Intelligent online detector for 10kV lightning arrester

Abstract

The invention discloses an intelligent online detector for a 10kV lightning arrester, and the detector comprises a data collection module which is used for obtaining a DC volt-ampere characteristic curve of the lightning arrester under different test conditions; the data preprocessing module is used for eliminating the linear or nonlinear trend of the volt-ampere characteristic curve and eliminating abnormal values; the piecewise linear fitting module is used for segmenting the volt-ampere characteristic curve and performing linear fitting on data of each segment to obtain linear parameters of each segment; the parameter optimization and separation module is used for optimizing the piecewise linear fitting result, separating the surface leakage current and the valve plate current data correction and calculation module, and performing correction calculation based on the optimized parameters to obtain key electrical parameters; and the state evaluation and early warning module is used for evaluating the operation state of the lightning arrester according to the corrected electrical parameters and providing early warning information. According to the method, instability caused by high-order derivative calculation can be avoided, and the accuracy and stability of an evaluation result are ensured.

Description

Technical Field

[0001] The invention relates to the technical field of lightning protection, and in particular to an intelligent online detector for a 10kV lightning arrester. Background Art

[0002] In modern power systems, metal oxide arresters (MOAs) are key overvoltage protection devices and are widely used in substations and transmission and distribution lines. However, as the operating time increases, arresters may age, become damp, and so on, leading to increased leakage current. In severe cases, they may even cause accidents such as overheating and explosion of the arrester. Therefore, it is extremely important to ensure the safe operation of the arrester and monitor its status in a timely manner.

[0003] Traditional arrester detection methods (Chinese patent, application number 202410560244.9) usually rely on the analysis of high-order derivatives of the volt-ampere characteristic curve. However, the calculation of high-order derivatives has high requirements on data quality and is easily affected by noise and abnormal data, resulting in unstable calculation results. Especially in field measurements, the accuracy and reliability of the calculation results are further reduced due to interference from factors such as complex environment and severe surface pollution. In addition, the calculation process of high-order derivatives is complicated, which easily leads to error accumulation in the analysis process, which is not conducive to accurate status assessment and early warning of the arrester. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a 10kV lightning arrester intelligent online detector.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] 10kV arrester intelligent online detector, including: data acquisition module, used to obtain the DC volt-ampere characteristic curve of the arrester under different test conditions;

[0007] Data preprocessing module, used to eliminate the linear or nonlinear trend of the volt-ampere characteristic curve and exclude abnormal values;

[0008] The segmented linear fitting module is used to segment the volt-ampere characteristic curve and perform linear fitting on the data of each segment to obtain the linear parameters of each segment;

[0009] Parameter optimization and separation module, used to optimize the piecewise linear fitting results and separate the surface leakage current and valve current

[0010] The data correction and calculation module performs correction calculation based on the optimized parameters to obtain key electrical parameters;

[0011] The status assessment and early warning module assesses the operating status of the arrester based on the corrected electrical parameters and provides early warning information.

[0012] Further: the data preprocessing module includes:

[0013] Apply a low-pass filter to remove high-frequency noise in the volt-ampere characteristic curve and set the filter cutoff frequency f c , and use the filter design formula to calculate the filter parameters:

[0014]

[0015] Where H(f) is the frequency response of the filter, f is the signal frequency, and f c is the cutoff frequency, n is the order of the filter;

[0016] Use polynomial fitting to remove linear or nonlinear trends in the volt-ampere characteristic curve and make the data smoother:

[0017] V fit (x) = a 2 x 2 +a 1 x+a 0

[0018] Among them, V fit (x) is the trend line after fitting, a 2 、a 1 、a 0 is the fitting coefficient;

[0019] The method based on standard deviation is used to detect outliers. The mean μ and standard deviation σ of the data are calculated, and the threshold of outliers is defined as μ±kσ, where k is the threshold coefficient;

[0020] Output volt-ampere characteristic curve data (V i ,I i ).

[0021] Further: the piecewise linear fitting module includes a piecewise algorithm module:

[0022] The sliding window method is used to calculate the volt-ampere characteristic curve data (V i ,I i ), set a window size w, and calculate the slope s in each window:

[0023]

[0024] where s j is the slope in the jth window, I j+w and I j are the current value in the window, V j+w and V j They are voltage values;

[0025] Calculate the change in the slope of the continuous window, determine the change point, and set a change threshold Δs:

[0026] Δs j =|s j -s j-1 |

[0027] If Δs j If it exceeds the set threshold Δs, the point is considered to be a change point;

[0028] Mark the detected slope change point as the starting point or ending point of the segment;

[0029] If the data change between two change points is less than a certain threshold, these points are merged into one segment and the merging threshold ε is set;

[0030] Merge judgment = Δs j <ε

[0031] If the conditions are met, these change points are merged into one segment;

[0032] According to the detected change points and merging rules, the starting point and ending point of each segment are output;

[0033] The information of the starting and ending points of each segment, including the voltage range of each segment [V start , V end ] and current range [I start , I end ].

[0034] Further: the piecewise linear fitting module includes a linear fitting module:

[0035] According to the output of the segmentation algorithm module, the data of each segment (V i ,I i ) is extracted; each segmented data includes [V start , V end ] to [I start , I end ] voltage and current values;

[0036] Perform linear regression on the data of each segment to obtain the linear relationship:

[0037] I=aV+b

[0038] Where a is the slope and b is the intercept;

[0039] The least squares method is used to solve the parameters of the linear model. For each segmented data (V i ,I i ), construct the loss function:

[0040]

[0041] Where n is the number of data points in the segment, and a and b are fitting parameters;

[0042] By solving the least squares equation, we can get the optimal a and b:

[0043]

[0044] Output the fitted linear parameters a and b for each segment, as well as the fitted straight line equation I=aV+b;

[0045] Output format:

[0046] I fit (V) = a j V+b j

[0047] Among them, a j and b j are the slope and intercept of the jth segment.

[0048] Further: the piecewise linear fitting module includes a fitting quality assessment module:

[0049] Calculate the residual of each data point, for each segmented data point (V i ,I i ), calculate the actual current value I i With the fitting current value I fit The residual between (V):

[0050] Residual i =I i -(a j V i +b j )

[0051] Among them, I i is the actual current value, (a j V i +b j is the fitted current value calculated by the fitting model;

[0052] Calculate the sum of squared residuals for all data points within each segment to measure the overall error of the fit:

[0053]

[0054] Among them, n j is the number of data points in the jth segment;

[0055] The root mean square error is calculated from the residual sum of squares as an indicator of the quality of the fit:

[0056]

[0057] Where RMSE j is the root mean square error of the jth segment;

[0058] The accuracy of the fitting is determined based on the calculated root mean square error value;

[0059] The RMSE value of each segment is output and compared with the preset threshold to determine whether the quality of the fit meets the requirements.

[0060] Further: the parameter optimization and separation module includes a nonlinear optimization module:

[0061] Construct an optimization objective function, which is used to measure the quality of the fitting results and is expressed as the sum of the residual squares of all segments:

[0062]

[0063] Where m is the number of segments, n j is the number of data points in the jth segment;

[0064] Defining constraints includes:

[0065] a j,min ≤a j ≤a j,max

[0066] b j,min ≤ b j ≤ b j,max

[0067] Among them, a j,min and a j,max are the minimum and maximum values ​​of the slope of the jth segment, b j,min and b j,max are the minimum and maximum values ​​of the intercept;

[0068] The fitting parameter a j and b j Encoded as chromosomes;

[0069] The objective function F is used as the fitness function;

[0070] Genetic operations, including selection, crossover, and mutation, generate new candidate solutions and update the population;

[0071] Fitness function calculation:

[0072]

[0073] Run the selected optimization algorithm to iteratively optimize the fitting parameters a according to the optimization objective function and constraints j and b j ;

[0074] Record the objective function value of each iteration and update the best solution;

[0075] According to the results of the optimization algorithm, the optimal linear fitting parameters for each segment are obtained. and

[0076] Output:

[0077]

[0078] Further: the parameter optimization and separation module includes a surface leakage current separation model:

[0079] Define the surface leakage current model, assuming that the relationship between the surface leakage current and voltage is nonlinear, and use a predefined nonlinear function to represent it:

[0080]

[0081] where c 1 、c 2 , and c 3 is a model parameter that characterizes the surface leakage current;

[0082] The valve current model is defined as follows:

[0083] I core (V)=I opt (V)-I surface (V)

[0084] Among them, I core (V) is the calculated valve current;

[0085] The parameter c of the surface leakage current model is optimized by an optimization algorithm. 1 、c 2 , and c 3 Fitting is performed to minimize the error of the surface leakage current model, and the objective function is:

[0086]

[0087] Among them, I surface,i is the current predicted by the surface leakage current model, I measured,i is the actual measured surface leakage current.

[0088] Further: the parameter optimization and separation module includes a parameter correction module:

[0089] Enter actual test data for calibration:

[0090] The measured voltage V i , the measured current I measured (V i );

[0091] Establish a correction objective function, which is used to measure the difference between the actual detection data and the model prediction data:

[0092]

[0093] Among them, I predicted (V i ) is the current predicted by the separation model:

[0094] I predicted (V i )=I surface (V i )+I core (V i )

[0095] The goal is to adjust the model parameters so that F calib minimize;

[0096] The parameters optimized during the calibration process include the parameters c of the surface leakage current model 1 、c 2 , and c 3 , and the parameters of the piecewise linear fit and

[0097] By iteratively updating the model parameters, the objective function F is minimized. calib , the update rule is:

[0098] δθ=-(J T J) -1 J T r

[0099] Among them, δθ is the parameter update amount, J is the Jacobian matrix, and r is the residual vector;

[0100] In each iteration, the predicted current I is calculated based on the current parameter values predicted (V i ), update the objective function value F calib ;

[0101] Update the parameters according to the optimization algorithm and repeat the iteration until the objective function converges to the minimum value;

[0102] The predicted current is recalculated using the corrected model parameters and compared with the actual measured data;

[0103] Output corrected parameters, including parameters of surface leakage current model and and the parameters of the piecewise linear fit and

[0104] Further: the data correction and calculation module includes:

[0105] Based on the corrected parameters, calculate the voltage value that makes the current 1mA:

[0106]

[0107] Among them, I 1maA =1mA;

[0108] Applying an additional correction factor to adjust the calculated voltage, the corrected formula is:

[0109]

[0110] Among them, K temp is the temperature correction coefficient, ΔT is the temperature deviation;

[0111] Based on the corrected voltage value, calculate the actual leakage current when the voltage is 0.75 times U1mA:

[0112]

[0113] Among them, I surface and I core are the current values ​​calculated by the surface leakage current model and the valve current model respectively;

[0114] Apply an additional correction factor to adjust the calculated leakage current value:

[0115]

[0116] Among them, k meas is the measurement correction factor, ΔM is the measurement error deviation;

[0117] Compare the calculated 1mA voltage and actual leakage current at 0.75 times U1mA with historical or experimental data to evaluate its accuracy. Compare the formula:

[0118]

[0119] Among them, U 1mA,hist and I 75％1mA,hist is the corresponding value in the historical data;

[0120] Outputs corrected voltage and leakage current data.

[0121] Compared with the prior art, the present invention has the following technical advances:

[0122] The prior art relies on calculating the high-order derivatives of the volt-ampere characteristic curve. Since the high-order derivatives are very sensitive to the noise and fluctuation of the data, the calculation results may be unstable, resulting in inaccurate assessment of the arrester state. The present application replaces the calculation of the high-order derivatives with a piecewise linear fitting module. The piecewise linear fitting decomposes the complex nonlinear volt-ampere characteristic curve into multiple linear segments for processing, avoiding the instability caused by the high-order derivative calculation and ensuring the accuracy and stability of the assessment results. The high-order derivatives have high requirements on the smoothness of the data. Any noise or outliers may significantly affect the derivative calculation results, reducing the reliability of the detection. The present application effectively removes the noise through the data preprocessing module, and further reduces the dependence on the smoothness of the data through piecewise linear fitting, making the system more robust to noise and outliers. The prior art analyzes the volt-ampere characteristics of the valve plate based on the high-order derivatives, which is easily affected by the surface leakage current, resulting in result deviation. The present application introduces a parameter optimization and separation module, which uses the optimization algorithm and model separation technology to accurately separate the surface leakage current and the valve plate current, further reduce the influence of surface contamination on the detection results, and ensure the accuracy of parameter calculation.

[0123] The existing technology relies on the analysis of high-order derivatives, which requires complex calculations and careful verification of the results. The process is cumbersome and prone to errors. By using piecewise linear fitting and optimization algorithms, the present application significantly simplifies the data processing process, avoids cumbersome high-order derivative calculations, and improves detection efficiency and ease of use. The results of high-order derivative calculations are limited to local analysis of the curve and fail to fully consider the overall operating status of the lightning arrester. The status assessment and early warning module of the present application comprehensively analyzes multiple parameters and uses the optimized model to perform global status assessment, providing more comprehensive status reports and early warning information, which helps to discover potential problems of lightning arresters at an earlier stage. BRIEF DESCRIPTION OF THE DRAWINGS

[0124] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0125] In the attached picture:

[0126] Figure 1 is a system structure diagram of the present invention; DETAILED DESCRIPTION

[0127] The following specific embodiments may be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments. The embodiments of the present invention will be described below in conjunction with the accompanying drawings.

[0128] like Figure 1 As shown, the present invention discloses a 10kV lightning arrester intelligent online detector, comprising:

[0129] Data acquisition module, used to obtain DC volt-ampere characteristic curve of the arrester under different test conditions;

[0130] Data preprocessing module, used to eliminate the linear or nonlinear trend of the volt-ampere characteristic curve and exclude abnormal values;

[0131] The segmented linear fitting module is used to segment the volt-ampere characteristic curve and perform linear fitting on the data of each segment to obtain the linear parameters of each segment;

[0132] Parameter optimization and separation module, used to optimize the piecewise linear fitting results and separate the surface leakage current and valve current

[0133] The data correction and calculation module performs correction calculation based on the optimized parameters to obtain key electrical parameters;

[0134] The status assessment and early warning module assesses the operating status of the arrester based on the corrected electrical parameters and provides early warning information.

[0135] The raw data provided by the data acquisition module is transmitted to the data preprocessing module for noise elimination and outlier processing.

[0136] The preprocessed data enters the piecewise linear fitting module for fitting analysis, replacing the calculation of high-order derivatives.

[0137] The fitting results are further optimized and separated into current parameters through the parameter optimization and separation module.

[0138] The optimized parameters are passed to the data correction and calculation module for correction calculation to obtain key electrical parameters.

[0139] The calculated parameters are used in the status assessment and early warning module to assess the arrester status and generate early warning information.

[0140] Specifically, the data acquisition module includes:

[0141] The sensor array, including high-precision voltage sensors and current sensors, is used to monitor the voltage and current of the arrester in real time. The sensor array collects the following specific raw data and transmits them to the data preprocessing module:

[0142] DC volt-ampere characteristic curve data: that is, the current value corresponding to different voltages, forming a complete volt-ampere characteristic curve, which is used to analyze the electrical characteristics of the arrester.

[0143] Voltage data: includes real-time voltage data under DC 1mA test conditions, which is used for subsequent correction and calculation.

[0144] Current data: includes the leakage current data actually measured during the test, especially the leakage current values ​​at different voltages.

[0145] The data acquisition card is responsible for converting the analog signals collected by the sensor into digital signals and transmitting them to the data processing module.

[0146] Specifically, the data preprocessing module includes:

[0147] The collected raw data is preliminarily processed to improve data quality and remove noise and outliers. The ultimate goal is to generate high-quality volt-ampere characteristic data and reduce reliance on higher-order derivatives.

[0148] 1. Filtering:

[0149] Apply a low-pass filter (such as Butterworth filter) to remove high-frequency noise from the signal and set the filter cutoff frequency f c , and use the filter design formula to calculate the filter parameters:

[0150]

[0151] Where H(f) is the frequency response of the filter, f is the signal frequency, and f c is the cutoff frequency and n is the order of the filter.

[0152] 2. Detrending:

[0153] Use polynomial fitting (such as quadratic fitting) to remove linear or nonlinear trends in the data and make the data smoother:

[0154] V fit (x) = a 2 x 2 +a 1 x+a 0

[0155] Among them, V fit (x) is the trend line after fitting, a 2 、a 1 、a 0 are the fitting coefficients.

[0156] 3. Outlier detection and processing:

[0157] The outliers are detected based on the standard deviation method. The mean μ and standard deviation σ of the data are calculated. The threshold of the outliers is defined as μ±kσ, where k is the threshold coefficient (usually 2 or 3). The outliers can be processed by correction or deletion.

[0158] Output: Smoothed and detrended volt-ampere characteristic curve data (V i ,I i ).

[0159] The data preprocessing module filters, removes trends and performs outlier detection on the DC volt-ampere characteristic curve data, voltage data and current data, and finally outputs high-quality volt-ampere characteristic data, corrected voltage data and current data.

[0160] Specifically, the piecewise linear fitting module includes a piecewise algorithm module:

[0161] 1. Data preparation

[0162] Based on the output of the data preprocessing module, the smoothed and detrended volt-ampere characteristic curve data (V i ,I i ), that is, the voltage and the corresponding current value.

[0163] 2. Change point detection

[0164] Calculate the change in slope of the data:

[0165] Use the sliding window method to calculate the local slope of the data, set a window size w, and calculate the slope s in each window:

[0166]

[0167] where s j is the slope in the jth window, I j+w and I j are the current value in the window, V j+w and V j are the voltage values ​​respectively.

[0168] Slope change detection:

[0169] Calculate the change in the slope of the continuous window, determine the change point, and set a change threshold Δs:

[0170] Δs j =|s j -s j-1 |

[0171] If Δs j If it exceeds the set threshold Δs, the point is considered to be a change.

[0172] 3. Determine the starting and ending points of the segment

[0173] Mark the detected slope change points as the start or end points of the segments.

[0174] If the data change between two change points is not significant (i.e. the slope change is less than a certain threshold), these points are merged into one segment and the merging threshold ε is set;

[0175] Merge judgment = Δs j <ε

[0176] If the conditions are met, these change points are merged into one segment.

[0177] 4. Output of segment start and end points

[0178] According to the detected change points and merging rules, the starting and ending points of each segment are output. These points will be used for subsequent linear fitting.

[0179] Output: information about the starting and ending points of each segment, including the voltage range of each segment [V start , V end ] and current range [I start , I end ].

[0180] The segmented algorithm module automatically determines the segmented starting and ending points of the volt-ampere characteristic curve through slope change detection and change point analysis. These segmented points will be used for linear fitting in the next step of the segmented linear fitting module, thereby replacing the reliance on high-order derivatives and performing accurate volt-ampere characteristic analysis.

[0181] The piecewise linear fitting module includes the linear fitting module:

[0182] 1. Input Data

[0183] Based on the segmentation algorithm module, the starting point and end point of the segmentation have been obtained. The data preprocessing module provides smooth and detrended volt-ampere characteristic curve data (V i ,I i ), including the voltage and current values ​​for each segment.

[0184] 2. Linear fitting steps

[0185] 1. Extract segmented data:

[0186] According to the output of the segmentation algorithm module, the data of each segment (V i ,I i ) is extracted. Each segmented data includes [V start , V end ] to [I start , Iend ] voltage and current values.

[0187] 2. Build a linear model:

[0188] Perform linear regression on the data of each segment to obtain the linear relationship:

[0189] I=aV+b

[0190] Where a is the slope and b is the intercept.

[0191] 3. Least squares fitting:

[0192] The least squares method is used to solve the parameters of the linear model. The goal of the least squares method is to minimize the residual sum of squares of the fitting function. For each segmented data (V i ,I i ), construct the loss function:

[0193]

[0194] Where n is the number of data points in the segment, and a and b are the fitting parameters.

[0195] 4. Solve for linear parameters:

[0196] By solving the least squares equation, we can get the optimal a and b. The equation can be obtained by analytical solution. The analytical solution formula is:

[0197]

[0198] 5. Fitting result output:

[0199] Output the fitted linear parameters a and b for each segment, as well as the fitted straight line equation I=aV+b;

[0200] Output format:

[0201] I fit (V) = a j V+b j

[0202] Among them, a j and b j are the slope and intercept of the jth segment.

[0203] The piecewise linear fitting module includes a fitting quality assessment module

[0204] 1. Input Data

[0205] Based on the linear fitting module, the linear fitting parameters (slope a and intercept b) of each segment and the fitting straight line equation have been obtained:

[0206] Ifit (V) = a j V+b j

[0207] Among them, a j and b j are the slope and intercept of the jth segment.

[0208] 2. Calculate the residual

[0209] Compute the residual for each data point:

[0210] For each segmented data point (V i ,I i ), calculate the actual current value I i With the fitting current value I fit The residual between (V):

[0211] Residual i =I i -(a j V i +b j )

[0212] Among them, I i is the actual current value, a j V i +b j ) is the fitted current value calculated by the fitting model.

[0213] Residual sum of squares:

[0214] Calculate the sum of squared residuals for all data points within each segment to measure the overall error of the fit:

[0215]

[0216] Among them, n j is the number of data points in the jth segment.

[0217] Calculate the root mean square error (RMSE)

[0218] The root mean square error is calculated from the residual sum of squares as an indicator of the quality of the fit:

[0219]

[0220] Where RMSE j is the root mean square error of the jth segment.

[0221] The root mean square error (RMSE) value and fit quality assessment result (such as "excellent" or "needs improvement") for each segment.

[0222] The fitting quality assessment module evaluates the accuracy of each piecewise linear fit by calculating the residual and root mean square error (RMSE). These evaluation results ensure the accuracy and reliability of the fitting process and provide a basis for subsequent analysis and processing.

[0223] Specifically, the parameter optimization and separation module includes a nonlinear optimization module:

[0224] 1. Input Data

[0225] Based on the piecewise linear fitting module, the linear fitting parameters a of each segment have been obtained. j and b j , that is, the linear relationship between current and voltage in each segment:

[0226] I fit (V) = a j V+b j

[0227] 2. Define optimization goals

[0228] Construct the optimization objective function:

[0229] The optimization objective function is used to measure the quality of the fitting results, usually the weighted sum of the fitting residuals. The objective function F can be expressed as the sum of the residual squares of all segments:

[0230]

[0231] Where m is the number of segments, n j is the number of data points in the jth segment.

[0232] Define constraints:

[0233] In practical applications, it is necessary to impose constraints on the optimization parameters, such as the physical range of the parameters or other prior knowledge restrictions. The constraints include:

[0234] a j,min ≤a j ≤a j,max

[0235] b j,min ≤ b j ≤ b j,max

[0236] Among them, a j,min and a j,max are the minimum and maximum values ​​of the slope of the jth segment, b j,min and b j,max are the minimum and maximum values ​​of the intercept.

[0237] 3. Implement optimization algorithms

[0238] Select an optimization algorithm:

[0239] Genetic Algorithm:

[0240] Coding: Fitting parameter a j and b j Encoded as chromosomes.

[0241] Fitness function: The objective function F is used as the fitness function.

[0242] Genetic operations: including selection, crossover and mutation, generating new candidate solutions, and updating the population.

[0243] Fitness function calculation:

[0244]

[0245] Execute the optimization process:

[0246] Run the selected optimization algorithm to iteratively optimize the fitting parameters a according to the optimization objective function and constraints j and b j .

[0247] The objective function value of each iteration is recorded and the best solution is updated.

[0248] 4. Output optimization results

[0249] According to the results of the optimization algorithm, the optimal linear fitting parameters for each segment are obtained. and

[0250] Output:

[0251]

[0252] The nonlinear optimization algorithm module uses a genetic algorithm to globally optimize the piecewise linear fitting results. By constructing an optimization objective function and imposing constraints, the optimization algorithm can improve the accuracy of current separation and reduce the impact of environmental factors, providing more accurate parameters for the subsequent surface leakage current separation model.

[0253] Parameter optimization and separation module including surface leakage current separation model

[0254] 1. Input Data

[0255] Based on the nonlinear optimization algorithm module, the optimized linear fitting parameters for each segment have been obtained. and That is, the optimized current-voltage linear relationship of each segment:

[0256]

[0257] 2. Establish a separation model

[0258] Define the surface leakage current model:

[0259] Assuming that the relationship between surface leakage current and voltage is nonlinear, it can be represented by a predefined nonlinear function, such as a logarithmic function or an exponential function. The surface leakage current model is:

[0260]

[0261] where c 1 、c 2 , and c 3 is a model parameter that characterizes the surface leakage current.

[0262] Define the valve current model:

[0263] The relationship between valve current and voltage is usually linear, which is given by the optimized piecewise linear fitting parameters. The valve current is:

[0264] I core (V)=I opt (V)-I surface (V)

[0265] Among them, I core (V) is the calculated valve current.

[0266] 3. Parameter estimation and model fitting

[0267] Fit the surface leakage current model:

[0268] The parameter c of the surface leakage current model is optimized by an optimization algorithm. 1 、c 2 , and c 3 Fitting is performed to minimize the error of the surface leakage current model, and the objective function is:

[0269]

[0270] Among them, I surface,i is the current predicted by the surface leakage current model, I measured,i is the actual measured surface leakage current.

[0271] Adjust the valve current model:

[0272] Calculation of surface leakage current I surface (V), the valve current at each voltage is obtained:

[0273] I core (V)=I opt (V)-I surface (V)

[0274] Ensure that the valve current model can accurately reflect the actual current value of the valve.

[0275] The surface leakage current separation model module separates the surface leakage current and the valve current by establishing the surface leakage current model and the valve current model. By fitting the model parameters and the correction process, the accuracy of the separation results is ensured, providing reliable data for the operation status analysis of the arrester.

[0276] The parameter optimization and separation module includes the parameter correction module:

[0277] 1. Input Data

[0278] Based on the surface leakage current separation model module, the separated current data has been obtained, including:

[0279] Surface leakage current I surface (V);

[0280] Valve current I core (V);

[0281] The relevant fitting parameter c 1 、c 2 , and c 3 (parameters of surface leakage current model);

[0282] Linear fitting parameters and (Optimization results for each segment);

[0283] In addition, actual test data must be input for correction: the measured voltage V i and the measured current I measured (V i )

[0284] 2. Correction objective function definition

[0285] Establish the correction objective function:

[0286] The calibration objective function is used to measure the difference between the actual detection data and the model prediction data:

[0287]

[0288] Among them, I predicted (V i ) is the current predicted by the separation model:

[0289] I predited (V i )=I surface (V i )+I core(V i )

[0290] The goal is to adjust the model parameters so that F calib minimize.

[0291] Define optimization parameters:

[0292] The parameters to be optimized during the calibration process include the parameters c of the surface leakage current model 1 、c 2 , and c 3 , and the parameters of the piecewise linear fit and

[0293] 3. Implement parameter correction

[0294] Optimization process: Update the model parameters iteratively to minimize the objective function F calib , update rules:

[0295] δθ=-(J T J) -1 J T r

[0296] Among them, δθ is the parameter update, J is the Jacobian matrix, and r is the residual vector.

[0297] 2. Perform correction iterations:

[0298] In each iteration, the predicted current I is calculated based on the current parameter values predicted (V i ), update the objective function value F calib , update the parameters according to the optimization algorithm, and repeat the iteration until the objective function converges to the minimum value.

[0299] 4. Verification and output

[0300] The predicted current is recalculated using the corrected model parameters and compared with the actual measured data.

[0301] Output corrected parameters, including parameters of surface leakage current model and and the parameters of the piecewise linear fit and

[0302] The parameter correction mechanism module adjusts the model parameters by defining and optimizing the correction objective function to improve the separation accuracy of surface leakage current and valve current.

[0303] Specifically, the data correction and calculation module includes:

[0304] 1. Input Data

[0305] Based on the parameter optimization and separation modules, the following have been obtained:

[0306] Corrected surface leakage current model parameters and

[0307] Corrected piecewise linear fitting parameters and

[0308] Corrected objective function F calib The minimum value of

[0309] In addition, the optimized current data and voltage data need to be input for further calculation of the target current value (eg, 1 mA).

[0310] 2. Voltage correction algorithm

[0311] Calculate 1mA voltage (U1mA):

[0312] Based on the corrected parameters, calculate the voltage value that makes the current 1mA:

[0313]

[0314] Among them, I 1maA =1mA.

[0315] Correction voltage value:

[0316] If further correction is required, additional correction factors (such as ambient temperature, material aging, etc.) can be applied to adjust the calculated voltage. The correction formula is:

[0317]

[0318] Among them, K temp is the temperature correction coefficient and ΔT is the temperature deviation.

[0319] 3. Leakage current calculation

[0320] Calculate the actual leakage current at 0.75 times U1mA (I75%U1mA). Based on the corrected voltage value, calculate the actual leakage current at 0.75 times U1mA:

[0321]

[0322] Among them, I surface and I core They are the current values ​​calculated by the surface leakage current model and the valve current model respectively.

[0323] 2. Correct the leakage current value:

[0324] Additional correction factors can be applied to adjust the calculated leakage current value, for example to account for measurement errors, correcting the formula:

[0325]

[0326] Among them, k meas is the measurement correction coefficient, and ΔM is the measurement error deviation.

[0327] 4. Evaluation of calibration results

[0328] Compare historical data:

[0329] Compare the calculated 1mA voltage and actual leakage current at 0.75 times U1mA with historical or experimental data to evaluate its accuracy. Compare the formula:

[0330]

[0331] Among them, U 1mA,hist and I 75％1mA,hist is the corresponding value in the historical data.

[0332] The data correction and calculation module uses voltage correction algorithms and leakage current calculations to accurately obtain the actual electrical parameters of the arrester. Through precise correction and calculation methods, it ensures that the 1mA voltage and the actual leakage current at 0.75 times U1mA truly reflect the operating status of the arrester, thereby providing reliable data support for further status assessment and maintenance decisions.

[0333] Specifically, the status assessment and early warning module includes:

[0334] 1. Input Data

[0335] Corrected 1mA voltage (U 1mA ) and 0.75 times the actual leakage current (I 75％1mA ), historical data, including previous test results and warning information, and environmental impact data, such as temperature, humidity and other external conditions.

[0336] 2. State Assessment Model

[0337] Multi-parameter analysis:

[0338] Combining the corrected electrical parameters with environmental factors, a multi-parameter analysis model is used to evaluate the current state of the arrester. The evaluation model:

[0339] Health Index (HI) = w 1 ×f(U 1mA )+w 2 ×g(I 75％1mA )+w3 ×h(environmental factors)

[0340] Among them, w 1 ,w 2 ,w 3 is the weight coefficient, f(U 1mA ),g(I 75％1mA ) are the impact functions of voltage and leakage current on health status, and h (environmental factors).

[0341] 2. Analysis of aging degree and moisture condition:

[0342] Based on the health index and the corrected parameters, the aging degree and moisture condition of the arrester are further analyzed. The analysis formula is:

[0343]

[0344] Among them, α, β, γ, δ are empirical coefficients.

[0345] 3. Threshold warning system

[0346] 1. Set the warning threshold:

[0347] Different warning thresholds are set according to the design standards and historical data of the lightning arrester.

[0348] According to the health index and analysis results, different levels of early warning information are generated to remind maintenance personnel to take corresponding measures.

[0349] 4. Historical data comparison

[0350] Data matching and comparison:

[0351] Match and compare current assessment results with historical data to identify degradation trends. Trend formula:

[0352]

[0353] in, is the current data, It is historical data.

[0354] Analyze whether the degradation trend is accelerating and combine it with the health index to predict the future state of the arrester.

[0355] The status assessment and early warning module achieves accurate status assessment and early warning of the arrester through the calculation and trend analysis of the health index. Combined with multi-parameter analysis and historical data comparison, it can detect potential failure risks in advance and help maintenance personnel perform effective preventive maintenance.

[0356] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of protection of the claims of the present invention.

Claims

1. 10kV lightning arrester intelligent online detector, characterized by: include: Data acquisition module, used to obtain DC volt-ampere characteristic curve of the arrester under different test conditions; Data preprocessing module, used to eliminate the linear or nonlinear trend of the volt-ampere characteristic curve and exclude abnormal values; The segmented linear fitting module is used to segment the volt-ampere characteristic curve and perform linear fitting on the data of each segment to obtain the linear parameters of each segment; Parameter optimization and separation module, used to optimize the piecewise linear fitting results and separate the surface leakage current and valve current The data correction and calculation module performs correction calculation based on the optimized parameters to obtain key electrical parameters; The status assessment and early warning module assesses the operating status of the arrester based on the corrected electrical parameters and provides early warning information.

2. The 10kV arrester intelligent online detector according to claim 1 is characterized in that: The data preprocessing module comprises: Apply a low-pass filter to remove high-frequency noise in the volt-ampere characteristic curve and set the filter cutoff frequency f c , and use the filter design formula to calculate the filter parameters: Where H(f) is the frequency response of the filter, f is the signal frequency, and f c is the cutoff frequency, n is the order of the filter; Use polynomial fitting to remove linear or nonlinear trends in the volt-ampere characteristic curve and make the data smoother: V fit (x)=a2x 2 +a1x+a0 Among them, V fit (x) is the trend line after fitting, a2, a1, and a0 are fitting coefficients; The method based on standard deviation is used to detect outliers. The mean μ and standard deviation σ of the data are calculated, and the threshold of outliers is defined as μ±kσ, where k is the threshold coefficient; Output volt-ampere characteristic curve data (V i ,I i ).

3. The 10kV arrester intelligent online detector according to claim 2 is characterized in that: The piecewise linear fitting module includes a piecewise algorithm module: The sliding window method is used to calculate the volt-ampere characteristic curve data (V i ,I i ), set a window size w, and calculate the slope s in each window: where s j is the slope in the jth window, I j+w and I j are the current value in the window, V j+w and V j They are voltage values; Calculate the change in the slope of the continuous window, determine the change point, and set a change threshold Δs: Δs j =|s j -s j-1 | If Δs j If it exceeds the set threshold Δs, the point is considered to be a change point; Mark the detected slope change point as the starting point or ending point of the segment; If the data change between two change points is less than a certain threshold, these points are merged into one segment and the merging threshold ε is set; Merge judgment = Δs j <ε If the conditions are met, these change points are merged into one segment; According to the detected change points and merging rules, the starting point and ending point of each segment are output; The information of the starting and ending points of each segment, including the voltage range of each segment [V start , V end ] and current range [I start , I end ].

4. The 10kV arrester intelligent online detector according to claim 3 is characterized in that: The piecewise linear fitting module includes a linear fitting module: According to the output of the segmentation algorithm module, the data of each segment (V i ,I i ) extracted; each segmented data includes [V start , V end ] to [I start , I end ] voltage and current values; Perform linear regression on the data of each segment to obtain the linear relationship: I=aV+b Where a is the slope and b is the intercept; The least squares method is used to solve the parameters of the linear model. For each segmented data (V i ,I i ), construct the loss function: Where n is the number of data points in the segment, and a and b are fitting parameters; By solving the least squares equation, we can get the optimal a and b: Output the fitted linear parameters a and b for each segment, as well as the fitted straight line equation I=aV+b; Output Format: I fit (V)=a j V+b j Among them, a j and b j are the slope and intercept of the jth segment.

5. The 10kV arrester intelligent online detector according to claim 4 is characterized in that: The piecewise linear fitting module includes a fitting quality assessment module: Calculate the residual of each data point, for each segmented data point (V i ,I i ), calculate the actual current value I i With the fitting current value I fit The residual between (V): Residual i =I i -(a j V i +b j ) Among them, I i is the actual current value, (a j V i +b j is the fitted current value calculated by the fitting model; Calculate the sum of squared residuals for all data points within each segment to measure the overall error of the fit: Among them, n j is the number of data points in the jth segment; The root mean square error is calculated from the residual sum of squares as an indicator of the quality of the fit: Where RMSE j is the root mean square error of the jth segment; The accuracy of the fitting is determined based on the calculated root mean square error value; The RMSE value of each segment is output and compared with the preset threshold to determine whether the quality of the fit meets the requirements.

6. The 10kV arrester intelligent online detector according to claim 5 is characterized in that: The parameter optimization and separation module includes a nonlinear optimization module: Construct an optimization objective function, which is used to measure the quality of the fitting results and is expressed as the sum of the residual squares of all segments: Where m is the number of segments, n j is the number of data points in the jth segment; Defining Constraints include: a j,min ≤a j ≤a j,max b j,min ≤b j ≤b j,max Among them, a j,min and a j,max are the minimum and maximum values ​​of the slope of the jth segment, b j,min and b j,max are the minimum and maximum values ​​of the intercept; The fitting parameter a j and b j Encoded as chromosomes; The objective function F is used as the fitness function; Genetic operations, including selection, crossover, and mutation, generate new candidate solutions and update the population; Fitness function calculation: Run the selected optimization algorithm to iteratively optimize the fitting parameters a according to the optimization objective function and constraints j and b j ; Record the objective function value of each iteration and update the best solution; According to the results of the optimization algorithm, the optimal linear fitting parameters for each segment are obtained. and Output:

7. The 10kV arrester intelligent online detector according to claim 6 is characterized in that: The parameter optimization and separation module includes a surface leakage current separation model: Define the surface leakage current model, assuming that the relationship between the surface leakage current and voltage is nonlinear, and use a predefined nonlinear function to represent it: Where c1, c2, and c3 are model parameters that represent the characteristics of surface leakage current; The valve current model is defined as follows: I core (V)=I opt (V)-I surface (V) Among them, I core (V) is the calculated valve current; The parameters c1, c2, and c3 of the surface leakage current model are fitted by an optimization algorithm to minimize the error of the surface leakage current model. The objective function is: Among them, I surface,i is the current predicted by the surface leakage current model, I measured,i is the actual measured surface leakage current.

8. The 10kV arrester intelligent online detector according to claim 7 is characterized in that: The parameter optimization and separation module includes a parameter correction module: Enter actual test data for calibration: The measured voltage V i , the measured current I measured (V i ); Establish a correction objective function, which is used to measure the difference between the actual detection data and the model prediction data: Among them, I predicted (V i ) is the current predicted by the separation model: I predicted (V i )=I surface (V i )+I core (V i ) The goal is to adjust the model parameters so that F calib minimize; The parameters optimized during the calibration process include the parameters c1, c2, and c3 of the surface leakage current model and the parameters of the piecewise linear fit. and By iteratively updating the model parameters, the objective function F is minimized. calib , the update rule is: δθ=-(J T J) -1 J T r Among them, δθ is the parameter update amount, J is the Jacobian matrix, and r is the residual vector; In each iteration, the predicted current I is calculated based on the current parameter values predicted (V i ), update the objective function value F calib ; Update the parameters according to the optimization algorithm and repeat the iteration until the objective function converges to the minimum value; The predicted current is recalculated using the corrected model parameters and compared with the actual measured data; Output corrected parameters, including parameters of surface leakage current model and and the parameters of the piecewise linear fit and 9. The 10kV arrester intelligent online detector according to claim 8, characterized in that: The data correction and calculation module includes: Based on the corrected parameters, calculate the voltage value that makes the current 1mA: Among them, I 1maA =1mA; Applying an additional correction factor to adjust the calculated voltage, the corrected formula is: Among them, K temp is the temperature correction coefficient, ΔT is the temperature deviation; Based on the corrected voltage value, calculate the actual leakage current when the voltage is 0.75 times U1mA: Among them, I surface and I core are the current values ​​calculated by the surface leakage current model and the valve current model respectively; Apply an additional correction factor to adjust the calculated leakage current value: Among them, k meas is the measurement correction factor, ΔM is the measurement error deviation; Compare the calculated 1mA voltage and actual leakage current at 0.75 times U1mA with historical or experimental data to evaluate its accuracy. Compare the formula: Among them, U 1mA,hist and I 75％1mA,hist is the corresponding value in the historical data; Outputs corrected voltage and leakage current data.

Images