Transformer partial discharge interference suppression method and system based on reference channel time domain gating and correlation analysis

Abstract

The invention relates to the technical field of power equipment state monitoring, in particular to a transformer partial discharge interference suppression method and system based on reference channel time domain gating and correlation analysis, and the method comprises the steps: synchronously collecting signal sequences of a main sensor and a reference sensor, and respectively carrying out the pulse detection to obtain a pulse event set; traversing a main channel pulse, if a pulse meeting a time correlation condition exists in a reference channel, determining that interference exists and removing the interference, otherwise, marking that the interference exists as suspected internal partial discharge; performing waveform correlation discrimination on the suspected pulse, if a normalized cross correlation coefficient of a main channel waveform and a reference channel waveform at a corresponding moment is greater than a preset threshold, determining interference, otherwise, determining real partial discharge; and finally collecting all real partial discharge pulses to form a partial discharge data set. Through a dual discrimination mechanism of time domain gating and correlation analysis, external interference is effectively eliminated while high sensitivity is kept, the defects of a traditional filtering method are overcome, and resource dependence of a complex algorithm is avoided.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring technology, specifically to a method and system for suppressing transformer partial discharge interference based on reference channel time-domain gating and correlation analysis. Background Technology

[0002] As a core component of the power system, the insulation condition of transformers directly affects the reliability and safety of the power grid. Partial discharge is one of the main manifestations of transformer insulation degradation, and online monitoring of partial discharge has become an important means of achieving predictive maintenance and preventing sudden failures. Among existing non-intrusive monitoring methods, the ultra-high frequency (UHF) method has been widely used in transformer partial discharge monitoring due to its advantages such as high sensitivity and relatively strong anti-interference ability, providing an effective technical approach for early detection of insulation defects.

[0003] However, UHF still faces severe challenges in the actual operation environment of substations. The main problems are that various strong interferences cause frequent false alarms and missed alarms in the monitoring system. On the one hand, the pulse signals generated by corona discharge, switching operation, etc. are highly similar to the real partial discharge in terms of time and frequency domain characteristics, and traditional filtering methods are difficult to effectively distinguish them. On the other hand, the real partial discharge signal is extremely weak after being attenuated by the transformer box and is often completely submerged by background noise, resulting in an extremely low signal-to-noise ratio.

[0004] Existing suppression methods mostly rely on complex signal processing (such as wavelet denoising, empirical mode decomposition, etc.) or artificial intelligence algorithms, which have limitations such as high consumption of computing resources, high requirements for model training data, and high hardware costs, making it difficult to achieve low-cost, high-reliability large-scale deployment in the field of power equipment monitoring. Summary of the Invention

[0005] To address the above technical problems, this invention provides a technical solution for a transformer partial discharge interference suppression method and system based on reference channel time-domain gating and correlation analysis.

[0006] The technical problem solved by this invention can be achieved by the following technical solution: A transformer partial discharge interference suppression method based on reference channel time-domain gating and correlation analysis, comprising: step S1, synchronously acquiring the main channel signal sequence of the main monitoring sensor and the reference channel signal sequence of the reference sensor; step S2, performing pulse detection on the main channel signal sequence and the reference channel signal sequence respectively to obtain the main channel pulse event set and the reference channel pulse event set; step S3, traversing the main channel pulses of the main channel pulse event set, if the main channel pulse is in the reference channel pulse event set... If a reference channel pulse that satisfies the time correlation condition exists, the main channel pulse is treated as an interference pulse and discarded; otherwise, the main channel pulse is marked as a suspected internal partial discharge pulse, and step S4 is executed; in step S4, the suspected internal partial discharge pulse is filtered for correlation, and it is determined whether the normalized cross-correlation coefficient between its main channel waveform and the reference channel waveform at the corresponding time is greater than a preset correlation threshold. If so, it is marked as an interference pulse; otherwise, it is marked as a real internal partial discharge pulse; in step S5, all events marked as real internal partial discharge pulses are collected to obtain the final partial discharge dataset.

[0007] Preferably, step S2 includes: step S21, calculating the background noise levels of the main channel signal sequence and the reference channel signal sequence respectively, to obtain the background noise levels of the main channel and the reference channel; step S22, setting the detection thresholds of the main channel and the reference channel respectively, as expressed by the following formula: , , in, The main channel detection threshold. As a reference channel detection threshold, This is an empirical coefficient. Main channel background noise level For reference channel background noise level; Step S23: Detect pulses in the main channel signal sequence and the reference channel signal sequence whose amplitude exceeds their corresponding preset detection thresholds, respectively, to obtain the main channel pulse event set and the reference channel pulse event set, as specifically expressed by the formula: , , in, Main channel pulse event set, For the reference channel pulse event set, The peak amplitude of the main channel pulse. The peak amplitude of the reference channel pulse. The peak arrival time of the main channel pulse is [time of arrival]. The peak arrival time of the reference channel pulse is denoted as .

[0008] Preferably, in step S21, a stable segment without pulse activity is selected in the main channel signal sequence and the reference channel signal sequence, and the root mean square value of the signal amplitude in the stable segment is calculated, which is used as the background noise level of the main channel and the background noise level of the reference channel, respectively.

[0009] Preferably, in step S3, the time correlation condition is: , in, The peak arrival time of the main channel pulse is [time of arrival]. The peak arrival time of the reference channel pulse. This is a preset time-controlled window.

[0010] Preferably, the preset time gate window is expressed by the formula: , in, The distance between the main monitoring sensor and the reference sensor is [missing information]. The speed at which electromagnetic waves propagate in the air. For margin.

[0011] Preferably, step S4 includes: step S41, extracting the complete waveform segment of the suspected internal partial discharge pulse in the main channel; step S42, extracting the reference channel waveform segment at the same time as the complete waveform segment of the main channel from the reference channel signal sequence; step S43, calculating the normalized cross-correlation coefficient between the complete waveform segment of the main channel and the reference channel waveform segment, specifically expressed by the formula:

[0012] in, To normalize the cross-correlation coefficient, The main channel complete waveform segment, For reference channel waveform segment, For delay time; Step S44: Determine whether the normalized cross-correlation coefficient is greater than a preset correlation threshold. If yes, mark the suspected internal partial discharge pulse as an interference pulse. If no, mark the suspected internal partial discharge pulse as a real internal partial discharge pulse.

[0013] Preferably, the normalized cross-correlation coefficient ranges from [-1, 1].

[0014] It also includes a transformer partial discharge interference suppression system based on reference channel time-domain gating and correlation analysis, which implements the transformer partial discharge interference suppression method based on reference channel time-domain gating and correlation analysis as described above. The system includes: a main monitoring sensor installed at a monitoring point on the transformer housing to capture electromagnetic signals, including internal partial discharge signals and external interference signals; a reference sensor installed outside the transformer housing facing the external space to capture spatial electromagnetic interference signals from outside the transformer; and a synchronous data acquisition and processing unit connected to the main monitoring sensor and the reference sensor to synchronously acquire signals from both sensors and execute the transformer partial discharge interference suppression method.

[0015] Preferably, the main monitoring sensor and the reference sensor are UHF sensors of the same model.

[0016] Preferably, the synchronous data acquisition and processing unit is a high-speed data acquisition card with at least two channels.

[0017] Beneficial effects: This invention, through a dual discrimination mechanism of reference channel time-domain gating and correlation analysis, effectively eliminates external interference pulses while retaining high sensitivity. It overcomes the shortcomings of traditional filtering methods in identifying similar signals and avoids the dependence of complex algorithms on computing resources and training data, providing a low-cost and highly reliable interference suppression solution for transformer partial discharge monitoring. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a flowchart illustrating step S2 of the present invention; Figure 3 This is a flowchart illustrating step S4 of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0021] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0022] Reference Figure 1 This invention provides a transformer partial discharge interference suppression method based on reference channel time-domain gating and correlation analysis, comprising: step S1, synchronously acquiring the main channel signal sequence of the main monitoring sensor. and the reference channel signal sequence of the reference sensor Step S2, respectively process the main channel signal sequence and the reference channel signal sequence Pulse detection is performed to obtain the main channel pulse event set. and reference channel pulse event set Step S3: Traverse the main channel pulse event set. Main channel pulse If the main channel pulse The reference channel pulse event set There exists a reference channel pulse that satisfies the time correlation condition. Then the main channel pulse Interference pulses are discarded; conversely, the main channel pulses are discarded. The pulse is marked as a suspected internal partial discharge pulse, and step S4 is executed; in step S4, the suspected internal partial discharge pulse is subjected to filtering correlation identification, and the normalized cross-correlation coefficient between its main channel waveform and the reference channel waveform at the corresponding time is determined. Is it greater than the preset correlation threshold? If yes, mark it as an interference pulse; otherwise, mark it as a real internal partial discharge pulse. Step S5: Collect all events marked as real internal partial discharge pulses. The final partial discharge dataset is obtained.

[0023] Specifically, in this embodiment of the invention, in response to the problems of frequent false alarms and missed alarms caused by strong interference at substation sites and the difficulty in large-scale deployment of complex algorithms, a spatial discrimination system is constructed by setting up a reference sensor, and a two-layer filtering mechanism of "time-domain gating initial screening + waveform correlation fine judgment" is adopted. This avoids the dependence on complex signal processing and big data AI models, and achieves effective suppression of pulse interference in extremely low signal-to-noise ratio environments, significantly improving monitoring reliability, while greatly reducing the computational complexity and hardware cost of the system.

[0024] Accordingly, the physical basis of the "time-domain gating initial screening" is as follows: interference pulses originating from outside the transformer have their signal sources located in open spaces. The electromagnetic waves generated travel through the air at near the speed of light, either directly or after a small amount of reflection, and arrive at the main monitoring sensor SM and the reference sensor SR, which is specifically arranged facing outwards, almost simultaneously or with a slight fixed delay. Therefore, the arrival times of the pulse peaks recorded by the two sensors will be highly similar. However, the actual internal partial discharge pulses have their signal sources hidden deep inside the transformer tank. The electromagnetic waves must penetrate and circulate through complex oil-paper barriers, windings, and tank walls before being captured by the SM. This process results in significant attenuation and delay. Furthermore, the path for the signal to continue leaking to the outside and be effectively received by the SR is longer and has extremely high losses, causing the SR to typically fail to detect effective pulses that are temporally correlated with the pulses inside the SM.

[0025] The purpose of the "waveform correlation precision judgment" is to identify residual interference after the initial screening, such as interference that is not precisely synchronized in the time domain but still has a certain correlation, or interference that is weakly coupled to the reference sensor through the gaps in the enclosure. This method calculates the normalized cross-correlation coefficient of the waveform segments of the suspected pulses on the main and reference channels. It takes advantage of the fact that the waveform similarity of external interference on the two channels is necessarily higher than that of the actual internal partial discharge, and finally eliminates the pseudo pulses with correlation coefficients higher than the threshold, thereby ensuring the accuracy of the final partial discharge dataset.

[0026] Furthermore, this proposal employs a sampling rate of no less than 1 GSa / s for strictly time-aligned parallel sampling of the main monitoring sensor SM and the reference sensor SR, ensuring that the amplitude-time trajectories of the same electromagnetic event on both channels can be accurately compared. After acquisition, "time-domain gating initial screening" and "waveform correlation fine judgment" are performed first to form high-confidence true internal partial discharge pulses, and then all events confirmed as "true internal partial discharge pulses" are... The collected data forms a final, clean partial discharge dataset. This dataset can be directly used to plot PRPD (Phase Resolved Partial Discharge) maps, track discharge quantity changes, and further use pattern recognition algorithms to diagnose defect types, providing reliable data support for transformer condition assessment and predictive maintenance.

[0027] As a preferred embodiment of the present invention, refer to Figure 2 Step S2 includes: Step S21, calculating the main channel signal sequence respectively. and the reference channel signal sequence The background noise level of the main channel is obtained from the background noise level. and the background noise level of the reference channel Step S22: Set the detection threshold for each main channel. and reference channel detection threshold The specific formula is expressed as follows: , , in, This is an empirical coefficient; Step S23: Detect the main channel signal sequence respectively. and the reference channel signal sequence Pulses with a mid-amplitude value exceeding their corresponding preset detection threshold are used to obtain the main channel pulse event set. and the reference channel pulse event set The specific formula is expressed as follows: , , in, The peak amplitude of the main channel pulse. The peak amplitude of the reference channel pulse. The peak arrival time of the main channel pulse is [time of arrival]. The peak arrival time of the reference channel pulse is denoted as .

[0028] Specifically, considering the time-varying characteristics of the electromagnetic environment and the gain differences of different sensor channels, using a uniform fixed threshold may lead to missed detection or false detection in a certain channel. In this embodiment of the invention, an adaptive threshold method based on the background noise level of each channel is adopted, that is, the detection threshold is calculated and set independently for each channel, thereby ensuring that both channels can achieve optimal pulse detection sensitivity and reliability under their own signal-to-noise ratio conditions.

[0029] This method allows the system to first accurately assess the instantaneous noise floor of each channel, and then set a reasonable and adjustable detection threshold for each channel relative to its own noise level. The empirical coefficient k is usually taken as 3 to 5 to achieve a balance between the detection probability and the false alarm probability. Finally, all significant pulse events are screened out from the two channel signals in parallel, and their key features (such as peak amplitude and peak arrival time) are recorded, providing a structured and comparable data basis for subsequent interference discrimination.

[0030] More specifically, considering that the background noise may contain random pulses or fluctuations with extremely small amplitudes, in order to ensure the accuracy of the noise level estimation, in this embodiment of the invention, in step S21, the noise level is determined from the main channel signal sequence... and the reference channel signal sequence Each stable region without pulse activity is selected (e.g., determined by observation or sliding window variance calculation), and the root mean square (RMS) value of the signal amplitude within the stable region is calculated to serve as the background noise level of the main channel. and the background noise level of the reference channel .

[0031] This method effectively eliminates the interference of the pulse event itself on noise estimation, thereby enabling the set detection threshold to truly reflect the noise state of the channel and improving the accuracy of pulse detection.

[0032] In a preferred embodiment of the present invention, the time correlation condition in step S3 is: , in, The peak arrival time of the main channel pulse is [time of arrival]. The peak arrival time of the reference channel pulse. This is a preset time-controlled window.

[0033] Specifically, considering that the external interference signal source is located in the external space, the electromagnetic waves it generates arrive at the main monitoring sensor SM and the reference sensor SR, which is specifically oriented outward, almost simultaneously or sequentially, in the air at a path close to the speed of light, directly or after a small amount of reflection. Since the SR is specifically oriented outward, it has a higher directional reception sensitivity to external interference, and its pulse arrival time is usually slightly earlier than or very close to the pulse arrival time of the SM. On the other hand, the internal partial discharge signal source is located inside the transformer, and its electromagnetic waves must first penetrate and circulate through the complex internal insulation structure of the transformer (oil-paper barrier, windings, etc.) before being captured by the SM. This process has already brought about significant attenuation and delay. If the signal is to continue to propagate to the outside of the transformer and be effectively received by the SR, it must penetrate the metal box wall again, resulting in extremely high path loss and extremely weak signal. As a result, the SR usually cannot detect the effective pulse corresponding to the pulse inside the SM.

[0034] Based on this physical principle, in this embodiment of the invention, by traversing the main channel pulse set... Each pulse in and in the reference channel pulse set The search examines whether a corresponding pulse that satisfies the time-gated condition exists. And for the pulse that finds its "time companion" in the reference channel. It is determined to be external interference and is eliminated; for pulses whose corresponding pulses cannot be found in the reference channel... If so, it is initially identified as a suspected internal partial discharge pulse and retained for the next processing stage.

[0035] This operation fully utilizes the essential differences between internal and external signal sources in their spatial propagation paths, and can efficiently filter out most strong external interferences (such as corona discharge, external spark discharge, switching operations, etc.) based on a clear physical mechanism. While significantly reducing the amount of data to be processed, it lays a reliable foundation for the final accurate identification of the actual partial discharge.

[0036] In a preferred embodiment of the present invention, the preset time gate window is expressed by the formula: , in, The distance between the main monitoring sensor and the reference sensor is [missing information]. The speed at which electromagnetic waves propagate in the air. For margin.

[0037] Specifically, because electromagnetic waves may encounter non-direct paths such as reflection and diffraction during propagation, and the measurement system itself has slight deviations in clock synchronization accuracy and response time, if only the theoretical direct wave time difference is used... Using a threshold can easily lead to missed correlations. Therefore, in this embodiment of the invention, an empirical time margin is introduced. (Typically set to 5~10 nanoseconds), set the time gating window to... This ensures that pulse pairs from the same external interference source, even with slight time differences due to path or system deviations, can be reliably captured. This effectively avoids misjudging real interference as internal partial discharge due to overly stringent gating conditions, significantly improving the robustness and reliability of the first-stage interference filtering.

[0038] As a preferred embodiment of the present invention, refer to Figure 3 Step S4 includes: Step S41, extracting the complete waveform segment of the suspected internal partial discharge pulse in the main channel of the main channel. Step S42: Extract the complete waveform segment of the main channel from the reference channel signal sequence. Reference channel waveform segment at the same time Step S43: Calculate the complete waveform segment of the main channel. With the reference channel waveform segment Normalized cross-correlation coefficient The specific formula is expressed as follows:

[0039] in, For delay time; Normalized cross-correlation coefficient The value range is [-1, 1]. The closer the value is to 1, the more similar the shapes of the two waveform segments are. Step S44: Determine the normalized cross-correlation coefficient. Is it greater than the preset correlation threshold? If yes, the suspected internal partial discharge pulse is marked as an interference pulse; otherwise, the suspected internal partial discharge pulse is marked as a real internal partial discharge pulse.

[0040] Specifically, in order to accurately identify residual interference that failed to be effectively eliminated by the first-stage time-domain gating due to complex propagation paths or weak signals (e.g., some interference that originates from the outside but has a slightly different arrival time, or whose signals do not exceed the detection threshold in the reference channel but still have waveform correlation), in this embodiment of the invention, waveform correlation analysis is further performed on each "suspected internal partial discharge pulse". The specific implementation process is as follows: first, the complete waveform segments of the main and reference channels at the pulse occurrence time are captured synchronously, and then the maximum normalized cross-correlation coefficient between the two under all possible relative delays is calculated. ,like Greater than the preset correlation threshold ,For example =0.5 indicates that the pulse has a similar waveform in both channels, which is consistent with the characteristics of external interference, so it is finally identified as an interference pulse; conversely, it indicates that the pulse has a unique response in the main channel and is independent of the reference channel, so it is finally identified as a real internal partial discharge pulse.

[0041] It is worth mentioning that, considering that the first-stage time-domain gating might miss some weak interference signals that, although not exceeding the detection threshold, are related to waveform characteristics, or interference signals that remain after distortion after penetrating the enclosure, even if the amplitude of the reference channel waveform is lower than the preset threshold, the waveform segment at the same moment as the main channel pulse is forcibly extracted, and its normalized cross-correlation coefficient is calculated. This design can more comprehensively capture those weak but morphology-related interference signals, effectively preventing such interference from being "missed" due to amplitude limitations and misjudged as real partial discharge, thereby further improving the robustness of interference suppression in low signal-to-noise ratio environments and the accuracy of the final identification results.

[0042] The present invention also includes a transformer partial discharge interference suppression system based on reference channel time-domain gating and correlation analysis, which implements the transformer partial discharge interference suppression method based on reference channel time-domain gating and correlation analysis as described above. The system includes: a main monitoring sensor SM, installed at a monitoring point on the transformer tank, for capturing electromagnetic signals, including internal partial discharge signals and external interference signals; a reference sensor SR, installed outside the transformer tank and facing the external space, for capturing spatial electromagnetic interference signals from outside the transformer; and a synchronous data acquisition and processing unit (Unit), connected to the main monitoring sensor SM and the reference sensor SR, for synchronously acquiring signals from both sensors and executing the transformer partial discharge interference suppression method.

[0043] Specifically, in this embodiment of the invention, the main monitoring sensor SM is preferably installed on the transformer tank wall at a location with high sensitivity to internal electromagnetic signals, such as an observation window, handhole, or reserved dedicated measurement flange; the reference sensor SR is installed near the transformer tank and at a certain spatial distance from the main sensor, for example, the spacing d is set to 0.5~2 meters, and its sensing direction is away from the transformer body and towards the external open space, so as to maximize its ability to capture external interference, while minimizing the direct coupling of internal transformer signals to it, thereby constructing a spatially separated "main monitoring-interference reference" dual-channel sensing system.

[0044] In a preferred embodiment of the present invention, the main monitoring sensor and the reference sensor are UHF sensors of the same model.

[0045] Specifically, in order to eliminate the systematic deviation caused by the difference in frequency response characteristics of the sensors themselves to the subsequent waveform correlation analysis, in this embodiment of the invention, the main monitoring sensor SM and the reference sensor SR must be of the same model, or at least have similar frequency bands (e.g., both covering the main UHF partial discharge signal band of 300MHz ~ 1500MHz) and sensitivity of ultra-high frequency (UHF) sensors, so as to ensure that the two channels can generate comparable response waveforms to the same electromagnetic event, providing a consistent signal basis for subsequent time domain correlation and correlation identification.

[0046] In a preferred embodiment of the present invention, the synchronous data acquisition and processing unit is a high-speed data acquisition card with at least two channels.

[0047] Specifically, in this embodiment of the invention, the core of the synchronous data acquisition and processing unit is a high-speed data acquisition card with at least two analog input channels. Each channel must have a nanosecond-level high-precision synchronous triggering and sampling clock, and the sampling rate is usually not less than 1 GSa / s, so as to ensure that the time domain details of the UHF pulse signal can be accurately captured and to provide a strictly synchronized data foundation for subsequent pulse arrival time comparison and waveform cross-correlation calculation.

[0048] More specifically, the synchronous data acquisition and processing unit also integrates a processing device for running all steps of the interference suppression algorithm. In a typical deployment case, the high-speed data acquisition card is installed in an industrial-grade embedded computer (i.e., an industrial control computer). This industrial control computer serves as a field monitoring terminal integrating data acquisition, processing, and storage, and it is pre-installed with and runs the interference suppression algorithm software that implements the method described in this invention. This integrated hardware design allows the system to complete the entire process from signal acquisition and interference filtering to partial discharge event extraction directly at the equipment site, without relying on a remote server for heavy data processing. This not only significantly reduces network transmission bandwidth requirements but also greatly improves the real-time performance and operational reliability of the entire monitoring system.

[0049] In summary, this invention successfully extracts the true partial discharge pulse inside the transformer under strong noise background by constructing a dual-channel spatial perception system of "main monitoring-interference reference" and innovatively adopting a two-layer progressive interference filtering logic that combines "time-domain gating initial screening" and "waveform correlation fine judgment".

[0050] Compared with the prior art, the present invention has achieved the following breakthroughs: 1) A fundamental improvement in interference identification accuracy: Through the dual physical discrimination mechanism of time-domain arrival time correlation and waveform morphology correlation, this invention can accurately distinguish between internal and external pulse signals that are highly similar in time-frequency characteristics, fundamentally solving the problem of high misjudgment rate of traditional single sensor systems or simple filtering algorithms, and achieving extremely low misjudgment rate elimination of external interference such as corona and switching operations.

[0051] 2) This method achieves dual optimization in monitoring reliability, reducing both false alarms and missed alarms: It not only significantly reduces the false alarm rate of the system by effectively eliminating interference, enabling maintenance personnel to focus on real internal defect alarms; but also, in a clean signal background, the system can safely set a lower detection threshold, thereby capturing early, weak partial discharge signals that were previously submerged by noise, significantly improving monitoring sensitivity and effectively preventing missed alarms caused by weak signals.

[0052] 3) Significant advantages in system cost and deployment convenience: The core hardware of this solution only requires the addition of one UHF sensor of the same type and one synchronous acquisition channel. Compared with solutions that rely on high-performance computing platforms or complex multiphysics monitoring systems, the incremental hardware cost is extremely low. Its algorithm is based on clear physical principles, has low computational complexity, does not require large-scale data training, and can run stably on ordinary industrial control computers, greatly reducing the total cost of ownership and deployment threshold of the system.

[0053] 4) It demonstrates excellent performance in terms of applicability and robustness: This method does not rely on a specific, pre-established interference sample library. Its discrimination criterion is the universal physical difference of internal and external signals in the spatial propagation path. Therefore, it has good adaptability and robustness to various unknown and variable interference types in different substation environments, and is easy to promote and apply quickly and effectively in different sites.

[0054] 5) It has generated a multiplier effect in terms of data value and system efficiency: The high-confidence partial discharge data output provides a reliable data foundation for subsequent accurate PRPD spectrum analysis, discharge type identification and fault trend prediction, thereby greatly improving the decision support capability and practical economic benefits of the entire condition monitoring and early warning system, and strongly supporting the construction of predictive maintenance and intelligent operation and maintenance system for power equipment.

[0055] The following is a detailed description using a typical embodiment: Example: An oil-immersed power transformer (model SFZ11-50000 / 110) in a 110kV substation experiences interference primarily from the following sources: continuous corona discharge on the high-voltage bushing (phase A) and strong transient pulse interference generated during operation by a disconnecting switch approximately 10 meters away from the transformer. The specific partial discharge interference suppression process is as follows: 1. System Deployment and Parameter Configuration A built-in UHF sensor with a frequency band of 300MHz-1.5GHz is installed at the dielectric window reserved on the transformer box wall as the main monitoring sensor SM. This position is close to the winding and has high sensitivity to receiving internal partial discharge signals; and LMR-400 low-loss coaxial cable is used for signal extraction.

[0056] Meanwhile, a UHF sensor of the same model as the main sensor was selected as the reference sensor SR. An independent, non-metallic (such as epoxy resin or nylon) mounting bracket was made for the reference sensor SR and fixed on a cement base 1.5 meters away from the transformer box wall. The signal was also led out using an LMR-400 coaxial cable.

[0057] The two coaxial cables are cut to the exact same length (e.g., 8.0 meters each) to ensure that the transmission delay introduced by the cables themselves is exactly the same, simplifying subsequent timestamp calibration.

[0058] Furthermore, a synchronous data acquisition and processing unit (Unit) is configured, which integrates a dual-channel synchronous data acquisition card and a processing unit. The key parameters of the dual-channel synchronous data acquisition card are configured as follows: sampling rate of 2 GSa / s (2 billion samples per second), vertical resolution of 12-bit, and inter-channel synchronization accuracy of less than 1 ns. The acquisition card is installed in an industrial-grade embedded computer (industrial control computer) as the processing unit, which is pre-installed with the interference suppression algorithm software of this invention. The signal of the main monitoring sensor SM is connected to the main channel CH1 of the acquisition card, and the signal of the reference sensor SR is connected to the reference channel CH2. At the same time, a 50Hz AC voltage signal is drawn from the secondary side of the potential transformer (PT) to the external trigger input port of the acquisition card as a precise phase reference for generating the phase-resolved partial discharge (PRPD) spectrum.

[0059] 2. Parameter Initialization In the algorithm software, the following initial parameters are set according to the field environment and hardware configuration: Main channel detection threshold and reference channel detection threshold The system sets the empirical coefficient k for the pulse detection threshold to 4 and automatically calculates the root mean square value of the background noise to obtain the background noise level of the main channel. and the background noise level of the reference channel and calculate , Time-controlled window : =20 ns, which is based on the distance d=1.5m between the main monitoring sensor and the reference sensor, and the theoretical propagation delay. =5ns, additional margin =15ns; Preset correlation threshold : Set to an empirical value of 0.6, which is the normalized cross-correlation coefficient of the two waveforms. If the value exceeds 0.6, they are considered to be of the same origin; waveform extraction window length: 150 nanoseconds, starting from the pulse peak point. To begin, take 50 nanoseconds forward and 100 nanoseconds backward to form a complete pulse waveform segment.

[0060] 3. Example of interference signal processing After the system starts running, the software in the industrial control computer processes the acquired data stream in real time. The following simulates the occurrence and processing of two typical events: Event A: External high-voltage bushing corona interference a1. Event Occurrence and Signal Acquisition: Due to surface contamination or burrs, the A-phase bushing on the high-voltage side of the transformer experiences continuous corona discharge under operating voltage, generating a series of UHF electromagnetic wave pulses.

[0061] a2. Signal Acquisition: The interference pulse was radiated through space and was almost simultaneously acquired by the main monitoring sensor SM and the reference sensor SR. Since SR was directly facing the bushing and unobstructed, the received signal energy was stronger. Specifically, the main channel CH1 detected one pulse. Its characteristics are: peak amplitude mV, arrival time ns; Reference channel CH2 detected a stronger pulse because it is directly facing the interference source and has a direct path. Its characteristic is peak amplitude mV, arrival time ns (slightly earlier than the main channel due to a better path).

[0062] a3. First-stage filtering (time-domain gating): The algorithm extracts the main channel impulse event set. Extract the main channel pulse from within and in the reference channel pulse event set Search within the time window ,Right now The pulse within; reference pulse successfully found. Its arrival time If ns is within the time window and meets the time correlation condition, the algorithm will immediately... This pulse is denoted as an interference pulse. They are discarded directly and no further processing is carried out.

[0063] Event B: Actual partial discharge occurs in the transformer's internal windings. b1. Incident Occurrence: Due to weak insulation inside the transformer's high-voltage winding, an electric field concentration occurs, triggering an electric tree-like discharge and generating a weak UHF partial discharge pulse.

[0064] b2. Signal Acquisition: After undergoing multiple refractions, reflections, and attenuations through complex paths such as oil gaps, insulating paperboard, and windings inside the transformer, the partial discharge signal finally escapes through the dielectric window in the tank wall and is captured by the main monitoring sensor SM; among them, the main channel CH1 detects a relatively weak pulse. Its characteristic is peak amplitude mV, arrival time Meanwhile, the signal energy has significantly attenuated after penetrating the transformer tank wall, and the electromagnetic wave energy leaking into the external space is extremely low. After propagating through 1.5 meters of air, the amplitude is already below the detection threshold when it reaches the reference sensor SR. Therefore, in Near the specified time, the data in reference channel CH2 exhibits background noise without pulses, and no pulse events exceeding the threshold are generated.

[0065] b3. First-stage filtering (time-domain gating): From the main channel pulse event set Extract the main channel pulse from within and in the reference channel pulse set Search time window The pulse within.

[0066] The search results were empty; no reference pulses matching the time correlation criteria were found. After passing through the first stage of filtering, pulses suspected of being internal partial discharge pulses are sent to the second stage for fine identification.

[0067] b4. Second-stage filtering (waveform correlation identification): The algorithm extracts data from the main channel CH1 data stream using... Extract waveform segments 150 nanoseconds in length, centered at ns. The waveform exhibits a typical partial discharge pulse pattern; the algorithm synchronously obtains data from the reference channel CH2 data stream within the exact same time period. Extract waveform segments This segment mainly consists of random background noise with no obvious pulse structure; calculation and Normalized cross-correlation coefficient Since one is a structured pulse and the other is random noise, their waveforms are completely different, and the calculation yields... ;Will correlation threshold with preset Comparison, because If the correlation criterion is not met, the algorithm will ultimately confirm that... This is a real internal partial discharge pulse.

[0068] 4. Data Output and Result Application All real internal partial discharge pulses Information such as peak amplitude, arrival time, and corresponding power frequency phase angle is stored in the partial discharge database to generate a clean PRPD spectrum; interference pulse These are then recorded in the interference event log for analysis purposes.

[0069] The above embodiments fully verify the effectiveness of the method described in this invention in the field environment of a substation with strong interference: the system quickly eliminates most external interference (such as bushing corona in event A) through "time-domain gating screening," and then accurately identifies weak real partial discharges (such as winding discharges in event B) that are submerged by noise through "waveform correlation fine judgment," finally outputting high-confidence partial discharge data. This method achieves highly reliable interference suppression without relying on complex models and large-scale data training by only adding a reference sensor and a low-cost configuration of dual-channel synchronous acquisition, combined with a two-stage filtering algorithm with low computational complexity. It provides a clean and reliable data foundation for transformer insulation condition assessment, significantly improving the practical value of the online monitoring system.

[0070] The above description is merely a preferred embodiment of the present invention and does not limit the implementation and protection scope of the present invention. Those skilled in the art should realize that any equivalent substitutions and obvious changes made based on the description and illustrations of the present invention should be included within the protection scope of the present invention.

Claims

1. A transformer partial discharge interference suppression method based on reference channel time domain gating and correlation analysis, characterized in that, The method comprises the following steps: Step S1, synchronously collecting a main channel signal sequence of a main monitoring sensor and a reference channel signal sequence of a reference sensor; Step S2, respectively performing pulse detection on the main channel signal sequence and the reference channel signal sequence to obtain a main channel pulse event set and a reference channel pulse event set; Step S3, traversing a main channel pulse of the main channel pulse event set, if there is a reference channel pulse in the reference channel pulse event set that meets a time correlation condition, regarding the main channel pulse as an interference pulse and removing it; 2. The method of claim 1, wherein, otherwise, regarding the main channel pulse as a suspected internal partial discharge pulse and performing Step S4; Step S4, performing filtering correlation discrimination on the suspected internal partial discharge pulse to determine whether a normalized cross-correlation coefficient of a main channel waveform of the suspected internal partial discharge pulse and a reference channel waveform at a corresponding time is greater than a preset correlation threshold, if yes, regarding the suspected internal partial discharge pulse as an interference pulse, otherwise, regarding the suspected internal partial discharge pulse as a real internal partial discharge pulse; Step S5, collecting all events that are marked as real internal partial discharge pulses to obtain a final partial discharge data set. ， ， wherein, is a main channel detection threshold, is a reference channel detection threshold, is an empirical coefficient, is a main channel background noise level, is a reference channel background noise level; The step S2 comprises the following steps: ， ， wherein, is a set of primary channel pulse events, is a set of reference channel pulse events, is a peak amplitude of the primary channel pulse, is a peak amplitude of the reference channel pulse, is a peak time of arrival of the primary channel pulse, is a peak time of arrival of the reference channel pulse.

3. The method of claim 2, wherein, Step S21, calculating background noise levels of the main channel signal sequence and the reference channel signal sequence respectively to obtain a main channel background noise level and a reference channel background noise level; Step S22, setting a main channel detection threshold and a reference channel detection threshold respectively, and the specific formula is as follows:

4. The method of claim 1, wherein, Step S23, detecting pulses with amplitudes exceeding corresponding preset detection thresholds in the main channel signal sequence and the reference channel signal sequence respectively to obtain the main channel pulse event set and the reference channel pulse event set, and the specific formula is as follows: ， wherein, is the peak arrival time of the main channel pulse, is the peak arrival time of the reference channel pulse, is a preset time gating window.

5. The method of claim 4, wherein, In the step S21, a stable section without pulse activity is selected in the main channel signal sequence and the reference channel signal sequence respectively, and a root mean square value of signal amplitudes in the stable section is calculated to be the main channel background noise level and the reference channel background noise level respectively. ， wherein, is the distance between the main monitoring sensor and the reference sensor, is the propagation speed of electromagnetic waves in air, is the margin.

6. The method of claim 1, wherein, In the step S3, the time correlation condition is as follows: ， wherein, is a normalized cross-correlation coefficient, is a main channel full waveform segment, is a reference channel waveform segment, is a delay time; and step S44, judging whether the normalized cross-correlation coefficient is greater than a preset correlation threshold value, if yes, marking the suspected internal localizing pulse as an interference pulse, if not, marking the suspected internal localizing pulse as a real internal localizing pulse.

7. The method of claim 6, wherein, The preset time gating window is expressed by a formula as follows: The step S4 comprises the following steps: Step S41, extracting a main channel complete waveform segment of the suspected internal partial discharge pulse in the main channel; Step S42, extracting a reference channel waveform segment at the same time as the main channel complete waveform segment in the reference channel signal sequence; Step S43, calculating a normalized cross-correlation coefficient of the main channel complete waveform segment and the reference channel waveform segment, and the specific formula is as follows: The normalized cross-correlation coefficient has a value range of [-1, 1].

8. A transformer partial discharge interference suppression system based on reference channel time domain gating and correlation analysis, characterized in that, The application discloses a transformer partial discharge interference suppression method based on a reference channel time domain gating and correlation analysis, and belongs to the technical field of partial discharge detection.

9. The partial discharge interference suppression system for transformer based on time domain gating of reference channel and correlation analysis of claim 8, wherein, The main monitoring sensor is installed on a monitoring point of a transformer tank and is used for capturing electromagnetic signals, wherein the electromagnetic signals include internal partial discharge signals and external interference signals; the reference sensor is installed outside the transformer tank and faces an external space, and is used for capturing space electromagnetic interference signals from outside the transformer; and the synchronous data acquisition and processing unit is connected with the main monitoring sensor and the reference sensor, is used for synchronously acquiring two-way sensor signals, and executes the transformer partial discharge interference suppression method.

10. The partial discharge interference suppression system for transformer based on time domain gating of reference channel and correlation analysis of claim 8, wherein, The main monitoring sensor and the reference sensor are UHF sensors of the same type. The synchronous data acquisition and processing unit is a high-speed data acquisition card with at least two channels.

Images