A method for diagnosing a transformer winding fault

Abstract

The present application relates to the field of electrical equipment test and converter transformer monitoring, and discloses a kind of diagnostic methods for transformer winding fault, comprising: obtaining 500 kilovolt and above DC converter transformer port transient voltage waveform and transient current waveform and DC bias current measurement value;When the absolute value of time derivative is greater than the voltage change rate threshold set between 0.5kV / μs to 5kV / μs, determine the step electric excitation wave trigger time and intercept window;Calculate high-frequency capacitance response factor and high-frequency inductance derivative factor;Transient electromagnetic coupling eigenvalue is calculated from current change rate;According to the static bias current measurement value of magnetic correction coefficient, algebraic compensation is carried out and winding inherent impedance scalar is calculated;The present application uses commutation voltage jump as excitation, offsets magnetization relaxation dynamic phase angle distortion, eliminates nonlinear impedance drift, realizes the online electrical test of DC converter transformer winding fault and improves fault diagnosis sensitivity.

Description

Technical Field

[0001] This invention belongs to the field of electrical equipment testing and converter transformer monitoring technology, and particularly relates to a diagnostic method for transformer winding faults. Background Technology

[0002] In 500kV and above ultra-high voltage direct current (UHVDC) transmission systems or large converter stations, the grid-side and valve-side windings of 500kV and above DC converter transformers are subjected to the combined effects of AC / DC electric fields and DC bias magnetism for extended periods. This can easily lead to localized geometric deformation, inter-strand short circuits, or loosening of the support blocks. Existing technologies typically employ offline frequency sweep response analysis to monitor winding conditions. This involves injecting a frequency sweep signal into the transformer ports after power outages and measuring the corresponding response characteristics. However, the power outage state cannot accurately reflect the DC bias magnetism saturation and electromagnetic stress distribution during transformer operation. Some winding displacements only occur under energized stress conditions and will elastically recover after power outages due to the disappearance of stress. Therefore, it is difficult to detect slight structural deformations. Online electrical variable measurement technology generally judges the equipment status by collecting voltage and current signals at the transformer port. The valve side of 500kV and above ultra-high voltage converter transformers is connected to the converter valve, and there are steep step voltage pulsations at the port for a long time. Traditional methods usually process them according to the linear superposition of frequency domain, regard these switching step waves as interference, remove high frequency ripples by filtering, and then extract the power frequency equivalent impedance. However, when high frequency step waves propagate in the winding, they are very sensitive to the slight changes in local distributed capacitance and inter-turn inductance. Filtering out high frequency step waves will also weaken the high frequency characteristics generated by slight geometric changes in the winding.

[0003] Existing DC bias compensation methods typically treat core saturation caused by DC bias as static inductance drift and compensate based on steady-state DC bias current and static correction coefficients. However, within the microsecond-level time window after a step voltage wave enters the winding, the changes in core magnetic domains exhibit magnetization relaxation hysteresis. This hysteresis causes a change in the response phase of high-frequency charge and flux accumulation at the port, resulting in impedance distortion related to the rate of change. Using only static correction coefficients is insufficient to eliminate the effects of hysteresis relaxation during transient processes. Residual phase angle distortion can mask the electrical characteristics corresponding to winding geometric deformation. In addition to the physical limitations at the hardware sensing and signal acquisition levels, existing diagnostic and control methods also face limitations at the software algorithm level. However, this method has its shortcomings. For example, Chinese invention patent application CN104991160A discloses a transformer winding fault detection and diagnosis method based on voltage and current characteristics. It injects high-frequency pulses by installing a capacitive coupling sensor and extracts the fundamental component using short-time Fourier transform, and then extracts elliptic feature parameters for comparison and identification. However, this method essentially relies on frequency domain averaging, and loses transient high-frequency details reflecting the small deformation of the winding during the algorithm noise reduction and fundamental extraction process. At the same time, its feature parameter extraction model cannot overcome the phase angle distortion interference caused by core hysteresis relaxation within the microsecond-level transient window, making it difficult to accurately distinguish early small deformations under strong AC / DC combined bias magnetic conditions.

[0004] Therefore, how to utilize the step wave generated by the commutation of the converter valve as an excitation to separate the influence of the dynamic magnetization relaxation of the iron core and the changes in the winding distribution parameters within the microsecond-level transient window, and to achieve accurate online measurement of the transformer's electrical parameters, is a technical problem that needs to be solved in the field of testing and monitoring of 500kV and above DC converter transformers. Summary of the Invention

[0005] To address the problems in the background art, the technical solution of the present invention is as follows: A method for diagnosing transformer winding faults, comprising the following steps: Step S1: Obtain the transient voltage waveform and transient current waveform at the port of the 500 kV and above DC converter transformer, and obtain the DC bias current measurement value. Step S2: Monitor the time derivative of the transient voltage waveform. When the absolute value of the time derivative is greater than the preset voltage change rate threshold, determine the trigger time of the step shock wave. The voltage change rate threshold is set in the range of 0.5kV / μs to 5kV / μs. Step S3: Starting from the trigger time, extract a time-domain sampling window with a duration of 20 μs; Step S4: Within the time-domain sampling window, calculate the ratio of the time integral value of the transient current waveform to the amplitude change of the transient voltage waveform to obtain the high-frequency capacitor response factor; and calculate the ratio of the time integral value of the transient voltage waveform to the amplitude change of the transient current waveform to obtain the high-frequency inductor derivative factor. Step S5: Within the time-domain sampling window, multiply the time-domain differential rate of change of the transient voltage waveform and the transient current waveform and calculate the time integral to obtain the transient electromagnetic coupling characteristic value. Step S6: Determine the static bias correction coefficient based on the DC bias current measurement value. Use the static bias correction coefficient and the transient electromagnetic coupling characteristic value to perform algebraic compensation on the high-frequency inductor derivative factor and the high-frequency capacitor response factor, and calculate the winding's inherent impedance scalar. Step S7: Calculate the relative deviation between the inherent impedance scalar of the winding and the health reference value, and output the transformer winding fault diagnosis result based on the relative deviation value.

[0006] Preferably, in step S7, the output of transformer winding fault diagnosis results based on the relative deviation value includes: when the relative deviation value is less than 0.02, the output result indicates that the winding structure of the 500 kV and above DC converter transformer is intact; when the relative deviation value is in the range of 0.02 to 0.05, the output result indicates that the winding of the 500 kV and above DC converter transformer has minor deformation warning; when the relative deviation value is greater than 0.05, the output result indicates that the winding of the 500 kV and above DC converter transformer has structural deformation and inter-turn short circuit fault.

[0007] Preferably, in step S4, obtaining the high-frequency capacitor response factor includes: calculating the time integral value of the absolute value of the transient current waveform within the time-domain sampling window, and dividing the time integral value by the difference between the maximum and minimum values ​​of the transient voltage waveform within the time-domain sampling window; obtaining the high-frequency inductor derivative factor includes: calculating the time integral value of the absolute value of the transient voltage waveform within the time-domain sampling window, and dividing the time integral value by the difference between the maximum and minimum values ​​of the transient current waveform within the time-domain sampling window.

[0008] Preferably, in step S5, obtaining the transient electromagnetic coupling characteristic value includes: multiplying the transient voltage waveform amplitude at each sampling moment within the time-domain sampling window by the time-domain differential rate of change at that sampling moment to obtain the transient charge product value at each sampling moment; and performing time-domain accumulation and integration on the transient charge product value to obtain the transient electromagnetic coupling characteristic value.

[0009] Preferably, step S1, which involves acquiring the transient voltage and transient current waveforms at the ports of a 500 kV or higher DC converter transformer, includes: simultaneously acquiring the transient current at the valve-side port and the transient current at the grid-side port of the 500 kV or higher DC converter transformer; and further includes the following sub-steps: step S11, calculating the time integral of the absolute value of the transient current at the valve-side port and the time integral of the absolute value of the transient current at the grid-side port within the time-domain sampling window; step S12, dividing the time integral of the absolute value of the transient current at the valve-side port by the time integral of the absolute value of the transient current at the grid-side port to calculate the cross-winding high-frequency charge transfer ratio factor.

[0010] Preferably, after step S6 and before step S7, the method further includes a step of correcting the winding intrinsic impedance scalar using the cross-winding high-frequency charge transfer ratio factor, including the following sub-steps: Step S61, multiplying the cross-winding high-frequency charge transfer ratio factor as a dimensionless gain coefficient with the winding intrinsic impedance scalar to generate the corrected winding intrinsic impedance scalar; Step S62, replacing the original winding intrinsic impedance scalar with the corrected winding intrinsic impedance scalar and inputting it into step S7.

[0011] Preferably, in step S1, obtaining the DC bias current measurement value includes reading real-time DC bias current data using a current transformer deployed at the neutral point of a 500 kV or higher DC converter transformer; the health reference value includes the historical average of the inherent impedance stored in advance under the same DC bias current measurement conditions when the 500 kV or higher DC converter transformer is in a fault-free state.

[0012] Preferably, while monitoring the time derivative of the transient voltage waveform in step S2, the following sub-steps are also included: Step S21, continuously calculate the time rate of change of the DC bias current measurement value; Step S22, compare the absolute value of the time rate of change with the set transient impact threshold of 50A / ms; Step S23, when the absolute value of the time rate of change is greater than 50A / ms, trigger the measurement protection state, suspend the update calculation of the winding inherent impedance scalar, and lock the current transformer winding fault diagnosis result for output.

[0013] Preferably, after triggering the measurement protection state in step S23, the following sub-steps are also included: step S231, continuously monitoring the absolute value of the time change rate of the DC bias current measurement value; step S232, when the absolute value of the time change rate decays to below 10A / ms and the duration exceeds 100ms, the measurement protection state is released, and the winding intrinsic impedance scalar is recalculated after the time domain sampling window is reopened.

[0014] Compared with existing technologies, the present invention provides a method for diagnosing transformer winding faults, which has the following advantages: 1. In the electrical testing and diagnosis of DC converter transformer winding faults, the step voltage generated by the commutation of the converter valve is used as the detection excitation, and the transient voltage and transient current characteristics of the port are extracted within a microsecond time window. Since the response of the winding distributed capacitance and inter-turn inductance to the high-frequency step wave is faster than the dynamic magnetization change of the iron core, the influence of DC bias and iron core saturation drift on the detection results can be reduced, the high-frequency response changes caused by slight geometric deformation of the winding are preserved, and the ability to identify early deformation of DC converter transformer windings under energized operation is improved.

[0015] 2. This invention combines transient electromagnetic coupling characteristic values ​​and static bias correction coefficients to compensate for high-frequency inductance derivative factors and high-frequency capacitor response factors. This reduces the interference of phase changes caused by core magnetization relaxation on impedance calculation. The resulting winding inherent impedance scalar better reflects the changes in electrical parameters caused by changes in winding structure. This reduces the possibility of misjudging the influence of core hysteresis as winding displacement and improves the accuracy of fault diagnosis.

[0016] 3. This invention uses the high-frequency transient currents on the valve side and the grid side to correct the inherent impedance scalar of the winding. The external electromagnetic interference experienced by both sides of the transformer can be weakened during the correction process, thereby reducing the impact of grid noise on the diagnostic results. Without increasing the hardware shielding structure, it can also improve the stability of feature recognition and online diagnosis in complex electromagnetic environments. Attached Figure Description

[0017] Figure 1 This is a flowchart of the calculation of the scalar value of the inherent impedance of the DC converter transformer winding in this invention; Figure 2 This is a diagnostic feature parameter diagram of the DC converter transformer winding of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0019] A method for diagnosing transformer winding faults includes the following steps: Step S1: Obtain the transient voltage waveform and transient current waveform at the port of the 500 kV and above DC converter transformer, and obtain the DC bias current measurement value. Step S2: Monitor the time derivative of the transient voltage waveform. When the absolute value of the time derivative is greater than the preset voltage change rate threshold, determine the trigger time of the step shock wave. The voltage change rate threshold is set in the range of 0.5kV / μs to 5kV / μs. Step S3: Starting from the trigger time, extract a time-domain sampling window with a duration of 20 μs; Step S4: Within the time-domain sampling window, calculate the ratio of the time integral value of the transient current waveform to the amplitude change of the transient voltage waveform to obtain the high-frequency capacitor response factor; and calculate the ratio of the time integral value of the transient voltage waveform to the amplitude change of the transient current waveform to obtain the high-frequency inductor derivative factor. Step S5: Within the time-domain sampling window, multiply the time-domain differential rate of change of the transient voltage waveform and the transient current waveform and calculate the time integral to obtain the transient electromagnetic coupling characteristic value. Step S6: Determine the static bias correction coefficient based on the DC bias current measurement value. Use the static bias correction coefficient and the transient electromagnetic coupling characteristic value to perform algebraic compensation on the high-frequency inductor derivative factor and the high-frequency capacitor response factor, and calculate the winding's inherent impedance scalar. Step S7: Calculate the relative deviation between the inherent impedance scalar of the winding and the health reference value, and output the transformer winding fault diagnosis result based on the relative deviation value.

[0020] Preferably, in step S7, the output of transformer winding fault diagnosis results based on the relative deviation value includes: when the relative deviation value is less than 0.02, the output result indicates that the winding structure of the 500 kV and above DC converter transformer is intact; when the relative deviation value is in the range of 0.02 to 0.05, the output result indicates that the winding of the 500 kV and above DC converter transformer has minor deformation warning; when the relative deviation value is greater than 0.05, the output result indicates that the winding of the 500 kV and above DC converter transformer has structural deformation and inter-turn short circuit fault.

[0021] Preferably, in step S4, obtaining the high-frequency capacitor response factor includes: calculating the time integral value of the absolute value of the transient current waveform within the time-domain sampling window, and dividing the time integral value by the difference between the maximum and minimum values ​​of the transient voltage waveform within the time-domain sampling window; obtaining the high-frequency inductor derivative factor includes: calculating the time integral value of the absolute value of the transient voltage waveform within the time-domain sampling window, and dividing the time integral value by the difference between the maximum and minimum values ​​of the transient current waveform within the time-domain sampling window.

[0022] Preferably, in step S5, obtaining the transient electromagnetic coupling characteristic value includes: multiplying the transient voltage waveform amplitude at each sampling moment within the time-domain sampling window by the time-domain differential rate of change at that sampling moment to obtain the transient charge product value at each sampling moment; and performing time-domain accumulation and integration on the transient charge product value to obtain the transient electromagnetic coupling characteristic value.

[0023] Preferably, step S1, which involves acquiring the transient voltage and transient current waveforms at the ports of a 500 kV or higher DC converter transformer, includes: simultaneously acquiring the transient current at the valve-side port and the transient current at the grid-side port of the 500 kV or higher DC converter transformer; and further includes the following sub-steps: step S11, calculating the time integral of the absolute value of the transient current at the valve-side port and the time integral of the absolute value of the transient current at the grid-side port within the time-domain sampling window; step S12, dividing the time integral of the absolute value of the transient current at the valve-side port by the time integral of the absolute value of the transient current at the grid-side port to calculate the cross-winding high-frequency charge transfer ratio factor.

[0024] Preferably, after step S6 and before step S7, the method further includes a step of correcting the winding intrinsic impedance scalar using the cross-winding high-frequency charge transfer ratio factor, including the following sub-steps: Step S61, multiplying the cross-winding high-frequency charge transfer ratio factor as a dimensionless gain coefficient with the winding intrinsic impedance scalar to generate the corrected winding intrinsic impedance scalar; Step S62, replacing the original winding intrinsic impedance scalar with the corrected winding intrinsic impedance scalar and inputting it into step S7.

[0025] Preferably, in step S1, obtaining the DC bias current measurement value includes reading real-time DC bias current data using a current transformer deployed at the neutral point of a 500 kV or higher DC converter transformer; the health reference value includes the historical average of the inherent impedance stored in advance under the same DC bias current measurement conditions when the 500 kV or higher DC converter transformer is in a fault-free state.

[0026] Preferably, while monitoring the time derivative of the transient voltage waveform in step S2, the following sub-steps are also included: Step S21, continuously calculate the time rate of change of the DC bias current measurement value; Step S22, compare the absolute value of the time rate of change with the set transient impact threshold of 50A / ms; Step S23, when the absolute value of the time rate of change is greater than 50A / ms, trigger the measurement protection state, suspend the update calculation of the winding inherent impedance scalar, and lock the current transformer winding fault diagnosis result for output.

[0027] Preferably, after triggering the measurement protection state in step S23, the following sub-steps are also included: step S231, continuously monitoring the absolute value of the time change rate of the DC bias current measurement value; step S232, when the absolute value of the time change rate decays to below 10A / ms and the duration exceeds 100ms, the measurement protection state is released, and the winding intrinsic impedance scalar is recalculated after the time domain sampling window is reopened.

[0028] Example 1: This example uses a 500kV DC converter transformer operating in a 500kV DC converter station for online diagnosis. The 500kV DC converter transformer is continuously operating at full load. The total harmonic distortion rate of the grid voltage is 12%, and the measured value of the DC bias current is 50A. The multi-band grid-side harmonic vibration components and the intrinsic vibration components generated by the winding stress deformation overlap in the frequency domain. It is difficult to identify the mechanical structural changes caused by winding loosening based solely on the amplitude of the vibration signal.

[0029] After the diagnosis begins, the electrical signal acquisition module synchronously acquires the transient voltage waveform and transient current waveform at the port of the 500kV DC converter transformer. The current transformer at the neutral point reads the real-time DC bias current data. The voltage input range of the electrical signal acquisition module is set to 0V to 10V, the analog-to-digital conversion accuracy is not less than 16 bits, and the sampling frequency is set according to the differential and integral operation requirements within the 20μs time domain sampling window, and is not less than 20kHz.

[0030] The central processing unit continuously calculates the time derivative of the transient voltage waveform. When the absolute value of the time derivative is greater than 1.5kV / μs, the current moment is determined as the trigger moment of the step shock wave. A time-domain sampling window with a length of 20μs is extracted from this trigger moment. The voltage change rate threshold of 1.5kV / μs is higher than the 500kV grid-side conventional power grid fluctuation limit of 0.5kV / μs, which is used to distinguish the step shock wave generated by the commutation of the converter valve from general power grid fluctuations. The sampling window length of 20μs is greater than the 15μs required for the high-frequency traveling wave to travel back and forth three times in the winding, and less than the 100μs start time of the core dynamic hysteresis relaxation, so as to preserve the high-frequency traveling wave response within the window.

[0031] Within this time-domain sampling window, the central processing unit integrates the absolute value of the transient current waveform over time and divides the resulting time integral value by the difference between the maximum and minimum values ​​of the transient voltage waveform to obtain the high-frequency capacitor response factor; simultaneously, it integrates the absolute value of the transient voltage waveform over time and divides the resulting time integral value by the difference between the maximum and minimum values ​​of the transient current waveform to obtain the high-frequency inductor derivative factor.

[0032] For each sampling moment within the time-domain sampling window, the central processing unit calculates the time-domain differential rate of change of the transient current waveform based on the adjacent transient current sampling values ​​and the corresponding sampling interval. This time-domain differential rate of change is multiplied by the amplitude of the transient voltage waveform at the same sampling moment to obtain the transient charge product value. The transient charge product values ​​within the window are accumulated and integrated according to the sampling time sequence to obtain the transient electromagnetic coupling characteristic value.

[0033] The central processing unit multiplies the 50A DC bias current collected by the neutral point current transformer by a proportional parameter pre-calibrated in the non-saturated linear region of the iron core to obtain a static bias correction coefficient of 0.05. Subsequently, the high-frequency capacitor response factor is multiplied by the static bias correction coefficient to obtain the capacitor compensation amount. The high-frequency inductance derivative factor is multiplied by the product of the transient electromagnetic coupling characteristic value and the static bias correction coefficient to obtain the inductance compensation amount. The capacitor compensation amount is subtracted from the high-frequency capacitor response factor to obtain the actual capacitor parameter. The inductance compensation amount is added to the high-frequency inductance derivative factor to obtain the actual inductance parameter. The actual inductance parameter is divided by the actual capacitor parameter, and the square root of the resulting ratio is taken to obtain the scalar value of the winding's inherent impedance.

[0034] The health reference value is the historical average value of the inherent impedance pre-stored when the DC converter transformer is in a fault-free state and the DC bias current measurement value is 50A. The central processing unit calculates the absolute value of the difference between the winding inherent impedance scalar and the health reference value, and then divides the absolute value by the health reference value to obtain the relative deviation value. When the relative deviation value is less than 0.02, the output winding structure is intact; when the relative deviation value is in the range of 0.02 to 0.05, the output winding has a small deformation warning result; when the relative deviation value is greater than 0.05, the output winding has structural deformation and inter-turn short circuit fault results.

[0035] While monitoring the transient voltage waveform, the central processing unit continuously calculates the time rate of change of the DC bias current measurement. The transient impact threshold of 50A / ms corresponds to the critical rate of change when the transformer enters deep bias saturation. When the absolute value of the time rate of change exceeds this threshold, the central processing unit pauses updating the winding intrinsic impedance scalar and maintains the current transformer winding fault diagnosis result. 10A / ms and 100ms are used as the recovery rate of change and duration criteria, respectively. When the absolute value of the time rate of change decays to below 10A / ms and continues to exceed 100ms, the measurement protection state is released, the time domain sampling window is reopened, and the winding intrinsic impedance scalar is calculated.

[0036] This embodiment also retains the electrical relaxation characteristic channel and the mechanical vibration characteristic channel for verifying the port transient electrical variable diagnosis results. The data processing unit uses the electrical signal acquisition module to obtain the electrical relaxation time after the pulse excitation applied to the bushing end screen of the DC converter transformer is released. The measured electrical relaxation time is subtracted from the pre-stored nominal relaxation time to obtain the relaxation time deviation. Then, the deviation is multiplied by the capacitance transformation constant to calculate the change in the inherent capacitance of the winding, and this is used as the electrical fingerprint signal.

[0037] The vibration sensor synchronously acquires the mechanical vibration signal of the transformer tank wall at a sampling rate of not less than 50kHz. The data processing unit converts the electrical fingerprint signal into a calibration coefficient array, performs point-by-point weighted product and inverse fast Fourier transform on the spectrum of the mechanical vibration signal; at the same time, using the change in the inherent capacitance of the winding as the calibration coefficient, the frequency domain cross-correlation calculation is performed on the mechanical vibration signal, and the electrical nonlinear state parameter corresponding to the electrical fingerprint signal is used as the decoupling and stripping condition for the harmonic components, separating the vibration component caused by the grid-side voltage harmonics from the mechanical vibration signal. The capacitive relaxation response is used to constrain the identification boundary of the mechanical vibration signal and correct the measurement deviation of the vibration channel.

[0038] When the pulse excitation of the bushing end screen cannot be applied due to external maintenance of the transformer, the central processing unit reads the transient operating voltage generated at the moment the transformer tap changer operates, uses it as an alternating excitation source and extracts the equivalent electrical relaxation characteristics. When calculating the decoupling and stripping results, if the energy terms of the extracted intrinsic spectral lines exceed the product of the preset turns ratio coefficient and the grid-side voltage within 30 consecutive sampling periods, a trigger signal is generated and a control command is output to the downstream actuator.

[0039] When multiple transformers are running in parallel, the station-side centralized communication controller receives global diagnostic tasks and calculates the task allocation results based on the idle rate of the processor core of each converter transformer diagnostic unit. When the total harmonic distortion rate of the grid voltage is less than 2% and the load variation rate is less than 5%, the diagnostic unit performs uplink calibration reconstruction and updates the initial baseline value using a sliding time window containing the latest 5 cycles of data to correct parameter drift caused by temperature rise. When the inherent resonant frequency offset exceeds the set threshold, the diagnostic unit issues a downlink locking alarm.

[0040] The data processing unit identifies frequency domain information through matrix eigenvalue decomposition based on the decoupled intrinsic vibration components and dynamically updated initial baseline values. It calculates the inherent resonant frequency offset of the three phase windings of the DC converter transformer. When the inherent resonant frequency of phase A winding deviates negatively by more than 5Hz relative to the initial baseline value, it determines that the phase winding has structural loosening. The unit then sends a fault feature vector containing the frequency offset, fault phase identifier, and current operating condition parameters to the local display terminal to display the location of mechanical defects and deformation trends inside the winding.

[0041] The electrical relaxation characteristic channel and the mechanical vibration characteristic channel form a mutual calibration loop. The two monitoring channels calibrate the dynamic data stream based on the correspondence between the capacitance relaxation response and the natural resonant frequency. During equipment operation, the zero-point drift of a single channel and the measurement deviation caused by changes in the environmental baseline are corrected. The relative deviation value output by the port transient electrical variable channel and the transformer winding fault diagnosis result are used as the main diagnostic results. The winding inherent capacitance change, inherent resonant frequency offset and fault phase identifier output by the electrical relaxation characteristic channel and the mechanical vibration characteristic channel are used for auxiliary verification.

[0042] Example 2: This example was conducted on a test platform equipped with a high-voltage AC / DC superposition physical simulation circuit. The test platform was equipped with an electrical signal acquisition module with an analog-to-digital conversion accuracy of not less than 16 bits and an input range of 0V to 10V. The resolution of the spectrum analysis module was 0.01Hz, and the maximum adjustment capability of the DC bias source was 100A. The physical simulation circuit simulated the substation operating conditions of a 500kV DC converter station. The 500kV DC converter transformer was continuously operating at full load. The measured value of the DC bias current was 50A, and the total harmonic distortion rate of the grid voltage was 12%.

[0043] The experiment used the method of this invention and the conventional amplitude monitoring method to process the same set of original signals. By releasing the mechanical preload of the transformer winding, four states with a preload reduction ratio of 0%, 10%, 20% and 30% were constructed in sequence, which correspond to the normal state with no loosening of the winding structure, the slightly loose state, the moderately loose state and the loose state, respectively.

[0044] The electrical signal acquisition module synchronously acquires transient voltage and transient current waveforms at the ports of 500 kV and above DC converter transformers, and reads the measured value of DC bias current. The sampling frequency is determined based on the variation range of the electrical relaxation time constant and the spectral distribution of high-frequency electromagnetic interference. When the electrical relaxation time constant shortens, the equivalent bandwidth of the signal increases accordingly, and the sampling frequency is adjusted to the upper limit of the set range. Under the condition that the total harmonic distortion rate of the grid voltage is 12%, the sampling frequency is set to 20 kHz to meet the requirements of the completeness of the acquisition of the relaxation characteristic waveform and the real-time computing load of the data processing unit.

[0045] The rise time of the commutation voltage step wave of the converter valve is less than 0.5μs. The charging and discharging response of the local distributed capacitance and inter-turn inductance of the winding occurs within 1μs to 5μs. The domain wall displacement and magnetization relaxation time constant of the iron core under the superimposed bias of AC and DC are greater than 1ms. The central processing unit determines the triggering time of the step shock wave based on the time derivative of the transient voltage waveform and extracts a 20μs time-domain sampling window from that time. Within this window, the magnetization saturation state of the iron core is treated as a quasi-static quantity. The high-frequency electrical signal at the port is used to calculate the high-frequency capacitor response factor, the high-frequency inductor derivative factor, and the transient electromagnetic coupling characteristic value.

[0046] During signal acquisition, Gaussian white noise with a signal-to-noise ratio of 20dB and power frequency electromagnetic interference with a frequency of 50Hz are superimposed at the signal acquisition input terminal. The central processing unit reads the 50A DC bias current collected by the neutral point current transformer, multiplies it by the proportional parameter calibrated in advance in the non-saturated linear region of the iron core, and obtains a static bias correction coefficient with a value of 0.05. Then, algebraic compensation is performed on the high-frequency capacitor response factor and the high-frequency inductance derivative factor, and the winding inherent impedance scalar is calculated based on the compensated actual inductance parameter and actual capacitance parameter.

[0047] The health baseline value is the historical average value of the inherent impedance pre-stored under the condition that the DC converter transformer is in a fault-free state and the DC bias current measurement value is the same. The central processing unit calculates the relative deviation between the winding inherent impedance scalar and the health baseline value, and uses 0.02 and 0.05 as the boundaries of the diagnostic interval: when the relative deviation value is less than 0.02, the output winding structure is intact; when the relative deviation value is in the range of 0.02 to 0.05, the output winding has a small deformation warning result; when the relative deviation value is greater than 0.05, the output winding has structural deformation and inter-turn short circuit fault results.

[0048] To verify the diagnostic results of the inherent impedance scalar of the winding, the test simultaneously collected the mechanical vibration signal of the transformer tank wall and used the electrical relaxation time after the pulse excitation was released to calculate the change in the inherent capacitance of the winding. The data processing unit used this change as a calibration coefficient to perform frequency domain cross-correlation calculation on the mechanical vibration signal, separated the vibration component caused by the grid side harmonics from the original vibration signal, and then calculated the decoupled inherent resonant frequency offset.

[0049] During the auxiliary determination process, the intrinsic spectral energy terms extracted within 30 consecutive sampling periods are compared with the product of the preset transformation ratio coefficient and the grid-side voltage. The determination relationship is as follows: ,in, This represents the intrinsic spectral energy term extracted from the decoupled mechanical vibration signal over 30 consecutive sampling periods. This indicates the preset ratio coefficient. The grid-side voltage is represented by a preset transformation ratio coefficient that includes the piezoelectric conversion constant of the vibration sensor and the electromechanical coupling matrix of the tank wall. Its value represents the upper limit of the equivalent mechanical vibration energy excited by a unit grid-side voltage fluctuation in the tank wall. Greater than and When the product is multiplied, the data processing unit generates a trigger signal and outputs control commands to the downstream actuator.

[0050] Under normal conditions where the winding preload decreases by 0%, the amplitude of the original vibration signal acquired by the conventional amplitude monitoring method is 4.2 m / s. 2The change in inherent capacitance of the winding obtained from electrical fingerprint analysis is 0.0 pF, and the shift in inherent resonant frequency is 0.0 Hz, indicating a normal output state. When processing the same state using the method of this invention, the amplitude of the original vibration signal is consistently 4.2 m / s. 2 The inherent capacitance change of the winding is 0.0pF, and the inherent resonant frequency offset after decoupling is 0.0Hz. The relative deviation between the inherent impedance scalar of the winding and the healthy reference value is within the range of intact winding structure. Both the main diagnostic results and the auxiliary verification results are normal.

[0051] When the winding preload decreases by 10%, the amplitude of the original vibration signal increases to 4.3 m / s. 2 The conventional amplitude monitoring method failed to analyze the change in the inherent capacitance of the winding, and the obtained change was 0.0pF. The inherent resonant frequency offset was 0.2Hz, which was still in a normal state. After processing with the method of this invention, the change in the inherent capacitance of the winding obtained by the electrical relaxation characteristic channel was -12.4pF, and the inherent resonant frequency offset obtained after frequency domain cross-correlation processing was -2.1Hz. The relative deviation of the inherent impedance scalar of the winding was still within the range of good winding structure. The main diagnostic results and auxiliary verification results were both in a normal state.

[0052] When the winding preload decreases by 20%, the original vibration signal amplitude is 4.5 m / s. 2 The change in inherent capacitance of the winding obtained by the conventional amplitude monitoring method is still 0.0pF, and the inherent resonant frequency offset is -0.5Hz, which does not reach the auxiliary judgment threshold of 5Hz, so the normal state continues to be output. After processing with the method of this invention, the change in inherent capacitance of the winding is -28.7pF, and the inherent resonant frequency offset after decoupling is -5.3Hz. The theoretical expected value derived from the capacitor relaxation response is 5.4Hz, and the difference between the two is 0.1Hz. The resolution of the spectrum analysis module is 0.01Hz. At this time, the relative deviation value of the inherent impedance scalar of the winding enters the small deformation warning range, and the main diagnostic link outputs a warning result of small deformation of the winding. The inherent resonant frequency offset of the auxiliary channel exceeds 5Hz, and the winding structure loosening alarm is output.

[0053] When the winding preload decreases by 30%, the original vibration signal amplitude is 4.6 m / s. 2 The conventional amplitude monitoring method yielded a winding inherent capacitance change of 0.0 pF and an inherent resonant frequency offset of -0.9 Hz, still indicating a normal output. After processing with the method of this invention, the winding inherent capacitance change was -45.1 pF, and the decoupled inherent resonant frequency offset was -8.6 Hz. The relative deviation of the winding inherent impedance scalar entered the structural deformation fault range. The main diagnostic link output the results of winding structural deformation and inter-turn short circuit faults, and the auxiliary channel synchronously output a winding structural loosening alarm.

[0054] As can be seen from the four sets of state data, the amplitude of the original vibration signal acquired by the conventional amplitude monitoring method is only 4.2 m / s. 2 Increased to 4.6 m / s 2 The harmonic vibration components on the grid side mask the intrinsic vibration changes caused by the changes in the winding structure. The inherent resonant frequency offset does not reach 5Hz under any fault gradient, so all outputs are in normal condition. After adopting the method of this invention, the change in the inherent capacitance of the winding changes from 0.0pF to -12.4pF, -28.7pF and -45.1pF as the preload decreases. The inherent resonant frequency offset after decoupling changes from 0.0Hz to -2.1Hz, -5.3Hz and -8.6Hz respectively. The 20% and 30% preload decrease states are both identified by the auxiliary channel as loose winding structure.

[0055] As the total harmonic distortion rate of the grid-side voltage increases from 2% to 12%, the conventional amplitude monitoring method cannot separate the harmonic vibration components. Its diagnostic error increases with the total harmonic distortion rate. After the total harmonic distortion rate of the grid-side voltage exceeds 14%, the capacitor relaxation response signal is overloaded and modulated. The electrical fingerprint signal enters the nonlinear saturation region, and the calculation error of frequency domain cross-correlation increases from 0.1Hz to 2.5Hz. The auxiliary channel then issues a downlink lockout alarm. Therefore, in this embodiment, the operating boundary of the total harmonic distortion rate of the grid-side voltage of the auxiliary electrical fingerprint channel is set to no more than 12%.

[0056] The two boundaries of the relative deviation value are determined based on finite element electromagnetic field simulation and actual short-circuit impact test data. When the relative deviation value is less than 0.02, the corresponding local deformation size of the winding is less than 1.5mm, which is within the normal range of thermal expansion and contraction and mechanical tolerance. When the relative deviation value is in the range of 0.02 to 0.05, the corresponding deformation size is 1.5mm to 5mm, which indicates a small geometric deformation caused by the loosening of the pad. When the relative deviation value is greater than 0.05, the corresponding deformation size is greater than 5mm or an inter-turn short circuit occurs, and the drift of the local distributed inductance and local distributed capacitance reaches more than 5%.

[0057] Under normal conditions where the winding is not loose, the conventional amplitude monitoring method does not perform matrix eigenvalue decomposition, and the processing time for a single data cycle is 6ms. When using the method of this invention, the processing time for a single data cycle is 14ms. After the test platform runs continuously for 72 hours, the system maintains diagnostic data output, and the local display terminal continuously updates the trend curve of the A-phase winding. After the physical simulation circuit switches to the loose state, the fault feature vector is assembled within 1.2s and sent to the downstream actuator by the station-end centralized communication controller. The downstream actuator cuts off the excitation source of the test circuit accordingly, and the system enters the downlink lockout alarm state.

[0058] Example 3: In this example, online diagnosis of a single-phase dual-winding 500kV DC converter transformer was performed at the operation site of a 500kV DC converter station. During operation, the transformer was simultaneously subjected to high-frequency AC harmonics and DC bias. The diagnostic device simultaneously acquired the transient voltage waveform, transient current waveform, and DC bias current measurement value at the port.

[0059] The diagnostic device continuously calculates the time derivative of the transient voltage waveform. When the absolute value of the time derivative is greater than 1.5 kV / μs, the current moment is determined as the trigger moment of the step shock wave. A time-domain sampling window of length 20 μs is extracted from this trigger moment. Within this window, the absolute value of the transient current waveform is integrated over time, and the integration result is divided by the difference between the maximum and minimum values ​​of the transient voltage waveform to obtain the high-frequency capacitor response factor. At the same time, the absolute value of the transient voltage waveform is integrated over time, and the integration result is divided by the difference between the maximum and minimum values ​​of the transient current waveform to obtain the high-frequency inductance derivative factor.

[0060] For each sampling moment within the time-domain sampling window, the diagnostic device calculates the time-domain differential rate of change of the transient current waveform. It multiplies this time-domain differential rate of change with the amplitude of the transient voltage waveform at the same sampling moment to obtain the transient charge product value. Then, it accumulates and integrates according to the sampling sequence to obtain the transient electromagnetic coupling characteristic value. The diagnostic device determines the static bias correction coefficient based on the DC bias current measurement value. Using the static bias correction coefficient and the transient electromagnetic coupling characteristic value, it performs algebraic compensation on the high-frequency inductor derivative factor and the high-frequency capacitor response factor, and calculates the winding inherent impedance scalar.

[0061] The health benchmark value is the historical average value of the inherent impedance pre-stored under the condition that the DC converter transformer is in a fault-free state and the DC bias current measurement value is the same. The diagnostic device calculates the relative deviation between the winding inherent impedance scalar and the health benchmark value. When the relative deviation value is less than 0.02, the output winding structure is intact. When the relative deviation value is in the range of 0.02 to 0.05, the output winding has a slight deformation warning. When the relative deviation value is greater than 0.05, the output winding has structural deformation and inter-turn short circuit fault.

[0062] During the monitoring of transient voltage waveforms, the diagnostic device also continuously calculates the time rate of change of the DC bias current measurement. When the absolute value of the time rate of change is greater than 50A / ms, the measurement protection state is triggered, the update calculation of the winding's inherent impedance scalar is paused, and the current transformer winding fault diagnosis result is maintained. When the absolute value of the time rate of change decays to below 10A / ms and continues to exceed 100ms, the measurement protection state is released, the time domain sampling window is reopened, and the winding's inherent impedance scalar is calculated.

[0063] To verify the diagnostic results of the transient electrical variables at the port, a piezoelectric vibration sensor is installed on the outer wall of the converter transformer tank at a position corresponding to the valve-side winding. The measurement resolution of this sensor is not less than 0.01 m / s². 2 The data sampling frequency is set to 50kHz. The core computing unit of the floating-point digital signal processor periodically reads 2048 floating-point voltage data output by the vibration sensor within the current sampling window through the data bus to form the original time-domain signal sequence.

[0064] The core computing unit performs frequency domain transformation on the original time-domain signal sequence, calculates the proportion of energy in each frequency band of the discrete spectrum to the total energy, and calculates the winding energy distribution entropy using the normalized energy value of each frequency band. For the spectrum array with normalized energy values ​​of 0.40, 0.30, 0.20 and 0.10 for the four frequency bands, after logarithmic transformation and weighted summation, the winding energy distribution entropy is 1.28. This index is used to help characterize the change in winding structure density with the energy distribution of the vibration spectrum.

[0065] The core computing unit compares the winding energy distribution entropy with a dynamic evaluation threshold, which is calculated using the following formula: ,in, To dynamically evaluate the threshold, This is a timeliness correction factor. Using the historical average as the baseline, the core calculation unit determines the aging correction factor based on a table looking up the cumulative operating years of the transformer: when the operating years are less than 2 years, Take 1.00; when the number of years of operation is 2 years or more but less than 5 years, Take 1.05; when the number of years of operation is 5 years or more, A value of 1.10 is used to compensate for the vibration response attenuation caused by thermal aging of the paper insulation.

[0066] When the winding energy distribution entropy is greater than the dynamic evaluation threshold of 1.50 for five consecutive data acquisition cycles, the auxiliary diagnostic channel determines that the valve-side winding has structural loosening and adjusts the fault status level from level 0 to level 1. When the fault status level is level 0, the sampling interval of the signal acquisition device is shortened from 60s to 10s to increase the number of samplings during transient overvoltage impacts. When the fault status level reaches level 1, the system sends a winding abnormality warning signal to the integrated automation system in the main control room and suspends the background spectrum refinement and calibration process.

[0067] In the dual twelve-pulse converter valve group consisting of six single-phase converter transformers, the top-level centralized management unit allocates spectrum analysis and calculation resources based on the DC bias current measurement values ​​collected in real time from each transformer node, and prioritizes processing transformer nodes with DC bias current measurement values ​​greater than 5A.

[0068] Historical mean baseline Adjustments are made according to a 168-hour update cycle. Each update applies a 0.95 attenuation weight to the old historical data to reduce baseline drift caused by long-term thermal aging of the insulating paper. When strong electromagnetic coupling noise in the field causes the signal-to-noise ratio of the vibration sensor to fall below 20dB, the auxiliary diagnostic channel suspends frequency domain Fourier transform and instead uses a time domain sliding window to extract the peak and effective values ​​of the original voltage signal. The ratio of the two is used as the peak factor to maintain the auxiliary monitoring of the vibration channel.

[0069] After receiving the winding abnormality warning signal, the integrated automation system in the main control room cuts off the harmonic abnormality injection source. Subsequently, the winding energy distribution entropy gradually decreases and stabilizes below the baseline of 1.10 in the subsequent monitoring cycle. The vibration amplitude of the transformer tank wall drops to within the safe design boundary. The system returns to normal load operation. The port transient electrical variable diagnostic channel continues to output the transformer winding fault diagnosis result based on the relative deviation between the winding inherent impedance scalar and the health baseline value. The vibration channel is used to assist in verifying the winding structure status.

[0070] Example 4: This example combines Figures 1 to 2 This paper describes a diagnostic method for transformer winding faults, such as... Figure 1 As shown, step S1 acquires transient voltage and current waveforms and measured values; step S2 determines the triggering time when the derivative is greater than 1.5kV / μs; step S3 extracts a 20μs time-domain sampling window; step S4 calculates the high-frequency capacitor response factor and inductance derivative factor; step S5 obtains the transient electromagnetic coupling characteristic value by calculating the time integral; step S6 calculates the winding's inherent impedance scalar using algebraic compensation; and step S7 calculates the deviation value and outputs the winding fault diagnosis result.

[0071] like Figure 2 As shown, the transformer winding fault diagnosis results are connected to waveform and measurement value acquisition, time-domain sampling window truncation, transient characteristic factor extraction, winding inherent impedance calculation, and multi-channel characteristic auxiliary verification. Waveform and measurement value acquisition connects transient voltage waveform, transient current waveform, and DC bias current measurement value. Time-domain sampling window truncation connects to measurement protection status, time-domain sampling window, and step shock wave triggering time. Transient characteristic factor extraction connects to high-frequency capacitor response factor, high-frequency inductance derivative factor, and transient electromagnetic coupling characteristic value. Winding inherent impedance calculation connects to winding inherent impedance scalar, cross-winding high-frequency charge transfer ratio factor, and static bias correction coefficient. Multi-channel characteristic auxiliary verification connects to winding inherent capacitance change, winding energy distribution entropy, and inherent resonant frequency offset.

[0072] Example 5: Under the calibration state of a 500kV DC converter transformer with no load and an ambient temperature of 20℃, an external excitation source injects a high-frequency pulse excitation signal into the high-voltage side winding end. A broadband response sensor located at the winding tap position synchronously collects the response current signal at the winding end. The calculation processor performs a Fourier transform on the response current signal to construct the winding frequency response reference matrix in the initial state.

[0073] The wideband response sensor continuously acquires the response current signal at the end of the winding five times. The processor discards the two acquisition results with the largest and smallest deviations, and takes the arithmetic mean of the remaining three acquisition results to obtain the initial characteristic baseline of the frequency response. A calibration lookup table containing 1024 frequency point characteristic values ​​is established in the designated address space of the system memory to store the initial characteristic baseline of the frequency response.

[0074] The health baseline value is the historical average of the inherent impedance pre-stored under the condition that the DC converter transformer is in a fault-free state and the DC bias current measurement value is the same. During the database construction and update process, the historical average of the inherent impedance is stored in segments according to the DC bias current measurement value from 0A to 100A with a step size of 5A. When the measured DC bias current falls between two adjacent step sizes, the calculation processor uses the linear interpolation method to calculate the corresponding historical average of the inherent impedance and uses a sliding window containing the most recent 10 historical sampling data for arithmetic mean smoothing.

[0075] When the 500kV DC converter transformer is put into actual operation and the grid-side voltage is 500kV, the electrical signal acquisition module acquires the transient voltage waveform and transient current waveform at the port, the current transformer at the neutral point reads the DC bias current measurement value, calculates the time derivative of the transient voltage waveform continuously monitored by the processor, determines the step shock triggering time when the absolute value of the time derivative is greater than 1.5kV / μs, and extracts a time domain sampling window with a length of 20μs from the triggering time.

[0076] Within the time-domain sampling window, the processor performs time integration on the absolute value of the transient current waveform and divides the integration result by the difference between the maximum and minimum values ​​of the transient voltage waveform to obtain the high-frequency capacitor response factor. Simultaneously, the processor performs time integration on the absolute value of the transient voltage waveform and divides the integration result by the difference between the maximum and minimum values ​​of the transient current waveform to obtain the high-frequency inductance derivative factor. For each sampling moment, the processor multiplies the time-domain differential rate of change of the transient current waveform with the amplitude of the transient voltage waveform at the same sampling moment to obtain the transient charge product value. Then, the processor performs time-domain cumulative integration on each transient charge product value to obtain the transient electromagnetic coupling characteristic value.

[0077] The processor determines the static bias correction coefficient based on the measured DC bias current value, and uses the static bias correction coefficient and transient electromagnetic coupling characteristic value to perform algebraic compensation on the high-frequency inductor derivative factor and high-frequency capacitor response factor, calculates the winding inherent impedance scalar, and then reads or interpolates the corresponding health reference value from the segmented stored data based on the measured DC bias current value, and calculates the relative deviation between the winding inherent impedance scalar and the health reference value.

[0078] When the relative deviation is less than 0.02, the output winding structure is intact; when the relative deviation is in the range of 0.02 to 0.05, the output winding has a minor deformation warning; when the relative deviation is greater than 0.05, the output winding has structural deformation and inter-turn short circuit fault.

[0079] The initial characteristic baseline of the frequency response is used to assist in verifying the above diagnostic results. The processor reads the initial characteristic baseline of the frequency response from the calibration lookup table and calculates the absolute deviation between the currently acquired frequency response characteristic value and the initial characteristic baseline of the frequency response at the same frequency point. The calculation formula is as follows: ,in, This is the absolute deviation of the winding frequency response. The frequency response characteristic value currently collected. This is the initial characteristic baseline of the frequency response at the same frequency point.

[0080] The processor calculates the absolute deviation of the winding frequency response. Compared to 3dB, the absolute deviation of the winding frequency response over three consecutive cycles When all values ​​are greater than or equal to 3dB, the system switches the data output port to a high level and drives the external alarm circuit to output an abnormal state indication signal. This abnormal state indication signal is used to assist in verifying the diagnostic results of the inherent impedance scalar of the winding, and does not replace the transformer winding fault diagnosis results output based on the relative deviation value.

[0081] Example 6: Before the intelligent 500kV DC converter transformer is put into operation for the first time, the pulse excitation source built into the transformer online monitoring system injects a standard impulse current signal into the grid-side winding. The high-frequency vibration acceleration sensor deployed on the surface of the DC converter transformer shell collects the transient vibration response signal of the grid-side winding under the excitation of this signal. The data processor performs time-frequency domain conversion on the transient vibration response signal and extracts the intrinsic frequency response feature vector under normal conditions. This feature vector contains 1024 frequency sampling points and is used to construct a baseline reference model for health status.

[0082] After the DC converter transformer enters the actual converter operation, the electrical signal acquisition module acquires the transient voltage waveform and transient current waveform at the port, the current transformer at the neutral point reads the DC bias current measurement value, the data processor continuously calculates the time derivative of the transient voltage waveform, and determines the step shock wave triggering time when the absolute value of the time derivative is greater than 1.5kV / μs, and extracts a time domain sampling window with a length of 20μs from the triggering time.

[0083] Within the time-domain sampling window, the data processor calculates the high-frequency capacitor response factor and the high-frequency inductor derivative factor, and multiplies the time-domain differential rate of change of the transient current waveform at each sampling moment with the corresponding transient voltage waveform amplitude. Then, it performs time-domain accumulation and integration on the obtained transient charge product value to obtain the transient electromagnetic coupling characteristic value. Subsequently, it determines the static bias correction coefficient based on the DC bias current measurement value, and uses the static bias correction coefficient and the transient electromagnetic coupling characteristic value to perform algebraic compensation on the high-frequency inductor derivative factor and the high-frequency capacitor response factor, and calculates the winding inherent impedance scalar.

[0084] The data processor reads the historical average value of the inherent impedance of the DC converter transformer under fault-free conditions and the same DC bias current conditions based on the current DC bias current measurement value, and uses it as the health reference value. The absolute value of the difference between the winding inherent impedance scalar and the health reference value is divided by the health reference value to obtain the relative deviation value. When the relative deviation value is less than 0.02, the output winding structure is intact; when the relative deviation value is in the range of 0.02 to 0.05, the output winding has a slight deformation warning; when the relative deviation value is greater than 0.05, the output winding has structural deformation and inter-turn short circuit fault.

[0085] During the port transient electrical variable diagnosis process, a pulse excitation source injects detection pulses at a fixed period, and a high-frequency vibration acceleration sensor collects measured vibration signals that fluctuate with the load under commutation conditions in real time. The data processor extracts the real-time frequency response feature vector from this signal. And call the health status baseline reference model to calculate and The Euclidean distance between them is used to obtain the deviation judgment value. : ,in, For frequency sampling point index, This represents the total number of frequency sampling points. Take 1024; For the real-time frequency response feature vector at the th The feature values ​​of each frequency sampling point The eigenvalues ​​of the intrinsic frequency response eigenvector at the corresponding frequency sampling point, and the deviation judgment value. Diagnostic results used to assist in verifying the inherent impedance scalar of the winding.

[0086] Safety threshold Based on the pre-calibrated normal fluctuation range in the health status baseline reference model, and to ensure that the auxiliary verification results are output within 3 seconds, the data processor calculates the deviation judgment value. A three-stage filtering sliding window algorithm based on a hardware timer is used to determine the deviation value within three consecutive monitoring periods. All are greater than or equal to the safety threshold. When the condition is met, the data processor drives the alarm relay on the local monitoring interface to close and outputs a maintenance conclusion for abnormal winding deformation on the monitoring interface; if the condition is not met, it outputs a conclusion that the operation is normal.

[0087] Based on the transformer winding fault diagnosis results and deviation judgment values ​​corresponding to the relative deviation values The corresponding auxiliary verification results determine whether the transformer is under maintenance or continues to operate. The auxiliary verification results are used to confirm changes in the winding structure and do not replace the transformer winding fault diagnosis results output based on the relative deviation between the inherent impedance scalar value and the health reference value.

[0088] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A method for diagnosing transformer winding faults, characterized in that, Includes the following steps: Step S1: Obtain the transient voltage waveform and transient current waveform at the port of the 500kV and above ultra-high voltage or extra-high voltage DC converter transformer, and obtain the DC bias current measurement value. Step S2: Monitor the time derivative of the transient voltage waveform. When the absolute value of the time derivative is greater than the preset voltage change rate threshold, determine the trigger time of the step shock wave. The voltage change rate threshold is set in the range of 0.5kV / μs to 5kV / μs. Step S3: Starting from the trigger time, extract a time-domain sampling window with a duration of 20μs; Step S4: Within the time-domain sampling window, calculate the ratio of the time integral value of the transient current waveform to the amplitude change of the transient voltage waveform to obtain the high-frequency capacitor response factor; and calculate the ratio of the time integral value of the transient voltage waveform to the amplitude change of the transient current waveform to obtain the high-frequency inductor derivative factor. Step S5: Within the time-domain sampling window, multiply the time-domain differential rate of change of the transient voltage waveform and the transient current waveform and calculate the time integral to obtain the transient electromagnetic coupling characteristic value. Step S6: Determine the static bias correction coefficient based on the DC bias current measurement value. Use the static bias correction coefficient and the transient electromagnetic coupling characteristic value to perform algebraic compensation on the high-frequency inductor derivative factor and the high-frequency capacitor response factor, and calculate the winding's inherent impedance scalar. Step S7: Calculate the relative deviation between the inherent impedance scalar of the winding and the health reference value, and output the transformer winding fault diagnosis result based on the relative deviation value.

2. The method for diagnosing transformer winding faults according to claim 1, characterized in that, In step S7, the transformer winding fault diagnosis results are output based on the relative deviation value, including: when the relative deviation value is less than 0.02, the winding structure of the 500 kV and above DC converter transformer is intact; when the relative deviation value is in the range of 0.02 to 0.05, the winding of the 500 kV and above DC converter transformer has a minor deformation warning result; when the relative deviation value is greater than 0.05, the winding of the 500 kV and above DC converter transformer has structural deformation and inter-turn short circuit fault results.

3. The method for diagnosing transformer winding faults according to claim 1, characterized in that, In step S4, obtaining the high-frequency capacitor response factor includes: calculating the time integral value of the absolute value of the transient current waveform within the time-domain sampling window, and dividing the time integral value by the difference between the maximum and minimum values ​​of the transient voltage waveform within the time-domain sampling window; obtaining the high-frequency inductor derivative factor includes: calculating the time integral value of the absolute value of the transient voltage waveform within the time-domain sampling window, and dividing the time integral value by the difference between the maximum and minimum values ​​of the transient current waveform within the time-domain sampling window.

4. The method for diagnosing transformer winding faults according to claim 1, characterized in that, In step S5, obtaining the transient electromagnetic coupling characteristic value includes: multiplying the transient voltage waveform amplitude at each sampling moment within the time-domain sampling window by the time-domain differential rate of change at that sampling moment to obtain the transient charge product value at each sampling moment; and performing time-domain accumulation and integration on the transient charge product value to obtain the transient electromagnetic coupling characteristic value.

5. The method for diagnosing transformer winding faults according to claim 1, characterized in that, Step S1, which involves acquiring the transient voltage and current waveforms at the ports of 500 kV and above DC converter transformers, includes: simultaneously acquiring the transient current at the valve-side port and the transient current at the grid-side port of the 500 kV and above DC converter transformer; it also includes the following sub-steps: Step S11, calculating the time integral of the absolute value of the transient current at the valve-side port and the time integral of the absolute value of the transient current at the grid-side port within the time domain sampling window; Step S12, dividing the time integral of the absolute value of the transient current at the valve-side port by the time integral of the absolute value of the transient current at the grid-side port to calculate the cross-winding high-frequency charge transfer ratio factor.

6. The method for diagnosing transformer winding faults according to claim 5, characterized in that, After step S6 and before step S7, the method further includes a step of correcting the winding intrinsic impedance scalar using the cross-winding high-frequency charge transfer ratio factor, including the following sub-steps: Step S61, the cross-winding high-frequency charge transfer ratio factor is used as a dimensionless gain coefficient and multiplied with the winding intrinsic impedance scalar to generate the corrected winding intrinsic impedance scalar. In step S62, the corrected winding inherent impedance scalar is used to replace the original winding inherent impedance scalar and input into step S7.

7. The method for diagnosing transformer winding faults according to claim 1, characterized in that, In step S1, obtaining the DC bias current measurement value includes reading real-time DC bias current data using a current transformer deployed at the neutral point of a 500 kV or higher DC converter transformer; the health reference value includes the historical average of the inherent impedance stored in advance under the same DC bias current measurement conditions when the 500 kV or higher DC converter transformer is in a fault-free state.

8. The method for diagnosing transformer winding faults according to claim 1, characterized in that, While monitoring the time derivative of the transient voltage waveform in step S2, the following sub-steps are also included: Step S21, continuously calculate the time rate of change of the DC bias current measurement value; Step S22, compare the absolute value of the time rate of change with the set transient impact threshold of 50A / ms; Step S23, when the absolute value of the time rate of change is greater than 50A / ms, trigger the measurement protection state, suspend the update calculation of the winding inherent impedance scalar, and lock the current transformer winding fault diagnosis result output.

9. A method for diagnosing transformer winding faults according to claim 8, characterized in that, After triggering the measurement protection state in step S23, the following sub-steps are also included: Step S231, continuously monitor the absolute value of the time change rate of the DC bias current measurement value; Step S232, when the absolute value of the time change rate decays to below 10A / ms and the duration exceeds 100ms, release the measurement protection state, and recalculate the winding intrinsic impedance scalar after reopening the time domain sampling window.

Images