Multi-parameter measurement system and method for power frequency non-partial discharge test transformer

Abstract

The invention discloses a multi-parameter measurement system and method for a power frequency non-partial discharge test transformer, and relates to the technical field of state monitoring of high-voltage test equipment and power equipment, and the system comprises a synchronous trigger unit which is connected to a power frequency voltage signal collection end and is used for detecting a zero crossing point of a power frequency voltage signal and generating a synchronous trigger signal, a power frequency period is divided into a plurality of sub-windows with equal phase angles. According to the multi-parameter measurement system and method for the power frequency non-partial discharge test transformer, the problem of phase mismatch of a power frequency signal and a high-frequency partial discharge signal in traditional multi-parameter measurement is effectively solved through a power frequency phase locked synchronous trigger mechanism and a dynamic noise suppression technology; and the positioning precision of the partial discharge source and the insulation defect diagnosis reliability are improved. The nanosecond-level time sequence error control is realized, and the power frequency coupling interference is inhibited while the details of the high-frequency pulse are kept by combining a phase correlation dynamic filtering strategy, so that the signal-to-noise ratio of the weak discharge signal is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of high voltage test equipment and power equipment status monitoring, in particular to a multi-parameter measurement system and method for a power frequency non-partial discharge test transformer. Background Art

[0002] In the operational status monitoring of transformers operating at power frequency without partial discharge testing, multi-parameter measurement is a core method for assessing their insulation performance and safety. Traditional methods typically rely on separate measurement units to independently collect power frequency voltage, current, and partial discharge signals. However, such solutions have significant drawbacks when collaboratively analyzing high-frequency partial discharge signals with power frequency signals. Due to the transient nature of high-frequency partial discharge pulses, their occurrence time is correlated with the phase of the power frequency cycle, directly affecting discharge type determination and location accuracy. Existing multi-parameter measurement systems generally employ time-sharing sampling or asynchronous clock control, resulting in a lack of strict synchronization between the time bases of the power frequency and high-frequency pulse signals. For example, when a partial discharge pulse occurs at different phase points in the power frequency waveform, each phase point may include a rising edge, peak, or falling edge, potentially representing different insulation defect modes. However, asynchronous sampling disrupts the temporal correspondence between the two, making it difficult to accurately map the pulse signal to a specific phase interval of the power frequency cycle. This phase mismatch not only leads to misjudgment of discharge characteristics but also introduces cumulative errors in time-difference location methods. This can further amplify location errors, especially for multi-winding transformers or complex insulation structures. Furthermore, the high amplitude of power-frequency signals can easily cause periodic interference in high-frequency measurement channels through electromagnetic coupling. Existing systems lack dynamic noise suppression mechanisms based on power-frequency phase synchronization, resulting in a reduced signal-to-noise ratio for partial discharge signals and the easy drowning out of weak discharge events. While existing technologies attempt to improve the accuracy of single-parameter detection through hardware integration or algorithm optimization, phase synchronization and coordinated processing of cross-frequency signals remain a key bottleneck restricting the reliability of multi-parameter joint diagnosis. Summary of the Invention

[0003] (1) Technical problems solved

[0004] In response to the shortcomings of the existing technology, the present invention provides a multi-parameter measurement system and method for a power frequency partial discharge-free test transformer, which solves the problem of how to achieve high-precision phase synchronization measurement of power frequency signals and high-frequency partial discharge signals to improve the accuracy of multi-parameter joint analysis and positioning reliability.

[0005] (2) Technical solution

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a multi-parameter measurement system for a power frequency non-partial discharge test transformer, comprising:

[0007] The synchronous trigger unit is connected to the power frequency voltage signal acquisition terminal and is used to detect the zero crossing point of the power frequency voltage signal and generate a synchronous trigger signal, dividing the power frequency cycle into multiple sub-windows with equal phase angles. The synchronous trigger unit monitors the waveform changes of the power frequency voltage signal in real time and generates a high-precision trigger pulse at the moment the voltage signal crosses the zero point, which serves as the reference starting point for the clock synchronization of the entire system. Subsequently, according to the preset phase angle interval, the power frequency cycle is divided into multiple equal-angle intervals, and the starting time of each sub-window is recursively determined by the zero crossing trigger signal. In order to deal with phase offsets caused by hardware transmission delays or signal distortion, the synchronous trigger unit has a built-in real-time feedback loop to dynamically collect the actual response time of the high-frequency signal acquisition module and compare it with the theoretical trigger time.

[0008] The dynamic filter bank, with its input connected to the high-frequency partial discharge signal acquisition channel and its output connected to the signal fusion module, is configured to dynamically switch the high-frequency channel's filtering mode based on the current phase angle and amplitude of the power frequency signal. The dynamic filter bank receives the power frequency signal's phase angle and amplitude data in real time and dynamically selects the high-frequency channel's filtering strategy based on the phase interval in which the current power frequency waveform resides. When the power frequency signal amplitude is in the phase interval near its peak, the filter bank automatically enhances the narrowband notch depth, generating a high-suppression notch band for the power frequency fundamental and major harmonic frequencies to block power frequency energy from interfering with the high-frequency partial discharge signal through electromagnetic coupling. In phase intervals with lower power frequency amplitudes, the filter bank switches to a wideband mode, filtering out only fixed-frequency noise excluding power frequency harmonics, preserving the full-band characteristics of the partial discharge pulse and preventing loss of high-frequency signal detail.

[0009] The dual-modal verification module includes a hardware verification circuit and a software mirror channel. The hardware verification circuit connects the synchronous trigger unit and the clock control terminal of the high-frequency signal acquisition module, while the software mirror channel is embedded in the signal processing unit. The dual-modal verification module ensures sampling synchronization and signal integrity through a collaborative mechanism of hardware and software. The hardware verification circuit captures the trigger signal from the synchronous trigger unit and the actual moment when the high-frequency signal acquisition module starts sampling in real time, calculates the time difference between the two, and converts it into an error voltage signal. When the time difference exceeds a preset threshold, the error voltage signal drives the phase compensation module to dynamically adjust the trigger time of the next cycle sub-window. The software mirror channel processes two data streams in parallel: one is the original asynchronous sampling data without synchronization optimization, and the other is the synchronization-optimized data after phase compensation and filtering. By comparing the differences in the time domain waveform and frequency domain energy distribution of the two data streams, the effectiveness of the synchronization mechanism is determined.

[0010] The signal fusion module receives the synchronized power frequency signal and the filtered high-frequency partial discharge signal, performs a spatiotemporal joint analysis, and outputs the discharge source location coordinates. The signal fusion module groups the high-frequency partial discharge pulses based on the synchronized power frequency phase label, maps each pulse to the corresponding power frequency period subwindow, and marks the phase interval to which it belongs. For the discharge pulse group within the same phase interval, the arrival time difference of the pulses received by each sensor is first calculated using the time difference method to generate the initial positioning candidate points. The candidate points are then weighted based on the power frequency amplitude characteristics of the phase interval. If there are multiple spatial positioning candidate points for the pulse group within the same phase interval and the distribution dispersion exceeds the threshold, it is determined that there are multiple discharge sources or signal reflection interference. The pulse waveform feature matching logic is activated, and parameters such as the pulse rising edge, pulse width, and spectrum main peak are extracted for cluster analysis to distinguish the signal categories of different discharge sources or interference sources.

[0011] The synchronization trigger unit uses multi-level clock tree distribution technology to send sampling instructions corresponding to the phase interval to the high-frequency signal acquisition module. The dynamic filter group adjusts the notch filter parameters based on the power frequency phase angle. The dual-modal verification module achieves synchronization calibration through real-time hardware verification and software data comparison. The synchronization trigger unit uses multi-level clock tree distribution technology to synchronously transmit the phase sub-window instructions divided by the power frequency period to all high-frequency signal acquisition modules, ensuring that the sampling start time of each module is strictly aligned.

[0012] Preferably, the synchronization triggering unit includes:

[0013] Zero-crossing detection circuit captures the zero-crossing point of the power frequency voltage signal in real time and outputs a trigger pulse;

[0014] The phase interval division module divides the power frequency cycle into multiple sub-windows according to the preset angle intervals and sends sampling instructions corresponding to the sub-windows to the high-frequency signal acquisition module;

[0015] The phase deviation compensation module dynamically adjusts the sending time of the sampling instruction according to the hardware delay parameters of the high-frequency signal acquisition module, so that the timing error is less than 10 nanoseconds.

[0016] Preferably, the phase interval division module synchronizes the starting time of each sub-window to the clock control end of the high-frequency signal acquisition module through a multi-level clock tree distribution technology, and embeds a real-time feedback loop in each sub-window to iteratively optimize the phase offset of the sampling instruction based on the error signal of the hardware verification circuit.

[0017] Preferably, the dynamic filter bank comprises:

[0018] an adaptive notch filter configured to enhance the narrowband notch depth in a phase interval where the power frequency amplitude is above a preset threshold;

[0019] a wideband filter configured to preserve the complete spectrum of the high frequency signal within a phase interval where the amplitude of the power frequency is below a preset threshold;

[0020] The filter mode switching logic dynamically selects a notch filter or a wideband filter to access the high-frequency signal channel based on the current phase angle and amplitude of the power frequency signal.

[0021] Preferably, the dynamic filter bank also includes a cross-band cross-correlation analysis module, which inputs the power frequency signal and the high-frequency partial discharge signal, extracts the time delay characteristics of the partial discharge pulse through joint analysis in the time and frequency domains, and uses a blind source separation algorithm to eliminate the residual interference of the power frequency harmonics.

[0022] Preferably, the dual-modal verification module performs the following operations:

[0023] The hardware verification circuit collects the actual start time difference between the synchronous trigger signal and the high-frequency sampling window in real time, and generates an error voltage signal to drive the phase deviation compensation module;

[0024] The software mirror channel compares the spectral energy distribution of the original asynchronously sampled data with the synchronously optimized data. If the energy difference in a specific frequency band exceeds a threshold, the secondary filtering logic of the dynamic filter bank is triggered.

[0025] Preferably, the signal fusion module includes:

[0026] A phase label generation unit groups partial discharge pulses according to the power frequency period sub-window and marks the phase interval to which they belong;

[0027] The phase-constrained positioning unit combines the time difference data and phase interval labels of multiple sensors to construct a weighted least squares positioning equation to correct the cumulative error caused by phase mismatch.

[0028] Preferably, in the phase constraint positioning unit, the time difference data of each partial discharge pulse is allocated a weight coefficient according to the phase interval to which it belongs, and the weight coefficient is inversely proportional to the power frequency amplitude of the phase interval.

[0029] Preferably, the software mirror channel adopts a redundant sampling strategy to store the time domain waveform and spectrum data of the high-frequency signal before and after the synchronization optimization, and determines whether to trigger system parameter recalibration through a preset difference threshold.

[0030] A multi-parameter measurement method for a power frequency non-partial discharge test transformer comprises the following steps:

[0031] S1: Detect the zero-crossing point of the power frequency voltage signal, generate a synchronous trigger signal and divide the power frequency cycle into sub-windows with equal phase angles;

[0032] S2: Start high-speed sampling of high-frequency partial discharge signals in each sub-window and correct hardware delays using a phase deviation compensation algorithm;

[0033] S3: Dynamically switches the notch filter mode of the high-frequency channel according to the current phase angle and amplitude of the power frequency signal to suppress power frequency coupled noise;

[0034] S4: The hardware verification circuit compares the trigger signal and the sampling window start time difference in real time to generate an error feedback signal to drive the compensation algorithm;

[0035] S5: Input the synchronized power frequency signal and high frequency signal into the spatiotemporal joint analysis model, and calculate the discharge source coordinates by combining the phase label and time difference data.

[0036] (3) Beneficial effects

[0037] The present invention provides a multi-parameter measurement system and method for power frequency partial discharge-free transformer testing. It has the following beneficial effects:

[0038] (1) The multi-parameter measurement system and method for the power frequency non-partial discharge test transformer effectively solves the phase mismatch problem between the power frequency signal and the high-frequency partial discharge signal in traditional multi-parameter measurement through the power frequency phase-locked synchronous trigger mechanism and dynamic noise suppression technology, thereby improving the positioning accuracy of the partial discharge source and the reliability of insulation defect diagnosis. Among them, the multi-level synchronous sampling architecture based on the power frequency zero-crossing trigger realizes nanosecond-level timing error control. Combined with the phase-related dynamic filtering strategy, it suppresses the power frequency coupling interference while retaining the high-frequency pulse details, thereby improving the signal-to-noise ratio of the weak discharge signal. The dual-modal verification mechanism ensures the robustness of the system in complex electromagnetic environments through cross-level collaboration of hardware verification and software mirroring. It is particularly suitable for long-term monitoring of multi-winding transformers or high-interference conditions, avoiding the misjudgment and data failure problems caused by asynchronous sampling in traditional methods.

[0039] (2) The multi-parameter measurement system and method of the power frequency non-partial discharge test transformer achieves deep fusion of multi-parameter data through a spatiotemporal joint analysis model and a phase-constrained positioning algorithm, which reduces the local discharge positioning error compared to the traditional time difference method, and supports the parallel identification of multiple discharge sources and intelligent filtering of interference signals. The system is compatible with multi-channel expansion and redundant design, automatically switches to the backup channel and triggers self-calibration in the event of hardware failure or signal anomaly, improving equipment availability and maintenance efficiency. In addition, the dynamic weight allocation and historical data interpolation mechanism enhance the system's adaptability to transient disturbances in the power grid, and solves the problem of measurement interruption caused by power frequency fluctuations or short-term loss of lock. It breaks through the limitations of traditional time-sharing sampling and static filtering, provides high-precision, high-reliability multi-dimensional data support for transformer insulation status assessment, and lays a technical foundation for intelligent fault warning and life prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic diagram of the overall framework of the present invention;

[0041] Figure 2 This is a control logic timing diagram of the present invention. DETAILED DESCRIPTION

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0043] See also Figure 1 and Figure 2 The present invention provides a technical solution: a multi-parameter measurement system for a power frequency non-partial discharge test transformer, comprising:

[0044] The synchronous trigger unit is connected to the power frequency voltage signal acquisition terminal and is used to detect the zero crossing point of the power frequency voltage signal and generate a synchronous trigger signal, dividing the power frequency cycle into multiple sub-windows with equal phase angles; the synchronous trigger unit monitors the waveform changes of the power frequency voltage signal in real time, and generates a high-precision trigger pulse at the moment the voltage signal crosses the zero point, which serves as the reference starting point for the clock synchronization of the entire system.

[0045] Subsequently, the power frequency cycle is divided into multiple equal-angle intervals according to the preset phase angle interval: each 10° phase angle is a sub-window, and the starting time of each sub-window is recursively determined in sequence by the zero-crossing trigger signal. In order to cope with the phase offset caused by hardware transmission delay or signal distortion, the synchronous trigger unit has a built-in real-time feedback loop to dynamically collect the actual response time of the high-frequency signal acquisition module and compare it with the theoretical trigger time. If a deviation is detected between the start time of the sampling window and the starting point of the theoretical phase interval, the trigger instructions of the subsequent sub-windows are fine-tuned through the phase compensation algorithm. For example, the sending time of the trigger signal is advanced or delayed in the next cycle so that the timing error converges to within ten nanoseconds.

[0046] Furthermore, when harmonic distortion or noise interference in the power frequency signal causes abnormal zero-crossing detection, the synchronous trigger unit automatically switches to prediction mode, interpolating and estimating the current zero-crossing location based on historical power frequency cycle data. Once the signal stabilizes, the real-time detection logic is reactivated to ensure the continuity and robustness of phase division. For multi-channel measurement scenarios, the synchronous trigger unit utilizes multi-level clock tree distribution technology to synchronously distribute phase sub-window division instructions for the same power frequency cycle to all high-frequency signal acquisition modules, avoiding inter-channel phase mismatches caused by distributed clock source drift.

[0047] The dynamic filter group has an input end connected to the high-frequency partial discharge signal acquisition channel and an output end connected to the signal fusion module, and is configured to dynamically switch the filtering mode of the high-frequency channel according to the current phase angle and amplitude of the power frequency signal; the dynamic filter group receives the phase angle and amplitude data of the power frequency signal in real time, and dynamically selects the filtering strategy of the high-frequency channel according to the phase interval of the current power frequency waveform.

[0048] When the power frequency signal amplitude is in the phase interval near the peak, the filter group automatically enhances the narrowband notch depth, generating a notch band with a high suppression ratio for the power frequency fundamental wave and main harmonic frequencies, so as to block the interference of the power frequency energy on the high-frequency partial discharge signal through electromagnetic coupling; while in the phase interval with lower power frequency amplitude, that is, the phase interval near the zero point, the filter group switches to wideband mode, filtering out only fixed-frequency noise other than the power frequency harmonics, retaining the full-band characteristics of the partial discharge pulse, and avoiding the loss of high-frequency signal details.

[0049] If the power frequency signal experiences instantaneous amplitude fluctuations due to sudden load changes or grid disturbances, the filter bank inserts transition filtering modes between adjacent phase intervals based on a preset amplitude-phase joint criterion, gradually adjusting the notch bandwidth and depth to prevent signal distortion caused by sudden changes in filtering parameters. For power frequency signals with harmonic distortion, the filter bank uses real-time spectrum analysis to identify the amplitude and phase distribution of the harmonic components and dynamically expands the number of notch frequencies to simultaneously suppress multi-band interference.

[0050] Furthermore, when anomalies in the phase correlation between partial discharge pulses and power frequency signals in high-frequency channels are detected, such as irregular distribution of pulse groups within the power frequency cycle, the filter bank activates cross-correlation analysis logic to extract the time delay characteristics of the pulse signal and the power frequency waveform. This, combined with a blind source separation algorithm, distinguishes true discharge pulses from residual noise, further optimizing the filtering effect. In multi-channel measurement scenarios, the filter bank coordinates the switching timing of filter parameters across channels based on the power frequency phase consistency of each acquisition channel, ensuring phase alignment for cross-channel signal analysis.

[0051] The dual-modal verification module includes a hardware verification circuit and a software mirror channel. The hardware verification circuit connects the synchronous trigger unit and the clock control terminal of the high-frequency signal acquisition module, while the software mirror channel is embedded in the signal processing unit. The dual-modal verification module ensures sampling synchronization and signal integrity through a collaborative mechanism of hardware and software. The hardware verification circuit captures the trigger signal from the synchronous trigger unit and the actual sampling start time of the high-frequency signal acquisition module in real time, calculates the time difference between the two, and converts it into an error voltage signal.

[0052] When the time difference exceeds a preset threshold, the error voltage signal drives the phase compensation module to dynamically adjust the triggering moment of the next cycle sub-window. For example, when a high-frequency sampling delay is detected, the hardware verification circuit sends an early trigger instruction to the clock control end to shorten the waiting time of the subsequent sub-window.

[0053] The software mirror channel processes two data channels in parallel: one is the original asynchronous sampling data without synchronization optimization, and the other is the synchronization-optimized data after phase compensation and filtering. By comparing the differences in the time domain waveform and frequency domain energy distribution of the two data channels, the effectiveness of the synchronization mechanism is judged.

[0054] If the energy difference in high-frequency bands above 1MHz exceeds the preset tolerance, it is determined to be insufficient noise suppression or synchronization mismatch, triggering the dynamic filter bank's secondary filtering logic or system parameter recalibration process. If the difference is within the tolerance, the current synchronization status is marked as valid. If the power frequency signal experiences an abnormal zero-crossing detection due to a transient disturbance in the power grid, the hardware verification circuit detects a continuous increase in the deviation between the trigger signal and the sampling window for multiple consecutive cycles. The software mirror channel automatically switches to historical phase data interpolation mode, generating a virtual trigger signal based on the phase interval division pattern of the previous normal cycle until the power frequency signal returns to stability.

[0055] For multi-channel systems, the software mirror channel synchronously compares the data spectrum consistency of all high-frequency acquisition channels before and after optimization. If the difference in a certain channel is significantly higher than that of other channels, it is determined that there is a hardware fault or electromagnetic interference anomaly in the channel, triggering an alarm and isolating the faulty channel data.

[0056] The signal fusion module receives the synchronized power frequency signal and the filtered high-frequency partial discharge signal, performs a joint spatiotemporal analysis, and outputs the discharge source location coordinates. The signal fusion module groups high-frequency partial discharge pulses based on the synchronized power frequency phase labels, maps each pulse to a corresponding power frequency period subwindow, and labels the phase interval to which it belongs. For groups of discharge pulses within the same phase interval, the arrival time difference of the pulses received by each sensor is calculated using the time difference method to generate initial positioning candidate points. The candidate points are then weighted based on the power frequency amplitude characteristics of the phase interval. For example, in phase intervals with high power frequency amplitudes, the weight coefficient of the corresponding sensor time difference data in this interval is reduced due to increased electromagnetic interference, while in phase intervals with low power frequency amplitudes, a higher weight is assigned.

[0057] If there are multiple spatial positioning candidate points for a pulse group within the same phase interval and the distribution dispersion exceeds the threshold, it is determined to be multiple discharge sources coexisting or signal reflection interference. The pulse waveform feature matching logic is activated, and parameters such as the pulse rising edge, pulse width, and spectrum main peak are extracted for cluster analysis to distinguish the signal categories of different discharge sources or interference sources.

[0058] For continuous discharge events across phase intervals, the module establishes a phase association chain, tracks the evolution of the discharge pulse within the power frequency cycle, and combines the transformer winding structure model to eliminate positioning conflicts caused by differences in signal propagation paths.

[0059] When a phase interval lacks a valid discharge pulse or the data confidence level is insufficient, the signal fusion module automatically interpolates historical data from adjacent phase intervals and corrects for jump errors in the positioning coordinates through spatiotemporal continuity verification. Ultimately, an iterative optimization algorithm is employed to output the spatial coordinates of the discharge source, integrating multi-dimensional information such as phase weights, time difference data, and waveform characteristics. The uncertainty range of the positioning result is also noted in the presence of residual noise.

[0060] The synchronization trigger unit uses multi-level clock tree distribution technology to send sampling instructions corresponding to the phase interval to the high-frequency signal acquisition module. The dynamic filter bank adjusts the notch filter parameters based on the power frequency phase angle. The dual-modal verification module achieves synchronization calibration through real-time hardware verification and software data comparison. The synchronization trigger unit uses multi-level clock tree distribution technology to synchronously transmit the phase sub-window instructions divided by the power frequency period to all high-frequency signal acquisition modules, ensuring that the sampling start time of each module is strictly aligned.

[0061] When the high-frequency module's sampling window starts late due to hardware transmission delay, the synchronous trigger unit embeds a compensation offset in the trigger instruction of the next phase sub-window based on the real-time feedback delay. For example, it sends the instruction in advance to offset the fixed delay, so that the timing error converges to within ten nanoseconds.

[0062] The dynamic filter group adjusts the notch filter parameters in real time based on the power frequency phase angle: in the phase interval with higher power frequency amplitude, such as the phase interval near the power frequency amplitude of 90°, the notch depth of the power frequency fundamental wave and its third harmonic is automatically enhanced, and the stopband width is expanded to suppress broadband coupled noise; in the phase interval with lower power frequency amplitude, such as the phase interval near the power frequency amplitude zero crossing point, it switches to wideband mode, filtering out only fixed interference in a specific frequency band and retaining the rising edge details of the high-frequency discharge pulse.

[0063] If the power frequency signal's phase angle jumps due to transient grid disturbances, the dynamic filter bank activates a transition filtering strategy, smoothly adjusting the notch parameters based on the phase change rate to avoid signal distortion caused by sudden changes in the filtering mode. The dual-modal verification module uses a hardware verification circuit to monitor the actual time difference between the trigger signal and the high-frequency sampling in real time. When the time difference exceeds a preset threshold for multiple consecutive cycles, a hardware-level compensation instruction is triggered to force correction of the clock tree distribution timing. Simultaneously, a software mirror channel compares the frequency domain energy distribution of the original asynchronous data with that of the synchronously optimized data. If the energy difference in the high-frequency band above 1MHz exceeds the tolerance, it is determined to be a synchronization failure or insufficient noise suppression, triggering a parameter reset of the dynamic filter bank and a self-test of the zero-crossing detection logic of the synchronous trigger unit.

[0064] For multi-channel systems, the hardware verification circuit synchronously verifies the trigger-sampling time difference of all channels. If the channel deviation is persistently abnormal, the channel data is isolated and the backup acquisition module is activated. The software mirror channel analyzes the phase consistency of the optimized data of each channel and downgrades the channel data that deviates from the mainstream phase distribution to prevent abnormal data from contaminating the positioning results.

[0065] When the zero-crossing detection of the power frequency signal fails due to harmonic distortion, the synchronous trigger unit switches to the prediction mode and generates a virtual zero-crossing point based on the interpolation of historical power frequency cycle data. The dynamic filter group synchronously switches to the wideband mode and enhances the anti-aliasing filtering. The dual-modal verification module marks the data in this stage as "transient mode" and recalibrates the system parameters after the signal returns to stability.

[0066] The synchronous trigger unit includes:

[0067] Zero-crossing detection circuit captures the zero-crossing point of the power frequency voltage signal in real time and outputs a trigger pulse;

[0068] The phase interval division module divides the power frequency cycle into multiple sub-windows according to the preset angle intervals and sends sampling instructions corresponding to the sub-windows to the high-frequency signal acquisition module;

[0069] The phase deviation compensation module dynamically adjusts the sending time of the sampling instruction according to the hardware delay parameters of the high-frequency signal acquisition module, so that the timing error is less than 10 nanoseconds.

[0070] It should be further explained that, in the specific implementation process, the synchronization trigger unit captures the waveform changes of the industrial frequency voltage signal in real time through the zero-crossing detection circuit. When the voltage crosses the zero point from negative to positive or from positive to negative, a high-precision trigger pulse is generated as the system synchronization reference.

[0071] The phase interval division module divides the power frequency cycle into a continuous sequence of sub-windows based on preset phase intervals, with each sub-window consisting of 10°. It then sends corresponding sampling start instructions to each high-frequency signal acquisition module. If the power frequency signal experiences zero-crossing detection anomalies due to harmonic distortion or noise interference, such as oscillation near the zero-crossing point, the trigger pulse generation logic automatically switches to prediction mode, interpolating the current zero-crossing position based on the phase pattern of the previous cycle and resuming real-time detection once the signal stabilizes.

[0072] The phase deviation compensation module collects the hardware response delay parameters of the high-frequency signal acquisition module in real time, such as ADC startup time and signal transmission delay, and dynamically adjusts the sending time of the sampling instruction. For example, when it is detected that the sampling start time of a sub-window lags behind the theoretical phase starting point by 5 nanoseconds, the trigger instruction of the next sub-window is sent 5 nanoseconds in advance, so that the timing error gradually converges to within 10 nanoseconds.

[0073] For multi-channel high-frequency acquisition modules, the phase interval division module uses multi-level clock tree distribution technology to synchronously transmit the trigger instructions of the same subwindow to all channels. If the timing deviation of a channel exceeds the threshold due to hardware differences, the trigger instructions of the channel are compensated for the offset separately to prevent the global clock adjustment from affecting the synchronization accuracy of other channels.

[0074] When the power signal frequency fluctuates slightly, the phase interval division module dynamically adjusts the sub-window time width to maintain a constant phase angle interval. If the frequency suddenly changes beyond the preset range of +2Hz or -2Hz, the system is triggered to relock the power frequency period and reset the sub-window division. The phase deviation compensation module also has a built-in historical error learning function, which calculates the delay characteristic distribution of the high-frequency module and preloads compensation values in the initial stage to accelerate error convergence.

[0075] The phase interval division module synchronizes the start time of each subwindow to the clock control terminal of the high-frequency signal acquisition module through multi-level clock tree distribution technology. A real-time feedback loop is embedded in each subwindow to iteratively optimize the phase offset of the sampling instruction based on the error signal of the hardware verification circuit. It should be further explained that in the specific implementation process, the phase interval division module synchronizes the start time of each subwindow to the clock control terminal of each high-frequency signal acquisition module through multi-level clock tree distribution technology, ensuring that all modules start sampling within the same power frequency phase interval.

[0076] If a high-frequency module's actual trigger time deviates from the theoretical value due to hardware variations, such as clock crystal oscillator deviation or transmission path delay, the real-time feedback loop collects the module's sampling start-up time difference, generates a phase offset error signal, and feeds it into the compensation algorithm. For example, if a module's start time in a 10° phase subwindow lags behind the theoretical value by 3 nanoseconds, the compensation algorithm will send the trigger command for that module in the next subwindow 3 nanoseconds earlier, while other modules continue to execute according to their original timing, thus achieving precise channel-level compensation.

[0077] In scenarios where communication delay jitter exists, such as long-distance transmission, the feedback loop uses a sliding average filter to process error signals over multiple consecutive cycles, distinguishing between fixed delays and random jitter components. Only the fixed portion is compensated to avoid excessive adjustments that introduce additional timing noise. When the phase offset of a high-frequency module continuously exceeds the preset safety threshold of 15 nanoseconds, the module automatically triggers an abnormality alarm and switches to the backup acquisition channel, while recording the offset characteristics of the faulty channel for subsequent diagnosis. If the power signal frequency fluctuates, causing the actual time length of the subwindow to change, the module dynamically adjusts the distribution frequency of the clock tree to maintain a constant phase angle interval. When the frequency mutation exceeds the system's tracking capability, the subwindow division is suspended and the power frequency period is relocked.

[0078] In a multi-channel system, a real-time feedback loop monitors the phase synchronization status of all channels simultaneously. If more than half of the channels deviate in the same direction, the global clock source is considered abnormal, triggering a clock tree self-test and switching to a redundant clock source. To prevent signal transmission in environments with high electromagnetic interference, the module inserts anti-interference relay units in the clock tree path to monitor signal integrity in real time. If clock pulse distortion is detected, error correction coding is enabled to reconstruct the original trigger instruction, ensuring the robustness of the sub-window division.

[0079] The dynamic filter bank includes:

[0080] an adaptive notch filter configured to enhance the narrowband notch depth in a phase interval where the power frequency amplitude is above a preset threshold;

[0081] a wideband filter configured to preserve the complete spectrum of the high frequency signal within a phase interval where the amplitude of the power frequency is below a preset threshold;

[0082] The filter mode switching logic dynamically selects a notch filter or a wideband filter to connect to the high-frequency signal channel based on the current phase angle and amplitude of the power frequency signal. It should be further explained that during the specific implementation process, the dynamic filter group monitors the phase angle and amplitude of the power frequency signal in real time. When the power frequency signal is in a phase range with a higher amplitude, such as the 90° to 150° range near the peak, the adaptive notch filter automatically increases the notch depth of the power frequency fundamental wave and its third and fifth harmonics, and expands the stopband width to suppress broadband coupling interference. At this time, the filtering mode of the high-frequency channel is mainly to suppress power frequency-related noise, while retaining the core frequency band of partial discharge pulses, such as 500kHz to 2MHz.

[0083] When the power-frequency signal enters a phase interval with lower amplitude, such as the 0° to 30° or 180° to 210° range near the zero crossing, the filter bank switches to wideband mode, filtering out only background noise in a fixed frequency band, including fixed-frequency switching noise, to avoid excessive filtering and loss of detail on the rising edge of high-frequency pulses. If the power-frequency signal experiences rapid amplitude fluctuations due to sudden load changes, the filter bank inserts a transition filtering mode between adjacent phase intervals based on a combined criterion of phase change rate and amplitude gradient, gradually adjusting the notch depth and bandwidth. For example, when the amplitude changes from high to low, the notch depth is linearly reduced in steps of 10° phase angle to prevent signal distortion caused by sudden changes in filter parameters.

[0084] For power frequency signals with harmonic distortion, such as those with a total harmonic distortion exceeding 5%, the filter bank uses real-time spectrum analysis to identify the amplitude and frequency of the main harmonic components, dynamically increasing the number of notch frequencies while suppressing interference from the fundamental and third and fifth harmonics. When an abnormal correlation between partial discharge pulses and the power frequency phase is detected in the high-frequency channel, such as pulse groups randomly distributed within the power frequency period, the filter bank initiates cross-band cross-correlation analysis to extract the time delay characteristics of the pulse signal and the power frequency waveform. Using blind source separation techniques, the filter bank distinguishes true discharge pulses from residual power frequency harmonic interference. For example, independent component analysis is used to isolate and filter out high-frequency noise components unrelated to the power frequency phase. In multi-channel measurement scenarios, the filter bank coordinates the filter parameter switching timing of different channels based on the power frequency phase consistency of each channel, such as whether the phase deviation is within a range of ±5°. This ensures that all channels use the same filtering mode within the same phase range, avoiding signal analysis inaccuracies caused by filtering differences between channels. If a channel fails to filter due to a hardware failure, such as insufficient notch depth, the channel data will be automatically isolated and switched to the redundant backup channel, and an alarm will be triggered to prompt maintenance.

[0085] The dynamic filter bank also includes a cross-band cross-correlation analysis module, which takes as input the power frequency signal and the high-frequency partial discharge signal. It extracts the time delay characteristics of the partial discharge pulses through a joint time-frequency domain analysis and employs a blind source separation algorithm to eliminate residual interference from the power frequency harmonics. It should be further explained that, in its implementation, the cross-band cross-correlation analysis module receives synchronized data streams of the power frequency signal and the high-frequency partial discharge signal and establishes a phase correlation model between the two through a joint time-frequency domain analysis. When the high-frequency pulse group exhibits a regular phase distribution within the power frequency cycle, the module extracts the time delay characteristics of the pulse signal relative to the power frequency zero crossing. For example, the module calculates the distribution variance of the pulse arrival times within the same phase interval. If the variance is below a threshold, it is considered a valid discharge event. If the variance is too high and there is no clear phase clustering, the blind source separation algorithm is activated to decompose the high-frequency signal into independent components that are both related and unrelated to the power frequency phase, suppressing the latter as noise interference.

[0086] For power frequency harmonic distortion scenarios, such as a sudden increase in the third harmonic amplitude, the cross-band cross-correlation analysis module uses spectral correlation analysis to identify interference components in the high-frequency signal that are coupled with the harmonic frequencies. The interference components include narrowband noise at the power frequency harmonic multiples. The corresponding notch filter parameters are dynamically generated and updated to the dynamic filter bank.

[0087] When the time delay characteristics of the detected partial discharge pulses are inconsistent with the power frequency phase change rate, such as the pulse group still maintains a fixed time interval when the power frequency fluctuates, the cross-band cross-correlation analysis module determines that it is an external interference signal, which includes switching operation noise, and marks such pulses for subsequent filtering.

[0088] In a multi-channel system, the cross-band cross-correlation analysis module synchronously analyzes the consistency of the delay characteristics of the high-frequency signals of each channel. If the delay distribution of a channel deviates significantly from other channels, such as the standard deviation exceeds twice the mean, it is determined that there is a hardware fault or local interference source in the channel, triggering channel data isolation and alarm.

[0089] For residual periodic noise, such as high-frequency carrier interference, the blind source separation algorithm separates Gaussian noise components that are incoherent with the power frequency based on the statistical independence of the signal and filters them out using an adaptive threshold. When a transient disturbance in the power grid causes the power frequency signal to lose lock for a short time, the module switches to the historical phase correlation model and predicts the current phase-delay mapping relationship based on the previous normal periodic data until the power frequency synchronization is restored and stabilized. If the separated signal still contains unidentified interference, and the unidentified interference includes random pulse noise, a secondary screening is performed based on the pulse waveform parameters, and only pulses that meet the typical characteristics of partial discharge are retained for positioning calculations; among them, the pulse waveform parameters include the rising edge slope and pulse width consistency.

[0090] The dual-modal authentication module performs the following operations:

[0091] The hardware verification circuit collects the actual start time difference between the synchronous trigger signal and the high-frequency sampling window in real time, and generates an error voltage signal to drive the phase deviation compensation module;

[0092] The software mirror channel compares the spectral energy distribution of the original asynchronously sampled data with the synchronously optimized data. If the energy difference in a specific frequency band exceeds a threshold, the secondary filtering logic of the dynamic filter bank is triggered. It should be further explained that during the specific implementation process, the dual-modal verification module uses a hardware verification circuit to collect the trigger signal output by the synchronous trigger unit and the actual sampling moment of the high-frequency signal acquisition module in real time, calculates the time difference between the two, and converts it into an error voltage signal. When the time difference exceeds the preset threshold of 10 nanoseconds, the error voltage signal drives the phase compensation module to embed an offset in the trigger instruction of the next power frequency cycle subwindow. For example, if a high-frequency sampling delay of 5 nanoseconds is detected, the trigger instruction for the next subwindow is sent 5 nanoseconds in advance, gradually converging the timing error.

[0093] For multi-channel systems, the hardware verification circuit independently monitors the trigger-sampling time difference of each channel. If the deviation of a channel continuously exceeds the threshold while other channels are normal, it is determined that there is a hardware fault in that channel, triggering data isolation and activating the backup channel.

[0094] The software mirror channel processes two data streams in parallel within the signal processing unit, including: the original asynchronously sampled data and the data that has been synchronously optimized and filtered, and compares the difference in spectral energy distribution between the two in the high-frequency band above 1MHz. If the high-frequency energy of the optimized data is significantly lower than that of the original data and the difference exceeds the tolerance of 30%, it is determined to be a synchronization failure or filter overload, triggering the dynamic filter group parameter reset and the zero-crossing detection logic self-test of the synchronous trigger unit; if the difference is within the tolerance but the local frequency band energy is abnormal, the secondary narrowband notch of the dynamic filter group is activated. When the zero-crossing detection of the power frequency signal is abnormal due to transient disturbances, the hardware verification circuit detects that the trigger deviation of multiple consecutive cycles continues to increase, and the software mirror channel switches to the historical phase data interpolation mode, generating a virtual trigger signal based on the phase law of the previous normal cycle, and marking the data in this stage as "transient mode" and reducing its positioning weight.

[0095] For multi-channel data, the software mirrors the channels synchronously to analyze the phase distribution consistency of the optimized data across all channels. If a channel's phase label deviates by ±15° from the mainstream distribution, it is determined to be affected by local interference, and its data is downgraded or excluded. If the hardware verification and software mirroring results conflict, such as if the hardware determines synchronization is normal but the software indicates excessive spectrum discrepancies, the signal fusion module prioritizes the software result and triggers a recalibration of all system parameters to ensure robustness against interference.

[0096] The signal fusion module includes:

[0097] A phase label generation unit groups partial discharge pulses according to the power frequency period sub-window and marks the phase interval to which they belong;

[0098] The phase-constrained positioning unit combines the time difference data from multiple sensors with phase interval labels to construct a weighted least squares positioning equation to correct for accumulated errors caused by phase mismatch. It should be further explained that, in its implementation, the signal fusion module groups high-frequency partial discharge pulses according to their phase intervals based on the synchronized power frequency phase labels. Initial positioning candidate points are generated using the multi-sensor time difference data from pulse groups within the same phase interval. Dynamic weighting is applied to candidate points based on the amplitude characteristics of the power frequency phase intervals. In phase intervals with higher power frequency amplitudes, such as near peaks, the weight of the time difference data from sensors susceptible to interference is reduced due to strong electromagnetic interference. In intervals with lower power frequency amplitudes, such as near zero crossings, the weight is increased to enhance positioning confidence. If multiple candidate points exist within the same phase interval and their spatial distribution dispersion exceeds a preset threshold, such as the physical size range of the transformer winding, pulse waveform feature matching logic is activated to extract the pulse rising edge slope, pulse width, and spectral peak parameters. Cluster analysis is then used to distinguish between true discharge sources and reflected interference signals. For continuous discharge events across phase intervals, the module tracks the phase evolution of the pulse group within the power frequency cycle and verifies the physical rationality of the positioning results based on the transformer winding structure model. For example, it eliminates abnormal coordinates where the signal propagation path passes through the insulation barrier.

[0099] When noise interference causes an insufficient number of valid pulses in a phase interval or the data confidence level falls below a threshold, the module calls upon historical positioning data from adjacent phase intervals, employing a spatiotemporal continuity interpolation algorithm to complete the coordinates of the current interval. This algorithm then optimizes and eliminates coordinate jump errors through gradient descent. If the multi-sensor time difference data within a phase interval is inconsistent, such as when some sensors fail to detect pulses, the anomalous channel data is dynamically ignored based on the sensor's historical reliability score, and time difference information from high-scoring sensors is prioritized for calculation. Ultimately, the positioning result, which integrates phase weights, time difference data, and waveform characteristics, is output through an iterative optimization algorithm. The positioning uncertainty range is also noted in the event of power frequency signal anomalies or synchronization failures. For scenarios with multiple discharge sources, phase-time difference joint clustering is used to separate pulse groups from different discharge sources, and the positioning coordinates of each cluster are independently calculated to avoid positioning ambiguity caused by signal aliasing.

[0100] In the phase-constrained positioning unit, the time-difference data of each partial discharge pulse is assigned a weight coefficient based on the phase interval to which it belongs. The weight coefficient is inversely proportional to the power frequency amplitude within that phase interval. It should be further explained that, in specific implementations, the phase-constrained positioning unit dynamically assigns a weight coefficient based on the power frequency amplitude within the phase interval to which each partial discharge pulse belongs. The weight coefficient is inversely proportional to the power frequency amplitude. This includes: in phase intervals with high power frequency amplitude, such as the 90° to 150° range near the peak, due to the strong power frequency electromagnetic coupling interference, the weight of the time-difference data of interference-prone sensors in this range is reduced. Interference-prone sensors include sensors near the high-voltage winding. For example, the weight coefficient is set to 0.3 to 0.5. In phase intervals with low power frequency amplitude, such as the 0° to 30° range near the zero crossing, the weight coefficient is increased to 0.8 to 1.0 to enhance the data contribution in low-interference environments. If the power frequency amplitude changes suddenly due to transient power grid disturbances, such as a short dip or surge, the weight coefficient is dynamically and smoothly adjusted based on the amplitude change rate to avoid positioning coordinate oscillation caused by weight jumps.

[0101] If multiple discharge source candidate points exist within the same phase interval, the system further weights the historical reliability scores of the corresponding sensors, such as the data consistency over the past 10 cycles, and prioritizes the time difference data from sensors with higher scores. If the power frequency amplitude in a phase interval is distorted due to measurement anomalies, such as sensor saturation, the weighting logic for that interval is automatically ignored, and the system switches to the historical weighting model for adjacent normal phase intervals. Positioning deviations are corrected through spatiotemporal continuity verification.

[0102] In scenarios with multiple discharge sources, the weight coefficient is also related to the waveform characteristics of the discharge pulses. For example, if a pulse group appears continuously in multiple phase intervals with consistent waveform parameters, its weight is increased to indicate the credibility of the stable discharge source. Waveform parameters include pulse width and rising edge. Conversely, if the pulse group appears only in a single high-amplitude phase interval with a chaotic waveform, the weight is reduced to suppress transient interference. If the system detects a discrepancy between the weight assignment results and the spectrum difference analysis of the software mirror channel, such as abnormal high-frequency energy corresponding to the high-weight interval, the weight logic self-check is triggered and the weight coefficient is recalculated using the clean signal after blind source separation to ensure positioning robustness.

[0103] The software mirror channel employs a redundant sampling strategy, storing the time-domain waveform and spectrum data of the high-frequency signal before and after synchronous optimization. A preset difference threshold is used to determine whether to trigger system parameter recalibration. It should be noted that, in its implementation, the software mirror channel redundantly samples the high-frequency partial discharge signal before and after synchronous optimization, storing the time-domain waveform and spectrum energy distribution of the original asynchronous data and the synchronously optimized data. When comparing the energy difference between the two in a specific frequency band (i.e., 1MHz to 3MHz), if the high-frequency energy attenuation of the optimized data exceeds a preset threshold of 40% and is not accompanied by a simultaneous improvement in power frequency harmonic suppression, it is determined to be synchronous overcompensation or filter overload, triggering the dynamic filter bank's secondary narrowband notch to restore signal integrity. If the energy difference falls below the threshold but an abnormal spike occurs in a localized frequency band—for example, if the energy difference falls below the threshold but an abnormal spike occurs at the third harmonic of the power frequency—the dynamic filter bank's harmonic tracking notch logic is activated to generate a supplementary notch filter for the specific frequency. For multi-channel systems, the mirror channel synchronously analyzes the energy difference consistency of each channel's optimized data within the same frequency band. If the difference in a channel significantly deviates from the others, such as by more than two standard deviations from the mean, it is determined that hardware synchronization failure or local interference exists in that channel. Its data is isolated and switched to a backup channel. If the synchronized optimization data becomes invalid due to a brief loss of power frequency signal lock, the mirror channel switches to historical data mode, generating virtual optimized data based on the spectral characteristics of the previous normal cycle. After signal recovery, it triggers a full system parameter recalibration to eliminate transient error accumulation. If the redundant sampled data exhibits periodic interference residues in the time domain waveform, such as pulse groups appearing at fixed positions within each power frequency cycle, the system uses a combination of phase label and power frequency amplitude correlation analysis to distinguish true discharge pulses from periodic noise, and enhances filtering effectiveness through a blind source separation algorithm. For persistent anomalous data in high-interference environments, the mirror channel initiates a continuous self-check cycle, updating the difference threshold and dynamically adjusting the filtering parameters each cycle until the spectral energy distribution returns to the steady-state range.

[0104] The multi-parameter measurement system for the power frequency non-partial discharge test transformer uses a synchronous trigger unit to capture the zero-crossing point of the power frequency voltage signal in real time, and uses this as the reference starting point for the clock synchronization of the entire system. Zero-crossing detection uses the dual criteria of waveform slope and threshold crossing. When the power frequency voltage crosses the zero point from negative to positive or from positive to negative and the slope meets the preset range, a trigger pulse signal is generated. If the power frequency signal oscillates near the zero-crossing point due to harmonic distortion, the prediction mode is triggered. The current zero-crossing point position is estimated by interpolation based on the phase law of the historical power frequency cycle, and the signal is switched back to real-time detection mode after it stabilizes. The synchronous trigger unit divides the power frequency cycle into multiple sub-windows with equal phase angles. The starting time of each sub-window is recursively determined by the zero-crossing trigger signal, and the trigger instruction is synchronously transmitted to all high-frequency signal acquisition modules through multi-level clock tree distribution technology. To address the hardware delay differences in high-frequency modules, the synchronous trigger unit has a built-in phase deviation compensation module, which collects the response time differences of each module in real time and dynamically adjusts the trigger instruction sending time of subsequent sub-windows. For example, if a module causes sampling lag due to a fixed delay, the trigger instruction of the next sub-window is advanced by a corresponding amount of time, so that the timing error converges to within ten nanoseconds.

[0105] The dynamic filter bank dynamically switches filtering modes based on the current phase angle and amplitude of the power frequency signal, i.e., it adjusts the filtering mode based on the gradual change in phase and amplitude. When the power frequency amplitude is in the phase range near the peak, the narrowband notch depth is increased to suppress the power frequency fundamental and harmonic interference; in the phase range near the zero crossing point with lower amplitude, it switches to a wideband mode to preserve the complete spectral characteristics of the high-frequency pulse. A gradual adjustment strategy is adopted during the switching of filtering parameters, smoothly transitioning the notch depth and bandwidth based on the phase change rate to avoid signal distortion caused by sudden parameter changes. For power frequency signals with harmonic distortion, the dynamic filter bank uses real-time spectrum analysis to identify the main harmonic components, dynamically expand the number of notch frequencies, and simultaneously suppress multi-band interference. When an abnormal correlation between the local discharge pulse and the power frequency phase in the high-frequency signal is detected, cross-band cross-correlation analysis is initiated to extract the pulse delay characteristics and combine blind source separation technology to distinguish the true discharge signal from residual noise.

[0106] The dual-modal verification module ensures system reliability through the collaborative operation of a hardware verification circuit and a software mirror channel. The hardware verification circuit compares the time difference between the synchronization trigger signal and the actual start time of the high-frequency sampling window in real time, generating an error feedback signal to drive iterative optimization of the compensation algorithm. The software mirror channel processes the original asynchronous data and the synchronized optimized data in parallel, comparing the energy distribution differences between the two in the time-frequency domain. If the energy difference in the high-frequency band exceeds a preset threshold, it is determined to be a synchronization failure or filtering anomaly, triggering a dynamic filter bank parameter reset and system self-test. For multi-channel systems, the hardware verification circuit independently monitors the synchronization deviation of each channel. If a channel is persistently abnormal, its data is isolated and a backup channel is activated. The software mirror channel analyzes the phase consistency of the optimized data of each channel and downgrades the data of channels that deviate from the mainstream distribution. If the power frequency signal is temporarily lost, the module switches to historical phase data interpolation mode, generates a virtual trigger signal, marks the transient data, and recalibrates the system parameters after the signal recovers.

[0107] The signal fusion module groups high-frequency partial discharge pulses based on synchronized power frequency phase labels and generates initial positioning candidate points based on multi-sensor time difference data. In phase intervals with higher power frequency amplitudes, the data weight of sensors susceptible to interference is reduced; in intervals with lower amplitudes, the weight is increased to enhance positioning confidence. If there are multiple candidate points in the same phase interval and the spatial discreteness is too large, the pulse waveform parameters are extracted for cluster analysis to distinguish between the true discharge source and the interference signal. For continuous discharge events across phase intervals, the pulse phase evolution law is tracked and the rationality of positioning is verified in combination with the transformer structure model. When the confidence level of the data in a phase interval is insufficient, the historical data of the adjacent interval is called for interpolation and completion, and the coordinate jump error is eliminated through spatiotemporal continuity verification. When the final positioning result is output, the uncertainty range is marked, and an early warning is triggered in the event of power frequency anomalies or synchronization failure.

[0108] A multi-parameter measurement method for a power frequency non-partial discharge test transformer comprises the following steps:

[0109] S1: Detect the zero-crossing point of the power frequency voltage signal, generate a synchronous trigger signal and divide the power frequency cycle into sub-windows with equal phase angles;

[0110] S2: Start high-speed sampling of high-frequency partial discharge signals in each sub-window and correct hardware delays using a phase deviation compensation algorithm;

[0111] S3: Dynamically switches the notch filter mode of the high-frequency channel according to the current phase angle and amplitude of the power frequency signal to suppress power frequency coupled noise;

[0112] S4: The hardware verification circuit compares the trigger signal and the sampling window start time difference in real time to generate an error feedback signal to drive the compensation algorithm;

[0113] S5: Input the synchronized power frequency signal and high frequency signal into the spatiotemporal joint analysis model, and calculate the discharge source coordinates by combining the phase label and time difference data.

[0114] It should be further explained that, in the specific implementation process, this method generates a synchronous trigger signal by real-time detection of the zero-crossing point of the power frequency voltage signal, divides the power frequency period into multiple equal-angle sub-windows based on a preset phase angle interval, and starts high-speed sampling of the high-frequency partial discharge signal in each sub-window.

[0115] When the high-frequency signal acquisition module's sampling window start time deviates from the theoretical phase interval due to hardware delay, the phase deviation compensation algorithm dynamically adjusts the trigger time of subsequent sub-windows. For example, when a fixed delay is detected, a trigger instruction is sent in advance to ensure continuous convergence of the timing error.

[0116] According to the current phase angle and amplitude of the power frequency signal, the filtering mode of the high-frequency channel is dynamically switched: in the phase range where the power frequency amplitude is higher than the preset threshold, the narrowband notch depth is enhanced to block the interference of the power frequency fundamental wave and harmonics; in the range where the amplitude is lower than the threshold, it switches to the wideband mode to retain the complete spectral characteristics of the partial discharge pulse.

[0117] If the power frequency signal amplitude changes rapidly due to transient fluctuations in the power grid, a gradual adjustment strategy is adopted in the filter mode switching process to prevent signal distortion caused by sudden parameter changes.

[0118] A hardware verification circuit compares the difference between the trigger signal and the actual start time of the sampling window in real time. When the deviation exceeds a threshold, an error feedback signal is generated to drive the iterative optimization of the compensation algorithm. At the same time, a software mirror channel compares the data spectrum energy distribution before and after synchronization optimization. If the energy difference in the high-frequency band (such as above 1MHz) exceeds the tolerance, secondary filtering of the dynamic filter bank or recalibration of system parameters is triggered.

[0119] For scenarios with multiple power sources, the time difference data is grouped and weighted in combination with phase tags in the spatiotemporal joint analysis. For example, the data weight of sensors susceptible to interference is reduced in phase intervals with higher power frequency amplitudes, while the weight is increased in intervals with lower amplitudes to enhance positioning reliability.

[0120] If there are multiple candidate positioning points in the same phase interval and the spatial discreteness is too large, pulse waveform feature matching is started to extract the rising edge slope, pulse width and spectrum main peak parameters for cluster analysis to distinguish the real discharge source from the reflected interference signal.

[0121] When the confidence level of data in a phase interval is detected to be insufficient, historical data from adjacent intervals is interpolated to complete the gap, and spatial and temporal continuity verification is used to eliminate positioning coordinate jump errors. The final output of the discharge source coordinates is annotated with the uncertainty range of the positioning result, and an early warning mechanism is triggered in the event of power frequency signal anomalies or synchronization failures.

[0122] The multi-parameter measurement method for power-frequency, non-partial-discharge test transformers optimizes positioning accuracy through a dynamic weighting mechanism. The weight coefficient is inversely proportional to the power-frequency amplitude and is quadratically weighted based on the sensor's historical reliability score. In scenarios with multiple discharge sources, pulse groups from different discharge sources are separated through phase-time difference clustering, and the coordinates of each cluster are independently calculated to prevent signal aliasing. The system supports redundant design and adaptive fault tolerance. In the event of hardware failure or signal anomaly, it automatically switches to a backup channel and triggers self-calibration to ensure long-term monitoring stability.

[0123] Through a power frequency phase-locked synchronous triggering mechanism and dynamic noise suppression technology, the phase mismatch problem between the power frequency signal and the high-frequency partial discharge signal in traditional multi-parameter measurements is effectively solved, improving the positioning accuracy of the partial discharge source and the reliability of insulation defect diagnosis. Specifically, the multi-level synchronous sampling architecture based on power frequency zero-crossing triggering achieves nanosecond-level timing error control. Combined with a phase-correlated dynamic filtering strategy, it suppresses power frequency coupling interference while retaining high-frequency pulse details, thereby improving the signal-to-noise ratio of weak discharge signals. The dual-modal verification mechanism ensures the robustness of the system in complex electromagnetic environments through cross-level collaboration of hardware verification and software mirroring. It is particularly suitable for long-term monitoring of multi-winding transformers or high-interference conditions, avoiding the misjudgment and data failure problems caused by asynchronous sampling in traditional methods.

[0124] Through the combined spatiotemporal analysis model and phase-constrained positioning algorithm, a deep fusion of multi-parameter data is achieved, reducing the error in localizing partial discharges compared to the traditional time-difference method. It also supports the parallel identification of multiple discharge sources and the intelligent filtering of interference signals. The system is compatible with multi-channel expansion and redundant design, automatically switching to backup channels and triggering self-calibration in the event of hardware failure or signal anomalies, improving equipment availability and maintenance efficiency. Furthermore, dynamic weight allocation and historical data interpolation mechanisms enhance the system's adaptability to transient grid disturbances, resolving measurement interruptions caused by power frequency fluctuations or short-term lockout. This system breaks through the limitations of traditional time-sharing sampling and static filtering, providing high-precision, high-reliability, multi-dimensional data support for transformer insulation status assessment and laying the technical foundation for intelligent fault warning and life prediction.

[0125] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0126] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A multi-parameter measurement system for power frequency non-partial discharge test transformer, characterized in that: include: The synchronous trigger unit is connected to the power frequency voltage signal acquisition terminal, and is used to detect the zero crossing point of the power frequency voltage signal and generate a synchronous trigger signal, dividing the power frequency cycle into multiple sub-windows with equal phase angles; A dynamic filter bank, with its input connected to the high-frequency partial discharge signal acquisition channel and its output connected to the signal fusion module, is configured to dynamically switch the filtering mode of the high-frequency channel based on the current phase angle and amplitude of the power frequency signal; The dual-modal verification module includes a hardware verification circuit and a software mirror channel. The hardware verification circuit is connected to the synchronous trigger unit and the clock control end of the high-frequency signal acquisition module, and the software mirror channel is embedded in the signal processing unit. The signal fusion module receives the synchronized power frequency signal and the filtered high-frequency partial discharge signal, performs a spatiotemporal joint analysis, and outputs the discharge source location coordinates; Among them, the synchronization trigger unit sends sampling instructions corresponding to the phase interval to the high-frequency signal acquisition module through multi-level clock tree distribution technology, the dynamic filter group adjusts the notch filter parameters based on the power frequency phase angle, and the dual-modal verification module realizes synchronization calibration through hardware real-time verification and software data comparison.

2. The multi-parameter measurement system for a power frequency non-partial discharge test transformer according to claim 1, characterized in that: The synchronization triggering unit includes: Zero-crossing detection circuit captures the zero-crossing point of the power frequency voltage signal in real time and outputs a trigger pulse; The phase interval division module divides the power frequency cycle into multiple sub-windows according to the preset angle intervals and sends sampling instructions corresponding to the sub-windows to the high-frequency signal acquisition module; The phase deviation compensation module dynamically adjusts the sending time of the sampling instruction according to the hardware delay parameters of the high-frequency signal acquisition module, so that the timing error is less than 10 nanoseconds.

3. The multi-parameter measurement system for a power frequency non-partial discharge test transformer according to claim 2, characterized in that: The phase interval division module synchronizes the start time of each subwindow to the clock control terminal of the high-frequency signal acquisition module through multi-level clock tree distribution technology, and embeds a real-time feedback loop in each subwindow to iteratively optimize the phase offset of the sampling instruction based on the error signal of the hardware verification circuit.

4. The multi-parameter measurement system for a power frequency non-partial discharge test transformer according to claim 1, characterized in that: The dynamic filter bank comprises: an adaptive notch filter configured to enhance the narrowband notch depth in a phase interval where the power frequency amplitude is above a preset threshold; a wideband filter configured to preserve the complete spectrum of the high frequency signal within a phase interval where the amplitude of the power frequency is below a preset threshold; The filter mode switching logic dynamically selects a notch filter or a wideband filter to access the high-frequency signal channel based on the current phase angle and amplitude of the power frequency signal.

5. The multi-parameter measurement system for power frequency non-partial discharge test transformer according to claim 4, characterized in that: The dynamic filter bank also includes a cross-band cross-correlation analysis module, which inputs the power frequency signal and the high-frequency partial discharge signal, extracts the time delay characteristics of the partial discharge pulse through joint analysis in the time and frequency domains, and uses a blind source separation algorithm to eliminate the residual interference of the power frequency harmonics.

6. The multi-parameter measurement system for power frequency non-partial discharge test transformer according to claim 1, characterized in that: The dual-modal verification module performs the following operations: The hardware verification circuit collects the actual start time difference between the synchronous trigger signal and the high-frequency sampling window in real time, and generates an error voltage signal to drive the phase deviation compensation module; The software mirror channel compares the spectral energy distribution of the original asynchronously sampled data with the synchronously optimized data. If the energy difference in a specific frequency band exceeds a threshold, the secondary filtering logic of the dynamic filter bank is triggered.

7. The multi-parameter measurement system for power frequency non-partial discharge test transformer according to claim 1, characterized in that: The signal fusion module includes: A phase label generation unit groups partial discharge pulses according to the power frequency period sub-window and marks the phase interval to which they belong; The phase-constrained positioning unit combines the time difference data and phase interval labels of multiple sensors to construct a weighted least squares positioning equation to correct the cumulative error caused by phase mismatch.

8. The multi-parameter measurement system for power frequency non-partial discharge test transformer according to claim 7, characterized in that: In the phase constraint positioning unit, the time difference data of each partial discharge pulse is allocated a weight coefficient according to the phase interval to which it belongs, and the weight coefficient is inversely proportional to the power frequency amplitude of the phase interval.

9. The multi-parameter measurement system for power frequency non-partial discharge test transformer according to claim 6, characterized in that: The software mirror channel adopts a redundant sampling strategy to store the time domain waveform and spectrum data of the high-frequency signal before and after synchronization optimization, and determines whether to trigger system parameter recalibration through a preset difference threshold.

10. A multi-parameter measurement method for a power frequency non-partial discharge test transformer, characterized in that: The steps include: S1: Detect the zero-crossing point of the power frequency voltage signal, generate a synchronous trigger signal and divide the power frequency cycle into sub-windows with equal phase angles; S2: Start high-speed sampling of high-frequency partial discharge signals in each sub-window and correct hardware delays using a phase deviation compensation algorithm; S3: Dynamically switches the notch filter mode of the high-frequency channel according to the current phase angle and amplitude of the power frequency signal to suppress power frequency coupled noise; S4: The hardware verification circuit compares the trigger signal and the sampling window start time difference in real time to generate an error feedback signal to drive the compensation algorithm; S5: Input the synchronized power frequency signal and high frequency signal into the spatiotemporal joint analysis model, and calculate the discharge source coordinates by combining the phase label and time difference data.

Images