A GIS disconnector mechanical state and insulation performance dynamic monitoring method and system

Abstract

The application discloses a GIS disconnector mechanical state and insulation performance dynamic monitoring method and system, relates to the technical field of power equipment monitoring, and comprises the following steps: in response to the GIS disconnector executing a closing and opening operation instruction, an elbow displacement signal and a shell acoustic fingerprint signal are acquired; according to the extreme points of the elbow displacement signal, a contact impact interval and a rebound monitoring interval are divided, and an impact time window and a rebound time window are constructed; based on the shell acoustic fingerprint signal in the impact time window, a mechanical impact trend curve is fitted, the shell acoustic fingerprint signal in the impact time window and the rebound time window is subtracted from the mechanical impact trend curve to obtain a residual acoustic fingerprint signal; the residual acoustic fingerprint signal is subjected to an independent component separation operation to acquire a rebound arc transient acoustic characteristic; and an insulation fault probability is calculated according to the transient acoustic characteristic. The application can eliminate mechanical impact interference and extract the hidden rebound arc characteristic.

Description

Technical Field

[0001] This application relates to the field of power equipment monitoring technology, and in particular to a method and system for dynamic monitoring of the mechanical condition and insulation performance of GIS disconnect switches. Background Technology

[0002] During long-term operation, the mechanical components of gas-insulated, fully enclosed switchgear (GIS) are prone to aging and deformation, and the insulation performance is also at risk of deterioration. Current technologies typically employ static X-ray imaging to inspect the mechanical position or capture partial discharge signals while the switch is energized to assess the insulation condition. However, many hidden defects of GIS disconnectors are only exposed during the opening and closing operations.

[0003] Traditional acoustic monitoring technology faces severe interference at the final stage of disconnector switch closing operation. During the final stage of GIS disconnector switch closing operation, mechanical impact occurs between the moving and stationary contacts. Within tens of milliseconds after the impact, the contact mechanism often bounces back, generating a weak, bouncing arc discharge between the contacts. This weak bouncing arc discharge is an early signal of potential insulation problems.

[0004] At this time, the acoustic sensor installed on the GIS casing simultaneously receives both the broadband, strong vibration acoustic signal generated by the mechanical impact and the weak acoustic signal generated by the bouncing arc. The broadband, strong vibration acoustic signal generated by the mechanical impact completely covers the weak acoustic signal generated by the bouncing arc in both the time and frequency domains. Conventional signal filtering methods cannot extract the weak arc acoustic features from the broadband, strong vibration acoustic signal. The masking effect of the broadband, strong vibration acoustic signal from the mechanical impact on the weak acoustic signal from the bouncing arc prevents the early, concealed discharge defects from being detected, causing the GIS disconnect switch to operate with the defect. Summary of the Invention

[0005] In view of the aforementioned problems, this application is hereby filed.

[0006] Therefore, this application provides a method and system for dynamic monitoring of the mechanical condition and insulation performance of GIS disconnect switches, which can solve the problem that the strong vibration sound of the closing impact of GIS disconnect switches masks the weak sound of the bouncing arc, making it impossible to extract the arc characteristics.

[0007] To solve the above-mentioned technical problems, this application provides the following technical solution: In a first aspect, this application provides a method for dynamic monitoring of the mechanical condition and insulation performance of a GIS disconnector, comprising: in response to the GIS disconnector executing a switching operation command, acquiring a crank arm displacement signal and a housing acoustic signature signal; Based on the extreme points in the crank arm displacement signal, the crank arm displacement signal and the shell acoustic pattern signal are divided into the contact impact interval and the bounce monitoring interval, and the impact time window and the bounce time window are constructed based on the contact impact interval and the bounce monitoring interval. Based on the shell acoustic signature signal fitting the mechanical impact trend curve within the impact time window, the amplitude value of the mechanical impact trend curve at the corresponding moment is subtracted from the shell acoustic signature signal within the impact time window and the bounce time window to obtain the residual acoustic signature signal. Independent component separation is performed on the residual acoustic signature signal to obtain the transient acoustic features belonging to the bouncing arc; The probability of insulation failure is calculated based on the transient acoustic characteristics of a bouncing arc.

[0008] Preferably, the step of dividing the crank arm displacement signal and the housing acoustic signature signal into a contact impact interval and a bounce monitoring interval based on the extreme points in the crank arm displacement signal, and constructing an impact time window and a bounce time window based on the contact impact interval and the bounce monitoring interval, includes: The time interval between two adjacent extreme points in the crank arm displacement signal is used as the basic motion interval. If the slope change of the crank arm displacement signal within the basic action range exceeds the preset slope threshold, the basic action range is determined as the contact impact range. The adjacent basic motion interval after the contact impact interval is defined as the bounce monitoring interval, and the displacement change within the bounce monitoring interval is extracted. Based on the contact impact interval and the bounce monitoring interval, an impact time window and a bounce time window are constructed, and the displacement change within the bounce monitoring interval is marked as the mechanical rebound displacement characteristic. Based on the characteristics of mechanical rebound displacement and the extreme values ​​of displacement within the contact impact range, an action state sequence is generated.

[0009] Preferably, the method of fitting the mechanical impact trend curve based on the shell acoustic signature signal within the impact time window, and subtracting the amplitude value of the mechanical impact trend curve at the corresponding moment from the shell acoustic signature signal within the impact time window and the bounce time window to obtain the residual acoustic signature signal includes: Obtain the initial time corresponding to the amplitude extreme point of the acoustic signature signal of the inner and outer shells within the impact time window, and record the absolute amplitude of the amplitude extreme point; Based on the envelope trend of the decrease in the amplitude of the shell acoustic signature signal after the initial moment, the corresponding amplitude decrease trend curve of mechanical impact is fitted and generated as the mechanical impact trend curve. Subtract the amplitude values ​​of the shell acoustic signature signal and the mechanical impact trend curve at the corresponding moments within the impact time window and bounce time window to generate residual acoustic signature signal. The completeness of mechanical impact interference removal is determined based on the amplitude distribution characteristics of the residual acoustic signature signal, and the completeness of mechanical impact interference removal is compared with the preset completeness threshold. In response to mechanical impact interference, if the integrity of the rejection exceeds a preset integrity threshold, the residual acoustic signature signal is output to the subsequent separation operation.

[0010] Preferably, the step of performing independent component separation on the residual acoustic signature signal to obtain the transient acoustic features belonging to the bouncing arc includes: The occurrence time and duration of mechanical rebound displacement characteristics are determined based on the crank arm displacement signal within the bounce time window; Perform independent component separation on the residual acoustic signature signal within the impact time window and bounce time window to generate multiple independent acoustic components; Based on the occurrence time and duration of the mechanical rebound displacement characteristics, candidate components with overlapping time distributions are selected from multiple independent acoustic components. Extract the high-frequency energy percentage from the candidate components and compare the high-frequency energy percentage with a preset percentage threshold; When the proportion of high-frequency energy is greater than a preset proportion threshold, the candidate component is identified as a transient acoustic feature belonging to the bouncing arc.

[0011] Preferably, the calculation of the insulation fault probability based on the transient acoustic characteristics belonging to the bouncing arc includes: Obtain the gas concentration signal in the gas chamber of the GIS disconnect switch, and calculate the difference between the gas concentration signal at the start and end of the bounce time window as the gas concentration increment signal. In response to a gas concentration increment signal exceeding a concentration difference threshold, the proportion of high-frequency energy belonging to the transient acoustic features of the bouncing arc within the corresponding bouncing time window is extracted. The product of the gas concentration increment signal and the high-frequency energy ratio is mapped to a preset probability range to generate the insulation fault probability. The monitoring log is updated based on the insulation fault probability, and the insulation fault probability is correlated with the gas concentration increment signal and stored in the monitoring log; In response to a gas concentration increment signal being less than or equal to a concentration difference threshold, the insulation fault probability is set to zero and recorded in the monitoring log.

[0012] Preferably, after calculating the insulation fault probability based on the transient acoustic characteristics belonging to a bouncing arc, the method further includes: In response to the GIS disconnector performing multiple opening and closing operations, the probability of insulation faults corresponding to each opening and closing operation is obtained. Based on the time sequence of each opening and closing operation, the difference between the probabilities of two adjacent insulation faults is calculated, and fault growth nodes with non-zero differences are extracted. Calculate the slope of the change in insulation failure probability with the number of operations based on the difference of all insulation failure probabilities, and obtain the positive or negative attribute of the slope. In response to the slope being positive and the slope being greater than the slope threshold, an early warning signal indicating insulation performance degradation is output. In response to the slope being negative or the slope being less than the slope threshold, the current monitoring status is maintained and the insulation fault probability for the next opening and closing operation is acquired.

[0013] Preferably, the independent component separation operation on the residual voiceprint signal includes: The background acoustic signature segment before the impact time window is obtained as the noise reference, and the signal-to-noise ratio of the residual acoustic signature signal within the bounce time window relative to the noise reference is calculated. In response to the signal-to-noise ratio attribute value being in the low signal-to-noise ratio range, the iteration step size of the independent component separation operation is reduced to the preset minimum step size to limit the mechanical residual vibration noise from being separated into independent components. In response to the signal-to-noise ratio attribute value being in the high signal-to-noise ratio range, the iteration step size of the independent component separation operation is increased to the preset maximum step size to accelerate the convergence of the bouncing arc component; Based on the reduced or increased iteration step size, perform iterative separation operation on the residual acoustic signature signal within the bounce time window until the mutual information of the separated independent components is lower than the mutual information threshold. Independent components that satisfy the condition that the mutual information is below the mutual information threshold are marked to generate independent acoustic components.

[0014] Preferably, the response to the slope being positive and the slope being greater than a slope threshold, outputting a warning signal indicating insulation performance degradation, includes: Obtain the feature combination consisting of insulation fault probability and slope, and extract the interval where the insulation fault probability is located in the feature combination; In response to the insulation failure probability being in the first probability range and the slope being greater than the slope threshold, a first-level warning signal indicating mild insulation degradation is output. In response to the insulation fault probability being in the second probability interval and the slope being greater than the slope threshold, a second-level warning signal indicating severe insulation degradation is output, wherein the second probability interval is greater than the first probability interval; Based on the first-level or second-level warning signal, maintenance suggestion information is generated and pushed to the monitoring terminal; In response to the insulation fault probability being in the third probability range, the warning signal output is shielded, where the third probability range is lower than the first probability range.

[0015] Preferably, after extracting the high-frequency energy percentage of the candidate components and comparing the high-frequency energy percentage with a preset percentage threshold, the method further includes: Obtain the energy distribution ratio of the candidate components in adjacent frequency bands, and combine the energy distribution ratio with the time-domain amplitude variation curve of the candidate components; The combined results of the energy distribution ratio and the time-domain amplitude variation curve are cross-correlated with the preset arc acoustic distribution ratio to generate cross-correlation coefficients. When the cross-correlation coefficient is greater than the correlation coefficient threshold, the candidate component is identified as a transient acoustic feature belonging to the bouncing arc. Based on the occurrence time of the transient acoustic characteristics belonging to the bouncing arc, the corresponding displacement deviation node is located in the crank arm displacement signal; Arc positioning information is generated based on displacement deviation nodes, and the arc positioning information is bound and stored with transient acoustic features belonging to bouncing arcs.

[0016] Secondly, this application also provides a dynamic monitoring system for the mechanical condition and insulation performance of a GIS disconnector, including: a signal synchronization acquisition module configured to acquire crank arm displacement signal and housing acoustic signal in response to the GIS disconnector executing the opening and closing operation command; The timing segmentation module is configured to divide the crank arm displacement signal and the housing acoustic pattern signal into a contact impact interval and a bounce monitoring interval based on the extreme points in the crank arm displacement signal, and to construct an impact time window and a bounce time window based on the contact impact interval and the bounce monitoring interval. The interference removal module is configured to fit the mechanical impact trend curve based on the shell acoustic signature signal within the impact time window, and subtract the amplitude value of the mechanical impact trend curve at the corresponding moment from the shell acoustic signature signal within the impact time window and the bounce time window to obtain the residual acoustic signature signal. The feature extraction module is configured to perform independent component separation on the residual acoustic signature signal to obtain transient acoustic features belonging to the bouncing arc; The fusion judgment module is configured to calculate the probability of insulation failure based on the transient acoustic characteristics of a bouncing arc.

[0017] Implementing this application has the following beneficial effects: This application provides a method and system for dynamic monitoring of the mechanical condition and insulation performance of GIS disconnect switches. Addressing the problem that the broadband strong vibration sound patterns generated by mechanical impact at the end of the closing phase of a GIS disconnect switch can mask the weak bouncing arc sound patterns generated by contact bounce, making it difficult to detect early hidden discharge defects, this application simultaneously acquires the crank arm displacement signal and the casing sound pattern signal. It accurately divides the contact impact interval and the bounce monitoring interval using the crank arm displacement extreme value, and after fitting the mechanical impact trend curve based on the impact time window, it removes strong interference components from the original sound patterns. Then, it performs independent component separation on the residual sound patterns, thereby effectively eliminating mechanical impact interference and extracting the masked transient acoustic features of the bouncing arc. Furthermore, it calculates the insulation fault probability, achieving dynamic joint monitoring of the mechanical condition and insulation performance of GIS disconnect switches, improving the ability to identify early insulation hazards, and preventing equipment from operating with defects before they are detected. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is an overall flowchart of a dynamic monitoring method for the mechanical condition and insulation performance of a GIS disconnector, as described in this application. Figure 2 This is a logic flowchart of a dynamic monitoring method for the mechanical condition and insulation performance of a GIS disconnector, as described in this application. Figure 3 This is a logic flowchart of signal processing and feature extraction for a dynamic monitoring method of mechanical status and insulation performance of a GIS disconnector, which is involved in this application. Figure 4 This is a logic flowchart of the insulation fault probability calculation and trend early warning of a dynamic monitoring method for the mechanical condition and insulation performance of a GIS disconnector, which is involved in this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0021] This application provides a method for dynamic monitoring of the mechanical condition and insulation performance of GIS disconnect switches.

[0022] Figure 1 A flowchart of a method for dynamic monitoring of the mechanical condition and insulation performance of GIS disconnect switches is shown, including: Step S1: In response to the GIS disconnecting switch executing the opening and closing operation command, acquire the crank arm displacement signal and the housing acoustic signal; Specifically, the operation of acquiring crank arm displacement signal and casing acoustic signal in response to the GIS disconnecting switch executing the opening and closing operation command is the data foundation of the entire monitoring process.

[0023] It should be noted that when the GIS disconnector receives the opening and closing operation command, the mechanical mechanism starts to move, and at this time, two key physical quantities need to be collected simultaneously.

[0024] In some embodiments, the crank arm displacement signal is acquired by a displacement sensor installed on the crank arm of the disconnector switch. The displacement sensor converts the rotation angle or linear travel of the crank arm into a voltage signal output, and the sampling frequency is set to no less than 1000 Hz to ensure that the displacement change at the moment of impact can be captured. The housing acoustic signal is acquired by an acoustic sensor installed on the GIS housing near the disconnector switch contacts. The acoustic sensor converts the vibration of the housing surface into an acoustic electrical signal output, and the sampling frequency is set to no less than 50000 Hz to meet the requirements for wideband vibration signal acquisition.

[0025] It is important to emphasize that the acquisition of the two signals is strictly synchronized on the time axis, with the initial moment of the crank arm movement as the zero point for alignment. For example, in a scenario of closing a 550 kV GIS disconnector switch, after the closing command is issued, the displacement sensor records the displacement curve of the crank arm from the initial position to the closing termination position at a sampling rate of 1000 Hz, and the acoustic signature sensor records the shell vibration waveform at a sampling rate of 50000 Hz. The two signals are sampled synchronously through a unified clock source to ensure the accurate correspondence between the subsequent acoustic signature characteristics and the mechanical action moment. After acquiring the two signals, they are transmitted to the signal processing unit for subsequent interval division and interference removal operations. In this application, the sampling frequencies of 1000 Hz and 50000 Hz are calculated using the Nyquist sampling law for the mechanical action spectrum and the arc acoustic signature spectrum of the disconnector switch, with a margin of 2.

[0026] Step S2: Based on the extreme points in the crank arm displacement signal, the crank arm displacement signal and the shell acoustic pattern signal are divided into the contact impact interval and the bounce monitoring interval, and the impact time window and the bounce time window are constructed based on the contact impact interval and the bounce monitoring interval. Understandably, in this application, the extreme points in the crank arm displacement signal are identified as the benchmark for segmenting the action phase. Then, the contact impact interval and the bounce monitoring interval are distinguished according to the slope characteristics between adjacent extreme points. Finally, the interval is mapped to the time axis to construct the corresponding time window, which can provide time range constraints for subsequent signal processing.

[0027] Preferably, the step of dividing the crank arm displacement signal and the housing acoustic signature signal into a contact impact interval and a bounce monitoring interval based on the extreme points in the crank arm displacement signal, and constructing an impact time window and a bounce time window based on the contact impact interval and the bounce monitoring interval, includes: Step S21: Obtain the time interval between two adjacent extreme points in the crank arm displacement signal as the basic motion interval; It is understandable that the operation of obtaining the time interval between two adjacent extreme points in the crank arm displacement signal as the basic motion interval aims to discretize the continuous displacement signal into multiple motion segments.

[0028] When the crank arm moves during the opening and closing process, if the displacement signal shows an alternating peak and trough pattern due to impact and rebound, the first-order difference is calculated on the acquired crank arm displacement signal. When the sign of the difference changes from positive to negative, the maximum point is identified. When the difference sign changes from negative to positive, the minimum point is identified. The time span between adjacent maximum and minimum points is extracted and defined as a basic motion interval. Extreme point identification eliminates glitches caused by sensor noise, retaining only extreme points where the amplitude change exceeds 1% of the overall displacement. Through operation, the entire displacement signal is divided into multiple interconnected basic motion intervals, each representing a unidirectional mechanical motion process, providing a basic unit for subsequent impact and bounce identification.

[0029] Step S22: In response to the change in the slope of the crank arm displacement signal within the basic action range exceeding a preset slope threshold, the basic action range is determined as the contact impact range. Specifically, the operation of determining the basic motion range as the contact impact range in response to the change in the slope of the crank arm displacement signal exceeding a preset slope threshold is the key to distinguishing impact actions from normal motion.

[0030] It should be noted that the average slope of the crank arm displacement signal within the basic motion interval is calculated as the ratio of the difference between the displacement value at the end of the interval and the displacement value at the beginning of the interval to the interval time length.

[0031] When the moving and stationary contacts collide, the displacement changes drastically in a very short time, causing the absolute value of the average slope to be much greater than that of the normal movement phase. The absolute value of the average slope is compared with a preset slope threshold. If it exceeds the preset slope threshold, the contact impact process corresponding to the basic movement interval is determined and marked as the contact impact interval.

[0032] To verify the impact of different parameter combinations on the processing effect in this step, the multivariate mapping table for contact impact interval determination shown in Table 1 is used for illustration: Table 1 Multivariate Mapping Table for Determining Contact Impact Zone

[0033] In this application, the preset slope threshold is obtained by statistically analyzing the slope distribution of historical normal closing actions and taking the upper limit of the 95% confidence interval, and is set to 2000 mm / s.

[0034] Step S23: Determine the adjacent basic action interval after the contact impact interval as the bounce monitoring interval, and extract the displacement change within the bounce monitoring interval. Furthermore, once the contact impact zone is determined, the rebound after the impact will inevitably follow the impact due to the continuity of mechanical motion.

[0035] Therefore, in this embodiment, the basic action interval immediately following the contact impact interval is extracted and defined as the bounce monitoring interval.

[0036] Within the bounce monitoring range, the maximum and minimum values ​​of the displacement signal are extracted, and the difference between the maximum and minimum values ​​is calculated. This difference is used as the displacement change within the bounce monitoring range. The displacement change reflects the physical amplitude of the contact rebound and is an important parameter for measuring the severity of mechanical rebound. It will subsequently be used as a mechanical rebound displacement feature in the generation of state sequences and the screening of arc features.

[0037] Step S24: Construct impact time window and bounce time window based on contact impact interval and bounce monitoring interval, and mark the displacement change in the bounce monitoring interval as mechanical rebound displacement feature; In detail, the operation of constructing impact time windows and bounce time windows based on the contact impact interval and bounce monitoring interval involves extracting the start and end times of the contact impact interval on the time axis to construct the impact time window; and extracting the start and end times of the bounce monitoring interval on the time axis to construct the bounce time window.

[0038] It should be noted that the two time windows are adjacent but do not overlap on the time axis. At the same time, the displacement change within the bounce monitoring interval extracted in S23 is formally marked as the mechanical rebound displacement feature. The mechanical rebound displacement feature is bound to the bounce time window and stored. In the subsequent independent component separation operation, it is used to filter the arc acoustic component that coincides with the mechanical rebound time, ensuring that the extracted acoustic features are consistent with the bounce action in temporal logic.

[0039] Step S25: Generate an action state sequence based on the mechanical rebound displacement characteristics and the displacement extreme values ​​within the contact impact range; It should be noted that the operation of generating the action state sequence based on the mechanical rebound displacement characteristics and the displacement extreme value within the contact impact range is to quantify the mechanical action into a numerical sequence.

[0040] Specifically, the maximum displacement within the contact impact zone is extracted as the impact extreme value. The impact extreme value and the mechanical rebound displacement characteristics are arranged in chronological order to generate an array containing two elements, namely the action state sequence. The action state sequence objectively records the impact amplitude and rebound amplitude, and is stored as a mechanical state file for this opening and closing operation, which is convenient for subsequent joint analysis with the insulation fault probability.

[0041] Step S3: Fit the mechanical impact trend curve based on the shell acoustic signature signal within the impact time window, and subtract the amplitude value of the mechanical impact trend curve at the corresponding moment from the shell acoustic signature signal within the impact time window and the bounce time window to obtain the residual acoustic signature signal. Understandably, this application first uses strong vibration data within the impact time window to fit a curve reflecting the decline law of mechanical impact acoustic signature, and then subtracts the corresponding amplitude of the fitted curve from the original acoustic signature within the impact and bounce time window, thereby eliminating the broadband strong interference caused by mechanical impact at the physical level and leaving residual signals that may contain weak arc characteristics.

[0042] Preferably, the method of fitting the mechanical impact trend curve based on the shell acoustic signature signal within the impact time window, and subtracting the amplitude value of the mechanical impact trend curve at the corresponding moment from the shell acoustic signature signal within the impact time window and the bounce time window to obtain the residual acoustic signature signal includes: Step S31: Obtain the initial time corresponding to the amplitude extreme point of the acoustic signature signal of the inner and outer shells within the impact time window, and record the absolute amplitude of the amplitude extreme point; To achieve accurate fitting, the sampling point with the largest absolute amplitude value is searched within the acoustic signature signal of the inner and outer shells within the impact time window. This point represents the moment when the mechanical impact produces the most intense vibration. The sampling moment corresponding to the maximum absolute amplitude value is recorded as the initial moment, and the maximum absolute amplitude value is also recorded as the absolute amplitude magnitude. The initial moment and the absolute amplitude magnitude constitute the starting coordinates and initial energy value of the impact vibration descent process, providing crucial boundary constraints for subsequent fitting of the descent curve. In this application, the search for the amplitude extremum point is obtained by traversing the absolute values ​​of all sampling points within the impact time window and taking the maximum value.

[0043] Step S32: Based on the trend of the extreme points of the decrease in the amplitude of the shell acoustic signal after the initial moment, a corresponding mechanical impact amplitude decrease trend curve is fitted and generated as the mechanical impact trend curve. Preferably, the acoustic signal generated by mechanical impact exhibits a physical characteristic where the amplitude decreases exponentially with time after the initial moment.

[0044] Furthermore, all local maxima of the voiceprint signal after the initial moment are extracted to form a discrete point sequence that reflects the amplitude decrease profile.

[0045] Furthermore, the least squares method is used to fit the discrete point sequence with an exponential function to generate a continuous amplitude decreasing trend curve.

[0046] In one implementation, the aforementioned downward trend curve can be obtained through the following specific calculation method, wherein: ; In the formula, The mechanical impact trend curve at time [time] The amplitude value, in volts; Indicates the absolute amplitude, measured in volts; This represents the descent rate parameter, in units of seconds. This represents a time variable, with the unit being seconds. Indicates the initial time, in seconds. Descent rate parameter. The solution is obtained by minimizing the sum of squared residuals.

[0047] For example, in a 550 kV GIS disconnector closing scenario, at the initial moment... For 0.05 seconds, the absolute amplitude is... The calculated descent rate parameter is 5 volts. If the value is 30 seconds, then the mechanical impact trend curve is... By performing a fitting operation, the complex broadband impact vibration is transformed into a smooth amplitude decreasing trend curve, providing an accurate baseline for subsequent subtraction of impact interference. In this application, the iterative convergence condition for the least squares fitting is set to the sum of squared residuals being less than the square of 0.01 volts. In this embodiment, the above convergence condition is obtained by taking 10 times the root mean square of the amplitude of the background noise level of the acoustic sensor.

[0048] Step S33: Subtract the amplitude values ​​of the shell acoustic signature signal and the mechanical impact trend curve at the corresponding moments within the impact time window and bounce time window to generate residual acoustic signature signal. It should be noted that after the mechanical impact trend curve is generated, for each sampling moment within the impact time window and the bounce time window, the original amplitude value of the shell acoustic signature signal and the amplitude value of the mechanical impact trend curve are obtained at that moment. The difference between the original amplitude value and the amplitude value of the mechanical impact trend curve is calculated, and the difference is used as the residual acoustic signature signal amplitude value at that moment.

[0049] In some embodiments, all sampling points within the impact time window and bounce time window are traversed to generate a complete residual acoustic signature signal. The subtraction operation eliminates the dominant trend of mechanical impact from the original acoustic signature, allowing the weak high-frequency bounce arc signal, which was originally masked by the low-frequency strong vibration, to stand out in the residual acoustic signature. For example, if the original amplitude value at a certain sampling moment is 4 volts and the amplitude value of the mechanical impact trend curve at the corresponding moment is 3.8 volts, then the amplitude value of the residual acoustic signature signal is 0.2 volts.

[0050] Step S34: Confirm the completeness of mechanical impact interference removal based on the amplitude distribution characteristics of the residual acoustic signature signal, and compare the completeness of mechanical impact interference removal with the preset completeness threshold. Understandably, in order to verify whether mechanical impact interference has been effectively eliminated, it is necessary to calculate the amplitude distribution characteristics of the residual acoustic signature signal.

[0051] When calculating the root mean square (RMS) value of the residual acoustic signature signal amplitude within the impact time window, the ratio of this RMS value to the RMS value of the original acoustic signature signal amplitude within the impact time window is defined as the completeness of mechanical impact interference removal. The closer the completeness of removal is to 0, the more thorough the removal.

[0052] Furthermore, the calculated completeness of mechanical impact interference removal is compared with a preset completeness threshold. In this application, the preset completeness threshold is set according to the proportion of residual signal energy when arc features are successfully extracted in historical data. The preset completeness threshold is set to 0.15, which means that the residual acoustic energy is required to be reduced to less than 15% of the original acoustic energy. The preset completeness threshold mentioned here is obtained by statistically analyzing the maximum percentage of residual energy when arc features are successfully extracted more than 10 times.

[0053] Step S35: In response to the mechanical impact interference removal integrity being greater than a preset integrity threshold, the residual acoustic signature signal is output to the subsequent separation operation. When the integrity of the rejection is less than or equal to the preset integrity threshold, it indicates that the mechanical impact interference has been effectively rejected. At this time, the processed residual acoustic signal is officially output to the subsequent independent component separation operation as input data for extracting the weak bouncing arc features.

[0054] Step S4: Perform independent component separation on the residual acoustic signature signal to obtain the transient acoustic features belonging to the bouncing arc; Understandably, step S4 is mainly for blind source separation of the residual acoustic signal after the main impact trend has been eliminated. By utilizing the statistical independence between signals, the superimposed mechanical residual noise and the weak acoustic signal generated by the bouncing arc are separated, thereby obtaining the pure transient acoustic characteristics of the arc.

[0055] Preferably, the step of performing independent component separation on the residual acoustic signature signal to obtain the transient acoustic features belonging to the bouncing arc includes: Step S41: Determine the occurrence time and duration of the mechanical rebound displacement characteristics based on the crank arm displacement signal within the bounce time window; Specifically, the operation of determining the occurrence time and duration of mechanical rebound displacement characteristics based on the crank arm displacement signal within the bounce time window is to provide a time reference for the subsequent screening of acoustic components.

[0056] It should be noted that in the crank arm displacement signal within the bounce time window, the mechanical rebound displacement characteristic is manifested as a distinct peak or trough.

[0057] In a preferred implementation, the initial zero-crossing point of the peak or trough is extracted as the occurrence time, and the final zero-crossing point of the peak or trough is extracted. The difference between the occurrence time and the end time is calculated as the duration. The occurrence time and duration define the physical time range of the contact bounce. Only acoustic components occurring within this time range meet the necessary conditions to become acoustic characteristics of a bouncing arc.

[0058] Preferably, the independent component separation operation on the residual voiceprint signal can be implemented through steps S411-S415, specifically including: Step S411: Obtain the background acoustic signature segment before the impact time window as the noise reference, and calculate the signal-to-noise ratio attribute value of the residual acoustic signature signal within the bounce time window relative to the noise reference. Furthermore, the operation of obtaining the background acoustic signature segment before the impact time window as the noise reference and calculating the signal-to-noise ratio of the residual acoustic signature signal within the bounce time window relative to the noise reference is a prerequisite for adaptively adjusting the separation step size.

[0059] In this embodiment, the voiceprint signal 50 milliseconds before the impact time window is extracted as the background voiceprint segment, and the root mean square value of the amplitude of the background voiceprint segment is calculated as the noise reference power.

[0060] Furthermore, the root mean square value of the residual acoustic signature signal amplitude within the bounce time window is calculated as the signal power.

[0061] Furthermore, the ratio of signal power to noise reference power is calculated, and the logarithm of the ratio is taken to obtain the signal-to-noise ratio attribute value.

[0062] For example, in a certain operation, the root mean square value of the background acoustic signature segment is 0.01 volts, and the root mean square value of the residual acoustic signature within the bounce time window is 0.1 volts. Therefore, the signal-to-noise ratio (SNR) attribute value is 10 multiplied by the logarithm of the ratio of 0.1 to 0.01 (base 10), which is 20 dB. The SNR attribute value reflects the prominence of the arc characteristics in the residual signal relative to the background noise. In this application, the SNR attribute value is used to guide the selection of the iteration step size for independent component separation operations. The aforementioned 50-millisecond truncation time is obtained by extracting the autocorrelation length of the environmental noise during the steady-state operation of the GIS equipment.

[0063] Step S412: In response to the signal-to-noise ratio attribute value being in the low signal-to-noise ratio range, the iteration step size of the independent component separation operation is reduced to a preset minimum step size to limit the mechanical residual vibration noise from being separated into independent components. Understandably, when the signal-to-noise ratio (SNR) value is in the low SNR range, it indicates that the acoustic signal generated by the bouncing arc is extremely weak and easily confused with mechanical residual vibration noise. To achieve high-precision separation, if the iteration step size is too large at this time, it will cause the separation process to diverge or the noise to be mistaken for an independent component.

[0064] In this application, the iteration step size of the independent component separation operation is reduced to a preset minimum step size. Reducing the step size allows for a more detailed search for the direction of weak components in the signal space, limiting the incorrect separation of mechanical residual noise into independent components and ensuring separation accuracy. The aforementioned low signal-to-noise ratio range is set to less than 15 dB, and the preset minimum step size is set to 0.01. In this embodiment, the low signal-to-noise ratio range and the preset minimum step size are obtained by performing step size optimization on simulated aliased signals with a signal-to-noise ratio below 15 dB.

[0065] Step S413: In response to the signal-to-noise ratio attribute value being in the high signal-to-noise ratio range, the iteration step size of the independent component separation operation is increased to the preset maximum step size to accelerate the convergence of the bouncing arc component. Specifically, when the signal-to-noise ratio (SNR) value is in the high SNR range, it indicates that the bouncing arc signal is relatively obvious and not easily masked by noise.

[0066] Preferably, in order to improve computational efficiency, the iteration step size of the independent component separation operation is increased to the preset maximum step size. Increasing the step size can accelerate the convergence speed of the independent component separation calculation process, shorten the processing time while ensuring the accuracy of the separation results, and meet the real-time requirements of monitoring.

[0067] For ease of understanding, the high signal-to-noise ratio range is set to greater than 25 dB, and the preset maximum step size is set to 0.15. In this application, the high signal-to-noise ratio range and the preset maximum step size are obtained by testing the maximum available step size that can be separated within the computing power limit time of the computing platform.

[0068] Step S414: Perform iterative separation operation on the residual acoustic signature signal within the bounce time window based on the reduced or increased iteration step size until the mutual information of the separated independent components is lower than the mutual information threshold. It should be noted that this application provides a method for performing iterative separation operations on residual acoustic signature signals within a bouncing time window based on decreasing or increasing the iteration step size, until the mutual information of the separated independent components is lower than the mutual information threshold, in order to solve the problem of strong vibrations masking weak arc acoustic signatures.

[0069] In some embodiments, the iterative separation operation employs a fast fixed-point computation method, constructing a multidimensional observation vector from the residual acoustic signature signal and its delayed copy, and extracting independent components by finding the direction with the greatest non-Gaussianity. For example, in each iteration, the separation matrix is ​​updated according to the step size adjusted in S412 or S413, and the mutual information of the separated components is calculated. Mutual information measures the statistical dependence between components; the closer the mutual information is to 0, the more thorough the separation.

[0070] In one implementation, the update process of the separation matrix in the aforementioned iterative separation operation can be obtained through the following specific calculation method, wherein: ; In the formula, Indicates the first The separation matrix after the nth iteration is dimensional dimensionless matrix; Indicates the first The separation matrix of the next iteration is dimensional dimensionless matrix; This represents the iteration step size after decreasing or increasing, and is a dimensionless scalar. Denotes the identity matrix as follows: dimensional dimensionless matrix; Represents the independent component vectors Perform non-linear operations element by element The resulting column vector is dimensional dimensionless vector; Indicates the first The independent component vectors separated in the next iteration are dimensional dimensionless vector; Represents independent component vectors The transpose of is dimensional dimensionless vector; The dimension of the observation vector is represented by the number of independent components. The observed signal is multiplied by the updated separation matrix to obtain new independent component vectors, and the mutual information of the independent component vectors obtained in two consecutive iterations is calculated. The iteration stops when the mutual information is below a mutual information threshold, and the current independent component vector is output.

[0071] In this application, because the weak arc discharge generated by the contact bounce at the end of the GIS disconnector switch closing is extremely short, usually between a few milliseconds and tens of milliseconds, and the arc acoustic signal exhibits broadband characteristics in the frequency domain, it is very easy to overlap with the mechanical residual noise present at the same time in the time and frequency domain. When the signal-to-noise ratio is extremely low, the conventional fixed step size separation method is very easy to get trapped in local extrema, resulting in the failure of arc component extraction. Therefore, when it is determined that the signal-to-noise ratio attribute value is in the low signal-to-noise ratio range and the current iteration number exceeds the preset number threshold, if the mutual information decrease rate is lower than the preset rate threshold, the iteration step size is further subdivided from the preset minimum step size, and it is gradually tested according to the step size gradient of 0.001. After each test, the kurtosis value of the independent component is calculated. When the kurtosis value is the maximum, the current test step size is locked as the actual execution step size to continue iterating. If the signal-to-noise ratio attribute value is in the high signal-to-noise ratio range, the preset maximum step size is directly used for iteration, and the iteration is forcibly terminated when the mutual information no longer decreases for 3 consecutive times to prevent overfitting noise. The aforementioned mutual information threshold was determined through experiments on the mixing and separation of a large number of pure arc signals and noise signals. In this application, the mutual information threshold is set to 0.05, the preset number threshold is set to 50 times, and the preset rate threshold is set to 0.001. The preset number threshold and preset rate threshold mentioned here are obtained by statistically analyzing the flat interval of the iterative convergence curve.

[0072] Step S415: Mark the independent components that satisfy the mutual information content being lower than the mutual information threshold to generate independent acoustic components. The process of marking independent components whose mutual information is below a threshold to generate independent acoustic components completes signal decoupling. When the mutual information of independent components is below the threshold, it indicates that the components have achieved statistical independence, and the mechanical residual noise and potential bouncing arc signals have been effectively decoupled. Independent components that meet the conditions are assigned unique identifiers to generate independent acoustic components. The marking operation includes the correspondence between components on the time axis, ensuring that each independent acoustic component can be aligned with a specific moment within the bouncing time window, providing a data foundation for subsequent filtering based on time overlap.

[0073] Step S42: Perform independent component separation on the residual acoustic signature signal within the impact time window and bounce time window to generate multiple independent acoustic components. The residual acoustic signals within the impact time window and the bounce time window are combined as input. After independent component separation in steps S411 to S415, the original aliased signal is decomposed into multiple independent components in the time and frequency domains. These multiple independent acoustic components include residual components from the mechanical impact, background noise components, and potential bounce arc components, which await further identification based on physical characteristics.

[0074] Step S43: Based on the occurrence time and duration of the mechanical rebound displacement characteristics, candidate components with overlapping time distributions are selected from multiple independent acoustic components. It is understandable that the bouncing arc must physically occur simultaneously with the contact rebound action; therefore, temporal overlap is the primary condition for selecting arc components. When calculating the energy concentration time period for each independent acoustic component, it is defined as the continuous time period where the absolute value of the amplitude is greater than 10% of the maximum amplitude.

[0075] Furthermore, the intersection of the energy concentration period and the occurrence time of the mechanical rebound displacement characteristic with the time period defined by the duration is calculated. If the length of the intersection exceeds 50% of the length of the energy concentration period of the independent acoustic component, the independent acoustic component is determined to overlap with the mechanical rebound in time distribution and is selected as a candidate component. This operation eliminates noise components existing outside the rebound period, significantly narrowing the target search range. In this application, the 50% threshold is obtained by statistically analyzing the lower limit of the overlap ratio between the time span of historical bouncing arc signals and the rebound duration.

[0076] Step S44: Extract the high-frequency energy ratio of the candidate components and compare the high-frequency energy ratio with the preset ratio threshold. Specifically, the acoustic signal energy generated by mechanical impact is mainly concentrated in the low-frequency band, while the acoustic signal generated by electric arc discharge has transient pulse characteristics and its energy is distributed in the higher frequency band.

[0077] It should be pointed out that the proportion of high-frequency energy is the key frequency domain characteristic that distinguishes electric arc from mechanical residual vibration.

[0078] In a preferred implementation, a Fourier transform is performed on the candidate components to calculate the ratio of energy above 20 kHz to the total energy, thus obtaining the high-frequency energy proportion. This high-frequency energy proportion is compared to a preset proportion threshold. If the high-frequency energy proportion is greater than the preset proportion threshold, it indicates that the candidate component possesses the frequency domain characteristics of arc acoustics. In this application, the preset proportion threshold is set to 0.4. The 20 kHz frequency division point and the 0.4 threshold are obtained through spectral statistics of historical arc acoustic data.

[0079] Preferably, after extracting the high-frequency energy percentage of the candidate components and comparing the high-frequency energy percentage with a preset percentage threshold, the method further includes: Step S441: Obtain the energy distribution ratio of the candidate component in the adjacent frequency band, and combine the energy distribution ratio with the time-domain amplitude variation curve of the candidate component; Furthermore, to further confirm the characteristics of the electric arc, it is necessary to analyze the spectral details of the candidate components. In this embodiment, the spectrum is divided into multiple adjacent frequency bands, for example, multiple sub-bands with a bandwidth of 5 kHz.

[0080] Furthermore, the energy ratio of each adjacent sub-band is calculated to generate an energy distribution ratio sequence. Simultaneously, the amplitude envelope variation of the candidate components in the time domain is extracted to form a time-domain amplitude variation curve.

[0081] Furthermore, the energy distribution ratio sequence and the time-domain amplitude variation curve are aligned on the time axis to form a combined feature vector. This combination operation integrates the signal's distribution relationship in the frequency domain with its transient variation characteristics in the time domain, providing a multi-dimensional comparison basis for subsequent matching with a standard arc template.

[0082] Step S442: Perform cross-correlation calculation on the combined result of the energy distribution ratio and the time-domain amplitude change curve and the preset arc acoustic distribution ratio to generate cross-correlation coefficients; In other words, this application provides a method for performing cross-correlation calculations on the combined results of the energy distribution ratio and the time-domain amplitude variation curve with a preset arc acoustic distribution ratio to generate a cross-correlation coefficient, in order to solve the problem that relying solely on frequency domain thresholds can easily lead to misjudgments.

[0083] To achieve accurate matching, the preset arc acoustic distribution ratio is based on a standard template extracted from a large number of measured bouncing arc acoustic patterns, which includes the typical frequency band energy ratios and time-domain pulse descent patterns of the arc. Cross-correlation calculations measure the waveform similarity between the candidate component combination results and the standard template.

[0084] In this application, the aforementioned cross-correlation coefficients can be calculated in the following manner: First, calculate the sum of covariances between the combined result and the preset arc acoustic distribution ratio. Then, calculate the ratio of the sum of covariances to the product of the standard deviation of the combined result and the standard deviation of the preset arc acoustic distribution ratio.

[0085] In an optional implementation, the above cross-correlation coefficient can be characterized as: ; In the formula, Represents the cross-correlation coefficient; The length of the combined result is represented in units of data. Indicates the first element in the combination result The value of each data point; This represents the average value of the combined results; Indicating the proportion of the preset electric arc acoustic distribution The value of each data point; This represents the average value of the preset acoustic distribution ratio of the electric arc. The calculation process iterates through the combined results and the standard template point by point, calculating the ratio of the covariance to the product of the respective standard deviations. Cross-correlation coefficient. The value ranges from -1 to 1. The closer the value is to 1, the more similar the candidate component is to the standard arc acoustic characteristics.

[0086] It is understandable that, in a specific computational instance, the combined result data length Given a value of 100, the numerator (sum of covariances) is 2.5, and the denominator (square root of the product of standard deviations) is 2.6. Therefore, the cross-correlation coefficient is... The correlation coefficient is approximately 0.96, indicating a high degree of correlation. The cross-correlation coefficient quantifies the similarity between candidate components and the real arc from the perspective of the joint time-frequency domain, which is more reliable than the proportion in a single frequency domain. In this application, the preset arc acoustic distribution ratio is obtained by extracting the energy proportion of the frequency band of the bouncing arc acoustic pattern confirmed by more than 50 actual measurements and the mean of the time-domain envelope.

[0087] Step S443: In response to the cross-correlation coefficient being greater than the correlation coefficient threshold, the candidate component is confirmed as a transient acoustic feature belonging to the bouncing arc. When the cross-correlation coefficient is greater than the correlation coefficient threshold, it indicates that the comprehensive characteristics of the candidate component in the time-frequency domain are highly consistent with the known characteristics of the bouncing arc, excluding interference from mechanical residual vibration or other random noise. At this point, the candidate component is formally confirmed as belonging to the transient acoustic characteristics of the bouncing arc. In this application, the correlation coefficient threshold is set to 0.85. The correlation coefficient threshold mentioned here is obtained by statistically analyzing the upper limit of the cross-correlation coefficient between non-arc components and the arc template in historical misjudgment events.

[0088] Step S444: Based on the occurrence time of the transient acoustic characteristics belonging to the bouncing arc, locate the corresponding displacement deviation node in the crank arm displacement signal. Specifically, after confirming the existence of the bouncing arc, it is necessary to establish a connection between the arc and the mechanical action. It should be noted that, based on the timing of the transient acoustic characteristics, the displacement value corresponding to the same time point should be searched in the crank arm displacement signal.

[0089] In a preferred implementation, since the electric arc occurs at the instant the contacts bounce and separate, the displacement signal typically deviates from the main trend line, forming a slight fluctuation. The displacement fluctuation point corresponding to this time point is defined as the displacement deviation node. Locating the displacement deviation node clarifies the exact location of the mechanical system at the moment the arc is generated, providing physical support for analyzing the correlation between the bounce amplitude and the arc intensity.

[0090] Step S445: Generate arc positioning information based on displacement deviation nodes, and bind and store the arc positioning information with the transient acoustic features belonging to the bouncing arc. The timestamps and displacement deviation amounts of the displacement deviation nodes are extracted and combined to generate arc location information. This arc location information is then bound to the corresponding transient acoustic features to form a complete arc event record. This bound storage facilitates the tracing of mechanical action details at the time of a specific arc event during subsequent historical data analysis, enabling joint diagnosis of mechanical condition and insulation performance.

[0091] Step S45: In response to the high-frequency energy ratio being greater than a preset ratio threshold, the candidate component is confirmed as a transient acoustic feature belonging to the bouncing arc. It should be noted that when the proportion of high-frequency energy calculated in step S44 is greater than the preset proportion threshold, it indicates that the spectral characteristics of the candidate component conform to the high-frequency characteristics of arc discharge.

[0092] In some embodiments, under a lightweight computation mode without complex time-frequency joint matching, candidate components that meet the conditions can be directly identified as transient acoustic features of a bouncing arc based on the determination result of the high-frequency energy proportion. For example, if the high-frequency energy proportion of a candidate component is calculated to be 0.6, which is greater than a preset proportion threshold of 0.4, it is directly identified as a transient acoustic feature of a bouncing arc. This operation provides a faster feature identification path, suitable for monitoring scenarios with extremely high real-time requirements and limited computational resources.

[0093] Step S5: Calculate the insulation fault probability based on the transient acoustic characteristics of a bouncing arc. Understandably, the purpose of this step is to convert the extracted transient acoustic features into intuitive insulation fault probability values, and combine them with the chemical features in the gas chamber to achieve a quantitative assessment of the degree of insulation performance degradation of the GIS disconnector.

[0094] Preferably, the step of calculating the insulation fault probability based on the transient acoustic characteristics belonging to a bouncing arc includes: Step S51: Obtain the gas concentration signal in the gas chamber of the GIS disconnect switch, and calculate the difference between the gas concentration signal at the start and end times of the bounce time window as the gas concentration increment signal. Understandably, the bouncing arc discharge causes the sulfur hexafluoride gas in the gas chamber to decompose, generating characteristic decomposition products such as sulfur tetrafluoride and hydrogen fluoride. When the gas concentration signal is acquired by a gas sensor installed in the gas chamber, the concentration values ​​at the start and end times of the bouncing time window are extracted, and the difference between the concentration value at the end time and the concentration value at the start time is calculated to obtain the gas concentration increment signal.

[0095] Furthermore, the gas concentration increment signal reflects the intensity of chemical changes under the influence of an electric arc and is an important physical parameter for assessing the destructive force of the arc. In this application, the gas concentration signal is acquired at a sampling rate of 1 Hz using a miniature electrochemical sensor array installed in the gas chamber.

[0096] Step S52: In response to the gas concentration increment signal being greater than the concentration difference threshold, extract the high-frequency energy proportion of the transient acoustic features belonging to the bouncing arc within the corresponding bouncing time window. Specifically, when the gas concentration increment signal is greater than the concentration difference threshold, it indicates that the arc discharge has caused observable chemical changes, the discharge energy is large, and insulation damage has actually occurred.

[0097] It should be noted that at this time, the high-frequency energy ratio of the transient acoustic features confirmed within the corresponding bounce time window is extracted. The high-frequency energy ratio reflects the electrical intensity of the arc discharge.

[0098] Furthermore, the proportion of high-frequency energy is extracted and used as a key input parameter for subsequent calculation of insulation fault probability. In this application, the concentration difference threshold is set to 0.5 μL per liter, which is obtained based on three times the detection limit of the gas sensor and the standard deviation of normal gas fluctuations.

[0099] Step S53: Based on the product of the gas concentration increment signal and the high-frequency energy ratio, the calculation result is mapped to a preset probability range to generate the insulation fault probability. Furthermore, this application provides a method for generating insulation fault probability by mapping the product calculation result of gas concentration increment signal and high frequency energy ratio to a preset probability interval, in order to solve the problem of inaccurate assessment of a single physical quantity.

[0100] In this embodiment, the destructive power of the electric arc depends not only on the electrical intensity of the discharge but also on its duration and chemical effects. The high-frequency energy percentage characterizes the electrical intensity, while the gas concentration increment signal characterizes the cumulative chemical effect. The product of these two factors reflects the overall threat posed by the electric arc to the insulation.

[0101] In this application, the insulation failure probability can be calculated as follows: This application first calculates the product of the gas concentration increment signal and the high-frequency energy ratio to obtain the comprehensive threat index.

[0102] Furthermore, the comprehensive threat index is input into the piecewise linear mapping function. Wherein: ; In the formula, Indicates the probability of insulation failure; This indicates the gas concentration increment signal, in microliters per liter. Indicates the proportion of high-frequency energy; This indicates the overall threat index, expressed in microliters per liter. This represents the concentration difference threshold, expressed in microliters per liter. This represents the comprehensive threat index corresponding to the lower limit of the preset probability interval, expressed in microliters per liter. This represents the comprehensive threat index corresponding to the upper limit of the preset probability range, expressed in microliters per liter.

[0103] For example, when the gas concentration increment signal It is 2 microliters per liter, with a high-frequency energy ratio of 2 microliters per liter. If the value is 0.5, then the overall threat index is... It is 1 microliter per liter; if It is 0.1 microliters per liter. If the value is 5 microliters per liter, then the probability of insulation failure is... The calculated value is 0.18. The mapping calculation fuses multidimensional features into probability values ​​between 0 and 1, intuitively reflecting the fault risk. This application presets a lower limit for the probability interval. and upper limit The data was obtained through simulation and measurement data of a large number of insulation defects developing into breakdown.

[0104] Step S54: Update the monitoring log according to the insulation fault probability, and associate the insulation fault probability with the gas concentration increment signal and store it in the monitoring log; The operation of updating the monitoring log based on the insulation fault probability achieves data persistence. After each insulation fault probability is calculated, the insulation fault probability value along with the corresponding gas concentration increment signal is written to the monitoring log. Correlated storage ensures that the result of each fault probability assessment can be traced back to the chemical characteristic changes at that time, which facilitates maintenance personnel to review historical data, analyze the insulation condition evolution process, and provide detailed data support for the formulation of maintenance plans.

[0105] Step S55: In response to the gas concentration increment signal being less than or equal to the concentration difference threshold, the insulation fault probability is set to zero and recorded in the monitoring log. When the gas concentration increment signal is less than or equal to the concentration difference threshold, it indicates that the energy of the bouncing arc is extremely low and has not caused gas decomposition; in this case, the insulation is not damaged. To eliminate false alarms caused by interference, the insulation fault probability is forcibly set to zero, and the zero value is recorded in the monitoring log, indicating that this operation did not cause any insulation hazards.

[0106] Preferably, after calculating the insulation fault probability based on the transient acoustic characteristics belonging to a bouncing arc, the method further includes: Step S6: In response to the GIS disconnecting switch performing multiple opening and closing operations, obtain the insulation fault probability corresponding to each opening and closing operation; To analyze the trend of insulation performance over time, probability data needs to be accumulated through multiple operations. Each time a GIS disconnector performs an opening / closing operation and completes the insulation fault probability calculation, the insulation fault probability value is bound to the operation timestamp and stored in a historical database. By traversing the historical database, the insulation fault probability sequence corresponding to each opening / closing operation, arranged in chronological order, is extracted and used as input for trend analysis.

[0107] Step S7: Based on the time sequence of each opening and closing operation, calculate the difference between the probabilities of two adjacent insulation faults, and extract the fault growth nodes where the difference is not zero. The first-order difference is calculated from the probability sequence of insulation faults, which is the probability of the subsequent operation minus the probability of the previous operation, resulting in a sequence of differences between the probabilities of two adjacent insulation faults. When the difference is not zero, it indicates that the insulation state has changed. The time points corresponding to the differences are extracted and defined as fault growth nodes. Fault growth nodes exclude stable periods where the probabilities do not change, focusing on the moments when insulation degradation abruptly changes.

[0108] Step S8: Calculate the slope of the change in insulation failure probability with the number of operations based on the difference of all insulation failure probabilities, and obtain the positive or negative attribute of the slope. Understandably, this is achieved by calculating the slope of the insulation failure probability at the extracted failure growth nodes using linear regression. To quantify the trend, a straight line is fitted using the least squares method with the number of operations as the independent variable and the insulation failure probability as the dependent variable. The slope of the resulting line reflects the rate of insulation degradation. In this application, the positive or negative attribute of the slope determines the direction of insulation condition development. If the slope is positive, it indicates that the probability of insulation failure increases with the number of operations, and the insulation performance is deteriorating. If the slope is negative, it indicates a decrease in the probability of insulation failure, possibly due to contact wear-in leading to arc weakening. Obtaining the positive or negative attribute of the slope provides a trend criterion for subsequent early warning decisions. In this application, the goodness-of-fit threshold for the aforementioned linear regression is set to 0.7. This goodness-of-fit threshold is obtained through statistical methods to ensure the reliability of the probability change trend.

[0109] Step S9: In response to the slope being positive and the slope being greater than the slope threshold, an early warning signal indicating insulation performance degradation is output. When the slope reflects an accelerated deterioration trend in insulation performance, a graded warning needs to be issued according to different severity levels. This step covers the complete logical link from probability interval determination to warning level classification and maintenance suggestion push.

[0110] Preferably, the step of outputting a warning signal indicating insulation performance degradation in response to a slope that is positive and greater than a slope threshold includes: Step S91: Obtain the feature combination consisting of insulation fault probability and slope, and extract the interval where the insulation fault probability is located in the feature combination; The latest calculated insulation fault probability is combined with its corresponding slope to form a feature combination. To differentiate the severity of insulation degradation, the range of insulation fault probabilities is divided into three intervals. The insulation fault probabilities in the feature combinations are extracted, and it is determined which probability interval they fall into. The division of probability intervals provides a quantitative criterion for subsequent graded early warning systems.

[0111] Step S92: In response to the insulation fault probability being in the first probability interval and the slope being greater than the slope threshold, a first-level warning signal indicating mild insulation degradation is output. When the insulation failure probability is within the first probability interval and the slope is greater than the slope threshold, it indicates that although the insulation has shown a deterioration trend, the current failure probability is still at a low level. At this time, a first-level warning signal indicating slight insulation deterioration is output to remind maintenance personnel to pay closer attention. In this application, the first probability interval is set to a range greater than 0.05 and less than 0.3, and the slope threshold is set to 0.02 per operation. The interval and threshold mentioned here are obtained through statistical fitting of the probability data of slight defects in the equipment.

[0112] Step S93: In response to the insulation fault probability being in the second probability interval and the slope being greater than the slope threshold, a second-level warning signal indicating severe insulation degradation is output, wherein the second probability interval is greater than the first probability interval. When the insulation failure probability is within the second probability interval and the slope is greater than the slope threshold, it indicates that the insulation degradation has reached a high level and the risk of breakdown is significant. At this point, a second-level warning signal indicating severe insulation degradation is output, requiring immediate power outage maintenance. The second probability interval is set to a range greater than or equal to 0.3 and less than 1.

[0113] Step S94: Generate maintenance suggestion information based on the first-level warning signal or the second-level warning signal, and push the maintenance suggestion information to the monitoring terminal; For different levels of early warning signals, corresponding maintenance recommendations are generated. For level one early warning signals, recommendations to strengthen patrols and live-line testing are generated; for level two early warning signals, recommendations to de-energize, disassemble, inspect, and replace contacts are generated. These maintenance recommendations are then pushed to the monitoring terminal via the communication network to guide on-site maintenance personnel in taking targeted measures, achieving closed-loop management.

[0114] Step S95: In response to the insulation fault probability being in the third probability interval, the warning signal output is blocked, wherein the third probability interval is lower than the first probability interval; When the probability of insulation failure falls within the third probability interval (which includes zero and extremely small probabilities), it indicates good insulation condition and negligible arcing effects. In this case, the warning signal output is disabled to avoid interfering with maintenance personnel and maintain the system's stable operation. The third probability interval is set to a range greater than or equal to 0 and less than or equal to 0.05.

[0115] Step S10: In response to the slope being negative or the slope being less than the slope threshold, maintain the current monitoring state and continue to acquire the insulation fault probability for the next opening and closing operation.

[0116] When the slope is negative, it indicates that the probability of insulation failure decreases with the number of operations; or when the slope is less than the slope threshold, it indicates that the insulation degradation rate is extremely slow. Both cases indicate that the current insulation condition is stable and there is no trend of accelerated deterioration. In this case, the current monitoring status is maintained, no warning is output, and the insulation failure probability of the next opening and closing operation is continuously acquired to continuously track the equipment status.

[0117] Figure 2 A logical flow of a dynamic monitoring method for the mechanical condition and insulation performance of a GIS disconnector switch is disclosed. First, in response to operation commands, the displacement of the crank arm and the acoustic signature of the casing are acquired synchronously, and impact and bounce time windows are constructed based on displacement extreme points. Simultaneously, a mechanical impact trend curve is fitted to the acoustic signature signal and discarded to generate residual acoustic signature signals. After integrity verification, independent component separation is performed. The separation process adaptively adjusts the iteration step size based on the signal-to-noise ratio to extract candidate components with overlapping time periods and high-frequency proportions exceeding a threshold. Further cross-correlation calculations confirm the transient acoustic characteristics of the bounce arc. Finally, the insulation fault probability is calculated by combining the gas concentration increment in the gas chamber, and the slope of each probability change determines whether to output a graded early warning signal, achieving accurate identification and early warning of early insulation defects.

[0118] Figure 3 This paper demonstrates the logical flow of signal processing and feature extraction in a dynamic monitoring method for the mechanical condition and insulation performance of GIS disconnect switches. First, the system responds to operation commands by acquiring crank arm displacement and casing acoustic signature signals, thereby dividing the action interval and constructing impact and bounce time windows. Then, within the impact time window, a mechanical impact trend curve is fitted, generating a residual acoustic signature signal. Whether to refit or output the signal is determined by whether the completeness of mechanical impact interference removal exceeds a preset threshold. The output residual acoustic signature signal undergoes independent component separation to generate candidate components. Further, by determining whether the proportion of high-frequency energy exceeds a preset proportion threshold, the transient acoustic characteristics belonging to a bounce arc are finally confirmed, thus achieving the extraction of weak arc signals from strong vibration interference.

[0119] Figure 4 The system demonstrates the logical flow of insulation fault probability calculation and trend early warning. First, it acquires the gas concentration increment signal and determines if it exceeds a concentration difference threshold. If so, it generates an insulation fault probability based on the product of the concentration increment and the proportion of high-frequency energy; otherwise, it sets the probability to zero. Next, it calculates the slope of the insulation fault probability change based on previous operations and determines if the slope is positive and greater than a threshold. If not, it maintains the current monitoring state; if so, it further determines the probability range. If it falls within the first probability range, a mild warning is output; if it falls within the second probability range, a severe warning is output; otherwise, the warning output is disabled, thus achieving tiered early warning and accurate assessment of insulation status.

[0120] This embodiment also provides a dynamic monitoring system for the mechanical condition and insulation performance of a GIS disconnector, including: a signal synchronization acquisition module, configured to acquire crank arm displacement signal and housing acoustic fingerprint signal in response to the GIS disconnector executing the opening and closing operation command; The timing segmentation module is configured to divide the crank arm displacement signal and the housing acoustic pattern signal into a contact impact interval and a bounce monitoring interval based on the extreme points in the crank arm displacement signal, and to construct an impact time window and a bounce time window based on the contact impact interval and the bounce monitoring interval. The interference removal module is configured to fit the mechanical impact trend curve based on the shell acoustic signature signal within the impact time window, and subtract the amplitude value of the mechanical impact trend curve at the corresponding moment from the shell acoustic signature signal within the impact time window and the bounce time window to obtain the residual acoustic signature signal. The feature extraction module is configured to perform independent component separation on the residual acoustic signature signal to obtain transient acoustic features belonging to the bouncing arc; The fusion judgment module is configured to calculate the probability of insulation failure based on the transient acoustic characteristics of a bouncing arc.

[0121] The above-mentioned unit modules can be embedded in the processor of the electronic device in hardware form or independent of it, or they can be stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0122] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0123] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for dynamic monitoring of the mechanical condition and insulation performance of a GIS disconnector, characterized in that, include: In response to the GIS disconnector switch executing the opening and closing operation command, the crank arm displacement signal and the housing acoustic signal are acquired; Based on the extreme points in the crank arm displacement signal, the crank arm displacement signal and the shell acoustic pattern signal are divided into the contact impact interval and the bounce monitoring interval, and the impact time window and the bounce time window are constructed based on the contact impact interval and the bounce monitoring interval. Based on the shell acoustic signature signal fitting the mechanical impact trend curve within the impact time window, the amplitude value of the mechanical impact trend curve at the corresponding moment is subtracted from the shell acoustic signature signal within the impact time window and the bounce time window to obtain the residual acoustic signature signal. Independent component separation is performed on the residual acoustic signature signal to obtain the transient acoustic features belonging to the bouncing arc; The probability of insulation failure is calculated based on the transient acoustic characteristics of a bouncing arc.

2. The method for dynamic monitoring of the mechanical condition and insulation performance of GIS disconnect switches as described in claim 1, characterized in that: The step of dividing the crank arm displacement signal and the housing acoustic signature signal into a contact impact interval and a bounce monitoring interval based on the extreme points in the crank arm displacement signal, and constructing an impact time window and a bounce time window based on the contact impact interval and the bounce monitoring interval, includes: The time interval between two adjacent extreme points in the crank arm displacement signal is used as the basic motion interval. If the slope change of the crank arm displacement signal within the basic action range exceeds the preset slope threshold, the basic action range is determined as the contact impact range. The adjacent basic motion interval after the contact impact interval is defined as the bounce monitoring interval, and the displacement change within the bounce monitoring interval is extracted. Based on the contact impact interval and the bounce monitoring interval, an impact time window and a bounce time window are constructed, and the displacement change within the bounce monitoring interval is marked as the mechanical rebound displacement characteristic. Based on the characteristics of mechanical rebound displacement and the extreme values ​​of displacement within the contact impact range, an action state sequence is generated.

3. The method for dynamic monitoring of the mechanical condition and insulation performance of GIS disconnect switches as described in claim 1, characterized in that: The method involves fitting the mechanical impact trend curve to the shell acoustic signature signal within the impact time window, and subtracting the amplitude value of the mechanical impact trend curve at the corresponding moment from the shell acoustic signature signal within the impact time window and the bounce time window to obtain the residual acoustic signature signal, including: Obtain the initial time corresponding to the amplitude extreme point of the acoustic signature signal of the inner and outer shells within the impact time window, and record the absolute amplitude of the amplitude extreme point; Based on the envelope trend of the decrease in the amplitude of the shell acoustic signature signal after the initial moment, the corresponding amplitude decrease trend curve of mechanical impact is fitted and generated as the mechanical impact trend curve. Subtract the amplitude values ​​of the shell acoustic signature signal and the mechanical impact trend curve at the corresponding moments within the impact time window and bounce time window to generate residual acoustic signature signal. The completeness of mechanical impact interference removal is determined based on the amplitude distribution characteristics of the residual acoustic signature signal, and the completeness of mechanical impact interference removal is compared with the preset completeness threshold. In response to mechanical impact interference, if the integrity of the rejection exceeds a preset integrity threshold, the residual acoustic signature signal is output to the subsequent separation operation.

4. The method for dynamic monitoring of the mechanical condition and insulation performance of GIS disconnect switches as described in claim 2, characterized in that: The step of performing independent component separation on the residual acoustic signature signal to obtain transient acoustic features belonging to the bouncing arc includes: The occurrence time and duration of mechanical rebound displacement characteristics are determined based on the crank arm displacement signal within the bounce time window; Perform independent component separation on the residual acoustic signature signal within the impact time window and bounce time window to generate multiple independent acoustic components; Based on the occurrence time and duration of the mechanical rebound displacement characteristics, candidate components with overlapping time distributions are selected from multiple independent acoustic components. Extract the high-frequency energy percentage from the candidate components and compare the high-frequency energy percentage with a preset percentage threshold; When the proportion of high-frequency energy is greater than a preset proportion threshold, the candidate component is identified as a transient acoustic feature belonging to the bouncing arc.

5. The method for dynamic monitoring of the mechanical condition and insulation performance of GIS disconnect switches as described in claim 2, characterized in that: The calculation of insulation fault probability based on the transient acoustic characteristics of bouncing arcs includes: Obtain the gas concentration signal in the gas chamber of the GIS disconnect switch, and calculate the difference between the gas concentration signal at the start and end of the bounce time window as the gas concentration increment signal. In response to a gas concentration increment signal exceeding a concentration difference threshold, the proportion of high-frequency energy belonging to the transient acoustic features of the bouncing arc within the corresponding bouncing time window is extracted. The product of the gas concentration increment signal and the high-frequency energy ratio is mapped to a preset probability range to generate the insulation fault probability. The monitoring log is updated based on the insulation fault probability, and the insulation fault probability is correlated with the gas concentration increment signal and stored in the monitoring log; In response to a gas concentration increment signal being less than or equal to a concentration difference threshold, the insulation fault probability is set to zero and recorded in the monitoring log.

6. The method for dynamic monitoring of the mechanical condition and insulation performance of GIS disconnect switches as described in claim 1, characterized in that: After calculating the insulation fault probability based on the transient acoustic characteristics belonging to bouncing arcs, the method further includes: In response to the GIS disconnector performing multiple opening and closing operations, the probability of insulation faults corresponding to each opening and closing operation is obtained. Based on the time sequence of each opening and closing operation, the difference between the probabilities of two adjacent insulation faults is calculated, and fault growth nodes with non-zero differences are extracted. Calculate the slope of the change in insulation failure probability with the number of operations based on the difference of all insulation failure probabilities, and obtain the positive or negative attribute of the slope. In response to the slope being positive and the slope being greater than the slope threshold, an early warning signal indicating insulation performance degradation is output. In response to the slope being negative or the slope being less than the slope threshold, the current monitoring status is maintained and the insulation fault probability for the next opening and closing operation is acquired.

7. The method for dynamic monitoring of the mechanical condition and insulation performance of GIS disconnect switches as described in claim 3, characterized in that: The independent component separation operation on the residual acoustic signature signal includes: The background acoustic signature segment before the impact time window is obtained as the noise reference, and the signal-to-noise ratio of the residual acoustic signature signal within the bounce time window relative to the noise reference is calculated. In response to the signal-to-noise ratio attribute value being in the low signal-to-noise ratio range, the iteration step size of the independent component separation operation is reduced to the preset minimum step size to limit the mechanical residual vibration noise from being separated into independent components. In response to the signal-to-noise ratio attribute value being in the high signal-to-noise ratio range, the iteration step size of the independent component separation operation is increased to the preset maximum step size to accelerate the convergence of the bouncing arc component; Based on the reduced or increased iteration step size, perform iterative separation operation on the residual acoustic signature signal within the bounce time window until the mutual information of the separated independent components is lower than the mutual information threshold. Independent components that satisfy the condition that the mutual information is below the mutual information threshold are marked to generate independent acoustic components.

8. The method for dynamic monitoring of the mechanical condition and insulation performance of GIS disconnect switches as described in claim 6, characterized in that: The response to a slope attribute being positive and the slope being greater than a slope threshold, outputting a warning signal indicating insulation performance degradation, includes: Obtain the feature combination consisting of insulation fault probability and slope, and extract the interval where the insulation fault probability is located in the feature combination; In response to the insulation failure probability being in the first probability range and the slope being greater than the slope threshold, a first-level warning signal indicating mild insulation degradation is output. In response to the insulation fault probability being in the second probability interval and the slope being greater than the slope threshold, a second-level warning signal indicating severe insulation degradation is output, wherein the second probability interval is greater than the first probability interval; Based on the first-level or second-level warning signal, maintenance suggestion information is generated and pushed to the monitoring terminal; In response to the insulation fault probability being in the third probability range, the warning signal output is shielded, where the third probability range is lower than the first probability range.

9. The method for dynamic monitoring of the mechanical condition and insulation performance of a GIS disconnector as described in claim 4, characterized in that: After extracting the high-frequency energy percentage of the candidate components and comparing the high-frequency energy percentage with a preset percentage threshold, the method further includes: Obtain the energy distribution ratio of the candidate components in adjacent frequency bands, and combine the energy distribution ratio with the time-domain amplitude variation curve of the candidate components; The combined results of the energy distribution ratio and the time-domain amplitude variation curve are cross-correlated with the preset arc acoustic distribution ratio to generate cross-correlation coefficients. When the cross-correlation coefficient is greater than the correlation coefficient threshold, the candidate component is identified as a transient acoustic feature belonging to the bouncing arc. Based on the occurrence time of the transient acoustic characteristics belonging to the bouncing arc, the corresponding displacement deviation node is located in the crank arm displacement signal; Arc positioning information is generated based on displacement deviation nodes, and the arc positioning information is bound and stored with transient acoustic features belonging to bouncing arcs.

10. A dynamic monitoring system for the mechanical condition and insulation performance of a GIS disconnector, employing the dynamic monitoring method for the mechanical condition and insulation performance of a GIS disconnector as described in any one of claims 1 to 9, characterized in that, include: The signal synchronization acquisition module is configured to acquire crank arm displacement signal and housing acoustic signal in response to the GIS disconnecting switch executing the opening and closing operation command; The timing segmentation module is configured to divide the crank arm displacement signal and the housing acoustic pattern signal into a contact impact interval and a bounce monitoring interval based on the extreme points in the crank arm displacement signal, and to construct an impact time window and a bounce time window based on the contact impact interval and the bounce monitoring interval. The interference removal module is configured to fit the mechanical impact trend curve based on the shell acoustic signature signal within the impact time window, and subtract the amplitude value of the mechanical impact trend curve at the corresponding moment from the shell acoustic signature signal within the impact time window and the bounce time window to obtain the residual acoustic signature signal. The feature extraction module is configured to perform independent component separation on the residual acoustic signature signal to obtain transient acoustic features belonging to the bouncing arc; The fusion judgment module is configured to calculate the probability of insulation failure based on the transient acoustic characteristics of a bouncing arc.

Images