A method for evaluating the leakage current spectrum deterioration of a surge arrester

Abstract

This invention discloses a method for assessing the spectral degradation of surge arresters' leakage current. It simultaneously acquires and preprocesses voltage and leakage current signals, constructs a dedicated harmonic coupling interference model based on the voltage signal, eliminates voltage-source interference components, and performs wideband, high-precision full-spectrum analysis on the de-interference current signal. Combining a condition-matched environmental interference feature library, it uses a hierarchical architecture to peel off three types of interference and extract pure degradation full-spectrum features. These features are then input into a nonlinear degradation correlation model with an adaptive feature weight allocation mechanism, outputting a continuously quantified degradation index, which is standardized and graded based on fixed interval thresholds. This method effectively isolates external interference, improves the accuracy and standardization of degradation assessment, and provides reliable support for condition-based maintenance of power grid equipment.

Description

Technical Field

[0001] This invention belongs to the field of power equipment condition monitoring and fault diagnosis technology, and relates to a method for assessing the degradation of the leakage current spectrum of surge arresters. Background Technology

[0002] Zinc oxide surge arresters are devices used for overvoltage protection in power systems. During long-term operation, these arresters can experience degradation issues such as aging of the valve plates, internal moisture absorption, and decreased insulation performance. This degradation directly reduces the safety margin of the power grid. Leakage current spectrum detection is currently the mainstream monitoring method for determining the operating status of surge arresters. This method relies on collecting the operating current signal of the equipment and combining it with spectrum analysis to extract harmonic characteristics, thereby reflecting the internal operating status of the surge arrester.

[0003] Current conventional assessment methods only collect a single signal of the surge arrester's leakage current, rely on Fast Fourier Transform (FFT) for spectrum analysis, and select single harmonic parameters as the basis for degradation judgment. Voltage harmonic components are prevalent in power grid operation. These harmonics can form coupling interference through the capacitive structure of the surge arrester, and this coupling interference can directly superimpose onto the leakage current. This results in the acquired harmonic characteristics being mixed with external voltage interference signals, failing to accurately reflect the characteristic changes caused by the degradation of the surge arrester itself.

[0004] Outdoor substations and complex power distribution scenarios are subject to various environmental factors. Environmental pollution, temperature and humidity changes, and electromagnetic coupling can all generate additional interference components in leakage current signals. Existing technologies lack targeted methods for distinguishing and separating environmental interference, making it impossible to differentiate between inherent equipment degradation signals and external environmental interference signals. This can easily lead to misjudgments or missed degradation detections.

[0005] Traditional spectrum analysis methods often employ fixed window functions and single-line correction algorithms, which are prone to spectral leakage and picket fence effects during the analysis process, making it difficult to accurately identify high-order weak harmonic components. The signal changes corresponding to early minor degradation of surge arresters are concentrated in the high-order harmonic range. Existing analysis methods cannot fully capture the characteristics of weak degradation, resulting in insufficient early degradation identification capability and a significant shortcoming in overall assessment sensitivity.

[0006] Conventional degradation assessment models use fixed parameter mapping relationships, which cannot adapt to the differences in signal response of surge arresters under different operating conditions, nor can they distinguish the spectral variation patterns corresponding to different degradation types. Unified judgment criteria are difficult to adapt to complex field operating conditions, further limiting the accuracy and stability of surge arrester degradation assessment results and failing to meet the actual needs of refined condition monitoring of power equipment. Summary of the Invention

[0007] To address the problems existing in the background technology, this invention proposes a method for evaluating the degradation of the leakage current spectrum of surge arresters.

[0008] The first aspect of this application provides a method for evaluating the spectral degradation of leakage current in surge arresters, including: The voltage and leakage current signals at the operating end of the surge arrester are acquired synchronously, and timing alignment and signal preprocessing are completed simultaneously. A dedicated harmonic coupling interference model is constructed based on the synchronous voltage signal to quantitatively solve the voltage homogeneous interference component in the leakage current. The voltage interference component is accurately removed from the original leakage current to obtain the voltage interference-free current signal, and wideband high-precision full spectrum analysis is performed on the voltage interference-free current signal. By combining the operating condition matching environmental interference feature library, environmental superimposed interference is adaptively stripped; pure degraded full-spectrum features unconstrained by external field interference are extracted. The pure degradation full-spectrum characteristics are input into the nonlinear degradation correlation model, and the quantitative degradation index is output and the level is classified.

[0009] Optionally, the signal preprocessing adopts a time-domain joint noise reduction architecture; synchronously couples baseline correction processing and timing error correction processing to eliminate the spectral aliasing problem caused by acquisition timing deviation.

[0010] Optionally, the harmonic coupling interference model adopts the capacitance differential response modeling method; it establishes a one-to-one correspondence mapping between voltage harmonics and interference current based solely on the inherent capacitive parameters of the surge arrester, thus isolating the influence of resistive parameter fluctuations on interference calculation.

[0011] Optionally, the wideband high-precision full spectrum analysis adopts a windowed multi-spectral line joint interpolation correction mode; it simultaneously corrects spectral leakage error and picket fence effect error, and fully preserves high-order weak degraded harmonic components.

[0012] Optionally, full spectrum analysis covers the interval from the fundamental wave to the eleventh consecutive odd harmonic; harmonic amplitude, harmonic phase, and harmonic distortion rate are collected simultaneously to construct a multi-dimensional spectral feature set.

[0013] Optionally, the working condition matching environmental interference feature library is built in multiple dimensions according to the degree of pollution, ambient temperature and humidity, and interphase electromagnetic coupling strength; and the corresponding benchmark interference sample is automatically matched according to the real-time working conditions on site.

[0014] Optionally, a hierarchical feature separation architecture can be used to distinguish interference types; and the superimposed signals from three different sources—voltage coupling interference, environmental condition interference, and random electromagnetic interference—can be separated in layers.

[0015] Optionally, the pure degradation full-spectrum characteristics retain only the inherent harmonic components generated by changes in the material properties of the surge arrester valve and internal moisture defects; and completely eliminate non-degradation harmonic signals introduced by external excitation.

[0016] Optionally, the nonlinear degradation correlation model adopts a feature weight adaptive allocation mechanism; different weights are assigned to different harmonic orders to match the signal response patterns of different degradation types of surge arresters.

[0017] Optionally, the degradation index is quantified using continuous numerical quantification; relying on fixed interval threshold boundaries, it achieves objective differentiation and standardized classification of degradation states. Compared with the prior art, the present invention has the following beneficial effects: First, by synchronously acquiring voltage and leakage current signals and constructing a dedicated harmonic coupling interference model, the voltage homogeneous interference components are accurately eliminated. Combined with a hierarchical feature separation architecture, environmental operating condition interference and random electromagnetic interference on site are stripped away, effectively isolating the influence of various external non-deterioration factors. This ensures that the extracted pure degradation full-spectrum features are specific and authentic, thereby improving the accuracy of degradation assessment from the root.

[0018] Secondly, the wideband high-precision full spectrum analysis covers the fundamental frequency to the eleventh consecutive odd harmonic range. Combined with the windowed multi-spectral line joint interpolation correction mode, it completely preserves the high-order weak degradation harmonic components, significantly enhances the ability to identify early degradation signals, solves the problem that traditional methods are difficult to capture early degradation characteristics, and provides early warning support for preventive maintenance of equipment.

[0019] Furthermore, the feature weight adaptive allocation mechanism of the nonlinear degradation correlation model can match the signal response patterns of different degradation types, continuously quantify the degradation index and standardize the grading of fixed interval thresholds, avoid the subjectivity and variability of human judgment, make the evaluation results uniform and comparable, and adapt to the surge arrester monitoring needs of different stations and different operating conditions.

[0020] Finally, the entire methodology is highly automated, requiring no manual intervention to complete the entire process from signal acquisition and processing to degradation classification. This simplifies the monitoring and evaluation process, reduces operation and maintenance costs, provides reliable data support for condition-based maintenance of power grid equipment, and helps improve the overall safety and stability of the power grid. Attached Figure Description

[0021] Figure 1 This is a flowchart of a method for evaluating the spectral degradation of surge arrester leakage current in one embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] In one embodiment, such as Figure 1 As shown, a method for evaluating the spectral degradation of leakage current in surge arresters is provided, and this method is applied to... Figure 1 Taking China as an example, the following specific steps will be used: S10: Synchronously acquire the voltage signal and leakage current signal at the operating end of the surge arrester, and simultaneously complete timing alignment and signal preprocessing.

[0024] Specifically, synchronous acquisition of the surge arrester's operating voltage and leakage current signals refers to using a unified sampling clock to drive the acquisition process of both types of signals, ensuring that the sampling times of the voltage and leakage current signals remain consistent. The entire acquisition process adopts a parallel acquisition mode, avoiding timing misalignment problems caused by time-division multiplexing, and can continuously acquire electrical data of the surge arrester at the same operating moment.

[0025] Synchronous timing alignment involves performing timing calibration on the raw acquired data to correct timing offsets caused by signal transmission delays and hardware sampling deviations. This process eliminates spectral aliasing issues caused by signal timeline misalignment, ensuring accurate correspondences of subsequent harmonic components and providing a reliable data foundation for subsequent interference component calculations and feature analysis.

[0026] Signal preprocessing comprises two categories. The first is joint noise reduction, which uses an improved wavelet threshold noise reduction method to filter high-frequency random noise and impulse interference within the signal, eliminating cluttered signals generated by the ambient electromagnetic environment. The second is baseline correction, which uses moving average filtering to smoothly correct signal baseline drift and suppress baseline shifts caused by temperature changes and fluctuations in hardware operating conditions.

[0027] The synergistic setting of timing alignment combined with dual-layer preprocessing can fully preserve the effective fundamental and harmonic components in both voltage and leakage current signals. This processing method can stably reduce invalid interference introduced by the complex acquisition environment, improve the purity of the original input signal, prevent invalid clutter signals from masking the weak electrical characteristics corresponding to the early degradation of the surge arrester, and ensure the orderly conduct of subsequent full-spectrum analysis and interference separation processes.

[0028] In one embodiment, in step S10, the signal preprocessing adopts a time-domain joint noise reduction architecture; the baseline correction processing and timing error correction processing are synchronously coupled to eliminate the spectral aliasing problem caused by the acquisition timing deviation.

[0029] Specifically, the time-domain joint noise reduction architecture used in signal preprocessing integrates two types of noise reduction methods into a collaborative working mode, simultaneously covering the dual suppression requirements of high-frequency noise and low-frequency drift. This architecture takes the original acquired signal as the processing object, first scanning the signal point by point through an improved wavelet threshold noise reduction algorithm to identify and filter high-frequency random noise and electromagnetic pulse interference. These interferences mostly come from external environments such as the start-up and shutdown of electrical equipment and electromagnetic radiation, and manifest as spike pulses and irregular fluctuations in the signal.

[0030] Simultaneously with noise reduction processing, baseline correction is applied. A moving average filtering algorithm dynamically adjusts the signal baseline, correcting baseline shifts caused by factors such as temperature changes and sensor drift due to prolonged operation. Baseline shifts cause the signal to fluctuate, masking true harmonic characteristics. Dynamic correction ensures the signal remains at a stable reference level, guaranteeing the accuracy of subsequent spectrum analysis.

[0031] Synchronous linkage with timing error correction processing is a key design feature of this architecture. Timing errors, stemming from sensor signal transmission delays and minor sampling clock deviations, can cause misalignments in the acquisition times of voltage and leakage current signals, leading to spectral aliasing and making it difficult to distinguish between harmonic components of different frequencies. Timing error correction establishes a precise timing correspondence by comparing the timestamp information of the two types of signals, calibrating the offset signal data frame by frame to ensure that each set of voltage and current data strictly corresponds to the same operating time.

[0032] For example, in outdoor substation scenarios, the dense equipment leads to a complex electromagnetic environment, making it easy for high-frequency noise to mix into the signals collected by sensors. At the same time, large temperature differences between day and night can cause baseline drift, and even slight delays in the sampling system can cause timing deviations. By adopting a joint time-domain noise reduction architecture, all three types of problems can be solved simultaneously in a single processing flow, eliminating the need for multiple steps.

[0033] This synchronous coupling processing method avoids the limitations of single processing methods, forming a closed loop for signal preprocessing. It effectively removes various interference impurities and corrects signal timing and baseline deviations, ensuring the purity and synchronization of voltage and leakage current signals from the source. It effectively eliminates spectral aliasing problems caused by acquisition timing deviations, providing a high-quality data foundation for subsequent harmonic analysis, interference model construction, and other steps, ensuring the reliability of subsequent calculation results.

[0034] S20: Based on the synchronous voltage signal, a dedicated harmonic coupling interference model is constructed to quantitatively solve the voltage homogeneous interference component in the leakage current.

[0035] Specifically, constructing a dedicated harmonic coupling interference model based on the synchronous voltage signal involves using the synchronously acquired voltage signal as the basic data source after time alignment and preprocessing, to conduct a complete analysis of harmonic information across the entire domain. First, spectral decomposition is performed on the clean and stable voltage signal to fully extract the fundamental operating parameters of the power grid's fundamental wave and all odd harmonics, thus obtaining a complete understanding of the harmonic composition rules of the operating voltage.

[0036] The dedicated harmonic coupling interference model uses the inherent capacitive electrical characteristics of the surge arrester as the modeling basis and establishes corresponding correlations by combining the electrical response laws of the capacitive load of power equipment. During the model construction process, only the inherent parameters corresponding to the equivalent capacitance of the surge arrester are fixed, and the dynamic changes of the resistive parameters of the valve plate are not introduced, which can avoid the adverse effects of the surge arrester's own resistive fluctuations on the interference calculation process.

[0037] In the specific implementation process, based on the variation law of each voltage harmonic, and combined with the electrical parameters corresponding to the harmonic order and the fundamental operating frequency of the power grid, the coupling current components generated in the surge arrester circuit when each type of voltage harmonic acts alone are derived one by one. The coupling current contents corresponding to all harmonic orders are integrated to form a complete and continuous set of voltage homogeneous interference, realizing the overall modeling of the interference components.

[0038] The voltage-source interference components in the leakage current are quantitatively calculated. A comprehensive calculation is performed based on a well-established harmonic coupling interference model, and the actual proportion of each interference component is accurately calculated according to electrical conduction laws. This calculation process follows a unified electrical conversion logic, ensuring a consistent standard for the calculation of various harmonic interference components, and achieving quantitative differentiation and precise location of interference components.

[0039] The coordinated modeling and calculation described above can accurately distinguish interference signals induced by voltage within the leakage current. It can isolate the coupling effects of grid voltage harmonics from mixed current data, preventing the superposition and confusion between voltage harmonic characteristics and arrester degradation characteristics. Simultaneously, it ensures the independence and stability of the interference component calculation process, providing accurate and reliable calculation basis for subsequent removal of voltage-related interference components, effectively avoiding evaluation biases caused by grid voltage fluctuations.

[0040] In one embodiment, in step S20, the harmonic coupling interference model adopts the capacitance differential response modeling method; a one-to-one correspondence mapping between voltage harmonics and interference current is established solely based on the inherent capacitive parameters of the surge arrester, thus isolating the influence of resistive parameter fluctuations on interference calculation.

[0041] Specifically, the harmonic coupling interference model uses a capacitance differential response modeling method, which is based on the capacitive electrical characteristics of the surge arrester to construct the associated logic. During the operation of the surge arrester, its equivalent capacitance is an inherent electrical parameter, which is minimally affected by the external environment and operating conditions. The current response of the capacitive load has a fixed differential relationship with the rate of voltage change, and this physical characteristic becomes the basis for modeling.

[0042] During the modeling process, only the inherent capacitive parameters of the surge arrester are extracted as calculation variables, without introducing the resistive parameters of the valve plate. The resistive parameters of the valve plate will dynamically change with factors such as the degree of deterioration of the surge arrester and fluctuations in operating voltage. If these parameters are included in the modeling, the interference calculation results will be affected by the changes in the state of the equipment itself, making it impossible to accurately separate voltage-source interference.

[0043] By utilizing the differential response relationship of capacitance, a one-to-one correspondence between voltage harmonics and interference currents is established. Specifically, the amplitude and frequency parameters of each voltage harmonic are used to obtain the amplitude and phase characteristics of the corresponding coupled interference current through differential calculations of the inherent capacitive parameters. This mapping relationship completely follows the electrical response law of capacitive loads, ensuring that the calculation process of interference current is only related to voltage harmonics and inherent capacitive parameters.

[0044] For example, in the scenario of surge arrester monitoring in a 110 kV substation, the grid voltage contains the fundamental frequency and multiple odd harmonics. These voltage harmonics will generate coupling currents through the equivalent capacitance of the surge arrester. By adopting the capacitance differential response modeling method, the interference current corresponding to each voltage harmonic can be accurately calculated based on the inherent capacitive parameters of the surge arrester, without being affected by fluctuations in the resistive parameters of the valve plate due to aging.

[0045] This modeling method fundamentally isolates the interference calculation from resistive parameter fluctuations, ensuring that the solution results for voltage-source interference components reflect only the coupling effects of grid voltage harmonics, independent of the arrester's own degradation state. This precise mapping relationship ensures the targetedness and accuracy of subsequent interference removal, avoiding ineffective interference separation due to parameter confounding, and laying a solid foundation for extracting pure degradation characteristics.

[0046] In one embodiment, in step S30, the wideband high-precision full spectrum analysis adopts a windowed multi-spectral line joint interpolation correction mode; the spectral leakage error and the picket fence effect error are corrected simultaneously, and the high-order weak deteriorated harmonic components are completely preserved.

[0047] Specifically, the wideband high-precision full-spectrum analysis employs a windowed multi-spectral line joint interpolation correction mode, which is a synergistic processing scheme integrating window function suppression and multi-spectral line interpolation correction. The wideband analysis is set to cover the analysis range from the fundamental frequency to the eleventh odd harmonic, ensuring that it can capture the high-order weak harmonic signals that may be generated by the early degradation of the surge arrester, and avoiding the omission of features due to the narrow bandwidth.

[0048] In the windowing process, a dedicated window function is used to preprocess the voltage interference removal current signal. The window function can smooth out abrupt changes at both ends of the signal and constrain the spread of side lobes during spectral analysis, thereby suppressing spectral leakage errors. Spectral leakage can cause harmonic components of different frequencies to overlap, blurring the true spectral characteristics and affecting the identification of higher-order weak harmonics.

[0049] Multi-spectral line joint interpolation correction addresses the picket fence effect error. The picket fence effect, stemming from the inherent characteristics of discrete sampling, prevents the accurate capture of harmonic components at certain frequencies, much like the blind spots in viewing an object through a fence. This correction method selects the main harmonic peak and its adjacent spectral lines as the analysis object, performing interpolation estimation through joint calculations of multi-spectral line data to restore the true harmonic parameters obscured by the picket fence effect.

[0050] For example, in monitoring the early, slight aging of surge arresters, the amplitude of high-order degraded harmonic signals in the leakage current is weak, and conventional spectrum analysis is easily affected by leakage errors and the picket fence effect, making them undetectable. By employing a windowed multi-spectral line joint interpolation correction mode, the window function suppresses spectral leakage, and multi-spectral line interpolation restores the missed weak harmonic information, enabling the complete capture of these key characteristics.

[0051] This simultaneous correction of two types of errors effectively solves the accuracy bottleneck of traditional spectrum analysis. It ensures wide bandwidth coverage across the entire spectrum while significantly improving the calculation accuracy of harmonic parameters, and can completely preserve high-order weak deterioration harmonic components. These weak components are directly related to the early deterioration state of the surge arrester; their complete extraction provides crucial data support for subsequent identification of pure deterioration features and assessment of deterioration severity, greatly improving the reliability of early deterioration identification.

[0052] S30: Accurately remove voltage interference components from the original leakage current to obtain the voltage interference-free current signal, and perform wideband high-precision full-spectrum analysis on the voltage interference-free current signal.

[0053] Specifically, after quantitatively calculating the voltage-source interference components, all calculated interference components are treated as a unified subtraction target and applied to the preprocessed original leakage current signal. The original leakage current signal simultaneously includes the arrester's operating current, the interference current generated by voltage coupling, and stray currents superimposed by the environment. Through the overall component cancellation processing method, the coupled interference content corresponding to each voltage harmonic is successively subtracted, achieving targeted elimination of interference components.

[0054] This elimination process maintains the overall timing structure of the current signal without alteration, removing only the invalid components derived from external voltage excitation while fully preserving the effective current components formed by the arrester's own electrical response. After targeted elimination, the original current signal, previously affected by mixed voltage harmonics, is purified, forming a structurally complete and continuous voltage-interference-free current signal. Removing coupling interference from external voltage reduces the superposition and doping of irrelevant electrical components, ensuring that the current signal's variation pattern closely matches the arrester's operating state.

[0055] After obtaining the voltage interference removal current signal, a wideband, high-precision full-spectrum analysis was immediately performed. The analysis process selected a reasonable signal analysis frequency band, fully covering the fundamental frequency band and higher odd harmonic frequency bands of the power grid, ensuring that the analysis range could accommodate all frequency characteristics generated during the arrester degradation process. The analysis operation employed a windowed multi-spectral-line joint interpolation correction processing mode, relying on a dedicated window function to constrain the spread range of the signal spectrum sidelobes and suppress spectral leakage problems generated during signal processing.

[0056] The correction method, combined with multi-spectral line joint interpolation, can correct the picket fence effect caused by discrete sampling calculations and compensate for the calculation bias in single-point spectral line analysis. The analysis phase simultaneously collects amplitude parameters, phase parameters, and harmonic distortion rate parameters corresponding to different frequency components, systematically constructing a multi-dimensional spectral data system.

[0057] This comprehensive analytical processing method enhances the ability to identify weak harmonic signals and stably restores the subtle frequency variation characteristics within current signals. The wide bandwidth analysis range avoids the omission of high-frequency degradation features, and the high-precision correction methods improve the accuracy of the spectral data, providing complete and accurate raw spectral data for subsequent environmental interference removal and pure degradation feature extraction.

[0058] In one embodiment, in step S30, the full spectrum analysis covers the interval from the fundamental wave to the eleventh consecutive odd harmonic; the three types of common-source spectrum parameters, namely harmonic amplitude, harmonic phase, and harmonic distortion rate, are collected simultaneously to construct a multi-dimensional spectrum feature set.

[0059] Specifically, the full spectrum analysis clearly covers the fundamental frequency to the eleventh consecutive odd harmonic range. The selection of this range is based on the frequency distribution pattern of arrester degradation signals. The characteristic signals generated by the degradation process of arrester varistors, such as aging and internal moisture absorption, are mainly concentrated in the fundamental frequency and the third, fifth, seventh, ninth, and eleventh odd harmonic frequency bands. Continuously covering this range ensures that no degradation-related frequency components are missed, avoiding the loss of key features due to frequency band breaks.

[0060] During the analysis process, three types of spectral parameters from the same source are collected simultaneously: harmonic amplitude, harmonic phase, and harmonic distortion rate. Harmonic amplitude reflects the intensity of the harmonic signal at the corresponding frequency and is directly related to the degree of degradation; harmonic phase reflects the temporal position relationship of the harmonic signal, and different types of degradation will cause specific phase shifts; harmonic distortion rate characterizes the degree of deviation of each harmonic from the fundamental frequency and can comprehensively reflect the overall distortion state of the signal.

[0061] The acquisition of the three types of parameters is based on the same set of voltage interference-free current signals, ensuring that the parameters are homogeneous and correlated, and avoiding parameter misalignment caused by different data sources. All acquired parameters are integrated in frequency order to construct a multi-dimensional spectral feature set. This set contains the three types of parameters corresponding to the fundamental wave to the eleventh odd harmonic, comprehensively covering key information in both the frequency and parameter dimensions.

[0062] For example, when monitoring a surge arrester that has been in operation for a long time, its internal varistors showed slight aging. After full-spectrum analysis covering the fundamental frequency to the eleventh odd harmonic range, a slight increase in the amplitude of the fifth harmonic was successfully captured. At the same time, the phase shift of the seventh harmonic and the change in the overall harmonic distortion rate were also detected. The synergistic feedback of these three types of parameters accurately outlined the degradation characteristics. If only a single parameter or a part of the harmonic range were collected, it would be impossible to fully identify this early aging signal.

[0063] This combination of continuous interval coverage and synchronous acquisition of multi-dimensional parameters breaks through the limitations of traditional single-parameter or limited harmonic interval analysis. The multi-dimensional spectral feature set can comprehensively and three-dimensionally characterize the electrical signal changes of the surge arrester from multiple dimensions such as intensity, phase, and distortion degree. This provides a rich and comprehensive data foundation for subsequent environmental interference stripping and pure degradation feature extraction, making the characterization of degradation state more accurate and comprehensive, and effectively improving the reliability and accuracy of subsequent degradation assessments.

[0064] S40: Combines the operating condition matching environmental interference feature library to adaptively remove superimposed environmental interference; extracts pure degraded full-spectrum features that are not constrained by external field interference.

[0065] Specifically, after completing wideband high-precision full-spectrum analysis, a complete set of current spectrum features is generated. This set still contains environmental superimposed interference caused by the on-site operating conditions. This invention pre-constructs an operating condition-matched environmental interference feature library. The feature library is classified and stored in multiple dimensions according to the on-site operating conditions. The classification dimensions include the degree of surface contamination, environmental temperature and humidity conditions, and the electromagnetic coupling strength between equipment phases.

[0066] The feature library contains standard spectrum data of surge arresters under normal operating conditions. Various data types correspond to different single and combined operating conditions, covering all common operating environments in power plants. During equipment operation in the field, real-time environmental operating condition information is collected, and the database content is automatically matched based on the real-time operating conditions to retrieve reference interference spectrum data that closely matches the field environment.

[0067] Based on the benchmark data obtained through matching, a hierarchical feature separation architecture is initiated to perform adaptive stripping of environmental superimposed interference. The separation process distinguishes different types of interference, separately classifying interphase electromagnetic interference, temperature and humidity-induced interference, and signal interference caused by dirt adhesion, and completes hierarchical removal according to the spectral distribution patterns of different interferences.

[0068] The separation process employs self-adaptive computational logic, avoiding a fixed and uniform subtraction standard. It can dynamically adjust the interference stripping intensity according to changes in on-site conditions, adapting to complex and ever-changing outdoor operating scenarios. After gradually stripping away all environmental superimposed interference components, the remaining spectral components no longer contain signal fluctuations induced by the external environment.

[0069] The final filtered and retained spectrum content is generated solely from changes in the internal state of the surge arrester itself, completely detached from the constraints of various external interference conditions, thus forming a pure degradation full-spectrum characteristic. The pure degradation full-spectrum characteristic centrally reflects the changes in the performance of the surge arrester varistor material, internal moisture defects, and electrical changes caused by the attenuation of the main body insulation, effectively isolating spectrum fluctuations caused by external factors other than the equipment itself.

[0070] This processing method effectively delineates the boundary between environmental interference characteristics and equipment degradation characteristics, preventing changes in environmental conditions from masking the true degradation signals of the surge arrester. Stable elimination of dynamically changing environmental superimposed interference ensures the specificity and authenticity of the extracted spectral features, effectively avoiding assessment biases caused by environmental factors and providing pure and reliable feature data for subsequent quantitative analysis of degradation levels.

[0071] In one embodiment, a hierarchical feature separation architecture is used to distinguish interference types; the superimposed signals from three different sources—voltage coupling interference, environmental condition interference, and random electromagnetic interference—are separated in layers.

[0072] Specifically, the hierarchical feature separation architecture is based on the source and spectral characteristics of the interference signal. It distinguishes different types of interference through hierarchical processing logic, ensuring that each level only performs stripping operations on specific interference. This architecture clearly classifies the interference signals superimposed on the leakage current into three categories: voltage-coupled interference, environmental condition interference, and random electromagnetic interference in the field. Each type of interference corresponds to an independent separation level and a dedicated processing strategy.

[0073] The first separation layer targets voltage-coupled interference. Based on a previously established dedicated harmonic-coupled interference model, and according to the mapping relationship between voltage harmonics and interference current, it selectively extracts and removes the coupling components generated by grid voltage harmonics from the leakage current. The spectral characteristics of voltage-coupled interference are consistent with those of grid voltage harmonics, exhibiting a clear frequency correspondence. This characteristic provides the foundation for accurate separation, ensuring that this layer removes only voltage-derived interference without affecting other signal components.

[0074] The second separation level focuses on environmental operating condition interference. Using a baseline interference sample obtained from an environmental interference feature library matched to operating conditions as a reference, it extracts and removes interference components caused by environmental factors by comparing the differences between the current spectral characteristics and the baseline sample. Environmental operating condition interference is directly related to the degree of pollution, temperature, humidity, and interphase coupling strength at the site. Its spectral characteristics fluctuate regularly with changes in operating conditions, allowing for targeted separation based on the baseline sample.

[0075] The third separation layer is specifically designed to handle random electromagnetic interference in the field. This type of interference originates from uncertain factors such as the start-up and shutdown of field equipment and electromagnetic radiation, and its spectral distribution is irregular and its duration is short. This layer employs a dynamic threshold recognition algorithm to monitor the instantaneous fluctuations of the spectral signal in real time, identify and capture the pulse characteristics and irregular spectral components of random electromagnetic interference, and complete interference stripping through signal smoothing processing to prevent it from masking the degradation characteristics.

[0076] For example, in urban substations, surge arresters operate in complex environments. Voltage harmonics from the power grid generate voltage coupling interference, construction work in the surrounding area causes environmental pollution and environmental interference, and radiation from nearby communication equipment creates random electromagnetic interference. By adopting a hierarchical feature separation architecture, voltage coupling interference is first stripped at the first level, then environmental interference is removed at the second level using matched reference samples, and finally, random electromagnetic interference is eliminated at the third level, gradually purifying the signal. If a hybrid separation method is used without distinguishing between interference types, it can easily lead to ineffective interference stripping or the accidental deletion of degraded features.

[0077] This layered processing model clearly defines the separation order and operational logic for various types of interference, avoiding mutual interference between different types of interference and ensuring that each type of interference can be accurately identified and effectively removed. The layered architecture makes the interference separation process more systematic and targeted, capable of comprehensively eliminating superimposed signals from three different sources, preserving the arrester's own degradation characteristics to the greatest extent, providing a clean signal foundation for subsequent pure degradation full-spectrum feature extraction, and improving the accuracy of degradation assessment.

[0078] In one embodiment, in step S40, the working condition matching environmental interference feature library is constructed by dividing the library into multiple dimensions according to the degree of pollution, ambient temperature and humidity, and interphase electromagnetic coupling strength; and the corresponding benchmark interference sample is automatically matched according to the real-time working conditions on site.

[0079] Specifically, the construction of the operating condition matching environmental interference feature library is based on the environmental factors that affect the leakage current of surge arresters. Specifically, it is divided into multiple dimensions according to three key dimensions: pollution level, ambient temperature and humidity, and phase-to-phase electromagnetic coupling strength.

[0080] The contamination level dimension is divided into different grade ranges based on the common on-site contamination conditions of power equipment, with each range corresponding to a specific surface contamination adhesion state; the environmental temperature and humidity dimension covers the temperature and humidity ranges that the equipment may encounter during operation, and is further subdivided into subdivided areas according to the combination of temperature and humidity; the interphase electromagnetic coupling strength dimension is divided into different strength level areas based on the coupling differences caused by factors such as equipment installation spacing and layout.

[0081] During the database construction process, for each multi-dimensional cross-region, normal surge arresters with good performance were selected for on-site testing. Leakage current spectrum data under different operating conditions were collected. These data only contain characteristics generated by environmental interference and do not involve any degradation-related signals. They are stored in the database as the benchmark interference samples corresponding to that region. The benchmark interference samples of all regions together constitute a complete operating condition-matched environmental interference feature database, ensuring that the data in the database can comprehensively cover various common operating conditions of power plants.

[0082] In practical applications, sensors deployed on-site collect real-time data on the pollution level, ambient temperature, ambient humidity, and phase-to-phase electromagnetic coupling strength of the surge arrester's operating environment, forming a real-time operating condition information set. This real-time operating condition information is then compared with the partitioning standards of a feature library. The system automatically retrieves and matches benchmark interference samples that perfectly match the real-time on-site operating conditions, achieving accurate matching without manual intervention.

[0083] For example, in substations located in humid coastal areas, surge arresters operate in environments with high humidity and a certain degree of salt spray pollution. Simultaneously, the dense arrangement of equipment results in strong inter-phase electromagnetic coupling. By collecting these environmental parameters in real time, a condition-matched environmental interference feature library can automatically match benchmark interference samples to corresponding coupling strength zones within high humidity and high pollution environments, providing a precise basis for subsequent environmental interference stripping. Using a unified benchmark library instead of a multi-dimensional partitioned library would fail to adapt to the interference characteristics under such specific operating conditions, leading to ineffective interference stripping.

[0084] This multi-dimensional partitioned database construction and real-time automatic matching design gives the environmental interference feature library a strong ability to adapt to different operating conditions. It can accurately locate the interference features corresponding to the real-time environment on site, avoiding the problem of mismatch between benchmark interference samples due to differences in operating conditions. This ensures that the benchmark data used for subsequent environmental interference stripping is targeted and accurate, laying a reliable foundation for extracting pure degradation full-spectrum features.

[0085] S50: Input the pure degradation full-spectrum characteristics into the nonlinear degradation correlation model, output the quantitative degradation index and complete the level classification.

[0086] Specifically, after completing the full extraction of pure degradation full-spectrum features, the processed feature data is uniformly imported into the nonlinear degradation correlation model. The nonlinear degradation correlation model is built based on multiple types of actual operating samples for preliminary training. The training samples include measured spectrum data of surge arresters with various degradation levels, covering different aging stages and internal defect types of the equipment.

[0087] The model incorporates an adaptive feature weight allocation mechanism to identify response differences in the spectrum of different harmonic orders. By combining the signal performance patterns of various internal degradations of the surge arrester, it autonomously assigns corresponding weights to each full-spectrum feature, differentiating the representational capabilities of different harmonic components for degradation states, and abandoning the single judgment mode with fixed weights.

[0088] After receiving the pure degradation full-spectrum features, the model performs global calculations based on its internally preset nonlinear mapping relationship. Combining the comprehensive change patterns of multi-dimensional spectral features, it integrates the weight contribution of each feature and generates a continuous quantitative degradation index. The quantitative degradation index can intuitively reflect the overall development degree of internal degradation of the surge arrester and fully reflect the comprehensive performance changes caused by valve aging or internal moisture.

[0089] Based on the obtained quantitative degradation indicators, standardized grading is carried out using pre-defined fixed numerical ranges. Different numerical ranges correspond to independent equipment operating states. The state is defined strictly according to the range to which the indicator belongs, and the equipment is distinguished in the following order: normal operating state, slight degradation state, obvious degradation state, and severe degradation state.

[0090] Nonlinear degradation correlation models can accurately reflect the nonlinear evolution of surge arrester degradation and adapt to the spectral variation characteristics corresponding to different defect types. Adaptive weight allocation enhances the model's ability to identify weak degradation signals and strengthens the adaptability of the assessment system. Continuous quantitative degradation indicators combined with standardized grading can unify the judgment criteria for equipment in different sites, eliminate human error in judgment, and ensure the objectivity and consistency of surge arrester degradation assessment results, accurately supporting the orderly implementation of condition-based maintenance of power equipment.

[0091] In one embodiment, the pure degraded full-spectrum feature retains only the inherent harmonic components generated by changes in the material properties of the surge arrester valve and internal moisture defects; it completely eliminates non-degraded harmonic signals introduced by external excitation.

[0092] Specifically, the full-spectrum characteristic of pure degradation is that after multiple layers of interference stripping, only harmonic components directly related to the internal state of the surge arrester remain. The root cause of these harmonic components is the change in the material properties of the surge arrester varistors and internal moisture defects. They are inherent electrical responses caused by the degradation of the equipment itself and are unrelated to the external environment and power grid excitation.

[0093] The varistor is a component of a surge arrester. During long-term operation, it undergoes material aging due to electrical and thermal stress, leading to changes in its nonlinear conductivity. This change in characteristics introduces specific frequency harmonic components into the leakage current, and the parameter changes of these components only with the degree of material aging, unaffected by external factors. Internal moisture defects can damage the internal insulation structure of the varistor, altering the current conduction path and also generating specific harmonic signals in the leakage current. These signals directly reflect the severity of the moisture ingress.

[0094] The extraction process of pure degradation full-spectrum characteristics is an effective way to eliminate non-degradation harmonic signals introduced by various external stimuli. Non-degradation harmonic signals include coupled interference signals generated by grid voltage harmonics, environmental interference signals caused by environmental pollution and temperature and humidity changes, and random interference signals caused by on-site electromagnetic radiation. These signals are all caused by external factors and are unrelated to the degradation state of the surge arrester itself, and must be completely removed through targeted processing.

[0095] For example, in mountainous substations, surge arresters may simultaneously face grid voltage harmonic interference, high humidity, and electromagnetic radiation interference from nearby transmission lines. After processing using a layered feature separation architecture, voltage-coupled interference from environmental conditions and random electromagnetic interference from the field are stripped away one by one. The remaining harmonic components are only generated by valve plate aging and internal moisture. If these external interference signals are not completely eliminated, the residual non-deterioration harmonics will superimpose with the deterioration features, leading to misjudgment of the equipment's true condition.

[0096] The formation of pure degradation full-spectrum characteristics effectively isolates external interference from internal degradation signals. It contains only inherent harmonic components caused by equipment degradation, thus accurately reflecting the internal state of the surge arrester. This pure characteristic form allows subsequent degradation correlation models to precisely capture the changing patterns of degradation signals, avoiding interference from irrelevant external signals. This provides the most direct and reliable data support for quantitative assessment and classification of degradation levels, ensuring the authenticity and accuracy of the assessment results.

[0097] In one embodiment, the nonlinear degradation correlation model employs a feature weight adaptive allocation mechanism; different weights are assigned to different harmonic orders to match the signal response patterns of different degradation types of surge arresters.

[0098] Specifically, the nonlinear degradation correlation model is designed with an adaptive feature weight allocation mechanism. This mechanism abandons the traditional single judgment logic of fixed weights and can dynamically adjust the weight ratio of different harmonics based on the signal performance of each harmonic in the full spectrum characteristics of pure degradation. The weight allocation is based on the strength of the correlation between each harmonic and different degradation types of the surge arrester. That is, harmonic components that are more sensitive to the characterization of the degradation state will be given higher weights, and vice versa, ensuring that the weight allocation is accurately matched with the signal response law.

[0099] The harmonic responses of surge arresters exhibit significant differences depending on the type of degradation. Aging of the valence material primarily causes changes in the parameters of low-order odd harmonics, while internal moisture defects are more likely to lead to significant fluctuations in high-order odd harmonics. The adaptive feature weight allocation mechanism, trained on a large number of samples, has pre-learned and stored the harmonic response patterns corresponding to different degradation types. It can identify the changing trends of each harmonic in the full-spectrum characteristics of current pure degradation, determine which degradation type's dominant feature it belongs to, and then assign differentiated weights accordingly.

[0100] For example, when the amplitude changes of the third and fifth harmonics are most significant in the full spectrum characteristics of pure degradation, the model will determine that the current degradation is likely mainly due to valve plate aging, and then increase the weight of the third and fifth harmonics to allow these key features to play a dominant role in the degradation assessment. If the seventh, ninth, eleventh and other higher harmonics show specific shifts, the model will identify it as degradation dominated by internal moisture, and correspondingly increase the weight of higher harmonics to ensure that the assessment results are consistent with the actual degradation type.

[0101] This differentiated weight allocation model allows the nonlinear degradation correlation model to accurately adapt to the signal response patterns of different degradation types, avoiding the neglect of some degradation features due to fixed weights. The model's sensitivity to various degradation signals is significantly improved, enabling it to capture weak but crucial changes in degradation characteristics, making the quantitative calculation of degradation degree more targeted and accurate. Simultaneously, the adaptive mechanism allows the model to adapt to degradation assessment needs under different operating conditions without manual parameter adjustment, significantly improving the model's practicality and adaptability, and providing a precise computational foundation for the subsequent output of quantitative degradation indicators.

[0102] In one embodiment, the degradation index is quantified using a continuous numerical quantification method; relying on fixed interval threshold boundaries, the objective distinction and standardized classification of degradation states are achieved.

[0103] Specifically, the quantitative degradation index adopts a continuous numerical quantification form, which can fully present the gradual change process of the arrester's degradation degree, rather than simply dividing it into discrete states. The changes in the numerical value are directly related to the parameter fluctuations of the pure degradation full-spectrum characteristics. Even slight changes in parameters such as the amplitude and phase harmonic distortion rate of each harmonic in the pure degradation characteristics will be reflected in the numerical changes of the quantitative degradation index, achieving a fine characterization of the degradation process.

[0104] This continuous numerical form breaks through the limitations of traditional discrete level determination. It can capture the entire process of changes in the surge arrester from normal state to slight degradation, obvious degradation and then to severe degradation. Whether it is the weak signal of early slight degradation or the significant change of later degradation, it can be accurately reflected by the continuous numerical value of the quantified degradation index, making the description of the degree of degradation more objective and detailed.

[0105] After obtaining the quantitative degradation indicators, standardized classification judgments are carried out based on pre-set fixed interval threshold boundaries. The fixed interval threshold boundaries are formulated based on a large amount of surge arrester degradation test data and actual operational failure cases, dividing the equipment into four independent numerical intervals, corresponding to the normal operating state, slight degradation state, obvious degradation state, and severe degradation state, respectively.

[0106] Once the threshold boundary value is set, it remains fixed and will not be arbitrarily adjusted due to changes in on-site operating conditions or equipment type, ensuring that the classification judgment of all surge arresters follows a unified standard. During the judgment process, the system automatically compares the calculated quantitative degradation index with the fixed interval threshold boundary, and directly outputs the corresponding degradation level according to the value range to which the index belongs, without the need for manual intervention.

[0107] For example, multiple surge arresters of different models in a regional power grid were tested using this method and obtained different continuous quantitative degradation indicators. Surge arresters with indicators falling within the lowest value range were judged to be in normal operating condition and required no special treatment; those with indicators falling within the slightly deteriorated range were marked as early warning conditions and required enhanced monitoring; those with indicators in the significantly deteriorated range were included in the key monitoring list and scheduled for regular testing; and those with indicators exceeding the highest threshold boundary were judged to be in a severely deteriorated state and required immediate shutdown and replacement.

[0108] The combination of continuous numerical quantification and fixed-interval threshold grading enables objective differentiation and standardized judgment of degradation states. Continuous numerical values ​​ensure the accuracy and continuity of degradation assessment, while fixed threshold boundaries avoid the subjectivity and variability of human judgment, making the degradation assessment results of surge arresters from different equipment and sites comparable and consistent. This standardized judgment method provides a clear basis for condition-based maintenance of power equipment, helping maintenance personnel quickly grasp the operating status of equipment, formulate targeted handling strategies, and ensure the safe and stable operation of the power grid.

[0109] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for assessing the degradation of the leakage current spectrum of a surge arrester, characterized in that, include: The voltage and leakage current signals at the operating end of the surge arrester are acquired synchronously, and timing alignment and signal preprocessing are completed simultaneously. A dedicated harmonic coupling interference model is constructed based on the synchronous voltage signal to quantitatively solve the voltage homogeneous interference component in the leakage current. The voltage interference component is accurately removed from the original leakage current to obtain the voltage interference-free current signal, and wideband high-precision full spectrum analysis is performed on the voltage interference-free current signal. By combining the operating condition matching environmental interference feature library, environmental superimposed interference is adaptively stripped; pure degraded full-spectrum features unconstrained by external field interference are extracted. The pure degradation full-spectrum characteristics are input into the nonlinear degradation correlation model, and the quantitative degradation index is output and the level is classified.

2. The method for assessing the leakage current spectrum degradation of surge arresters according to claim 1, characterized in that, Signal preprocessing adopts a time-domain joint noise reduction architecture; synchronously coupled baseline correction processing and timing error correction processing eliminate the spectral aliasing problem caused by acquisition timing deviation.

3. The method for assessing the leakage current spectrum degradation of surge arresters according to claim 1, characterized in that, The harmonic coupling interference model adopts the capacitance differential response modeling method; it establishes a one-to-one correspondence between voltage harmonics and interference current based solely on the inherent capacitive parameters of the surge arrester, thus isolating the influence of resistive parameter fluctuations on interference calculation.

4. The method for assessing the leakage current spectrum degradation of surge arresters according to claim 1, characterized in that, Wideband high-precision full spectrum analysis adopts a windowed multi-spectral line joint interpolation correction mode; synchronously corrects spectral leakage error and picket fence effect error, and fully preserves high-order weak degraded harmonic components.

5. The method for assessing the leakage current spectrum degradation of surge arresters according to claim 1, characterized in that, Full spectrum analysis covers the interval from the fundamental wave to the eleventh consecutive odd harmonic; three types of common-source spectrum parameters, namely harmonic amplitude, harmonic phase, and harmonic distortion rate, are collected simultaneously to construct a multi-dimensional spectrum feature set.

6. The method for assessing the leakage current spectrum degradation of surge arresters according to claim 1, characterized in that, The environmental interference feature library with operating conditions matching is divided into multiple dimensions according to the degree of pollution, ambient temperature and humidity, and interphase electromagnetic coupling strength; the corresponding benchmark interference sample is automatically matched according to the real-time operating conditions on site.

7. The method for evaluating the leakage current spectrum degradation of a surge arrester according to claim 1, characterized in that, A hierarchical feature separation architecture is adopted to distinguish interference types; the superimposed signals from three different sources, namely voltage coupling interference, environmental condition interference, and random electromagnetic interference, are separated in layers.

8. The method for evaluating the leakage current spectrum degradation of a surge arrester according to claim 1, characterized in that, The pure degradation full-spectrum characteristics retain only the inherent harmonic components generated by changes in the material properties of the surge arrester valve and internal moisture defects; it completely eliminates non-degradation harmonic signals introduced by external excitation.

9. The method for assessing the leakage current spectrum degradation of a surge arrester according to claim 1, characterized in that, The nonlinear degradation correlation model adopts an adaptive feature weight allocation mechanism; it assigns differentiated weights to different harmonic orders to match the signal response patterns of different degradation types of surge arresters.

10. The method for evaluating the leakage current spectrum degradation of a surge arrester according to claim 1, characterized in that, The degradation index adopts a continuous numerical quantification form; relying on fixed interval threshold boundaries, it realizes the objective differentiation and standardized classification of degradation status.

Images