Composite insulator residual life prediction method and system based on machine learning

Abstract

The invention discloses a machine learning-based composite insulator residual life prediction method and system, and the method comprises the steps: obtaining composite insulator samples under different operation years and different environment conditions, and determining the residual life based on the operation years; testing the composite insulator sample to obtain performance parameter data; determining an initial model for predicting the residual life of the composite insulator based on a machine learning algorithm, and training and optimizing the initial model for predicting the residual life of the composite insulator based on the residual life and the performance parameter data to obtain an optimal model for predicting the residual life of the composite insulator; and using the composite insulator residual life prediction optimal model to predict the life of a to-be-detected composite insulator. According to the method, multiple performance parameters such as the voltage grade, the rated tension, the hydrophobicity and the hardness are comprehensively considered, standardization and normalization processing are carried out, interference between different data characteristics is eliminated, the accuracy and the stability of the model are ensured, and errors and limitation of manual analysis are effectively avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of insulator performance evaluation, and more particularly, to a method and system for predicting the remaining service life of a composite insulator based on machine learning. Background Art

[0002] Composite insulators play a vital role in power systems, primarily isolating and supporting high-voltage conductors to ensure the stable operation of power equipment. As power equipment ages, composite insulators are affected by environmental factors such as temperature fluctuations, humidity, contaminant deposition, and mechanical stress, leading to degradation and consequently affecting their performance and service life. Manifestations of aging include surface damage, decreased hydrophobicity, and weakened mechanical strength, all of which can lead to insulator failure, breakdown, or compromise the safety and stability of power systems. Therefore, accurately assessing the degree of aging and remaining life of composite insulators is crucial to ensuring power system stability.

[0003] Currently, the field of remaining life prediction for composite insulators still faces the challenge of an imperfect methodology. Prediction methods with practical engineering value are limited, and existing models generally suffer from insufficient parameter dimensionality. Most methods establish evaluation systems based on only two or three key parameters, making it difficult to fully characterize the multi-physics coupling effects of material aging. This lack of parameters can lead to misinterpretation of microscopic degradation mechanisms (such as decreased silicone rubber crosslinking and interfacial debonding), directly impacting the reliability of prediction results.

[0004] Therefore, a method for predicting the remaining life of composite insulators based on machine learning is needed. Summary of the Invention

[0005] The present invention proposes a method and system for predicting the remaining service life of a composite insulator based on machine learning to solve the problem of how to efficiently and accurately predict the service life of a composite insulator.

[0006] In order to solve the above problem, according to one aspect of the present invention, a method for predicting the remaining service life of a composite insulator based on machine learning is provided, the method comprising:

[0007] Obtaining composite insulator samples under different operating years and different environmental conditions, and determining the remaining life based on the operating years;

[0008] Testing the composite insulator sample to obtain performance parameter data;

[0009] Determining an initial model for predicting the remaining service life of a composite insulator based on a machine learning algorithm, and training and optimizing the initial model for predicting the remaining service life of a composite insulator based on the remaining service life and performance parameter data to obtain an optimal model for predicting the remaining service life of a composite insulator;

[0010] The composite insulator remaining life prediction optimal model is used to predict the life of the composite insulator to be tested.

[0011] Preferably, the performance parameter data include: voltage level data, rated tensile force data, appearance inspection data, mechanical failure load data, hydrophobicity data, hardness data, tracking data, sheathed core rod water diffusion data, water boiling steep slope data and top sealing data.

[0012] Preferably, the method further comprises:

[0013] Before training and optimizing the initial model for predicting the remaining life of the composite insulator based on the remaining life and performance parameter data, the remaining life is standardized to be converted into a form that conforms to a normal distribution, and the performance parameter data is normalized.

[0014] Preferably, the method uses Z-score standardization to standardize the remaining life to convert it into a form that conforms to a normal distribution, including:

[0015]

[0016] Among them, Z is the standardized data, X is the original data, μ is the mean of the data, and σ is the standard deviation of the data. After standardization by Z-score, the data will be converted into a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0017] Preferably, the machine learning algorithm includes but is not limited to: XGBoost, Catboost, random forest and gradient boosting decision tree algorithms.

[0018] According to another aspect of the present invention, a system for predicting the remaining service life of a composite insulator based on machine learning is provided, the system comprising:

[0019] a sample acquisition unit, configured to acquire composite insulator samples under different operating years and different environmental conditions, and determine the remaining life based on the operating years;

[0020] a performance parameter data acquisition unit, configured to test the composite insulator sample to acquire performance parameter data;

[0021] a model determination unit, configured to determine an initial model for predicting the remaining service life of a composite insulator based on a machine learning algorithm, and to train and optimize the initial model for predicting the remaining service life of the composite insulator based on the remaining service life and performance parameter data to obtain an optimal model for predicting the remaining service life of the composite insulator;

[0022] The prediction unit is used to use the composite insulator remaining life prediction optimal model to predict the life of the composite insulator to be tested.

[0023] Preferably, the performance parameter data include: voltage level data, rated tensile force data, appearance inspection data, mechanical failure load data, hydrophobicity data, hardness data, tracking data, sheathed core rod water diffusion data, water boiling steep slope data and top sealing data.

[0024] Preferably, the system further comprises:

[0025] A preprocessing unit is used to standardize the remaining service life to convert it into a form that conforms to a normal distribution, and to normalize the performance parameter data before training and optimizing the initial model for predicting the remaining service life of the composite insulator based on the remaining service life and performance parameter data.

[0026] Preferably, the preprocessing unit uses Z-score standardization to standardize the remaining life to convert it into a form that conforms to a normal distribution, including:

[0027]

[0028] Among them, Z is the standardized data, X is the original data, μ is the mean of the data, and σ is the standard deviation of the data. After standardization by Z-score, the data will be converted into a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0029] Preferably, in the model determination unit, the machine learning algorithm includes but is not limited to: XGBoost, Catboost, random forest and gradient boosting decision tree algorithms.

[0030] The present invention provides a method and system for predicting the remaining life of composite insulators based on machine learning, comprising: obtaining composite insulator samples with different operating years and different environmental conditions, and determining the remaining life based on the operating years; testing the composite insulator samples to obtain performance parameter data; determining an initial model for predicting the remaining life of the composite insulator based on a machine learning algorithm, and training and optimizing the initial model for predicting the remaining life of the composite insulator based on the remaining life and performance parameter data to obtain an optimal model for predicting the remaining life of the composite insulator; and using the optimal model for predicting the remaining life of the composite insulator to predict the life of the composite insulator to be tested. The present invention comprehensively considers multiple performance parameters such as voltage level, rated tension, hydrophobicity, hardness, etc., and performs standardization and normalization processing to eliminate interference between different data features, ensure the accuracy and stability of the model, and constructs a prediction model through a machine learning algorithm. It can process large-scale, high-dimensional data, optimize the model's prediction ability, and effectively avoid the errors and limitations of manual analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] A more complete understanding of exemplary embodiments of the present invention may be obtained by referring to the following drawings:

[0032] Figure 1 Flowchart of a method 100 for predicting the remaining service life of a composite insulator based on machine learning according to an embodiment of the present invention;

[0033] Figure 2 Schematic diagram of the structure of a composite insulator remaining life prediction system 200 based on machine learning according to an embodiment of the present invention. DETAILED DESCRIPTION

[0034] Exemplary embodiments of the present invention will now be described with reference to the accompanying drawings. However, the present invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to provide a thorough and complete disclosure of the present invention and to fully convey the scope of the present invention to those skilled in the art. The terminology used in the exemplary embodiments shown in the accompanying drawings is not intended to limit the present invention. In the accompanying drawings, identical elements are denoted by the same reference numerals.

[0035] Unless otherwise specified, the terms used herein (including technical terms) have the meanings commonly understood by those skilled in the art. In addition, it is understood that terms defined in commonly used dictionaries should be understood to have the same meanings as those in the context of the relevant fields, and should not be understood as idealized or overly formal meanings.

[0036] The application of machine learning can significantly improve analysis speed and prediction accuracy while avoiding errors inherent in human analysis. This makes the remaining life prediction method for composite insulators based on machine learning of great practical significance. By developing a machine learning model that integrates multi-dimensional experimental data, it is possible to not only accurately assess the current condition of composite insulators but also predict their future aging trends and remaining life. This provides a scientific basis for the maintenance and replacement of power equipment, thereby improving the reliability and safety of power systems.

[0037] The present invention's machine learning-based method for predicting the remaining life of composite insulators begins by collecting samples of composite insulators from different manufacturers and with varying years of operation. The remaining life is calculated based on the years of operation, and relevant performance parameter data is obtained experimentally. This performance parameter data is then standardized and normalized. Next, a suitable machine learning model is selected, and the model structure and training parameters are set. The model is then trained using the training set, and cross-validation is used to optimize the model to prevent overfitting. After training, the model performance is evaluated using an independent test set, and the parameters are adjusted based on the evaluation results, ultimately resulting in an accurate remaining life prediction model.

[0038] Figure 1 FIG. 1 is a flow chart of a method 100 for predicting the remaining life of a composite insulator based on machine learning according to an embodiment of the present invention. Figure 1 As shown, the machine learning-based composite insulator remaining life prediction method provided by an embodiment of the present invention comprehensively considers multiple performance parameters such as voltage level, rated tension, hydrophobicity, and hardness, and performs standardization and normalization processing to eliminate interference between different data features, ensuring the accuracy and stability of the model. The prediction model is constructed through a machine learning algorithm, which can process large-scale, high-dimensional data, optimize the model's predictive capabilities, and effectively avoid the errors and limitations of manual analysis. The machine learning-based composite insulator remaining life prediction method 100 provided by an embodiment of the present invention begins at step 101. In step 101, composite insulator samples with different operating years and different environmental conditions are obtained, and the remaining life is determined based on the operating years.

[0039] In step 102, the composite insulator sample is tested to obtain performance parameter data.

[0040] Preferably, the performance parameter data include: voltage level data, rated tensile force data, appearance inspection data, mechanical failure load data, hydrophobicity data, hardness data, tracking data, sheathed core rod water diffusion data, water boiling steep slope data and top sealing data.

[0041] In this method, we first collect composite insulator samples from different operating years and environmental conditions. The samples should cover different regions, voltage levels, string types, manufacturers, and temperature and humidity environments. This ensures that the experimental data obtained is broadly representative and diverse, thereby enhancing the generalization capabilities of the model.

[0042] Secondly, the remaining life of composite insulators was calculated based on the lifespan thresholds and years of operation in different cities. Specifically, the country was divided into humid and hot regions, arid regions, and other regions based on annual precipitation and temperature. Statistical analysis determined the lifespan thresholds for composite insulators in each region to be 15 years for humid and hot regions, 30 years for arid regions, and 25 years for other regions. The remaining lifespan was then calculated based on the years of operation of the composite insulators.

[0043] The collected composite insulator samples were then subjected to a series of standardized experimental tests to determine various performance parameters. Key test items included voltage level, rated tensile strength, visual inspection, mechanical failure load, hydrophobicity, hardness, tracking, water diffusion on the sheathed core rod, water boiling slope, and top seal. These experimental parameters reflect the performance of composite insulators under different environmental conditions and serve as essential data for predicting their degree of aging and remaining life.

[0044] In step 103, an initial model for predicting the remaining service life of a composite insulator is determined based on a machine learning algorithm, and the initial model for predicting the remaining service life of a composite insulator is trained and optimized based on the remaining service life and performance parameter data to obtain an optimal model for predicting the remaining service life of a composite insulator.

[0045] Preferably, the method further comprises:

[0046] Before training and optimizing the initial model for predicting the remaining life of the composite insulator based on the remaining life and performance parameter data, the remaining life is standardized to be converted into a form that conforms to a normal distribution, and the performance parameter data is normalized.

[0047] Preferably, the method uses Z-score standardization to standardize the remaining life to convert it into a form that conforms to a normal distribution, including:

[0048]

[0049] Among them, Z is the standardized data, X is the original data, μ is the mean of the data, and σ is the standard deviation of the data. After standardization by Z-score, the data will be converted into a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0050] Preferably, the machine learning algorithm includes but is not limited to: XGBoost, Catboost, random forest and gradient boosting decision tree algorithms.

[0051] In step 104, the composite insulator remaining life prediction optimal model is used to predict the life of the composite insulator to be tested.

[0052] In the present invention, the remaining life and experimental parameters can be standardized and normalized. First, the remaining life is used as the target variable and converted into a form that conforms to the normal distribution through standardization, so as to ensure that the model can better understand and predict the target value and avoid the deviation of the data distribution from affecting the training effect. For performance parameter data, since the performance parameter data has different dimensions and numerical ranges, it needs to be normalized to ensure that each feature contributes equally to the model training and to prevent certain features from dominating the model training due to their large numerical ranges. Standardization and normalization can make the data more in line with the requirements of the machine learning algorithm and improve the stability and accuracy of the model.

[0053] Since remaining life data can fluctuate significantly, this can make it difficult for some models to converge effectively during training. To ensure that it meets the training requirements of machine learning models, it needs to be normalized. The purpose of normalization is to transform the data into a normal distribution, thereby avoiding prediction errors caused by large data range differences.

[0054] Therefore, the present invention adopts the Z-score standardization method to process the remaining life, and the formula is as follows:

[0055]

[0056] Where X is the original data, μ is the mean of the data, and σ is the standard deviation of the data.

[0057] After Z-score standardization, the data will be transformed into a standard normal distribution with a mean of 0 and a standard deviation of 1, which helps the machine learning algorithm to learn and optimize better.

[0058] At the same time, different features of experimental data usually have different dimensions and numerical ranges. In order to avoid the excessive impact of certain features on the model training process due to scale differences, it is necessary to normalize the data of each performance parameter. The present invention adopts the Min-Max normalization method, and the formula is as follows:

[0059]

[0060] Among them, X min and X max are the minimum and maximum values ​​of a feature in all data.

[0061] The normalized data range is compressed to [0, 1], ensuring that the impact of each feature on the model is compared at the same scale. This can avoid unfair effects on the model training process due to differences in data dimensions.

[0062] In the present invention, a suitable machine learning algorithm is selected for model training based on the processed data. Depending on the characteristics of the task, the model is constructed based on different machine learning algorithms, including but not limited to: XGBoost, Catboost, random forest, and gradient boosted decision tree (GBDT). Next, the selected algorithm is trained on the training set to optimize the model parameters, and cross-validation and other methods are used to reduce overfitting and improve the generalization ability of the model. After training, a model that can accurately predict the remaining life of composite insulators is obtained, providing a basis for subsequent evaluation and optimization.

[0063] After model training, evaluate the model using an independent test set. Comprehensively assess the model's predictive capabilities by calculating various metrics, such as precision, recall, and F1 score. Based on the evaluation results, adjust the model. Further optimize model performance by adjusting hyperparameters, increasing the amount of training data, or selecting other appropriate features.

[0064] Finally, after model training, evaluation, and adjustment, an accurate prediction model for the remaining life of composite insulators was obtained. This model can be widely used in power systems to predict the remaining life of composite insulators, providing a scientific basis for the maintenance, inspection, and replacement of power equipment, thereby improving the stability and safety of power systems.

[0065] XGBoost is an algorithm based on gradient boosting decision trees. It uses the boosting method to improve prediction accuracy by combining multiple weak classifiers. XGBoost's advantages lie in its efficiency, scalability, and ability to handle large-scale data. It can automatically select the optimal feature subset and has strong anti-overfitting capabilities.

[0066] The goal of XGBoost is to minimize the loss function, combined with the regularization term, the formula is as follows:

[0067]

[0068] in, is the loss function, which measures the difference between the model prediction and the actual label, Ω(f k ) is a regularization term that controls the model complexity.

[0069] Catboost, a gradient boosting tree algorithm developed by Yandex, is particularly well-suited for processing categorical variables. Compared to traditional decision tree methods, Catboost handles categorical data more efficiently, avoiding the need for manual encoding of categorical features in traditional methods. Catboost also utilizes a novel data sorting method to effectively improve prediction accuracy.

[0070] CatBoost calculates the residuals of each tree and sums them up weightedly to obtain the prediction result. Its formula is the same as XGBoost.

[0071] Random forest is an ensemble learning method that improves prediction accuracy by constructing multiple decision trees and performing voting. Each tree is trained by randomly sampling samples and features from the data. Because each tree is relatively independent, random forests are highly resistant to overfitting and can effectively handle high-dimensional data.

[0072] The prediction of random forest is an integration of multiple decision trees. Assuming there are T trees, the final prediction result is the average of all tree prediction results (regression problem) or the voting result (classification problem). The formula is as follows:

[0073]

[0074] where f t (x) is the prediction of the t-th tree, T is the total number of trees, is the final prediction result.

[0075] GBDT is an efficient prediction model that reduces model error by gradually adding decision trees. Each step minimizes the residual error of the previous step and is optimized using the gradient descent algorithm. GBDT is particularly suitable for handling complex nonlinear problems and is often seen in machine learning competitions. Its main disadvantages are complex parameter adjustment, long training time, and sensitivity to outliers. The iterative formula of GBDT uses the gradient of the loss function for updates:

[0076] F t+1 (x) = F t (x)+ρ t h t (x)

[0077] Among them, F t (x) is the model of round t, h t (x) is the decision tree in round t, ρ t is the learning rate.

[0078] The method of the present invention constructs a relatively complete composite insulator status evaluation index system by integrating parameter data of multiple dimensions, synchronously captures the surface degradation and body performance attenuation processes, and realizes a relatively accurate prediction of the remaining service life of the composite insulator.

[0079] The following examples illustrate the embodiments of the present invention.

[0080] In a specific embodiment of the present invention, the process of predicting the remaining service life of a composite insulator includes:

[0081] Composite insulators of varying service life and age were collected from various environments and conditions to ensure that the sample covered a wide range of environmental conditions and operating scenarios. After sample collection, necessary experimental tests were conducted to determine a series of performance parameters, such as voltage level, rated tensile strength, visual inspection, mechanical failure load, hydrophobicity, hardness, tracking, sheathed core rod water diffusion, water boiling slope, and top seal. This experimental data served as feature input for the subsequent prediction model. The remaining life of composite insulators was then calculated based on their life thresholds and service life in different regions.

[0082] After the experiment is complete, all data is preprocessed. First, the different features are standardized and normalized. Remaining life, the target variable, is standardized to conform to a normal distribution, ensuring data stability and consistency. Experimental parameters are normalized to map all feature data to the same numerical range, eliminating dimensional differences between features and ensuring that each contributes equally to model training.

[0083] The processed data is divided into a training set and a test set, where the training set accounts for 85% and the test set accounts for 15%.

[0084] Use XGBoost to analyze the data. XGBoost is an ensemble learning method that improves the predictions of the previous model by gradually adding new decision trees. Each new tree attempts to correct the residual errors left by the previous tree, thereby improving the accuracy of the overall model. During this process, XGBoost uses gradient boosting techniques to update the model. In each iteration, the structure of the new tree is adjusted based on the error of the previous model, so that the model can better fit the data. The basic steps include:

[0085] 1. Initialization: The XGBoost model starts with a constant value F0(x), which is usually the mean of the target values ​​or the most frequent category. In regression problems, the mean of the target values ​​is usually chosen as the initial prediction.

[0086] 2. Iterative tree building: In each iteration t, the model performs the following steps:

[0087] Compute the residual for each sample.

[0088] To optimize the loss function, XGBoost calculates the negative gradient direction of each sample as the target value for the next tree training.

[0089] Use this gradient information to train a new decision tree.

[0090] Add this new tree to the existing model and update the model's prediction results. At this time, the contribution of each tree is controlled by the learning rate η, that is:

[0091] F t (x) = F t-1 (x)+η·h t (x)

[0092] Among them, h t (x) is the output of the t-th tree, and η is the learning rate, which is used to control the contribution of the new tree to the final model.

[0093] 3. Loss Function Optimization: XGBoost optimizes the model by minimizing the loss function L. The choice of loss function varies depending on the task; for regression tasks, the mean squared error (MSE) is commonly used. In each iteration, XGBoost updates the tree structure using a gradient descent algorithm along the negative gradient of the loss function until the loss function reaches its minimum.

[0094] After training, use Optuna to optimize model parameters. Optuna offers efficient hyperparameter optimization, automatically adjusting hyperparameters and finding the optimal model configuration. Using methods like Bayesian optimization, Optuna intelligently explores the hyperparameter space, avoiding the redundant computations that grid search can encounter.

[0095] After optimization, the optimal parameters were obtained. Under these parameters, the average prediction error of the model can reach 1.76 years. The optimal parameters are shown in Table 1.

[0096] Table 1 Optimal model parameters

[0097] max_depth learning_rate subsample colsample_bytree gamma reg_alpha reg_lambda 5 0.03647 0.2178 0.5398 0.1798 0.1893 0.0156

[0098] Finally, the composite insulator remaining life prediction optimal model is used to predict the life of the composite insulator to be tested.

[0099] Figure 2 FIG. 2 is a structural diagram of a composite insulator remaining life prediction system 200 based on machine learning according to an embodiment of the present invention. Figure 2 As shown, a composite insulator remaining life prediction system 200 based on machine learning provided in an embodiment of the present invention includes: a sample acquisition unit 201, a performance parameter data acquisition unit 202, a model determination unit 203 and a prediction unit 204.

[0100] Preferably, the sample acquisition unit 201 is used to acquire composite insulator samples under different operating years and different environmental conditions, and determine the remaining service life based on the operating years.

[0101] Preferably, the performance parameter data acquisition unit 202 is used to test the composite insulator sample to acquire performance parameter data.

[0102] Preferably, the performance parameter data include: voltage level data, rated tensile force data, appearance inspection data, mechanical failure load data, hydrophobicity data, hardness data, tracking data, sheathed core rod water diffusion data, water boiling steep slope data and top sealing data.

[0103] Preferably, the model determination unit 203 is used to determine an initial model for predicting the remaining life of a composite insulator based on a machine learning algorithm, and to train and optimize the initial model for predicting the remaining life of a composite insulator based on the remaining life and performance parameter data to obtain an optimal model for predicting the remaining life of a composite insulator.

[0104] Preferably, the system further comprises:

[0105] A preprocessing unit is used to standardize the remaining life to convert it into a form that conforms to a normal distribution, and to normalize the performance parameter data before training and optimizing the initial model for predicting the remaining life of the composite insulator based on the remaining life and performance parameter data.

[0106] Preferably, the preprocessing unit uses Z-score standardization to standardize the remaining life to convert it into a form that conforms to a normal distribution, including:

[0107]

[0108] Among them, Z is the standardized data, X is the original data, μ is the mean of the data, and σ is the standard deviation of the data. After standardization by Z-score, the data will be converted into a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0109] Preferably, in the model determination unit 203, the machine learning algorithm includes but is not limited to: XGBoost, Catboost, random forest and gradient boosting decision tree algorithms.

[0110] Preferably, the prediction unit 204 is configured to use the composite insulator remaining life prediction optimal model to predict the life of the composite insulator to be tested.

[0111] The composite insulator remaining life prediction system 200 based on machine learning in an embodiment of the present invention corresponds to the composite insulator remaining life prediction method 100 based on machine learning in another embodiment of the present invention, and will not be repeated here.

[0112] The present invention has been described with reference to a few embodiments. However, it is apparent to those skilled in the art that other embodiments than the ones disclosed above are equally within the scope of the present invention.

[0113] Generally, all terms used in this disclosure are to be interpreted according to their ordinary meaning in the art, unless explicitly defined otherwise herein. All references to "a / the / the [device, component, etc.]" are to be interpreted openly as referring to at least one instance of the device, component, etc., unless explicitly stated otherwise. The steps of any method disclosed herein do not necessarily need to be performed in the exact order disclosed, unless explicitly stated otherwise.

[0114] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0115] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0116] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0117] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the present invention.

Claims

1. A method for predicting the remaining life of composite insulators based on machine learning, characterized in that: The method comprises: Obtaining composite insulator samples under different operating years and different environmental conditions, and determining the remaining life based on the operating years; Testing the composite insulator sample to obtain performance parameter data; Determining an initial model for predicting the remaining service life of a composite insulator based on a machine learning algorithm, and training and optimizing the initial model for predicting the remaining service life of a composite insulator based on the remaining service life and performance parameter data to obtain an optimal model for predicting the remaining service life of a composite insulator; The composite insulator remaining life prediction optimal model is used to predict the life of the composite insulator to be tested.

2. The method according to claim 1, characterized in that The performance parameter data include: voltage level data, rated tensile data, appearance inspection data, mechanical failure load data, hydrophobicity data, hardness data, tracking data, sheathed core rod water diffusion data, water boiling steep slope data and top sealing data.

3. The method according to claim 1, characterized in that The method further comprises: Before training and optimizing the initial model for predicting the remaining life of the composite insulator based on the remaining life and performance parameter data, the remaining life is standardized to be converted into a form that conforms to a normal distribution, and the performance parameter data is normalized.

4. The method according to claim 3, characterized in that The method uses Z-score standardization to standardize the remaining life to convert it into a form that conforms to a normal distribution, including: Among them, Z is the standardized data, X is the original data, μ is the mean of the data, and σ is the standard deviation of the data. After standardization by Z-score, the data will be converted into a standard normal distribution with a mean of 0 and a standard deviation of 1.

5. The method according to claim 1, wherein The machine learning algorithms include, but are not limited to, XGBoost, Catboost, random forest, and gradient boosting decision tree algorithms.

6. A composite insulator remaining life prediction system based on machine learning, characterized in that: The system comprises: a sample acquisition unit, configured to acquire composite insulator samples under different operating years and different environmental conditions, and determine the remaining life based on the operating years; a performance parameter data acquisition unit, configured to test the composite insulator sample to acquire performance parameter data; a model determination unit, configured to determine an initial model for predicting the remaining service life of a composite insulator based on a machine learning algorithm, and to train and optimize the initial model for predicting the remaining service life of the composite insulator based on the remaining service life and performance parameter data to obtain an optimal model for predicting the remaining service life of the composite insulator; The prediction unit is used to use the composite insulator remaining life prediction optimal model to predict the life of the composite insulator to be tested.

7. The system according to claim 6, characterized in that The performance parameter data include: voltage level data, rated tensile data, appearance inspection data, mechanical failure load data, hydrophobicity data, hardness data, tracking data, sheathed core rod water diffusion data, water boiling steep slope data and top sealing data.

8. The system according to claim 6, wherein: The system further comprises: A preprocessing unit is used to standardize the remaining life to convert it into a form that conforms to a normal distribution, and to normalize the performance parameter data before training and optimizing the initial model for predicting the remaining life of the composite insulator based on the remaining life and performance parameter data.

9. The system according to claim 8, characterized in that The preprocessing unit uses Z-score standardization to standardize the remaining life to convert it into a form that conforms to a normal distribution, including: Among them, Z is the standardized data, X is the original data, μ is the mean of the data, and σ is the standard deviation of the data. After standardization by Z-score, the data will be converted into a standard normal distribution with a mean of 0 and a standard deviation of 1.

10. The system according to claim 6, wherein: In the model determination unit, the machine learning algorithm includes but is not limited to: XGBoost, Catboost, random forest and gradient boosting decision tree algorithms.

Images