Surge arrester leakage current monitoring device and method based on multi-sensor fusion

Abstract

The application provides a multi-sensor fusion arrester leakage current monitoring device and method, relates to the technical field of arrester monitoring, and the method collects a leakage current signal generated along an arrester ground path through a tunnel magnetoresistance sensor and combines an auxiliary sensor to collect a transient current signal; when the transient signal meets a trigger condition, a pseudo signal compensated by temperature is generated by calling a pseudo modeling module, and the leakage current is corrected; then, time-frequency analysis is performed on the corrected signal, steady-state and transient aging characteristics are extracted, environmental data is input into a trend analysis model, and an arrester aging evaluation result is generated. The method can obtain high-precision and traceable leakage current data in the environment of sensor distortion and multi-source interference, effectively enhances the accuracy and practicality of arrester state evaluation, and is suitable for operation and maintenance scenes in intelligent substations and power transmission and distribution systems.

Description

Technical Field

[0001] This application relates to the field of surge arrester monitoring technology, and more specifically, to surge arrester leakage current monitoring equipment and method using multi-sensor fusion. Background Technology

[0002] In power systems, surge arresters are critical lightning protection devices, and monitoring their leakage current is an important means of assessing their operational status. Current technologies primarily employ traditional series-connected piezoelectric or electromechanical monitoring devices, directly connected to the surge arrester circuit to measure leakage current and the number of lightning strikes. However, these methods have significant drawbacks: series connection reduces overall system reliability, and a failure of the monitoring device may lead to arrester failure; simultaneously, under transient current surges, sensors are prone to saturation distortion, introducing spurious harmonic components and interfering with the accurate extraction of leakage current signals; furthermore, environmental factors such as temperature, humidity, and moisture pollution can further couple interference, resulting in high noise and low accuracy in monitoring data, making high-frequency, real-time monitoring impossible. While existing non-contact electromagnetic induction methods avoid the risks of series connection, they are still affected by the nonlinear distortion of the sensor itself and environmental noise, making it difficult to obtain high-precision data, and lack a long-term traceability mechanism for historical data, making it difficult to reliably assess the overall aging status of the surge arrester.

[0003] Therefore, it is necessary to solve how to obtain high-precision, long-term traceable surge arrester leakage current data under conditions of transient current surges, sensor saturation distortion, and environmental interference, and to reliably assess the overall aging status of surge arresters based on this data. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this application provides a multi-sensor fusion-based surge arrester leakage current monitoring device and method.

[0005] Firstly, this application provides a method for monitoring the leakage current of surge arresters using multi-sensor fusion, including:

[0006] The leakage current signal generated along the ground leakage path of the surge arrester is collected using a tunnel magnetoresistive sensor.

[0007] The transient current signal flowing through the arrester body or the down conductor is collected by an auxiliary sensor;

[0008] In response to at least one transient feature of the transient current signal satisfying a preset trigger condition, an artifact signal corresponding to the transient feature is determined using a preset distortion model; the distortion model is used to characterize the distortion response of the tunnel magnetoresistive sensor under different transient features.

[0009] based on the artifact signal, correcting the leakage current signal to generate a corrected leakage current signal; performing time domain and / or frequency domain feature analysis on the corrected leakage current signal to determine an aging feature, and collecting environmental data associated with the aging feature;

[0010] For the aging feature and the environmental data, a trend analysis model deployed in a cloud platform is used for analysis to generate monitoring result information.

[0011] Optionally, the distortion model is a distortion signature matrix fixed in the firmware of the edge computing unit;

[0012] The edge computing unit is in communication connection with the tunnel magnetoresistance sensor and the auxiliary sensor.

[0013] The distortion signature matrix represents the mapping relationship between the amplitude and rise rate of the input current and the frequency component and amplitude of the corresponding artifact harmonic.

[0014] Optionally, the use of a preset distortion model to determine the artifact signal corresponding to the transient feature includes:

[0015] Based on the transient feature, at least two discrete calibration points adjacent to the transient feature in the distortion signature matrix are found, and the reference artifact signals of the discrete calibration points are obtained respectively;

[0016] The current operating temperature of the tunnel magnetoresistance sensor is obtained.

[0017] The reference artifact signals of the at least two discrete calibration points are taken as interpolation references, the transient feature is taken as interpolation input, nonlinear interpolation is performed to generate an interpolated artifact signal.

[0018] Based on the current operating temperature, a temperature compensation function established in advance is used to perform temperature compensation on the interpolated artifact signal to generate a temperature-compensated artifact signal.

[0019] Optionally, the generation of the corrected leakage current signal includes:

[0020] The temperature-compensated artifact signal is subtracted from the leakage current signal to obtain a residual signal.

[0021] Based on the transient feature, a time-varying response scale factor representing the magnetic saturation state recovery process of the tunnel magnetoresistance sensor after a transient current impact is determined;

[0022] The time-varying response scale factor is determined by a dynamic response model constructed in advance, and the dynamic response model is constructed based on the magnetic saturation and recovery characteristics of the tunnel magnetoresistance sensor under different amplitude and rise rate of the transient current impact.

[0023] multiplying the residual signal by the time-varying response scaling factor to perform amplitude dynamic compensation on the residual signal to obtain the corrected leakage current signal.

[0024] Optionally, the determining the time-varying response scaling factor characterizing the recovery process of the magnetic saturation state of the tunnel magnetoresistance sensor after the transient current impulse comprises:

[0025] determining, based on the transient feature and the current operating temperature, a distortion signature matrix, and determining an artifact index corresponding to the transient feature by using the distortion signature matrix;

[0026] inputting the artifact index into a pre-established saturation level classification table to determine a saturation level corresponding to the artifact index;

[0027] based on the saturation level, selecting a recovery base function matching the saturation level from a pre-constructed recovery curve function library;

[0028] calculating a dynamic parameter used to represent a time-domain shape of the recovery base function according to at least one of a continuous value of the transient feature and the artifact index;

[0029] determining the time-varying response scaling factor based on the recovery base function and the dynamic parameter.

[0030] Optionally, the determining the artifact index corresponding to the transient feature by using the distortion signature matrix comprises:

[0031] generating a temperature-compensated artifact signal by using the distortion signature matrix, wherein the artifact signal comprises a plurality of harmonic components of different frequencies;

[0032] performing Fourier transform on the temperature-compensated artifact signal to determine an amplitude of each harmonic component;

[0033] determining a corresponding saturation contribution weight coefficient for each harmonic component according to a pre-set harmonic weight library;

[0034] multiplying the amplitude of each harmonic component by the corresponding saturation contribution weight coefficient thereof, and summing the multiplication results to determine the artifact index.

[0035] Optionally, the determining the aging feature comprises:

[0036] performing wavelet packet decomposition on the corrected leakage current signal to obtain a plurality of sub-band coefficients, wherein each sub-band coefficient corresponds to a frequency segment and a time interval;

[0037] From the plurality of sub-band coefficients, a first group of sub-band coefficients corresponding to a preset steady-state aging frequency band is extracted, and an energy value of the first group of sub-band coefficients is calculated to determine a steady-state aging feature representing a long-term aging state of the surge arrester;

[0038] From the plurality of sub-band coefficients, a second group of sub-band coefficients belonging to a transient frequency band is identified, and whether an energy value of the second group of sub-band coefficients in a target time interval exceeds a set threshold value is judged to determine a transient aging feature representing a partial discharge or current impulse response;

[0039] The steady-state aging feature and the transient aging feature are combined to generate an aging feature for representing a comprehensive aging state of the surge arrester.

[0040] Optionally, the wavelet packet decomposition performed on the corrected leakage current signal to obtain the plurality of sub-band coefficients comprises:

[0041] A saturation recovery index is calculated based on the transient feature, and the saturation recovery index is compared with a preset recovery threshold value to judge whether the corrected leakage current signal is in a saturation recovery period after a transient current impulse, and a judgment result is generated;

[0042] In response to the judgment result indicating that it is in the saturation recovery period, a first wavelet mother function corresponding to the recovery period is queried and selected from a preset wavelet base function library, otherwise a second wavelet mother function corresponding to a non-recovery period is selected, and the selected wavelet mother function is generated;

[0043] A first decomposition level is set based on a target frequency band corresponding to the steady-state aging feature, and a second decomposition level is set based on a target frequency band corresponding to the transient aging feature, and an asymmetric wavelet packet decomposition structure is generated, wherein the first decomposition level is different from the second decomposition level;

[0044] The selected wavelet mother function and the asymmetric wavelet packet decomposition structure are applied to perform wavelet packet decomposition on the corrected leakage current signal to generate the plurality of sub-band coefficients.

[0045] Optionally, the environment data associated with the aging feature comprises:

[0046] A current temperature value of the tunnel magnetoresistance sensor is collected by a temperature sensor, and a current humidity value of the tunnel magnetoresistance sensor is collected by a humidity sensor;

[0047] The current temperature value and the current humidity value are matched with a collection timestamp of the aging feature to generate the environment data associated with the aging feature.

[0048] In a second aspect, the present application provides a multi-sensor fusion surge arrester leakage current monitoring device, comprising:

[0049] The acquisition module is configured to acquire, by a tunnel magnetoresistance sensor, a leakage current signal generated along a leakage path of a lightning arrester to ground; and acquire, by an auxiliary sensor, a transient current signal flowing through the lightning arrester body or a down conductor;

[0050] The processing module is configured to, in response to at least one transient characteristic of the transient current signal satisfying a preset triggering condition, determine a artifact signal corresponding to the transient characteristic by using a preset distortion model, and the distortion model is used to characterize a distortion response of the tunnel magnetoresistance sensor under different transient characteristics.

[0051] The analysis module is configured to correct the leakage current signal based on the artifact signal to generate a corrected leakage current signal, and perform time domain and / or frequency domain characteristic analysis on the corrected leakage current signal to determine an aging characteristic and acquire environmental data associated with the aging characteristic.

[0052] The cloud output module is configured to analyze the aging characteristic and the environmental data by using a trend analysis model deployed in a cloud platform to generate monitoring result information.

[0053] Compared with the prior art, the tunnel magnetoresistance sensor is used to realize non-contact leakage current acquisition, the auxiliary sensor is used to capture a transient current signal, a preset distortion model is used to determine and correct artifact interference in real time, and the high precision and reliability of monitoring data are ensured. Meanwhile, time-frequency domain characteristic analysis is performed on the corrected signal, an aging characteristic is extracted, and associated environmental data is acquired and uploaded to a cloud platform for comprehensive evaluation by using a trend analysis model, so that dynamic tracking and long-term tracing of the deterioration trend of the lightning arrester are realized. Compared with the prior art, the method effectively avoids the reliability risk of traditional series measurement, reduces artifact caused by transient impact and environmental interference, improves the accuracy and real-time performance of data acquisition, supports whole life cycle monitoring, and provides more scientific fault diagnosis and maintenance strategies through cloud-edge collaborative mechanism. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 A flowchart of a lightning arrester leakage current monitoring method provided by an embodiment of the present application is provided.

[0055] Figure 2 A flowchart of a method for determining an artifact signal corresponding to a transient characteristic is provided.

[0056] Figure 3 A flowchart of a method for generating a corrected leakage current signal is provided.

[0057] Figure 4 A schematic diagram of a lightning arrester leakage current monitoring device provided by an embodiment of the present application is provided.

[0058] Figure 5 An installation schematic diagram of a digital monitoring device of a lightning arrester is provided in the embodiments of the present application.

[0059] The figure reference: 10, acquisition module; 20, processing module; 30, analysis module; 40, cloud output module; 100, tower installation plate; 200, iron tower angle steel; 300, antenna; 400, M10 fixing bolt; 500, detection and early warning device; 600, lightning arrester insulating down lead; 700, antenna connection. DETAILED DESCRIPTION

[0060] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments.

[0061] Referring to Figure 1 The figure reference: 10, acquisition module; 20, processing module; 30, analysis module; 40, cloud output module; 100, tower installation plate; 200, iron tower angle steel; 300, antenna; 400, M10 fixing bolt; 500, detection and early warning device; 600, lightning arrester insulating down lead; 700, antenna connection.

[0062] S101: Collecting a leakage current signal generated along a leakage path of a lightning arrester to ground through a tunnel magnetoresistance sensor;

[0063] S102: Collecting a transient current signal flowing through the lightning arrester body or the down lead through an auxiliary sensor;

[0064] S103: In response to at least one transient feature of the transient current signal satisfying a preset trigger condition, determining an artifact signal corresponding to the transient feature by using a preset distortion model; the distortion model is used to characterize the distortion response of the tunnel magnetoresistance sensor under different transient features;

[0065] S104: Correcting the leakage current signal based on the artifact signal to generate a corrected leakage current signal; performing time domain and / or frequency domain feature analysis on the corrected leakage current signal to determine an aging feature, and collecting environmental data associated with the aging feature;

[0066] S105: Using a trend analysis model deployed in a cloud platform to analyze the aging feature and the environmental data to generate monitoring result information.

[0067] For the above S101:

[0068] In a specific implementation, this step uses a tunnel magnetoresistance sensor (TMR sensor) as a non-contact current sensing element for collecting the leakage current of the surge arrester. The TMR sensor works on the basis of the tunnel magnetoresistance effect, and its internal structure contains a magnetic sensitive layer. When the external magnetic field changes, the resistance value of the magnetic sensitive layer changes accordingly, thereby indirectly reflecting the magnetic field strength generated by the current. The sensor can be mechanically installed on the surge arrester support and precisely positioned so that its sensing ring is in close contact with the grounding downlead of the surge arrester but is not directly connected in series in the circuit, thereby avoiding interference with the original electrical performance and reliability of the surge arrester.

[0069] During the collection process, the TMR sensor sensing ring is first wrapped around the ground leakage path of the surge arrester, for example, fixed near the counter at the bottom of the surge arrester through an insulating support, ensuring that the vertical distance between the sensing ring and the leakage current path is maintained between 5-10 mm to obtain sufficient magnetic field coupling strength.

[0070] When the surge arrester is in operation, the leakage current flows along the ground path, generating a corresponding magnetic field that acts on the magnetic sensitive layer of the TMR sensor, causing a change in its resistance. Subsequently, the signal conditioning circuit inside the sensor converts the resistance change into a voltage signal, for example, amplifies the change through a Wheatstone bridge configuration, and outputs an analog voltage signal. The analog voltage signal is then converted into a digital signal by an analog-to-digital converter, and the sampling frequency can be set to 1000-5000 times per second to capture the subtle fluctuations of microampere-level leakage current.

[0071] For example, in the scenario of monitoring a surge arrester in a substation with a voltage level of 10 kV, a commercially available TMR sensor with high sensitivity can be selected, such as a model with a sensitivity higher than 10 mV per Gauss. The sensor is mechanically fixed to the support of the surge arrester and connected to the input port of the edge computing unit through a signal line. Under normal operating conditions of the surge arrester, its leakage current usually fluctuates between 40 μA and 5 mA. The digital signal output by the sensor can accurately reflect the electrical waveform of the leakage current, which includes the fundamental component and several harmonic components.

[0072] In the above manner, this step realizes non-intrusive collection of the leakage current signal, providing basic data for subsequent signal processing.

[0073] For the above S102:

[0074] In a specific implementation, this step uses an auxiliary sensor as a transient current sensing element to collect the transient current signal flowing through the lightning arrester body or down conductor. The auxiliary sensor works on the principle of electromagnetic induction, for example, a Rogowski coil or similar high-frequency current transformer can be selected, and its internal structure includes a toroidal coil. When the transient current passes through the lightning arrester, a proportional voltage signal will be induced in the surrounding coil, thereby achieving non-contact detection of large current transients. The sensor is installed near the lightning arrester body or down conductor and is positioned by clamping or fixing the support to ensure that the coil surrounds the current path but does not directly access the circuit to maintain electrical isolation and safety of the system.

[0075] During the collection process, the coil of the auxiliary sensor is first wrapped around the lightning arrester body or down conductor, for example, fixed to the discharge end of the lightning arrester by an insulating clamp, ensuring that the distance between the coil and the current path is maintained at an appropriate distance, for example, 3 to 8 millimeters, to obtain sufficient electromagnetic coupling strength while avoiding mechanical interference.

[0076] For example, when the lightning arrester is subjected to a transient current impact, such as a lightning strike event, the current flows along the body or down conductor, generating a rapidly changing magnetic field that passes through the coil to induce a voltage signal. Subsequently, the integral circuit or signal conditioning module inside the sensor converts the voltage signal into a measurable current waveform signal, for example, by removing high-frequency noise through low-pass filtering, and outputs an analog signal. The analog signal is then converted to a digital signal by an analog-to-digital converter, with a sampling frequency set to tens of thousands to hundreds of thousands of times per second to capture the rising edge and peak characteristics of the transient current.

[0077] For example, in the monitoring scenario of a line lightning arrester with a voltage level of 35 kilovolts, a commercial auxiliary sensor with a wide dynamic range can be selected, such as a model with a response range covering thousands of amperes to hundreds of thousands of amperes. The sensor is installed on the lightning arrester down conductor through a fixed support and connected to the input port of the edge computing unit through a shielded cable. Under lightning transient conditions, the current intensity usually fluctuates above several thousand amperes, and the digital signal output by the sensor can reflect the amplitude, duration, and waveform details of the transient current. The collection process is carried out in an event-triggered mode, and the sensor is only activated when a current mutation is detected to reduce power consumption and prolong the life of the device.

[0078] In the above manner, this step realizes the rapid collection of transient current signals, providing real-time data support for subsequent signal processing.

[0079] For the above S103:

[0080] In a specific implementation, this step involves real-time analysis of the transient current signal by the edge computing unit, which activates a distortion model to determine a corresponding artifact signal when the signal meets certain conditions. The artifact signal represents the distortion component that the tunnel magnetoresistance sensor can introduce under transient events, and the distortion model is pre-constructed to characterize the sensor's response characteristics under various transient features, such as magnetic field distortion effects under different current intensities or change rates. This model can be generated through laboratory calibration data and stored in the storage medium of the edge computing unit to support fast queries.

[0081] During the response process, at least one transient feature is first extracted from the transient current signal collected by the auxiliary sensor, such as calculating the peak intensity or rising slope of the signal through a digital signal processing module.

[0082] Specifically, the processor of the edge computing unit performs window analysis on the digital signal, such as taking the data of the last few sampling points, calculating the change rate or maximum value from the baseline to the peak. Then, the extracted transient feature is compared with the preset trigger condition, which can be pre-set in the firmware, such as when the peak intensity exceeds a certain reference level or the rising slope is higher than the normal working condition, it is considered to meet the condition. This comparison operation can be realized through a simple threshold judgment circuit or software algorithm, ensuring that the response time is within milliseconds.

[0083] Once the transient feature meets the trigger condition, the system determines the artifact signal using the pre-set distortion model. Specifically, the edge computing unit queries the distortion response data matching the current transient feature from the model, such as searching for the corresponding entry according to the extracted peak intensity and rising slope, which describes the distortion waveform that the sensor can produce under similar conditions. Subsequently, the artifact signal is generated through interpolation or mapping, such as using the query data as a reference to adjust the waveform parameters in combination with the current transient feature to reflect the nonlinear response of the sensor. This process is executed on the local processor to avoid network delay.

[0084] For example, in a scenario where a high-voltage line arrester is struck by lightning, the edge computing unit extracts the rising slope and peak intensity of the transient current signal captured by the auxiliary sensor, and if these features exceed the pre-set reference level, the system immediately queries the distortion model to generate the corresponding artifact signal, which can be used for subsequent correction of the leakage current data.

[0085] For example, the distortion model can be constructed during the production phase through laboratory tests, for example using a pulse current generator to simulate current impulses of different transient characteristics, such as gradually increasing current peaks from thousands of amperes to hundreds of thousands of amperes, and recording the response data of the TMR sensor under each impulse, including waveform changes caused by magnetic field distortion. These data can be collected through multiple repeated tests, for example at least dozens of impulse tests in a controlled environment to cover common transient scenarios. Subsequently, the recorded data is organized in the form of a lookup table, for example each row in the table corresponds to a combination of transient characteristics, such as peak intensity and rise slope, and each column stores the corresponding distortion response parameters. The table is finally compiled into the firmware of the edge computing unit for runtime query. For example, during the construction process, a commercial pulse generator device can be used to apply a simulated lightning waveform to the sensor, and after collecting the response, the table entries can be generated through data processing software to ensure that the model covers the actual application range.

[0086] For the above S104:

[0087] In a specific implementation, this step first corrects the leakage current signal based on the artifact signal to remove the distortion components introduced by transient interference, generating a corrected leakage current signal; then, time domain and / or frequency domain feature analysis is performed on the corrected signal to extract key indicators reflecting the aging state of the surge arrester, and determine the aging characteristics; finally, environmental data associated with the aging characteristics is collected to support subsequent comprehensive analysis. This process is mainly performed on the edge computing unit, which utilizes its processor and storage resources to achieve efficient computation.

[0088] In the correction process, the artifact signal is first taken as a reference to remove the corresponding distortion components from the leakage current signal. Specifically, the signal processing module of the edge computing unit adjusts the digital signal of the leakage current point by point, for example by comparing the corresponding parts of the leakage current waveform and the artifact waveform, and deducting the distortion influence point by point, so as to restore the true characteristics of the original leakage current. This adjustment can be achieved through digital filtering or waveform alignment to ensure that the corrected signal is closer to the actual electrical response of the surge arrester. This correction operation is completed locally to reduce the delay, and the corrected leakage current signal is output for subsequent use.

[0089] Next, time domain and / or frequency domain feature analysis is performed on the corrected leakage current signal to determine the aging characteristics. Specifically, the edge computing unit applies signal analysis algorithms, for example time domain analysis can check the peak value, average value or resistive component of the waveform, and frequency domain analysis can extract harmonic components such as the amplitude and phase of the fundamental, third harmonic or other key frequencies through transformation methods. These characteristics reflect the aging degree of the surge arrester, for example, an increase in harmonics may indicate deterioration of the valve piece. This analysis process can be performed in segments, first dividing the window in the time domain to extract local characteristics, then converting in the frequency domain to quantify the frequency distribution, and finally combining into a set of aging characteristics.

[0090] Then, environmental data associated with the aging features are collected. Specifically, environmental parameters around the surge arrester are measured in real time by environmental sensors integrated in the device, such as temperature sensors and humidity sensors. This collection is synchronized with the analysis of aging features, for example, the current temperature and humidity values are recorded at the same time as the features are extracted, and these values are stored in association with the feature data. This environmental data is used to assist in interpreting the aging features, for example, a high temperature and high humidity environment can amplify the leakage current change.

[0091] For example, in a substation surge arrester monitoring scenario, after correcting the leakage current signal, the system extracts harmonic features as aging indicators through time-frequency analysis, and collects environmental temperature and humidity at the same time, and uploads these data together to support trend evaluation.

[0092] In the above manner, this step realizes signal correction, feature determination and data association, and provides comprehensive input for monitoring result generation.

[0093] For the above S105:

[0094] In specific implementation, this step receives the aging features and environmental data uploaded by the edge computing unit through the cloud platform, and uses the trend analysis model deployed on the platform for comprehensive processing to generate monitoring result information. The monitoring result information can include surge arrester degradation trend evaluation, potential risk warning or maintenance suggestion, etc., to support remote operation and maintenance decision. This process utilizes the computing resources and storage capacity of the cloud platform to realize deep fusion and long-term analysis of data.

[0095] In the analysis process, the edge computing unit first extracts aging features representing the operating state of the surge arrester, such as spectral change features revealed by harmonic components, and combines environmental parameters reflecting the on-site operating conditions, including temperature and humidity values, to encapsulate the above information into a structured data packet. This data packet is transmitted to the cloud platform through an Internet of Things communication module, which can realize remote communication based on fourth or fifth generation cellular network. Each data packet contains a collection timestamp and device identification information to ensure traceability in subsequent data analysis process. After receiving the data, the cloud platform first performs data preprocessing operations, including checking data integrity and removing outliers. Subsequently, the processed data is input into the trend analysis model. The model has been pre-deployed in the cloud server and is constructed based on a historical sample library, which is used to analyze the correlation between aging features and environmental factors.

[0096] Specifically, the analysis module of the cloud platform performs time series processing on the input data, e.g. comparing current aging features with historical records to identify patterns of change such as increasing trends in leakage current harmonic amplitudes. Meanwhile, environmental data is fused, e.g. considering the impact of temperature rise on features, to adjust evaluation weights through correlation analysis. The model can be run hierarchically, first calculating local trends in short-term windows, then extending to full-life cycle data for global evaluation, and finally generating monitoring result information such as descriptive reports of degradation levels or predicted remaining life. The results can be pushed to users through web interfaces or mobile applications.

[0097] For example, in a continuously monitored substation surge arrester scenario, after the cloud platform receives uploaded aging features and environmental data, it compares recent data with historical baselines through a trend analysis model, finds the correlation between harmonic increase and humidity rise, and generates result information such as "moderate degradation, recommended maintenance", and records it as a device health profile.

[0098] For example, the construction of the trend analysis model can be carried out during the initialization phase of the cloud platform, e.g. selecting samples from historical monitoring data such as aging features and environmental parameter records over the past few months or years as a training basis. These samples can be collected through batch uploading of devices, e.g. including thousands of data points of leakage current features and corresponding temperature and humidity. Subsequently, through the data processing module of the cloud server, the correlation between features and degradation patterns is established, e.g. grouping analysis of feature change trends under different environmental conditions, and forming an updateable analysis framework. This framework supports regular sample replenishment, e.g. importing new data every quarter to adjust parameters, ensuring that the model adapts to various scenarios.

[0099] For example, during the construction process, standard data analysis tools can be used to process samples to identify patterns such as the correlation between harmonic increase and humidity, and form decision rules for generating evaluation reports.

[0100] Specifically, the model can use time series analysis structures such as autoregressive integrated moving average models to predict future degradation by smoothing historical leakage current trends; or support vector regression structures to map aging features to high-dimensional spaces to fit non-linear relationships; or incorporate long short-term memory network structures to capture long-term dependency patterns and fuse environmental data as input features; in addition, harmonic analysis structures can also be combined to extract frequency components of leakage current such as the third harmonic to quantify the change trend of resistive components. These structures can be used in combination according to the data size, e.g. in small sample scenarios, prefer harmonic analysis, and in large samples, extend to neural networks to achieve flexible application from short-term local evaluation to full-life cycle prediction.

[0101] In this way, the application realizes non-contact leakage current collection through a tunneling magnetoresistance sensor, captures transient current signals in combination with an auxiliary sensor, determines and corrects artifacts in real time using a preset distortion model, ensures high precision and reliability of monitoring data, and at the same time, performs time-frequency domain feature analysis on the corrected signal, extracts aging characteristics, collects associated environmental data, uploads to a cloud platform, and uses a trend analysis model for comprehensive evaluation to realize dynamic tracking and long-term tracing of the deterioration trend of the lightning arrester. Compared with the prior art, the method effectively avoids the reliability risk of traditional series measurement, reduces artifacts caused by transient impact and environmental interference, improves the accuracy and real-time performance of data collection, supports life cycle monitoring, and provides more scientific fault diagnosis and maintenance strategies through cloud edge collaboration mechanism.

[0102] Optionally, in order to solve the problem of how to efficiently store and query the distortion model under the transient current signal triggering condition, so as to accurately characterize the nonlinear response characteristics of the tunneling magnetoresistance sensor, thereby reducing the calculation delay and improving the reliability of the artifact signal determination.

[0103] By designing the distortion model as a distortion signature matrix and solidifying it in the firmware of the edge computing unit, the principle of this method is that the matrix structure allows fast lookup of the mapping relationship, reduces real-time calculation overhead, and improves response speed; at the same time, solidification storage ensures stable operation of the model on low-power devices, avoiding external access delay; communication connection realizes seamless transmission of sensor data, supporting local processing; the mapping relationship captures the correspondence between input current characteristics and artifact harmonics, improving correction accuracy, thereby improving the stability and efficiency of the monitoring system as a whole.

[0104] In specific implementation, the distortion model adopts a distortion signature matrix form, which is a pre-constructed data table used to represent the mapping relationship between the amplitude and rise rate of the input current and the frequency component and amplitude of the corresponding artifact harmonics. Specifically, the rows of the matrix can correspond to different amplitude ranges of the input current, such as hierarchical entries from small amplitude to large amplitude, and the columns correspond to the classification of the rise rate, such as levels from slow change to rapid impact, and each cell stores the corresponding artifact harmonic data, such as the amplitude estimate of a specific frequency. The matrix is constructed through a laboratory calibration process, for example, using a current simulation device to generate various input current scenarios, recording sensor responses, and organizing the data into a table form.

[0105] The distortion signature matrix is solidified in the firmware of the edge computing unit. Specifically, the edge computing unit is an embedded processor module, such as a low-power chip based on a microcontroller, which compiles the matrix data into a read-only memory during the production stage, such as by programming tools to burn the firmware, ensuring direct access during runtime without dynamic loading. This solidification method allows the matrix to be available immediately after device startup, supporting offline operation.

[0106] The edge computing unit is communicatively connected with the tunneling magnetoresistance sensor and the auxiliary sensor. Specifically, the edge computing unit is connected with the sensors through a wired interface such as a serial peripheral interface or an analog-to-digital conversion interface, for example, the tunneling magnetoresistance sensor output is connected to the input pin of the unit, and the auxiliary sensor is accessed through a dedicated channel to ensure real-time data transmission. The connection supports bidirectional communication, for example, the unit can send a control signal to the sensor to adjust the sampling mode.

[0107] For example, in a line arrester monitoring scenario, when the transient current rise rate is high, the edge computing unit queries the solidification matrix, finds the corresponding row and column according to the input amplitude and rise rate, and obtains the frequency component and amplitude data of the artifact harmonic for subsequent processing. This method ensures that the processing time is completed within a short time.

[0108] In this way, the practicability of the distortion model is enhanced, and reliable support is provided for the monitoring method.

[0109] Optionally, referring to Figure 2 A flowchart of a method for determining an artifact signal corresponding to a transient feature provided by the embodiment of the present application, comprising steps S201-S204, wherein:

[0110] S201: Based on the transient feature, at least two discrete calibration points adjacent to the transient feature in the distortion signature matrix are found, and the reference artifact signals of the discrete calibration points are obtained respectively;

[0111] S202: The current operating temperature of the tunneling magnetoresistance sensor is obtained;

[0112] S203: The reference artifact signals of the at least two discrete calibration points are used as interpolation references, the transient feature is used as interpolation input, nonlinear interpolation is performed, and an interpolated artifact signal is generated;

[0113] S204: Based on the current operating temperature, the interpolated artifact signal is temperature compensated using a pre-established temperature compensation function, and a temperature-compensated artifact signal is generated.

[0114] Further, considering how to accurately generate an artifact signal when the transient feature does not completely match the discrete points of the distortion signature matrix, and considering the dynamic influence of the operating temperature, the accuracy and adaptability of signal compensation are improved.

[0115] The present embodiment is implemented by searching adjacent calibration points based on the transient feature, acquiring temperature, performing nonlinear interpolation and temperature compensation, the principle of which lies in that searching adjacent points provides reliable reference, nonlinear interpolation adapts to continuously changing transient feature, and temperature compensation corrects environmental induced deviation, thereby improving the generation accuracy of artifact signal as a whole, reducing distortion error, and enhancing the robustness of the system under variable temperature conditions.

[0116] In a specific implementation, the processor of the edge computing unit takes the transient feature such as current amplitude and rise rate as a query key, locates the closest entries in the matrix, for example, by comparing the feature values with the classification ranges of the matrix rows and columns, and selects two or more points closest in value. These points correspond to pre-calibrated distortion response data, and the reference artifact signal such as the waveform template of a specific harmonic is extracted from the matrix.

[0117] Secondly, the ambient temperature around the sensor is measured in real time by a temperature sensing element integrated in the sensor module or device, such as a thermistor or a digital temperature sensor. This temperature value is transmitted to the edge computing unit through an internal interface, for example, collected every few seconds to reflect the current working condition.

[0118] Then, the processor processes the reference signal using an interpolation algorithm, for example, taking the waveform data of adjacent points as endpoints, adjusting the curve fitting according to the relative position of the transient feature, such as weighted average and smooth transition of harmonic components in the amplitude dimension, so as to generate a continuous artifact signal that adapts to the current feature. This interpolation ensures smooth transition of the signal and avoids jumps between discrete points.

[0119] Further, the compensation function is pre-stored in the firmware, for example, by querying the correction coefficient table with the temperature value, scaling and adjusting the amplitude or frequency components of the interpolated signal, such as slightly amplifying certain harmonic components at high temperature, to compensate for the thermal effect of the sensor material. This function is constructed based on calibration experimental data, ensuring that the compensated signal is closer to the actual distortion.

[0120] For example, the construction of the temperature compensation function can be completed through laboratory tests during the production phase, for example, testing the sensor response in a controlled temperature environment, such as gradually changing the temperature from a lower level to a higher level, and recording the signal deviation data at each temperature, including the change in amplitude or frequency components. These data can be collected through multiple repeated tests, for example, at least dozens of experiments under fixed input conditions to cover the common temperature range. Subsequently, the recorded data is organized into a compensation reference form, for example, a correction coefficient table or a mapping relationship is generated, which is finally integrated into the firmware for runtime use. For example, during the construction process, a thermostat device can be used to apply a standard signal to the sensor, and after collecting the response, the compensation parameters are generated through data processing software to ensure that the function covers the actual application range.

[0121] Specifically, the compensation function can adopt a linear compensation structure to adjust the artifact signal amplitude by a temperature coefficient, for example, assuming that the deviation is proportional to the temperature, querying a preset coefficient table to scale and correct the signal; or adopt a polynomial compensation structure, taking temperature as input, and calculating the influence of quadratic or higher order terms step by step to fit the nonlinear deviation; or integrate a lookup table compensation structure, mapping temperature to compensation values in stages, for example, one column for every few temperature units, quickly querying and applying correction; in addition, a reference compensation structure can also be combined, using the reference signal and the temperature-dependent output for comparison to dynamically adjust the signal component. These structures can be combined according to temperature sensitivity, for example, preferentially linear compensation in mild change scenarios, and expansion to polynomials in wide ranges, to achieve flexible application from simple adjustment to precise correction.

[0122] For example, in the transient impulse scenario of the lightning arrester of the substation, when the transient feature is located between the matrix points, the system finds the adjacent calibration points, obtains the reference signal, generates the preliminary artifact waveform by interpolation, and then adjusts the amplitude according to the current temperature to output the compensation signal for correction.

[0123] In this way, the optimized artifact determination process provides accurate support for the monitoring method.

[0124] Optionally, referring to Figure 3 A flowchart of a method for generating a corrected leakage current signal provided by the embodiment of the application includes steps S301-S303, wherein:

[0125] S301: subtracting the temperature-compensated artifact signal from the leakage current signal to obtain a residual signal;

[0126] S302: determining a time-varying response scale factor representing the recovery process of the magnetic saturation state of the tunnel magnetoresistance sensor after the transient current impulse based on the transient feature; wherein the time-varying response scale factor is determined by a pre-constructed dynamic response model, and the dynamic response model is constructed based on the magnetic saturation and recovery characteristics of the tunnel magnetoresistance sensor under transient current impulses of different amplitudes and rise rates;

[0127] S303: multiplying the residual signal by the time-varying response scale factor to dynamically compensate the amplitude of the residual signal to obtain the corrected leakage current signal.

[0128] For how to further correct the leakage current signal after the tunnel magnetoresistance sensor is magnetically saturated by a transient current impulse, to compensate for the amplitude deviation in the saturation recovery process, so as to restore the true characteristics of the signal and improve the accuracy of the overall monitoring.

[0129] The present embodiment is implemented by generating a residual signal, determining a time-varying response scaling factor and performing amplitude dynamic compensation. In this way, the residual signal first removes the temperature-related artifact distortion, and the time-varying response scaling factor represents the dynamic process of saturation recovery, which ensures that the compensation adapts to transient changes through pre-constructed model queries and adjustments, thereby reducing signal distortion caused by saturation effects and improving the accuracy of correction and the robustness of the system.

[0130] In a specific implementation, the temperature-compensated artifact signal is subtracted from the leakage current signal to obtain a residual signal. Specifically, the processor of the edge computing unit first aligns the leakage current signal and the artifact signal on the time axis, for example by comparing the starting points or peak positions of the signals to ensure synchronization of corresponding sampling points. Then, a subtraction operation is performed for each sampling point, that is, the corresponding value of the artifact signal is deducted from the value of the leakage current, which can be performed point by point by a digital signal processing software module, for example by loading signal arrays in the processor memory and looping through each element to complete the deduction, thereby extracting the residual signal that removes the initial distortion. The residual signal retains the core information of the leakage current, but may still be affected by saturation recovery.

[0131] Further, based on the transient feature, a time-varying response scaling factor is determined, which represents the magnetic saturation state recovery process of the tunnel magnetoresistance sensor after a transient current impact.

[0132] Specifically, the factor is a time-varying scaling value used to adjust the residual amplitude, and the processor retrieves the recovery curve from the pre-set data according to the transient feature such as the current amplitude and rise rate, for example by comparing the feature value with known impact scenarios to generate a factor sequence that gradually decays from the impact time. This determination process takes into account the gradual nature of recovery, for example from a larger adjustment in the high saturation period to a value close to unity in the stable period.

[0133] wherein the time-varying response scaling factor is determined by a pre-constructed dynamic response model based on the magnetic saturation and recovery characteristics of the tunnel magnetoresistance sensor under different amplitude and rise rate transient current impacts.

[0134] Specifically, the model can be generated through testing in the laboratory stage, for example using a current simulation device to apply multiple impact levels and recording the output change curve of the sensor from saturation to recovery, for example by collecting signal samples within several seconds after the impact and analyzing the process of amplitude decay from the peak to stability. These curve data can be stored in categories, for example grouped by amplitude, each group containing corresponding records of recovery time and scaling coefficient. The model is finally integrated into the edge computing unit to support fast matching based on transient features.

[0135] Finally, the residual signal is multiplied by the time-varying response scaling factor to perform amplitude dynamic compensation on the residual signal to obtain the corrected leakage current signal.

[0136] Specifically, the processor applies a multiplication operation on each sample point of the residual signal with the factor of the corresponding time point, for example, using a larger factor to amplify the signal part at the initial stage of recovery to compensate for the amplitude compression caused by saturation, and then gradually reducing the factor until stable. This multiplication can be performed by an array operation module to ensure uniform adjustment of the entire signal waveform in the time dimension.

[0137] For example, in a scenario where a lightning arrester is subjected to a moderate intensity lightning strike, the artifact is first subtracted to obtain the residual, and then a factor sequence from 1.5 to 1.0 is determined based on the transient characteristics, which is multiplied point by point with the residual to output the compensated corrected signal for subsequent feature analysis.

[0138] Optionally, how to accurately classify the saturation level based on the transient characteristics and temperature after the transient current impulse, and generate a time-varying response scaling factor to adapt to the variability of the sensor recovery process, thereby improving the adaptability of the correction and the reliability of the overall monitoring.

[0139] The present embodiment realizes the determination of the artifact index, the classification of the saturation level, the selection of the recovery base function, the calculation of the dynamic parameter, and finally the determination of the factor. The principle of this method is that the artifact index quantifies the distortion degree, the classification table provides a hierarchical judgment, and the function library and the dynamic parameter allow flexible modeling, thereby capturing the nonlinear dynamics of recovery, improving the accuracy of the factor, reducing the residual error of saturation, and enhancing the correction efficiency of the system in various environments.

[0140] In specific implementation, the process of determining the time-varying response scaling factor includes the following operations. First, based on the transient characteristics and the current operating temperature, the distortion signature matrix is used to determine the artifact index corresponding to the transient characteristics. Specifically, the processor of the edge computing unit combines the transient characteristics such as current amplitude and rise rate with the temperature value as a joint input, and performs an extended query in the distortion signature matrix, for example, locates the relevant rows and columns and adjusts the reference value, to generate a comprehensive index that reflects the overall strength of the artifact. This determination can be achieved through a temperature correction mechanism of the matrix, for example, first querying the basic artifact data, and then applying temperature-related adjustment to generate the final index.

[0141] Secondly, the artifact index is input into a pre-established saturation level classification table to determine the saturation level corresponding to the artifact index. Specifically, the classification table is a pre-constructed reference table, for example, containing multiple level entries, each corresponding to a range threshold of the artifact index. The processor compares the artifact index with the threshold in the table, for example, matches the category from low saturation to high saturation level by level, and outputs the corresponding level. The table is established based on experimental data, for example, by collecting the corresponding relationship between artifacts and saturation degree through simulation impact test, and arranging into a query format.

[0142] Based on the saturation level, a recovery base function matching the saturation level is selected from a pre-constructed recovery curve function library. Specifically, the function library is a pre-stored set, for example, containing multiple curve templates, each corresponding to a saturation level. The processor retrieves the library according to the level key value, for example, selects the basic curve describing the decay from the saturation peak to the stable state. The library is constructed in the laboratory stage, for example, by recording the recovery path through multiple impact tests and storing as template data.

[0143] According to at least one of the continuous value of the transient feature and the artifact index, a dynamic parameter is calculated for representing the time-domain morphology of the recovery base function. Specifically, the processor uses the transient feature and the artifact index as input to adjust the parameter, for example, by proportional scaling or offset calculation, to generate the coefficient of the modified recovery curve, such as the time-domain extension or amplitude adjustment value. The calculation can be realized by pre-set adjustment rules, for example, mapping the feature value to the parameter range.

[0144] Based on the recovery base function and the dynamic parameter, the time-varying response scaling factor is determined. Specifically, the processor applies the dynamic parameter to the base function, for example, adjusts the shape and duration of the curve, to generate a complete time series factor that gradually changes from the impact start. The determination ensures that the factor reflects the actual recovery process.

[0145] For example, in a medium impact scenario, after determining the artifact index, it is classified as medium saturation, the corresponding base function is selected, the parameter adjustment curve is calculated according to the feature, and a factor sequence gradually decreasing from a higher value is generated for compensation.

[0146] Optionally, for the artifact processing based on the transient feature and the temperature, how to extract and quantify the saturation-related artifact index from the distortion signature matrix to support accurate modeling of the recovery process, thereby improving the adaptability of the time-varying response scaling factor.

[0147] The embodiment realizes the generation of the temperature-compensated artifact signal, the determination of the harmonic amplitude through Fourier transform, the determination of the weight coefficient, and the product summation, the principle of which is that the compensation signal integrates the temperature effect, the transform extracts the frequency component, the weight library and the summation mechanism quantify the saturation contribution, thereby generating a comprehensive index, improving the model's representation ability for nonlinear distortion, reducing the calculation error, and enhancing the correction efficiency of the system in a variable environment.

[0148] In a specific implementation, the process of determining the artifact index by using the distortion signature matrix includes the following operations. First, a temperature-compensated artifact signal is generated by using the distortion signature matrix, wherein the artifact signal includes harmonic components of different frequencies. Specifically, the processor of the edge computing unit extracts preliminary artifact data based on the transient feature query matrix, such as locating the entries corresponding to the amplitude and the rise rate, and obtaining the reference harmonic information; then, a temperature compensation mechanism is applied to adjust these data, such as querying a correction table according to the current temperature, and performing amplitude scaling on each harmonic component, to generate a compensation signal containing the fundamental wave and multiple harmonics. The generation process ensures that the signal reflects the distortion under actual temperature conditions.

[0149] Secondly, a Fourier transform is performed on the temperature-compensated artifact signal to determine the amplitude of each harmonic component. Specifically, the processor performs frequency domain conversion on the compensation signal through a digital signal processing module, such as decomposing the time series signal into frequency components, and calculating the intensity values of harmonics such as the third and fifth harmonics one by one. The transform can be realized through a preloaded algorithm library, such as processing the signal window in segments to extract the amplitude list.

[0150] According to a preset harmonic weight library, a corresponding saturation contribution weight coefficient is determined for each harmonic component. Specifically, the weight library is a pre-constructed table, such as containing the contribution factors of each harmonic frequency, and the processor queries the library according to the harmonic number, such as assigning higher weights to low-order harmonics and lower weights to high-order harmonics, to reflect their relative influence on saturation. The library is based on experimental data, such as recording the harmonic contribution through saturation test.

[0151] The amplitude of each harmonic component is multiplied by its corresponding saturation contribution weight coefficient, and the multiplication results are summed to determine the artifact index. Specifically, the processor performs multiplication on each harmonic, such as multiplying the amplitude by the weight, and accumulates all the results to generate a single index value. The summation can be completed through a loop operation module to ensure that the index comprehensively reflects the saturation degree.

[0152] Optionally, in the lightning arrester leakage current monitoring scene, how to accurately distinguish the steady-state and transient-state aging characteristics from the correction signal to accurately capture the long-term degradation such as valve aging and the short-term impact such as partial discharge, so as to realize the reliable representation of the comprehensive aging state of the lightning arrester.

[0153] The embodiment realizes the determination of the aging feature by obtaining sub-band coefficients through wavelet packet decomposition, extracting and calculating energy of a steady-state frequency band, identifying and judging a transient-state frequency band threshold, and combining to generate the aging feature. The principle of the method is that the wavelet packet decomposition provides high-resolution time-frequency analysis, the energy calculation quantifies the steady-state degradation, the threshold judgment highlights the transient-state anomaly, and the combination ensures comprehensive features, thereby improving the sensitivity and accuracy of monitoring, supporting timely early warning in power scenarios with frequent lightning or complex environment, and avoiding the limitations of traditional single analysis.

[0154] In a specific implementation, the process of determining the aging feature includes the following operations. First, wavelet packet decomposition is performed on the corrected leakage current signal to obtain a plurality of sub-band coefficients, each of which corresponds to a frequency band and a time interval. Specifically, the processor of the edge computing unit selects a suitable wavelet basis function, for example, selects a commonly used wavelet family based on the non-stationary characteristics of the signal, performs multi-level decomposition on the corrected signal, for example, gradually divides the signal into low-frequency and high-frequency sub-bands, and each level of decomposition generates a coefficient representing the energy distribution of a specific frequency band. The decomposition process starts from the original signal, recursively applies a filter bank to separate frequency components, for example, low-pass filtering retains low-frequency details, and high-pass filtering captures high-frequency transients, and finally generates a tree-structured multi-layer coefficient, each of which corresponds to a frequency interval within a time window.

[0155] Secondly, from the plurality of sub-band coefficients, a first group of sub-band coefficients corresponding to a preset steady-state aging frequency band is extracted, and the energy value of the first group of sub-band coefficients is calculated to determine a steady-state aging feature representing the long-term aging state of the lightning arrester. Specifically, the processor selects relevant sub-bands from the coefficient tree according to the preset frequency band, such as the low-frequency layer, for example, the total energy index is obtained by processing the values in the coefficient group, which reflects the long-term degradation such as the cumulative effect of aging or moisture. The preset frequency band is set based on historical experimental data, for example, focusing on the lower part of the fundamental wave and harmonics to quantify the steady-state feature.

[0156] From the plurality of sub-band coefficients, a second group of sub-band coefficients belonging to a transient frequency band is identified, and a determination is made as to whether an energy value of the second group of sub-band coefficients in a target time interval exceeds a set threshold value, to determine a transient aging feature representing a partial discharge or current impulse response. Specifically, the processor selects high frequency sub-bands from the coefficients, e.g. corresponding to transient portions other than power frequency, which reflect fast responses to lightning strikes or partial discharges. Then, in a target time interval, e.g. a short window after an impulse occurs, the energy of each sub-band coefficient is calculated and compared with a set threshold value, which is preset based on a normal signal baseline, e.g. if the energy exceeds the threshold value, it is marked as a transient anomaly. This determination can be performed on a sub-band-by-sub-band basis to isolate impulse responses. Specifically, the processor independently performs energy calculation and threshold comparison on each sub-band in the second group of sub-band coefficients, e.g. selects a high frequency sub-band first, calculates the sum of its energy in the target time interval, and then compares it with the preset threshold value, if it exceeds, it is marked as an abnormal response; repeat this process for all sub-bands in the group, so as to isolate the impulse event at a specific frequency, e.g. a transient caused by partial discharge, without affecting the analysis of other sub-bands. This sub-band-by-sub-band approach ensures the refinement of feature extraction, supporting accurate diagnosis in complex signal environments.

[0157] The steady-state aging feature and the transient aging feature are combined to generate an aging feature representing a comprehensive aging state of the surge arrester. Specifically, the processor integrates the steady-state and transient indicators, e.g. by concatenating or weightedly combining them into a vector form, and the combination takes into account the complementarity of the two, e.g. the steady-state provides a trend baseline and the transient highlights sudden events, so as to form a comprehensive feature set. This generation ensures that the feature is used for subsequent evaluation.

[0158] For example, in a monitoring scenario of a substation surge arrester subjected to lightning strikes, after wavelet packet decomposition of the corrected signal, the energy of the low frequency sub-bands is extracted as a steady-state feature, the high frequency sub-bands are identified to determine a threshold value to capture partial discharge transients, and then the aging vector is generated by combination, for cloud trend analysis. This approach can effectively distinguish between aging accumulation and transient impulse in a high voltage line scenario in a humid environment, supporting timely maintenance decisions.

[0159] In this way, the feature determination process is refined, and the comprehensiveness of the monitoring method is improved.

[0160] Optionally, in the actual scenario of surge arrester leakage current monitoring, such as high voltage lines or substation environments where lightning strikes frequently, how to adaptively adjust the wavelet packet decomposition process to cope with the non-stationary characteristics of the saturated recovery period signal, so as to optimize the time-frequency resolution between the steady-state and transient frequency bands, and accurately extract the aging feature.

[0161] The embodiment judges the recovery period by calculating the saturation recovery index, selects the mother function in response to the judgment, sets different levels to generate an asymmetric structure, and realizes the implementation by applying the decomposition to generate sub-band coefficients. The principle of this method is that the saturation index and threshold value judge and identify the signal state, the adaptive mother function and structure enhance the high-frequency resolution for the recovery period, the preset library and level setting provide flexibility, thereby improving the adaptability and efficiency of decomposition, reducing noise interference, improving the accuracy of aging feature extraction in complex power scenarios, and supporting timely degradation evaluation.

[0162] In a specific implementation, the process of performing wavelet packet decomposition on the corrected leakage current signal to obtain a plurality of sub-band coefficients includes the following operations. First, a saturation recovery index is calculated based on the transient feature, and the saturation recovery index is compared with a preset recovery threshold to determine whether the corrected leakage current signal is in the saturation recovery period after the transient current impact, and a judgment result is generated.

[0163] Specifically, the processor of the edge computing unit extracts key parameters such as impact intensity or duration from the transient feature, calculates a comprehensive recovery index through an internal processing module, for example, generates a value reflecting the saturation depth in combination with the characteristic value. The index calculation can be achieved by preloading rules, for example, weighting the transient parameters to quantify the recovery demand. Subsequently, the index is compared with a preset recovery threshold, which is set based on experimental data, for example, a reference level is determined through historical impact test, if the index is higher than the threshold, it is judged as recovery period, and a binary result such as "yes" or "no" is output. The threshold is pre-stored in the firmware to support adjustment according to the scene.

[0164] In response to the judgment result indicating that it is in the saturation recovery period, a first wavelet mother function corresponding to the recovery period is selected from a preset wavelet base function library, otherwise a second wavelet mother function corresponding to the non-recovery period is selected, and the selected wavelet mother function is generated.

[0165] Specifically, the wavelet base function library is a pre-constructed set, for example, containing a plurality of mother function templates, each template is labeled as applicable to a specific state, and the processor queries the library as a key value according to the judgment result, for example, selecting a function template sensitive to high-frequency transient in the recovery period, otherwise selecting a template optimized for low-frequency steady state. The library is constructed through signal test in the design stage, for example, the performance of different functions on the saturation signal is evaluated and stored in categories. This selection ensures that the function matches the current signal characteristics.

[0166] A first decomposition level is set based on a target frequency band corresponding to the steady-state aging feature, and a second decomposition level is set based on a target frequency band corresponding to the transient aging feature, and an asymmetric wavelet packet decomposition structure is generated, wherein the first decomposition level is different from the second decomposition level.

[0167] Specifically, the processor sets a shallow level for a target frequency band of steady-state features, such as a low frequency range, for example, limits the decomposition depth to retain time domain details; for a high frequency target band of transient features, a deeper level is set to improve frequency resolution. This setting is achieved by adjusting the decomposition tree, for example, in a tree structure, low frequency branches are assigned few layers, and high frequency branches are assigned more layers, forming an asymmetric tree. This structure is generated based on preset rules, for example, according to the frequency band priority to dynamically allocate the number of layers.

[0168] The selected wavelet mother function and the asymmetric wavelet packet decomposition structure are applied to perform wavelet packet decomposition on the corrected leakage current signal to generate the plurality of sub-band coefficients.

[0169] Specifically, the processor loads the mother function as the filter base to recursively decompose the signal, for example, first applies low-pass and high-pass filtering to separate the first level sub-band, and then continues to subdivide the high frequency part more layers under the guidance of the asymmetric structure. This performs a step-by-step processing of the signal array, and finally outputs a coefficient set, each coefficient corresponding to a specific frequency-time interval.

[0170] For example, in a monitoring scenario where a high-voltage line arrester is struck by lightning, the saturation index is calculated to determine the recovery period, a high-frequency sensitive mother function is selected, a low-frequency shallow layer and a high-frequency deep layer structure is set, and the corrected signal is decomposed to generate coefficients for extracting post-lightning transient aging such as partial discharge characteristics, while retaining the steady-state degradation trend. This method can effectively separate noise and improve the timeliness of evaluation in a substation in a humid and thunderstorm environment.

[0171] In the above manner, the decomposition process is optimized, and the adaptability of the monitoring method is improved.

[0172] Optionally, in the process of monitoring the leakage current of the arrester, how to collect environmental data in real time and accurately associate it with aging characteristics to reflect the influence of temperature and humidity on aging indicators, thereby improving the accuracy and traceability of comprehensive evaluation.

[0173] The present embodiment collects the current values through temperature and humidity sensors, and generates associated data by matching the time stamp of the aging characteristics. The principle of this method is that synchronous acquisition captures the transient environment, and time stamp matching ensures data correspondence, thereby enhancing the fusion of features and environment, improving the context relevance of degradation analysis, and supporting reliable diagnosis in a variable power environment.

[0174] In a specific implementation, the process of collecting environmental data associated with the aging characteristics includes the following operations. First, the current temperature value of the tunnel magnetoresistance sensor is collected by a temperature sensor, and the current humidity value of the tunnel magnetoresistance sensor is collected by a humidity sensor.

[0175] In particular, the temperature sensor can be selected from a thermistor or an integrated digital component, such as installed near the tunnel magnetoresistance sensor module, sensing the ambient temperature through resistance variation or digital interface; the humidity sensor can be selected from a capacitive or resistive component, such as fixed inside the device housing, measuring the moisture content in the air. The acquisition process is parallel to the aging characteristic analysis, such as triggered by the edge computing unit, the processor sends a read instruction to the sensor, acquires the instantaneous value, and converts it into processable data through analog-to-digital conversion or direct digital output. This operation ensures that the value reflects the current working condition of the sensor, such as automatic acquisition every few seconds to match the real-time monitoring requirements.

[0176] Secondly, the current temperature value and the current humidity value are matched with the acquisition timestamp of the aging characteristic, generating environmental data associated with the aging characteristic.

[0177] In particular, the processor stores the timestamp of the aging characteristic in the memory, such as a numerical value representing the time of feature extraction, and then attaches the temperature and humidity values to the same or nearest time label, such as ensuring matching within a short interval by comparing the time difference. This matching can be achieved through data structures, such as creating a record entry containing the characteristic and environmental values, with the timestamp as the key value link. The generation process outputs the associated dataset, such as a package containing temperature, humidity, and corresponding characteristics, for uploading or local storage.

[0178] For example, in a substation arrester monitoring scenario, after collecting temperature and humidity values, match them with the timestamp of the aging characteristic to form an associated record like "Time X: Temperature Y Celsius, Humidity Z Percentage, Aging Characteristic W", which is used for cloud analysis of the impact of the environment on degradation.

[0179] In this way, the data acquisition process is optimized, and the integrity of the monitoring method is improved.

[0180] Based on the same inventive concept, the embodiments of the present application also provide a multi-sensor fusion arrester leakage current monitoring device corresponding to the multi-sensor fusion arrester leakage current monitoring method. Since the principle of solving problems in the device of the embodiments of the present application is similar to the above-mentioned multi-sensor fusion arrester leakage current monitoring method of the embodiments of the present application, the implementation of the device can be referred to the implementation of the method, and the repeated parts will not be described again.

[0181] Reference Figure 4 The multi-sensor fusion arrester leakage current monitoring device provided by the embodiments of the present application is shown in the following schematic diagram:

[0182] The acquisition module 10 is used to acquire the leakage current signal generated along the leakage path of the arrester to ground through the tunnel magnetoresistance sensor; and acquire the transient current signal flowing through the arrester body or the downlead through the auxiliary sensor;

[0183] The processing module 20 is configured to, in response to the at least one transient feature of the transient current signal satisfying a preset triggering condition, determine an artifact signal corresponding to the transient feature by using a preset distortion model, wherein the distortion model is used to characterize the distortion response of the tunnel magnetoresistance sensor under different transient features.

[0184] The analysis module 30 is configured to correct the leakage current signal based on the artifact signal to generate a corrected leakage current signal, and perform time domain and / or frequency domain feature analysis on the corrected leakage current signal to determine an aging feature and collect environmental data associated with the aging feature.

[0185] The cloud output module 40 is configured to analyze the aging feature and the environmental data by using a trend analysis model deployed in a cloud platform to generate monitoring result information.

[0186] Referring to Figure 5 A lightning arrester digital monitoring device installation schematic diagram provided by the embodiment of the present application includes: a tower-mounted installation plate 100, a tower angle steel 200, an antenna 300, M10 fixing bolts 400, a detection and early warning device 500, a lightning arrester insulating downlead 600, and an antenna connection 700. The tower-mounted installation plate 100 is used to fix the detection and early warning device 500 on the tower structure to provide a stable installation base. The tower angle steel 200 is used as a structural component of the tower to support the overall installation and provide mechanical strength. The antenna 300 is used for wireless communication transmission of monitoring data, such as 4G or 5G signal transmission. The M10 fixing bolts 400 are used to fasten the detection and early warning device 500 and the tower-mounted installation plate 100 to ensure firm installation. The detection and early warning device 500 is a core monitoring module, integrating sensors and a computing unit, and is used to collect and process leakage current signals. The lightning arrester insulating downlead 600 is connected to the lightning arrester body and is used to guide the leakage current and access the sensing ring of the monitoring device. The antenna connection 700 is the interface between the antenna 300 and the detection and early warning device 500, ensuring reliable signal connection, and can be fixed by copper wire binding to enhance stability. The schematic diagram shows the distributed deployment of the device on the tower, supporting non-contact monitoring and remote data transmission.

[0187] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or make equivalent replacements to some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for monitoring the leakage current of a surge arrester by means of multi-sensor fusion, characterized in that, include: The leakage current signal generated along the ground leakage path of the surge arrester is collected by the tunnel magnetoresistive sensor; The transient current signal flowing through the arrester body or the down conductor is collected by an auxiliary sensor; In response to at least one transient feature of the transient current signal satisfying a preset trigger condition, an artifact signal corresponding to the transient feature is determined using a preset distortion model; the distortion model is used to characterize the distortion response of the tunnel magnetoresistive sensor under different transient features; wherein, the distortion model is a distortion signature matrix, and the artifact signal includes multiple harmonic components of different frequencies; the distortion signature matrix characterizes the mapping relationship between the amplitude and rate of rise of the input current and the frequency components and amplitudes of the corresponding harmonic components; Based on the artifact signal, the leakage current signal is corrected to generate a corrected leakage current signal; time-domain and / or frequency-domain feature analysis is performed on the corrected leakage current signal to determine aging characteristics, and environmental data associated with the aging characteristics are collected; The aging characteristics and environmental data are analyzed using a trend analysis model deployed on a cloud platform to generate monitoring results.

2. The multi-sensor fused surge arrester leakage current monitoring method of claim 1, wherein, The distortion model is a distortion signature matrix embedded in the edge computing unit firmware; The edge computing unit is communicatively connected to the tunnel magnetoresistive sensor and the auxiliary sensor.

3. The multi-sensor fused surge arrester leakage current monitoring method of claim 2, wherein, The step of determining the artifact signal corresponding to the transient feature using a preset distortion model includes: Based on the transient feature, at least two discrete calibration points adjacent to the transient feature are found in the distortion signature matrix, and the reference artifact signals of the discrete calibration points are obtained respectively. Obtain the current operating temperature of the tunnel magnetoresistive sensor; Using the reference artifact signals of the at least two discrete calibration points as the interpolation reference, and the transient characteristics as the interpolation input, nonlinear interpolation is performed to generate the interpolated artifact signal. Based on the current operating temperature, the interpolated artifact signal is temperature compensated using a pre-established temperature compensation function to generate a temperature-compensated artifact signal.

4. The multi-sensor fused surge arrester leakage current monitoring method of claim 3, wherein, The generation of the corrected leakage current signal includes: The residual signal is obtained by subtracting the temperature-compensated artifact signal from the leakage current signal. Based on the transient characteristics, a time-varying response scaling factor is determined to characterize the recovery process of the tunnel magnetoresistive sensor from magnetic saturation after a transient current impact. The time-varying response scaling factor is determined by a pre-constructed dynamic response model, which is based on the magnetic saturation and recovery characteristics of the tunnel magnetoresistive sensor under transient current impacts of different amplitudes and rise rates. The residual signal is multiplied by the time-varying response scaling factor to perform dynamic amplitude compensation on the residual signal, thereby obtaining the corrected leakage current signal.

5. The multi-sensor fused surge arrester leakage current monitoring method of claim 4, wherein, The time-varying response scaling factor characterizing the recovery process of the tunnel magnetoresistive sensor from magnetic saturation after a transient current impact includes: Based on the transient characteristics and the current operating temperature, the artifact index corresponding to the transient characteristics is determined using the distortion signature matrix; inputting the artifact index into a pre-established saturation level classification table to determine a saturation level corresponding to the artifact index; selecting a recovery base function matching the saturation level from a pre-constructed recovery curve function library based on the saturation level; calculating a dynamic parameter for representing a time-domain morphology of the recovery base function according to at least one of a continuous value of the transient feature and the artifact index; determining the time-varying response scale factor based on the recovery base function and the dynamic parameter.

6. The multi-sensor fused surge arrester leakage current monitoring method of claim 5, wherein, The determining the artifact index corresponding to the transient feature by using the distortion signature matrix comprises: generating a temperature-compensated artifact signal by using the distortion signature matrix; performing Fourier transform on the temperature-compensated artifact signal to determine an amplitude of each harmonic component; determining a corresponding saturation contribution weight coefficient for each harmonic component according to a pre-set harmonic weight library; multiplying the amplitude of each harmonic component with the corresponding saturation contribution weight coefficient thereof respectively, and summing the multiplication results to determine the artifact index.

7. The multi-sensor fused surge arrester leakage current monitoring method of claim 1, wherein, The determining the aging feature comprises: performing wavelet packet decomposition on the corrected leakage current signal to obtain a plurality of sub-band coefficients, wherein each sub-band coefficient corresponds to a frequency segment and a time interval; extracting a first group of sub-band coefficients corresponding to a pre-set steady-state aging frequency segment from the plurality of sub-band coefficients, and calculating an energy value of the first group of sub-band coefficients to determine a steady-state aging feature representing a long-term aging state of the surge arrester; identifying a second group of sub-band coefficients belonging to a transient frequency segment from the plurality of sub-band coefficients, and determining whether an energy value of the second group of sub-band coefficients in a target time interval exceeds a set threshold to determine a transient aging feature representing a partial discharge or current impulse response; combining the steady-state aging feature and the transient aging feature to generate an aging feature representing a comprehensive aging state of the surge arrester.

8. The multi-sensor fused surge arrester leakage current monitoring method of claim 7, wherein, The performing wavelet packet decomposition on the corrected leakage current signal to obtain a plurality of sub-band coefficients comprises: calculating a saturation recovery index based on the transient feature, and comparing the saturation recovery index with a pre-set recovery threshold to determine whether the corrected leakage current signal is in a saturation recovery period after a transient current impulse, and generating a determination result; in response to the determination result indicating that it is in the saturation recovery period, querying and selecting a first wavelet mother function corresponding to the recovery period from a pre-set wavelet base function library, or selecting a second wavelet mother function corresponding to a non-recovery period, and generating the selected wavelet mother function; setting a first decomposition level based on a target frequency segment corresponding to the steady-state aging feature, and setting a second decomposition level based on a target frequency segment corresponding to the transient aging feature to generate an asymmetric wavelet packet decomposition structure, wherein the first decomposition level is different from the second decomposition level; applying the selected wavelet mother function and the asymmetric wavelet packet decomposition structure to perform wavelet packet decomposition on the corrected leakage current signal to generate the plurality of sub-band coefficients.

9. The multi-sensor fused surge arrester leakage current monitoring method of claim 1, wherein, The collecting environmental data associated with the aging feature comprises: collecting a current temperature value of the tunnel magnetoresistance sensor through a temperature sensor, and collecting a current humidity value of the tunnel magnetoresistance sensor through a humidity sensor; Matching the current temperature value and the current humidity value with a collection time stamp of the aging feature generates environmental data associated with the aging feature.

10. A multi-sensor fused surge arrester leakage current monitoring device, characterized by, Comprise: A collection module is configured to collect a leakage current signal generated along a leakage path of a lightning arrester to ground through a tunnel magnetoresistance sensor; Collect a transient current signal flowing through the lightning arrester body or the down conductor through an auxiliary sensor; A processing module is configured to determine an artifact signal corresponding to the transient feature by using a preset distortion model in response to the at least one transient feature of the transient current signal satisfying a preset trigger condition; the distortion model is used to characterize the distortion response of the tunnel magnetoresistance sensor under different transient features; wherein the distortion model is a distortion signature matrix, and the artifact signal includes a plurality of harmonic components of different frequencies; the distortion signature matrix characterizes the mapping relationship between the amplitude and the rise rate of the input current and the frequency component and the amplitude of the corresponding harmonic component; An analysis module is configured to correct the leakage current signal based on the artifact signal to generate a corrected leakage current signal; perform time domain and / or frequency domain feature analysis on the corrected leakage current signal to determine an aging feature, and collect environmental data associated with the aging feature; A cloud output module is configured to analyze the aging feature and the environmental data by using a trend analysis model deployed in a cloud platform to generate monitoring result information.

Images