Transient current waveform capturing method for transformer short circuit withstand capability test

Abstract

The present application belongs to the technical field of transformer, and particularly relates to a transient current waveform capturing method for transformer short circuit withstand capability test, comprising: obtaining the transient current derivative of each sampling point of the main circuit; establishing a macroscopic reference derivative analytical model and approximating fitting with the transient current derivative, stripping the power frequency alternating current and transient direct current background components, and obtaining the initial wideband residual; calculating the stress mapping value according to the transient current and the current absolute peak value, and dividing the rising stage and the decay stage based on the first-order difference; respectively extracting the residual feature strength of the two stages to calculate the elasticity factor; combining the elasticity factor, the initial wideband residual and the stress mapping value to calculate the health early warning index, so as to alarm the transformer winding mechanical looseness. The present application processes data in the derivative domain, avoids weakening the high-frequency characteristics by integral reduction, weakens the external test bus lag interference by using the elasticity factor, and realizes the early warning of the early hidden trouble of winding looseness under complex working conditions.

Description

Technical Field

[0001] This invention relates to the field of transformer technology. More specifically, this invention relates to a method for capturing transient current waveforms for transformer short-circuit withstand capability tests. Background Technology

[0002] In the engineering evaluation of short-circuit withstand capability tests of power transformers, real-time monitoring of the mechanical integrity of the windings is a core component of the fault prediction and health management system. At the moment of a short-circuit test, the transformer windings experience an inrush current tens of times greater than the rated value, generating enormous axial and radial electrodynamic forces within the windings. If the transformer has inherent problems such as insufficient clamping force or loose insulation pads during manufacturing or long-term operation, these electrodynamic forces will directly induce mechanical displacement of the windings. This displacement causes slight changes in local leakage inductance, which in turn creates a high-frequency, weak-amplitude modulation effect on the main circuit current. Therefore, accurately capturing and identifying these early transient signals is fundamental to assessing the degradation trend of the transformer's short-circuit withstand capability and formulating health management strategies.

[0003] Chinese patent document CN102323503B, authorized by patent number CN102323503B, provides a method for detecting transformer inrush current distortion based on Rogowski coils. This method uses Rogowski coils to acquire differential voltage signals reflecting the rate of change of current, performs Fourier transform on the signal to calculate the effective values ​​of each harmonic, and distinguishes between inrush current and internal fault current using a distortion degree criterion. However, this patent document uses Fourier transform to process transient signals, and its algorithm relies on the assumptions of signal periodicity and stationarity. Short-circuit test current is a typical non-stationary decay process, and the transform processing leads to ambiguity in the time-domain abrupt change characteristics. Furthermore, this scheme only calculates a single distortion degree index based on the total harmonics, failing to establish a correspondence between the rate of change of current and the mechanical stress state of the windings, and cannot eliminate background interference caused by the mechanical oscillation of the external busbar at the test station, making it difficult to meet the requirements for accurate identification of early mechanical hazards.

[0004] Chinese patent application CN107478949A discloses an online diagnostic method and system for transformer winding deformation. This method injects a swept-frequency signal via a Rogowski coil using magnetic field induction, collects the excitation and response currents, constructs a transfer function to obtain the frequency response curve, and compares the curves against industry standards to determine the winding deformation. However, this patent application primarily relies on external signal injection and frequency domain curve comparison, focusing on extracting static features after changes in winding distribution parameters. During the strong impact transient process of a short-circuit test, it cannot directly utilize the impact current waveform generated by the test itself. Furthermore, frequency response curve comparison requires a long data window and a stable operating environment, making it difficult to capture the rapid transient high-frequency characteristics caused by mechanical loosening during a short circuit. Therefore, its applicability in dynamic fault prediction scenarios is limited.

[0005] In existing technologies, although some solutions attempt to assess transformer condition by calculating transient harmonic distortion or injecting sweep signals for frequency response comparison, these solutions have significant limitations in capturing and extracting features from transient currents during short-circuit tests. Conventional Rogowski coil measurement schemes typically require a hardware integrator or digital integration operation at the back end to obtain a clear current amplitude waveform. From a signal processing perspective, integration significantly attenuates high-frequency components, weakening the high-frequency micro-vibration characteristics that truly indicate winding loosening during waveform reconstruction, thus reducing the sensitivity of the detection system to early-stage problems. Furthermore, large-capacity short-circuit tests involve complex mechanical interference. At the moment of impact, the external test busbar, carrying a huge current, experiences violent mechanical oscillations, and these oscillations exhibit significant inertial hysteresis. Existing analysis methods struggle to distinguish between the synchronous micro-vibrations of the winding itself and the hysteretic oscillations of the external busbar at the feature level, easily misinterpreting environmental background interference as internal mechanical faults. This leads to discrepancies between pre-diagnosis and health management results, limiting the practical application of short-circuit test waveform capture technology in transformer health management engineering. Summary of the Invention

[0006] To address the technical problems of existing solutions where integral restoration weakens high-frequency micro-vibration characteristics and makes it difficult to distinguish between winding micro-vibrations and external interference, leading to inaccurate early hazard identification, this invention provides a transient current waveform capture method for transformer short-circuit withstand capability testing, comprising: S1: Obtain the transient current derivatives at each sampling point in the main circuit of the transformer short-circuit withstand capability test; S2: Based on the differentiation of the classical transient equation of the short-circuit circuit, establish a macroscopic benchmark derivative analytical model in the derivative domain; S3: Approximate the transient current derivatives with the time-discrete macroscopic benchmark derivative analytical model to remove the background components of power frequency AC and transient DC, and obtain the initial broadband residual characterizing the mechanical vibration of the transformer winding; S4: Calculate the stress mapping value based on the transient current and absolute peak current obtained from each sampling point, and divide the rising and decaying stages of the electromagnetic mechanical force based on the first-order difference result of the stress mapping value; S5: Extract the residual characteristic intensity in the rising and decaying stages respectively to calculate the elasticity factor, and calculate the health warning index by combining the elasticity factor, the initial broadband residual, and the stress mapping value; S6: In response to the health warning index being greater than the preset benchmark warning index threshold, output an alarm signal characterizing the mechanical loosening of the transformer winding.

[0007] This invention avoids the attenuation effect of traditional integral restoration on high-frequency components by directly obtaining the transient current derivative in the derivative domain and fitting and comparing it with the macroscopic benchmark derivative analytical model, thus fully preserving the high-frequency micro-vibration characteristics carrying the precursors of winding loosening. Simultaneously, based on the dynamic evolution law of electromagnetic mechanical forces, it divides the process into rising and declining stages and extracts the corresponding residual characteristic intensities. The difference in residual characteristic intensities between the two stages effectively distinguishes between internal winding synchronous oscillations and the hysteretic interference of external test busbar swaying. Finally, the health warning index obtained by combining the elasticity factor, initial broadband residual, and stress mapping value can accurately distinguish between internal real damage and external background disturbances, reducing the false alarm rate under complex test conditions and improving the accuracy of early detection of winding mechanical loosening hazards. This provides a reliable basis for the pre-diagnosis and health management of transformer short-circuit withstand capability.

[0008] Preferably, the step of obtaining the transient current derivative of each sampling point in the main circuit of the transformer short-circuit withstand capability test includes: acquiring the original induced voltage of each sampling point through a broadband Rogowski coil set in the main circuit; dividing the original induced voltage of each sampling point by the inherent mutual inductance coefficient of the broadband Rogowski coil to obtain the transient current derivative of the corresponding sampling point.

[0009] Preferably, the macroscopic benchmark derivative analytical model satisfies the expression: In the formula, An analytical model representing the macroscopic baseline derivative that varies with time; Indicates the peak value of the power frequency current; Indicates the power frequency angular frequency; Indicates time; Indicates the initial phase angle at closing; Indicates the impedance angle of the main circuit; Indicates the initial amplitude of the DC component; The base of the natural logarithm; Indicates the system's equivalent resistance; Indicates the system's equivalent inductance; This represents the cosine function.

[0010] This invention constructs a macroscopic benchmark derivative analytical model that includes AC rate of change components and DC attenuation rate of change components. In the derivative domain, it reconstructs the theoretical electrical evolution trajectory of a short-circuit loop under healthy conditions. This model strictly follows the physical laws of transient processes, providing a highly matched theoretical benchmark for the alignment of subsequent measured signals. It ensures the accuracy of background electrical component stripping and avoids the erroneous retention of abnormal features or the erroneous deletion of normal fluctuations caused by benchmark trajectory deviation.

[0011] Preferably, the step of approximating the transient current derivative with the analytical model of the macroscopic reference derivative after time discretization includes: extracting the transient current derivative after the short circuit occurs to form a measured sequence of transient current derivatives; using the analytical model of the macroscopic reference derivative after replacing the continuous time variable with discrete time sampling points as the objective function, employing a nonlinear least squares fitting algorithm to perform multi-parameter regression optimization on the measured sequence of transient current derivatives, and iteratively solving for the optimal solution set of parameters such as the system equivalent inductance, system equivalent resistance, initial phase angle of closing, peak power frequency current, and initial amplitude of the DC component.

[0012] Preferably, obtaining the initial broadband residual characterizing the mechanical vibration of the transformer winding includes: substituting the optimal solution set of parameters into the analytical model of the macroscopic reference derivative after time discretization to obtain the macroscopic reference derivative at each discrete time point; and subtracting the transient current derivative at each sampling point from the macroscopic reference derivative at the corresponding discrete time point to obtain the initial broadband residual.

[0013] This invention obtains the macroscopic reference derivative at each time point by substituting the optimal solution set of fitted parameters into the discretized theoretical model, and subtracts it point by point from the measured transient current derivative. This process removes the inherent power frequency fluctuations and attenuated DC background in the main circuit, and directly highlights the high-frequency micro-fluctuations that deviate from the theoretical trajectory in the derivative domain. This allows the initial broadband residual characterizing the winding mechanical displacement to be completely preserved, thereby improving the signal-to-noise ratio of subsequent feature extraction.

[0014] Preferably, the step of calculating the stress mapping value based on the transient current and absolute peak current of each sampling point obtained by fitting includes: substituting the optimal solution set of parameters into the classical transient equation to generate the transient current corresponding to the discrete time point; extracting the maximum value among the absolute values ​​of the transient currents corresponding to all discrete time points as the absolute peak current; and taking the square of the ratio between the transient current and the absolute peak current as the stress mapping value.

[0015] Preferably, the division of the electromagnetic mechanical force into a rising stage and a declining stage based on the first-order difference result of the stress mapping value includes: performing a first-order backward difference operation on the stress mapping value to obtain a first-order difference value; obtaining the sampling point number corresponding to each discrete time point where the first-order difference value is greater than or equal to 0, forming a first set to characterize the rising stage; and obtaining the sampling point number corresponding to each discrete time point where the first-order difference value is less than 0, forming a second set to characterize the declining stage.

[0016] This invention performs a first-order backward difference operation on the stress mapping value and strictly divides the sampling time point into the stress intensification interval and the stress reduction interval according to the positive and negative signs of the difference result. It accurately captures the dynamic evolution trajectory of electromagnetic mechanical force during short-circuit impact, provides a clear time boundary for subsequent evaluation of the mechanical response characteristics of different stress stages, and effectively supports the identification of the difference between internal synchronous oscillation and external inertial hysteresis.

[0017] Preferably, the extraction of residual feature intensity during the ascent and decline phases includes: extracting the maximum value among the absolute values ​​of the macroscopic baseline derivatives corresponding to all discrete time points as the absolute peak value of the derivative; and calculating the residual feature intensity during the ascent phase. ; Calculate the residual characteristic strength during the decline period: In the formula, Indicates the intensity of residual characteristics during the climbing phase; Indicates the intensity of residual characteristics during the decline period; Indicates the first Initial broadband residuals at each sampling point; Indicates the absolute peak value of the derivative; Represents the first set; This represents the second set.

[0018] This invention uses the limiting rate of change of the theoretical benchmark derivative as a normalized scale to transform the alternating positive and negative residual fluctuations in each stage into a constant positive intensity scalar and accumulate them in segments. This eliminates the numerical cancellation phenomenon of high-frequency signals on the time axis and truly records the intensity of mechanical vibration accumulated during the electromagnetic force rise and decay periods, providing a reliable data basis for subsequent comparison of the response differences between the two stages.

[0019] Preferably, the elasticity factor satisfies the expression: if ,but ;like Then set ;in, Indicates the elasticity factor; Indicates the intensity of residual characteristics during the decline period; This indicates the intensity of the residual characteristics during the climbing period.

[0020] This invention measures the temporal symmetry of mechanical response during force evolution by calculating the relative deviation of residual characteristic intensity between the decay and rise phases and converting it into an elasticity factor. The swaying of the external test busbar exhibits significant hysteresis asymmetry due to mechanical inertia, while the synchronous oscillation caused by loosening of the internal winding tends to be symmetrically distributed. The elasticity factor utilizes this physical difference to automatically weaken the weight of external hysteresis interference and highlight the true internal damage characteristics.

[0021] Preferably, the calculation of the health early warning index includes: calculating the amplitude of the basic damage characteristic: In the formula, Indicates the amplitude of the basic damage characteristics; This represents the total number of samples in the measured sequence of transient current derivatives; Indicates the first Initial broadband residuals at each sampling point; Indicates the absolute peak value of the derivative; This represents the stress mapping value; the product of the elasticity factor and the amplitude of the basic damage characteristic is used as the health warning index.

[0022] This invention converts high-frequency alternating oscillations into a positive scalar by calculating the square of the ratio of the initial broadband residual to the absolute peak value of the derivative, thus avoiding numerical cancellation during signal accumulation. It then multiplies this value by the stress mapping value and performs a square root operation to obtain the amplitude of the basic damage characteristics. This amplifies the true structural damage characteristics at the extreme stress moment while transforming it into a stable linear physical scale. Finally, it combines the elasticity factor to obtain a health warning index, effectively eliminating asymmetric hysteresis interference caused by external test busbar swaying. This ensures that the health warning index can purely and accurately reflect the true physical damage state inside the transformer, improving the reliability of the warning judgment.

[0023] The beneficial effects of this invention are as follows: By directly obtaining the transient current derivative in the derivative domain and approximating it with the macroscopic benchmark derivative analytical model, this invention avoids the attenuation effect of conventional integral restoration on high-frequency components, thus fully preserving the mechanical vibration characteristics carrying the precursors of winding loosening; by using differential operations to strip away the power frequency AC and transient DC background components to obtain the initial broadband residual, and dividing the rise and fall stages of electromagnetic mechanical force according to the first-order difference results of stress mapping values; by extracting the residual characteristic intensity in the two stages to calculate the elasticity factor, the difference between the inertial hysteresis of internal winding synchronous oscillation and external test busbar sway is effectively distinguished and weakened; finally, by combining the elasticity factor, the initial broadband residual, and the stress mapping value to calculate the health warning index, the real mechanical damage characteristics at high stress moments are reasonably amplified, and the random interference during low stress periods is naturally suppressed. The output health warning index can stably exceed the preset benchmark warning index threshold to trigger an alarm, improving the accuracy of early identification of winding mechanical loosening under short-circuit test conditions. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating the transient current waveform capture method for transformer short-circuit withstand capability testing in this invention. Figure 2 A schematic diagram comparing the transient current derivative sequence with the macroscopic benchmark derivative analytical model; Figure 3 A schematic diagram showing the initial broadband residual waveform and the stages of force evolution; Figure 4 This is a schematic diagram showing the evolution trajectory of transient current and the stress mapping value. Figure 5 This is a schematic diagram comparing the distribution of residual characteristic intensity during the rise and the residual characteristic intensity during the decline phase. Figure 6 This is a schematic diagram illustrating the final determination results of the health warning index and the benchmark warning index threshold. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0027] This invention discloses a transient current waveform capture method for transformer short-circuit withstand capability testing, referring to... Figure 1 This includes steps S1 to S6: S1: Obtain the transient current derivative at each sampling point.

[0028] It should be noted that conventional measurement methods in the main circuit of transformer short-circuit withstand capability tests typically include hardware integration, which causes high-frequency characteristics resulting from mechanical deformation of the windings to be suppressed by low-pass filtering during signal reconstruction. To overcome the smoothing effect of traditional integrator circuits on high-frequency components, this invention fully utilizes the inherent differential response characteristics of the Rogowski coil, avoiding integration reconstruction and directly retaining the acquired signal in the derivative domain. Since the gain of differential operations on high-frequency components is significantly higher than that on low-frequency components, this method can effectively highlight the high-frequency mechanical vibration characteristics masked by conventional methods in the numerical dimension, providing a high signal-to-noise ratio foundational sequence for subsequent defect identification.

[0029] Specifically, a broadband Rogowski coil is installed in the main circuit of the transformer short-circuit withstand capability test. The output terminal of the broadband Rogowski coil is connected to a high-speed, high-resolution data acquisition card, and the original induced voltage is acquired through the high-speed, high-resolution data acquisition card.

[0030] Based on Faraday's law of electromagnetic induction and the working principle of the Rogowski coil, the output voltage of the coil and the rate of change of the primary current satisfy a strictly linear proportional relationship, the existing continuous domain formula of which is: ,in Representing continuous time The instantaneous induced voltage output by the lower Rogowski coil; This represents the inherent mutual inductance coefficient of a broadband Rogowski coil; This represents the transient current flowing through the primary side of the transformer's main circuit; Indicates transient current A tiny change within an infinitesimal time interval; This represents the corresponding time element. In the mathematical description of the continuous domain, the rate of change of current... This is strictly equivalent to the first derivative of the transient current with respect to time. Discretizing the continuous time according to the data acquisition frequency corresponds to the... The voltage sample value at each sampling time is denoted as The transient current derivative at that moment, i.e., the rate of change of current in the discrete sequence, is denoted as... Substituting into the continuity relation, we get Due to mutual inductance coefficient To determine the non-zero constants specified in the factory multi-frequency calibration, divide both sides of the equation by... By rearranging the terms algebraically, we can find that the derivative of the transient current satisfies the following expression: ; In the formula, Indicates the first Transient current derivative at each sampling point; Indicates the first The original induced voltage at each sampling point; This represents the inherent mutual inductance of a broadband Rogowski coil, obtained through factory multi-frequency calibration using a standard high-frequency current source. When the original induced voltage When the amplitude of the mutual inductance coefficient increases due to high-frequency oscillations, the mutual inductance coefficient... The transient current derivative obtained while keeping it constant. By scaling up proportionally, this invention enhances the numerical dimension of the masked high-frequency microseismic components by directly utilizing division scaling transformation instead of conventional hardware integration operations, thus ensuring the signal-to-noise ratio of data for subsequent feature analysis and fault prediction.

[0031] S2: Establish a macroscopic benchmark derivative analytical model.

[0032] Specifically, according to the transient analysis theory of power systems, the current in a short-circuit loop containing resistance and inductance during the closing transient is... It satisfies the following publicly available classical transient equations: ; In the formula, Indicates time The corresponding transient current; Indicates the peak value of the power frequency current; Represents the power frequency angular frequency, satisfying , Represents pi (π). This indicates the rated operating frequency of the AC power grid to which the transformer is connected; Indicates time; Indicates the initial phase angle at closing; Indicates the initial amplitude of the DC component; The base of the natural logarithm; Indicates the system's equivalent resistance; Indicates the system's equivalent inductance; Represents the sine function; This represents the impedance angle of the main circuit, satisfying... , This represents the arctangent function. The classical transient equations describe the evolution of macroscopic electrical signals under healthy and lossless conditions of a transformer. In the classical transient equations, the transient current... From AC steady-state components and DC attenuation component Superimposed structure, due to the system's equivalent resistance and system equivalent inductance All were positive values, and as time went on... The increase of corresponds to the negative exponent term. Gradual decay leads to DC component Gradually decays, eventually becoming a transient current. It approaches a pure sinusoidal alternating state.

[0033] To establish a benchmark in the derivative domain, the above classical transient equations are analyzed in time to obtain the macroscopic benchmark derivative analytical model: ; In the formula, An analytical model representing the macroscopic baseline derivative that varies with time; Indicates the peak value of the power frequency current; Indicates the power frequency angular frequency; Indicates time; Indicates the initial phase angle at closing; Indicates the impedance angle of the main circuit; Indicates the initial amplitude of the DC component; The base of the natural logarithm; Indicates the system's equivalent resistance; Indicates the system's equivalent inductance; This represents the cosine function. The macroscopic baseline derivative consists of the AC rate of change component. and DC attenuation rate component Composition, over time Add DC attenuation rate component Rapid decay, the mathematical operation mechanism of analytical differentiation makes the alternating rate of change component... The amplitude increased the power frequency angular frequency The multiplier term significantly enhances the periodic fluctuation characteristics of AC signals in the derivative domain. The macroscopic benchmark derivative analytical model establishes a theoretical comparison benchmark in the derivative domain, providing mathematical support for the subsequent differential stripping of abnormal waveforms.

[0034] S3: Fit the macroscopic benchmark derivative to obtain the initial broadband residual.

[0035] It should be noted that, since the actual transient current derivatives collected contain macroscopic power frequency AC and transient DC background components, it is impossible to directly extract the weak mechanical vibration characteristics from the absolute value of the transient current derivatives. Therefore, this invention uses the Levenberg-Marquardt (LM) nonlinear least squares fitting algorithm to approximate and align the transient current derivatives with the macroscopic benchmark derivative analytical model, and uses a subtraction operation to remove the aforementioned power frequency AC and transient DC background components, thereby separating the initial broadband residuals that characterize the mechanical vibration and structural loosening of the transformer windings.

[0036] Specifically, starting from the initial moment of the short circuit in the transformer short circuit withstand capability test, the transient current derivatives of the first three cycles are extracted to form a measured sequence of transient current derivatives. Here, a cycle represents the process of the fundamental AC current of the power grid completing one complete sinusoidal periodic change, and its time span is obtained by calculating the reciprocal of the rated operating frequency. The first three cycles cover the core time window in which the electrodynamic impact and transient decay are most severe after the short circuit occurs.

[0037] Using the existing LM nonlinear least squares fitting algorithm, and with the time-discrete macroscopic benchmark derivative analytical model as the objective function, multi-parameter regression optimization is performed on the measured transient current derivative sequence. The time-discrete macroscopic benchmark derivative analytical model refers to replacing the continuous time variables in the macroscopic benchmark derivative analytical model with discrete time sampling points that match the actual hardware data acquisition frequency, thereby converting the continuous mathematical analytical expression into a discrete digital sequence to strictly match the data length and structure of the measured transient current derivative sequence.

[0038] During the iterative approximation process, the system's equivalent inductance is... System equivalent resistance Initial phase angle of closing Peak power frequency current and the initial amplitude of the DC component They are all treated as free undetermined parameters and iteratively solved to obtain the optimal set of parameter solutions. Then, they are input into the analytical model of the macroscopic benchmark derivative after time discretization to generate the macroscopic benchmark derivative corresponding to each discrete time point.

[0039] For example, Figure 2 This paper presents a comparison in the time domain between the transient current derivative sequence captured by a broadband Rogowski coil and the macroscopic benchmark derivative analytical model constructed through parameter fitting optimization in a transformer short-circuit withstand capability test. Figure 2As can be seen, the macroscopic benchmark derivative analytical model generated based on the optimal solution set of parameters encapsulates the real macroscopic background components of the system, such as the periodic fluctuations of power frequency AC and the evolution of transient DC attenuation in the measured signal. Meanwhile, the transient current derivative sequence clearly superimposes the high-frequency violent oscillation characteristics caused by the short-circuit impact on the winding structure on its macroscopic envelope contour.

[0040] Furthermore, the initial broadband residual is obtained based on the transient current derivative at each sampling point and the corresponding macroscopic reference derivative at discrete time points: ; In the formula, Indicates the first Initial broadband residuals at each sampling point; Indicates the first Transient current derivative at each sampling point; Representing discrete time points Corresponding macroscopic baseline derivative. Initial broadband residual. Derivative of the measured transient current Subtract the macroscopic baseline derivative of the theoretical fit The initial broadband residual obtained by subtracting the measured signal from the theoretical healthy reference when the measured signal deviates from the reference due to winding loosening is obtained. The absolute value increases accordingly. This subtraction operation removes the macroscopic power frequency AC and transient DC background components of the main circuit, while retaining the abnormal fluctuation characteristics related to mechanical micro-vibrations.

[0041] S4: Calculate stress mapping values ​​to divide the rising and falling phases.

[0042] It should be noted that since the initial broadband residual includes not only the loosening characteristics of the transformer's internal windings, but also the hysteresis interference component generated by the swaying of the external test busbar at the test station, directly using the amplitude of the initial broadband residual for judgment is prone to serious false alarms. Therefore, this invention extracts the intensity of abnormal features at different stages of change based on the changing law of the magnitude of the electromagnetic physical force exerted on the transformer to evaluate the transformer's health status.

[0043] Specifically, the optimal set of parameters obtained in step S3 is substituted into the classical transient equation to generate the transient current corresponding to each discrete time point. The absolute value of the transient current corresponding to each discrete time point is calculated to form a sequence of absolute values ​​of transient current. All values ​​in the sequence of absolute values ​​of transient current are compared and traversed to select the maximum value as the absolute peak value of the current.

[0044] According to Ampere's law of motion, the electromagnetic mechanical force experienced by the transformer windings under short-circuit conditions... It is proportional to the square of the transient current flowing through the main circuit, and its existing physical relationship is as follows: ,in, The structure constants for winding geometry and leakage flux distribution are: Representing discrete time points The corresponding transient current. To eliminate the influence of short-circuit current amplitude fluctuations on the performance indicators and to construct dimensionless characteristics, peak electromagnetic force is used. As a normalization benchmark, the structure constant is eliminated by dividing the two equations. The stress mapping value is derived as follows: ; In the formula, Indicates the first Stress mapping values ​​at each sampling point; Representing discrete time points The corresponding transient current; This indicates the absolute peak value of the current. This invention is based on transient current. and the absolute peak current This maps the changing patterns of electrical signals to the distribution patterns of electromagnetic mechanical force intensity at the mechanical level. When the transient current... When the amplitude increases, the stress mapping value The rapid quadratic increase objectively reflects the underlying principle that the intensity of electromagnetic mechanical force is proportional to the square of the instantaneous current.

[0045] Furthermore, a first-order backward difference operation is performed on the stress mapping values ​​to obtain the first-order difference values: ,in, Indicates the first The first-order difference value of each sampling point Indicates the first Stress mapping values ​​at each sampling point Indicates the first The stress mapping values ​​at each sampling point. It is important to note that, to avoid array out-of-bounds errors or reading invalid, dirty memory data at the discretized sequence starting point due to tracing back to historical states, and to ensure that the system origin strictly conforms to the true physical initial state where the electromagnetic mechanical force inside the transformer is objectively zero before the short-circuit impact, initial cold-start parameters are set before calculating the first-order difference values. .

[0046] Obtain the first-order difference value The sampling point indices corresponding to each discrete time point that are greater than or equal to 0 constitute the first set. Obtain the first-order difference value The sampling point indices corresponding to each discrete time point less than 0 constitute the second set. When the first difference value When the value is greater than or equal to 0, it indicates that the electromagnetic mechanical force borne by the transformer winding is in a state of gradual increase or intensification, or is in a critical state of extreme force; when the first-order difference value When the value is less than 0, it indicates that the electromagnetic mechanical force borne by the transformer winding is in a state of gradual decay. This invention rigorously divides the different physical evolution stages of the electromagnetic mechanical force's rise and fall through time-difference.

[0047] For example, Figure 3 The diagram shows the initial broadband residual waveform obtained by subtracting the macroscopic reference derivative from the measured sequence of transient current derivatives, and the time intervals within which the electromagnetic mechanical force rise and decay periods are strictly defined based on the first-order difference value. Figure 3 As can be seen, after the subtraction stripping operation, the macroscopic power frequency AC and transient DC bias background in the main circuit are eliminated. The extracted initial broadband residual waveform exhibits the characteristic of high-frequency violent oscillation around the absolute zero axis. The waveform evolution effect proves that the differential technology of this invention eliminates conventional electrical interference and successfully extracts a pure high-frequency micro-vibration waveform, which can accurately characterize the real mechanical displacement response caused by the micro-loosening of the transformer's internal structure or the swaying of the external test busbar.

[0048] Figure 4 This illustrates the dynamic synchronous correspondence between the transient current evolution trajectory and the stress mapping value distribution. From... Figure 4 As can be seen, after short-circuit closing, the transient current exhibits a unique alternating transient process of asymmetric decay to symmetric steady state. The corresponding stress mapping value is distributed in the range of values ​​greater than or equal to 0. Its pulsation frequency is objectively twice the alternating frequency of the transient current, and it accurately anchors and amplifies the position of each absolute peak and absolute trough of the transient current. This evolution effect intuitively restores the objective underlying law in the physical law of Ampere's force that the intensity of the electromagnetic mechanical force is proportional to the square of the transient current in the main circuit. This proves that the stress mapping value constructed by this invention can accurately identify the moment of extreme stress in the test process.

[0049] S5: Calculate the elasticity factor and health early warning index.

[0050] The absolute values ​​of the macroscopic baseline derivatives at each discrete time point are calculated to form a sequence of absolute values ​​of the baseline derivatives. All values ​​in the sequence are compared and traversed to select the maximum value as the absolute peak value of the derivative.

[0051] According to the discrete signal energy theorem, the physical strength of a discrete sequence is equal to the sum of the squares of the amplitudes at each sampling point, and its current formula is: ,in, This represents the total energy scalar of a discrete signal. Represents discrete-time index The signal amplitude sampling points at these locations. To unify the units and eliminate absolute numerical differences, the initial broadband residual is... Divided by the absolute peak value of the derivative After normalization, the result is substituted into the basic formula for discrete signal energy. Based on the set of electromagnetic force evolution stages defined in step S4, the summation domain is restricted, and the expression satisfying the residual characteristic intensity of the rising phase and the residual characteristic intensity of the declining phase is derived: ; ; In the formula, Indicates the intensity of residual characteristics during the climbing phase; Indicates the intensity of residual characteristics during the decline period; Indicates the first Initial broadband residuals at each sampling point; Indicates the absolute peak value of the derivative; Represents the first set; Represents the second set. Intensity of residual characteristics during the ascent period. Strength of residual characteristics during the decay period These correspond to the cumulative residual characteristic intensity during the rising and falling phases of the electromagnetic mechanical force, respectively. The absolute peak value of the derivative represents the extreme value of the highest rate of change in the macroscopic benchmark evolution process, locking in the ultimate benchmark physical scale throughout the transient process. This scale is used to normalize the initial broadband residual, thereby converting the absolute deviation into a relative distortion ratio. Due to the initial broadband residual... It exhibits a high-frequency alternating positive and negative fluctuation characteristic over time. Directly algebraically summing the amplitudes would cause the positive and negative values ​​to cancel each other out on the time axis, making it impossible to accurately measure the severity of the mechanical vibration. Therefore, this invention, based on the discrete signal energy theorem, transforms the alternating instantaneous amplitude into a scalar that is always positive through squaring, so that mechanical displacement fluctuations in all directions can be converted into accumulative physical work; when the initial broadband residual... When the amplitude of the fluctuation increases, the characteristic intensity of the residual during the rising period is obtained due to the squaring of the numerical values ​​and the summation over the time series. or the intensity of residual characteristics during the decline period All of these will increase rapidly. This invention provides physical comparison benchmark data for subsequent elimination of mechanical hysteresis by extracting the residual fluctuation characteristic intensity at different stages of force evolution.

[0052] For example, Figure 5 This diagram illustrates a comparison of the residual characteristic intensity distributions during the ascent and decline phases. From... Figure 5As can be seen, the value of the residual characteristic intensity during the decay period is significantly greater than that during the rise period. The two exhibit a strong asymmetry in physical response on a numerical scale. The difference in the distribution of the residual characteristic intensity during the rise period and the residual characteristic intensity during the decay period reflects the severe operating condition characteristics of the giant test busbar connected externally, which begins to generate severe residual mechanical vibration due to the large mechanical physical hysteresis inertia when the internal electromagnetic mechanical force is rapidly unloaded and decayed in the transformer short-circuit withstand capacity test environment. This confirms the necessity of extracting the residual characteristic intensity during the rise period and the residual characteristic intensity during the decay period to identify external mechanical hysteresis interference.

[0053] Furthermore, the elasticity factor is calculated based on the residual characteristic strength during the ascent and the residual characteristic strength during the decline phases: ; In the formula, Indicates the elasticity factor; Indicates the intensity of residual characteristics during the decline period; This represents the residual characteristic strength during the ascent phase. The elasticity factor measures the symmetry of the characteristic strength response by using the relative deviation between the residual characteristic strength during the decline phase and the residual characteristic strength during the ascent phase. Due to the significant mechanical and physical hysteresis in the external test busbar motion, the residual mechanical vibration increases substantially during the decline phase of the electromagnetic mechanical force, thus affecting the residual characteristic strength during the decline phase. Significantly greater than the residual characteristic strength during the climbing period When the absolute value of the difference between the residual characteristic intensity during the decline period and the residual characteristic intensity during the rise period increases, The overall increase makes the elasticity factor Decrease; if synchronous oscillation occurs in the internal winding structure, the residual characteristic intensity during the decline period tends to be consistent with the residual characteristic intensity during the rise period, and the elasticity factor... It approaches 1. It should be noted that when the residual characteristic strength during the climbing period... Strength of residual characteristics during the decay period When both are 0, it indicates that the measured signal and the theoretical analytical model completely coincide and the transformer is in an absolutely ideal healthy state. To prevent calculation anomalies where the denominator is 0, the elasticity factor is directly set to 1.

[0054] Furthermore, based on the structural dynamics vibration energy theorem and the fatigue damage accumulation principle of variable amplitude loads, the microscopic damage intensity of transformer windings under dynamic electromagnetic excitation is not a linear superposition of single signal amplitudes, but rather determined by the coupling of instantaneous electromagnetic driving force intensity and mechanical vibration response energy. To establish an objective damage assessment benchmark, two existing physical fundamental analytical formulas are first introduced: First, based on the discrete vibration energy theorem, the mechanical vibration energy scalar at the sampling point is proportional to the square of the normalized vibration response amplitude, and its existing basic formula is: In the formula, Indicates the first Mechanical vibration energy at each sampling point; This represents a proportionality constant related to the winding structure stiffness and mass distribution. This represents the normalized vibration response amplitude. Secondly, based on the stress-energy coupled damage model, the instantaneous damage intensity of a structure under transient impact is product of the instantaneous normalized driving force intensity and the vibration response energy. The existing engineering evaluation formula is as follows: In the formula, Indicates the first Instantaneous damage intensity at each sampling point; Indicates the fatigue characteristic coefficient of the material; This represents the normalized driving force intensity.

[0055] To strictly align the above theoretical model with the experimental sequence of this invention, an equivalent mapping of physical quantities is performed: the stress mapping values ​​constructed in step S4 are... Equivalent mapping to normalized driving force intensity This mapping strictly follows the fundamental principle of Ampere's law that the electromagnetic mechanical force is proportional to the square of the current; it normalizes the initial broadband residual. The equivalent mapping is the normalized vibration response amplitude. This mapping characterizes the relative mechanical distortion ratio after stripping away the macroscopic electrical background. Substituting the above mapping variables into the existing formula for instantaneous damage intensity and combining the comprehensive structural constants... The dimensionless instantaneous damage intensity sequence obtained satisfies the relation .

[0056] Furthermore, in order to eliminate structural constants The interference with the engineering threshold setting is addressed, and the discrete and rapidly changing instantaneous damage intensity sequence within the entire sampling period is transformed into a stable linear physical scale, based on the standard definition of root mean square (RMS) energy restoration in signal processing. ,in This represents the root mean square effective value of a discrete signal sequence, used to characterize the average energy level of an alternating signal. Substituting the dimensionless instantaneous damage intensity sequence into this standard formula and performing time-domain accumulation and dimension reduction, the basic damage characteristic amplitude is derived, satisfying the expression: ; In the formula, Indicates the amplitude of the basic damage characteristics; This represents the total number of samples in the measured sequence of transient current derivatives; Indicates the first Initial broadband residuals at each sampling point; Indicates the absolute peak value of the derivative; Represents the stress mapping value. Initial broadband residual. The high-frequency oscillation signal, exhibiting alternating positive and negative values, cannot be directly algebraically summed, resulting in physical cancellation of the values ​​on the time axis. This makes it impossible to accurately measure the severity of the winding oscillation. Therefore, this invention addresses this issue by adjusting the initial broadband residual... With the absolute peak value of the derivative The ratio is squared to convert the alternating oscillation amplitude into a constant positive intensity scalar, when the initial broadband residual... As the fluctuation amplitude increases, the obtained intensity scalar increases by a square factor, objectively reflecting the physical response released by mechanical vibration. The microscopic loosening of the transformer's internal structure is a passive response process; its induced mechanical characteristics only manifest under the excitation of a large electromagnetic mechanical force. Residual fluctuations occurring at low stress levels are usually due to environmental electromagnetic noise interference. This invention addresses this by multiplying the intensity scalar by the stress mapping value. Weighting is applied so that the intensity scalar characterizing mechanical vibration is mapped to the stress value. The peak stress moment is significantly amplified numerically, which corresponds to the actual physical damage characteristics caused by the massive short-circuit impact force tearing the internal structure, while also making the strength scalar in the stress mapping value... The moment of weak force approaching 0 is forcibly attenuated. This forced attenuation eliminates random background electromagnetic noise interference unrelated to the structural force, thereby ensuring that the finally extracted abnormal features originate from the actual internal structural physical damage induced by the huge short-circuit impact force. This invention transforms the discrete and high-frequency drastic instantaneous intensity sequence into a stable effective amplitude by averaging the weighted intensity values ​​over the entire sampling period and then performing a square root operation. It should be noted that this square root operation aims to restore the sum of squares to a linear scale, making it easier to set an intuitive judgment threshold.

[0057] Furthermore, to eliminate the asymmetric interference introduced by the inertial lag of the external busbar, the elasticity factor is... The adaptive suppression coefficient is multiplied by the amplitude of the basic damage characteristic to finally derive the health warning index: ; In the formula, Indicates a health warning index; Indicates the elasticity factor; This represents the amplitude of the basic damage characteristic. The present invention multiplies the amplitude of the basic damage characteristic by the elasticity factor. Using the elasticity factor By using the zero-touch characteristic under external test busbar swaying interference, and physically shielding the false fluctuation signals induced by the mechanical displacement of the external test busbar, the resulting health warning index accurately identifies and characterizes the real damage characteristics caused by internal structural defects in the transformer.

[0058] S6: Outputs a mechanical loosening alarm signal based on the health warning index.

[0059] Specifically, the health warning index is compared with a preset benchmark warning index threshold. In response to the health warning index being greater than the benchmark warning index threshold, the control system outputs an alarm signal indicating that the transformer winding has mechanical loosening or weakening of clamping force.

[0060] Among these methods, health warning index samples were obtained from multiple tests conducted on standard healthy transformers of the same model during routine transformer short-circuit withstand capacity tests. The arithmetic mean and standard deviation of the health warning index samples were calculated, and the final value of the arithmetic mean plus three times the standard deviation was used as the benchmark warning index threshold. The reason for this choice is that the test index of healthy transformers under standard operating conditions objectively follows a normal distribution, based on the well-known 3... The criterion is that the probability of test data falling within the arithmetic mean plus or minus three standard deviations when the system is in normal physical condition reaches 99.73%, making data falling outside this range an extremely low probability event. Therefore, this invention sets the arithmetic mean plus three standard deviations as the limit tolerance boundary for a healthy state, thereby ensuring the scientific validity and engineering robustness of the benchmark early warning index threshold setting. In other embodiments, implementers can set the benchmark early warning index threshold according to the manufacturer's factory tolerance standards for short-circuit withstand capability of a specific batch of transformers.

[0061] For example, Figure 6 The comparison between the final health warning index and the benchmark warning index threshold is shown. Although the measured broadband residual contains a large hysteresis interference component caused by the swaying of the external test busbar, the final output health warning index is forcibly suppressed and safely kept below the benchmark warning index threshold after the identification and product attenuation effect of the elasticity factor, without triggering a false alarm signal. This judgment effect verifies that the comprehensive evaluation model constructed in this invention can deterministically eliminate external pseudo-distortion components, lock in the true health status of the equipment, and thus successfully avoid false judgments that are very easy to occur in traditional conventional test data analysis.

Claims

1. A method for capturing transient current waveforms for transformer short-circuit withstand capability tests, characterized in that, include: S1: Obtain the transient current derivatives at each sampling point in the main circuit of the transformer short-circuit withstand capability test; S2: Based on the differentiation of the classical transient equations of short-circuit loops, establish a macroscopic benchmark derivative analytical model in the derivative domain; S3: The transient current derivative is approximated and fitted to the analytical model of the macroscopic benchmark derivative after time discretization in order to remove the background components of power frequency AC and transient DC and obtain the initial broadband residual characterizing the mechanical vibration of the transformer winding. S4: Calculate the stress mapping value based on the transient current and absolute peak current of each sampling point obtained by fitting, and divide the rise and fall phases of the electromagnetic mechanical force based on the first-order difference result of the stress mapping value. S5: Extract the residual characteristic intensity in the climbing and declining phases respectively to calculate the elasticity factor, and combine the elasticity factor, initial broadband residual and stress mapping value to calculate the health warning index; S6: In response to a health warning index exceeding a preset baseline warning index threshold, output an alarm signal indicating mechanical loosening of the transformer windings.

2. The transient current waveform capture method for transformer short-circuit withstand capability testing according to claim 1, characterized in that, The acquisition of transient current derivatives at each sampling point in the main circuit of the transformer short-circuit withstand capability test includes: The original induced voltage at each sampling point is acquired by a broadband Rogowski coil set in the main circuit; the transient current derivative at the corresponding sampling point is obtained by dividing the original induced voltage at each sampling point by the inherent mutual inductance coefficient of the broadband Rogowski coil.

3. The transient current waveform capture method for transformer short-circuit withstand capability testing according to claim 1, characterized in that, The macroscopic benchmark derivative analytical model satisfies the expression: ； In the formula, An analytical model representing the macroscopic baseline derivative that varies with time; Indicates the peak value of the power frequency current; Indicates the power frequency angular frequency; Indicates time; Indicates the initial phase angle upon closing; Indicates the impedance angle of the main circuit; Indicates the initial amplitude of the DC component; The base of the natural logarithm; Indicates the system's equivalent resistance; Indicates the system's equivalent inductance; This represents the cosine function.

4. The transient current waveform capture method for transformer short-circuit withstand capability testing according to claim 3, characterized in that, The process of approximating the transient current derivative with the analytical model of the time-discrete macroscopic benchmark derivative includes: The transient current derivatives after a short circuit occur are used to construct a measured sequence of transient current derivatives; Using the macroscopic benchmark derivative analytical model after replacing continuous-time variables with discrete-time sampling points as the objective function, a nonlinear least squares fitting algorithm is used to perform multi-parameter regression optimization on the measured sequence of transient current derivatives, and iteratively solves for the optimal solution set of parameters such as system equivalent inductance, system equivalent resistance, initial phase angle of closing, peak power frequency current, and initial amplitude of DC component.

5. The transient current waveform capture method for transformer short-circuit withstand capability testing according to claim 4, characterized in that, The acquisition of the initial broadband residual characterizing the mechanical vibration of the transformer winding includes: Substituting the optimal set of parameters into the analytical model of the macroscopic reference derivative after time discretization, we obtain the macroscopic reference derivative at each discrete time point; subtracting the transient current derivative at each sampling point from the corresponding macroscopic reference derivative at the discrete time point, we obtain the initial broadband residual.

6. The transient current waveform capture method for transformer short-circuit withstand capability testing according to claim 4, characterized in that, The calculation of stress mapping values ​​based on the transient current and absolute peak current at each sampling point obtained through fitting includes: Substitute the optimal set of parameters into the classical transient equation to generate the transient current corresponding to the discrete time point; extract the maximum value among the absolute values ​​of the transient currents corresponding to all discrete time points as the absolute peak value of the current; and take the square of the ratio between the transient current and the absolute peak value of the current as the stress mapping value.

7. The transient current waveform capture method for transformer short-circuit withstand capability testing according to claim 5 or 6, characterized in that, The first-order difference result based on the stress mapping value divides the electromagnetic mechanical force into a rising phase and a declining phase, including: First-order backward difference operation is performed on the stress mapping value to obtain the first-order difference value; the sampling point number corresponding to each discrete time point with the first-order difference value greater than or equal to 0 is obtained to form the first set to represent the climbing stage; the sampling point number corresponding to each discrete time point with the first-order difference value less than 0 is obtained to form the second set to represent the decay stage.

8. The transient current waveform capture method for transformer short-circuit withstand capability testing according to claim 7, characterized in that, The extraction of residual feature intensities during the ascent and decline phases, respectively, includes: The maximum value among the absolute values ​​of the macroscopic baseline derivative at all discrete time points is taken as the absolute peak value of the derivative; Calculate the characteristic strength of the residuals during the ascent period: ; Calculate the residual characteristic strength during the decay period: ; In the formula, Indicates the intensity of residual characteristics during the climbing phase; Indicates the intensity of residual characteristics during the decline period; Indicates the first Initial broadband residuals at each sampling point; Indicates the absolute peak value of the derivative; Represents the first set; This represents the second set.

9. The transient current waveform capture method for transformer short-circuit withstand capability test according to claim 8, characterized in that, The elasticity factor satisfies the expression: like ,but ;like Then set ;in, Indicates the elasticity factor; Indicates the intensity of residual characteristics during the decline period; This indicates the intensity of the residual characteristics during the climbing phase.

10. The transient current waveform capture method for transformer short-circuit withstand capability test according to claim 8, characterized in that, The calculation of the health early warning index includes: Calculate the amplitude of the basic damage characteristics: In the formula, Indicates the amplitude of the basic damage characteristics; This represents the total number of samples in the measured sequence of transient current derivatives; Indicates the first Initial broadband residuals at each sampling point; Indicates the absolute peak value of the derivative; Represents the stress mapping value; The product of the elasticity factor and the amplitude of the basic damage characteristics is used as the health early warning index.

Images