A high-voltage switch cabinet partial discharge on-line monitoring method and system

Abstract

The application discloses a high-voltage switch cabinet partial discharge on-line monitoring method and system, and relates to the technical field of power equipment insulation state monitoring. In view of the problems that the monitoring device has no power frequency voltage reference, the phase spectrum diagram cannot be established, and the periodic interference false alarm occurs frequently, the method records the arrival time and amplitude of the transient ground voltage pulse; the candidate folding period is scanned in the search interval covering the power frequency nominal period, the maximum of the weighted synthesis vector module value of the folding phase angle and its double angle is taken as the synchronism score, the estimated power frequency period and phase reference are locked, and the frequency deviation is re-estimated and tracked in sections; the power frequency related discharge is determined according to the score threshold, and the phase resolution discharge spectrum diagram is reconstructed, otherwise, it is determined as non-synchronous interference; according to the trend of the discharge pulse rate and amplitude quantile value, the monitoring platform is reported in stages through the Internet of Things link, and is reviewed and disposed. The application can output the standard phase spectrum diagram and quantitatively distinguish the interference without any voltage access, and is suitable for the large-scale on-line monitoring of high-voltage switch cabinets.

Description

Technical Field

[0001] This invention relates to the field of electrical equipment insulation condition monitoring technology, and more specifically, to a method for online monitoring of partial discharge in high-voltage switchgear, as well as a system, electronic equipment and computer-readable storage medium for implementing the method. Background Technology

[0002] Before insulation defects in high-voltage switchgear completely break down, they often exist for several months in the form of partial discharge. The high-frequency pulses generated by partial discharge will induce transient ground voltage signals on the metal wall of the cabinet. Therefore, attaching transient ground voltage sensors (TEV sensors) to the outer wall of the cabinet is the most popular sensing method for online monitoring of partial discharge in switchgear - the installation is uninterrupted, the cost per point is low, and it is suitable for large-scale deployment of rows of switchgear in the power distribution room.

[0003] According to DL / T 393 "Test Procedures for Condition-Based Maintenance of Power Transmission and Transformation Equipment", live-line testing and online monitoring of switchgear are routine items in condition-based maintenance. To transform detected pulses into diagnostic data, the industry standard practice is to plot a phase-resolved discharge spectrum: with the phase of the power frequency voltage as the horizontal axis and the discharge amplitude as the vertical axis, the accumulated pulse points show different insulation defects exhibiting characteristic clustered distributions on the spectrum. Textbooks such as "High Voltage Testing Technology" have long established the strong correlation between discharge phase and power frequency voltage, and the knowledge of identifying defect types based on spectrum morphology has been accumulated in the industry for over twenty years. The problem is that the horizontal axis of the spectrum requires a power frequency phase reference, which is naturally inaccessible to monitoring devices mounted on the cabinet wall. Existing online devices address this issue along two paths. One approach is to use an external reference: either by drawing lines from the secondary side of the voltage transformer, or by installing a dedicated power frequency zero-crossing detector and wirelessly broadcasting synchronization pulses to each monitoring node. This requires wiring or adding master-slave synchronization hardware when deploying multiple cabinets, necessitating power outages for wiring work, and the cost and workload increase linearly with the number of cabinets. Another approach completely abandons the phase reference: the device uses a freely running internal clock as a reference, analyzing only the amplitude and frequency of pulses or their clustering patterns on a certain self-running time axis to distinguish between discharge and noise. This eliminates the need for wiring, but at the cost of the output being disconnected from existing phase spectrum knowledge systems. Furthermore, it lacks effective means to distinguish between periodic interference commonly found in the field, such as inverter switching noise and carrier crosstalk, resulting in a high false alarm rate. Maintenance personnel often lose trust in the device after repeated ineffective alarms, leading to its frequent silencing.

[0004] Through repeated comparisons and verifications at the power distribution room, the inventors discovered a fact overlooked by both routes: the discharge pulse sequence itself carries power frequency phase information. Since the discharge is concentrated in a specific phase interval of the power frequency cycle, the flow at the moment of pulse arrival will inevitably exhibit clustering after folding according to the power frequency cycle. Conversely, by searching along the candidate cycle for the most significant folding pattern of clustering, the power frequency cycle and phase reference can be recovered from the pulse flow without any voltage input. Engineering this simple idea requires overcoming two hurdles. First, the power grid frequency is not always equal to the nominal value, but drifts slowly around it, folding at a fixed 20-millisecond cycle. When the frequency deviation is 0.01Hz, the phase drift rate reaches 3.6 degrees per second, accumulating to 36 degrees in 10 seconds, and turning what should be a clear discharge cluster into a ring in 60 seconds. Therefore, the folding period must be estimated segment by segment rather than taking the nominal value. Secondly, the typical form of partial discharge in switchgear is one cluster each in the positive and negative half-cycles, with the two clusters separated by approximately 180 degrees. If the phase concentration after folding is directly measured using a conventional double-angle composite vector, the two cluster vectors will be in opposite directions and almost completely cancel each other out, resulting in severe distortion of the concentration index. Therefore, a separate metric must be constructed specifically for this dual-cluster structure. Without overcoming these two hurdles, the recovery of the phase reference from the pulse current will remain merely theoretical.

[0005] Therefore, there is a need for an online monitoring method for partial discharge in high-voltage switchgear that does not require any power frequency voltage reference, can recover the phase reference and reconstruct the standard phase-resolved discharge spectrum using only the transient ground voltage pulse current, and can quantitatively distinguish between power frequency-related discharges and periodic interferences. Summary of the Invention

[0006] In view of this, the present invention proposes an online monitoring method for partial discharge in high-voltage switchgear, as well as a system, electronic equipment and computer-readable storage medium for implementing the method, aiming to solve the problems of phase spectrum diagrams being impossible to establish and periodic interference false alarms occurring when the monitoring device has no voltage reference, so that the monitoring device without wiring can also output standard spectrum diagrams compatible with existing spectrum criteria.

[0007] According to one aspect of the present invention, a method for online monitoring of partial discharge in a high-voltage switchgear is provided. The method includes: step 101, a monitoring device installed on the high-voltage switchgear cabinet detects discharge pulses through a transient ground voltage sensor and records the arrival time and amplitude of each discharge pulse. The arrival time is given by the local clock of the monitoring device, and the monitoring device is not connected to a power frequency voltage reference; step 102, for the arrival time of each discharge pulse within a sliding statistical window, candidate folding periods are selected one by one within a preset search interval covering the nominal power frequency period. The remainder of each arrival time divided by the candidate folding period is converted into a folding phase angle. The amplitude-weighted first composite vector magnitude and second composite vector magnitude are calculated using the folding phase angle and its double angle, respectively. The one of the two, which is not lower than the other, is taken as the synchronization of the candidate folding period. The synchronization score is used to determine the candidate folding period with the highest synchronization score among all candidates as the estimated power frequency period, and the folding phase angle under the estimated power frequency period is used as the phase reference for each discharge pulse; Step 103: when the synchronization score corresponding to the estimated power frequency period is not lower than the preset synchronization threshold, it is determined that there is power frequency related discharge within the sliding statistical window, and a phase-resolved discharge spectrum is generated according to the phase reference and amplitude accumulation; when the synchronization score is lower than the preset synchronization threshold, it is determined to be asynchronous interference and is not included in the discharge statistics; Step 104: the discharge pulse rate and amplitude quantile are statistically analyzed for the time period in which power frequency related discharge is determined, and the warning level, attention level or normal level are defined according to preset conditions. The monitoring device reports the spectrum, characteristic quantity or statistical summary to the monitoring platform according to the level through the Internet of Things communication link, and the monitoring platform reviews and generates disposal suggestions.

[0008] In one optional implementation, the candidate folding period is scanned in two levels of coarseness and fineness, and the sliding statistical window is divided into multiple sub-segments for re-estimation to track the power grid frequency drift. The amplitude weighting of the synthesized vector is taken as the proportion of each pulse amplitude to the sum of amplitudes within the window, so that the large amplitude discharge dominates the synchronization judgment. After generating the spectrum, the number of phase clusters, the cluster spacing and the phase width within the cluster can be extracted and compared with the pre-stored template to distinguish between tip discharge, floating discharge and surface discharge. When it is determined to be asynchronous interference, it is folded twice in the interference search interval that deviates from the nominal power frequency period to estimate the interference repetition frequency for interference source investigation. The classification conditions are defined according to the growth rate of the discharge pulse rate and the amplitude quantile value relative to their respective baseline values.

[0009] According to another aspect of the present invention, a partial discharge online monitoring system for high-voltage switchgear is provided. The system includes a monitoring device installed on the cabinet and a monitoring platform connected to it via an Internet of Things (IoT) communication link. The monitoring device includes a pulse acquisition module, a period search module, a discharge discrimination module, and a statistical reporting module, which respectively undertake pulse detection and time recording, candidate period scanning and phase reference locking, discharge and interference discrimination and spectrum generation, and hierarchical statistics and reporting. The monitoring platform includes a verification module and a processing module.

[0010] According to another aspect of the present invention, an electronic device is provided, the electronic device including a processor and a memory, the memory storing a computer program, which, when executed by the processor, implements the online monitoring method for partial discharge of high-voltage switchgear as described in any of the above implementations.

[0011] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program for executing the online monitoring method for partial discharge of high-voltage switchgear described in any of the above implementations.

[0012] Based on the method and system provided in the above embodiments of the present invention, the monitoring device is freed from dependence on voltage transformer wiring and synchronization hardware. The phase reference is recovered from the discharge pulse current itself, thus saving the cost of wiring construction and synchronization devices for multi-cabinet deployments. The recovered phase reference allows the spectrum to return to the industry-standard phase resolution form, and the spectrum criteria accumulated over more than 20 years can be directly used. The double-angle composite vector fills in the blind zone of the positive and negative half-cycle dual-cluster structure with a double angle, and the synchronization score transforms the distinction between periodic interferences such as discharge and inverter noise into a quantitative criterion with an adjustable threshold, thereby reducing false alarms. Thus, the embodiments of the present invention use pure algorithmic means to supplement the missing phase dimension of the wiring-free monitoring device.

[0013] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0014] The above and other objects, features, and advantages of the present invention will become more apparent from the more detailed description of the embodiments of the invention in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0015] Figure 1 A flowchart of an online partial discharge monitoring method for high-voltage switchgear according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the online partial discharge monitoring system for high-voltage switchgear according to an embodiment of the present invention; Figure 3 This is a schematic diagram comparing the deviation at different frequencies with the cumulative phase drift under a window length when folding at a fixed nominal period. Figure 4 This is a schematic diagram of the folded phase histogram of the discharge pulse in the embodiment; Figure 5 This is a schematic diagram illustrating the synchronization score scanning along the candidate folding cycle in the embodiment; Figure 6This is a schematic diagram showing the comparison between the first and second composite vector magnitudes of the discharge window and the interference window in the embodiment; Figure 7 This is a schematic diagram illustrating the transition of discharge states between the normal level, the attention level, and the warning level. Figure 8 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention; Figure 9 This is a schematic diagram illustrating the hierarchical reporting and interaction process between the monitoring device and the monitoring platform in this embodiment. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. The search interval width, scanning interval, synchronization threshold, statistical window and segment duration, as well as the number of pulses, frequency, and score values ​​in the embodiments described below, are all illustrative examples and do not constitute a limitation on the scope of protection of this invention. Those skilled in the art can adjust the corresponding parameters according to the on-site conditions, and the adjusted implementation methods still fall within the scope of protection of this invention.

[0017] Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described exemplary embodiments are merely a part of the embodiments of the present invention, and not all of the embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein. Furthermore, the term "and / or" in the present invention is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone.

[0018] Figure 1 This is a flowchart of an online partial discharge monitoring method for high-voltage switchgear according to an embodiment of the present invention. Figure 1 As shown, the method includes steps 101 to 104, which are executed by a monitoring device installed on the high-voltage switchgear cabinet and a remote monitoring platform.

[0019] In step 101, the monitoring device detects discharge pulses using a transient ground voltage sensor and records the arrival time and amplitude of each discharge pulse. The transient ground voltage sensor is a capacitively coupled probe, magnetically attached to the metal surface of the front panel or side wall of the switchgear. When an insulation defect discharges inside the cabinet, the transient voltage pulse induced in the cabinet wall is coupled by the sensor, bandpass filtered, and envelope detected before being sent to the comparator. If the pulse exceeds the detection threshold, a pulse event is formed. The monitoring device uses its local clock to stamp the arrival time of this event and simultaneously records the envelope peak value as the amplitude, which is commonly expressed in decibels and millivolts. The local clock is maintained by a common crystal oscillator within the monitoring device, without the need for high-precision time synchronization—subsequent steps perform a search estimation for the period, and the frequency difference of the crystal oscillator, on the order of twenty parts per million, will be naturally absorbed by the search process. This also means that no time synchronization is required between multiple monitoring devices, and each device independently completes its own phase recovery. Optionally, a monitoring device can simultaneously connect to two sensors on the front panel and side wall of the same cabinet via multiple sensor interfaces, or one sensor on each of two adjacent cabinets. Each sensor executes this method independently, and the judgment results of adjacent cabinets can be used for horizontal comparison by the monitoring platform.

[0020] Two types of boundary situations on the acquisition side should also be handled. First, pulse storms near the detection threshold: When discharge is severe or interference breaks out, the number of pulse events per unit time may surge. The monitoring device records a maximum of 2000 events in each statistical window according to the principle of first-come, first-served. The excess events are only accumulated and not stored in detail to prevent storage and computing power from being overwhelmed by instantaneous surges. Second, clock correction: When the device receives remote time correction from the platform, the statistical window that crosses the time correction time is invalidated. It is re-accumulated after the self-time correction is completed to avoid human time jumps contaminating the folding calculation. Third, self-diagnosis of sensor mounting looseness: When the magnetic surface of the sensor is loose or the coupling is poor, the detected pulse will be interrupted for a long time or the amplitude will collapse. The monitoring device counts the pulse flow and amplitude level of several consecutive reporting cycles. When it deviates significantly from the previous pattern, it reports the device self-test abnormality and prompts the maintenance personnel to come to the site to reset the sensor to avoid "no alarm" being misread as "no discharge". It is deliberately emphasized here that the monitoring device is not connected to the power frequency voltage reference, does not have leads from the voltage transformer, and does not have any zero-crossing detection or synchronization receiving hardware. The entire device has no electrical connection with the primary equipment or secondary circuits. This is precisely the deployment premise that distinguishes this method from the external reference route. The selection of the detection threshold and filter frequency band are mature technologies for transient ground voltage detection and will not be elaborated here.

[0021] In step 102, the monitoring device performs candidate period scanning and folding on the pulse arrival times within the sliding statistical window to restore the estimated power frequency period and phase reference. The sliding statistical window refers to all pulse events within the most recent period participating in a period search; the window length is typically 10 seconds, but can be adjusted between 5 and 30 seconds, progressing according to the acquisition schedule. The preset search interval refers to the period range for candidate folding period values, covering 0.998 to 1.002 times the nominal power frequency period, which for a 50Hz system is 19.96 milliseconds to 20.04 milliseconds, sufficient to cover the normal fluctuation range of the power grid frequency from 49.9Hz to 50.1Hz. The candidate folding period refers to the folding period value obtained one by one within the preset search interval. For each candidate folding period, take the remainder of the candidate value at each arrival time within the window, divide the remainder by the candidate value and multiply by 360 degrees to obtain the folding phase angle of the pulse. Intuitively speaking, it is to cut the entire time axis into segments according to the candidate periods and stack them together to see where the pulse falls within the period.

[0022] Here's a concrete example of the conversion: Suppose a pulse arrives at 37.254682 seconds after the device is powered on. The candidate folding period is 20.0072 milliseconds. Dividing the arrival time by the candidate period gives 1862 integer periods, with a remainder of 1.2756 milliseconds. Dividing the remainder by the candidate period and multiplying by 360 degrees gives a folding phase angle of approximately 23.0 degrees. If the same arrival time is folded according to the nominal period of 20.0000 milliseconds, the remainder is 14.6820 milliseconds, and the phase angle is 264.3 degrees. The difference of only 7.2 microseconds in the candidate period, accumulated over 1862 integer periods, causes a phase misalignment of more than 240 degrees on this single pulse, demonstrating the extent to which the period value is amplified by the window length.

[0023] The metric for judging whether the stacking is correct is the magnitude of the composite vector weighted by amplitude. Each pulse is considered as a vector whose azimuth angle above the unit circle is equal to its folded phase angle. The vector sum is calculated using the proportion of the pulse amplitude to the total amplitude within the window as the weight. The magnitude of the resulting composite vector is between 0 and 1: close to 1 when the pulse phase is highly concentrated, and close to 0 when it is uniformly distributed. The reason for using amplitude weighting is to take into account the presence of sporadic noise pulses with very low amplitude in the background. This allows the true discharge with high amplitude to dominate the score, making it difficult for low-amplitude noise to influence the judgment.

[0024] Using only the aforementioned double-angle magnitude has a pitfall specifically designed for partial discharge: a typical form of partial discharge in switchgear involves two clusters of discharge occurring in each of the positive and negative half-cycles of the power frequency, with the two clusters approximately 180 degrees apart in phase and their vector directions being exactly opposite, almost canceling each other out. After repeated testing, the inventor found that the magnitude of the combined vector at the single-angle of the discharge window was only 0.014, which would be misjudged as no discharge at any reasonable threshold. The solution is to double the phase angle of each fold and then calculate the combined vector again—the two clusters 180 degrees apart overlap at the double angle, resulting in a calculated double-angle magnitude of 0.922, accurately reflecting the concentration. Therefore, this method calculates two magnitudes simultaneously for each candidate folding cycle. The first combined vector magnitude refers to the magnitude obtained by weighting the amplitudes using the single-angle of each folding phase angle as the azimuth angle; the second combined vector magnitude refers to the magnitude obtained by weighting the amplitudes using the double-angle of each folding phase angle as the azimuth angle. For example: Within the sliding statistical window, a large number of discharge pulses are divided into two major clusters: Cluster 1: Phase concentrated at θ≈70° (power frequency positive half-cycle discharge) Cluster 2: The phase is concentrated at θ≈250° (power frequency negative half-cycle discharge, which is 180° out of phase with 70°). All pulses are weighted by amplitude, and the total weight of the two clusters is approximately equal to that of each other (after merging, the equivalent weights w1=0.5, w2=0.5, and the final magnitude after the superposition of multiple pulses in the field is 0.922, which is close to the theoretical maximum value of 1). Key phase transition: double angle Cluster 1 Double angle: 2θ1 = 2 × 70° = 140° Cluster 2 Double angle: 2θ2 = 2 × 250° = 500°, minus 360° Periodic equivalent: 500° - 360° = 140°; The two clusters of double angles completely overlap, and their vectors are superimposed in the same direction, so they will not cancel each other out.

[0025] Let the weights of the two equivalent impulses be w1=0.5 and w2=0.5; Summation of double-angle components: X2 = w1cos(2θ1) + w2cos(2θ2) Y2=w1sin(2θ1)+w2sin(2θ2) The magnitude of the double-angle composite vector: M2= .

[0026] cos140°≈-0.766044; sin140°≈0.642788; X 2=0.5×cos140°+0.5×cos140° =(0.5+0.5)×cos140°=1×(-0.766044)=-0.766044; Y2=0.5×sin140°+0.5×sin140° =(0.5+0.5)×sin140°=1×0.642788=0.642788; Theoretical ideal magnitude (pulse is completely concentrated with no dispersion) M 2理论 = = =1.0 The above are the theoretical values ​​for ideal perfect clustering. However, three loss factors exist in the field, causing the modulus to decrease from 1.0 to 0.922: 1. Each cluster of pulses is not strictly concentrated at a single angle, but has a phase width of about 30°. The pulse vector has a slight angular offset, which slightly cancels out after superposition. 2. A small amount of low-amplitude random noise pulses are mixed in within the window. The noise vector is evenly distributed, slightly reducing the length of the synthesized vector. 3. The power grid frequency drifts slightly, and there is a slight phase shift within the sub-segment, so the clusters cannot be completely overlapped.

[0027] After considering all on-site attenuation factors, the final measured values ​​are as follows: M2=0.922.

[0028] Two clusters with a double angle of 70° and 250° phase, differing by 180°, cancel each other out significantly due to their opposing vectors. X1=0.5×cos70°+0.5×cos250°=0; Y1=0.5×sin70°+0.5×sin250°=0; M 1理论 = =0.

[0029] The synchronization score refers to the larger of the first and second synthetic vector magnitudes, which serves as a quantitative measure of the pulse phase concentration under the candidate folding period: a single-cluster discharge can be identified by obtaining a high magnitude value from a double-angle synthetic vector, while a double-cluster discharge can be identified by obtaining a high magnitude value from a double-angle synthetic vector, and neither type is missed.

[0030] The candidate folding period with the highest synchronization score is determined as the estimated power frequency period. The phase reference refers to the folded phase angle obtained after folding each discharge pulse according to the estimated power frequency period—it assigns each pulse a relative phase coordinate corresponding to the power frequency voltage period, and the horizontal axis of the subsequent spectrum is based on it. Figure 5As shown, the score, along with the scanning curve of the candidate period, exhibits a sharp peak at the actual power frequency period. A deviation of several microseconds from this peak results in a significant drop in score, demonstrating the search's resolving power. It's worth explaining the source of this resolving power: even a difference of only a few tenths of a microsecond between the candidate period and the actual period accumulates over hundreds of power frequency periods within the window, resulting in a considerable phase misalignment. This causes the folded clusters to disperse, leading to a drop in score—the period error is amplified into a phase error by the window length, which is the physical mechanism behind the accurate peak finding.

[0031] In terms of computational implementation, modulo, conversion, and vector accumulation are all fixed-point operations, and sine and cosine operations can be performed by looking up tables. The computational cost for each candidate folding period is proportional to the number of pulses within the window. The computational cost of the entire scan is multiplied by the number of candidates. There is no iterative convergence process, and the time consumption is fixed and controllable. It is also necessary to prevent a boundary case: when the number of pulses within the window is too small, the statistical fluctuation of the magnitude of the synthesized vector is large. In extreme cases, when there are only 2 pulses within the window, the magnitude can easily be accidentally too high. Therefore, a minimum pulse number threshold is set. When the number of pulses within the window is less than the threshold (typically 20), this window does not perform period search or make any judgment, and waits for the pulses to continue to accumulate. The scan is performed in two levels, coarse and fine, to save computational cost: first, a coarse scan of the entire interval is performed at a first scan interval of 4 microseconds to obtain the initial selected period. Then, a fine scan is performed at a second scan interval of 0.2 microseconds within the fine scan interval centered on the initial selected period. The period error of 0.2 microseconds accumulates to a phase error of about 1.8 degrees within a 10-second window, which has no visible impact on the spectral morphology. The first scan interval can be between 2 microseconds and 8 microseconds, and the second scan interval can be between 0.1 microseconds and 0.5 microseconds, depending on the computing power of the device.

[0032] The time-varying nature of the power grid frequency must also be addressed. Frequency drift means that there is no fixed period that is accurate for the entire window, such as... Figure 3 As shown, when folded according to a fixed nominal period, a frequency deviation of 0.01Hz corresponds to a phase drift rate of 3.6 degrees per second, accumulating to 36 degrees in 10 seconds. A deviation of 0.05Hz accumulates to 180 degrees in 10 seconds. The longer the window, the more thorough the blurring. That is, the larger the value of the sliding statistical window (the duration of the statistical pulse), the greater the cumulative phase shift caused by small frequency drifts in the power grid, resulting in a more scattered discharge phase cluster and a more blurred spectrum. To address this, the sliding statistical window is divided into multiple segments according to a preset segment duration, typically 2 seconds per segment. Each segment performs the above scan independently and obtains its own phase reference. The frequency within a segment is approximately constant, and a deviation of 0.018Hz accumulates to only 13 degrees in 2 seconds, having a slight impact on the cluster morphology. The spectra of each segment are aligned according to their respective phase references and then merged for accumulation. The statistics of the long window are thus established on the basis of no blurring. The segment duration can be between 1 and 5 seconds, with shorter values ​​used for weak power grids with drastic frequency fluctuations.

[0033] In step 103, the monitoring device distinguishes between power frequency-related discharges and asynchronous interference using synchronicity scores as a quantitative criterion. Power frequency-related discharges refer to discharge events whose occurrence time maintains a stable phase relationship with the power frequency voltage cycle; partial discharges caused by insulation defects fall into this category. Asynchronous interference refers to pulse events that do not have a stable phase relationship with the power frequency cycle, including inverter switching noise, power line carrier crosstalk, and random burst noise. When the estimated synchronicity score corresponding to the power frequency cycle is not lower than a preset synchronicity threshold, it is determined that a power frequency-related discharge exists within the sliding statistical window. Each pulse is accumulated into a phase-resolved discharge spectrum based on phase reference and amplitude. The phase-resolved discharge spectrum is a two-dimensional spectrum with the folded phase angle as the horizontal axis and the pulse amplitude as the vertical axis, accumulating pulse counts and amplitude distribution according to phase bins. It is isomorphic to the PRPD spectrum used in the industry, and existing spectrum discrimination experience can be directly applied. When the synchronicity score is lower than the preset synchronicity threshold, the pulses within the window are determined to be asynchronous interference and are not included in the discharge statistics. The phase bins of the spectrum are typically divided into 72 bins of 5 degrees each. The bin width can be adjusted between 2 and 10 degrees: if the bins are too wide, the cluster shape will be flattened; if they are too narrow, the count of each bin will be sparse and the shape will be fuzzy. 5 degrees is a suitable compromise for the common 30-degree cluster width.

[0034] It is also necessary to explain the situation of multiple discharge sources coexisting. If two defects discharge simultaneously within the same cabinet—for example, surface discharge along the cable compartment superimposed with busbar suspension discharge—both pulses maintain a stable phase relationship with the power frequency. The period search and synchronization judgment are unaffected, and the scores remain significant. The only difference is that the spectrum shows more than two clusters of superimposed patterns. In this case, the spectral characteristics no longer match the single template. The monitoring device will label the type as composite discharge and report it at one level higher. The monitoring platform will retrieve the pulse details, perform amplitude layer separation, and then conduct manual analysis. Multiple sources coexisting are a few complex operating conditions. Returning them to humans rather than forcing machine judgment is a more prudent engineering choice. The preset synchronization threshold can be between 0.25 and 0.6, typically 0.4. The calibration method is to collect several interference windows on a cabinet without discharge background, and then inject several discharge windows with simulated discharge sources. The score distribution of the two types of windows is statistically analyzed and the boundary is taken. From the perspective of the physical magnitude of the two types of scores, this boundary has sufficient safety margin: the score of the phase-concentrated discharge window is close to 1, while the score of the uniformly distributed interference window is inversely proportional to the square root of the number of pulses within the window according to the statistical fluctuation law, and is only around 0.1 at the level of hundreds of pulses. The two types of distributions hardly overlap, and the threshold of 0.4 falls within a wide gap between the two, and is not sensitive to changes in the number of pulses and the field noise level—this is also why this threshold can be uniformly configured throughout the station without needing to be tuned to each cabinet; although the inverter switching noise itself is strictly periodic, its repetition period is far from the power frequency search range, and is evenly distributed under the power frequency candidate period, such as Figure 6As shown, the measured magnitudes of the first and second synthetic vectors of the interference window are only 0.003 and 0.002, respectively, which are significantly different from the 0.922 of the discharge window, indicating that the threshold has sufficient margin.

[0035] Optionally, after determining that it is asynchronous interference, the monitoring device performs cycle folding again on the arrival times of the same batch within a preset interference search interval that deviates from the nominal power frequency period. If the interference comes from periodic sources such as inverter switches or carrier signals, it will be highly clustered under its own repetition period. When there is a folding period with a synchronization score not lower than the preset interference threshold, the reciprocal of the folding period is determined as the interference repetition frequency and reported. The interference repetition frequency refers to the pulse repetition frequency of the asynchronous interference source itself, which directly corresponds to the switching frequency or carrier frequency of the interfering equipment. Maintenance personnel can use this to locate the interference source by comparing the duty log with the nameplate of the on-site equipment. The preset interference search interval typically covers 0.05 milliseconds to 10 milliseconds, corresponding to a repetition frequency range of 20kHz to 100Hz; the preset interference threshold can be the same value as the preset synchronization threshold. The value of interference identification is not limited to avoiding false alarms: by finding and managing the interference source, the signal-to-noise environment of the monitoring point is fundamentally improved, which cannot be achieved by simply suppressing false alarms with a threshold.

[0036] In step 104, the monitoring device statistically analyzes the discharge pulse rate and amplitude quantile for the time periods where power frequency-related discharges are determined to exist, and reports them in a tiered manner. The discharge pulse rate refers to the number of pulses included in the spectrum per unit time; the amplitude quantile refers to the 95th percentile of the amplitude of all discharge pulses within the window. Using quantiles rather than maximum values ​​is to exclude interference from occasional large spikes. Ticking is based on relative changes rather than absolute values. This is because the background discharge levels of different cabinets vary greatly—cabinets in damp cable silos operate with low-level discharges year-round, and absolute thresholds either report frequently for them or lose sensitivity for clean cabinets. Increases relative to the baseline are reliable signals of deterioration progress. Time periods meeting preset warning conditions are designated as warning levels, time periods meeting only preset attention conditions are designated as attention levels, and other time periods are designated as normal levels. Preset warning conditions and preset attention conditions are defined based on the increase in discharge pulse rate and amplitude quantile relative to their respective baseline values. The baseline value is the median of the corresponding statistics within the most recent baseline period, which can be 7 to 30 days, typically 14 days. At the initial stage of equipment commissioning, the baseline period is not yet fully accumulated. During this stage, only statistics are accumulated, and no classification judgment is made. If an early warning occurs during the baseline period and is subsequently addressed through maintenance, the platform issues a baseline reset command. The baseline is then re-accumulated from the moment the maintenance is completed to prevent statistics from the deterioration period from contaminating the new baseline. Typically, an increase of more than 50% in both statistics relative to the baseline value is considered an early warning, and an increase of more than 30% in a single statistic is considered a concern. The growth threshold can be adjusted according to the importance of the cabinet. Figure 7As shown, the discharge state migrates between the three levels as the statistics evolve. To migrate to a higher level, two consecutive statistical windows must meet the conditions to avoid level fluctuations caused by single-window fluctuations.

[0037] Reporting data granularity is differentiated by level: Warning level reports phase-resolved discharge spectra and detailed discharge pulse information, including the arrival time, folded phase angle, and amplitude of each pulse, for platform verification and manual recalculation when necessary; Attention level reports spectral features and two statistical quantities; Normal level reports only statistical summaries according to a preset reporting cycle. The summary includes four values: discharge pulse rate, amplitude quantile, current synchronicity score, and estimated power frequency period, less than 100 bytes per report, with a typical reporting cycle of 4 hours. Including the estimated power frequency period in the summary also serves another purpose: the platform can horizontally compare the estimated values ​​of various monitoring points within the same time period. Under normal circumstances, all stations tend to converge; persistent deviations at individual points indicate clock or sensor link abnormalities, effectively providing additional device self-diagnostic information. The spectral features include the number of phase clusters, the spacing between phase clusters, and the phase width within each cluster. Specifically, phase clusters are formed by merging adjacent phase bins whose counts continuously exceed the background level in the spectral phase histogram. The background level is twice the median of all phase bin counts. The number of clusters is the number of clusters obtained by merging. The cluster spacing is the difference between the center phases of adjacent clusters, and the intra-cluster phase width is the phase span covered by the cluster. All three feature quantities are directly read from the histogram without any fitting. Optionally, the monitoring device compares the spectral feature quantities with pre-stored discharge type templates one by one to determine the discharge type, at least distinguishing between tip discharge, floating discharge, and surface discharge: tip discharge is a single cluster with a narrow cluster; floating discharge is two clusters of equal height with a cluster spacing of approximately 180 degrees; and surface discharge is two clusters of unequal height and wide clusters. Template comparison is performed by matching feature quantities to intervals one by one, without introducing training models. Tuning and troubleshooting are transparent to maintenance personnel. The IoT communication link can be NB-IoT, Cat.1, or the existing industrial Ethernet in the power distribution room. Networking and transmission are mature technologies and will not be elaborated here. After receiving the report, the monitoring platform reviews it and generates handling suggestions. The key points of the review are described together with Example 1.

[0038] Figure 2 This is a schematic diagram of the structure of an online partial discharge monitoring system for high-voltage switchgear according to an embodiment of the present invention. Figure 2 As shown, the system includes a monitoring device 10 installed on the high-voltage switchgear cabinet and a monitoring platform 20 connected to the monitoring device 10 via an Internet of Things (IoT) communication link. The division of the internal modules of the monitoring device 10 is a schematic logical functional division; in practice, they can be integrated into the same processing chip. The monitoring device 10 includes a pulse acquisition module 11, a period search module 12, a discharge discrimination module 13, and a statistical reporting module 14. The pulse acquisition module 11 is used to execute... Figure 1Step 101 in the illustrated embodiment. The periodic search module 12 executes step 102. The discharge discrimination module 13 executes step 103. The statistical reporting module 14 executes step 104. The monitoring platform 20 includes a review module 21 and a handling module 22. The review module 21 reviews the reported data, and the handling module 22 generates handling suggestions and tracks the closed loop. In one possible implementation, the monitoring device 10 uses a low-power microcontroller as its core, is magnetically mounted on the cabinet door or side, and is powered by a secondary power supply or battery inside the cabinet; the monitoring platform 20 is deployed at the maintenance station or dispatch server, and interfaces with the alarm subsystem of the substation integrated automation system. The handling suggestions are directly linked to the existing maintenance work order process. The system provided in this embodiment is used to execute the methods described in any of the aforementioned implementations, achieving the same technical effect. The processing mechanisms of each module will not be repeated.

[0039] Example The following combination Figure 4 , Figure 5 , Figure 6 and Figure 9 Taking a medium-voltage switchgear in the 10kV switch room of a 110kV substation in southern China as an example, this paper demonstrates the complete execution process of this method from interference identification to discharge early warning. A capacitively coupled transient ground voltage sensor is mounted on the front panel of the switchgear. The monitoring device is attached to the switchgear door. The sliding statistical window is set to 10 seconds, the segment duration is set to 2 seconds, the search interval is 19.96 milliseconds to 20.04 milliseconds, the coarse scan interval is 4 microseconds, the fine scan interval is 0.2 microseconds, the synchronization threshold is set to 0.4, and the statistical baseline period is set to 14 days. The device is connected to the monitoring platform through the Cat.1 gateway in the distribution room.

[0040] The first 14 days after the device is put into operation is the baseline accumulation period. The monitoring device only archives and counts data and does not make any classification judgments. During this period, the baseline value of the discharge pulse rate of the cabinet is set at 4.0 pulses per second and the baseline value of the 95th percentile of the amplitude is set at 12.1 dBmV. This background level corresponds to slight and stable surface activity in the cable compartment, which is a common condition of this batch of cabinets and does not constitute a defect.

[0041] One day during the day, a frequency converter was being debugged at a nearby pumping station. The monitoring device detected 160 pulses within a certain statistical window. The periodic search module scanned within the power frequency search range. The score curve was flat throughout, with the first composite vector magnitude of the peak window being only 0.003 and the second composite vector magnitude only 0.002. The synchronization score of 0.003 was far below the threshold of 0.4, indicating asynchronous interference, and was not included in the discharge statistics. Subsequently, within the interference search range of 0.05 milliseconds to 10 milliseconds, the signal folded twice, reaching a high score of 0.99 at 0.125 milliseconds. The interference repetition frequency was locked at 8kHz and reported. Maintenance personnel compared this frequency with the field equipment and confirmed the source was the switching noise of the pumping station's frequency converter. A filter was then installed on its power supply side. The device had zero discharge alarms throughout the day—based on the amplitude threshold, these interference pulses with relatively high amplitudes were sufficient to trigger repeated false alarms. Figure 6 As shown, the interference window and the discharge window differ by more than two orders of magnitude in their synthesized vector magnitudes. The quantification criterion transforms the previously manual experience-based visual assessment of whether a signal resembles a discharge into a single threshold comparison at the device level. The day after the filter was installed, a retest showed that the pulse count at the same window position decreased from 160 to 7, and the synchronicity score of the residual pulses remained below 0.03. The effectiveness of the interference mitigation was directly confirmed by the monitoring data. Statistics for this period subsequently returned to normal and were incorporated into the baseline. The device required no manual parameter adjustments throughout this interference event, and its no-tuning characteristic was verified on-site.

[0042] At 9:14 PM one evening after the rainy season that same year, 132 pulses were detected within the same statistical window. A coarse scan showed a clear main peak around 20.008 milliseconds, while a fine scan estimated the power frequency period to be 20.0072 milliseconds, corresponding to a grid frequency of 49.982 Hz—a deviation of 0.018 Hz from the nominal value. This deviation, when folded according to the fixed nominal period, represents a drift rate of 6.5 degrees per second. Figure 6 As shown, the magnitude of the second composite vector calculated at the main peak of the score is 0.922, while the magnitude of the first composite vector is only 0.014, which is a typical fingerprint of positive and negative half-cycle dual-cluster discharge; the synchronization score of 0.922 exceeds the threshold, indicating the presence of power frequency correlated discharge. Figure 4 As shown, the folded phase histogram accumulated according to the phase reference shows two clusters: about 58 pulses are concentrated between 65 degrees and 95 degrees, about 55 pulses are concentrated between 245 degrees and 275 degrees, and the rest are scattered. The number of clusters is 2, the cluster spacing is about 180 degrees, the two clusters are at the same height and the phase width within the cluster is about 30 degrees, which matches the floating discharge template item by item. The discharge type is initially judged to be floating discharge.

[0043] Over the next three weeks, the discharge statistics of the cabinet continued to climb: the discharge pulse rate increased from 4.1 pulses per second to 7.8 pulses per second and then to 13.2 pulses per second, while the 95th percentile of the amplitude rose from 12.4 dBmV to 18.6 dBmV. In the third week, the increases in both statistics relative to the baseline values ​​reached approximately 220% and 50% respectively, meeting the preset warning conditions for two consecutive statistical windows, and the status shifted from the "concern" level to the "warning" level. Figure 9 As shown, the monitoring device immediately reports the phase-resolved discharge spectrum and pulse details; the verification module checks three things in turn: the dual-cluster morphology of the spectrum remains stable across multiple consecutive statistical windows (excluding occasional events), the synchronicity score is consistently higher than 0.9 (excluding critical misjudgment), and the synchronous summary of other cabinets in the same room does not rise synchronously (excluding spatial external interference; external electromagnetic interference will be picked up by the sensors of multiple cabinets at the same time, while defects in the cabinet will only appear in this cabinet). After all three items are verified, the warning is confirmed, and the handling module generates a handling suggestion of "floating discharge, it is recommended to shut down the power within two weeks to check the busbar connection hardware" and pushes it to the maintenance team. Based on the handling recommendations, the maintenance team conducted a partial disassembly inspection of the cabinet during the planned power outage window. Focusing on the floating type indicated by the spectrum, they inspected the metal connectors and found an loose equalizing pad on the top of the busbar support insulator. There were visible black discharge marks between the pad and the hardware, consistent with the floating discharge pattern. After replacement and tightening, the baseline was reset via the platform. Within two weeks, the new baseline was established, and the discharge pulse rate dropped to 0.6 pulses per second and the amplitude percentile dropped to 8.3 dBmV, which was even lower than the background level before the maintenance. Some of the discharge activity previously included in the background level actually originated from this pad. The complete elimination of the defect took it away along with the background level.

[0044] It is worth adding that the cross-sectional comparison within the same room provided reverse evidence in this incident: during the three-week evolution period, the statistics of the two cabinets adjacent to the cabinet remained stable, and the estimated power frequency cycle was consistent with that of the cabinet, while the discharge statistics did not show any synchronous increase. This indicates that the incident was indeed due to a defect inside the cabinet rather than environmental interference. This spatial information does not require any coordination between devices and can be obtained simply by the platform collecting summaries from various points.

[0045] As a comparative example 1, the same batch of pulses from that night were folded directly according to the nominal period of 20 milliseconds without periodic search. Due to the continuous accumulation of the 0.018Hz frequency deviation at a rate of 6.5 degrees per second, the two clusters within the 10-second window were each widened by about 65 degrees, and the magnitude of the second composite vector dropped from 0.922 to 0.723. Although it could still pass the threshold, the discrimination margin had been eroded. If the window was extended to 60 seconds to improve statistical confidence, the cumulative drift reached 390 degrees, and the two clusters were completely flattened, so that the magnitude of the second composite vector was only 0.050. According to the same threshold, all discharges would be missed. However, this method, under the 60-second long window, relies on re-estimating the 2-second segments one by one and aligning and merging them according to their respective phase references. The magnitude of the second composite vector is maintained at 0.906. Thus, both long window accumulation and spectral clarity are achieved. This is the value of the segment re-estimation process.

[0046] As a comparative example 2, retaining the periodic search but canceling the double angle and using only the first synthetic vector magnitude for synchronization scoring: the score of the night discharge window was only 0.014, which is lower than any reasonable threshold, resulting in the entire window missing the floating discharge of the double cluster morphology; since the double cluster morphology such as floating and surface-mounted discharges in the switch cabinet is a common defect type with great harm, using only the circle with a double angle is almost unusable in this scenario, so the introduction of the double angle is not icing on the cake but a matter of success or failure.

[0047] Looking at the two comparative examples together reveals another layer of relationship: the former fails due to an inaccurate phase reference—because the period is not aligned, even the best concentration scale cannot measure concentration, resulting in distortion of both the spectrum and the score; the latter fails due to a mismatched scale—even if the period is aligned, the scale may not detect the dual clusters, leading to an accurate reference but a flawed judgment. Period search and double-angle measurement each address one of the failure points. This method combines both to be effective in real-world applications; separating either half would not work. This also explains the mechanistic relationship between the two processes.

[0048] The pulse count, score, and frequency values ​​in the above embodiments are illustrative examples based on the actual measured load patterns on site. The specific values ​​may vary for different sites, but the clear contrast between the cancellation of dual-cluster discharge at one-fold angle, its manifestation at two-fold angle, and the blurring of fixed-period folding with window length and segment re-estimation remains unchanged.

[0049] In terms of deployment, all new processing in this method resides in the monitoring device as firmware. Existing transient ground voltage monitoring devices already connected to the grid can acquire the capabilities of this method through remote firmware upgrades, without the need for on-site modifications. New monitoring points can be installed using conventional methods. Neither scenario involves power outages or wiring work. Compared to the external reference route, if a distribution room with 40 cabinets were to have synchronous lines laid in each cabinet or synchronous transceiver hardware added, the wiring would need to be carried out in batches in conjunction with power outages. However, with this method, the points can be magnetically attached to each cabinet, and the entire room can be covered within one working day. This deployment difference will be magnified many times over in the hundreds of distribution rooms in a county-level distribution network.

[0050] Figure 8 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Figure 8 As shown, the electronic device 80 includes a processor 81 and a memory 82. The processor 81 can be a microcontroller or digital signal processor in the monitoring device. The computational complexity of the remainder and vector accumulation during periodic scanning is proportional to the number of pulses and candidates within the window. The calculation of over a hundred pulses in a single window and hundreds of candidates across two scanning stages can be completed within one second on a microcontroller with a clock frequency of 100MHz. The memory 82 includes one or more combinations of volatile and non-volatile memory, storing a computer program. When executed by the processor 81, this computer program implements the online partial discharge monitoring method for high-voltage switchgear described in any of the above implementations of this invention. Optionally, the electronic device 80 also includes an input device 83 and an output device 84, such as a sensor interface and an IoT communication module, with each component interconnected via a bus. For simplicity, Figure 8 Only some components relevant to this invention are shown.

[0051] In addition to the methods, systems, and electronic devices described above, embodiments of the present invention can also be computer program products, comprising computer program instructions that, when executed by a processor, cause the processor to perform the steps in the online monitoring method for partial discharge in high-voltage switchgear described in the foregoing sections of this specification. Embodiments of the present invention can also be computer-readable storage media storing computer program instructions that, when executed by a processor, cause the processor to perform the steps in the aforementioned methods. The computer-readable storage medium can be random access memory, read-only memory, flash memory, or any suitable combination thereof.

[0052] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions described in the various embodiments of the present invention.

Claims

1. A method for online monitoring of partial discharge in high-voltage switchgear, characterized in that, include: Step 101: The monitoring device installed on the high-voltage switchgear cabinet detects the discharge pulse through a transient ground voltage sensor and records the arrival time and amplitude of each discharge pulse. The arrival time is given by the local clock of the monitoring device. The monitoring device is not connected to the power frequency voltage reference. Step 102: For the arrival time of each discharge pulse within the sliding statistical window, select candidate folding periods one by one within a preset search interval covering the nominal power frequency period. Take the remainder of each arrival time with respect to the candidate folding period and convert it into a folding phase angle. Calculate the first and second composite vector magnitude values ​​weighted by the folding phase angle and its double angle, respectively. Take the one that is not lower than the other as the synchronization score of the candidate folding period. Determine the candidate folding period with the highest synchronization score among all candidates as the estimated power frequency period, and use the folding phase angle under the estimated power frequency period as the phase reference for each discharge pulse. Step 103: When the synchronicity score corresponding to the estimated power frequency cycle is not lower than the preset synchronicity threshold, it is determined that there is power frequency related discharge within the sliding statistical window, and a phase-resolved discharge spectrum is generated according to the phase reference and amplitude accumulation. When the synchronization score is lower than the preset synchronization threshold, the discharge pulse in the sliding statistical window is determined to be asynchronous interference and is not included in the discharge statistics. Step 104: For the time periods where power frequency-related discharges are determined to exist, the discharge pulse rate and amplitude quantile are statistically analyzed. The time periods that meet the preset warning conditions are designated as warning level, the time periods that only meet the preset attention conditions are designated as attention level, and the remaining time periods are designated as normal level. The monitoring device reports to the monitoring platform in stages via the Internet of Things communication link: the early warning level reports the phase-resolved discharge spectrum and discharge pulse details, the attention level reports the spectrum characteristics, and the normal level reports the statistical summary according to the preset reporting cycle. The monitoring platform reviews the reports and generates handling suggestions.

2. The method according to claim 1, characterized in that, The preset search interval covers 0.998 to 1.002 times the nominal power frequency cycle; Candidate folding periods are selected by two-level scanning: first, a coarse scan is performed within a preset search interval at the first scanning interval to obtain the initial selected period, and then a fine scan is performed within a fine scan interval centered on the initial selected period at a second scanning interval lower than the first scanning interval. The sliding statistical window is divided into multiple segments according to the preset segment duration, and step 102 is re-executed for each segment to track the slow drift of the power grid frequency.

3. The method according to claim 1, characterized in that, When calculating the magnitude of the first and second composite vectors, the weight of each discharge pulse is taken as the proportion of the amplitude of that discharge pulse to the sum of the amplitudes of all discharge pulses within the sliding statistical window.

4. The method according to claim 1, characterized in that, After generating a phase-resolved discharge spectrum, the number of phase clusters, the spacing between phase clusters, and the phase width within the clusters are extracted as spectral features. The spectral features are then compared with pre-stored discharge type templates to determine the discharge type, which must at least distinguish between tip discharge, floating discharge, and surface discharge.

5. The method according to claim 1, characterized in that, After determining that it is asynchronous interference, the cycle folding is performed again for each arrival time within the preset interference search interval that deviates from the nominal power frequency cycle. When there is a folding cycle with a synchronization score not lower than the preset interference threshold, the reciprocal of the folding cycle is determined as the interference repetition frequency, and the interference repetition frequency is reported to the monitoring platform for interference source investigation.

6. The method according to claim 1, characterized in that, The amplitude quantile is taken as the 95th percentile of the amplitude of all discharge pulses within the sliding statistical window; The preset warning conditions and preset attention conditions are defined according to the growth rate of the discharge pulse rate and amplitude quantile relative to their respective baseline values. The baseline value is the median of the corresponding statistics within the most recent baseline period.

7. The method according to claim 1, characterized in that, The preset synchronization threshold ranges from 0.25 to 0.

6.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 7.

9. A partial discharge online monitoring system for high-voltage switchgear, the system being used to implement the partial discharge online monitoring method for high-voltage switchgear according to any one of claims 1-7, characterized in that, This includes monitoring devices installed on the cabinet of high-voltage switchgear and a monitoring platform connected to the monitoring devices via an Internet of Things (IoT) communication link; The monitoring device includes: a pulse acquisition module, used to detect discharge pulses and record their arrival time and amplitude via a transient ground voltage sensor, with the arrival time provided by a local clock; The period search module is used to scan candidate folding periods within a preset search interval covering the nominal power frequency period. The synchronization score is determined by the magnitude of the amplitude-weighted composite vector of the folding phase angle and the magnitude of the amplitude-weighted composite vector of its double angle being not lower than the other. This locks the estimated power frequency period and the phase reference of each discharge pulse. The discharge discrimination module is used to compare the synchronization score with a preset synchronization threshold, distinguish between power frequency-related discharge and asynchronous interference, and generate a phase-resolved discharge spectrum when it is determined that power frequency-related discharge exists. The statistical reporting module is used to statistically analyze the discharge pulse rate and amplitude quantile and classify the warning level, attention level or normal level, and report the spectrum, characteristic quantity or statistical summary to the monitoring platform according to the level. The monitoring platform includes a review module and a handling module. The review module is used to review the reported data, and the handling module is used to generate handling suggestions and track the closed loop.

10. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program that, when executed by the processor, implements the method of any one of claims 1 to 7.

Images