Data processing method for current transformer test

Abstract

The invention belongs to the technical field of mutual inductor verification, and particularly relates to a data processing method for testing a current mutual inductor, so as to solve the technical problem that an existing verification method cannot truly reflect the comprehensive error of the mutual inductor, and the method comprises the following steps: S1, obtaining the current waveform data of a tested current mutual inductor; s2, deducting the exponential decay aperiodic component from the original current waveform data to obtain corrected current waveform data; s3, reconstructing to obtain de-noised current waveform data; s4, acquiring fundamental waves of the primary side current and the secondary side current and setting amplitude and phase angles of subharmonics; s5, calculating an initial ratio error and an initial angle error of the current transformer; s6, generating a ratio error compensation vector and an angular difference compensation vector; and S7, respectively carrying out vector summation on the initial ratio error and the initial angle error with the ratio error compensation vector and the angle error compensation vector to obtain the final ratio error and the final angle error of the current transformer. And the test accuracy and reliability of the ratio difference and the angle difference of the current transformer are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of transformer calibration, and in particular relates to a data processing method for current transformer testing. Background Art

[0002] Current transformers are key devices used for measurement, protection, and monitoring in power systems. Their function is to convert high primary currents into lower secondary currents at a specified ratio for use by secondary equipment such as instrumentation and relay protection. The accuracy of current transformer measurement performance is directly related to the power grid's energy metering, fault diagnosis, and safe and stable operation. Therefore, regular and accurate on-site calibration of current transformers is crucial. Traditional current transformer calibration methods rely primarily on standard current transformers and transformer calibrators, using a comparison method. This method typically requires standard power-frequency sinusoidal steady-state conditions. The test system is bulky, requires a high-power current source, and is complex to wire and operate on-site. This is labor-intensive and inefficient. This method cannot effectively simulate and evaluate the dynamic performance of current transformers under complex operating conditions such as transient faults and harmonic pollution that may be encountered in real power grids. Consequently, the test results fail to fully reflect the transformer's true operating characteristics.

[0003] With the rapid development of digital signal processing and computer technology, current transformer testing methods based on digital sampling have emerged. This method acquires the instantaneous waveforms of the primary and secondary currents and analyzes them using algorithms to obtain the ratio and angle differences. However, existing digital processing methods still face numerous technical bottlenecks. First, during transient characteristic testing, the primary current often contains an exponentially decaying non-periodic component. If left unprocessed, this can severely interfere with subsequent spectrum analysis algorithms, such as the Fourier transform, resulting in spectrum leakage and affecting the accuracy of fundamental and subharmonic parameter extraction. Second, the acquired signal inevitably contains random noise, such as white noise and power frequency interference, which reduces the signal-to-noise ratio and affects measurement accuracy. Furthermore, the classic fast Fourier transform (FFT) algorithm suffers from fence effects and spectrum leakage. Its frequency resolution is limited by the sampling time and number of points, making it difficult to perform high-resolution and precise analysis of the signal spectrum, resulting in errors in amplitude and phase angle calculations. More critically, most methods only consider the fundamental component to calculate errors, ignoring the impact of the transformer core's nonlinear characteristics. In fact, under harmonic or transient high current conditions, each harmonic component will affect the main magnetic flux of the iron core, generating additional error components. Simply calculating the ratio difference and angle difference based on the fundamental wave cannot truly reflect the comprehensive error of the transformer. Therefore, how to accurately compensate for the nonlinear error caused by harmonics is the core technical difficulty in improving the accuracy of transformer testing. Summary of the Invention

[0004] The present invention provides a data processing method for current transformer testing, which solves the technical problem that the existing calibration method cannot truly reflect the comprehensive error of the transformer.

[0005] To solve the above problems, the present invention provides a method for processing data for current transformer testing, which adopts the following technical solution: a method for processing data for current transformer testing, comprising the following steps: S1, obtaining current waveform data of the current transformer under test in transient or steady-state conditions, where the current waveform data is discrete sampling results of the primary and secondary sides; S2, fitting the exponential decay non-periodic component in the original current waveform data using the least squares method, and then deducting the exponential decay non-periodic component from the original current waveform data to obtain corrected current waveform data; S3, using symlet wavelet to perform multi-layer decomposition on the corrected current waveform data, and then process the wavelet coefficients by thresholding based on Stein's unbiased risk estimation to reconstruct the denoised current waveform data; S4, using Chirp Z transform to perform spectrum analysis on the denoised current waveform data to obtain the amplitude and phase angle of the fundamental wave and the set subharmonic of the primary and secondary side currents; S5, calculating the initial ratio difference and initial angle difference of the current transformer according to the fundamental wave amplitude and phase angle of the primary and secondary side currents; S6, based on a preset harmonic flux compensation model, takes the fundamental amplitude of the secondary side current and the amplitudes and phase angles of the third, fifth, and seventh harmonics as input to generate a ratio difference compensation vector and an angle difference compensation vector; S7 , performing vector summation on the initial ratio difference and the initial angle difference, the ratio difference compensation vector and the angle difference compensation vector, respectively, to obtain final current transformer ratio difference and angle difference.

[0006] Furthermore, in S1 , the current signals of the primary side and the secondary side are synchronously collected at a set sampling frequency through a synchronous data acquisition card, and the collection lasts for several power frequency cycles.

[0007] Furthermore, S2 includes the following steps: Create original current waveform data The mathematical model is expressed as: ; in, 、 、 is the sine amplitude, angular frequency, and initial phase of the sine component; 、 is the initial amplitude and time constant of the exponentially decaying non-periodic component; For the The moment of each sampling point; The parameters are obtained by iterative solution using nonlinear least squares method. and The optimal estimated value of , which minimizes the sum of squares of the current waveform data output by the model and the actual current waveform data; From the original current waveform data The exponential decay non-periodic component obtained after deducting the fitting The corrected current waveform data is obtained.

[0008] Furthermore, the wavelet decomposition step uses the sym8 wavelet as the wavelet basis function to perform a 5-layer decomposition on the corrected current waveform data to obtain the approximate coefficients and detail coefficients of each layer.

[0009] Furthermore, threshold processing and reconstruction include the following steps: For the detail coefficients of layers 1 to 5, the optimal threshold of each layer is adaptively calculated based on the Stein unbiased risk estimation criterion; The soft threshold function is used to denoise the detail coefficients of each layer; The processed detail coefficients of each layer and the approximate coefficients of the 5th layer are reconstructed by wavelet to obtain the denoised waveform data.

[0010] Furthermore, Chirp Z transform is used to perform high-resolution spectrum analysis, including the following steps: Nominal frequency for fundamental wave and each harmonic , respectively set to A narrowband analysis frequency range centered on and with a preset width; In the narrowband analysis frequency range, Chirp Z transform is used to perform spectrum refinement calculation to obtain a high-resolution discrete spectrum; In the discrete spectrum, the amplitude and phase angle of the fundamental wave and the set subharmonic are solved by performing quadratic interpolation on the discrete spectrum peak point and its adjacent points.

[0011] Furthermore, in S5, the initial ratio difference Calculated using the following formula: ; in, is the nominal change, 、 are the fundamental amplitudes of the primary and secondary currents respectively; Initial angle difference Calculated using the following formula: ; in, 、 are the fundamental phase angles of the primary and secondary currents respectively.

[0012] Furthermore, according to the preset harmonic flux compensation model, the ratio difference compensation vector and the angle difference compensation vector are calculated and generated, including the following steps: the fundamental amplitude of the secondary side current is converted to , and the amplitude and phase angle of the 3rd, 5th and 7th harmonics as input, the ratio difference compensation is calculated by the following formula and angular difference compensation : ; ; in, is the fundamental amplitude of the secondary side current; 、 、 are the amplitudes of the 3rd, 5th, and 7th harmonics on the secondary side respectively; 、 、 are the phase angles of the 3rd, 5th, and 7th harmonics on the secondary side relative to the fundamental wave; 、 、 They are respectively the proportional coefficients of the 3rd, 5th and 7th harmonic contrast difference compensation preset in the harmonic flux compensation model; 、 、 They are respectively the proportional coefficients of the 3rd, 5th and 7th harmonic diagonal difference compensation preset in the harmonic flux compensation model.

[0013] Furthermore, the coefficients in the harmonic flux compensation model are calibrated by the following steps: Apply a standard current containing fundamental wave and set harmonic components to the current transformer under test, and synchronously measure the actual ratio difference and angle difference generated by it using a standard instrument; The multiple linear regression algorithm is used to fit the amplitude and phase angle of each harmonic on the secondary side of the measured current transformer with the actual error measured by the standard instrument, and the coefficient that minimizes the difference between the compensation amount calculated by the harmonic flux compensation model and the actual error is obtained.

[0014] Furthermore, the calculated initial ratio difference is used as the real part, and the initial angle difference is converted into radians as the imaginary part to form an initial error complex vector; The ratio difference compensation vector is used as the real part, and the angle difference compensation vector is converted into radians as the imaginary part, and added to the real part and imaginary part of the initial error complex vector respectively, and the complex vector summation operation is performed to obtain the final error vector; The real part of the error vector is the final ratio difference after compensation, and the imaginary part is the final angle difference after compensation.

[0015] The beneficial effects are as follows: compared with the existing technology, the data processing method proposed in the present invention removes the non-periodic components through the least squares method and combines wavelet and Stein unbiased risk estimation threshold for denoising, thereby purifying the original sampling signal and laying a high-precision foundation for subsequent analysis; secondly, the Chirp Z transform is used to replace the traditional FFT to achieve high-resolution extraction of fundamental and harmonic parameters, overcoming the calculation deviation caused by spectrum leakage and fence effect. The present invention establishes a harmonic flux compensation model, and uses the secondary side harmonics as the key input to compensate for the error introduced by the nonlinearity of the core, solving the problem that the existing technology cannot accurately evaluate the performance of the transformer under harmonic or transient conditions by relying solely on fundamental wave analysis. This method improves the test accuracy and reliability of the ratio difference and angle difference of the current transformer through precise signal processing and accurate error compensation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Flowchart of the data processing method for current transformer testing. DETAILED DESCRIPTION

[0017] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Those skilled in the art should know that the embodiments described below are part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0018] An embodiment of the data processing method for current transformer testing provided by the present invention: like Figure 1 As shown, the data processing method for current transformer testing includes the following steps: S1, obtaining current waveform data of the current transformer under test in transient or steady-state conditions, where the current waveform data is discrete sampling results of the primary and secondary sides.

[0019] In an optional embodiment, the current signals of the primary and secondary sides are synchronously collected at a set sampling frequency (e.g., 10 kHz) by a synchronous data acquisition card for several power frequency cycles to ensure that the acquired current waveform data is complete and the phase relationship is accurate.

[0020] S2, using the least square method to fit the exponential decay non-periodic component in the original current waveform data, and then deducting the exponential decay non-periodic component from the original current waveform data to obtain corrected current waveform data.

[0021] In an optional embodiment, S2 includes the following steps: Create original current waveform data , The mathematical model is expressed as: ; in, 、 、 is the sine amplitude, angular frequency, and initial phase of the sine component; 、 is the initial amplitude and time constant of the exponentially decaying non-periodic component; For the The moment of each sampling point; The parameters are obtained by iterative solution using nonlinear least squares method. and The optimal estimated value of , which minimizes the sum of squares of the current waveform data output by the model and the actual current waveform data; From the original current waveform data The exponential decay non-periodic component obtained after deducting the fitting The corrected current waveform data is obtained.

[0022] Specifically, when a short circuit occurs in the power system, the fault current often contains a DC component that decays exponentially. This DC component interferes with the subsequent Fourier analysis, resulting in inaccurate harmonic measurement. A mathematical model is established that can simultaneously describe the AC sinusoidal component and the DC decay component. For example, a sampling point Current value It is modeled as the sum of an ideal sine wave and an exponential decay term.

[0023] Through the nonlinear least squares method, a series of current data collected are 、 ···、 Fitting with mathematical model. Nonlinear least squares method continuously adjusts parameters through iterative calculation. (initial amplitude) and parameters (time constant), the goal is to minimize the sum of squares of the errors between the model calculation value and the actual measurement value at all sampling points. For example, the optimal initial amplitude may be obtained through calculation Equal to 50A, time constant Based on these optimal parameters, the value of the exponentially decaying non-periodic component at each sampling moment is calculated and subtracted from the original current waveform data to obtain a pure AC waveform data, laying the foundation for subsequent accurate wavelet analysis and spectrum analysis.

[0024] In S3, the symlet wavelet is used to perform multi-layer decomposition on the corrected current waveform data, and the wavelet coefficients are processed by thresholding based on the Stein unbiased risk estimation to obtain the denoised current waveform data after reconstruction.

[0025] In an optional embodiment, the wavelet decomposition step uses the sym8 wavelet as the wavelet basis function to perform a 5-layer decomposition on the corrected current waveform data to obtain approximate coefficients and detail coefficients at each layer.

[0026] Specifically, the sym8 wavelet basis function was chosen due to its good symmetry and tight support, resulting in minimal phase distortion and error when decomposing and reconstructing signals, making it suitable for processing transient signals in power systems. Wavelet decomposition of the current waveform data after subtracting the non-periodic components actually involves passing the current waveform data through a series of high-pass and low-pass filters. The low-pass filter obtains the approximate coefficient A, and the high-pass filter obtains the detail coefficient D.

[0027] The 5-layer decomposition of the current waveform data is essentially to separate the current waveform data into frequency bands of different frequencies step by step. Specifically, if the sampling frequency of the original current waveform data is 12.8kHz, the detail coefficient D1 obtained by the first-layer decomposition corresponds to the highest frequency noise of 3.2kHz-6.4kHz. The detail coefficient D2 obtained by the second-layer decomposition corresponds to the highest frequency noise of 1.6kHz-3.2kHz. Similarly, the detail coefficient D5 of the fifth layer corresponds to the highest frequency noise of 200Hz-400Hz. The approximate coefficient A5 of the fifth layer corresponds to the main low-frequency component of 0-200Hz, which contains key information about the fundamental wave and harmonics. Through this layered approach, the noise of different frequency bands is clearly separated, which facilitates the subsequent refined noise processing for each frequency band.

[0028] In an optional embodiment, the threshold processing and reconstruction includes the following steps: For the detail coefficients of layers 1 to 5, the optimal threshold of each layer is adaptively calculated based on the Stein Unbiased Risk Estimation (SURE) criterion; The soft threshold function is used to denoise the detail coefficients of each layer; The processed detail coefficients of each layer and the approximate coefficients of the 5th layer are reconstructed by wavelet to obtain the denoised waveform data.

[0029] Specifically, to effectively remove noise while maximally preserving useful signals, an adaptive threshold determination method is employed. The Stan Unbiased Risk Estimation (SURE) criterion automatically calculates the optimal threshold based on the data characteristics of each layer's detail coefficients, without requiring prior knowledge of the noise intensity. For example, for the first layer's detail coefficients, which contain a significant amount of Gaussian white noise, the SURE criterion will calculate a higher optimal threshold (e.g., 0.9); for the fourth layer's detail coefficients, which have less noise, a lower optimal threshold (e.g., 0.15) will be obtained.

[0030] After determining the optimal threshold for each layer, a soft thresholding function is applied. This function sets coefficients with absolute values ​​less than the optimal threshold to zero and shrinks coefficients with absolute values ​​greater than the optimal threshold toward zero by a threshold value. Compared to the hard thresholding method, which simply sets the value to zero, this method smoothes the reconstructed signal and avoids excessive oscillations. The thresholded detail coefficients of all layers (D1 to D5) are combined with the unprocessed approximation coefficients of the fifth layer, A5, which contain the main signal information. The final, clean current waveform data is reconstructed using an inverse wavelet transform.

[0031] S4, using Chirp Z transform to perform spectrum analysis on the denoised current waveform data to obtain the amplitude and phase angle of the fundamental wave and the set subharmonic of the primary and secondary side currents.

[0032] In an optional embodiment, high-resolution spectrum analysis is performed using Chirp Z transform, including the following steps: Nominal frequency for fundamental wave and each harmonic , respectively set to A narrowband analysis frequency range centered on and with a preset width; In the narrowband analysis frequency range, Chirp Z transform is used to perform spectrum refinement calculation to obtain a high-resolution discrete spectrum; In the discrete spectrum, the amplitude and phase angle of the fundamental wave and the set subharmonic are solved by performing quadratic interpolation on the discrete spectrum peak point and its adjacent points.

[0033] Specifically, the frequency resolution of conventional Fourier transforms (FFTs) is limited by the sampling duration and number of points, making it impossible to accurately capture harmonics that deviate from the nominal frequency. The Chirp Z transform (CZT) is used to achieve local amplification of the spectrum. For example, for a nominal 50Hz fundamental wave, a narrowband analysis frequency range of 49Hz to 51Hz can be set. The CZT can calculate 1,000 spectrum points within this narrowband analysis frequency range of only 2Hz, achieving a frequency resolution of 0.002Hz. This is much higher than the resolution that FFT can provide, making it possible to accurately identify fundamental frequency offsets caused by power grid fluctuations.

[0034] Even after CZT refinement, the true spectrum peak may fall between two calculated frequency points. To further improve accuracy, quadratic interpolation is used. After finding a peak point in the high-resolution spectrum calculated by CZT, the peak point and its two adjacent points on its left and right are taken, for a total of three points. A parabola is fitted from these three points, and the frequency, amplitude, and phase angle corresponding to the vertex of the parabola are considered to be the most accurate estimate of the harmonic component. For example, if the CZT spectrum peaks at 50.014Hz, quadratic interpolation can correct it to a more accurate frequency of 50.0146Hz and the corresponding amplitude and phase angle at 50.0146Hz.

[0035] S5, calculating the initial ratio difference and initial angle difference of the current transformer according to the fundamental wave amplitude and phase angle of the primary and secondary side currents.

[0036] In an optional embodiment, the initial ratio difference Calculated using the following formula: ; in, is the nominal change, 、 are the fundamental amplitudes of the primary and secondary currents respectively.

[0037] Initial angle difference Calculated using the following formula: ; in, 、 are the fundamental phase angles of the primary and secondary currents respectively.

[0038] S6, based on the preset harmonic flux compensation model, takes the fundamental amplitude of the secondary side current and the amplitudes and phase angles of the 3rd, 5th, and 7th harmonics as input to generate a ratio difference compensation vector and an angle difference compensation vector.

[0039] In an optional embodiment, according to a preset harmonic flux compensation model, the ratio difference compensation vector and the angle difference compensation vector are calculated and generated, including the following steps: the fundamental amplitude of the secondary side current is converted to , and the amplitude and phase angle of the 3rd, 5th and 7th harmonics as input, the ratio difference compensation is calculated by the following formula and angular difference compensation : ; ; in, is the fundamental amplitude of the secondary side current; 、 、 are the amplitudes of the 3rd, 5th, and 7th harmonics on the secondary side respectively; 、 、 are the phase angles of the 3rd, 5th, and 7th harmonics on the secondary side relative to the fundamental wave; 、 、 They are respectively the proportional coefficients of the 3rd, 5th and 7th harmonic contrast difference compensation preset in the harmonic flux compensation model; 、 、 They are respectively the proportional coefficients of the 3rd, 5th and 7th harmonic diagonal difference compensation preset in the harmonic flux compensation model.

[0040] Specifically, the measurement error of current transformers is primarily caused by the nonlinear magnetization characteristics of the core. The presence of harmonic components exacerbates this nonlinearity, leading to increased error. The harmonic flux compensation model aims to quantify the additional error introduced by major harmonic components such as the third, fifth, and seventh order. The harmonic flux compensation model decomposes the impact of these harmonic components on the total flux into two components: one component is in phase with the fundamental flux and corresponds to the cause of ratio error; the other component is orthogonal to the fundamental flux and corresponds to the cause of angle error.

[0041] For example, the secondary side fundamental current is obtained through spectrum analysis 5A, 3rd harmonic is 0.4A and its phase angle relative to the fundamental wave At the same time, assuming the preset compensation coefficient is 0.01, is 0.008. Then, considering only the third harmonic, the ratio difference compensation component it produces is: ; The angular difference compensation component is: .

[0042] The total ratio difference compensation can be obtained by linearly superimposing the compensation components generated by the 3rd, 5th and 7th harmonics. and angular difference compensation , used to correct the measurement results of the mutual inductor.

[0043] In an optional embodiment, the coefficients in the harmonic flux compensation model are calibrated by the following steps: Apply a standard current containing fundamental wave and set harmonic components to the current transformer under test, and synchronously measure the actual ratio difference and angle difference generated by it using a standard instrument; The multiple linear regression algorithm is used to fit the amplitude and phase angle of each harmonic on the secondary side of the measured current transformer with the actual error measured by the standard instrument, and the coefficient that minimizes the difference between the compensation amount calculated by the harmonic flux compensation model and the actual error is obtained.

[0044] Specifically, the coefficients in the harmonic flux compensation model are 、 The current transformer is specific to each instrument and needs to be calibrated experimentally. During the calibration process, a programmable high-power current source is used to inject a precisely known current signal into the current transformer under test. For example, a synthetic current consisting of a 50Hz fundamental wave superimposed with 10% of the third harmonic and 5% of the fifth harmonic can be injected. At the same time, a standard power quality analyzer or transformer calibrator will accurately measure the actual ratio difference and angle difference generated by the current transformer under this working condition. For example, the ratio difference is measured to be 0.1% and the angle difference is 5. ′ .

[0045] By changing the amplitude and phase of the injected harmonics and performing multiple measurements, a rich set of experimental data is obtained. Each set of data contains the amplitude and phase angle information of each input harmonic and the corresponding measured error. Using this data as input, a multivariate linear regression algorithm is applied. The goal of the multivariate linear regression algorithm is to solve a set of optimal coefficients. 、 、 、 、 、 , so that the theoretical error calculated by substituting the harmonic data into the compensation formula minimizes the overall deviation from the error values ​​measured at all experimental points. This set of calculated coefficients is then fixed into the measurement device for online real-time compensation.

[0046] S7 , performing vector summation on the initial ratio difference and the initial angle difference, the ratio difference compensation vector and the angle difference compensation vector, respectively, to obtain final current transformer ratio difference and angle difference.

[0047] In an optional embodiment, the calculated initial ratio difference is used as the real part, and the initial angular difference is converted into radians as the imaginary part to form an initial complex error vector. The ratio difference compensation vector is used as the real part, and the angular difference compensation vector is converted into radians as the imaginary part. These are added to the real and imaginary parts of the initial complex error vector, respectively, to perform a complex vector summation operation to obtain a final error vector. The real part of the error vector is the final ratio difference after compensation, and the imaginary part is the final angular difference after compensation.

[0048] In addition, in the description of this specification, “a plurality of” means at least two, for example, two, three or more, etc., unless otherwise clearly and specifically defined.

Claims

1. A data processing method for current transformer testing, characterized in that: The following steps are involved: S1, obtaining current waveform data of the current transformer under test in transient or steady-state conditions, where the current waveform data is discrete sampling results of the primary and secondary sides; S2, fitting the exponential decay non-periodic component in the original current waveform data using the least squares method, and then deducting the exponential decay non-periodic component from the original current waveform data to obtain corrected current waveform data; S3, using symlet wavelet to perform multi-layer decomposition on the corrected current waveform data, and then process the wavelet coefficients by thresholding based on Stein's unbiased risk estimation to reconstruct the denoised current waveform data; S4, using Chirp Z transform to perform spectrum analysis on the denoised current waveform data to obtain the amplitude and phase angle of the fundamental wave and the set subharmonic of the primary and secondary side currents; S5, calculating the initial ratio difference and initial angle difference of the current transformer according to the fundamental wave amplitude and phase angle of the primary and secondary side currents; S6, based on a preset harmonic flux compensation model, takes the fundamental amplitude of the secondary side current and the amplitudes and phase angles of the third, fifth, and seventh harmonics as input to generate a ratio difference compensation vector and an angle difference compensation vector; S7 , performing vector summation on the initial ratio difference and the initial angle difference, the ratio difference compensation vector and the angle difference compensation vector, respectively, to obtain final current transformer ratio difference and angle difference.

2. The data processing method for current transformer testing according to claim 1, characterized in that: In S1, the current signals of the primary and secondary sides are synchronously collected at the set sampling frequency through the synchronous data acquisition card and last for several power frequency cycles.

3. The data processing method for current transformer testing according to claim 1, characterized in that: S2 includes the following steps: Create original current waveform data The mathematical model is expressed as: ; in, 、 、 is the sine amplitude, angular frequency, and initial phase of the sine component; 、 is the initial amplitude and time constant of the exponentially decaying non-periodic component; For the The moment of each sampling point; The parameters are obtained by iterative solution using nonlinear least squares method. and The optimal estimated value of , which minimizes the sum of squares of the current waveform data output by the model and the actual current waveform data; From the original current waveform data The exponential decay non-periodic component obtained after deducting the fitting The corrected current waveform data is obtained.

4. The data processing method for current transformer testing according to claim 1, characterized in that: The wavelet decomposition step uses sym8 wavelet as the wavelet basis function to perform a 5-layer decomposition on the corrected current waveform data to obtain the approximate coefficients and detail coefficients of each layer.

5. The data processing method for current transformer testing according to claim 4, characterized in that: Thresholding and reconstruction includes the following steps: For the detail coefficients of layers 1 to 5, the optimal threshold of each layer is adaptively calculated based on the Stein unbiased risk estimation criterion; The soft threshold function is used to denoise the detail coefficients of each layer; The processed detail coefficients of each layer and the approximate coefficients of the 5th layer are reconstructed by wavelet to obtain the denoised waveform data.

6. The data processing method for current transformer testing according to claim 1, characterized in that: Using Chirp Z transform to perform high-resolution spectrum analysis includes the following steps: Nominal frequency for fundamental wave and each harmonic , respectively set to A narrowband analysis frequency range centered on and with a preset width; In the narrowband analysis frequency range, Chirp Z transform is used to perform spectrum refinement calculation to obtain a high-resolution discrete spectrum; In the discrete spectrum, the amplitude and phase angle of the fundamental wave and the set subharmonic are solved by performing quadratic interpolation on the discrete spectrum peak point and its adjacent points.

7. The data processing method for current transformer testing according to claim 1, characterized in that: In S5, the initial ratio difference Calculated using the following formula: ; in, is the nominal change, 、 are the fundamental amplitudes of the primary and secondary currents respectively; Initial angle difference Calculated using the following formula: ; in, 、 are the fundamental phase angles of the primary and secondary currents respectively.

8. The data processing method for current transformer testing according to claim 1, characterized in that: According to the preset harmonic flux compensation model, the ratio difference compensation vector and the angle difference compensation vector are calculated and generated, including the following steps: the fundamental amplitude of the secondary side current is converted to , and the amplitude and phase angle of the 3rd, 5th and 7th harmonics as input, the ratio difference compensation is calculated by the following formula and angular difference compensation : ; ; in, is the fundamental amplitude of the secondary side current; 、 、 are the amplitudes of the 3rd, 5th, and 7th harmonics on the secondary side respectively; 、 、 are the phase angles of the 3rd, 5th, and 7th harmonics on the secondary side relative to the fundamental wave; 、 、 They are respectively the proportional coefficients of the 3rd, 5th and 7th harmonic contrast difference compensation preset in the harmonic flux compensation model; 、 、 They are respectively the proportional coefficients of the 3rd, 5th and 7th harmonic diagonal difference compensation preset in the harmonic flux compensation model.

9. The data processing method for current transformer testing according to claim 1, characterized in that: The coefficients in the harmonic flux compensation model are calibrated by the following steps: Apply a standard current containing fundamental wave and set harmonic components to the current transformer under test, and synchronously measure the actual ratio difference and angle difference generated by it using a standard instrument; The multiple linear regression algorithm is used to fit the amplitude and phase angle of each harmonic on the secondary side of the measured current transformer with the actual error measured by the standard instrument, and the coefficient that minimizes the difference between the compensation amount calculated by the harmonic flux compensation model and the actual error is obtained.

10. The data processing method for current transformer testing according to any one of claims 1 to 9, characterized in that: The calculated initial ratio difference is used as the real part, and the initial angle difference converted into radians is used as the imaginary part to form an initial error complex vector; The ratio difference compensation vector is used as the real part, and the angle difference compensation vector is converted into radians as the imaginary part, and added to the real part and imaginary part of the initial error complex vector respectively, and the complex vector summation operation is performed to obtain the final error vector; The real part of the error vector is the final ratio difference after compensation, and the imaginary part is the final angle difference after compensation.

Images