the Bayesian optimization algorithm, and train the cable insulation state evaluation model based on the first cable characteristic parameters and their corresponding first cable insulation states to obtain a trained complete cable insulation state evaluation model;

[0108] A model evaluation module 504, configured to evaluate the cable characteristic parameters to be evaluated based on the trained complete cable insulation state evaluation model.

[0109] The cable insulation state evaluation device 500 provided in the above embodiments can implement the technical solutions described in the embodiments of the above cable insulation state evaluation method. The specific implementation principles of the above modules or units can be referred to the corresponding content in the embodiments of the above cable insulation state evaluation method, which will not be elaborated here.

[0110] As Figure 6 shown, the present invention also correspondingly provides an electronic device 600. The electronic device 600 includes a processor 601, a memory 602, and a display 603. Figure 6 Only some components of the electronic device 600 are shown, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be alternatively implemented.

[0111] In some embodiments, the processor 601 may be a central processing unit (CPU), a microprocessor, or other data processing chips, which are used to run the program code stored in the memory 602 or process data, such as the cable insulation state evaluation method in the present invention.

[0112] In some embodiments, the processor 601 may be a single server or a server group. The server group may be centralized or distributed. In some embodiments, the processor 601 may be local or remote. In some embodiments, the processor 601 may be implemented on a cloud platform. In one embodiment, the cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-cloud, etc., or any combination of the above.

[0113] In some embodiments, the memory 602 may be an internal storage unit of the electronic device 600, such as a hard disk or a memory of the electronic device 600. In some other embodiments, the memory 602 may also be an external storage device of the electronic device 600, such as a plug-in hard disk equipped on the electronic device 600, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc.

[0114] Furthermore, the memory 602 may also include both the internal storage unit of the electronic device 600 and the external storage device. The memory 602 is used to store the application software installed in the electronic device 600 and various types of data.

[0115] In some embodiments, the display 603 may be an LED display, a liquid crystal display, a touch liquid crystal display, and an OLED (Organic Light-Emitting Diode) toucher, etc. The display 603 is used to display the information in the electronic device 600 and to display a visual user interface. The components 601-603 of the electronic device 600 communicate with each other through a system bus.

[0116] In one embodiment, when the processor 601 executes the cable insulation state evaluation program in the memory 602, the following steps may be implemented:

[0117] Obtain cable operation data, where the cable operation data includes cable characteristic parameters and their corresponding cable insulation states;

[0118] Perform permutations and combinations on the cable characteristic parameters to obtain a plurality of feature subsets, and pre-train the plurality of feature subsets based on the native lightweight gradient boosting machine model to obtain the first cable characteristic parameters;

[0119] Build a cable insulation status evaluation model, optimize the hyperparameters of the cable insulation status evaluation model based on the Bayesian optimization algorithm, and train the cable insulation status evaluation model based on the first cable characteristic parameters and their corresponding first cable insulation status to obtain a well-trained cable insulation status evaluation model;

[0120] Evaluate the cable characteristic parameters to be evaluated based on the well-trained cable insulation status evaluation model.

[0121] It should be understood that when the processor 601 executes the cable insulation status evaluation program in the memory 602, in addition to the above functions, other functions can also be realized. For specific details, reference can be made to the description of the corresponding method embodiments above.

[0122] Furthermore, the embodiments of the present invention do not specifically limit the type of the electronic device 600. The electronic device 600 can be a portable electronic device such as a mobile phone, a tablet computer, a personal digital assistant (PDA), a wearable device, a laptop computer, etc. Exemplary embodiments of the portable electronic device include, but are not limited to, portable electronic devices equipped with IOS, android, microsoft or other operating systems. The above portable electronic devices can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (such as a touch panel). It should also be understood that in some other embodiments of the present invention, the electronic device 600 may not be a portable electronic device, but a desktop computer with a touch-sensitive surface (such as a touch panel).

[0123] Correspondingly, the embodiments of the present application also provide a computer-readable storage medium. The computer-readable storage medium is used to store computer-readable programs or instructions. When the programs or instructions are executed by a processor, the steps or functions in the cable insulation status evaluation method provided by the above method embodiments can be realized.

[0124] Those skilled in the art can understand that all or part of the processes of implementing the above method embodiments can be completed by instructing relevant hardware (such as a processor, a controller, etc.) through a computer program. The computer program can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium is a magnetic disk, an optical disk, a read-only memory or a random access memory, etc.

[0125] The above has introduced in detail the method, device, electronic equipment and storage medium for evaluating the cable insulation state provided by the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for evaluating the insulation state of a cable, characterized in that, Including: Obtain cable operation data, where the cable operation data includes cable characteristic parameters and their corresponding cable insulation states; Arrange and combine the cable characteristic parameters to obtain multiple feature subsets, and pre-train the multiple feature subsets based on the original lightweight gradient boosting machine model to obtain the first cable characteristic parameters; Construct a cable insulation state evaluation model, optimize the hyperparameters of the cable insulation state evaluation model based on the Bayesian optimization algorithm, and train the cable insulation state evaluation model based on the first cable characteristic parameters and their corresponding first cable insulation states to obtain a well-trained cable insulation state evaluation model; Evaluate the cable characteristic parameters to be evaluated based on the well-trained cable insulation state evaluation model.

2. The cable insulation status evaluation method according to claim 1, characterized in that The cable characteristic parameters include cable operation environment parameters, cable operation years, cable partial discharge amount, and cable visual state conditions; The cable insulation states include good state, general state, signs of aging, relatively serious aging, and complete aging.

3. The cable insulation state evaluation method according to claim 1, characterized in that The pre-training of the multiple feature subsets based on the original lightweight gradient boosting machine model to obtain the first feature parameters includes: Pre-train the multiple feature subsets based on the original lightweight gradient boosting machine model to obtain the K-fold cross-validation accuracy of each feature subset; Determine the feature subset with the highest K-fold cross-validation accuracy as the first feature parameter.

4. The cable insulation state evaluation method according to claim 3, characterized in that, The pre-training of the multiple feature subsets based on the original lightweight gradient boosting machine model includes: Initialize the predicted values of each feature subset; Iteratively calculate the probabilities of each feature subset in each category, and determine the predicted values of each feature subset based on the probabilities; Determine the negative gradient of each feature subset based on the predicted values and true values of each feature subset, where the negative gradient is the target that the new tree needs to fit in each iteration process; Determine the leaf node values after splitting the leaf nodes based on the negative gradients of each feature subset, and optimize the original lightweight gradient boosting machine model based on the leaf node values.

5. The cable insulation state evaluation method according to claim 4, characterized in that The objective to be fitted by the new tree is as follows: where y i,k is the true probability of the feature parameter x ij , p k,m-1 (x ij ) is the predicted probability of the feature parameter x ij , i is the number of samples, j is the number of features per sample, m is the number of iterations, and k is the number of classes; Optimizing the original lightweight gradient boosting machine model based on the leaf node values includes: Among them, F k (X) is the optimized lightweight gradient boosting machine model, η is the learning rate of the model, I is the index of the sample set of a certain leaf node, h is the number of the h-th leaf node on the decision tree, H is the maximum number of leaf nodes set on the decision tree, γ h,k,m is the leaf node value after the leaf node is split, m is the number of iterations, and M is the maximum number of iterations set.

6. The cable insulation state evaluation method according to claim 1, characterized in that The hyperparameters of the cable insulation state evaluation model include the maximum tree depth, the number of leaf nodes, the number of iterations, the data sampling ratio, the learning rate, the L1 regularization term coefficient, and the L2 regularization term coefficient.

7. The cable insulation state evaluation method according to claim 6, wherein The optimization of the hyperparameters of the cable insulation state evaluation model based on the Bayesian optimization algorithm includes: Define the objective function of Bayesian optimization, where the input of the Bayesian optimization objective function is the hyperparameters of the cable insulation state evaluation model, and the output of the Bayesian optimization objective function is the accuracy of the K-fold cross-validation of the feature-selection lightweight gradient boosting machine; Perform Gaussian process processing on the objective function to obtain the posterior probability of the objective function; According to the mean and covariance of the posterior probability, use the acquisition function to update the parameters of the objective function to obtain the optimal hyperparameters.

8. A cable insulation state evaluation device, characterized in that, Including: A data acquisition module for obtaining cable operation data, where the cable operation data includes cable characteristic parameters and their corresponding cable insulation states; A data pre-training module, which is used to perform permutations and combinations on the cable characteristic parameters to obtain multiple feature subsets, and pre-train the multiple feature subsets based on a native lightweight gradient boosting machine model to obtain first cable characteristic parameters; A model training module, which is used to construct a cable insulation state evaluation model, optimize the hyperparameters of the cable insulation state evaluation model based on the Bayesian optimization algorithm, and train the cable insulation state evaluation model based on the first cable characteristic parameters and their corresponding first cable insulation states to obtain a trained complete cable insulation state evaluation model; A model evaluation module, which is used to evaluate the cable characteristic parameters to be evaluated based on the trained complete cable insulation state evaluation model.

9. An electronic device, characterized in that, It includes a memory and a processor, wherein, The memory is used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the cable insulation state evaluation method according to any one of claims 1 to 7 above.

10. A computer-readable storage medium, characterized in that, It is used to store computer-readable programs or instructions, and when the programs or instructions are executed by the processor, the steps in the cable insulation state evaluation method according to any one of claims 1 to 7 above can be implemented.

Images