Full-period monitoring method for gas insulated switchgear by using multi-source data fusion

Abstract

The invention relates to the technical field of partial discharge detection, in particular to a gas insulated switchgear full-period monitoring method using multi-source data fusion, and the method comprises the steps: obtaining an ultrahigh frequency signal and a power frequency voltage signal of each monitoring period during the operation of a gas insulated switchgear in real time; calculating the de-noising energy degree of each moment; acquiring all initial discharge moments of the ultrahigh-frequency signal by using the variation trend of the de-noising energy degrees at different moments; extracting a phase angle of each discharge initial moment in the power frequency voltage signal to obtain a PRPD atlas; the discharge density is calculated, the distribution deviation degree of each monitoring period is calculated, and the discharge difference degree of each monitoring period is determined; and the characteristic vector is obtained, and the partial discharge of the gas insulated switchgear is detected by using a classification algorithm. According to the invention, the partial discharge defect of the gas insulated switchgear can be identified more accurately, so that the potential fault hidden danger of equipment can be found in time.

Description

Technical Field

[0001] This application relates to the field of partial discharge detection technology, specifically to a method for full-cycle monitoring of gas-insulated switchgear using multi-source data fusion. Background Technology

[0002] Gas-insulated switchgear is a key piece of equipment in the power system, and its operational reliability is directly related to the safety and stability of the power grid. Due to metal burrs on the metal casing or internal surface, metal particles generated during operation, contamination on the surface of the insulator, and loose internal equipment parts, partial discharge may occur in gas-insulated switchgear during operation. The partial discharge activity inside may lead to a decrease in insulation performance. If not detected in time, it may cause equipment failure or even power grid accidents.

[0003] The ultra-high frequency detection method uses phase-resolved partial discharge maps for pattern recognition to determine the type of partial discharge fault in gas-insulated switchgear. However, under strong noise background interference, it cannot accurately separate the real partial discharge pulse from the strong electromagnetic noise background. Secondly, the morphological features of the map are complex and varied, making it difficult to fully capture the multi-dimensional information contained in the map and effectively quantify the essential differences between different types of partial discharge defects. This results in low accuracy of the classification model in identifying the type of partial discharge defect in gas-insulated switchgear. Summary of the Invention

[0004] To address the aforementioned technical issues, a full-cycle monitoring method for gas-insulated switchgear using multi-source data fusion is provided to resolve existing problems.

[0005] The solution to the technical problem presented in this application is to provide a method for full-cycle monitoring of gas-insulated switchgear using multi-source data fusion, comprising the following steps: Real-time acquisition of UHF and power frequency voltage signals during each monitoring cycle of gas-insulated switchgear operation; The method analyzes the cumulative energy deviation of the signal amplitude at different times in the UHF signal and calculates the denoising energy level at each time. Using the changing trend of the denoising energy level at different times, it obtains all initial discharge times of the UHF signal. It extracts the phase angle of each initial discharge time in the power frequency voltage signal and, combined with the signal amplitude of the UHF signal, obtains the partial discharge distribution (PRPD) spectrum. It analyzes the density of discharge activity in the neighborhood of each data point in the PRPD spectrum, calculates the discharge density, and, considering the high frequency characteristics and distribution skewness of the discharge density, calculates the distribution deviation for each monitoring period. Combining the distribution range and concentration of data points under different discharge intensities in the PRPD spectrum, it determines the discharge difference for each monitoring period. Based on the steepness of the discharge intensity distribution at different phase angles in the PRPD spectrum, as well as the discharge density and the aforementioned skewness, and combining the distribution deviation and discharge difference, it obtains feature vectors and uses a classification algorithm to detect partial discharge in gas-insulated switchgear.

[0006] Preferably, the calculation of the denoised energy at each time point includes: Calculate the square of the signal amplitude at each moment in the UHF signal, and denote it as the instantaneous energy; calculate the mean of the instantaneous energy at all moments in the UHF signal, and denote it as the average energy. Calculate the product of the time sequence and the average energy at any given moment in the UHF signal, and use it as the reference cumulative energy at that given moment; The denoising energy level is the sum of the differences between the instantaneous energy at any given time and all previous times and the baseline cumulative energy at any given time.

[0007] Preferably, acquiring all initial discharge moments of the UHF signal includes: Obtain the minimum value of the denoised energy at all times in the UHF signal; take the time corresponding to the first minimum value as the initial time of the first discharge; take the last time in the preset time after the initial time of the first discharge as the segmentation point; remove the signal amplitude of all times before the segmentation point from the UHF signal as the newly segmented UHF signal; for the newly segmented UHF signal, recalculate the denoised energy at all times in the newly segmented UHF signal, and obtain the time corresponding to the first minimum value as the initial time of the second discharge, and so on, to obtain all the initial times of discharge of the UHF signal.

[0008] Preferably, the process of obtaining the PRPD map is as follows: Each initial discharge moment corresponds to one discharge event. The maximum signal amplitude in the UHF signal after each initial discharge moment within a preset time is obtained as the discharge amount of one discharge event. Based on the discharge amount and phase angle of all discharge events in each monitoring period, a PRPD spectrum is plotted. In the PRPD spectrum, the horizontal axis represents the phase, the vertical axis represents the discharge amount, and the color of the data point represents the number of discharges under a specific phase angle and discharge amount.

[0009] Preferably, the discharge density is the sum of the number of discharges corresponding to all data points in the neighborhood of each data point in the PRPD spectrum.

[0010] Preferably, the degree of skewness is the skewness of the discharge density of all data points within the PRPD spectrum.

[0011] Preferably, the distribution deviation is the product of the maximum discharge density of all data points in the PRPD spectrum and the degree of skewness.

[0012] Preferably, determining the discharge difference in each monitoring cycle includes: Divide the vertical axis of the PRPD graph into multiple intervals; count the number of all data points in the PRPD graph that fall within each interval. Arrange the quantities of all interval ranges in ascending order of their corresponding discharge amounts to form a quantity distribution column; Find the maximum length of consecutive non-zero elements in a statistical distribution. Calculate the sum of the position corresponding to the maximum value in the quantity distribution column and the maximum length; The discharge difference is the product of the sum and the distribution deviation.

[0013] Preferably, the method for measuring the steepness is as follows: Select the maximum discharge quantity among all data points at each phase angle in the PRPD spectrum; calculate the kurtosis of the maximum discharge quantity at all phase angles as the steepness.

[0014] Preferably, the process of obtaining the feature vector is as follows: Obtain the maximum and minimum discharge densities of all data points in the PRPD map; combine the maximum and minimum discharge densities, skewness, distribution deviation, discharge difference, and steepness of each monitoring period into a feature vector.

[0015] This application has at least the following beneficial effects: This application calculates the denoising energy level at each moment by analyzing the cumulative energy deviation of the signal amplitude in UHF signals. Its advantages lie in utilizing the difference between the stability of noise energy and the suddenness of discharge energy to highlight the true discharge pulse from the stable noise, improving the sensitivity and anti-interference capability for identifying weak discharge signals from strong noise, thus enabling more accurate identification of real discharge activity. Obtaining all initial moments of discharge in the UHF signal, its advantage lies in solving the problem of difficulty in separating multiple consecutive discharge events through an iterative signal segmentation strategy, ensuring that each discharge pulse can be captured independently and completely, avoiding misjudgments caused by pulse overlap. Calculating the discharge density, its advantage lies in assessing the concentrated area of ​​discharge activity. Calculating the distribution deviation of each monitoring cycle, its advantage lies in... This study reflects the asymmetry in the distribution of discharge activity in the PRPD spectrum and the density of discharges; it determines the discharge difference in each monitoring period, and its beneficial effect lies in the coupling characteristics of discharge activity in phase and amplitude distribution, assessing whether the discharge is concentrated in phase and whether it is both high and concentrated in amplitude, thus achieving information coupling in two dimensions: phase and amplitude; it obtains feature vectors and uses classification algorithms to detect partial discharges in gas-insulated switchgear, and its beneficial effect lies in comprehensively evaluating the distribution pattern of the PRPD spectrum through multiple dimensions of information, extracting complex and implicit PRPD spectrum morphological information and using classification algorithms for identification, which can more accurately identify partial discharge defects in gas-insulated switchgear, thereby timely discovering potential equipment faults and ensuring the stability and safety of power grid operation. Attached Figure Description

[0016] The following section provides a more detailed description of the full-cycle monitoring method for gas-insulated switchgear using multi-source data fusion, in conjunction with the accompanying drawings.

[0017] Figure 1 A flowchart illustrating the steps of the full-cycle monitoring method for gas-insulated switchgear using multi-source data fusion provided in this application embodiment; Figure 2 A flowchart illustrating the steps of the method for obtaining discharge difference provided in this application embodiment. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description, in conjunction with the accompanying drawings and embodiments, provides a more comprehensive overview of the full-cycle monitoring method for gas-insulated switchgear utilizing multi-source data fusion proposed in this application. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0020] Please see Figure 1 The diagram illustrates a flowchart of a method for full-cycle monitoring of gas-insulated switchgear using multi-source data fusion, according to an embodiment of this application. The method includes the following steps: Step 1: Real-time acquisition of UHF signals and power frequency voltage signals during each monitoring cycle of the gas-insulated switchgear.

[0021] Gas-insulated substations (GIS) are enclosed electrical devices that house multiple electrical components within a metal casing filled with SF6 inert gas. As key equipment in substations, their performance and quality directly impact the reliability of the power grid. GIS faults are primarily caused by insulation problems, with partial discharge being a significant reason for insulation degradation. Partial discharge, also known as electrical discharge, occurs when only a portion of the insulation between conductors is broken down under the influence of an electric field. This is mainly due to uneven electric field distribution caused by external voltage, leading to small-scale charge accumulation and release within the insulating components, resulting in localized gas discharge or breakdown of the insulating medium.

[0022] Partial discharge can have a series of physical and chemical effects on the interior and surrounding space of GIS equipment. These include the sparks generated by the discharge, the sounds caused by the discharge breaking down the air, the changes in electrical state due to changes in the electric field, and the resulting mechanical vibrations. Furthermore, cavitation or bubbles formed during the manufacturing process of insulating components, cracks caused by thermal expansion and contraction during operation, sharp burrs and scratches on the surface of high-voltage conductors, improperly connected conductors and freely moving metal particles during operation, dust particles mixed in, and loose installation components. All of these can trigger partial discharge phenomena and seriously endanger the safe operation of the equipment.

[0023] Currently, commonly used partial discharge detection methods include ultra-high frequency (UHF) method, pulse current method, and ultrasonic detection method. Among them, the principle of UHF method for partial discharge detection is as follows: when partial discharge occurs inside the GIS, it will emit electromagnetic waves to the surroundings. The detection is performed by capturing the electromagnetic waves released by the GIS equipment. UHF method has extremely high sensitivity and anti-electromagnetic interference capability, which can achieve accurate location of the fault source.

[0024] Based on the above analysis, an ultra-high frequency sensor is built into the gas-insulated switchgear to detect the partial discharge signal inside the gas-insulated switchgear in real time, and the ultra-high frequency signal of each monitoring cycle is recorded in real time by a high sampling digital oscilloscope. The operating voltage of the gas-insulated switchgear is collected synchronously using a voltage transformer, and the power frequency voltage signal of each monitoring cycle is recorded in real time using a high-sampling digital oscilloscope. In this embodiment, the sampling frequency of the high-sampling digital oscilloscope is 20GS / s, and the duration of the monitoring period is 10s. As for other implementation methods, the implementer can set them according to the actual situation.

[0025] Thus, the ultra-high frequency signal and power frequency voltage signal for each monitoring cycle are obtained.

[0026] Step 2: Analyze the deviation of the cumulative energy of the signal amplitude at each moment in the UHF signal, and calculate the denoising energy at each moment; use the changing trend of the denoising energy at different moments to obtain the initial moment of all discharges of the UHF signal.

[0027] Electromagnetic interference signals present in the operating environment of gas-insulated switchgear generate noise. This noise typically has low energy, but when partial discharge occurs, the signal energy generated by the partial discharge is usually much higher than the background noise energy, and it increases sharply at the moment of discharge. Therefore, since the partial discharge energy is much higher than the noise energy, the denoising energy is calculated by analyzing the deviation of the instantaneous energy of the signal amplitude at different times in the UHF signal. Specifically: Calculate the square of the signal amplitude at each moment in the UHF signal, and denot it as the instantaneous energy; Calculate the mean of the instantaneous energy at all times in the UHF signal, and use it as the average energy. Calculate the product of the time sequence and the average energy at any given moment in the UHF signal, and use it as the reference cumulative energy at that given moment; The sum of the differences between the instantaneous energy at any given moment and all moments prior to the UHF signal and the reference cumulative energy at any given moment is taken as the denoising energy level at each moment in the UHF signal. In this embodiment, the UHF signal of the first The formula for calculating the denoised energy at time step is:

[0028] in, For the UHF signal, the first The noise reduction energy level at any given moment. For the UHF signal, the first The signal amplitude at time [time]. Indicates instantaneous energy. For the first The timing sequence corresponding to each moment This represents the average energy.

[0029] It should be noted that the aforementioned cumulative energy reflects the energy accumulated at the time of... The cumulative energy expected at the average energy level at any given time will be far greater than the energy generated by noise after partial discharge of the gas-insulated switchgear. This will cause a sharp increase in the energy intensity at the time of discharge. The denoising energy level reflects the difference between the actual observed energy accumulation and the expected energy accumulation in the UHF signal. The larger the denoising energy level, the higher the signal amplitude at that time, and the more likely that the signal is in a partial discharge state at that time.

[0030] Furthermore, during non-discharge periods, the denoising energy level typically shows a decreasing trend. However, if a partial discharge occurs, the denoising energy level will suddenly jump at the discharge moment. Therefore, each discharge will produce a distinct trough in the denoising energy level, i.e., a minimum point, which corresponds to the start time of the discharge. But since the denoising energy level is a cumulative change, the denoising energy level at subsequent moments will accumulate the energy from the previous discharge, causing the subsequently calculated denoising energy level to show a decreasing trend, making it impossible to identify other partial discharge events.

[0031] Based on this analysis, by detecting the extreme points of the denoised energy at all times, the initial moment of each discharge is identified. After removing signals from the discharge process, iterative analysis is performed to obtain the start time of all discharges, specifically: Obtain the minimum point of the denoising energy at all times in the UHF signal; take the time corresponding to the first minimum point as the initial time of the first discharge; In this embodiment, the AMPD algorithm (Automatic multiscale-based peak detection) is used to obtain the minimum point. The AMPD algorithm is a well-known technology and will not be described in detail here.

[0032] The last moment within the preset time after the initial moment of the first discharge is used as the dividing point; In this embodiment, since the duration of a single partial discharge is usually less than 1µs, the preset duration is set to 1µs. As for other implementation methods, the implementer can set it according to the actual situation.

[0033] Remove the signal amplitude values ​​from all times before the segmentation point from the UHF signal to obtain the newly segmented UHF signal; For the newly cut UHF signal, the denoising energy of all moments in the newly cut UHF signal is recalculated, and the moment corresponding to the first minimum point is obtained as the initial moment of the second discharge. This process continues until the minimum point can no longer be found, at which point the iteration ends, thus obtaining all the initial discharge moments of the UHF signal. Thus, all the initial discharge moments of the UHF signal have been obtained.

[0034] Step 3: Extract the phase angle of each discharge initial moment in the power frequency voltage signal, and combine it with the signal amplitude of the UHF signal to obtain the PRPD spectrum; analyze the density of discharge activity of each data point in the PRPD spectrum in the neighborhood, calculate the discharge density, and calculate the distribution deviation of each monitoring period by combining the high frequency of the discharge density and the degree of distribution skewness.

[0035] Furthermore, the Phase Resolved Partial Discharge (PRPD) analysis mode is the most widely used. It identifies the types of partial discharge defects within a GIS by measuring and statistically analyzing the distribution characteristics of discharge counts and discharge signal amplitudes within the periodic phase interval, and then plotting PRPD maps. Specifically: Extract the phase angle at each discharge initial moment from the power frequency voltage signal; In this embodiment, the phase angle is extracted using the zero-crossing detection method, which is a well-known technique and will not be described in detail here.

[0036] Each initial discharge moment corresponds to one discharge event; The maximum signal amplitude in the UHF signal after each discharge initial moment within a preset time period is obtained as the discharge amount of a single discharge event. Based on the discharge amount and phase angle of all discharge events in each monitoring period, a three-dimensional PRPD map is plotted. In the PRPD map, the X-axis represents the phase, the Y-axis represents the discharge amount, and the Z-axis represents the number of discharges at a specific phase angle and discharge amount. It should be noted that the creation of PRPD maps is a well-known technique and will not be elaborated upon here.

[0037] Secondly, partial discharge defects include various types such as tip discharge, free particle discharge, and suspended electrode discharge. Different types of partial discharge defects exhibit different characteristics in PRPD patterns. For example, tip discharge typically presents symmetrical discharge clusters near the peak phases of the positive and negative half-cycles, with a large number of discharges, high amplitude, and low dispersion, and uniform discharge time intervals, often occurring near the voltage peak. Free particle discharge has a more dispersed partial discharge pulse phase distribution, fewer discharges, average discharge quantity, and is distributed throughout the entire phase, with unstable discharge time intervals, and the PRPD pattern may show randomly distributed discharge points. Suspended electrode discharge is basically symmetrical in the positive and negative half-cycles and has a relatively high discharge amplitude, with relatively stable discharge.

[0038] Based on the above analysis, by analyzing the concentration and distribution trend of partial discharge activity in voltage phase and amplitude, the distribution deviation is calculated, specifically as follows: The projection of the three-dimensional PRPD pattern onto the XOY plane represents the two-dimensional PRPD pattern. In the two-dimensional PRPD pattern, the X-axis represents the phase, the Y-axis represents the discharge quantity, and the color of each data point represents the number of discharges. The sum of the number of discharges corresponding to all data points in the neighborhood of each data point in the two-dimensional PRPD map is used as the discharge density. In this embodiment, the neighborhood radius is 5, and the sum of the values ​​on the Z-axis corresponding to all data points distributed in the neighborhood of each data point is used as the discharge density.

[0039] Calculate the skewness of the discharge density for all data points; It should be noted that the calculation of skewness is a well-known technique and will not be elaborated upon here.

[0040] It should be noted that the discharge density of all data points is sorted in ascending order according to phase angle. Among them, the discharge density of the same phase angle is sorted in ascending order according to signal amplitude before the deviation is calculated.

[0041] It should be noted that the skewness reflects the degree of skewness of the PRPD spectrum to assess the bias in the phase-amplitude distribution of discharge activity.

[0042] The product of the maximum discharge density of all data points in the two-dimensional PRPD spectrum and the aforementioned skewness is used as the distribution deviation for each monitoring period. It should be noted that a higher discharge density indicates more discharges occurring at this phase angle and signal amplitude, reflecting a more concentrated distribution of partial discharge activity at this voltage phase and amplitude. A higher skewness indicates a less asymmetrical distribution of discharge density in the phase direction, while a lower skewness indicates a symmetrical distribution of discharge density in the phase direction. A higher maximum discharge density indicates highly concentrated discharge activity during the monitoring period. The distribution deviation reflects the distribution of partial discharge activity; a higher deviation indicates a more concentrated and asymmetrical distribution of partial discharge processes in the PRPD spectrum during the monitoring period, and a higher discharge intensity.

[0043] Thus, the distribution deviation for each monitoring period is obtained.

[0044] Step 4: By analyzing the distribution range and concentration of data points at different discharge intensities in the PRPD spectrum, and combining the distribution deviation, determine the discharge difference for each monitoring period; based on the steepness of the discharge intensity distribution at different phase angles in the PRPD spectrum, as well as the discharge density and the aforementioned skewness, and combining the distribution deviation and discharge difference, obtain the feature vector, and use a classification algorithm to detect the partial discharge of the gas-insulated switchgear.

[0045] Furthermore, since the amplitude variation distribution of different partial discharge modes in the PRPD spectrum varies, and the number of discharge activities distributed across different signal amplitude ranges differs for different discharge defect types, the discharge difference is calculated by analyzing the distribution of discharge activities under different signal amplitudes in the PRPD spectrum and combining the distribution deviation. Specifically: Divide the Y-axis in the two-dimensional PRPD map into multiple interval ranges; In this embodiment, the Y-axis is divided into 20 intervals. As for other implementation methods, the implementer can set them according to the actual situation.

[0046] The number of data points distributed across each interval range within the two-dimensional PRPD map is counted; the number of all interval ranges is arranged in ascending order according to the corresponding discharge amount to form a quantity distribution column; Find the maximum length of consecutive non-zero elements in a statistical distribution. It should be noted that, for ease of understanding, we assume the quantity distribution is as follows: The cases where elements are consecutively non-zero are: , ,in, The length is 2. The length is 4, therefore the maximum length is 4.

[0047] The sum of the position corresponding to the maximum value in the quantity distribution column and the maximum length is calculated, and the product of the sum and the distribution deviation is used as the discharge difference degree for each monitoring cycle. In this embodiment, the first Taking the first monitoring cycle as an example, then the first... The formula for calculating the discharge difference across monitoring cycles is:

[0048] in, For the first Discharge variability across monitoring cycles For the first Distribution deviation over each monitoring period For the first The maximum length of the quantity distribution column corresponding to each monitoring period The middle is the first The position of the maximum value in the quantity distribution column corresponding to each monitoring cycle.

[0049] It should be noted that a larger maximum length indicates a wider amplitude range of discharge activity distribution, while a smaller maximum length indicates a narrower amplitude range of discharge activity distribution, meaning the discharge activity is more concentrated in a specific amplitude region. A larger positional order indicates that the most concentrated area of ​​discharge is in the high amplitude region, while a smaller positional order indicates that the most concentrated area of ​​discharge is in the low amplitude region. The discharge difference reflects the coupling characteristics of discharge activity in phase and amplitude distribution, assessing whether the discharge is concentrated in phase and whether it is both high and concentrated in amplitude. A larger value indicates that the discharge activity exhibits highly concentrated, anomalous, and non-random characteristics in both phase and amplitude distribution.

[0050] The flowchart of the method for obtaining discharge difference provided in this application embodiment is as follows: Figure 2 As shown.

[0051] Furthermore, the degree of abrupt changes in the phase distribution of discharge intensity in the two-dimensional PRPD spectrum was analyzed, specifically: Select the maximum discharge quantity among all data points at each phase angle within the two-dimensional PRPD spectrum; calculate the kurtosis of the maximum discharge quantity at all phase angles; It should be noted that the calculation of kurtosis is a well-known technique and will not be elaborated upon here.

[0052] It should be noted that the kurtosis reflects the degree of prominence of the shape distribution of the PRPD spectrum compared to the normal distribution, and assesses the sharpness of the maximum discharge quantity in the phase distribution. The larger the kurtosis, the more concentrated the maximum discharge quantity is on a very small number of phases. The smaller the kurtosis, the flatter and more uniform the distribution of the maximum discharge quantity in the phase.

[0053] Secondly, based on the distribution deviation, discharge difference, and kurtosis of each monitoring cycle, the partial discharge situation of the gas-insulated switchgear is detected, specifically as follows: Obtain the maximum and minimum discharge densities for all data points in the PRPD map; The maximum discharge density, minimum discharge density, skewness, distribution deviation, discharge difference, and kurtosis of each monitoring period are used to form a feature vector; It should be noted that the maximum discharge density in the eigenvector reflects the concentration of the discharge in the densest region on the phase-amplitude plane, the minimum discharge density reflects the concentration of the discharge in the sparsest region on the phase-amplitude plane, the skewness reflects the asymmetry of the distribution of discharge frequency, the distribution deviation reflects the concentration and distribution symmetry of discharge activity, the discharge difference reflects the distribution characteristics of discharge in amplitude, and the kurtosis reflects the sharpness of the discharge intensity distribution in phase.

[0054] We collected ultra-high frequency signals and power frequency voltage signals corresponding to different partial discharge defect types in historical periods, and obtained feature vectors respectively using the above methods to form a training set. It should be noted that different types of partial discharge defects can include various types such as tip discharge, hole discharge, floating electrode discharge, and free particle discharge.

[0055] The training set is used as input to the classification algorithm to train the classification algorithm; In this embodiment, multiple ultra-high frequency signals and power frequency voltage signals corresponding to different partial discharge defect types are collected. The data volume for each partial discharge defect type is 100. Data of 100 gas-insulated switchgear units during normal operation are also collected. Feature vectors are obtained for each unit and a training set is formed. As for other implementation methods, the implementer can set them according to the actual situation.

[0056] In this embodiment, a decision tree algorithm is used for training. The decision tree algorithm is a well-known technology and will not be described in detail here. As other implementation methods, implementers may use other methods of existing technology, such as neural network models. This embodiment does not impose any special restrictions on this.

[0057] The feature vectors of each monitoring period are used as input to the classification algorithm, and the partial discharge defect type is output to enable real-time detection of partial discharge in gas-insulated switchgear.

[0058] When partial discharge is detected in gas-insulated switchgear, maintenance personnel can inspect and repair the gas-insulated switchgear according to the type of partial discharge defect, thereby improving the maintenance efficiency of gas-insulated switchgear.

[0059] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0060] 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 specification.

[0061] 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 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. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of this application, without departing from the content of the technical solution of this application, shall fall within the protection scope of the technical solution of this application.

Claims

1. A method for full-cycle monitoring of gas-insulated switchgear using multi-source data fusion, characterized in that, The method includes: Real-time acquisition of UHF and power frequency voltage signals during each monitoring cycle of gas-insulated switchgear operation; The method analyzes the cumulative energy deviation of the signal amplitude at different times in the UHF signal and calculates the denoising energy level at each time. Using the changing trend of the denoising energy level at different times, it obtains all initial discharge times of the UHF signal. It extracts the phase angle of each initial discharge time in the power frequency voltage signal and, combined with the signal amplitude of the UHF signal, obtains the partial discharge distribution (PRPD) spectrum. It analyzes the density of discharge activity in the neighborhood of each data point in the PRPD spectrum, calculates the discharge density, and, considering the high frequency characteristics and distribution skewness of the discharge density, calculates the distribution deviation for each monitoring period. Combining the distribution range and concentration of data points under different discharge intensities in the PRPD spectrum, it determines the discharge difference for each monitoring period. Based on the steepness of the discharge intensity distribution at different phase angles in the PRPD spectrum, as well as the discharge density and the aforementioned skewness, and combining the distribution deviation and discharge difference, it obtains feature vectors and uses a classification algorithm to detect partial discharge in gas-insulated switchgear.

2. The method for full-cycle monitoring of gas-insulated switchgear using multi-source data fusion as described in claim 1, characterized in that, The calculation of the denoised energy at each time step includes: Calculate the square of the signal amplitude at each moment in the UHF signal, and denote it as the instantaneous energy; calculate the mean of the instantaneous energy at all moments in the UHF signal, and denote it as the average energy. Calculate the product of the time sequence and the average energy at any given moment in the UHF signal, and use it as the reference cumulative energy at that given moment; The denoising energy level is the sum of the differences between the instantaneous energy at any given time and all previous times and the baseline cumulative energy at any given time.

3. The method for full-cycle monitoring of gas-insulated switchgear using multi-source data fusion as described in claim 1, characterized in that, The acquisition of all initial discharge moments of the UHF signal includes: Obtain the minimum value of the denoised energy at all times in the UHF signal; take the time corresponding to the first minimum value as the initial time of the first discharge; take the last time in the preset time after the initial time of the first discharge as the segmentation point; remove the signal amplitude of all times before the segmentation point from the UHF signal as the newly segmented UHF signal; for the newly segmented UHF signal, recalculate the denoised energy at all times in the newly segmented UHF signal, and obtain the time corresponding to the first minimum value as the initial time of the second discharge, and so on, to obtain all the initial times of discharge of the UHF signal.

4. The method for full-cycle monitoring of gas-insulated switchgear using multi-source data fusion as described in claim 1, characterized in that, The process of obtaining the PRPD map is as follows: Each initial discharge moment corresponds to one discharge event. The maximum signal amplitude in the UHF signal after each initial discharge moment within a preset time is obtained as the discharge amount of one discharge event. Based on the discharge amount and phase angle of all discharge events in each monitoring period, a PRPD spectrum is plotted. In the PRPD spectrum, the horizontal axis represents the phase, the vertical axis represents the discharge amount, and the color of the data point represents the number of discharges under a specific phase angle and discharge amount.

5. The method for full-cycle monitoring of gas-insulated switchgear using multi-source data fusion as described in claim 4, characterized in that, The discharge density is the sum of the number of discharges corresponding to all data points in the neighborhood of each data point in the PRPD spectrum.

6. The method for full-cycle monitoring of gas-insulated switchgear using multi-source data fusion as described in claim 1, characterized in that, The degree of skewness refers to the skewness of the discharge density of all data points within the PRPD spectrum.

7. The method for full-cycle monitoring of gas-insulated switchgear using multi-source data fusion as described in claim 1, characterized in that, The distribution deviation is the product of the maximum discharge density of all data points within the PRPD spectrum and the degree of skewness.

8. The method for full-cycle monitoring of gas-insulated switchgear using multi-source data fusion as described in claim 4, characterized in that, Determining the discharge difference in each monitoring cycle includes: Divide the vertical axis of the PRPD map into multiple intervals; count the number of all data points in the PRPD map that fall within each interval; arrange the numbers in all intervals in ascending order of the corresponding discharge amount to form a quantity distribution column; The maximum length of consecutive non-zero elements in the quantity distribution column is calculated; the sum of the position corresponding to the maximum value in the quantity distribution column and the maximum length is calculated. The discharge difference is the product of the sum and the distribution deviation.

9. The method for full-cycle monitoring of gas-insulated switchgear using multi-source data fusion as described in claim 4, characterized in that, The method for measuring steepness is as follows: select the maximum discharge quantity among all data points at each phase angle in the PRPD spectrum; calculate the kurtosis of the maximum discharge quantity at all phase angles as the steepness.

10. The method for full-cycle monitoring of gas-insulated switchgear using multi-source data fusion as described in claim 9, characterized in that, The process of obtaining the feature vector is as follows: obtain the maximum discharge density and minimum discharge density of all data points in the PRPD spectrum; and combine the maximum discharge density, minimum discharge density, skewness, distribution deviation, discharge difference and steepness of each monitoring period to form the feature vector.

Images