Method and device for constructing residual life prediction model of composite insulator

Abstract

The invention discloses a method and a device for constructing a residual life prediction model of a composite insulator. The method comprises the following steps: acquiring performance parameters of composite insulators with different operating years; performing quantification processing and preprocessing on the performance parameters of the composite insulators with different operating years to obtain a training set and a test set; training the deep learning model by adopting the training set, and optimizing a model gradient by using an optimizer through minimizing a regression loss value to obtain a trained deep learning model; testing the trained deep learning model by adopting a test set, and adjusting hyper-parameters of the trained deep learning model according to a test result; and for the deep learning model after hyper-parameter adjustment, returning to the model training and testing steps until the model reaches a preset performance index. Through the method and the device provided by the embodiment of the invention, a model for objectively and automatically predicting the residual life of the composite insulator is provided, the efficiency is remarkably improved, and the deviation caused by human factors is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of insulator life prediction, in particular to a method and device for constructing a composite insulator residual life prediction model. BACKGROUND

[0002] Composite insulators are a kind of key equipment widely used in power systems, which mainly provide electrical insulation and support conductors to ensure the safe operation of power transmission lines and power equipment. Compared with traditional porcelain insulators and glass insulators, composite insulators use high polymer materials such as silicone rubber as the outer insulation layer, which has excellent pollution flashover resistance, strong anti-aging ability and light weight, so it can still maintain stable electrical and mechanical properties in harsh environments.

[0003] Due to the long-term exposure of composite insulators to complex environmental conditions, such as high voltage, electric field stress, ultraviolet radiation, heat and humidity, pollution and mechanical stress, the material properties will gradually degrade, which will affect the electrical and mechanical properties of the insulator. Life assessment of composite insulators can predict the failure risk of insulators in advance, avoid sudden failures caused by degradation of insulation performance or reduction of mechanical strength, and reduce the risk of line tripping, short circuit or large area power outage. Secondly, through life assessment, the replacement time of the insulator can be reasonably arranged to avoid waste of resources caused by premature replacement, and to reduce safety hazards caused by late replacement and improve the efficiency of power equipment.

[0004] The service life of composite insulators is affected by many performance factors, including hydrophobicity, electrical performance, mechanical strength, anti-aging ability, pollution flashover resistance and thermal stability. These properties interact with each other to determine the long-term service reliability and service life of composite insulators. At present, the life assessment of composite insulators mainly relies on experimental testing, online monitoring and numerical simulation methods. Laboratory accelerated aging test is one of the most commonly used life assessment methods, which simulates long-term service environment to study the changes of key indicators such as hydrophobicity, electrical performance and mechanical performance of the insulator, and then estimates its service life. However, there are some differences between laboratory environment and actual working conditions, and it is difficult to completely reproduce the real aging process; field online monitoring can obtain the running state of composite insulators in real time, but online monitoring is greatly affected by external environment, such as climate change and pollution accumulation, which may cause fluctuations in monitoring data, and the deployment cost of monitoring equipment is high.

[0005] In summary, although the existing life assessment methods can reflect the aging state of composite insulators to some extent, they have limitations and cannot provide comprehensive and accurate life prediction alone. SUMMARY

[0006] In view of this, the present application proposes a method and device for constructing a composite insulator residual life prediction model, aiming to solve one or more of the technical problems mentioned in the background section.

[0007] In the first aspect, the embodiments of the present application provide a method for constructing a composite insulator residual life prediction model, which comprises: acquiring performance parameters of composite insulators with different operating life; performing quantitative processing and preprocessing on the performance parameters of the composite insulators with different operating life to obtain a training set and a test set; training a deep learning model using the training set, minimizing the regression loss value, optimizing the model gradient using an optimizer, and obtaining a trained deep learning model; testing the trained deep learning model using the test set, and adjusting the hyperparameters of the trained deep learning model according to the test results; returning to the model training and testing steps for the deep learning model with adjusted hyperparameters until the model reaches a preset performance indicator.

[0008] Further, the performance parameters include: appearance, mechanical damage load, hydrophobicity, hardness, umbrella skirt tear strength, electric leakage tracking, stress corrosion, water diffusion of core rod with sheath, water diffusion of core rod without sheath, water boiling steep wave, temperature rise test, dissection test, end sealing, dye penetration, rated mechanical load resistance, load resistance test data, and power frequency resistance test data.

[0009] Further, the quantitative processing and preprocessing of the performance parameters of the composite insulators with different operating life to obtain the training set and the test set comprises: performing quantitative processing on part of the performance parameters of the composite insulators with different operating life according to a preset rule, and merging the quantitatively processed performance parameters with the remaining performance parameters that have not been quantitatively processed to obtain quantitatively processed performance parameters; standardizing the quantitatively processed performance parameters to a standard normal distribution to obtain standardized data; and dividing the standardized data into a training set and a test set according to a preset proportion.

[0010] Further, the deep learning model is obtained in the following manner: defining a neural network model in a deep learning framework, and determining the neural network model to be at least one of a multi-layer perceptron, a convolutional neural network, a recurrent neural network, and a self-attention model; selecting the training settings of the model according to the determined neural network model, determining the number of neurons, the learning rate, the batch size, the loss function, and the optimizer of each layer, and obtaining the deep learning model.

[0011] Furthermore, the trained deep learning model is tested using a test set, and the hyperparameters of the trained deep learning model are adjusted according to the test results, including: inputting the test set into the trained deep learning model, and selecting the level with the highest probability from the output probability distribution of each aging level as the prediction result; based on the prediction results, calculating the mean square error, root mean square error, mean absolute error and determination coefficient of the model on the test set to evaluate the model performance; and adjusting the hyperparameters of the model according to the evaluated model performance.

[0012] In a second aspect, an embodiment of the present invention also provides a device for constructing a composite insulator remaining life prediction model, the device comprising: an acquisition unit for acquiring performance parameters of composite insulators of different operating years; a processing unit for quantifying and preprocessing the performance parameters of the composite insulators of different operating years to obtain a training set and a test set; a training unit for training a deep learning model using the training set, optimizing the model gradient using an optimizer by minimizing the regression loss value, and obtaining a trained deep learning model; a testing unit for testing the trained deep learning model using the test set, and adjusting the hyperparameters of the trained deep learning model according to the test results; for the deep learning model after the hyperparameters are adjusted, returning to the training unit and the testing unit until the model reaches the preset performance indicators.

[0013] Furthermore, the performance parameters include: appearance, mechanical failure load, hydrophobicity, hardness, shed tearing strength, tracking, stress corrosion, water diffusion of sheathed core rod, water diffusion of unsheathed core rod, water boiling steep wave, temperature rise test, autopsy test, end seal, dye penetration, rated mechanical load tolerance, load tolerance test data and power frequency withstand voltage test data.

[0014] Furthermore, the processing unit is also used to: quantify some of the performance parameters of the composite insulators with different operating years according to preset rules, and merge them with the remaining unquantified performance parameters to obtain quantized performance parameters; standardize the quantized performance parameters to a standard normal distribution to obtain standardized data; and divide the standardized data into a training set and a test set according to a preset ratio.

[0015] Furthermore, the deep learning model is obtained in the following manner: defining a neural network model in a deep learning framework, and determining that the neural network model is at least one of a multi-layer perceptron, a convolutional neural network, a recurrent neural network, and a self-attention model; according to the determined neural network model, selecting the training settings of the model, determining the number of neurons in each layer, the learning rate, the batch size, the loss function, and the optimizer, to obtain a deep learning model.

[0016] Further, the test unit is further configured to: input the test set into the trained deep learning model, select a probability highest level from the output probability distribution of each aging level as a prediction result; based on the prediction result, calculate mean square error, root mean square error, mean absolute error and determination coefficient of the model on the test set to evaluate the model performance; and adjust the hyperparameters of the model according to the evaluated model performance.

[0017] In a third aspect, an embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method provided by any of the above embodiments.

[0018] In a fourth aspect, an embodiment of the present application further provides an electronic device, which comprises: a processor; a memory for storing executable instructions of the processor; and the processor is configured to read the executable instructions from the memory and execute the instructions to implement the method provided by any of the above embodiments.

[0019] The method and device for constructing a composite insulator residual life prediction model provided by the embodiment of the present application quantize and preprocess the performance parameters of composite insulators with different operating years to obtain a training set and a test set, train a deep learning model using the training set, obtain a trained deep learning model by minimizing the regression loss value and optimizing the model gradient using an optimizer, test the trained deep learning model using the test set, adjust the hyperparameters of the trained deep learning model according to the test result, and return the deep learning model with adjusted hyperparameters to the model training and testing step until the model reaches a preset performance indicator. Compared with the traditional manual analysis method, the composite insulator residual life prediction model constructed by the embodiment provides an automatic process of a more objective and consistent evaluation standard, significantly improves the efficiency of data analysis, and reduces the deviation caused by human factors. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 An exemplary flowchart of the method for constructing a composite insulator residual life prediction model according to an embodiment of the present application is shown;

[0021] Figure 2 A schematic diagram of a Transformer model structure according to an embodiment of the present application is shown;

[0022] Figure 3 A structural schematic diagram of the device for constructing a composite insulator residual life prediction model according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0023] Reference will now be made to the drawings to describe the exemplary embodiments of the present application in greater detail. The present application may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and fully convey the scope of the application to those skilled in the art. The terminology used in the description presented herein is not intended to be interpreted in any limited or restrictive manner simply because it is being used in conjunction with a detailed description of particular embodiments. In the drawings, like reference numerals refer to like elements throughout.

[0024] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

[0025] Figure 1 An exemplary flowchart of a method for constructing a residual life prediction model of a composite insulator according to an embodiment of the present application is shown.

[0026] As shown in Figure 1 The method comprises the following steps:

[0027] Step S101: Obtain performance parameters of composite insulators with different service life.

[0028] Further, the performance parameters include: appearance, mechanical breakdown load, hydrophobicity, hardness, shed tear resistance, tracking, stress corrosion, water diffusion with sheath, water diffusion without sheath, steep wave after boiling, temperature rise test, cross-section test, end sealing, dye penetration, rated mechanical load resistance, load resistance test data for 96 hours, and power frequency (wet) withstand voltage test data.

[0029] Specifically, performance parameters of composite insulators with different service life are collected, including: appearance, mechanical breakdown load, hydrophobicity, hardness, shed tear resistance, tracking, stress corrosion, water diffusion with sheath, water diffusion without sheath, steep wave after boiling, temperature rise test, cross-section test, end sealing, dye penetration, rated mechanical load resistance, load resistance test for 96 hours, and power frequency (wet) withstand voltage test. These insulator samples should cover samples with different service life, different operating environment, and different voltage grade to ensure the diversity and representativeness of the data.

[0030] For the samples of composite insulators, additional information about their operating environment is collected, including temperature, humidity, pollution level, rainfall, etc., and it is used as part of the deep learning input features.

[0031] Step S102: Quantize and preprocess the performance parameters of composite insulators with different service life to obtain a training set and a test set.

[0032] Further, step S102 comprises:

[0033] After quantifying part of the performance parameters of the composite insulators with different service life according to the preset rule, the quantified performance parameters are combined with the remaining performance parameters which are not quantified to obtain quantified performance parameters;

[0034] The quantified performance parameters are standardized to a standard normal distribution to obtain standardized data;

[0035] The standardized data is divided into a training set and a test set according to a preset proportion.

[0036] Specifically, the performance parameters of part of the composite insulators are quantified: Table 1 shows the quantification standard of the appearance of the insulator according to an embodiment of the present application. The appearance is usually described by words such as “powdering” and “damage”. The appearance is quantified to a range of [0, 1] by the standard shown in Table 1. Part of the performance parameters (stress corrosion, water boiling steep wave, temperature rise test, section test, end sealing, dye penetration, rated mechanical load resistance, 96h load resistance test, and power frequency (wet) withstand voltage test) are determined by “qualified” and “unqualified”. At this time, “qualified” is quantified as 1, and “unqualified” is quantified as 0, so as to be distinguished.

[0037] Table 1

[0038]

[0039] Data standardization: Since the data distribution ranges of different features are different, all features are standardized to a standard normal distribution by standardization, and the data set is divided into a training set and a test set according to a ratio of 8:2.

[0040] Since it is difficult to ensure that all performance index experiments are complete in the historical performance detection experimental data of the composite insulator samples, but the historical data still includes valuable information. In order to ensure the data amount of the data set, unknown performance characteristics can be filled by mathematical methods.

[0041] First, the performance parameters of the composite insulators are quantified, and the performance parameters described by words are quantified to numbers. Then, it is confirmed that there is no missing or abnormal value in the sample performance data. If there is a missing value, a mathematical method is selected for filling according to the situation (such as using an average value, a median value, or an interpolation method).

[0042] All performance data is converted to a standard normal distribution using Z-score standard deviation, as shown in equation (1).

[0043]

[0044] where X is the input data, μ is the mean, and σ is the variance, X new is the standardized data.

[0045] Step S103: The training set is used to train the deep learning model, the regression loss value is minimized, the model gradient is optimized using the optimizer, and the trained deep learning model is obtained.

[0046] Further, the deep learning model is obtained in the following manner:

[0047] A neural network model is defined in a deep learning framework, and the neural network model is determined to be at least one of a multilayer perceptron, a convolutional neural network, a recurrent neural network, and a self-attention model.

[0048] According to the determined neural network model, the training settings of the model are selected, the number of neurons in each layer, the learning rate, the batch size, the loss function, and the optimizer are determined, and the deep learning model is obtained.

[0049] Specifically, a neural network model is defined in a deep learning framework (TensorFlow or PyTorch), and the model structure is determined. For the composite insulator depth and life prediction task, it can be regarded as a regression task, and deep learning models such as multilayer perceptron (MLP), convolutional neural network (CNN), recurrent neural network (RNNs), or self-attention model (Transformer) can be used to solve the regression task. Using a variety of models in combination to form a hybrid model can improve prediction accuracy, especially in complex data distribution and strong nonlinear relationship. For example, MLP and Transformer are combined, Transformer focuses on global patterns, MLP processes local features, and the combination of the two can reduce information loss and reduce the risk of overfitting of a single model.

[0050] After determining the model structure, the training settings of the model need to be selected, the number of neurons in each layer, the learning rate, the batch size, the loss function, and the optimizer are determined.

[0051] The training set established in the foregoing is used to train the deep learning model. During the training process, the model minimizes the regression loss value, optimizes the model gradient using the optimizer, and finally determines the weights of the model to obtain the best remaining life prediction effect.

[0052] Next, a deep learning model is established for the remaining life prediction of the composite insulator. Different performance parameter data is used as input features, the remaining life of the composite insulator is used as model output, and the Transformer model is used for training, Figure 2 A schematic diagram of a Transformer model structure according to an embodiment of the present application is shown.

[0053] The Transformer uses multi-head self-attention mechanism, as shown in equations (2)-(4).

[0054] Q = W Q ·x, K = W K ·x, V = W V ·x (2)

[0055]

[0056] O Attention = [A1, A2, …, A n ]W O (4)

[0057] Where Query (Q), Key (K) and Value (V) are three matrices that can map the correlation between different features, calculated by input x and trainable parameter matrix W Q , W K and W V . d k is the dimension of K, used to scale the dot product of Q and K to stabilize the gradient in the training process, A is the self-attention calculated by Q, K, V, and finally multiple parallel self-attention is spliced and multiplied by an additional trainable matrix W O as the output of the module.

[0058] The self-attention mechanism in the Transformer allows global interaction of information, can effectively model long-distance dependencies, and is not limited to a fixed size of local receptive field, but can model global features in the entire input sequence range. In addition, the calculation process of the model is highly parallelized, benefiting from the independent calculation units of the self-attention mechanism and the feedforward neural network, which can fully utilize GPU acceleration for training. Overall, the Transformer combines long-range dependency modeling capability, high parallelization, global feature learning, multi-level feature expression, and stable gradient optimization, and is one of the most important models in the current deep learning field.

[0059] Using the deep learning framework pytorch to build the model Transformer: First, the libraries needed to build and train the model need to be imported; Next, define a Transformer model containing an input layer, a self-attention layer and an output layer, the Transformer can set the size of the parameter scale, the larger the parameter quantity, the stronger the expression ability of the model in theory, but it will also increase the training difficulty and the calculation cost. Among them, the relu activation function is added after the input layer to introduce a nonlinear part in the model.

[0060] Next, define the loss function and optimizer required during model training. In regression tasks, the commonly used loss function is Mean Squared Error (MSE), as shown in equation (5).

[0061]

[0062] where y and are the true value and predicted value, respectively, and n is the number of samples during training. MSE calculates the squared error between the predicted value and the true value, and takes the average, which is sensitive to large errors.

[0063] The gradient of the loss function with respect to the model parameters is calculated by the backpropagation algorithm, and then the parameters are updated using the optimization algorithm Momentum, as shown in equations (6) and (7).

[0064] θ t = β t-1 + ηg t (6)

[0065] θ t = θ t-1 + v t (7)

[0066] where η is the learning rate, controls the step size, and β is the momentum hyperparameter, usually taking a value of 0.9 or close to 1. v t represents the momentum at the current time, which is equivalent to the cumulative gradient information. The Momentum algorithm combines the gradient information of the current and past multiple time steps, allowing the optimization process to maintain a faster speed in the valley region, thereby avoiding stagnation in small gradient regions.

[0067] Step S104: Test the trained deep learning model using the test set, and adjust the hyperparameters of the trained deep learning model based on the test results.

[0068] Further, step S104 includes:

[0069] Input the test set into the trained deep learning model, and select the highest probability level from the output probability distribution of each aging level as the prediction result;

[0070] Based on the prediction result, calculate the mean squared error, root mean squared error, mean absolute error, and coefficient of determination of the model on the test set to evaluate the model performance;

[0071] Adjust the hyperparameters of the model based on the evaluated model performance.

[0072] Specifically, the trained model is evaluated using the test set, and the model outputs the probability distribution of each aging level. The level with the highest probability is selected as the prediction result. The mean square error, root mean square error, mean absolute error, and determination coefficient of the model on the test set are calculated to evaluate the performance of the model. Among them, the evaluation index is also the mean square error shown in formula (5). The performance of the model on the test set is analyzed, and according to the test results, the hyperparameters of the model, such as the learning rate and the batch size, are adjusted to improve the performance of the model.

[0073] The relationship between the input and output of the model is analyzed using a visualization analysis method, and the performance characteristics in the input that are not important or less important are deleted, so that the model can better learn the real pattern of the data, improve the generalization ability on new data, and at the same time, reduce the computing overhead, reduce the training time, speed up the model convergence speed, and reduce the storage and inference cost.

[0074] The hyperparameters can be adjusted by training strategies or optimization methods such as grid search, random search, or Bayesian optimization to further improve the performance of the model.

[0075] Further, for the deep learning model after adjusting the hyperparameters, the model training and testing steps are returned again until the model reaches the preset performance indicators, including:

[0076] Specifically, the model is retrained using the adjusted hyperparameters and tested on the test set. The above steps are iterated, and the hyperparameters of the model are further adjusted according to the test results and feedback information until the performance reaches the preset performance indicators, and the final deep learning model is obtained as the composite insulator remaining life prediction model.

[0077] Using the model to predict the remaining life of the composite insulator, visualization methods such as Matplotlib, Seaborn, or Plotly can be used to display the relationship between the input insulator performance parameters and the output remaining life in the deep learning model.

[0078] In the above embodiment, the performance parameters of the composite insulator with different operating years are quantified and preprocessed to obtain the training set and the test set. The training set is used to train the deep learning model, the model gradient is optimized using the optimizer by minimizing the regression loss value, and the trained deep learning model is obtained. The trained deep learning model is tested using the test set, and the hyperparameters of the trained deep learning model are adjusted according to the test results. For the deep learning model after adjusting the hyperparameters, the model training and testing steps are returned again until the model reaches the preset performance indicators. Compared with the traditional manual analysis method, the composite insulator remaining life prediction model constructed in this embodiment provides an automatic process of a more objective and consistent evaluation standard, which significantly improves the efficiency of data analysis and reduces the bias caused by human factors.

[0079] Figure 3 A schematic structural diagram of an apparatus for constructing a composite insulator remaining life prediction model according to an embodiment of the present invention is shown.

[0080] like Figure 3 As shown, the device includes:

[0081] An acquisition unit 301 is used to acquire performance parameters of composite insulators of different service years;

[0082] The processing unit 302 is used to quantify and preprocess the performance parameters of composite insulators with different service years to obtain a training set and a test set;

[0083] A training unit 303 is configured to train the deep learning model using the training set, optimize the model gradient using an optimizer by minimizing the regression loss value, and obtain a trained deep learning model;

[0084] A testing unit 304 is configured to test the trained deep learning model using a test set and adjust hyperparameters of the trained deep learning model based on the test results.

[0085] For the deep learning model after hyperparameter adjustment, return to the training unit and test unit until the model reaches the preset performance indicators.

[0086] Furthermore, the performance parameters include: appearance, mechanical failure load, hydrophobicity, hardness, shed tearing strength, tracking, stress corrosion, water diffusion of sheathed core rod, water diffusion of unsheathed core rod, water boiling steep wave, temperature rise test, autopsy test, end seal, dye penetration, rated mechanical load tolerance, load tolerance test data and power frequency withstand voltage test data.

[0087] Furthermore, the processing unit 302 is further configured to:

[0088] After quantifying some of the performance parameters of composite insulators with different service years according to preset rules, the performance parameters are combined with the remaining unquantified performance parameters to obtain quantified performance parameters;

[0089] Normalize the quantitative performance parameters to the standard normal distribution to obtain standardized data;

[0090] The standardized data is divided into training set and test set according to the preset ratio.

[0091] Furthermore, the deep learning model is obtained in the following way:

[0092] define a neural network model in a deep learning framework, and determine that the neural network model is at least one of a multilayer perceptron, a convolutional neural network, a recurrent neural network, and a self-attention model;

[0093] According to the determined neural network model, select the training settings of the model, determine the number of neurons in each layer, the learning rate, the batch size, the loss function, and the optimizer, and obtain the deep learning model.

[0094] Further, the test unit 304 is also used for:

[0095] input the test set into the trained deep learning model, and select the highest probability level from the output probability distribution of each aging level as the prediction result;

[0096] Based on the prediction result, calculate the mean square error, root mean square error, mean absolute error, and determination coefficient of the model on the test set to evaluate the model performance;

[0097] According to the evaluated model performance, adjust the hyperparameters of the model.

[0098] The above embodiment quantizes and preprocesses the performance parameters of composite insulators with different service life, obtains a training set and a test set, trains a deep learning model using the training set, optimizes the model gradient using an optimizer by minimizing the regression loss value, obtains a trained deep learning model, tests the trained deep learning model using the test set, adjusts the hyperparameters of the trained deep learning model according to the test result, and returns to the model training and testing step for the deep learning model with adjusted hyperparameters until the model reaches a preset performance indicator. Compared with the traditional manual analysis method, the composite insulator residual life prediction model constructed in this embodiment provides an automatic process of a more objective and consistent evaluation standard, significantly improves the efficiency of data analysis, and reduces the deviation caused by human factors.

[0099] It should be noted that the device provided in the above embodiment is only exemplified by the division of the above functional modules when realizing its functions, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above described functions. In addition, the device and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process is described in detail in the method embodiments, which will not be repeated here.

[0100] The embodiment of the application also provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the method for constructing a composite insulator residual life prediction model provided in each of the above embodiments.

[0101] The embodiment of the present application also provides an electronic device, comprising: a processor; a memory for storing processor-executable instructions; the processor is used for reading the executable instructions from the memory and executing the instructions to realize the method for constructing the residual life prediction model of the composite insulator provided by each of the above embodiments.

[0102] The present application has been described with reference to a few embodiments. However, those skilled in the art will appreciate that other embodiments than those described herein are possible, and that the scope of the application is not limited to the embodiments described herein. The scope of the present application is defined by the appended patent claims.

[0103] Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the technical field, unless explicitly defined otherwise herein. All references to "a / an / the [device, component, etc.] are to be interpreted to mean "one or more" of the entity unless otherwise indicated. The steps of any method disclosed herein do not have to be performed in the exact order disclosed unless explicitly stated.

[0104] Those skilled in the art will appreciate that embodiments of the present application can be readily used as a method, a system or a computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer program code thereon for use by or in connection with an instruction execution system. For the purposes of this description, a computer-usable or computer readable storage medium can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.

[0105] The present application is described with reference to the accompanying drawings, which show example embodiments of the present application. In the drawings: the flowcharts and / or block diagrams illustrate the architecture, functionality, and operation of possible implementations of apparatuses (systems), methods and computer program products according to various embodiments of the present application. In this regard, each flowchart and / or block diagram can represent a method, a system, and / or a computer program product that can be implemented by a processor, such as a processor of a special purpose computer, an embedded processor, or a general purpose computer. Each flowchart and / or block diagram can also represent a processor that can be programmed to carry out the functions, acts, and / or instructions of the flowcharts and / or block diagrams. Figure 1 one or more functions specified in the flow or flows and / or block or blocks. Figure 1 one or more functions specified in the flow or flows and / or block or blocks.

[0106] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow Figure 1 one or more functions specified in the flow or flows and / or block or blocks. Figure 1the function specified in the one or more blocks.

[0107] These computer program instructions can also be loaded into computer or other programmable data processing devices, so that a series of operation steps are performed on the computer or other programmable data processing devices to generate computer-implemented processes, so that the instructions executed on the computer or other programmable devices provide processes for implementing the flow Figure 1 one or more flows and / or blocks Figure 1 the steps of the function specified in the one or more blocks.

[0108] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, but not to limit it, although the above embodiments of the present application have been described in detail, those skilled in the art should understand: the specific embodiments of the present application can be modified or replaced by the same, without departing from the spirit and scope of the present application, any modification or equivalent replacement, which should be covered in the protection scope of the claims of the present application.

Claims

1. A method for constructing a composite insulator remaining life prediction model, characterized in that: The method comprises: Obtain performance parameters of composite insulators with different service years; Quantifying and preprocessing the performance parameters of the composite insulators with different service years to obtain a training set and a test set; The deep learning model is trained using the training set. By minimizing the regression loss value and optimizing the model gradient using the optimizer, the trained deep learning model is obtained. Use the test set to test the trained deep learning model and adjust the hyperparameters of the trained deep learning model based on the test results; For the deep learning model after hyperparameter adjustment, return to the model training and testing steps until the model reaches the preset performance indicators.

2. The method according to claim 1, characterized in that The performance parameters include: appearance, mechanical failure load, hydrophobicity, hardness, shed tear resistance, tracking, stress corrosion, water diffusion of sheathed core rod, water diffusion of unsheathed core rod, water boiling steep wave, temperature rise test, autopsy test, end seal, dye penetration, rated mechanical load tolerance, load tolerance test data and power frequency withstand voltage test data.

3. The method according to claim 1, characterized in that The performance parameters of the composite insulators with different service years are quantified and preprocessed to obtain a training set and a test set, including: After quantifying some of the performance parameters of the composite insulators of different service years according to a preset rule, the performance parameters are combined with the remaining unquantified performance parameters to obtain quantified performance parameters; Normalize the quantitative performance parameters to the standard normal distribution to obtain standardized data; The standardized data is divided into training set and test set according to the preset ratio.

4. The method according to claim 1, wherein The deep learning model is obtained in the following way: Define a neural network model in a deep learning framework and determine that the neural network model is at least one of a multi-layer perceptron, a convolutional neural network, a recurrent neural network, and a self-attention model; Based on the determined neural network model, the training settings of the model are selected, and the number of neurons in each layer, learning rate, batch size, loss function and optimizer are determined to obtain a deep learning model.

5. The method according to claim 1, wherein The method of testing the trained deep learning model using the test set and adjusting the hyperparameters of the trained deep learning model according to the test results includes: The test set is input into the trained deep learning model, and the level with the highest probability is selected from the output probability distribution of each aging level as the prediction result; Based on the prediction results, the mean square error, root mean square error, mean absolute error, and coefficient of determination of the model on the test set are calculated to evaluate the model performance; Based on the evaluated model performance, tune the model's hyperparameters.

6. A device for constructing a composite insulator remaining life prediction model, characterized in that: The device comprises: An acquisition unit, used to acquire performance parameters of composite insulators of different service years; a processing unit, configured to quantify and preprocess the performance parameters of the composite insulators of different service years to obtain a training set and a test set; The training unit is used to train the deep learning model using the training set. By minimizing the regression loss value and optimizing the model gradient using the optimizer, the trained deep learning model is obtained. A testing unit is used to test the trained deep learning model using a test set and adjust the hyperparameters of the trained deep learning model based on the test results; For the deep learning model after hyperparameter adjustment, return to the training unit and test unit until the model reaches the preset performance indicators.

7. The device according to claim 6, characterized in that The performance parameters include: appearance, mechanical failure load, hydrophobicity, hardness, shed tear resistance, tracking, stress corrosion, water diffusion of sheathed core rod, water diffusion of unsheathed core rod, water boiling steep wave, temperature rise test, autopsy test, end seal, dye penetration, rated mechanical load tolerance, load tolerance test data and power frequency withstand voltage test data.

8. The device according to claim 6, characterized in that The processing unit is further configured to: After quantifying some of the performance parameters of the composite insulators of different service years according to a preset rule, the performance parameters are combined with the remaining unquantified performance parameters to obtain quantified performance parameters; Normalize the quantitative performance parameters to the standard normal distribution to obtain standardized data; The standardized data is divided into training set and test set according to the preset ratio.

9. The device according to claim 6, characterized in that The deep learning model is obtained in the following way: Define a neural network model in a deep learning framework and determine that the neural network model is at least one of a multi-layer perceptron, a convolutional neural network, a recurrent neural network, and a self-attention model; Based on the determined neural network model, the training settings of the model are selected, and the number of neurons in each layer, learning rate, batch size, loss function and optimizer are determined to obtain a deep learning model.

10. The device according to claim 6, characterized in that The testing unit is further used for: The test set is input into the trained deep learning model, and the level with the highest probability is selected from the output probability distribution of each aging level as the prediction result; Based on the prediction results, the mean square error, root mean square error, mean absolute error, and coefficient of determination of the model on the test set are calculated to evaluate the model performance; Based on the evaluated model performance, tune the model's hyperparameters.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

12. An electronic device comprising: processor; a memory for storing instructions executable by the processor; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the method according to any one of claims 1 to 5.

Images