Lightning arrester residual service life prediction method based on digital-analog linkage

Abstract

The invention discloses a lightning arrester residual service life prediction method based on digital-analog linkage. The lightning arrester residual service life prediction method comprises the steps of obtaining real-time monitoring data of a to-be-predicted lightning arrester; preprocessing the lightning arrester real-time monitoring data to be predicted to obtain input data; inputting the input data into a pre-constructed lightning arrester residual service life prediction model based on an XGBoost algorithm, and outputting to obtain the predicted residual service life of the lightning arrester; based on the target Wiener process model, obtaining a confidence interval of the predicted residual service life of the lightning arrester; and taking the output residual service life of the lightning arrester as a predicted value, and judging whether the predicted value exceeds a third harmonic current set value or not. According to the invention, through digital-analog linkage, i.e., combination of a physical model and a data driving model, precision, intelligentization and stabilization of lightning arrester residual service life prediction are realized, defects of a traditional feature determination method are effectively made up, and adaptability of a data driving method in a complex power grid environment is improved at the same time.

Description

Technical Field

[0001] The present invention relates to the field of arrester service life prediction, and in particular to a method for predicting the remaining service life of an arrester based on digital-analog linkage. Background Art

[0002] Metal oxide surge arresters (MOSAs) play a vital role in protecting power grid equipment from surge voltage damage. In modern power systems, MOSA health monitoring and service life prediction are key technologies for ensuring the long-term stable operation of power equipment. However, due to the complexity of the grid operating environment, the variability of external interference, and the nonlinear evolution of equipment aging, accurate MOSA service life prediction still faces many challenges.

[0003] Currently, the aging process of MOSA is typically assessed by monitoring key status parameters such as leakage current and temperature. These indicators can reflect the health status and degradation trends of the device. Traditional prediction methods rely primarily on empirical models or statistical methods, such as adaptive neural-fuzzy inference systems (ANFIS) and support vector regression (SVR). However, these methods often exhibit insufficient prediction accuracy and adaptability when faced with multi-source heterogeneous data, high-dimensional nonlinear relationships, and complex environmental interference. Furthermore, traditional prediction methods struggle to dynamically adjust the prediction model, making the prediction results susceptible to data noise, which affects the stability and reliability of long-term monitoring.

[0004] In recent years, with the rapid development of data-driven methods and machine learning technologies, the use of efficient regression models to accurately predict the service life of MOSAs has become a research hotspot. The XGBoost algorithm, in particular, due to its powerful modeling capabilities and automatic feature selection, can better capture device aging characteristics in complex data environments, improving prediction accuracy and generalization. However, relying solely on a single data-driven approach still has significant shortcomings: 1) While the XGBoost algorithm, as a purely data-driven model, can learn device aging trends from historical data, its predictive performance is highly dependent on data quality, potentially leading to misjudgments or lags in non-stationary aging processes; 2) A single third harmonic feature cannot fully characterize the physical aging mechanism of lightning arresters and is susceptible to environmental interference such as grid harmonic pollution and temperature fluctuations, resulting in reduced model stability and robustness. Summary of the Invention

[0005] In order to solve the above-mentioned technical problems, the present invention provides a method for predicting the remaining service life of a lightning arrester based on digital-analog linkage. By combining a data-driven model with a physical model, a high-precision and high-stability lightning arrester service life prediction model is constructed to improve the adaptability and robustness of the model in a complex power grid environment.

[0006] The technical solution adopted by the present invention is: A method for predicting the remaining service life of a lightning arrester based on digital-analog linkage, comprising: S1. Obtaining real-time monitoring data of the arrester to be predicted; S2. Preprocessing the real-time monitoring data of the lightning arrester to be predicted to obtain input data; S3. Input the input data into a pre-built arrester remaining service life prediction model based on the XGBoost algorithm, and output a predicted value of the arrester remaining service life; S4. Based on the target Wiener process model, the confidence interval of the predicted value of the remaining service life of the lightning arrester is output; S5. Based on the predicted value of the remaining service life of the lightning arrester and the set value, determine the working status of the lightning arrester and formulate a corresponding maintenance strategy.

[0007] Furthermore, the real-time monitoring data of the lightning arrester to be predicted in S1 includes the leakage current, ambient temperature, vibration amplitude, and operating voltage of the lightning arrester. The specific acquisition process is as follows: 1) Collecting leakage current: The arrester's leakage current signal is collected using a magnetic current sensor with a resolution of 0.01 mA and a portable digital oscilloscope with a sampling rate of 1 Msamples / s. The data format is set to time series. 2) Collecting ambient temperature: The temperature inside the arrester is collected as the ambient temperature by a high-precision temperature sensor installed on the surface of the arrester, and the sampling frequency is once per hour; 3) Collecting vibration amplitude: The vibration signal of the lightning arrester is collected by a piezoelectric vibration sensor with a sensitivity of 0.01g installed on the base of the lightning arrester. The sampling rate is 10kHz. The data is filtered and the average vibration amplitude is obtained; 4) Collecting working voltage: Measure the actual operating voltage at both ends of the arrester through a voltage sensor and use the daily average of the actual operating voltage as the working voltage.

[0008] Furthermore, S2 includes: S21. Perform feature extraction on the collected leakage current signal based on Fast Fourier Transform (FFT) to obtain the third harmonic of the leakage current. The calculation formula is as follows: (1) In formula (1), is the third harmonic of the leakage current, is the real Fourier coefficient of the third harmonic, is the imaginary Fourier coefficient of the third harmonic; S22. Normalize the third harmonic of the leakage current, the ambient temperature, the vibration amplitude, and the operating voltage to obtain input data. The calculation formula for the normalization is as follows: (2) In formula (2), is the normalized value after normalization, is the original data in the dataset, is the minimum value in the data set, is the maximum value in the data set; the data set is a data set formed by the third harmonic of the leakage current, the ambient temperature, the vibration amplitude and the operating voltage; S23. Define the input data format as: (3) In formula (3), is the third harmonic of the leakage current, is the ambient temperature, is the vibration amplitude, is the operating voltage.

[0009] Furthermore, the arrester remaining service life prediction model based on the XGBoost algorithm pre-built in S3 includes the following steps: S31. Obtain multi-dimensional arrester data: 1) Obtain and pre-process historical arrester data at different operating times, including leakage current data, ambient temperature data, vibration amplitude data, and operating voltage data; 2) Extract features from leakage current data based on Fast Fourier Transform (FFT) to obtain the third harmonic data of leakage current; 3) Based on Pandas, the third harmonic current, vibration amplitude, operating voltage, and ambient temperature are standardized to obtain the target data; 4) Divide the target data into 70% training set, 20% validation set and 10% test set; S32. Construct a prediction model for the remaining service life of a lightning arrester based on a gradient boosting tree, and train it to learn the nonlinear relationship between the third harmonic of the leakage current, vibration amplitude, ambient temperature, and operating voltage and the remaining service life of the lightning arrester, specifically including the following: 1) Determine the objective function: (4) in, (5) (6) In formulas (4) to (6), is the objective function; is the loss function; is the regularization term; is the true value; is the predicted value; is the number of leaf nodes in the current tree; is the square of the weight of the leaf node, 、 is the regularization parameter; 2) Build a decision tree and calculate the gradient and the split gain of each node , the calculation formula is as follows: (7) (8) In formulas (7) to (8), 、 are the gradient sums of the left and right child nodes respectively, 、 are the sum of the second-order gradients of the left and right child nodes, is the regularization parameter; if the gain is less than , no splitting is performed; 3) Perform multiple rounds of iterative training and update the predicted value after each round of iterative training. The formula is as follows: (9) In formula (9), For samples In the The predicted value in the round iteration, For samples In the The predicted value in the round iteration, is the learning rate, In the t-th decision tree, input feature The corresponding predicted incremental value; where the sample is a set of data in the training set; S33. Input the test set into the trained model and output preliminary data on the predicted value of the remaining service life of the arrester.

[0010] Furthermore, the historical arrester data is pre-processed including: 3) Use wavelet denoising or bandpass filtering to remove high-frequency noise to eliminate errors caused by grid interference; 4) Eliminate extreme abnormal data through outlier detection method to improve the credibility of data; 3) All data are organized in time series to meet the needs of subsequent model training and prediction.

[0011] Furthermore, the multi-round iterative training process of the model is as follows: 1) Determine the initial hyperparameters based on the training set and target error range; 2) Define the training round k=1 and the maximum number of training iterations K; 3) Use the training set to train the model and obtain preliminary prediction results; 4) Dynamically adjust hyperparameters based on the validation set, including learning rate, tree depth, and regularization parameters; 5) Determine whether the maximum number of training iterations K has been reached: If not, continue optimizing the model's hyperparameters; if the training converges or reaches the maximum number of iterations, stop training and save the final optimized model.

[0012] Furthermore, S4 includes the following specific steps: S41. Collecting and preprocessing degradation data of the arrester to obtain target degradation data, wherein the degradation data includes historical remaining service life data, aging rate data, and degradation increment data of the arrester; S42, constructing a Wiener process model and performing parameter estimation to obtain a target Wiener process model; S43. Calculate the aging rate and degradation increment of the arrester based on the outputted predicted value of the remaining service life of the arrester; S44. The predicted value of the remaining service life of the lightning arrester, the aging rate and the degradation increment are input into the target Wiener process model as target degradation data to obtain the confidence interval of the predicted value of the remaining service life of the lightning arrester, which is used to judge the reliability of the predicted value of the remaining service life of the lightning arrester.

[0013] Furthermore, the parameter estimation in S42 includes: 1) Set initial conditions and iteration termination criteria: define the drift coefficient and diffusion coefficient , initialized by the mean or empirical value of historical data; set the maximum number of iterations and the convergence threshold ; 2) Iteratively optimize the parameters of the Wiener process model based on the degradation data of the arrester: Based on the degradation data of the arrester, the current parameters are calculated by the maximum likelihood estimation method. The likelihood function is: (10) In formula (10), is the degradation increment, is the time interval, is the aging rate, is the random fluctuation term in the aging process; among them, the parameters are updated by the gradient ascent method: (11) In formula (11), is the random fluctuation term in the aging process of the k+1th iteration, is the random fluctuation term in the aging process of the kth iteration, is the learning rate, It is the log-likelihood function and needs to be adjusted dynamically to strike a balance between convergence speed and stability; 3) Convergence judgment and dynamic adjustment: If the parameter change meets or reaches the maximum number of iterations K, the iteration ends and proceeds to the next step; otherwise, the learning rate is adjusted based on the current parameter estimation error: the adaptive learning rate algorithm Adagrad is used, and step 2) is repeated to further optimize the parameters; 4) Output the final optimized parameters and verify the model: ① Verify whether the residual sequence meets the independent and identically distributed assumption through degradation path residual analysis; ② Calculate the confidence interval coverage through Monte Carlo simulation to ensure that the confidence interval of the parameter estimate can effectively cover the true value.

[0014] Furthermore, in step 2), the ABC-Gibbs method can also be used: ① construct an approximate posterior distribution and alternately update y and y through Gibbs sampling; ② use the mean square error between the simulated data and the observed data of the degradation path as the acceptance criterion to ensure the stability of the parameter estimation.

[0015] Furthermore, S5 includes: S51. Based on the outputted predicted value of the remaining service life of the lightning arrester and the confidence interval of the predicted value of the remaining service life of the lightning arrester, determine whether the outputted predicted value of the remaining service life of the lightning arrester is reliable: if it is reliable, proceed to the next step; otherwise, directly determine that the working state is dangerous; S52. Determine whether the outputted predicted value of the remaining service life of the lightning arrester exceeds the set value of the third harmonic of the leakage current: if so, determine that the working state of the lightning arrester is dangerous and formulate a corresponding maintenance strategy; if not, determine that the working state of the lightning arrester is normal and perform regular monitoring.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1) Use non-invasive measurement devices to ensure the quality of data collection; 2) Collect the third harmonic, vibration amplitude, ambient temperature, and operating voltage of the arrester. Multi-dimensional features comprehensively characterize the arrester aging status and improve the accuracy of the arrester remaining service life prediction model; 3) The XGBoost algorithm is used to build a lightning arrester remaining service life prediction model. This algorithm can not only handle complex nonlinear relationships, but also has the ability to automatically select features and efficiently model. Compared with traditional feature determination methods, it has higher prediction accuracy and stronger adaptability. 4) Introducing the Wiener degradation process model into the constructed arrester remaining service life prediction model can effectively describe the uncertainty of arrester aging and provide a confidence interval for the arrester remaining service life prediction. This makes the arrester remaining service life prediction no longer a single fixed value, but a dynamic assessment including upper and lower bounds, which is used for intelligent operation and maintenance decision-making and optimization of equipment replacement cycle; 5) By combining the physical model with the data-driven model, the physical degradation mechanism data characteristics of the arrester are optimized, so that the model can maintain high prediction stability even when the data is incomplete or the operating conditions vary greatly; 6) The present invention achieves accurate, intelligent and stable life prediction of lightning arresters by integrating the strategy of multi-dimensional data + XGBoost algorithm + Wiener process, effectively making up for the shortcomings of traditional feature determination methods, while improving the adaptability of data-driven methods in complex power grid environments, providing a more scientific and reliable life prediction solution, and providing important technical support for the intelligent operation and maintenance of power grid equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention will be further described below with reference to the accompanying drawings and specific embodiments: Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 This is a trend diagram of the third harmonic of the leakage current in the present invention; Figure 3 It is the ambient temperature change trend diagram in the present invention; Figure 4 It is a vibration amplitude variation trend diagram in the present invention; Figure 5 This is a graph showing a trend of operating voltage changes in the present invention; Figure 6 It is a training flow chart of the arrester remaining service life prediction model based on the XGBoost algorithm in the present invention; Figure 7 It is the training accuracy optimization result of the arrester remaining service life prediction model constructed based on the XGBoost algorithm in the present invention; Figure 8 It is a schematic diagram of the confidence interval results obtained by the Wiener process model in the present invention. DETAILED DESCRIPTION Example

[0018] like Figure 1As shown, a method for predicting the remaining service life of a lightning arrester based on digital-analog linkage includes: S1. Obtaining real-time monitoring data of the arrester to be predicted; Specifically, the real-time monitoring data of the lightning arrester to be predicted includes the leakage current, ambient temperature, vibration amplitude, and operating voltage of the lightning arrester. The specific acquisition process is as follows: 1) Collecting leakage current: The arrester's leakage current signal is collected using a magnetic current sensor with a resolution of 0.01 mA and a portable digital oscilloscope with a sampling rate of 1 Msamples / s. The data format is set to time series. 2) Collecting ambient temperature: The temperature inside the arrester is collected as the ambient temperature by a high-precision temperature sensor installed on the surface of the arrester, and the sampling frequency is once per hour; 3) Collecting vibration amplitude: The vibration signal of the lightning arrester is collected by a piezoelectric vibration sensor with a sensitivity of 0.01g installed on the base of the lightning arrester. The sampling rate is 10kHz. The data is filtered and the average vibration amplitude is obtained; 4) Collecting the working voltage: Measure the actual operating voltage at both ends of the arrester through a voltage sensor and use the daily average of the actual operating voltage as the working voltage; Among them, some real-time monitoring data of the lightning arrester to be predicted collected at different time points are shown in Table 1: Table 1 Real-time monitoring data of some arresters to be predicted collected at different time points Time point Leakage current (μA) Vibration amplitude (g) Ambient temperature (°C) Operating voltage (V) 1 7.21 0.12 45.87 205.32 2 5.34 0.15 50.23 210.45 3 3.12 0.09 40.56 208.76 4 6.51 0.2 60.34 215.21 5 4.73 0.26 70.02 220.11 Note: Leakage current is collected using a non-invasive measurement method to avoid interference with equipment operation. A magnetic current sensor is used to measure the arrester's leakage current time domain signal. The magnetic current sensor has a resolution of 0.01mA and can capture tiny current changes. A portable digital oscilloscope is used for recording to ensure the accuracy of signal collection. The sampling rate is set to 1Msamples / s to ensure high resolution of the data. Vibration amplitude, ambient temperature, and operating voltage are collected using a direct contact measurement method.

[0019] S2. Preprocessing the real-time monitoring data of the lightning arrester to be predicted to obtain input data; Specifically, S21, based on the fast Fourier transform (FFT), the collected leakage current signal is subjected to feature extraction to obtain the third harmonic of the leakage current, which is calculated as follows: (1) In formula (1), is the third harmonic of the leakage current, is the real Fourier coefficient of the third harmonic, is the imaginary Fourier coefficient of the third harmonic; Among them, some of the third harmonic data extracted are shown in Table 2 (retain two decimal places): Table 2 Extracted third harmonic data of leakage current Time point Third harmonic of leakage current (μA) 1 13.14 2 12.52 3 14.08 4 15.26 5 16.19 S22. Normalize the third harmonic of the leakage current, the ambient temperature, the vibration amplitude, and the operating voltage to obtain input data. The calculation formula for the normalization is as follows: (2) In formula (2), is the normalized value after normalization, is the original data in the dataset, is the minimum value in the data set, is the maximum value in the data set; the data set is a data set formed by the third harmonic of the leakage current, the ambient temperature, the vibration amplitude and the operating voltage; Among them, some input data after normalization are shown in Table 3: Table 3 Part of the input data after normalization Time point Third harmonic of leakage current (μA) Vibration amplitude (g) Ambient temperature (°C) Operating voltage (V) 1 1 0.18 0.18 0 2 0.54 0.35 0.33 0.35 3 0 0 0 0.23 4 0.83 0.65 0.67 0.67 5 0.39 1 1 1 S23. Define the input data format as: (3) In formula (3), is the third harmonic of the leakage current, is the ambient temperature, is the vibration amplitude, is the operating voltage; Note: After data collection is complete, feature extraction of the signal is required to obtain key variables that characterize the aging status of the lightning arrester. The core content of feature extraction is to calculate the third harmonic of the leakage current, which is an important indicator of the degradation of the lightning arrester's insulation performance. After feature extraction is completed, statistical analysis of the third harmonic data at different operating times is performed. For example, for a lightning arrester that has been in operation for 8 years, its third harmonic is generally distributed between 6μA and 11μA, with a median of 8.5μA. This data can be used to establish a lightning arrester aging database and further used as training data for subsequent machine learning models.

[0020] S3. Input the input data into a pre-built arrester remaining service life prediction model based on the XGBoost algorithm, and output a predicted value of the arrester remaining service life; Specifically, the pre-built arrester remaining service life prediction model based on the XGBoost algorithm includes the following steps: S31. Obtain multi-dimensional arrester data: 1) Obtain and pre-process historical arrester data at different operating times, including leakage current data, ambient temperature data, vibration amplitude data, and operating voltage data; 2) Extract features from leakage current data based on Fast Fourier Transform (FFT) to obtain the third harmonic data of leakage current; 3) Based on Pandas, the third harmonic current, vibration amplitude, operating voltage, and ambient temperature are standardized to obtain the target data; 4) Divide the target data into 70% training set, 20% validation set and 10% test set; Among them, historical arrester data and preprocessing include: ① Use wavelet denoising or bandpass filtering to remove high-frequency noise to eliminate errors caused by grid interference; ② Eliminate extreme abnormal data through outlier detection method to improve the credibility of data; ③ All data are organized according to time series to meet the needs of subsequent model training and prediction; like Figures 2 to 5 The following graphs show the dynamic trends of the third harmonic of the arrester's leakage current, vibration amplitude, ambient temperature, and operating voltage. Data on the third harmonic, vibration amplitude, ambient temperature, and operating voltage of the leakage current, which operated between 2000 and 2020, are selected as historical arrester data. The third harmonic of the arrester's leakage current, vibration amplitude, ambient temperature, and operating voltage interact with each other and jointly determine the arrester's aging rate. The operating voltage fluctuates between 200V and 220V, influenced by factors such as grid load changes and short-term overvoltages. The overall upward trend in ambient temperature reflects the heat accumulation effect of long-term equipment operation. The nonlinear growth of the third harmonic leakage current further indicates that the arrester's aging process is not uniform, but rather accelerates during specific periods, possibly due to the effects of high temperature or voltage fluctuations. Therefore, accurately assessing the lifespan of an arrester based solely on a single variable is difficult. A comprehensive analysis of multiple factors can provide a more reliable health assessment.

[0021] like Figure 6 As shown, the multi-round iterative training process of the model is as follows: 1) Based on the training set and the target error range, determine the initial hyperparameters: the initial learning rate is set to 0.1, the maximum tree depth is set to limit the depth of a single tree to prevent overfitting, and the typical value is 6. The subsample ratio is used to control the random sampling ratio in each round of training, and the typical value is 0.8. 2) Define the training round k=1 and the maximum number of training iterations K: Start training with the first round and set the total number of training rounds. This is determined based on the training set and the target error range, with a typical value of 100-500. 3) Use the training set to train the model and obtain preliminary prediction results: Use the initially set hyperparameters to train the model. In each round of training, fit the residual between the current predicted value and the true value to build a new regression tree. Use the gradient update rule to optimize the predicted value and gradually reduce the model error. 4) Dynamically adjust hyperparameters based on the validation set, including learning rate, tree depth, and regularization parameters, to improve the generalization ability of the model and prevent overfitting or underfitting; 5) Determine whether the maximum number of training iterations K has been reached: If not, continue to optimize the model's hyperparameters; if the training converges or reaches the maximum number of iterations, stop training and save the final optimized model; In addition, after each round of training, the training round k is updated to k+1, and the model training step is returned to continue optimizing the model parameters and structure; like Figure 7 The figure shows the training accuracy optimization results of the arrester remaining service life prediction model based on the XGBoost algorithm. As the number of training rounds increases, the error gradually decreases and stabilizes after 40 rounds, indicating that the model has successfully learned the characteristic pattern of arrester aging and effectively fitted the complex relationship between vibration, ambient temperature, operating voltage and the remaining service life of the arrester. The rapid convergence of the training error shows that the arrester remaining service life prediction model based on the XGBoost algorithm can efficiently capture the nonlinear characteristics of the data and achieve better prediction results in fewer training rounds, making it have the potential for rapid deployment in actual power grid operation and maintenance; S32. Construct a prediction model for the remaining service life of a lightning arrester based on a gradient boosting tree, and train it to learn the nonlinear relationship between the third harmonic of the leakage current, vibration amplitude, ambient temperature, and operating voltage and the remaining service life of the lightning arrester, specifically including the following: 1) Determine the objective function: (4) in, (5) (6) In formulas (4) to (6), is the objective function; is the loss function; is the regularization term; is the true value; is the predicted value; is the number of leaf nodes in the current tree; is the square of the weight of the leaf node, 、 is the regularization parameter; 2) Build a decision tree and calculate the gradient and the split gain of each node , the calculation formula is as follows: (7) (8) In formulas (7) to (8), 、 are the gradient sums of the left and right child nodes respectively, 、 are the sum of the second-order gradients of the left and right child nodes, is the regularization parameter; if the gain is less than , no splitting is performed; 3) Perform multiple rounds of iterative training and update the predicted value after each round of iterative training. The formula is as follows: (9) In formula (9), For samples In the The predicted value in the round iteration, For samples In the The predicted value in the round iteration, is the learning rate, In the t-th decision tree, input feature The corresponding predicted incremental value; where the sample is a set of data in the training set; S33, inputting the test set into the trained model, and outputting preliminary data on the predicted value of the remaining service life of the arrester; In addition, after the model training is completed, the cross-validation method is used to optimize the hyperparameters to ensure that the model has good generalization ability; Description: After completing data preprocessing and feature extraction, a lightning arrester remaining service life prediction model based on a combination of data-driven methods and physical modeling methods is established. The model includes an input layer, a feature extraction layer, an XGBoost modeling layer, a digital-analog linkage optimization layer, and an output layer; a prediction model based on a gradient boosting tree is constructed to learn the nonlinear relationship between leakage current, operating voltage, and ambient temperature on the remaining service life of the lightning arrester; Wiener process modeling is introduced in the digital-analog linkage optimization layer to enhance the stability of the lightning arrester remaining service life prediction model.

[0022] S4. Based on the Wiener process model, the confidence interval of the predicted value of the remaining service life of the lightning arrester is obtained; Specifically, S41, collecting and preprocessing degradation data of the arrester to obtain target degradation data, wherein the degradation data includes historical remaining service life data, aging rate data, and degradation increment data of the arrester; S42, constructing a Wiener process model and performing parameter estimation to obtain a target Wiener process model; S43. Calculate the aging rate and degradation increment of the arrester based on the outputted predicted value of the remaining service life of the arrester; S44, using the predicted value of the remaining service life of the lightning arrester, the aging rate, and the degradation increment as target degradation data, inputting them into the target Wiener process model, and obtaining a confidence interval of the predicted value of the remaining service life of the lightning arrester, which is used to determine the reliability of the predicted value of the remaining service life of the lightning arrester; The parameter estimation in S42 includes: 1) Set initial conditions and iteration termination criteria: define the drift coefficient and diffusion coefficient , initialized by the mean or empirical value of historical data; set the maximum number of iterations and the convergence threshold ; 2) Iteratively optimize the parameters of the Wiener process model based on the degradation data of the arrester: Based on the degradation data of the arrester, the current parameters are calculated by the maximum likelihood estimation method. The likelihood function is: (10) In formula (10), is the degradation increment, is the time interval, is the aging rate, is the random fluctuation term in the aging process; among them, the parameters are updated by the gradient ascent method: (11) In formula (11), is the random fluctuation term in the aging process of the k+1th iteration, is the random fluctuation term in the aging process of the kth iteration, is the learning rate, It is the log-likelihood function and needs to be adjusted dynamically to strike a balance between convergence speed and stability; 3) Convergence judgment and dynamic adjustment: If the parameter change meets or reaches the maximum number of iterations K, the iteration ends and proceeds to the next step; otherwise, the learning rate is adjusted based on the current parameter estimation error: the adaptive learning rate algorithm Adagrad is used, and step 2) is repeated to further optimize the parameters; 4) Output the final optimized parameters and verify the model: ① Verify whether the residual sequence meets the independent and identically distributed assumption through degradation path residual analysis; ② Calculate the confidence interval coverage through Monte Carlo simulation to ensure that the confidence interval of the parameter estimate effectively covers the true value; In addition, in step 2), the ABC-Gibbs method can also be used: ① construct an approximate posterior distribution and alternately update and through Gibbs sampling; ② use the mean square error between the simulated data and the observed data of the degradation path as the acceptance criterion to ensure the stability of the parameter estimation.

[0023] like Figure 8 The figure shows the confidence interval for the predicted remaining useful life of a lightning arrester based on the Wiener process model. The black solid line represents the predicted trend, and the green area represents the 95% confidence interval. The confidence interval gradually widens over time, indicating that the uncertainty of the prediction increases with time. This uncertainty modeling approach can help power grid operators formulate more scientific maintenance plans, conduct maintenance in advance, avoid sudden arrester failures, and improve power grid reliability.

[0024] S5. Based on the predicted value of the remaining service life of the arrester and the set value, determine the working status of the arrester and formulate corresponding maintenance strategies: Specifically, S51, based on the outputted predicted value of the remaining service life of the lightning arrester and the confidence interval of the predicted value of the remaining service life of the lightning arrester, determine whether the outputted predicted value of the remaining service life of the lightning arrester is reliable: if reliable, proceed to the next step; otherwise, directly determine that the working state is dangerous; S52. Determine whether the output predicted value of the remaining service life of the lightning arrester exceeds the set value of the third harmonic of the leakage current: if it exceeds, the working state of the lightning arrester is determined to be dangerous and a corresponding maintenance strategy is formulated; if it does not exceed, the working state of the lightning arrester is determined to be normal and regular monitoring is required; wherein, the set value of the third harmonic of the leakage current is 100μA.

[0025] The constructed arrester remaining useful life prediction model was ultimately deployed to an online monitoring system, which monitors the arrester's health status in real time and dynamically updates the prediction results. Through the intelligent operation and maintenance system, the prediction results are used to make equipment maintenance decisions: the model is input for real-time predictions; the prediction results are regularly updated to generate operation reports. For arresters with a predicted lifespan below the safety threshold, early replacement is recommended to prevent grid failures. Furthermore, by combining historical data, the arrester replacement cycle can be optimized and a more scientific operation and maintenance plan can be developed, thereby improving the reliability and stability of the power system.

[0026] The embodiments described above are merely descriptions of preferred implementations of the present invention and are not intended to limit the scope of the present invention. Without departing from the principles and essence of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the scope of protection determined by the claims of the present invention.

Claims

1. A method for predicting the remaining service life of a lightning arrester based on digital-analog linkage, characterized in that: include: S1. Obtaining real-time monitoring data of the arrester to be predicted; S2. Preprocessing the real-time monitoring data of the lightning arrester to be predicted to obtain input data; S3. Input the input data into a pre-built arrester remaining service life prediction model based on the XGBoost algorithm, and output a predicted value of the arrester remaining service life; S4. Based on the target Wiener process model, the confidence interval of the predicted value of the remaining service life of the lightning arrester is output; S5. Based on the predicted value of the remaining service life of the lightning arrester and the set value, determine the working status of the lightning arrester and formulate a corresponding maintenance strategy.

2. The method for predicting the remaining service life of a lightning arrester based on digital-analog linkage according to claim 1 is characterized in that: The real-time monitoring data of the lightning arrester to be predicted in S1 include the arrester's leakage current, ambient temperature, vibration amplitude, and operating voltage. The specific acquisition process is as follows: 1) Collecting leakage current: The arrester's leakage current signal is collected using a magnetic current sensor with a resolution of 0.01 mA and a portable digital oscilloscope with a sampling rate of 1 Msamples / s. The data format is set to time series. 2) Collecting ambient temperature: The temperature inside the arrester is collected as the ambient temperature by a high-precision temperature sensor installed on the surface of the arrester, and the sampling frequency is once per hour; 3) Collecting vibration amplitude: The vibration signal of the lightning arrester is collected by a piezoelectric vibration sensor with a sensitivity of 0.01g installed on the base of the lightning arrester. The sampling rate is 10kHz. The data is filtered and the average vibration amplitude is obtained; 4) Collecting working voltage: Measure the actual operating voltage at both ends of the arrester through a voltage sensor and use the daily average of the actual operating voltage as the working voltage.

3. The method for predicting the remaining service life of a lightning arrester based on digital-analog linkage according to claim 2, characterized in that S2 include: S21. Perform feature extraction on the collected leakage current signal based on Fast Fourier Transform (FFT) to obtain the third harmonic of the leakage current. The calculation formula is as follows: （1） In formula (1), is the third harmonic of the leakage current, is the real Fourier coefficient of the third harmonic, is the imaginary Fourier coefficient of the third harmonic; S22. Normalize the third harmonic of the leakage current, the ambient temperature, the vibration amplitude, and the operating voltage to obtain input data. The calculation formula for the normalization is as follows: （2） In formula (2), is the normalized value after normalization, is the original data in the dataset, is the minimum value in the data set, is the maximum value in the data set; the data set is a data set formed by the third harmonic of the leakage current, the ambient temperature, the vibration amplitude and the operating voltage; S23. Define the input data format as: （3） In formula (3), is the third harmonic of the leakage current, is the ambient temperature, is the vibration amplitude, is the operating voltage.

4. The method for predicting the remaining service life of a lightning arrester based on digital-analog linkage according to claim 1 is characterized in that: The XGBoost algorithm-based remaining useful life prediction model for lightning arresters pre-built in S3 includes the following steps: S31. Obtain multi-dimensional arrester data: 1) Obtain and pre-process historical arrester data at different operating times, including leakage current data, ambient temperature data, vibration amplitude data, and operating voltage data; 2) Extract features from leakage current data based on Fast Fourier Transform (FFT) to obtain the third harmonic data of leakage current; 3) Based on Pandas, the third harmonic current, vibration amplitude, operating voltage, and ambient temperature are standardized to obtain the target data; 4) Divide the target data into 70% training set, 20% validation set and 10% test set; S32. Construct a prediction model for the remaining service life of a lightning arrester based on a gradient boosting tree, and train it to learn the nonlinear relationship between the third harmonic of the leakage current, vibration amplitude, ambient temperature, and operating voltage and the remaining service life of the lightning arrester, specifically including the following: 1) Determine the objective function: （4） in, （5） （6） In formulas (4) to (6), is the objective function; is the loss function; is the regularization term; is the true value; is the predicted value; is the number of leaf nodes in the current tree; is the square of the weight of the leaf node, 、 is the regularization parameter; 2) Build a decision tree and calculate the gradient and the split gain of each node , the calculation formula is as follows: （7） （8） In formulas (7) to (8), 、 are the gradient sums of the left and right child nodes respectively, 、 are the sum of the second-order gradients of the left and right child nodes, is the regularization parameter; if the gain is less than , no splitting is performed; 3) Perform multiple rounds of iterative training and update the predicted value after each round of iterative training. The formula is as follows: （9） In formula (9), For samples In the The predicted value in the round iteration, For samples In the The predicted value in the round iteration, is the learning rate, In the t-th decision tree, input feature The corresponding predicted incremental value; where the sample is a set of data in the training set; S33. Input the test set into the trained model and output preliminary data on the predicted value of the remaining service life of the arrester.

5. The method for predicting the remaining service life of a lightning arrester based on digital-analog linkage according to claim 4 is characterized in that: The preprocessing of historical arrester data includes: 3) Use wavelet denoising or bandpass filtering to remove high-frequency noise to eliminate errors caused by grid interference; 4) Eliminate extreme abnormal data through outlier detection method to improve the credibility of data; 3) All data are organized in time series to meet the needs of subsequent model training and prediction.

6. The method for predicting the remaining service life of a lightning arrester based on digital-analog linkage according to claim 4 is characterized in that: The multi-round iterative training process of the model is as follows: 1) Determine the initial hyperparameters based on the training set and target error range; 2) Define the training round k=1 and the maximum number of training iterations K; 3) Use the training set to train the model and obtain preliminary prediction results; 4) Dynamically adjust hyperparameters based on the validation set, including learning rate, tree depth, and regularization parameters; 5) Determine whether the maximum number of training iterations K has been reached: If not, continue to optimize the model's hyperparameters; if the training converges or reaches the maximum number of iterations, stop training and save the final optimized model.

7. The method for predicting the remaining service life of a lightning arrester based on digital-analog linkage according to claim 1 is characterized in that: S4 includes the following steps: S41, collecting and preprocessing degradation data of the arrester to obtain target degradation data, wherein the degradation data includes historical remaining service life data, aging rate data, and degradation increment data of the arrester; S42, constructing a Wiener process model and performing parameter estimation to obtain a target Wiener process model; S43. Calculate the aging rate and degradation increment of the arrester based on the outputted predicted value of the remaining service life of the arrester; S44. The predicted value of the remaining service life of the lightning arrester, the aging rate and the degradation increment are input into the target Wiener process model as target degradation data to obtain the confidence interval of the predicted value of the remaining service life of the lightning arrester, which is used to judge the reliability of the predicted value of the remaining service life of the lightning arrester.

8. The method for predicting the remaining service life of a lightning arrester based on digital-analog linkage according to claim 7 is characterized in that: The parameter estimation in S42 includes: 1) Set initial conditions and iteration termination criteria: define the drift coefficient and diffusion coefficient , initialized by the mean or empirical value of historical data; set the maximum number of iterations and the convergence threshold ; 2) Iteratively optimize the parameters of the Wiener process model based on the degradation data of the arrester: Based on the degradation data of the arrester, the current parameters are calculated by the maximum likelihood estimation method. The likelihood function is: （10） In formula (10), is the degradation increment, is the time interval, is the aging rate, is the random fluctuation term in the aging process; among them, the parameters are updated by the gradient ascent method: （11） In formula (11), is the random fluctuation term in the aging process of the k+1th iteration, is the random fluctuation term in the aging process of the kth iteration, is the learning rate, It is the log-likelihood function and needs to be adjusted dynamically to strike a balance between convergence speed and stability; 3) Convergence judgment and dynamic adjustment: If the parameter change meets or reaches the maximum number of iterations K, the iteration ends and proceeds to the next step; otherwise, the learning rate is adjusted based on the current parameter estimation error: the adaptive learning rate algorithm Adagrad is used, and step 2) is repeated to further optimize the parameters; 4) Output the final optimized parameters and verify the model: ① Verify whether the residual sequence meets the independent and identically distributed assumption through degradation path residual analysis; ② Calculate the confidence interval coverage through Monte Carlo simulation to ensure that the confidence interval of the parameter estimate can effectively cover the true value.

9. The method for predicting the remaining service life of a lightning arrester based on digital-analog linkage according to claim 8, characterized in that: In step 2), the ABC-Gibbs method can also be used: ① construct an approximate posterior distribution and alternately update and through Gibbs sampling; ② use the mean square error between the simulated data and the observed data of the degradation path as the acceptance criterion to ensure the stability of the parameter estimation.

10. The method for predicting the remaining service life of a lightning arrester based on digital-analog linkage according to claim 1, wherein S5 include: S51. Based on the outputted predicted value of the remaining service life of the lightning arrester and the confidence interval of the predicted value of the remaining service life of the lightning arrester, determine whether the outputted predicted value of the remaining service life of the lightning arrester is reliable: if it is reliable, proceed to the next step; otherwise, directly determine that the working state is dangerous; S52. Determine whether the outputted predicted value of the remaining service life of the lightning arrester exceeds the set value of the third harmonic of the leakage current: if so, determine that the working state of the lightning arrester is dangerous and formulate a corresponding maintenance strategy; if not, determine that the working state of the lightning arrester is normal and perform regular monitoring.

Images