A voltage transformer anti-harmonic control method, device, medium and equipment

Abstract

The application discloses a voltage transformer resonance elimination control method, device, medium and equipment. The method comprises the following steps: synchronously collecting a voltage transformer secondary side instantaneous voltage and a primary side instantaneous current, and outputting a pure characteristic voltage sequence and a pure characteristic current sequence; performing phase space reconstruction on the pure characteristic voltage sequence and judging a resonance type label; constructing a frequency band weighted observation matrix according to the resonance type label, taking a pure characteristic current component of a corresponding frequency band as an observation input, and inversely outputting a time-varying inductance through improved Kalman filtering; substituting the inverted inductance into an equivalent circuit to draw a dynamic reference trajectory; obtaining a pure abnormal deviation amount by real-time difference between a measured phasor trajectory and the dynamic reference trajectory, calculating a damping resistance value and outputting a control signal to drive an execution device to input the damping resistance until the voltage transformer resonance is eliminated. The application can overcome the defect that a traditional fixed power frequency reference fails in physical reference under a transient strong distortion, and improve the resonance elimination success rate by synchronizing the dynamic reference trajectory with the core state in real time.

Description

Technical Field

[0001] This application belongs to the field of power system automation and overvoltage protection technology, specifically relating to a voltage transformer harmonic suppression control method, device, medium and equipment. Background Technology

[0002] Electromagnetic voltage transformers (PTs), due to the nonlinear magnetization characteristics of their cores, are prone to ferroresonant overvoltages when disturbances occur in the power system, thus seriously threatening the safe operation of the power system. Typical disturbance conditions in the power system include: single-phase grounding faults (including metallic grounding and arcing grounding), sudden closing of unloaded busbars or long lines, switching operations of voltage transformers, single-phase open circuits, severe three-phase load asymmetry, lightning strikes, and severe fluctuations in system voltage.

[0003] The basic process of existing harmonic elimination methods is as follows: Fast Fourier Transform (FFT) is used to extract spectral features to determine the resonance type, and then a fixed 50Hz power frequency sine wave is used as the phasor reference to apply the corresponding fixed damping resistor. However, this method has the following drawbacks: First, the effective operation of FFT requires the signal to be periodically stationary. However, ferromagnetic resonant transient signals exhibit continuous frequency shifts and drastic amplitude fluctuations, making them typical non-stationary signals. Under these conditions, FFT is prone to severe spectral leakage, leading to misjudgment of the resonance type.

[0004] Secondly, the phasor reference of existing harmonic elimination methods is a pre-set fixed power frequency constant, which cannot be adjusted in real time according to the core saturation. When the core is highly saturated, this fixed reference is severely out of sync with the actual physical state of the system, causing the control deviation calculated based on it to be distorted, resulting in the control commands losing accurate reference.

[0005] Third, misjudgment of the type and mismatch of the reference lead to a mismatch between the damping resistor value and the actual resonance state. In addition, the FFT requires the accumulation of multiple cycles of data, which results in a serious lag in control and a low success rate of harmonic elimination.

[0006] In summary, the shortcomings of existing methods lie in the fact that the overall framework is based on the implicit assumption that "the signal can be approximated as a steady-state sine wave". This assumption does not hold under transient strong distortion conditions. Therefore, there is an urgent need for a harmonic elimination method that does not rely on the steady-state assumption and is dynamically synchronized with the core saturation in real time. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the purpose of this application is to provide a method, device, medium, and equipment for harmonic suppression control of voltage transformers. This application dynamically reconstructs a virtual steady-state reference phasor trajectory that is synchronized with the time-varying saturation degree of the iron core by real-time inversion, and then subtracts the measured trajectory from the reference in real time. Adaptive closed-loop suppression is performed only for pure abnormal deviations, so that the control reference can match the current state, thereby overcoming the defects of electromagnetic transformer action lag and high misjudgment rate caused by existing methods.

[0008] To achieve the above objectives, this application provides the following technical solution: A voltage transformer harmonic elimination control method based on phasor trajectory is disclosed. The method includes: continuously and synchronously acquiring the instantaneous voltage on the secondary side and the instantaneous current on the primary side of the voltage transformer, removing the power frequency fundamental component, and outputting a pure characteristic voltage sequence and a pure characteristic current sequence; reconstructing the phase space of the pure characteristic voltage sequence to output a resonance type label, constructing a frequency band weighted observation matrix based on the resonance type label, and using the pure characteristic current component of the corresponding frequency band as the observation input for improved Kalman filtering to invert and output a time-varying inductance; substituting the inverted time-varying inductance into the equivalent circuit of the voltage transformer to inversely deduce the voltage phasor corresponding to the current core state, and plotting a dynamic reference trajectory that is synchronized with the core saturation fluctuation in real time point by point; subtracting the measured phasor trajectory from the dynamic reference trajectory in real time to obtain the pure abnormal deviation; adaptively calculating the damping resistance value based on the pure abnormal deviation and the resonance type label to output a control signal; and driving the actuator to operate based on the control signal until the voltage transformer resonance is eliminated.

[0009] Optionally, the step of reconstructing the phase space of the pure characteristic voltage sequence to output a resonance type label includes: determining the embedding dimension and delay time of the phase space; reconstructing the pure characteristic voltage sequence based on the embedding dimension and delay time of the phase space to obtain a reconstructed attractor; extracting the nonlinear characteristic quantity of the reconstructed attractor, and determining the resonance type label based on the characteristic quantity.

[0010] Optionally, the step of extracting the nonlinear feature quantity of the reconstructed attractor and determining the resonance type label based on the feature quantity includes: calculating the correlation dimension and the maximum Lyapunov exponent of the reconstructed attractor; comparing the correlation dimension and the maximum Lyapunov exponent with a set threshold, and determining the resonance type label based on the comparison result.

[0011] Optionally, the step of comparing the correlation dimension and the maximum Lyapunov exponent with a set threshold, and determining the resonance type label based on the comparison result, includes: if the correlation coefficient is in a first preset interval and the maximum Lyapunov exponent is less than zero, the resonance type is determined to be low-frequency resonance; if the correlation coefficient is in a second preset interval and the maximum Lyapunov exponent is approximately equal to zero, the resonance type is determined to be fundamental frequency resonance; if the correlation coefficient is in a third preset interval and the maximum Lyapunov exponent is greater than zero, the resonance type is determined to be high-frequency resonance.

[0012] Optionally, the step of constructing a frequency-weighted observation matrix based on the resonance type label and using the pure characteristic current component of the corresponding frequency band as the observation input for the improved Kalman filter to invert the time-varying inductance includes: separating the pure characteristic current sequence into current components of different frequency bands, calculating the instantaneous energy of each current component and its proportion in the total energy, determining whether to activate the improved Kalman filter inversion based on the proportion, and if activated, proceeding to the next step; otherwise, exiting the inversion operation for this period and waiting for the next sampling period to re-determine the energy proportion; constructing a frequency-weighted observation matrix based on the resonance type label; and weighting the observation residuals based on the frequency-weighted observation matrix to invert the time-varying inductance.

[0013] Optionally, the step of substituting the time-varying inductance output from the inversion into the equivalent circuit of the voltage transformer, back-deriving the voltage phasor corresponding to the current core state, and plotting the dynamic reference trajectory that is synchronized with the core saturation fluctuation in real time point by point includes: substituting the time-varying inductance into the equivalent circuit of the primary side of the voltage transformer to construct the time-varying equivalent impedance at the current moment; calculating the amplitude of the time-varying equivalent impedance and obtaining the normalized complex rotation factor based on the amplitude; obtaining the voltage phasor that the system should have when operating stably under the current core state based on the normalized complex rotation factor and the rated phase voltage of the distribution network; and obtaining the dynamic reference trajectory by plotting the voltage phasor point by point.

[0014] Optionally, the step of obtaining the pure abnormal deviation by subtracting the measured phasor trajectory from the dynamic reference trajectory in real time, adaptively calculating the damping resistance value based on the pure abnormal deviation value and the resonance type label to output a control signal, and driving the actuator to operate based on the control signal until the voltage transformer resonance is eliminated includes: obtaining the measured phasor trajectory and subtracting it from the dynamic reference trajectory to obtain the pure abnormal deviation value; calculating the amplitude of the pure abnormal deviation value and performing trend prediction to obtain the deviation prediction value; calculating the adaptive damping resistance value based on the deviation prediction value and the resonance type label; and converting the adaptive damping resistance value into a control signal to drive the actuator to engage the damping resistance to achieve resonance elimination.

[0015] This application also provides a voltage transformer harmonic elimination control device based on phasor trajectory. The device includes: an acquisition and output module for continuously and synchronously acquiring the instantaneous voltage on the secondary side and the instantaneous current on the primary side of the voltage transformer, removing the power frequency fundamental component, and outputting a pure characteristic voltage sequence and a pure characteristic current sequence; a reconstruction and inversion module for performing phase space reconstruction on the pure characteristic voltage sequence to output a resonance type label, constructing a frequency band weighted observation matrix based on the resonance type label, and using the pure characteristic current component of the corresponding frequency band as the observation input for improved Kalman filtering to invert and output a time-varying inductance; a plotting module for substituting the inverted time-varying inductance into the equivalent circuit of the voltage transformer, back-deriving the voltage phasor corresponding to the current core state, and plotting a dynamic reference trajectory that is synchronized with the core saturation fluctuation in real time; and a resonance elimination module for obtaining the pure abnormal deviation by subtracting the measured phasor trajectory from the dynamic reference trajectory in real time, adaptively calculating the damping resistance value based on the pure abnormal deviation and the resonance type label to output a control signal, and driving the actuator to operate based on the control signal until the voltage transformer resonance is eliminated.

[0016] This application also provides a computer storage medium storing computer instructions for performing the method as described in the preceding claim.

[0017] This application also provides an electronic device, the electronic device comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the method as described in any of the preceding claims.

[0018] Compared with the prior art, the beneficial effects of this application are as follows: This application enables accurate identification of resonance types through phase space reconstruction. Constructing a frequency band weighted observation matrix based on resonance type labels can effectively reduce the time-varying inductor inversion error. Furthermore, the improved Kalman filter point-by-point recursive mechanism significantly compresses the control response delay, ultimately improving the harmonic elimination success rate. Attached Figure Description

[0019] Figure 1 This is a schematic flowchart of a voltage transformer harmonic elimination control method provided in this application; Figure 2 This is a time-domain comparison diagram of the original voltage signal and the pure characteristic voltage sequence output after adaptive notch filtering. Figure 3 This is a time-domain comparison diagram of the original current signal and the pure characteristic current sequence output after adaptive notch filtering. Figure 4 A schematic diagram of the phasor trajectory of a pure characteristic voltage sequence in the complex plane; Figure 5 A schematic diagram of the phasor trajectory of a pure characteristic current sequence on the complex plane; Figure 6 This is a schematic diagram comparing the harmonic elimination control performance of this application with that of the traditional fixed-reference FFR analysis method; Figure 7 This diagram illustrates the comparison of recognition accuracy under different resonance types. Detailed Implementation

[0020] Specific embodiments of this application will now be described in detail with reference to the accompanying drawings. While specific embodiments of this application are shown in the drawings, it should be understood that this application can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of this application and to fully convey the scope of this application to those skilled in the art.

[0021] To facilitate understanding of the embodiments of this application, the following will provide further explanation and description with reference to the accompanying drawings and specific embodiments, and the drawings do not constitute a limitation on the embodiments of this application.

[0022] Figure 1 This application provides an exemplary embodiment of a voltage transformer harmonic elimination control method based on phasor trajectories, such as... Figure 1 As shown, the method includes the following steps: S100: Continuously and synchronously acquires the instantaneous voltage on the secondary side and the instantaneous current on the primary side of the voltage transformer, removes the power frequency fundamental component, and outputs the pure characteristic voltage sequence and the pure characteristic current sequence. In this step, this embodiment synchronously acquires the open delta voltage on the secondary side of the voltage transformer at a sampling frequency of 10kHz. and a single instantaneous side current After analog-to-digital conversion, discrete voltage and current digital sequences are obtained. In these discrete voltage and current digital sequences, the energy of the 50Hz power frequency fundamental component is much greater than the characteristic component generated by the ferromagnetic resonance. To prevent this power frequency fundamental component from overshadowing the characteristic component generated by the ferromagnetic resonance in subsequent analysis, this embodiment feeds the discrete voltage and current digital sequences into an adaptive notch filter. This filter operates at the power frequency... As the notch center, it can accurately remove the fundamental frequency component and interference in the nearby frequency band, and output a pure characteristic voltage sequence containing only resonance distortion information. and pure characteristic current sequence .

[0023] Figure 2This is a time-domain comparison diagram of the original voltage signal (containing a 50Hz power frequency component) and the pure characteristic voltage sequence output after adaptive notch filtering. In the diagram, the power frequency component in the original voltage signal has a large amplitude and exhibits a regular 50Hz sinusoidal oscillation pattern, which masks the resonance distortion characteristics. In the filtered pure characteristic voltage sequence, the power frequency component is effectively suppressed, the waveform amplitude is significantly reduced, and it exhibits an irregular distortion pattern caused by ferromagnetic resonance, thus clearly reflecting the resonance characteristic information.

[0024] Figure 3 This diagram illustrates the time-domain comparison between the original current signal (containing a 50Hz power frequency component) and the pure characteristic current sequence output after adaptive notch filtering. In the diagram, the original current signal is dominated by the 50Hz power frequency component, with a regular waveform and a large amplitude. After filtering, the power frequency component is accurately removed, the amplitude of the output pure characteristic current sequence is significantly reduced, the non-power frequency distortion component related to resonance is retained, and the waveform exhibits obvious non-periodic and fluctuating characteristics, thus providing a clean data foundation for subsequent resonance analysis and control.

[0025] Figure 4 and Figure 5 The phasor trajectories of pure characteristic voltage sequences and pure characteristic current sequences on the complex plane are shown respectively. The plotting method is as follows: First, Hilbert transform is performed on the original voltage or current sampling sequence to construct an analytical signal, thereby obtaining the instantaneous amplitude (real part) and orthogonal component (imaginary part) corresponding to each sampling moment. Then, the phasor trajectory is plotted point by point in time sequence with the real part as the horizontal axis and the imaginary part as the vertical axis. Figure 4 and Figure 5 The evolution of voltage and current phasors over time is clearly presented. Circles mark the starting points of the trajectory, and squares mark the ending points. The spiraling trend of the trajectory reflects the dynamic changes of the phasors over time. The geometric shape of this phasor trajectory contains rich system characteristic information, including: the ellipticity (ratio of major to minor axis) reflects the amplitude ratio or damping ratio between characteristic components; the rotation direction (clockwise or counterclockwise) indicates the positive or negative sequence nature of the characteristic frequency components or the phase lead / lag relationship; and the number of loops the trajectory passes around the origin is closely related to the composition of the system's resonant frequency and the beat frequency effect or modal coupling between the frequency components. By analyzing the elliptical shape, rotation direction, and number of loops of this phasor trajectory, the frequency composition and phase evolution of the resonant voltage, as well as the oscillation mode and phase change trend of the resonant current, can be intuitively identified. This provides measured phasor evidence for subsequent identification of resonance types and abnormal deviations.

[0026] S200: For pure characteristic voltage sequences Phase space reconstruction is performed to output resonance type labels. Based on the resonance type labels, a frequency band weighted observation matrix is ​​constructed, and the pure characteristic current components of the corresponding frequency bands are used as the observation inputs for the improved Kalman filter to invert the time-varying inductance. S300: Substitute the time-varying inductance output from the inversion into the equivalent circuit of the voltage transformer, reverse the voltage phasor corresponding to the current core state, and draw the dynamic reference trajectory that is synchronized with the core saturation fluctuation in real time point by point. S400: The measured phasor trajectory is subtracted from the dynamic reference trajectory in real time to obtain the pure abnormal deviation. Based on the pure abnormal deviation and the resonance type label, the damping resistance value is adaptively calculated to output the control signal. The actuator is driven to operate based on the control signal until the voltage transformer resonance is eliminated.

[0027] This embodiment first utilizes adaptive notch filtering to remove power frequency interference from the original voltage and current signals, thus preventing strong power frequency components from overwhelming the resonant characteristic signals and providing a clean data foundation for subsequent analysis. Second, phase space reconstruction is used to identify the resonance type, overcoming the spectral leakage defect of traditional spectrum analysis under non-steady-state distortion, ensuring the robustness of resonance type identification, and constructing a frequency band weighted observation matrix accordingly. This allows the improved Kalman filter to invert the time-varying inductance based solely on the dominant frequency current component, eliminating cross-frequency error contamination and enabling accurate tracking of the real-time saturation state of the core. Furthermore, this embodiment substitutes the inverted inductance into the equivalent circuit to construct a dynamic reference trajectory synchronized with the core saturation fluctuations. This solves the fundamental defect of the traditional fixed power frequency reference failing under transient strong distortion conditions. Finally, the measured trajectory is subtracted from the dynamic reference point by point to extract the pure resonance abnormal component after eliminating saturation distortion and implement adaptive closed-loop suppression. This ensures that the control command is always based on an accurate physical reference system, overcoming the problems of lag and high misjudgment rate of traditional methods, thereby improving the harmonic elimination success rate of the voltage transformer.

[0028] In another exemplary embodiment, in step S200, the pure characteristic voltage sequence To perform phase space reconstruction to output resonance type labels, the following steps are included: S201: Determine the embedding dimension of the phase space and delay time ; In this step, the mutual information method is first used to determine the phase space delay time. Specifically, this includes: targeting pure characteristic voltage sequences Calculate its relationship with the delay in turn. The sequence after sampling points Mutual information between (The greater the mutual information, the more information the two sequences share.) If the height is too small, there will be redundant heights between adjacent points. If the information is too large, the information becomes independent and loses its correlation. Then, the mutual information content is plotted. With delay time The changing curve, and the mutual information content is selected. First drop to its maximum value Time corresponding to As the optimal delay time (when mutual information) First drop to its maximum value At this time, the information redundancy between the reconstructed coordinates has been fully eliminated, while the system dynamic correlation has not been completely lost, allowing the phase space to achieve an optimal balance between "information independence" and "correlation preservation," ensuring that the reconstructed coordinates maintain sufficient independence without losing the intrinsic correlation of the dynamic system. Furthermore, in the delay time... Based on the established parameters, this embodiment uses the spurious nearest neighbor method to determine the embedding dimension of the phase space. Specifically, this involves: starting with m=1, progressively increasing the dimension, and at each dimension, calculating whether the nearest neighbor of each point in the reconstructed phase space remains the nearest neighbor after the dimension increases (if it is no longer the nearest neighbor after the dimension increases, it is called a "false nearest neighbor"). When the dimension increases to a certain value, and the proportion of false nearest neighbors first drops below a preset threshold (e.g., 1% or 5%), this dimension is the optimal embedding dimension of the phase space. Its physical significance is that when the dimension is sufficiently high, the geometry of the reconstructed attractor is fully unfolded, and the trajectories of different states no longer produce "false intersections" in the low-dimensional space due to projection dimensionality reduction (i.e., two points that were originally far apart in the real high-dimensional space are mistakenly identified as nearby points due to projection into the low-dimensional space), ensuring that the proximity relationships in the phase space truly reflect the proximity relationships of the system states. It should be noted that the determination of both parameters was completed in one go during the offline phase using a large number of known types of mixed experimental and simulation datasets.

[0029] S202: Embedding Dimension Based on Phase Space and delay time For pure characteristic voltage sequences Perform reconstruction to obtain reconstruction attractors; In this step, this embodiment is based on the embedding dimension of the phase space. and delay time Based on the following formula, the pure characteristic voltage sequence Reconstructed A set of points in a dimensional phase space :

[0030] In this point set, each sampling time k corresponds to a point in the phase space. All points are connected in chronological order to form a trajectory curve in m-dimensional space, which is the reconstructed attractor.

[0031] The aforementioned reconstruction attractor is formed by connecting all delay vector points in chronological order. It is the trajectory of the system's state evolution in the reconstructed phase space. The set shape of this trajectory (including geometric features such as ellipticity, degree of distortion, sharp angles, concavity, and number of rings) directly describes the motion law of the ferromagnetic resonant system in the phase space.

[0032] Different resonance types correspond to different trajectory shapes. An ideal sine wave presents a smooth ellipse; low-frequency resonances exhibit distorted or concave trajectories due to waveform clipping distortion; and high-frequency resonances, due to complex superposition of oscillations, present trajectories with multiple intertwined loops or chaotic dispersion. Therefore, the type of ferromagnetic resonance can be identified by analyzing the geometric characteristics of the trajectory.

[0033] S203: Extract the nonlinear characteristic quantity of the reconstructed attractor and determine the resonance type label based on the characteristic quantity.

[0034] In this step, after completing the phase space reconstruction and obtaining the point set in the m-dimensional phase space, this embodiment first uses the Grassberger-Procaccia algorithm to calculate the correlation dimension D of the reconstructed attractor. This calculation process includes: for different radii r, statistically calculating the correlation integrals of phase points with distances less than r. ,exist and In the curve, a linear region at an intermediate scale is selected for straight-line fitting. The slope of this region is used as the correlation dimension D to quantitatively describe the geometric complexity of the reconstructed attractor. A smaller correlation dimension D indicates that the motion of the reconstructed attractor tends to be more periodic and ordered. Furthermore, this embodiment uses the Wolfalgorithm algorithm to calculate the maximum Lyapunov exponent. This includes: tracking the average divergence rate of multiple nearest neighbor pairs over time in the reconstructed phase space, the sign of which directly reflects the system's sensitivity to initial disturbances. Wherein, if... This indicates that the disturbance converges and the system tends to stabilize during periodic motion; This indicates that the disturbance is diverging and the system is in a high-frequency complex oscillation state.

[0035] Furthermore, the calculated correlation dimension D and the maximum Lyapunov exponent are... The threshold value is compared with the offline threshold library as follows: Where the D value is within the first preset range (approximately 1.0~1.8) and... If the value is less than 0, it is determined to be a low-frequency resonance; if the D value is within the second preset range (approximately 1.5~2.2) and (A slightly negative or near-zero value) indicates fundamental frequency resonance; if the D value is within the third preset range (approximately 2.5~4.0) and A value greater than 0 indicates high-frequency resonance; while the D value corresponding to the normal state without resonance is the smallest (approximately 0.5~1.2). If the value is significantly negative, a resonance type label will be output based on this comprehensive judgment. .

[0036] It is important to understand that the above discrimination method does not rely on the stable sinusoidal assumption required by FFT (Fast Fourier Transform), nor does it require any specific circuit parameters. This is because the correlation dimension D and the maximum Lyapunov exponent are intrinsic nonlinear invariants extracted after phase space reconstruction. Their calculations focus on the geometric structure and divergence / convergence rate of the voltage waveform's own evolution trajectory, rather than the frequency domain energy distribution or circuit impedance parameters. Therefore, regardless of whether the waveform satisfies the periodic stationarity condition or whether the circuit parameters are known, the aforementioned characteristic quantities can be extracted from the reconstructed attractor, enabling accurate identification of the resonance type label.

[0037] In another exemplary embodiment, step S200, which involves constructing a frequency band weighted observation matrix based on the resonance type label and using the pure characteristic current component of the corresponding frequency band as the observation input for the improved Kalman filter to invert the time-varying inductance, includes the following steps: S2001: Transform the pure characteristic current sequence The current components are separated into different frequency bands. The instantaneous energy of each current component and its proportion in the total energy are calculated. Based on the proportion, it is determined whether to activate the improved Kalman filter inversion. If it is determined to be activated, step S2002 is executed; otherwise, the inversion operation of this cycle is exited and the energy proportion is determined again in the next sampling cycle. In this step, this embodiment will use the pure characteristic current sequence Simultaneously, the current is fed into bandpass filters with passbands of 5 to 25 Hz (low frequency), 45 to 55 Hz (fundamental frequency), and 100 to 500 Hz (high frequency), separating the current components in the three frequency bands, which are denoted as follows: , , Then, the instantaneous energy of the current components in each frequency band was calculated. and its proportion in total energy And when the resonance type label When the instantaneous energy percentage of the corresponding frequency band is greater than the threshold (e.g., 60%), the subsequent improved Kalman wave inversion can be activated; otherwise, the operation will be forcibly terminated in this cycle.

[0038] S2002: Based on resonance type tag Construct a frequency band weighted observation matrix; In this step, this embodiment relies on the resonance type label. And construct the frequency band weighted observation matrix according to the following rules. The matrix consists of 1×3 row vectors, with the three vectors corresponding to the weighting coefficients of the low-frequency, fundamental, and high-frequency components, respectively. , , :

[0039]

[0040]

[0041] In the above weighting coefficients, 1 indicates that the frequency band component is enabled to participate in the Kalman filter observation update, and 0 indicates that the frequency band component is blocked.

[0042] The physical meaning of this observation matrix is ​​expressed as follows: In subsequent Kalman filter update steps, the weighted observation residuals are... It only includes current component information at the current resonant frequency; residual components in non-resonant frequency bands are forcibly cleared to zero. Specifically, when the resonant type label... When it is low frequency, ( This represents the weighted observation residuals in the low-frequency band. This represents the measured low-frequency current component. This indicates that the model predicts the low-frequency current component (estimated from state priors), and the residual components of the fundamental and high-frequency currents are forcibly cleared to zero and do not participate in state correction; when the resonance type label... When it is the fundamental frequency, ( This represents the weighted observation residual in the fundamental frequency band. This represents the measured fundamental frequency current component. This indicates that the model predicts the fundamental frequency current component (estimated from state priors), with only the fundamental frequency current residual participating in the correction; when the resonance type label... High frequency, ( This represents the weighted observation residuals in the high-frequency band. This represents the measured high-frequency current component. This means that the model predicts the high-frequency current component (estimated by state prior), and only the high-frequency current residual participates in the correction, thereby cutting off the path of mutual contamination of errors across the entire frequency band.

[0043] S2003: Weight the observation residuals based on the frequency band weighted observation matrix to invert the output time-varying inductance.

[0044] In this step, this embodiment uses a state vector Construct a discrete nonlinear state-space model of the voltage transformer, where... Indicates instantaneous current. This represents the core flux linkage (an intermediate variable describing the magnetization state of the core and determining whether the core enters the saturation region). This represents the time-varying inductance to be inverted.

[0045] The recursive process of Kalman filtering includes the following steps: First, predict the state estimate of the previous time step. Based on the system state transition model, calculate the prior state estimate at the current moment. and its error covariance :

[0046]

[0047] in, Represents the state transition function. Represents the state transition matrix. Represents the process noise covariance matrix. This indicates transpose.

[0048] Secondly, the frequency band weighted observation matrix A Kalman filter update stage is introduced to update the measured current. and model predicted current Perform synchronous weighting to obtain weighted observation residuals that contain only the dominant frequency component information. ( The measured current vector, For the observation matrix, (For model prediction of current vector), where, Based on the resonance type label construction, its function is to retain the residual components in the dominant frequency band and force the residual components in non-dominant frequency bands to zero. Subsequently, the Kalman gain is calculated. and weighted observation residuals Multiplied by Kalman gain The prior state estimate is then corrected to obtain the posterior state estimate. ,in, This represents the estimated value of the instantaneous current. This represents the estimated value of the iron core flux linkage. This represents the estimated value of the time-varying inductance. This indicates transpose.

[0049] Through the above update steps, the state correction of the Kalman filter depends only on the current component information corresponding to the current resonant frequency, thereby reversing the output of a time-varying inductor that is strongly correlated with the current resonant frequency and is not contaminated by errors in other frequency bands.

[0050] Furthermore, from posterior state estimation Extracting the third dimension component This is used as the time-varying inductance estimate at the current sampling moment and output to the next step for calculating the dynamic reference trajectory. Then, let... Return to the prediction step in step S2003 and repeat the prediction and update process until the voltage transformer resonance is eliminated and the inversion is terminated.

[0051] In summary, the above steps first select a frequency band by using the resonance type label, then use a bandpass filter to extract the frequency band component from the pure characteristic current, and finally feed this component as the only reliable observation basis to the Kalman filter, forcing it to correct the state estimate only based on the residual of the main frequency current, thereby retrieving the time-varying inductor that is strongly correlated with the current resonant main frequency and is not contaminated by errors in other frequency bands.

[0052] In another exemplary embodiment, step S300, which involves substituting the time-varying inductance output from the inversion into the equivalent circuit of the voltage transformer, deducing the voltage phasor corresponding to the current core state, and plotting a dynamic reference trajectory that is synchronized with the core saturation fluctuation in real time, includes the following steps: S301: Estimate the time-varying inductance value Substituting the equivalent circuit of the primary side of the voltage transformer, we construct the time-varying equivalent impedance at the current moment. :

[0053] in, The DC resistance of the primary winding (reflects the copper loss of the coil, which is given by the manufacturer's parameters for the current transformer). It is the power frequency angular frequency. The equivalent capacitance to ground. This is the magnetizing reactance, reflecting the effect of core saturation on impedance. For resistance, The imaginary unit indicates that the voltage on an inductor leads the current by 90°, and the voltage on a capacitor lags the current by 90°.

[0054] The physical meaning of the above time-varying equivalent impedance is: under the current core saturation level and under the action of a fixed power frequency of 50Hz, the equivalent impedance seen from the primary side of the voltage transformer decreases as the estimated value of the time-varying inductance decreases (the deeper the saturation, the lower the equivalent impedance).

[0055] S302: Calculate time-varying equivalent impedance amplitude And based on the magnitude, a normalized complex rotation factor is obtained. ; It should be noted that this factor is a complex number with a modulus of 1. Its real part reflects the proportion of the resistive component of the impedance, and its imaginary part reflects the proportion of the inductive / capacitive component of the impedance. Its argument represents the phase shift angle of the voltage relative to the current under the current core condition.

[0056] S303: Based on normalized complex rotation factor And the rated phase voltage of the distribution network, to obtain the voltage phasors that the system should have when operating stably under the current core condition; In this step, this embodiment uses the effective value of the rated phase voltage of the distribution network. (A known constant) is used as input, multiplied by the normalized complex rotation factor obtained in step S302. To obtain the voltage phasor that the system should have for stable operation under the current core condition. :

[0057] It should be noted that the working principle of a voltage transformer is as follows: voltage is applied to the primary side → magnetic flux is generated in the iron core → voltage is induced on the secondary side. When the saturation level of the iron core changes, the magnetizing inductance changes, causing a shift in the phase angle of the equivalent impedance, and the phase relationship between the induced voltage on the secondary side and the primary side voltage changes accordingly. This normalized complex rotation factor can accurately describe this phase shift. Multiplying it by the rated voltage yields the phasor (including amplitude and phase information) that the secondary side voltage should have under the current iron core saturation state. Compared with a fixed power frequency reference, this phasor can be updated point by point according to the iron core state, rather than being a pre-set fixed constant.

[0058] S304: Based on point-by-point plotting of voltage phasors, a dynamic reference trajectory is obtained.

[0059] In this step, the voltage phasor calculated at each sampling time k is plotted on the complex plane, where the horizontal axis represents the real part (resistive component) and the vertical axis represents the imaginary part (inductive / capacitive component). By connecting the points sequentially in time order, the dynamic reference trajectory can be obtained.

[0060] in, This represents the dynamic baseline trajectory.

[0061] In the above dynamic reference trajectory, as k increases, the time-varying inductor is updated once per sampling period, and the voltage phasor changes point by point accordingly. The entire trajectory extends continuously on the complex plane to track the fluctuations in the core saturation level in real time, forming a dynamic reference system synchronized with the core state. This trajectory is output to the next step to be used for point-by-point subtraction with the measured phasor trajectory.

[0062] In another exemplary embodiment, in step S400, the step of obtaining the pure abnormal deviation by subtracting the measured phasor trajectory from the dynamic reference trajectory in real time, adaptively calculating the damping resistance value based on the pure abnormal deviation and the resonance type label to output a control signal, and driving the actuator to operate based on the control signal until the voltage transformer resonance is eliminated, includes the following steps: S401: Obtain the measured phasor trajectory and subtract it from the dynamic reference trajectory to obtain the pure abnormal deviation. In this step, the pure characteristic voltage sequence obtained in S100 is... Performing a Hilbert transform yields an analytic signal, from which the measured phasor trajectory on the complex plane is obtained. :

[0063] in, Represents the Hilbert transform. This represents an imaginary number. The measured phasor trajectory reflects the actual evolution path of the voltage phasor in the complex plane of the real system, including all voltage distortion information caused by ferromagnetic resonance.

[0064] Furthermore, the measured phasor trajectory With dynamic reference trajectory Real-time subtraction on the same complex plane yields the pure outlier deviation. :

[0065] The effect of the above subtraction is that core saturation itself causes normal waveform distortion (i.e., waveform changes accompanying the current saturation state). This distortion exists simultaneously in both the measured trajectory and the dynamic reference trajectory, and the subtraction can automatically cancel it out. This refers to the abnormal voltage component generated by ferroresonance after eliminating the effects of saturation. In contrast, traditional fixed references cannot distinguish between distortion caused by saturation and distortion caused by resonance, resulting in a persistent systematic deviation in control commands.

[0066] S402: Calculate the pure abnormal deviation. The amplitude is determined and a trend is predicted to obtain the deviation prediction value; In this step, the pure abnormal deviation can be calculated using the following formula. amplitude :

[0067] Furthermore, a fourth-order autoregressive model is used to predict the short-term trend of this amplitude, and to infer the predicted deviation values ​​for the next P sampling points (e.g., P=5). To compensate for delays in the actions of the implementing agencies.

[0068] S403: Based on Deviation Prediction Value and Resonance Type Label The adaptive damping resistance value is calculated. In this step, the current resonance type label is... Deviation prediction value Current time-varying inductance estimate Input the adaptive damping calculation model to obtain the adaptive damping resistance value:

[0069] in, For adaptive damping resistance value, This is the angular frequency corresponding to the resonance type; for example, 2 is taken for low frequencies. ×25, with the base frequency taken as 2 ×50, high frequency takes 2 ×150, The proportionality coefficients related to the resonance type were determined through offline training. This is the deviation correction coefficient, ensuring that the greater the deviation, the stronger the damping.

[0070] S404: Converts the adaptive damping resistor value into a control signal to drive the actuator to engage the damping resistor to achieve harmonic elimination.

[0071] In this step, the calculated adaptive damping resistance value will be... After being limited and mapped to the actuator range, the signal is converted into a control signal (such as duty cycle or switching command) for the actuator (such as a thyristor switching resistor or a PWM adjustable resistor). This signal is output to the actuator and engages the damping resistor. The damping resistor is then switched to the open delta winding (secondary side) of the voltage transformer or between the primary neutral point and ground (primary side) to form a path for dissipating resonant energy. It should be noted that the specific switching position must be selected based on the system wiring method: The open delta switching method connects the damping resistor in parallel across the open delta on the secondary side of the voltage transformer. This method is suitable for distribution networks where the neutral point is not grounded or grounded via an arc suppression coil. This method has minimal impact on primary side equipment and provides a fast control response. The primary side neutral point switching method connects the damping resistor in series between the primary neutral point and ground. This method is suitable for systems where the neutral point is directly grounded or grounded via a small resistor. This method provides a more direct damping effect but requires higher resistor withstand voltage and heat dissipation.

[0072] Finally, repeat the above steps in each sampling period and continuously monitor the amplitude of pure abnormal deviations. ,when If the voltage transformer resonance is less than the threshold (e.g., 5% of the rated voltage) for N consecutive sampling cycles (e.g., N=20, corresponding to 2 power frequency cycles), the resonance is considered eliminated, and the control loop is exited; otherwise, closed-loop control continues until the voltage transformer resonance is eliminated. It should be explained here that the threshold is set to 5% of the rated voltage based on the following considerations: Firstly, according to "DL / T 620-2019 Overvoltage Protection and Insulation Coordination of AC Electrical Installations" and "Q / GDW 415-2010..." The "Technical Specification for Ferroresonant Suppression Devices in Distribution Networks" stipulates that the domestic power industry has long used 5% as a common criterion for successful resonance suppression. Secondly, when the resonance voltage amplitude drops below 5% of the rated voltage, the resonance energy has decayed to 1 / 400 of the initial value (energy is proportional to the square of the amplitude), and the threat to equipment insulation and system safety is negligible. Thirdly, the 5% threshold provides sufficient safety margin for the control loop, avoiding system oscillations caused by frequent switching of damping resistors due to measurement noise or minor disturbances, and preventing residual energy from being amplified again under subsequent disturbances when the threshold is too high (e.g., 10%), leading to resonance re-ignition. Fourthly, the measurement accuracy of the secondary side of conventional voltage transformers is typically in the range of 0.5% to 1%, and the superposition of measurement noise and AD quantization error usually does not exceed 2% to 3%. The 5% threshold ensures that the judgment result is significantly higher than the noise floor, guaranteeing the reliability of the criterion. Considering all these factors, 5% is an engineering threshold that achieves a reasonable balance between industry standards, physical energy decay, control robustness, and detection accuracy.

[0073] To verify the effectiveness of the solution described in this application, the following two treatment groups were set up for comparative experiments: T0 (Traditional Fixed Reference FFT Analysis Method): The traditional fast Fourier transform is used to extract the voltage / current spectrum characteristics. A fixed 50Hz power frequency sine wave is used as the phasor reference. The resonance type is determined based on the peak value of the spectrum, and then the corresponding fixed damping resistor is applied.

[0074] T1 (This application): The voltage transformer harmonic elimination control method based on phasor trajectory proposed in this application is adopted.

[0075] This experiment employs five-fold cross-validation, which involves randomly dividing 80 sets of mixed datasets consisting of measured and simulated data into five subsets, each containing 16 sets of data. Four subsets (64 sets in total) are selected sequentially as the training set, and the remaining subset (16 sets in total) is used as the validation set. This process is repeated five times, and the mean and standard deviation of the evaluation metrics are then calculated. Four evaluation metrics are used: ① Resonance type identification accuracy (%), which refers to the proportion of test samples that correctly identify low-frequency / fundamental-frequency / high-frequency resonance types; ② Time-varying inductance inversion error (%), which refers to the error in the inverted inductance... ③ Average relative error between the actual inductance value calibrated in the offline experiment; ④ Control response delay (ms), which refers to the time interval from the appearance of the resonance characteristic to the output of the control quantity; ⑤ Ferromagnetic resonance cancellation success rate (%), which refers to the proportion of test samples whose resonance amplitude decays to less than 5% of the rated voltage within 2 power frequency cycles after the control action.

[0076] Table 1 shows the comparison results of harmonic suppression control performance between the traditional fixed-reference FFR analysis method (T0) and the method of this application (T1). Figure 6 As shown.

[0077] Table 1 Comparison of Harmonic Suppression Control Performance of Different Methods

[0078] First, it should be noted that the four evaluation indicators in Table 1 are set based on the following criteria: The resonance type identification accuracy is used to verify the effectiveness of the nonlinear feature extraction method based on phase space reconstruction under strongly distorted waveforms, because accurate identification of the resonance main frequency type is a prerequisite for the subsequent construction of the frequency band weighted observation matrix and the adaptive calculation of the damping resistor. If the type is misidentified, all subsequent control links will fail; the time-varying inductor inversion error is used to verify whether the frequency band weighted observation matrix can effectively eliminate the mutual contamination of errors across the entire frequency band, because the inversion accuracy of the time-varying inductor directly determines the accuracy of the subsequent construction of the dynamic reference trajectory; the control response delay is used to verify the delay advantage of the point-by-point recursive mechanism compared with the traditional window function accumulation mechanism, because the transient process of ferromagnetic resonance can usually damage the equipment within hundreds of milliseconds. If the delay is too long, the control loses its timeliness; the ferromagnetic resonance harmonic elimination success rate is used to verify whether the complete closed loop from type identification to adaptive control can reliably converge, and it is also the final criterion for measuring the engineering practical value of the entire method. The above four indicators form a progressive evaluation system from local link verification to overall effect verification. The first three indicators independently quantify and assess the key sub-links in the method, while the last indicator comprehensively evaluates the final control effect of the whole method. This system can both identify weak links and evaluate overall effectiveness, thus fully covering all the innovative points of the core technical solution of this application.

[0079] In addition, from Table 1 and Figure 6It can be seen that the resonance type identification accuracy of this application (T1) is 94.7%, which is 16.4% higher than the traditional method (T0). This is due to the nonlinear characteristic quantity based on phase space reconstruction, which does not rely on precise circuit parameters during discrimination and can still maintain high robustness under severe waveform distortion conditions. In contrast, the traditional fast Fourier transform suffers from frequent misjudgments due to spectral leakage under non-steady-state distortion. The time-varying inductance inversion error is reduced from 22.7% to 13.4%, and the absolute value of the error is reduced by 9.3 percentage points. This is because this application dynamically constructs a weighted observation matrix based on the resonance type label and forces only the main frequency current component as the observation basis, eliminating the mutual contamination of errors across the entire frequency band. The control response delay is significantly reduced from 146ms to 28ms (a reduction of about 80.8%). The core reason is that the improved Kalman filter recursively updates the state once per sampling period (0.1ms), without waiting for the data window to accumulate. In contrast, the traditional fast Fourier transform method requires the acquisition of multiple cycles (80ms to 200ms) before it can be calculated. The success rate of harmonic elimination increased from 72.9% to 94.6%, an improvement of 21.7%. This is attributed to the fact that the dynamic reference trajectory constructed in this application can be synchronized with the core saturation in real time, and the control loop only implements closed-loop targeted control for pure abnormal deviations after eliminating the influence of saturation, thereby solving the problems of parameter mismatch and action lag under the traditional fixed reference.

[0080] This application further examines the identification performance of the proposed method under different resonance types, and the comparison results are shown in Table 2 and below. Figure 7 As shown: Table 2 Comparison of identification accuracy for different resonance types

[0081] From Table 2 and Figure 7 It can be seen that among the three resonance types, the recognition accuracy of this application (T1) is significantly better than that of the traditional method (T0). Among them, the improvement in low-frequency resonance is more significant (from 71.2% to 93.8%, an improvement of 22.6%). This is because the waveform distortion of low-frequency resonance is the most serious, and the spectral leakage problem of traditional FFT is particularly prominent in the low-frequency band. However, the discrimination method of this application based on phase space reconstruction is not affected by frequency shift. The recognition accuracy of fundamental frequency resonance is the highest (96.2%). Since it is closest to the power frequency, both methods perform well in this scenario, but this application still maintains an advantage of about 10%. The recognition accuracy of high-frequency resonance is improved from 76.4% to 94.0%, an improvement of 17.6%. This is due to the fact that the nonlinear feature quantity extracted by phase space reconstruction in this application can effectively characterize the dynamic characteristics of high-frequency complex oscillations.

[0082] In another exemplary embodiment, this application also provides a voltage transformer harmonic elimination control device based on phasor trajectory. The device includes: an acquisition and output module, used to continuously and synchronously acquire the instantaneous voltage on the secondary side and the instantaneous current on the primary side of the voltage transformer, remove the power frequency fundamental component, and output a pure characteristic voltage sequence and a pure characteristic current sequence; a reconstruction and inversion module, used to perform phase space reconstruction on the pure characteristic voltage sequence to output a resonance type label, construct a frequency band weighted observation matrix based on the resonance type label, and use the pure characteristic current component of the corresponding frequency band as the observation input for improved Kalman filtering to invert and output a time-varying inductance; a plotting module, used to substitute the inverted time-varying inductance into the equivalent circuit of the voltage transformer, back-deduce the voltage phasor corresponding to the current core state, and plot a dynamic reference trajectory that is synchronized with the core saturation fluctuation in real time; and a resonance elimination module, used to obtain a pure abnormal deviation by subtracting the measured phasor trajectory from the dynamic reference trajectory in real time, adaptively calculate the damping resistance value based on the pure abnormal deviation and the resonance type label to output a control signal, and drive the actuator to operate based on the control signal until the voltage transformer resonance is eliminated.

[0083] In another exemplary embodiment, this application also provides a computer storage medium storing computer instructions for executing the voltage transformer harmonic suppression control method as described in the preceding embodiments.

[0084] In another exemplary embodiment, this application also provides an electronic device, the electronic device comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the voltage transformer harmonic suppression control method as described in the preceding embodiments.

[0085] The above embodiments are only for illustrating the technical concept and features of this application, and are intended to enable those skilled in the art to understand the content of this application and implement it accordingly. They should not be construed as limiting the scope of protection of this application. All equivalent changes or modifications made in accordance with the spirit and essence of this application should be included within the scope of protection of this application.

Claims

1. A voltage transformer harmonic elimination control method based on phasor trajectories, characterized in that, The method includes: The instantaneous voltage on the secondary side and the instantaneous current on the primary side of the voltage transformer are continuously and synchronously acquired. After removing the power frequency fundamental component, the pure characteristic voltage sequence and the pure characteristic current sequence are output. The phase space of the pure characteristic voltage sequence is reconstructed to output the resonance type label. Based on the resonance type label, a frequency band weighted observation matrix is ​​constructed. The pure characteristic current component of the corresponding frequency band is used as the observation input of the improved Kalman filter to invert the time-varying inductance. Substitute the time-varying inductance output from the inversion into the equivalent circuit of the voltage transformer, deduce the voltage phasor corresponding to the current core state, and plot the dynamic reference trajectory that is synchronized with the core saturation fluctuation in real time point by point. The pure abnormal deviation is obtained by subtracting the measured phasor trajectory from the dynamic reference trajectory in real time. Based on the pure abnormal deviation and the resonance type label, the damping resistance value is adaptively calculated to output the control signal. The actuator is driven to operate based on the control signal until the voltage transformer resonance is eliminated.

2. The method according to claim 1, characterized in that, The phase space reconstruction of the pure characteristic voltage sequence to output a resonance type label includes: Determine the embedding dimension and delay time of the phase space; The pure feature voltage sequence is reconstructed based on the embedding dimension and delay time of the phase space to obtain the reconstructed attractor; Extract the nonlinear characteristic of the reconstructed attractor and determine the resonance type label based on the characteristic.

3. The method according to claim 2, characterized in that, The extraction of nonlinear characteristic quantities of the reconstructed attractor and the determination of resonance type labels based on these characteristic quantities include: Calculate the correlation dimension and the maximum Lyapunov exponent of the reconstruction attractor; The resonance type label is determined by comparing the correlation dimension and the maximum Lyapunov exponent with a set threshold.

4. The method according to claim 3, characterized in that, The comparison based on the correlation dimension and the maximum Lyapunov exponent with a set threshold, and the determination of the resonance type label based on the comparison result, includes: If the correlation coefficient is within the first preset range and the maximum Lyapunov exponent is less than zero, the resonance type is determined to be low-frequency resonance. If the correlation coefficient is in the second preset range and the maximum Lyapunov exponent is approximately zero, then the resonance type is determined to be fundamental frequency resonance. If the correlation coefficient is in the third preset interval and the maximum Lyapunov exponent is greater than zero, then the resonant type is determined to be high-frequency resonance.

5. The method according to claim 1, characterized in that, The process of constructing a frequency band weighted observation matrix based on resonance type labels, and using the pure characteristic current component of the corresponding frequency band as the observation input for the improved Kalman filter, to invert the time-varying inductance, includes: The pure characteristic current sequence is separated into current components of different frequency bands. The instantaneous energy of each current component and its proportion in the total energy are calculated. Based on the proportion, it is determined whether to activate the improved Kalman filter inversion. If it is determined to be activated, the next step is executed; otherwise, the inversion operation of this cycle is exited and the energy proportion is determined again in the next sampling cycle. Constructing a band-weighted observation matrix based on resonance type labels; The observation residuals are weighted based on the frequency band weighted observation matrix to invert the time-varying inductance.

6. The method according to claim 1, characterized in that, The process of substituting the time-varying inductance output from the inversion into the equivalent circuit of the voltage transformer, reversing the voltage phasor corresponding to the current core state, and plotting a dynamic reference trajectory that is synchronized with the core saturation fluctuation in real time point by point includes: Substitute the time-varying inductor into the equivalent circuit of the primary side of the voltage transformer to construct the time-varying equivalent impedance at the current moment; Calculate the magnitude of the time-varying equivalent impedance and obtain the normalized complex twitch factor based on the magnitude; Based on the normalized complex rotation factor and the rated phase voltage of the distribution network, the voltage phasor that the system should have for stable operation under the current core condition is obtained. A dynamic reference trajectory is obtained by plotting the voltage phasors point by point.

7. The method according to claim 1, characterized in that, The process of obtaining the pure abnormal deviation by subtracting the measured phasor trajectory from the dynamic reference trajectory in real time, adaptively calculating the damping resistance value based on the pure abnormal deviation and the resonance type label to output a control signal, and driving the actuator to operate based on the control signal until the voltage transformer resonance is eliminated includes: Obtain the measured phasor trajectory and subtract it from the dynamic reference trajectory to obtain the pure anomaly deviation. Calculate the magnitude of the pure abnormal deviation and perform trend prediction to obtain the deviation prediction value; The adaptive damping resistance value is calculated based on the deviation prediction value and the resonance type label; The adaptive damping resistor value is converted into a control signal to drive the actuator to engage the damping resistor in order to achieve harmonic elimination.

8. A voltage transformer harmonic elimination control device based on phasor trajectories, characterized in that, The device includes: The acquisition and output module is used to continuously and synchronously acquire the instantaneous voltage on the secondary side and the instantaneous current on the primary side of the voltage transformer, and after removing the power frequency fundamental wave component, output the pure characteristic voltage sequence and the pure characteristic current sequence. The reconstruction and inversion module is used to reconstruct the phase space of the pure characteristic voltage sequence to output the resonance type label, construct the frequency band weighted observation matrix based on the resonance type label, and use the pure characteristic current component of the corresponding frequency band as the observation input of the improved Kalman filter to invert the time-varying inductance. The plotting module is used to substitute the time-varying inductance output by the inversion into the equivalent circuit of the voltage transformer, reverse the voltage phasor corresponding to the current core state, and plot the dynamic reference trajectory that is synchronized with the core saturation fluctuation in real time point by point. The resonance elimination module is used to obtain the pure abnormal deviation by subtracting the measured phasor trajectory from the dynamic reference trajectory in real time. Based on the pure abnormal deviation and the resonance type label, the damping resistance value is adaptively calculated to output a control signal. The control signal drives the actuator to operate until the voltage transformer resonance is eliminated.

9. A computer storage medium, characterized in that, The computer storage medium stores computer instructions for performing the method as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, The electronic device includes: Memory, used to store computer programs; A processor for executing the computer program to implement the method as described in any one of claims 1 to 7.

Images