Lightning arrester state identification method and system based on current and temperature signal combined diagnosis, terminal and medium

Abstract

The invention discloses a lightning arrester state identification method and system based on current and temperature signal combined diagnosis, a terminal and a medium. The method comprises the following steps: acquiring basic data of a lightning arrester based on an operation voltage effect; extracting a resistive current signal based on the total current signal, and decomposing the resistive current signal by adopting a fast Fourier algorithm to obtain each harmonic peak value of the resistive current; based on the temperature data of different position points on the surface of the lightning arrester, performing inversion speculation to obtain internal temperature data of the lightning arrester, and calculating to obtain a temperature rise value of a valve plate of the lightning arrester; and based on the peak value of the total current, the peak value of each harmonic of the resistive current and the temperature rise value of the lightning arrester valve plate, key characteristic quantities and auxiliary characteristic quantities representing the health state of the lightning arrester are extracted and serve as input of a lightning arrester state recognition model for lightning arrester state prediction. The method can accurately, comprehensively and online evaluate and identify the state of the lightning arrester in real time.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of lightning arrester defect diagnosis, and particularly relates to a lightning arrester state recognition method and system based on joint diagnosis of current and temperature signals, a terminal and a medium. BACKGROUND

[0002] A lightning arrester is an important power equipment for limiting overvoltage in a power grid, which can protect the safe operation of a power system and equipment. The nonlinear volt-ampere characteristic of the lightning arrester valve body makes it flow through a very small current when working at normal voltage; but under the action of lightning or operating overvoltage, the resistance of the lightning arrester valve body drops sharply, which can quickly release the lightning or operating overvoltage energy, thereby playing an important role in reducing the damage of overvoltage to power grid equipment. In the running process, the lightning arrester is inevitably affected by many comprehensive factors such as changes in environmental temperature and humidity, lightning and operating overvoltage impact, etc. The internal valve piece of the lightning arrester will be damp, aged and deteriorated, and the outside will be polluted, thereby losing the protection of power equipment. Therefore, accurate diagnosis of the health status of the lightning arrester and identification of the defect type can help to discover lightning arrester abnormalities in time and realize early warning of lightning arrester failure.

[0003] At present, the diagnosis of the state of the operating lightning arrester is mainly realized by collecting lightning arrester resistive current and surface temperature at a fixed period and analyzing the change trend according to DL / T 393-2021 “Power Transmission and Transformation Equipment State Maintenance Test Regulations”, but a large number of practices have found that this method also has some problems in application: first, it is difficult to realize online diagnosis. The lightning arrester is frequently impacted during the lightning season and when the circuit breaker is switched, and the offline live collection data cannot track the operating state of the lightning arrester in real time. Second, no specific threshold value is proposed for the safety limit of the lightning arrester resistive current and surface temperature, which cannot reflect the gradual deterioration process and safety margin of the lightning arrester, and cannot further distinguish the type or degree of abnormal state of the lightning arrester. In recent years, the diagnosis and analysis of lightning arrester failure has gradually shifted from offline data collection to intelligent online diagnosis, and thus the diagnosis method based on model training has been widely applied. In the prior art, some models only judge the working state of the lightning arrester according to the collected leakage current data, the training data type is single, the model precision is low, the algorithm of some models is easily affected by data interference, which leads to limited prediction ability, and some models have the problem of data not easy to fit due to complex training.

[0004] In summary, there is an urgent need for an online evaluation and recognition method that can accurately, timely and comprehensively evaluate the state of the lightning arrester. SUMMARY

[0005] To overcome the shortcomings of existing technologies, this invention provides a surge arrester status identification method, system, terminal, and medium based on joint diagnosis of current and temperature signals, in order to solve the technical problems of unclear criteria, easy misjudgment, and inability to accurately identify the abnormal status category of surge arresters in existing surge arrester status diagnosis methods.

[0006] In a first aspect, the present invention provides a surge arrester condition identification method based on joint diagnosis of current and temperature signals, comprising:

[0007] Acquire the full current signal of the surge arrester under operating voltage and the temperature data at different locations on the surface of the surge arrester;

[0008] The resistive current signal is extracted based on the full current signal, and the resistive current signal is decomposed using the fast Fourier algorithm to obtain the peak values ​​of each harmonic of the resistive current.

[0009] Based on the temperature data at different locations on the surface of the surge arrester, the internal temperature data of the surge arrester is obtained by inversion and estimation to obtain the maximum temperature value and normal temperature value of the surge arrester valve. The maximum temperature rise value and normal temperature rise value of the surge arrester valve are then calculated.

[0010] Based on the peak value of the total current and the peak values ​​of each harmonic of the resistive current, the maximum temperature rise of the surge arrester valve, and the normal temperature rise of the surge arrester valve, key and auxiliary characteristic quantities that characterize the health status of the surge arrester are extracted.

[0011] The key features and auxiliary features are used as inputs to the constructed and trained arrester state recognition model for prediction, and the arrester state type and / or state degree are output.

[0012] Furthermore, the peak values ​​of each harmonic of the resistive current include the peak value of the fundamental resistive current, the peak value of the third harmonic resistive current, the peak value of the fifth harmonic resistive current, and the peak value of the seventh harmonic resistive current; the temperature data at different locations on the surface of the surge arrester include the highest surface temperature and the normal surface temperature of the surge arrester.

[0013] Furthermore, based on the extraction of the resistive current signal from the full current signal, the resistive current signal is decomposed using the Fast Fourier Transform algorithm to obtain the peak values ​​of each harmonic of the resistive current. The specific process is as follows:

[0014] Extracting resistive current from a full-current signal;

[0015] Perform a Fourier operation on the resistive current signal of at least two cycles every N sampling points to obtain the peak value of the fundamental resistive current, the peak value of the third harmonic resistive current, the peak value of the fifth harmonic resistive current, and the peak value of the seventh harmonic resistive current.

[0016] Furthermore, based on the extraction of the resistive current signal from the full current signal, the resistive current signal is decomposed using the Fast Fourier Transform algorithm. After obtaining the peak values ​​of each harmonic of the resistive current, the following is also included:

[0017] Based on the surface insulation contamination of surge arresters, the peak value of the total current Peak value of fundamental resistive current Third harmonic resistive current peak value Fifth harmonic resistive current peak value Peak value of the seventh harmonic resistive current The data has been corrected, and the specific corrected data is as follows:

[0018]

[0019]

[0020]

[0021]

[0022]

[0023] In the formula, , , , , These are the peak values ​​of the total current, the fundamental resistive current, the third harmonic resistive current, the fifth harmonic resistive current, and the seventh harmonic resistive current, respectively, after correction based on the surface insulation contamination of the surge arrester. , , , , These are the peak values ​​of the total current, fundamental resistive current, third harmonic resistive current, fifth harmonic resistive current, and seventh harmonic resistive current before correction based on the surface insulation contamination of the surge arrester. , , , , This is the soiling correction factor; Ambient air humidity; The equivalent salt density is applied to the outer insulation surface of the surge arrester.

[0024] Furthermore, the formula for calculating the maximum temperature rise of the surge arrester varistor is as follows:

[0025]

[0026] The formula for calculating the normal temperature rise of surge arrester varistors is:

[0027]

[0028] in, This refers to the highest temperature value of the surge arrester valve plate; This refers to the normal temperature value of the surge arrester valve plate. Surge arrester operating ambient temperature; The maximum temperature rise of the surge arrester varistor; This represents the normal temperature rise value of the surge arrester valve plate.

[0029] Furthermore, key characteristic quantities include peak total current. Peak value of fundamental resistive current Valve plate maximum temperature rise Auxiliary characteristic quantities include the rate of change of total current. Rate of change of resistive current Third harmonic characteristic coefficient RC characteristic coefficient Valve plate temperature rise rate Valve plate temperature rise fluctuation rate The specific calculation formula is as follows:

[0030]

[0031]

[0032]

[0033] in, , , These are the peak value of the total current, the peak value of the fundamental resistive current, and the average temperature rise of the valence plate, obtained when the operating voltage is applied to the surge arrester sample under normal conditions. This represents the peak value of the total current of the surge arrester. This represents the peak value of the fundamental resistive current. The peak value of the third harmonic resistive current, The peak value of the fifth harmonic resistive current , The peak value of the seventh harmonic resistive current; The maximum temperature rise of the surge arrester varistor; This represents the normal temperature rise value of the surge arrester valve plate.

[0034] Furthermore, the surge arrester condition identification model is obtained by applying the rated operating voltage at power frequency to surge arrester samples in different health states through experimental methods, acquiring the current and temperature electrothermal response time series data of the surge arrester under the action of the operating voltage, extracting key feature quantities and auxiliary feature quantities characterizing the health state of the surge arrester, preprocessing them to construct training set data for various health states, and using the actual state labels of the surge arrester and the training set data to train and optimize the parameters of the CatBoost model.

[0035] Furthermore, the training process of the surge arrester state recognition model is as follows:

[0036] The feature values ​​of the preprocessed training set data are randomly sorted in multiple groups, and the model is initialized; the preprocessing includes outlier replacement and normalization.

[0037] During the Boosting main loop phase, for each iteration, the residual of the current sample is calculated, and a symmetric binary tree model with a preset depth is trained based on the residual. At each leaf node, a splitting scheme that minimizes the loss function is automatically selected, and each leaf node corresponds to a subset of samples and a predicted value. At the same time, the prediction function of this round is weighted according to the learning rate and accumulated into the previous round model to form a new prediction function. The residual is the negative gradient of the loss function in the output direction of the current model.

[0038] The Softmax multi-class cross-entropy loss function value is calculated as the criterion for convergence. Early stopping is triggered when the loss function does not decrease significantly in a preset number of iterations, or training is terminated when the number of iterations reaches a preset upper limit, thus completing the CatBoost model training.

[0039] Secondly, the present invention provides a surge arrester condition identification system based on joint diagnosis of current and temperature signals, comprising:

[0040] Data acquisition module: used to acquire the full current signal of the surge arrester under the action of the operating voltage and the temperature data at different locations on the surface of the surge arrester;

[0041] Data processing module: used to extract resistive current signal based on full current signal, decompose resistive current signal using fast Fourier algorithm to obtain peak values ​​of each harmonic of resistive current; used to inversely infer internal temperature data of arrester based on temperature data at different locations on the arrester surface to obtain the maximum temperature value and normal temperature value of arrester valence, and then calculate and obtain the maximum temperature rise value and normal temperature rise value of arrester valence;

[0042] Model prediction module: Based on the peak value of the total current and the peak values ​​of each harmonic of the resistive current, the maximum temperature rise of the arrester valve, and the normal temperature rise of the arrester valve, it extracts key and auxiliary features that characterize the health status of the arrester and inputs them into the constructed and trained arrester status recognition model for prediction, and outputs the arrester's status type and / or status degree.

[0043] Furthermore, the peak values ​​of each harmonic of the resistive current include the peak value of the fundamental resistive current, the peak value of the third harmonic resistive current, the peak value of the fifth harmonic resistive current, and the peak value of the seventh harmonic resistive current; the surface temperature data of the surge arrester includes the highest surface temperature and the normal surface temperature of the surge arrester.

[0044] Furthermore, based on the extraction of the resistive current signal from the full current signal, the resistive current signal is decomposed using the Fast Fourier Transform algorithm to obtain the peak values ​​of each harmonic of the resistive current. The specific process is as follows:

[0045] Extracting resistive current from a full-current signal;

[0046] Perform a Fourier operation on the resistive current signal of at least two cycles every N sampling points to obtain the peak value of the fundamental resistive current, the peak value of the third harmonic resistive current, the peak value of the fifth harmonic resistive current, and the peak value of the seventh harmonic resistive current.

[0047] Furthermore, based on the extraction of the resistive current signal from the full current signal, the resistive current signal is decomposed using the Fast Fourier Transform algorithm. After obtaining the peak values ​​of each harmonic of the resistive current, the following is also included:

[0048] Based on the surface insulation contamination of surge arresters, the peak value of the total current Peak value of fundamental resistive current Third harmonic resistive current peak value Fifth harmonic resistive current peak value Peak value of the seventh harmonic resistive current The data has been corrected, and the specific corrected data is as follows:

[0049]

[0050]

[0051]

[0052]

[0053]

[0054] In the formula, , , , , These are the peak values ​​of the total current, the fundamental resistive current, the third harmonic resistive current, the fifth harmonic resistive current, and the seventh harmonic resistive current, respectively, after correction based on the surface insulation contamination of the surge arrester. , , , , These are the peak values ​​of the total current, fundamental resistive current, third harmonic resistive current, fifth harmonic resistive current, and seventh harmonic resistive current before correction based on the surface insulation contamination of the surge arrester. , , , , This is the soiling correction factor; Ambient air humidity; The equivalent salt density is applied to the outer insulation surface of the surge arrester.

[0055] Furthermore, the formula for calculating the maximum temperature rise of the surge arrester varistor is as follows:

[0056]

[0057] The formula for calculating the normal temperature rise of surge arrester varistors is:

[0058]

[0059] in, This refers to the highest temperature value of the surge arrester valve plate; This refers to the normal temperature value of the surge arrester valve plate. Surge arrester operating ambient temperature; The maximum temperature rise of the surge arrester varistor; This represents the normal temperature rise value of the surge arrester valve plate.

[0060] Furthermore, the key characteristic quantity includes the peak value of the total current. Peak value of fundamental resistive current Valve plate maximum temperature rise Auxiliary characteristic quantities include the rate of change of total current. Rate of change of resistive current Third harmonic characteristic coefficient RC characteristic coefficient Valve plate temperature rise rate Valve plate temperature rise fluctuation rate The specific calculation formula is as follows:

[0061]

[0062]

[0063]

[0064] in, , , The peak value of the total current, the peak value of the fundamental resistive current, and the average temperature rise of the valence plate were obtained when the operating voltage was applied to the surge arrester sample under normal conditions. This represents the peak value of the total current of the surge arrester. This represents the peak value of the fundamental resistive current. The peak value of the third harmonic resistive current, The peak value of the fifth harmonic resistive current , The peak value of the seventh harmonic resistive current; The maximum temperature rise of the surge arrester varistor; This represents the normal temperature rise value of the surge arrester valve plate.

[0065] Furthermore, the surge arrester condition identification model is obtained by applying the rated operating voltage at power frequency to surge arrester samples in different health states through experimental methods, acquiring the current and temperature electrothermal response time series data of the surge arrester under the action of the operating voltage, extracting key feature quantities and auxiliary feature quantities characterizing the health state of the surge arrester from the data, preprocessing them to construct training set data for various health states, and using the actual state labels of the surge arrester and the training set data to train and optimize the parameters of the CatBoost model.

[0066] Furthermore, the training process of the surge arrester state recognition model is as follows:

[0067] The feature values ​​of the preprocessed training set data are randomly sorted in multiple groups, and the model is initialized; the preprocessing includes outlier replacement and normalization.

[0068] During the Boosting main loop phase, for each iteration, the residual of the current sample is calculated, and a symmetric binary tree model with a preset depth is trained based on the residual. At each leaf node, a splitting scheme that minimizes the loss function is automatically selected, and each leaf node corresponds to a subset of samples and a predicted value. At the same time, the prediction function of this round is weighted according to the learning rate and accumulated into the previous round model to form a new prediction function. The residual is the negative gradient of the loss function in the output direction of the current model.

[0069] The Softmax multi-class cross-entropy loss function value is calculated as the criterion for convergence. Early stopping is triggered when the loss function does not decrease significantly in a preset number of iterations, or training is terminated when the number of iterations reaches a preset upper limit, thus completing the CatBoost model training.

[0070] Thirdly, the present invention provides an electronic terminal comprising a processor and a memory, the memory storing a computer program, the processor invoking the computer program to execute the steps of the method described above.

[0071] Fourthly, the present invention provides a readable storage medium storing a computer program that, when invoked by a processor, performs the steps of the method described above.

[0072] This invention provides a surge arrester status identification method, system, terminal, and medium based on joint diagnosis of current and temperature signals. Compared with the prior art, the method has the following advantages:

[0073] By applying external pressure to surge arresters in different health states and conducting thermoelectric time-series data tests, the mapping relationship between the actual operating data and health state of the surge arresters was determined, ensuring the validity of the model training sample data. Based on the thermoelectric time-series data, key and auxiliary features characterizing surge arrester fault information were identified as training set data. The CatBoost model for surge arrester condition diagnosis and analysis was iteratively trained, and the model showed better robustness and faster training and prediction speeds, improving the efficiency and accuracy of surge arrester condition evaluation. Attached Figure Description

[0074] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0075] Figure 1 This is a flowchart illustrating the surge arrester status identification method based on joint diagnosis of current and temperature signals provided in an embodiment of the present invention.

[0076] Figure 2 This is a schematic diagram of the total current amplitude and phase angle distribution of the surge arrester under different health status types provided in the embodiments of the present invention;

[0077] Figure 3 This is a schematic diagram of the internal temperature distribution of a surge arrester under different health conditions provided in the embodiments of the present invention, wherein (a) is the normal state, (b) is internally damp, (c) the insulating sleeve is aged, and (d) is wet and dirty.

[0078] Figure 4 This is a flowchart of the CatBoost classification algorithm training process provided in an embodiment of the present invention;

[0079] Figure 5This is a schematic diagram comparing the confusion matrices of different classification models provided in this embodiment of the invention;

[0080] Figure 6 This is a schematic diagram illustrating the verification of the CatBoost test set confusion matrix results provided in an embodiment of the present invention. Detailed Implementation

[0081] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0082] Example 1

[0083] like Figure 1 As shown, this embodiment provides a surge arrester status identification method based on joint diagnosis of current and temperature signals, wherein the order of the steps is not limited, including:

[0084] S1: Acquire the full current signal and surface temperature data of the surge arrester under the operating voltage.

[0085] In practice, the total current signal of the surge arrester leakage can be directly measured from the grounding lead of the surge arrester using a current monitoring sensor (such as a Hall coil); the temperature data at different locations on the surface of the surge arrester can be directly measured using an infrared precision thermometer.

[0086] S2: Extract the resistive current signal based on the full current signal, and use the fast Fourier algorithm to decompose the resistive current signal to obtain the peak values ​​of each harmonic of the resistive current; among which, the peak values ​​of each harmonic of the resistive current include the peak value of the fundamental resistive current, the peak value of the third harmonic resistive current, the peak value of the fifth harmonic resistive current, and the peak value of the seventh harmonic resistive current.

[0087] S21: The specific steps for extracting resistive current from a full-current signal are as follows:

[0088] The first peak value in the half-wave of the full-current signal is detected and set as the peak value of the capacitive current. Then, the resistive current is calculated using the formula:

[0089]

[0090] in, It is a full current signal; It is a resistive current signal; This represents the peak value of the capacitive current. For any time; For the corresponding The moment; This is the full current cycle.

[0091] S22: For resistive current signals of at least two cycles, every... Perform a Fourier operation on each sampling point to obtain the peak value of the fundamental resistive current. Third harmonic resistive current peak value Fifth harmonic resistive current peak value Peak value of the seventh harmonic resistive current The specific method is as follows:

[0092] Discrete values ​​of the resistive current signal are acquired within one cycle, denoted as... , It is a discrete value sequence. , on data The Fourier transform operation is performed using the following formula:

[0093]

[0094] In the formula, Represents a complex unit; This represents the number of sampled data points. Different values ​​correspond to different orders of harmonic components, such as Let 1 represent the fundamental frequency. Take 3 to represent the third harmonic; For discrete frequencies, The Fourier coefficients are in complex form and contain... The peak value and phase information of the subharmonic, the peak value is .

[0095] The total current and resistive current of a surge arrester increase linearly with the increase of the degree of pollution on its outer surface. The degree of pollution and the equivalent salt density of the external insulation are directly related to the ambient humidity. In order to reduce the interference of the pollution current of the external insulation of the surge arrester on the test results of the total current and resistive current, S23 is also included.

[0096] S23: Impact of total current peak value on surface insulation contamination of surge arrester Peak value of fundamental resistive current Third harmonic resistive current peak value Fifth harmonic resistive current peak value Peak value of the seventh harmonic resistive current The data has been corrected, and the specific corrected data is as follows:

[0097]

[0098]

[0099]

[0100]

[0101]

[0102] In the formula, , , , , These are the peak values ​​of the total current, the fundamental resistive current, the third harmonic resistive current, the fifth harmonic resistive current, and the seventh harmonic resistive current, respectively, after correction based on the surface insulation contamination of the surge arrester. , , , , This is the soiling correction factor; Ambient air humidity; This is the equivalent salt density of the external insulation surface of the surge arrester. For surface-clean surge arresters, The value is 0.

[0103] S3: The infrared thermometry method is used to obtain the temperature data at different locations on the surface of the surge arrester. Based on the difference between the internal and external temperatures of the surge arrester, the inversion method is used to infer the temperature data of the internal valve plate of the surge arrester, and the maximum temperature value and normal temperature value of the surge arrester valve plate are obtained. Combined with the operating environment temperature value of the surge arrester, the maximum temperature rise value and normal temperature rise value of the surge arrester valve plate are calculated.

[0104] Specifically, using infrared thermometry, N temperature points are collected sequentially and uniformly on the surface of the surge arrester from top to bottom along its height direction, and these points are denoted as follows: , ... Based on the temperature difference between the inside and outside of the surge arrester, the temperature data of the valve plate at the corresponding height position inside the surge arrester is obtained using an inversion method. , ... ,in , This is the temperature point sequence number. , This is a correction factor for the temperature difference between the inside and outside of the surge arrester. The maximum temperature value of the valve plate inside the surge arrester is calculated as follows: The normal temperature value of the internal valve plate of the surge arrester is calculated as follows: .

[0105] Furthermore, the formula for calculating the maximum temperature rise of the surge arrester varistor is as follows:

[0106]

[0107] The formula for calculating the normal temperature rise of surge arrester varistors is:

[0108]

[0109] in, Surge arrester operating ambient temperature; The maximum temperature rise of the surge arrester varistor; This represents the normal temperature rise value of the surge arrester valve plate.

[0110] The health status of surge arresters is assessed and analyzed using the total current and external insulation surface temperature. Figure 2 The distribution of the total current amplitude and phase angle was measured after applying operating voltage to two surge arrester samples in different health states. It can be seen from the distribution diagram that when the surge arrester is in different health states or even when different parts of the same surge arrester are in different states, the total current and phase angle test values ​​have obvious overlap. Therefore, it is difficult to further distinguish the fault type and fault degree of the surge arrester using the total current and phase angle. Figure 3 To illustrate the internal temperature distribution of two surge arrester samples under rated voltage during long-term operation in different health states, taking localized moisture damage to the valve plates or aging of the outer insulation of the lower arrester as an example, compared to normal operation, the abnormal parts of the arrester will show a significant temperature rise, or a significant temperature difference between the upper and lower arrester samples. Therefore, unlike commonly used state quantities such as the total current amplitude and surface hot spot temperature of surge arresters, this document introduces key characteristic quantities and auxiliary characteristic quantities to determine the state type and degree of the surge arrester.

[0111] S4: Based on the peak value of the total current, the peak values ​​of each harmonic of the resistive current, the maximum temperature rise of the surge arrester valve, and the normal temperature rise of the surge arrester valve, extract key and auxiliary characteristic quantities characterizing the health status of the surge arrester. The key characteristic quantity includes the peak value of the total current. Peak value of fundamental resistive current Valve plate maximum temperature rise The key characteristic quantities include current time-series data and temperature time-series data; the auxiliary characteristic quantities include the total current change rate. Rate of change of resistive current Third harmonic characteristic coefficient RC characteristic coefficient Valve plate temperature rise rate Valve plate temperature rise fluctuation rate The specific calculation formula is as follows:

[0112]

[0113]

[0114]

[0115] in, , , These are the peak value of the total current of the surge arrester, the peak value of the fundamental resistive current, and the average temperature rise of the valve plate, obtained when the operating voltage is applied to the surge arrester sample under normal conditions. They also serve as the initial values ​​of the state quantities described under different state degradation labels.

[0116] S5: Input the key features and auxiliary features into the constructed and trained arrester state recognition model for prediction, and output the arrester state type and / or state degree.

[0117] Specifically, the surge arrester condition identification model is developed by applying the rated operating voltage at power frequency to surge arrester samples in different health states experimentally. This yields time-series data on the current and temperature electrothermal responses of the surge arresters under the operating voltage. Key and auxiliary features characterizing the health state of the surge arresters are extracted from this data. After preprocessing, training sets for various health states are constructed. The CatBoost model is then trained and its parameters optimized using the actual state labels of the surge arresters and the training set data. In practice, the type of model is not limited; in addition to the CatBoost model, neural network models such as linear neural networks (e.g., FitcNet) can also be used.

[0118] More specifically, the health status of surge arrester samples includes, but is not limited to, surge arrester dampness, surge arrester aging, and surge arrester contamination. Specifically, surge arrester dampness can be addressed by performing damp heat treatment on the valve plates of each surge arrester sample; surge arrester aging can be addressed by performing impulse current treatment on the valve plate outer insulation of each surge arrester sample; and surge arrester contamination can be addressed by performing external insulation contamination treatment on each surge arrester sample. In this embodiment, taking surge arrester dampness, surge arrester aging, and surge arrester contamination as examples, a degradation label is assigned to each type of health status. For example, 1 represents surge arrester dampness, 2 represents surge arrester aging, and 3 represents surge arrester contamination.

[0119] By adjusting the moisture absorption time during the wet heat treatment of the surge arrester varistor, the moisture state of the surge arrester can be further classified; by adjusting the number of current surges during the impulse current treatment of the surge arrester varistor's outer insulation, the aging state of the surge arrester can be further classified; by adjusting the salt density level during the pollution treatment of the surge arrester's outer insulation, the pollution accumulation state of the surge arrester can be further classified. The judgment of the surge arrester's health status uses one or more criteria, without limitation, and can be adjusted according to the actual situation. In this specific implementation, the criterion is the reference voltage of the surge arrester at 1mA DC, or... The DC leakage current is measured. The classification of various health states is achieved by setting multiple health thresholds. The number of health thresholds is not limited and can be selected according to the actual scenario requirements. In this embodiment, taking the surge arrester's moisture absorption as an example, the moisture absorption time during the valve plate's damp heat treatment process can be adjusted to ensure that the reference voltage of the surge arrester under 1mA DC is within acceptable limits. The voltage health thresholds are set as follows: less than or equal to the first voltage health threshold, greater than the first voltage health threshold but less than or equal to the second voltage health threshold, greater than the second voltage health threshold but less than or equal to the third voltage health threshold, and so on, until the upper limit of the voltage health threshold is reached by the surge arrester's nominal voltage health threshold. 40% of the value; or The DC leakage current is greater than the first current health threshold and less than or equal to the second current health threshold, greater than the second current health threshold and less than or equal to the third current health threshold, greater than the third current health threshold and less than or equal to the fourth current health threshold, and so on, until the upper limit of the current health threshold is reached by the surge arrester's nominal value. Five times the value. The values ​​of each voltage health threshold and each current health threshold are not limited and can be adjusted according to the actual needs of the scenario.

[0120] After preprocessing the key and auxiliary features of surge arresters under different health conditions, a training set is constructed. Among them, key characteristic quantities include the peak total current. Peak value of fundamental resistive current Valve plate temperature rise Auxiliary characteristic quantities include the rate of change of total current. Rate of change of resistive current Third harmonic characteristic coefficient RC characteristic coefficient Valve plate temperature rise rate Valve plate temperature rise fluctuation rate .

[0121] according to Number of samples, The eigenvalues ​​can be used to construct the feature matrix of the training set: .

[0122] Set the state prediction probability of the CatBoost model under different arrester state degradation labels. The output is:

[0123]

[0124] In the formula, , , , These correspond to the predicted probabilities under the following conditions: normal surge arrester, surge arrester damp, surge arrester aging, and surge arrester contaminated.

[0125] More specifically, preprocessing includes outlier substitution and normalization. Specifically:

[0126] First, outliers are identified using the standard deviation method. If... For the first of all sample data The mean of the eigenvalues, For the first The standard deviation of each eigenvalue For the first The first sample There are eigenvalues. When a certain eigenvalue satisfies... When it is identified as an outlier, it is adopted Replace the feature value.

[0127] Then, the entropy weighting method is used to objectively assign weights to each input feature value in the training set, increasing the contribution of the feature values ​​to the arrester fault classification results. The model's sensitivity is enhanced by calculating the proportional coefficients of each feature variable. The entropy weighting method is a typical objective weighting method in multi-index decision-making, capable of assessing the importance of features through differences in sample information distribution. (Information entropy) The smaller the value, the greater the variation in the distribution of the feature among samples, meaning the richer the information content. Among these, information entropy... The calculation is as follows:

[0128]

[0129] .

[0130] The degree of dispersion of each feature across all samples is statistically analyzed. The greater the difference, the higher the amount of information contained in the feature, and the better it can distinguish different fault states.

[0131] The difference is converted into feature importance scores, and the importance scores of all features are normalized to obtain objective weights, where each weight ranges from 0 to 1, and the sum of all weights is 1. The first... The objective weights of the eigenvalues ​​are denoted as . This can be expressed as a formula:

[0132]

[0133] in, This represents the total number of key and auxiliary feature values.

[0134] Specifically, the training and parameter optimization of the CatBoost model involves using the training set data as input and the predicted output probabilities as results. During training, the predicted output probabilities of the surge arrester state are compared with the actual state probabilities, and the network parameters of the CatBoost model are continuously adjusted based on the comparison results. In this embodiment, the parameters of the CatBoost model are set as follows during training:

[0135]

[0136] like Figure 4 As shown, the CatBoost model is trained using an ordered boosting mechanism. This step uses the preprocessed training set as input to iteratively train the classification model and determine the final parameters. The training process includes four stages: data sorting and preparation, residual calculation, fitted base learner, model update, and convergence determination.

[0137] 1) Data sorting and preparation

[0138] Generate ordered arrangement: Randomly arrange the training samples to ensure that the order of samples within the same group is fixed, which is used for sequential learning and verification;

[0139] Model initialization: Set the cumulative prediction function to the zero function, denoted as . Where 0 represents the 0th iteration and h is the number of arrester state degradation tags. After completing the above steps, proceed with sequential iteration.

[0140] 2) Residual calculation

[0141] In the At the start of the iteration round, the previous model applied to the 1st iteration. Sample data The internal score for surge arresters belonging to the h-class condition is recorded as follows: The scores for all categories are converted into predicted probabilities using the Softmax function, i.e., the sample data. The probability of a surge arrester belonging to the h-th type of deterioration state Represented as:

[0142]

[0143] in, This represents the total number of key and auxiliary feature values.

[0144] Further sample The true status label is If the sample The corresponding surge arrester's actual condition degradation label is ,but ,otherwise For example, when the sample For the operating data of the surge arrester valve plate when it is damp, then , , .

[0145] Therefore, the residual for this round is defined as:

[0146]

[0147] In the formula, For the first Samples in round iteration The state degradation label is The residual value under the health condition of the surge arrester represents the difference between the predicted value and the actual value.

[0148] 3) Base fit learner

[0149] The residual obtained in the previous step is used as the learning objective to train a symmetric binary tree of a preset depth (6 in this embodiment) as the basis learner for this round. During training:

[0150] 1) By traversing candidate features and thresholds, the splitting scheme that can minimize the current loss is automatically selected;

[0151] 2) After the tree grows, it forms 64 leaf nodes, each corresponding to a subset of samples and a leaf value;

[0152] 3) When any new sample is input, it is first assigned to a leaf node by the tree, and the output of the tree in this round is the leaf value of that leaf node;

[0153] 4) Update the cumulative model and determine convergence.

[0154] The output of the base learner in this round is used as the correction amount for this round. After being weighted by the learning rate, it is added to the cumulative model in the previous round to obtain a new cumulative prediction function:

[0155]

[0156] in, This is the output of the base learner in this round. For learning rate, This represents the cumulative output after the t-th iteration.

[0157] To avoid overfitting and improve training efficiency, the following evaluation and stopping rules are executed after each round of model updates:

[0158] 1) Model evaluation metrics. Calculate the "Softmax multi-class cross-entropy" on the reserved validation set as a unified evaluation metric to measure the current model's comprehensive ability to distinguish between different classes.

[0159] 2) Save the best record. If the validation metric in this round is better than the historical best value, mark the current model parameters as the "best checkpoint" (save the weights and key hyperparameters).

[0160] 3) Early Stop Setting. (No limitation is imposed; in this embodiment, the number of patience rounds is set to 50.) When the validation metric does not show effective improvement within a preset number of iterations (in this embodiment, the preset number of iterations is 50 rounds) (improvement less than the set threshold is considered no improvement), early stop is triggered, the training process terminates, and the model parameters corresponding to the optimal checkpoint are automatically rolled back.

[0161] 4) Upper limit termination condition. If early stopping is not triggered, but the cumulative iterations reach the preset upper limit of 500 rounds, training will be terminated directly, and the historical best checkpoint will be used as the final model.

[0162] Taking timely maintenance measures based on the early warning information output by the model can prevent surge arresters from malfunctioning or being damaged. Using the above method, the condition of surge arresters can be predicted based on online monitoring data, providing a reference for power plant operators. This invention is simple, highly operable, easy to implement in a computer program, and readily deployable in the field. It can effectively predict the development trend of surge arrester conditions and has significant value for promotion and utilization.

[0163] To further illustrate the method, a specific embodiment is provided below:

[0164] A 220kV zinc oxide surge arrester was designed to withstand conditions including moisture absorption, aging, and contamination. Data was collected and processed under operating voltage, resulting in 290 training datasets. After data processing, the datasets were input into a CatBoost model for training. The softmax function was used to convert the score for each health status category into a probability vector, yielding the final predicted label:

[0165]

[0166] Table 2 shows the results of state prediction for some surge arrester samples with different health conditions. It can be seen that the prediction results are in good agreement with the actual conditions.

[0167]

[0168] Additionally, a preset value for the first state can be set. Second state preset value ,in > ,like Then it is determined that the surge arrester is in a high-risk state; if If so, the surge arrester is judged to be in a low-risk state; if If the probability vector is positive, the surge arrester is determined to be in a normal state. The specific fault types output based on the calculated probability vector are shown in Table 3.

[0169]

[0170] Furthermore, the training set data was used to perform 5-fold cross-validation on both the CatBoost model and the baseline neural network (fitcnet) model. Figure 4 As can be seen, compared to the FitCNet model, CatBoost improves accuracy by 3.4 percentage points and reduces the error rate from 4.1% to 0.7%; the macro-average F1 score also improves by 3.3 percentage points, demonstrating better class balance and lower bias. Figure 4 A comparison of the confusion matrices of CatBoost and FitCNet also shows that the CatBoost model significantly reduces cross-reactions in adjacent operating conditions such as "upper / lower section wet and dirty" and "upper / lower section damp".

[0171]

[0172] Additionally, 74 test sets without arrester condition degradation labels were input into the trained CatBoost model for inference, as shown in the attached figure. Figure 6 As shown, the predicted labels are completely consistent with the actual labels, with an overall accuracy of 100%, further verifying that the model can still maintain zero misclassifications on data that does not participate in cross-validation.

[0173] Example 2

[0174] This embodiment provides a surge arrester condition identification system based on joint diagnosis of current and temperature signals, including:

[0175] Data acquisition module: used to acquire the total current signal of the surge arrester under the action of the operating voltage and the temperature signal at different locations on the surface of the surge arrester;

[0176] Data processing module: used to extract resistive current signal based on full current signal, decompose resistive current signal using fast Fourier algorithm to obtain peak values ​​of each harmonic of resistive current; used to invert the internal valve data of surge arrester based on temperature data at different locations on the surface of surge arrester to obtain the maximum temperature value and normal temperature value of surge arrester valve, and calculate the maximum temperature rise value and normal temperature rise value of surge arrester valve.

[0177] Model prediction module: Based on the peak value of the total current and the peak values ​​of each harmonic of the resistive current, the maximum temperature rise of the arrester valve, and the normal temperature rise of the arrester valve, it extracts key and auxiliary features that characterize the health status of the arrester and inputs them into the constructed and trained arrester status recognition model for prediction, and outputs the arrester's status type and / or status degree.

[0178] Furthermore, the peak values ​​of each harmonic of the resistive current include the peak value of the fundamental resistive current, the peak value of the third harmonic resistive current, the peak value of the fifth harmonic resistive current, and the peak value of the seventh harmonic resistive current; the internal temperature data of the surge arrester includes the highest temperature of the surge arrester valve plate and the normal temperature of the valve plate; the surface temperature data of the surge arrester includes the highest surface temperature of the surge arrester and the normal surface temperature.

[0179] Furthermore, based on the extraction of the resistive current signal from the full current signal, the resistive current signal is decomposed using the Fast Fourier Transform algorithm to obtain the peak values ​​of each harmonic of the resistive current. The specific process is as follows:

[0180] Extracting resistive current from a full-current signal;

[0181] Perform a Fourier operation on the resistive current signal of at least two cycles every N sampling points to obtain the total current peak, fundamental resistive current peak, third harmonic resistive current peak, fifth harmonic resistive current peak, and seventh harmonic resistive current peak.

[0182] Furthermore, based on the extraction of the resistive current signal from the full current signal, the resistive current signal is decomposed using the Fast Fourier Transform algorithm. After obtaining the peak values ​​of each harmonic of the resistive current, the following is also included:

[0183] Based on the surface insulation contamination of surge arresters, the peak value of the total current Peak value of fundamental resistive current Third harmonic resistive current peak value Fifth harmonic resistive current peak value Peak value of the seventh harmonic resistive current The data has been corrected, and the specific corrected data is as follows:

[0184]

[0185]

[0186]

[0187]

[0188]

[0189] In the formula, , , , , These are the peak values ​​of the total current, the fundamental resistive current, the third harmonic resistive current, the fifth harmonic resistive current, and the seventh harmonic resistive current, respectively, after correction based on the surface insulation contamination of the surge arrester. , , , , These are the peak values ​​of the total current, fundamental resistive current, third harmonic resistive current, fifth harmonic resistive current, and seventh harmonic resistive current before correction based on the surface insulation contamination of the surge arrester. , , , , This is the soiling correction factor; Ambient air humidity; The equivalent salt density is applied to the outer insulation surface of the surge arrester.

[0190] Furthermore, the formula for calculating the maximum temperature rise of the surge arrester varistor is as follows:

[0191]

[0192] The formula for calculating the normal temperature rise of surge arrester varistors is:

[0193]

[0194] in, This refers to the highest temperature value of the surge arrester valve plate; This refers to the normal temperature value of the surge arrester valve plate. Surge arrester operating ambient temperature; The maximum temperature rise of the surge arrester varistor; This represents the normal temperature rise value of the surge arrester valve plate.

[0195] Furthermore, key characteristic quantities include peak total current. Peak value of fundamental resistive current Valve plate maximum temperature rise Auxiliary characteristic quantities include the rate of change of total current. Rate of change of resistive current Third harmonic characteristic coefficient RC characteristic coefficient Valve plate temperature rise rate Valve plate temperature rise fluctuation rate The specific calculation formula is as follows:

[0196]

[0197]

[0198]

[0199] in, , , These are the peak value of the total current, the peak value of the fundamental resistive current, and the average temperature rise of the valence plate, obtained when the operating voltage is applied to the surge arrester sample under normal conditions. This represents the peak value of the total current of the surge arrester. This represents the peak value of the fundamental resistive current. The peak value of the third harmonic resistive current, The peak value of the fifth harmonic resistive current , The peak value of the seventh harmonic resistive current; The maximum temperature rise of the surge arrester varistor; This represents the normal temperature rise value of the surge arrester valve plate.

[0200] Furthermore, the surge arrester condition identification model is obtained by applying the rated operating voltage at power frequency to surge arrester samples in different health states through experimental methods, acquiring the current and temperature electrothermal response time series data of the surge arrester under the action of the operating voltage, extracting key feature quantities and auxiliary feature quantities characterizing the health state of the surge arrester, preprocessing them to construct training set data for various health states, and using the actual state labels of the surge arrester and the training set data to train and optimize the parameters of the CatBoost model.

[0201] Furthermore, the training process of the surge arrester state recognition model is as follows:

[0202] The feature values ​​of the preprocessed training set data are randomly sorted in multiple groups, and the model is initialized; the preprocessing includes outlier replacement and normalization.

[0203] During the Boosting main loop phase, for each iteration, the residual of the current sample is calculated, and a symmetric binary tree model with a preset depth is trained based on the residual. At each leaf node, a splitting scheme that minimizes the loss function is automatically selected, and each leaf node corresponds to a subset of samples and a predicted value. At the same time, the prediction function of this round is weighted according to the learning rate and accumulated into the previous round model to form a new prediction function. The residual is the negative gradient of the loss function in the output direction of the current model.

[0204] The Softmax multi-class cross-entropy loss function value is calculated as the criterion for convergence. Early stopping is triggered when the loss function does not decrease significantly in a preset number of iterations, or training is terminated when the number of iterations reaches a preset upper limit, thus completing the CatBoost model training.

[0205] Secondly, the present invention provides a surge arrester condition identification system based on joint diagnosis of current and temperature signals, comprising:

[0206] Data acquisition module: used to acquire the full current signal of the surge arrester under the action of the operating voltage and the temperature data at different locations on the surface of the surge arrester;

[0207] Data processing module: used to extract resistive current signal based on full current signal, decompose resistive current signal using fast Fourier algorithm to obtain peak values ​​of each harmonic of resistive current; used to inversely infer internal temperature data of arrester based on temperature data at different locations on the arrester surface to obtain the maximum temperature value and normal temperature value of arrester valence, and then calculate and obtain the maximum temperature rise value and normal temperature rise value of arrester valence;

[0208] Model prediction module: Based on the peak value of the total current and the peak values ​​of each harmonic of the resistive current, the maximum temperature rise of the arrester valve, and the normal temperature rise of the arrester valve, it extracts key and auxiliary features that characterize the health status of the arrester and inputs them into the constructed and trained arrester status recognition model for prediction, and outputs the arrester's status type and / or status degree.

[0209] Furthermore, the peak values ​​of each harmonic of the resistive current include the peak value of the fundamental resistive current, the peak value of the third harmonic resistive current, the peak value of the fifth harmonic resistive current, and the peak value of the seventh harmonic resistive current; the surface temperature data of the surge arrester includes the highest surface temperature and the normal surface temperature of the surge arrester.

[0210] Furthermore, based on the extraction of the resistive current signal from the full current signal, the resistive current signal is decomposed using the Fast Fourier Transform algorithm to obtain the peak values ​​of each harmonic of the resistive current. The specific process is as follows:

[0211] Extracting resistive current from a full-current signal;

[0212] Perform a Fourier operation on the resistive current signal of at least two cycles every N sampling points to obtain the total current peak, fundamental resistive current peak, third harmonic resistive current peak, fifth harmonic resistive current peak, and seventh harmonic resistive current peak.

[0213] Furthermore, based on the extraction of the resistive current signal from the full current signal, the resistive current signal is decomposed using the Fast Fourier Transform algorithm. After obtaining the peak values ​​of each harmonic of the resistive current, the following is also included:

[0214] Based on the surface insulation contamination of surge arresters, the peak value of the total current Peak value of fundamental resistive current Third harmonic resistive current peak value Fifth harmonic resistive current peak value Peak value of the seventh harmonic resistive current The data has been corrected, and the specific corrected data is as follows:

[0215]

[0216]

[0217]

[0218]

[0219]

[0220] In the formula, , , , , These are the peak values ​​of the total current, the fundamental resistive current, the third harmonic resistive current, the fifth harmonic resistive current, and the seventh harmonic resistive current, respectively, after correction based on the surface insulation contamination of the surge arrester. , , , , These are the peak values ​​of the total current, fundamental resistive current, third harmonic resistive current, fifth harmonic resistive current, and seventh harmonic resistive current before correction based on the surface insulation contamination of the surge arrester. , , , , This is the soiling correction factor; Ambient air humidity; The equivalent salt density is applied to the outer insulation surface of the surge arrester.

[0221] Furthermore, the formula for calculating the maximum temperature rise of the surge arrester varistor is as follows:

[0222]

[0223] The formula for calculating the normal temperature rise of surge arrester varistors is:

[0224]

[0225] in, This refers to the highest temperature value of the surge arrester valve plate; This refers to the normal temperature value of the surge arrester valve plate. Surge arrester operating ambient temperature; The maximum temperature rise of the surge arrester varistor; This represents the normal temperature rise value of the surge arrester valve plate.

[0226] Furthermore, the key characteristic quantity includes the peak value of the total current. Peak value of fundamental resistive current Valve plate maximum temperature rise Auxiliary characteristic quantities include the rate of change of total current. Rate of change of resistive current Third harmonic characteristic coefficient RC characteristic coefficient Valve plate temperature rise rate Valve plate temperature rise fluctuation rate The specific calculation formula is as follows:

[0227]

[0228]

[0229]

[0230] in, , , The peak value of the total current, the peak value of the fundamental resistive current, and the average temperature rise of the valence plate were obtained when the operating voltage was applied to the surge arrester sample under normal conditions. This represents the peak value of the total current of the surge arrester. This represents the peak value of the fundamental resistive current. The peak value of the third harmonic resistive current, The peak value of the fifth harmonic resistive current , The peak value of the seventh harmonic resistive current; The maximum temperature rise of the surge arrester varistor; This represents the normal temperature rise value of the surge arrester valve plate.

[0231] Furthermore, the surge arrester condition identification model is obtained by applying the rated operating voltage at power frequency to surge arrester samples in different health states through experimental methods, acquiring the current and temperature electrothermal response time series data of the surge arrester under the action of the operating voltage, extracting key feature quantities and auxiliary feature quantities characterizing the health state of the surge arrester from the data, constructing training set data for various health states after preprocessing, and using the actual state label of the surge arrester and the training set data to train and optimize the parameters of the CatBoost model.

[0232] Furthermore, the training process of the surge arrester state recognition model is as follows:

[0233] The feature values ​​of the preprocessed training set data are randomly sorted in multiple groups, and the model is initialized; the preprocessing includes outlier replacement and normalization.

[0234] During the Boosting main loop phase, for each iteration, the residual of the current sample is calculated, and a symmetric binary tree model with a preset depth is trained based on the residual. At each leaf node, a splitting scheme that minimizes the loss function is automatically selected, and each leaf node corresponds to a subset of samples and a predicted value. At the same time, the prediction function of this round is weighted according to the learning rate and accumulated into the previous round model to form a new prediction function. The residual is the negative gradient of the loss function in the output direction of the current model.

[0235] The Softmax multi-class cross-entropy loss function value is calculated as the criterion for convergence. Early stopping is triggered when the loss function does not decrease significantly in a preset number of iterations, or training is terminated when the number of iterations reaches a preset upper limit, thus completing the CatBoost model training.

[0236] Example 3

[0237] This embodiment provides an electronic terminal including a processor and a memory, wherein the memory stores a computer program, and the processor invokes the computer program to perform the steps of the method described above.

[0238] Example 4

[0239] This embodiment provides a readable storage medium storing a computer program that, when invoked by a processor, performs the steps of the method described above.

[0240] It should be understood that, in the embodiments of the present invention, the processor may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. The memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.

[0241] The readable storage medium is a computer-readable storage medium, which can be an internal storage unit of the controller described in any of the foregoing embodiments, such as the controller's hard drive or memory. The readable storage medium can also be an external storage device of the controller, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the controller. Further, the readable storage medium can include both the controller's internal storage unit and external storage devices. The readable storage medium is used to store the computer program and other programs and data required by the controller. The readable storage medium can also be used to temporarily store data that has been output or will be output.

[0242] Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned readable storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0243] It is understood that the same or similar parts in the above embodiments can be referred to each other, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0244] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A surge arrester condition identification method based on joint diagnosis of current and temperature signals, characterized in that, include: Acquire the full current signal of the surge arrester under operating voltage and the temperature data at different locations on the surface of the surge arrester; The resistive current signal is extracted based on the full current signal, and the resistive current signal is decomposed using the fast Fourier algorithm to obtain the peak values ​​of each harmonic of the resistive current. Based on the temperature data at different locations on the surface of the surge arrester, the internal temperature data of the surge arrester is obtained by inversion and estimation to obtain the maximum temperature value and normal temperature value of the surge arrester valve. The maximum temperature rise value and normal temperature rise value of the surge arrester valve are then calculated. Based on the peak value of the total current and the peak values ​​of each harmonic of the resistive current, the maximum temperature rise of the surge arrester valve, and the normal temperature rise of the surge arrester valve, key and auxiliary characteristic quantities that characterize the health status of the surge arrester are extracted. The key features and auxiliary features are used as inputs to the constructed and trained arrester state recognition model for prediction, and the arrester state type and / or state degree are output.

2. The method according to claim 1, characterized in that, The peak values ​​of each harmonic of the resistive current include the peak value of the fundamental resistive current, the peak value of the third harmonic resistive current, the peak value of the fifth harmonic resistive current, and the peak value of the seventh harmonic resistive current; the temperature data of different locations on the surface of the surge arrester include the highest surface temperature and the normal surface temperature.

3. The method according to claim 2, characterized in that, The process of extracting resistive current signals based on full current signals and decomposing the resistive current signals using the Fast Fourier Transform algorithm to obtain the peak values ​​of each harmonic of the resistive current is as follows: Extracting resistive current from a full-current signal; Perform a Fourier operation on the resistive current signal of at least two cycles every N sampling points to obtain the peak value of the fundamental resistive current, the peak value of the third harmonic resistive current, the peak value of the fifth harmonic resistive current, and the peak value of the seventh harmonic resistive current.

4. The method according to claim 3, characterized in that, The resistive current signal is extracted based on the total current signal. The resistive current signal is then decomposed using the Fast Fourier Transform algorithm. After obtaining the peak values ​​of each harmonic of the resistive current, the following steps are also taken: Based on the surface insulation contamination of surge arresters, the peak value of the total current Peak value of fundamental resistive current Third harmonic resistive current peak value Fifth harmonic resistive current peak value Peak value of the seventh harmonic resistive current The data has been corrected, and the specific corrected data is as follows: ； ； ； ； ； In the formula, , , , , These are the peak values ​​of the total current, the fundamental resistive current, the third harmonic resistive current, the fifth harmonic resistive current, and the seventh harmonic resistive current, respectively, after correction based on the surface insulation contamination of the surge arrester. , , , , These are the peak values ​​of the total current, fundamental resistive current, third harmonic resistive current, fifth harmonic resistive current, and seventh harmonic resistive current before correction based on the surface insulation contamination of the surge arrester. , , , , This is a correction factor for the degree of filth. Ambient air humidity; The equivalent salt density is applied to the outer insulation surface of the surge arrester.

5. The method according to claim 2, characterized in that, The formula for calculating the maximum temperature rise of the surge arrester varistor is: ； The formula for calculating the normal temperature rise of surge arrester varistors is: ； in, This refers to the highest temperature value of the surge arrester valve plate; This refers to the normal temperature value of the surge arrester valve plate. Surge arrester operating ambient temperature; The maximum temperature rise of the surge arrester varistor; This represents the normal temperature rise value of the surge arrester valve plate.

6. The method according to claim 1, characterized in that, Key characteristic quantities include peak total current. Peak value of fundamental resistive current Valve plate maximum temperature rise ; Auxiliary characteristic quantities include the rate of change of total current. Rate of change of resistive current Third harmonic characteristic coefficient RC characteristic coefficient Valve plate temperature rise rate Valve plate temperature rise fluctuation rate The specific calculation formula is as follows: ； ； ； in, , , The peak value of the total current, the peak value of the fundamental resistive current, and the average temperature rise of the valence plate were obtained when the operating voltage was applied to the surge arrester sample under normal conditions. This represents the peak value of the total current of the surge arrester. This represents the peak value of the fundamental resistive current. The peak value of the third harmonic resistive current, The peak value of the fifth harmonic resistive current , The peak value of the seventh harmonic resistive current; The maximum temperature rise of the surge arrester varistor; This represents the normal temperature rise value of the surge arrester valve plate.

7. The method according to claim 1, characterized in that, The surge arrester condition identification model is obtained by applying the rated operating voltage at power frequency to surge arrester samples in different health states through experimental methods, acquiring the current and temperature electrothermal response time series data of the surge arrester under the operating voltage, extracting key and auxiliary features characterizing the health state of the surge arrester, preprocessing them to construct training set data for various health states, and using the actual state labels of the surge arrester and the training set data to train and optimize the parameters of the CatBoost model.

8. The method according to claim 7, characterized in that, The training process for the surge arrester state recognition model is as follows: The feature values ​​of the preprocessed training set data are randomly sorted in multiple groups, and the model is initialized; the preprocessing includes outlier replacement and normalization. During the Boosting main loop phase, for each iteration, the residual of the current sample is calculated, and a symmetric binary tree model with a preset depth is trained based on the residual. At each leaf node, a splitting scheme that minimizes the loss function is automatically selected, and each leaf node corresponds to a subset of samples and a predicted value. At the same time, the prediction function of this round is weighted according to the learning rate and accumulated into the previous round model to form a new prediction function. The residual is the negative gradient of the loss function in the output direction of the current model. The Softmax multi-class cross-entropy loss function value is calculated as the criterion for convergence. Early stopping is triggered when the loss function does not decrease significantly in a preset number of iterations, or training is terminated when the number of iterations reaches a preset upper limit, thus completing the CatBoost model training.

9. A surge arrester condition identification system based on joint diagnosis of current and temperature signals, characterized in that, include: Data acquisition module: used to acquire the full current signal of the surge arrester under the action of the operating voltage and the temperature data at different locations on the surface of the surge arrester; Data processing module: used to extract resistive current signal based on full current signal, and use fast Fourier algorithm to decompose resistive current signal to obtain peak values ​​of each harmonic of resistive current; used to inversely infer internal temperature data of arrester based on temperature data at different locations on the arrester surface to obtain the maximum temperature value and normal temperature value of arrester valence, and calculate the maximum temperature rise value and normal temperature rise value of arrester valence; Model prediction module: Based on the peak value of the total current and the peak values ​​of each harmonic of the resistive current, the maximum temperature rise of the surge arrester valve, and the normal temperature rise of the surge arrester valve, it extracts key and auxiliary features that characterize the health status of the surge arrester and uses them as inputs. These inputs are then fed into the constructed and trained surge arrester status recognition model for prediction, and the output is the state type / or state degree of the surge arrester.

10. The system according to claim 9, characterized in that, The peak values ​​of each harmonic of the resistive current include the peak value of the fundamental resistive current, the peak value of the third harmonic resistive current, the peak value of the fifth harmonic resistive current, and the peak value of the seventh harmonic resistive current; the temperature data of different locations on the surface of the surge arrester include the highest surface temperature and the normal surface temperature.

11. The system according to claim 10, characterized in that, The process of extracting resistive current signals based on full current signals and decomposing the resistive current signals using the Fast Fourier Transform algorithm to obtain the peak values ​​of each harmonic of the resistive current is as follows: Extracting resistive current from a full-current signal; Perform a Fourier operation on the resistive current signal of at least two cycles every N sampling points to obtain the peak value of the fundamental resistive current, the peak value of the third harmonic resistive current, the peak value of the fifth harmonic resistive current, and the peak value of the seventh harmonic resistive current.

12. The system according to claim 11, characterized in that, The resistive current signal is extracted based on the total current signal. The resistive current signal is then decomposed using the Fast Fourier Transform algorithm. After obtaining the peak values ​​of each harmonic of the resistive current, the following steps are also taken: Based on the surface insulation contamination of surge arresters, the peak value of the total current Peak value of fundamental resistive current Third harmonic resistive current peak value Fifth harmonic resistive current peak value Peak value of the seventh harmonic resistive current The data has been corrected, and the specific corrected data is as follows: ； ； ； ； ； In the formula, , , , , These are the peak values ​​of the total current, the fundamental resistive current, the third harmonic resistive current, the fifth harmonic resistive current, and the seventh harmonic resistive current, respectively, after correction based on the surface insulation contamination of the surge arrester. , , , , These are the peak values ​​of the total current, fundamental resistive current, third harmonic resistive current, fifth harmonic resistive current, and seventh harmonic resistive current before correction based on the surface insulation contamination of the surge arrester. , , , , This is a correction factor for the degree of filth. Ambient air humidity; The equivalent salt density is applied to the outer insulation surface of the surge arrester.

13. The system according to claim 10, characterized in that, The formula for calculating the maximum temperature rise of the surge arrester varistor is: ； The formula for calculating the normal temperature rise of surge arrester varistors is: ； in, This refers to the highest temperature value of the surge arrester valve plate; This refers to the normal temperature value of the surge arrester valve plate. Surge arrester operating ambient temperature; The maximum temperature rise of the surge arrester varistor; This represents the normal temperature rise value of the surge arrester valve plate.

14. The system according to claim 9, characterized in that, Key characteristic quantities include peak total current. Peak value of fundamental resistive current Valve plate maximum temperature rise ; Auxiliary characteristic quantities include the rate of change of total current. Rate of change of resistive current Third harmonic characteristic coefficient RC characteristic coefficient Valve plate temperature rise rate Valve plate temperature rise fluctuation rate The specific calculation formula is as follows: ； ； ； in, , , These are the peak value of the total current, the peak value of the fundamental resistive current, and the average temperature rise of the valence plate, obtained when the operating voltage is applied to the surge arrester sample under normal conditions. This represents the peak value of the total current of the surge arrester. This represents the peak value of the fundamental resistive current. The peak value of the third harmonic resistive current, The peak value of the fifth harmonic resistive current , The peak value of the seventh harmonic resistive current; The maximum temperature rise of the surge arrester varistor; This represents the normal temperature rise value of the surge arrester valve plate.

15. The system according to claim 9, characterized in that, The surge arrester condition identification model is obtained by applying the rated operating voltage at power frequency to surge arrester samples in different health states through experimental methods, acquiring the current and temperature electrothermal response time series data of the surge arrester under the operating voltage, extracting key and auxiliary features characterizing the health state of the surge arrester, preprocessing them to construct training set data for various health states, and using the actual state labels of the surge arrester and the training set data to train and optimize the parameters of the CatBoost model.

16. The system according to claim 15, characterized in that, The training process for the surge arrester state recognition model is as follows: The feature values ​​of the preprocessed training set data are randomly sorted in multiple groups, and the model is initialized; the preprocessing includes outlier replacement and normalization. During the Boosting main loop phase, for each iteration, the residual of the current sample is calculated, and a symmetric binary tree model with a preset depth is trained based on the residual. At each leaf node, a splitting scheme that minimizes the loss function is automatically selected, and each leaf node corresponds to a subset of samples and a predicted value. At the same time, the prediction function of this round is weighted according to the learning rate and accumulated into the previous round model to form a new prediction function. The residual is the negative gradient of the loss function in the output direction of the current model. The Softmax multi-class cross-entropy loss function value is calculated as the criterion for convergence. Early stopping is triggered when the loss function does not decrease significantly in a preset number of iterations, or training is terminated when the number of iterations reaches a preset upper limit, thus completing the CatBoost model training.

17. An electronic terminal, characterized in that: It includes a processor and a memory, the memory storing a computer program, the processor calling the computer program to perform the steps of the method according to any one of claims 1-8.

18. A readable storage medium, characterized in that: A computer program is stored, which, when invoked by a processor, performs the steps of the method according to any one of claims 1-8.

Images