A high-precision performance test processing method and system for low-impedance voltage transformer

Abstract

The present invention relates to the field of electrical variable testing technology, and in particular to a high-precision performance test processing method and system for a low-impedance voltage transformer. The method comprises the following steps: constructing a controllable test excitation signal and synchronously collecting the corresponding transformer primary voltage and transformer secondary voltage; performing spatial electromagnetic elimination based on the transformer primary voltage and the transformer secondary voltage, while performing frequency domain feature decomposition and signal instantaneous offset calculation to construct a transformer signal feature matrix; obtaining environmental parameters corresponding to the low-impedance voltage transformer; constructing a neural network model based on Bayesian optimization and predicting and outputting a performance parameter set corresponding to frequency response and load characteristics; generating a calibration compensation coefficient by comparing the performance parameter set with a preset standard threshold, and performing test evaluation to generate a high-precision performance test report corresponding to the low-impedance voltage transformer. The present invention can achieve high-precision performance testing of low-impedance voltage transformers.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical variable testing, and in particular to a high-precision performance testing processing method and system for a low-impedance voltage transformer. Background Art

[0002] Low-impedance voltage transformers (VTs) are widely used in power systems to convert high-voltage signals into standard low-voltage signals for use in measurement, protection, and control systems. Due to the high-voltage environment and importance of power systems, the accuracy and stability of low-impedance voltage transformers are crucial to the safe operation of power equipment. Therefore, high-precision performance testing of low-impedance voltage transformers is crucial to ensuring the normal operation of power systems. However, traditional low-impedance voltage transformer testing methods typically rely on static test equipment, performing simple measurements by comparing voltage signals and output signals to evaluate their operating performance. While these methods can reflect the performance of the voltage transformer to a certain extent, they are subject to significant errors due to factors such as the test environment, equipment accuracy, and manual operation. These errors can lead to performance deviations of the voltage transformer in actual applications, thereby reducing the efficiency of voltage transformer performance testing. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide a high-precision performance test processing method and system for a low-impedance voltage transformer to solve at least one of the above technical problems.

[0004] To achieve the above object, a high-precision performance test processing method for a low-impedance voltage transformer is provided, comprising the following steps:

[0005] Step S1: Generate a corresponding controllable test excitation signal through a tunable standard voltage source and a wide-band load network, and apply the controllable test excitation signal to a low-impedance voltage transformer to synchronously collect the corresponding transformer primary voltage and transformer secondary voltage; perform spatial electromagnetic cancellation based on the transformer primary voltage and transformer secondary voltage to generate the transformer original test signal;

[0006] Step S2: performing frequency domain feature decomposition on the original test signal of the mutual inductor to obtain frequency domain features of the mutual inductor signal corresponding to amplitude error, phase error, and harmonic distortion; performing instantaneous offset calculation on the original test signal of the mutual inductor to obtain instantaneous offset of the mutual inductor signal; constructing a mutual inductor signal feature matrix based on the frequency domain features of the mutual inductor signal and the instantaneous offset of the mutual inductor signal;

[0007] Step S3: Obtaining environmental parameters corresponding to the low-impedance voltage transformer; constructing a neural network model based on Bayesian optimization, and inputting the transformer signal feature matrix and environmental parameters to train a corresponding transformer performance joint prediction model to predict and output a performance parameter set corresponding to the frequency response and load characteristics;

[0008] Step S4: Generate a calibration compensation coefficient by comparing the performance parameter set with the preset standard threshold value, and inject the calibration compensation coefficient into the transformer performance joint prediction model to iteratively optimize the test until the error converges and test and evaluate, and generate a high-precision performance test report corresponding to the low-impedance voltage transformer.

[0009] Furthermore, step S1 includes the following steps:

[0010] Step S11: Generate a sinusoidal excitation signal with an adjustable fundamental frequency by using a tunable standard voltage source to cover the rated operating frequency band corresponding to the low-impedance voltage transformer;

[0011] Step S12: Building a broadband load network including resistive, inductive, and capacitive components, and configuring a load impedance adjustment for a sinusoidal excitation signal corresponding to an adjustable fundamental frequency based on the broadband load network to generate a corresponding controllable test excitation signal;

[0012] Step S13: applying a controllable test excitation signal to the low-impedance voltage transformer, and synchronously collecting the corresponding transformer primary side voltage and transformer secondary side voltage using a symmetrical Rogowski coil and a high-precision resistor divider;

[0013] Step S14: Design a differential sampling and common-mode suppression network circuit based on a symmetrical Rogowski coil and a high-precision resistor divider to perform spatial electromagnetic elimination on the primary side voltage and the secondary side voltage of the transformer to generate an original transformer test signal.

[0014] Furthermore, step S14 includes the following steps:

[0015] Step S141: simulating the magnetic field distribution generated by the primary side voltage of the transformer and the corresponding coupling condition of the symmetrical Rogowski coil according to the law of electromagnetic induction for the symmetrical Rogowski coil, and calculating the mutual inductance coefficient of the symmetrical Rogowski coil in all directions to generate a spatial distribution field of the mutual inductance coefficient of the Rogowski coil;

[0016] Step S142: calculating the input impedance and output impedance of the high-precision resistor divider at different frequencies based on the high-precision resistor divider and taking into account the parasitic parameters of the resistor elements and the frequency variation, and constructing a corresponding resistor divider impedance frequency response matrix according to the different frequencies;

[0017] Step S143: Decomposing the interference electromagnetic field corresponding to the low-impedance voltage transformer into common-mode interference and differential-mode interference, and determining the distribution of the interference electromagnetic field in the space surrounding the low-impedance voltage transformer to generate an electromagnetic interference modal decomposition vector field, wherein the vector of each spatial point in the field includes the magnitude and direction of the common-mode interference and the differential-mode interference;

[0018] Step S144: Designing a differential sampling and common-mode suppression network circuit based on the spatial distribution field of the Rogowski coil mutual inductance coefficient and the impedance frequency response matrix of the resistor divider in combination with the electromagnetic interference modal decomposition vector field, thereby optimizing the symmetrical Rogowski coil layout by utilizing the symmetrical structure corresponding to the symmetrical Rogowski coil to suppress common-mode interference and differential-mode interference, selecting appropriate impedance parameters based on the impedance frequency response corresponding to the high-precision resistor divider to match the differential sampling, and designing a corresponding filter circuit for the electromagnetic interference mode;

[0019] Step S145: Connect the differential sampling and common-mode suppression network circuit to the sampling loops corresponding to the primary-side voltage and the secondary-side voltage of the transformer to perform spatial electromagnetic elimination on their output signals, so as to extract the effective signal components corresponding to the primary-side voltage and the secondary-side voltage of the transformer through differential sampling, and use the common-mode suppression network to suppress the common-mode noise and differential-mode noise generated by spatial electromagnetic interference. At the same time, the processed signal is filtered and amplified and modulated to remove residual high-frequency noise and low-frequency drift, and generate the original transformer test signal, which includes the primary-side signal and the secondary-side signal.

[0020] Furthermore, step S2 includes the following steps:

[0021] Step S21: performing spectrum transformation and decomposition on the original test signal of the mutual inductor to generate amplitude distribution and phase distribution corresponding to the fundamental wave and each harmonic component in the primary and secondary side signals;

[0022] Step S22: performing amplitude and phase error analysis on the amplitude distribution and phase distribution corresponding to the fundamental wave components in the primary and secondary side signals to obtain frequency domain characteristics of the mutual inductor signal corresponding to the amplitude error and phase error;

[0023] Step S23: performing harmonic distortion evaluation analysis on the amplitude distribution and phase distribution corresponding to each harmonic component in the primary and secondary side signals to obtain the frequency domain characteristics of the mutual inductor signal corresponding to the harmonic distortion;

[0024] Step S24: calculating the instantaneous offset of the original test signal of the transformer to obtain the instantaneous offset of the transformer signal;

[0025] Step S25: constructing a mutual inductor signal characteristic matrix according to the mutual inductor signal frequency domain characteristics and the mutual inductor signal instantaneous offset.

[0026] Furthermore, step S22 includes the following steps:

[0027] Performing time-frequency synchronization processing on the amplitude distribution and phase distribution corresponding to the fundamental wave components in the primary and secondary side signals to obtain the amplitude distribution and phase distribution corresponding to the primary and secondary side signal components in the same time-frequency range;

[0028] Amplitude error statistics are performed between the amplitude distributions corresponding to the primary and secondary side signal components in the same time-frequency range to obtain the frequency domain characteristics of the mutual inductor signal corresponding to the amplitude error between the primary and secondary side signals;

[0029] The phase error is calculated based on the phase distributions corresponding to the primary and secondary side signal components in the same time-frequency range to obtain the frequency domain characteristics of the mutual inductor signal corresponding to the phase error between the primary and secondary side signals.

[0030] Furthermore, step S23 includes the following steps:

[0031] Step S231: performing harmonic proportion statistics calculation based on the amplitude distribution and phase distribution corresponding to each harmonic component in the primary and secondary side signals to obtain the proportion between each harmonic component of the primary and secondary side signals;

[0032] Step S232: constructing a vector field for each harmonic component vector based on the coupling relationship between adjacent harmonic component vectors by treating the amplitude distribution and phase distribution corresponding to each harmonic component in the primary and secondary side signals as harmonic component vectors having corresponding amplitude and phase attributes, and calculating the coupling relationship between adjacent harmonic component vectors, thereby generating harmonic component vector fields corresponding to the primary and secondary side signals;

[0033] Step S233: Calculate the harmonic energy proportion and harmonic energy transfer path of each harmonic component in the entire signal cycle according to the harmonic component vector field corresponding to the primary and secondary side signals;

[0034] Step S234: quantifying the amplitude-phase coupling distortion rate of the amplitude distribution and phase distribution corresponding to each harmonic component in the primary and secondary side signals based on the harmonic energy proportion corresponding to each harmonic component in the entire signal cycle and the harmonic energy transfer path, to obtain the total distortion rate of the harmonic components corresponding to the primary and secondary side signals;

[0035] Step S235: performing harmonic distortion evaluation analysis based on the proportion of each harmonic component of the primary and secondary side signals and the corresponding total distortion rate of the harmonic components to obtain the frequency domain characteristics of the mutual inductor signal corresponding to the harmonic distortion.

[0036] Furthermore, step S24 includes the following steps:

[0037] Step S241: obtaining the mutual inductor signal distribution corresponding to each spatial point through the original test signal of the mutual inductor, and calculating the signal distribution curvature corresponding to each spatial point according to the mutual inductor signal distribution corresponding to each spatial point;

[0038] Step S242: constructing a curvature tensor field of the transformer original test signal based on the corresponding signal distribution curvature at each spatial point to generate a transformer test signal curvature tensor field;

[0039] Step S243: determining a signal curvature mutation region of the original transformer test signal based on the transformer test signal curvature tensor field to obtain a corresponding signal mutation region frame on the transformer test signal;

[0040] Step S244: Calculate the instantaneous offset of the signal according to the signal mutation region frame corresponding to the mutual inductor test signal to obtain the instantaneous offset of the mutual inductor signal.

[0041] Furthermore, step S244 includes the following steps:

[0042] Calculate the signal energy flow density distribution corresponding to the mutual inductor test signal at each time point through the signal mutation region frame corresponding to the mutual inductor test signal;

[0043] Performing energy divergence calculation according to the signal energy flow density distribution corresponding to the mutual inductor test signal at each time point to obtain the signal energy divergence value corresponding to the mutual inductor test signal at each time point;

[0044] Determine the instantaneous offset position of the signal on the signal mutation region frame corresponding to the mutual inductor test signal based on the signal energy divergence value corresponding to each time point of the mutual inductor test signal, so as to calibrate the position of the drastic change of the signal energy flow determined by the signal energy divergence value as the instantaneous offset position of the corresponding signal;

[0045] Based on the instantaneous offset position of the corresponding signal, the signal instantaneous offset between the primary side and secondary side signal distributions at the same position on the original test signal of the mutual inductor is calculated to obtain the instantaneous offset of the mutual inductor signal.

[0046] Furthermore, step S3 includes the following steps:

[0047] Step S31: Acquire environmental parameters corresponding to the low-impedance voltage transformer, including temperature, humidity, and vibration frequency;

[0048] Step S32: by designing a three-layer fully connected neural network, including an input layer with 25 dimensions, a hidden layer with 128 nodes, and an output layer with 2 dimensions;

[0049] Step S33: By using the Bayesian optimization algorithm to search for the optimal hyperparameters corresponding to the three-layer fully connected neural network, including the learning rate, regularization coefficient and the number of hidden layer nodes, and constructing the neural network model based on Bayesian optimization with the root mean square error as the optimization target, the mutual inductor signal feature matrix and environmental parameters are input at the same time and the corresponding mutual inductor performance joint prediction model is obtained through 5-fold cross-validation training. The attention mechanism is also introduced to give higher weights to the performance parameters corresponding to the frequency response and load characteristics, so that the output layer predicts and outputs a set of performance parameters corresponding to the frequency response and load characteristics.

[0050] Furthermore, the present invention also provides a high-precision performance test processing system for a low-impedance voltage transformer, which is used to execute the high-precision performance test processing method for a low-impedance voltage transformer as described above. The high-precision performance test processing system for a low-impedance voltage transformer includes:

[0051] The transformer test signal excitation module is used to generate a corresponding controllable test excitation signal through a tunable standard voltage source and a wide-band load network, and apply the controllable test excitation signal to the low-impedance voltage transformer to synchronously collect the corresponding transformer primary and secondary voltages; perform spatial electromagnetic cancellation based on the transformer primary and secondary voltages to generate the transformer original test signal;

[0052] The transformer characteristic matrix construction module is used to perform frequency domain feature decomposition on the original transformer test signal to obtain the transformer signal frequency domain features corresponding to the amplitude error, phase error, and harmonic distortion; calculate the instantaneous offset of the original transformer test signal to obtain the instantaneous offset of the transformer signal; and construct the transformer signal characteristic matrix based on the transformer signal frequency domain features and the transformer signal instantaneous offset;

[0053] The transformer performance prediction module is used to obtain the environmental parameters corresponding to the low-impedance voltage transformer. A neural network model based on Bayesian optimization is constructed, and the transformer signal feature matrix and environmental parameters are input to train the corresponding transformer performance joint prediction model, which predicts and outputs a set of performance parameters corresponding to the frequency response and load characteristics.

[0054] The performance calibration test evaluation module is used to generate a calibration compensation coefficient by comparing the performance parameter set with the preset standard threshold value, and inject the calibration compensation coefficient into the transformer performance joint prediction model to iteratively optimize the test until the error converges and test and evaluate, thereby generating a high-precision performance test report corresponding to the low-impedance voltage transformer.

[0055] Beneficial effects of the present invention:

[0056] 1. The high-precision performance test processing method for low-impedance voltage transformer proposed in the present invention, compared with the existing technology, has the beneficial effect of generating corresponding controllable test excitation signals through a tunable standard voltage source and a wide-band load network to ensure that the low-impedance voltage transformer can accurately simulate the signal response in the actual working environment when receiving test signals of different frequencies and amplitudes. By applying these controllable test excitation signals to the low-impedance voltage transformer, the voltages on the primary and secondary sides of the transformer can be synchronously collected. The key to this step is to use spatial electromagnetic elimination technology to extract the original test signal of the transformer. By eliminating the interference components in the signal, it is ensured that the acquired data can reflect the true performance of the transformer and avoid external factors from misleading the test results. This process can accurately restore the voltage behavior of the transformer during operation, which is helpful for subsequent signal processing and analysis, and provides a reliable data basis for further evaluation of the transformer performance. Secondly, by performing frequency domain feature decomposition on the original test signal of the transformer, the frequency domain features of the signal, such as amplitude error, phase error, and harmonic distortion, are extracted. These frequency domain features can provide in-depth insights for subsequent signal analysis, revealing the differences in the response of the transformer in different frequency bands. These error and distortion information directly reflect the performance defects of the transformer and can help engineers make a more accurate assessment of the working status of the transformer. In addition, the instantaneous offset of the signal is calculated to accurately capture the changes in the signal in the time domain, which is crucial for the precise calibration of the transformer. By calculating the instantaneous offset of the primary and secondary side signals, the time delay or performance deviation of the transformer during operation can be identified, which helps in the subsequent construction of the transformer signal feature matrix and performance optimization. The instantaneous offset of the signal can provide data support for optimizing transformer design and performance improvement. Then, by obtaining the environmental parameters of the low-impedance voltage transformer and constructing a neural network model based on Bayesian optimization, the neural network model can combine the signal feature matrix of the transformer and the environmental parameters, and obtain a joint prediction model of the transformer performance through training. Through the Bayesian optimization method, the neural network can effectively predict the performance of the transformer in a high-dimensional complex space, thereby providing data support for the subsequent tuning and performance improvement of the transformer. The core goal of model training is to achieve accurate prediction of key performance parameters such as frequency response and load characteristics. Through these predictions, engineers can quickly identify the performance of the transformer in different working environments, especially in complex or dynamic environments. It can more accurately predict and optimize the working state of the transformer, thereby improving the performance test efficiency of the low-impedance voltage transformer.Finally, by comparing the predicted performance parameter set with the preset standard threshold, a calibration compensation coefficient is generated. This process is an important step in optimizing the performance of the transformer. By generating the calibration compensation coefficient, the performance of the transformer can be accurately adjusted. The compensation coefficient can further optimize its accuracy in the design of the transformer and eliminate performance fluctuations caused by manufacturing errors or environmental changes. Through iterative optimization of the joint prediction model of the transformer performance, it is ensured that the error is gradually reduced during the calibration process and ultimately the error convergence is achieved. This process can improve the accuracy and repeatability of the performance test and ultimately generate a high-precision performance test report. This report can not only reflect the precise performance of the transformer in various working environments, but also greatly improve the reliability and stability of the transformer, and enhance its application value in the power system.

[0057] 2. The high-precision performance test and processing system for low-impedance voltage transformer proposed in the present invention is composed of a transformer test signal excitation module, a transformer characteristic matrix construction module, a transformer performance prediction module and a performance calibration test evaluation module. It can realize the high-precision performance test and processing method of any low-impedance voltage transformer described in the present invention, and is used to combine the operations between computer programs running on each module to realize the high-precision performance test and processing method of low-impedance voltage transformer. The internal structures of the system cooperate with each other, which can greatly reduce repetitive work and manpower investment, and can quickly and effectively provide a more accurate and efficient high-precision performance test and processing process for low-impedance voltage transformer, thereby simplifying the operating process of the high-precision performance test and processing system for low-impedance voltage transformer. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0059] Figure 1 A schematic flow chart of the steps of a high-precision performance test processing method for a low-impedance voltage transformer according to the present invention;

[0060] Figure 2 for Figure 1 Detailed step flow diagram of step S1;

[0061] Figure 3 for Figure 2 Detailed step flow chart of step S14. DETAILED DESCRIPTION

[0062] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0063] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0064] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0065] To achieve this, please refer to Figures 1 to 3 The present invention provides a high-precision performance test processing method for a low-impedance voltage transformer, the method comprising the following steps:

[0066] Step S1: Generate a corresponding controllable test excitation signal through a tunable standard voltage source and a wide-band load network, and apply the controllable test excitation signal to a low-impedance voltage transformer to synchronously collect the corresponding transformer primary voltage and transformer secondary voltage; perform spatial electromagnetic cancellation based on the transformer primary voltage and transformer secondary voltage to generate the transformer original test signal;

[0067] Step S2: performing frequency domain feature decomposition on the original test signal of the mutual inductor to obtain frequency domain features of the mutual inductor signal corresponding to amplitude error, phase error, and harmonic distortion; performing instantaneous offset calculation on the original test signal of the mutual inductor to obtain instantaneous offset of the mutual inductor signal; constructing a mutual inductor signal feature matrix based on the frequency domain features of the mutual inductor signal and the instantaneous offset of the mutual inductor signal;

[0068] Step S3: Obtaining environmental parameters corresponding to the low-impedance voltage transformer; constructing a neural network model based on Bayesian optimization, and inputting the transformer signal feature matrix and environmental parameters to train a corresponding transformer performance joint prediction model to predict and output a performance parameter set corresponding to the frequency response and load characteristics;

[0069] Step S4: Generate a calibration compensation coefficient by comparing the performance parameter set with the preset standard threshold value, and inject the calibration compensation coefficient into the transformer performance joint prediction model to iteratively optimize the test until the error converges and test and evaluate, and generate a high-precision performance test report corresponding to the low-impedance voltage transformer.

[0070] In the embodiment of the present invention, please refer to Figure 1 FIG. 1 is a flow chart showing the steps of a high-precision performance test processing method for a low-impedance voltage transformer according to the present invention. In this example, the high-precision performance test processing method for a low-impedance voltage transformer includes the following steps:

[0071] Step S1: Generate a corresponding controllable test excitation signal through a tunable standard voltage source and a wide-band load network, and apply the controllable test excitation signal to a low-impedance voltage transformer to synchronously collect the corresponding transformer primary voltage and transformer secondary voltage; perform spatial electromagnetic cancellation based on the transformer primary voltage and transformer secondary voltage to generate the transformer original test signal;

[0072] In an embodiment of the present invention, an arbitrary function generator of model AFG3101C is selected as a tunable standard voltage source, whose frequency output range is 0.1μHz to 50MHz and voltage output range is 0V to 10Vpp (peak-to-peak value). For a low-impedance voltage transformer with a rated operating frequency band of 45Hz-65Hz, the generator output mode is set to a sine wave, the starting frequency is set to 45Hz, the frequency adjustment step value is 0.1Hz, the end frequency is 65Hz, and the output voltage peak-to-peak value is set to 5V. At the same time, a wide-band load network is constructed, and 10 metal foil resistors (accuracy ±0.1%, temperature coefficient 5ppm / °C) with different resistance values ​​(10Ω, 100Ω, 1kΩ, etc.) are selected, 5 ferrite core wound inductors with different inductance values ​​(10μH, 100μH, etc.), and 5 polypropylene film capacitors with different capacitances (10nF, 100nF, etc.) are selected and connected in combination through a switch matrix composed of 20 high-precision electronic relays to connect the low-impedance voltage transformer to the sine wave. The primary side of the impedance voltage transformer is connected to a loop formed by a tunable standard voltage source and a broadband load network. The secondary side is connected to subsequent signal processing circuitry. The primary voltage is collected using a symmetrical Rogowski coil with 1000 turns, a 5 cm circular cross-sectional radius, and an average radius of 20 cm, which is tightly wrapped around the primary conductor. The secondary voltage is collected using a high-precision resistor divider consisting of 1000 kΩ, consisting of 10 metal film resistors with a nominal resistance of 100 kΩ and an accuracy of ±0.01%. The collected voltage signal first passes through a π-type filter structure consisting of four 10 nF polypropylene film capacitors and two 100 μH ferrite core inductors, then through a first-order RC low-pass filter circuit (1 kΩ resistor and 10 μF capacitor, cutoff frequency 16 Hz). Finally, it passes through a non-inverting proportional amplifier circuit (using an LM324 operational amplifier, power supply voltage ±12 V, feedback resistor 9 kΩ, input resistor 1 kΩ, and amplification factor of 10) to generate the transformer's original test signal.

[0073] Step S2: performing frequency domain feature decomposition on the original test signal of the mutual inductor to obtain frequency domain features of the mutual inductor signal corresponding to amplitude error, phase error, and harmonic distortion; performing instantaneous offset calculation on the original test signal of the mutual inductor to obtain instantaneous offset of the mutual inductor signal; constructing a mutual inductor signal feature matrix based on the frequency domain features of the mutual inductor signal and the instantaneous offset of the mutual inductor signal;

[0074] In this embodiment of the present invention, a digital oscilloscope with a sampling rate of 100kHz and 16-bit resolution is used to acquire the original test signal of the transformer, obtaining a discrete signal sequence with a duration of 1 second and 100,000 sampling points. The signal is then input into a dedicated signal processing device with a built-in fast Fourier transform module for a 1024-point FFT transform, which determines the amplitude and phase of the fundamental wave and each harmonic component. Given a transformer rated voltage ratio of 1000:1, the primary and secondary signal amplitudes are compared at a fundamental frequency of 50Hz to calculate the amplitude error. For example, if the primary amplitude is 98V and the secondary amplitude is 0.097V, the theoretical secondary amplitude is 0.098V. This yields the amplitude error. The phase difference between the primary and secondary fundamental waves is then calculated to obtain the phase error. The total distortion rate of each harmonic component is then calculated, and the contribution of each harmonic is analyzed to assess the harmonic distortion. A second-order difference approximation method is used to calculate the distribution curvature of each spatial point of the signal, construct a curvature tensor field, set a curvature mutation threshold (e.g., the mean plus three times the standard deviation, assumed to be 0.01), and determine the signal mutation region frame. Within a signal mutation region frame (e.g., sampling points 4990-5010), the 5000th sampling point is taken, and the amplitude and phase of the primary and secondary signals are compared. The amplitude offset and phase offset are calculated, and the instantaneous offset of the mutual inductor signal is then calculated according to a specific formula. Finally, the amplitude error, phase error, harmonic distortion-related data, and instantaneous offset are organized into a 25-dimensional mutual inductor signal feature matrix with four rows and multiple columns.

[0075] Step S3: Obtaining environmental parameters corresponding to the low-impedance voltage transformer; constructing a neural network model based on Bayesian optimization, and inputting the transformer signal feature matrix and environmental parameters to train a corresponding transformer performance joint prediction model to predict and output a performance parameter set corresponding to the frequency response and load characteristics;

[0076] In an embodiment of the present invention, environmental parameters are collected by using a temperature sensor (measuring range of -55°C to +125°C, accuracy of ±0.5°C), a humidity sensor (measuring range of 0% to 100% RH, accuracy of ±3% RH) and a vibration sensor (measuring range of ±16g, resolution of 0.003g / LSB). The three sensors are connected to a data acquisition instrument via an SPI interface and collected synchronously at a sampling rate of 10Hz. Within a 1-hour test period, 3600 sets of environmental parameter data are collected and synchronized with the mutual inductor signal feature matrix according to the timestamp. A three-layer fully connected neural network is designed. The 25-dimensional input layer corresponds to the 25 feature dimensions of the transformer signal feature matrix. The 128 nodes in the hidden layer use the ReLU activation function. The 2-dimensional output layer corresponds to the two performance parameters of frequency response and load characteristics. The Bayesian optimization algorithm is used to optimize the hyperparameters. The learning rate range is 0.0001 to 0.01 (logarithmic scale), the regularization coefficient range is 0.00001 to 0.001 (logarithmic scale), and the number of hidden layer nodes range is 64 to 256 (integer). The root mean square error is used as the optimization target. Initially, 10 sets of hyperparameters are randomly sampled, and 50 iterations are performed. An attention layer is added between the hidden layer and the output layer. Different weights are assigned according to the degree of correlation between the features and the target performance parameters. The model is trained using 5-fold cross-validation. Finally, a joint prediction model for transformer performance is obtained, which can output a set of performance parameters corresponding to frequency response and load characteristics.

[0077] Step S4: Generate a calibration compensation coefficient by comparing the performance parameter set with the preset standard threshold value, and inject the calibration compensation coefficient into the transformer performance joint prediction model to iteratively optimize the test until the error converges and test and evaluate, and generate a high-precision performance test report corresponding to the low-impedance voltage transformer.

[0078] In an embodiment of the present invention, standard thresholds for frequency response and load characteristics are pre-set, such as a frequency response error range of ±2% and a load characteristic error range of ±3%. The performance parameter set output by the transformer performance joint prediction model is compared with the standard thresholds. If the frequency response prediction value is 102%, exceeding the upper limit of the standard threshold, a calibration compensation coefficient of -2% is calculated; if the load characteristic prediction value is 104%, exceeding the upper limit of the standard threshold, a calibration compensation coefficient of -3% is calculated. The calibration compensation coefficient is then injected into the transformer performance joint prediction model, and training and testing are re-performed. During the iterative process, the prediction value is compared with the standard threshold after each iteration, and a new error is calculated. When the error change is less than 0.1% for five consecutive iterations, the error is considered to have converged. For example, after 10 iterations, the frequency response prediction value stabilizes at 99.8%, and the load characteristic prediction value stabilizes at 99.5%, meeting the error convergence conditions. Finally, the original test signal, signal feature matrix, environmental parameters, performance parameter set, iterative optimization process, and other data from the entire test process are compiled into a report, detailing the test results of the low-impedance voltage transformer on various performance indicators, generating a high-precision performance test report.

[0079] Furthermore, step S1 includes the following steps:

[0080] Step S11: Generate a sinusoidal excitation signal with an adjustable fundamental frequency by using a tunable standard voltage source to cover the rated operating frequency band corresponding to the low-impedance voltage transformer;

[0081] Step S12: Building a broadband load network including resistive, inductive, and capacitive components, and configuring a load impedance adjustment for a sinusoidal excitation signal corresponding to an adjustable fundamental frequency based on the broadband load network to generate a corresponding controllable test excitation signal;

[0082] Step S13: applying a controllable test excitation signal to the low-impedance voltage transformer, and synchronously collecting the corresponding transformer primary side voltage and transformer secondary side voltage using a symmetrical Rogowski coil and a high-precision resistor divider;

[0083] Step S14: Design a differential sampling and common-mode suppression network circuit based on a symmetrical Rogowski coil and a high-precision resistor divider to perform spatial electromagnetic elimination on the primary side voltage and the secondary side voltage of the transformer to generate an original transformer test signal.

[0084] As an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 Detailed step flow diagram of step S1 in FIG. 1 , in this embodiment, step S1 includes the following steps:

[0085] Step S11: Generate a sinusoidal excitation signal with an adjustable fundamental frequency by using a tunable standard voltage source to cover the rated operating frequency band corresponding to the low-impedance voltage transformer;

[0086] In this embodiment of the present invention, an arbitrary function generator (AFG3101C) is selected as a tunable standard voltage source. This generator has high-precision waveform generation capabilities, a frequency output range of 0.1μHz to 50MHz, and a voltage output range of 0V to 10Vpp (peak-to-peak). For a low-impedance voltage transformer with a rated operating frequency range of 45Hz-65Hz, the arbitrary function generator's output mode is set to a sine wave. The knob and numeric keys are used to set the starting frequency to 45Hz, the frequency adjustment step to 0.1Hz, and the end frequency to 65Hz. The output voltage is set to 5V peak-to-peak to ensure that the signal strength meets the test requirements and does not damage the transformer. After the settings are completed, the arbitrary function generator is activated to output a sinusoidal excitation signal with an adjustable fundamental frequency according to preset parameters. The digital synthesis module within the generator, based on direct digital synthesis (DDS) technology, generates a stable and continuously frequency-adjustable sine wave signal by precisely controlling the increment of the phase accumulator. At a frequency of 45Hz, the waveform of the output signal is stable, with the peak-to-peak value strictly maintained at 5V. As the frequency gradually increases to 65Hz in steps of 0.1Hz, the amplitude fluctuation of the signal is controlled within ±0.05V throughout the entire process, thus covering the rated operating frequency band corresponding to the low-impedance voltage transformer and providing a stable and adjustable excitation source for subsequent tests.

[0087] Step S12: Building a broadband load network including resistive, inductive, and capacitive components, and configuring a load impedance adjustment for a sinusoidal excitation signal corresponding to an adjustable fundamental frequency based on the broadband load network to generate a corresponding controllable test excitation signal;

[0088] In an embodiment of the present invention, a broadband load network is constructed and a plurality of standard electronic components are selected. The resistance components are metal foil resistors with an accuracy of ±0.1% and a temperature coefficient of 5ppm / °C. The resistance values ​​include 10Ω, 100Ω, 1kΩ, 10kΩ, etc., with a total of 10 resistors with different resistance values; the inductance components are inductors wound with ferrite cores, with inductance values ​​covering 10μH, 100μH, 1mH, etc., with a total of 5 different specifications; the capacitance components are polypropylene film capacitors, with capacitance values ​​including 10nF, 100nF, 1μF, etc., with a total of 5 different capacities. These resistive, inductive, and capacitive components are combined and connected through a switch matrix composed of 20 high-precision electronic relays, which can realize rapid switching and combination of components to generate controllable measurement of specific load impedance. Taking the test excitation signal as an example, when simulating a purely resistive 100Ω load, two 50Ω metal foil resistors are connected in series to the circuit by controlling the switch matrix. To simulate an RLC parallel load consisting of a 100μH inductor and a 100nF capacitor, the corresponding inductor and capacitor components are connected in parallel by switching. For sinusoidal excitation signals between 45Hz and 65Hz, the load network is configured sequentially according to different test requirements. For example, at 45Hz, the load is set to a 50Ω resistor connected in series with a 100μH inductor. As the frequency changes, the component combination is dynamically adjusted to make the load impedance continuously adjustable over a wide bandwidth. In this way, the sinusoidal excitation signal is configured with load impedance adjustment based on the wideband load network, generating a controllable test excitation signal that meets the test requirements and provides diverse load conditions for accurately testing the performance of low-impedance voltage transformers.

[0089] Step S13: applying a controllable test excitation signal to the low-impedance voltage transformer, and synchronously collecting the corresponding transformer primary side voltage and transformer secondary side voltage using a symmetrical Rogowski coil and a high-precision resistor divider;

[0090] In this embodiment of the present invention, a low-impedance voltage transformer is connected to the test circuit, with its primary side connected to a controllable test excitation signal output and its secondary side connected to a subsequent signal processing circuit. A symmetrical Rogowski coil with 1000 turns, a 5-centimeter ring cross-sectional radius, and an average radius of 20 centimeters is used to collect the transformer's primary voltage. This Rogowski coil exhibits excellent linearity and stability within its rated operating frequency band. The coil is tightly wrapped around the transformer's primary conductor to ensure that the magnetic field generated by the primary current fully passes through the coil. The output of the Rogowski coil is connected to the input of a signal conditioning circuit. At the same time, a 1000kΩ high-precision resistor divider, consisting of ten series-connected metal film resistors with a nominal resistance of 100kΩ and an accuracy of ±0.01%, is used to collect the transformer's secondary voltage. The high-voltage end of the resistor divider is connected to the transformer's secondary output terminal, while the low-voltage end is grounded. A scaled-down voltage signal is obtained from the center tap and fed into the signal conditioning circuit. For example, when the controllable test excitation signal is a 50Hz, 100V amplitude sine wave, the actual primary input voltage is 100V. The symmetrical Rogowski coil induces a signal proportional to the primary voltage, which is converted and output to the signal conditioning circuit. The theoretical output voltage of the transformer's secondary side is 0.1V, which is scaled down by the high-precision resistor divider and output to the signal conditioning circuit. This enables simultaneous acquisition of the transformer's primary and secondary voltages, providing raw data for subsequent signal processing.

[0091] Step S14: Design a differential sampling and common-mode suppression network circuit based on a symmetrical Rogowski coil and a high-precision resistor divider to perform spatial electromagnetic elimination on the primary side voltage and the secondary side voltage of the transformer to generate an original transformer test signal.

[0092] In an embodiment of the present invention, a differential sampling and common-mode suppression network circuit is designed based on the signals collected by a symmetrical Rogowski coil and a high-precision resistor divider. Two 1kΩ metal film resistors with completely identical parameters and an accuracy of ±0.01% are used as differential sampling resistors. The primary side voltage signal output by the symmetrical Rogowski coil and the secondary side voltage signal output by the high-precision resistor divider are respectively connected to one end of the two differential sampling resistors. The other ends of the two resistors are connected to ground, and the differential signal is obtained from the connection point. To suppress spatial electromagnetic interference, a common-mode suppression network is connected after the differential sampling circuit. The common-mode suppression network consists of four 10nF polypropylene film capacitors and two 100μH ferrite core inductors forming a π-type filter structure. The two inductors are connected in series to the two lines of the differential signal, and the four capacitors are connected between the differential line and ground. At the same time, to address possible low-frequency noise, a first-order RC low-pass filter circuit is added, consisting of a 1kΩ resistor and a 10μF capacitor, with a cutoff frequency set to 16Hz. When the input signal contains common-mode noise with an amplitude of 5V and a frequency of 500Hz, and differential-mode noise with an amplitude of 2V and a frequency of 300Hz, after differential sampling and common-mode suppression network circuit processing, the common-mode noise is reduced to 0.05V, and the differential-mode noise is reduced to 0.1V. Finally, the signal amplitude is amplified 10 times by the in-phase proportional amplifier circuit. The amplification circuit is constructed using an LM324 operational amplifier with a power supply voltage of ±12V. The amplification factor is determined by the ratio of the feedback resistor to the input resistor (the feedback resistor is 9kΩ and the input resistor is 1kΩ). After the above processing, a clean and accurate transformer original test signal is generated. The amplitude and phase relationship of the primary and secondary side signals accurately reflect the performance of the low-impedance voltage transformer, providing reliable data for high-precision performance testing.

[0093] Further, as an embodiment of the present invention, refer to Figure 3 As shown, Figure 2 Detailed step flow diagram of step S14 in the embodiment, step S14 includes the following steps:

[0094] Step S141: simulating the magnetic field distribution generated by the primary side voltage of the transformer and the corresponding coupling condition of the symmetrical Rogowski coil according to the law of electromagnetic induction for the symmetrical Rogowski coil, and calculating the mutual inductance coefficient of the symmetrical Rogowski coil in all directions to generate a spatial distribution field of the mutual inductance coefficient of the Rogowski coil;

[0095] In this embodiment of the present invention, an analysis is performed based on the law of electromagnetic induction for a symmetrical Rogowski coil. Taking a symmetrical Rogowski coil with 1000 turns, a circular cross-sectional radius of 5 cm, and an average radius of 20 cm as an example, it is assumed that the primary side of the transformer is a straight conductor carrying a sinusoidal AC current with an amplitude of 100 A and a frequency of 50 Hz. Finite element analysis software is used to divide the space containing the Rogowski coil and the primary conductor into 100,000 tetrahedral mesh elements. Input the material properties in the software. The skeleton of the Rogowski coil is non-magnetic plastic, and the relative magnetic permeability is set to 1. The coil conductor is copper, and the conductivity is set to 5.8×0.0000001S / m. Set the boundary conditions and set the outer boundary of the calculation area to the parallel boundary of the magnetic flux. By solving the Maxwell equations, simulate the magnetic field distribution generated by the primary current of the transformer. It is observed that the magnetic field is axisymmetric in the annular area of ​​the Rogowski coil. When calculating the mutual inductance coefficient of the Rogowski coil in various directions, 36 points are selected at equal intervals on the circumference of the Rogowski coil, and each point corresponds to a direction. Take one of the points as an example. In this direction, by changing the magnitude and frequency of the primary current, the change of the induced voltage of the Rogowski coil is recorded. According to the mutual inductance coefficient formula (in is the induced electromotive force, is the primary side current), the mutual inductance coefficient values ​​are calculated in all 36 directions, and these mutual inductance coefficient values ​​are filled into the three-dimensional grid according to the spatial position distribution to generate the spatial distribution field of the mutual inductance coefficient of the Rogowski coil, which intuitively presents the change of the mutual inductance coefficient in space.

[0096] Step S142: calculating the input impedance and output impedance of the high-precision resistor divider at different frequencies based on the high-precision resistor divider and taking into account the parasitic parameters of the resistor elements and the frequency variation, and constructing a corresponding resistor divider impedance frequency response matrix according to the different frequencies;

[0097] In an embodiment of the present invention, an analysis is performed based on a high-precision resistor divider. Taking a 1000kΩ high-precision resistor divider composed of 10 metal film resistors with a nominal resistance of 100kΩ connected in series as an example, the parasitic parameters corresponding to the resistor elements are considered. The parasitic capacitance of each resistor is 1pF, and the parasitic inductance is 1nH. Within the frequency variation range of 10Hz-100kHz, 10,000 frequency points are selected at intervals of 10Hz. For each frequency point, an equivalent circuit model including the resistor divider and its parasitic parameters is constructed using circuit analysis software. According to circuit theory, the input impedance and output impedance corresponding to the high-precision resistor divider at this frequency are calculated by solving the node voltage equation and the loop current equation. For example, at a frequency of 100 Hz, after calculation, the input impedance is 999.9 + j0.01 Ω, and the output impedance is 0.1 + j0.001 Ω. The input impedance and output impedance values ​​corresponding to all frequency points are organized into a 2 × 10000 matrix. The first row records the real and imaginary parts of the input impedance, and the second row records the real and imaginary parts of the output impedance. The corresponding resistor divider impedance frequency response matrix is ​​constructed according to different frequencies to clearly show the change characteristics of the resistor divider impedance with frequency.

[0098] Step S143: Decomposing the interference electromagnetic field corresponding to the low-impedance voltage transformer into common-mode interference and differential-mode interference, and determining the distribution of the interference electromagnetic field in the space surrounding the low-impedance voltage transformer to generate an electromagnetic interference modal decomposition vector field, wherein the vector of each spatial point in the field includes the magnitude and direction of the common-mode interference and the differential-mode interference;

[0099] In an embodiment of the present invention, by analyzing the interference electromagnetic field around a low-impedance voltage transformer, assuming that the low-impedance voltage transformer is in an environment with interference from industrial equipment, an electromagnetic field measuring device is used to construct a three-dimensional spatial grid of 1 meter × 1 meter × 1 meter at intervals of 0.1 meters around the transformer, obtaining a total of 1000 spatial points. A three-axis electromagnetic probe is used to measure the electric field strength and magnetic field strength in three orthogonal directions at each spatial point to obtain the electromagnetic field vector of each point. The measured interference electromagnetic field is decomposed into common-mode interference and differential-mode interference. The symmetrical component method is used to decompose the electromagnetic field vector of each spatial point into a common-mode component and a differential-mode component. For example, at the spatial point (0.3, 0.4, 0.5), the measured electric field strength vector is (10, 5, 3) V / m. After decomposition, the common-mode electric field strength component is (6, 6, 6) V / m, and the differential-mode electric field strength component is (4, -1, -3) V / m; the magnetic field strength vector is (2, 1, 0.5) A / m, the common-mode magnetic field strength component is (1.17, 1.17, 1.17) A / m, and the differential-mode magnetic field strength component is (0.83, -0.17, -0.67) A / m. The magnitude and direction information of the common-mode interference and differential-mode interference at all spatial points are organized into a three-dimensional vector field to generate an electromagnetic interference modal decomposition vector field, providing data support for subsequent interference suppression.

[0100] Step S144: Designing a differential sampling and common-mode suppression network circuit based on the spatial distribution field of the Rogowski coil mutual inductance coefficient and the impedance frequency response matrix of the resistor divider in combination with the electromagnetic interference modal decomposition vector field, thereby optimizing the symmetrical Rogowski coil layout by utilizing the symmetrical structure corresponding to the symmetrical Rogowski coil to suppress common-mode interference and differential-mode interference, selecting appropriate impedance parameters based on the impedance frequency response corresponding to the high-precision resistor divider to match the differential sampling, and designing a corresponding filter circuit for the electromagnetic interference mode;

[0101] In an embodiment of the present invention, a differential sampling and common-mode suppression network circuit is designed based on the previously generated spatial distribution field of the Rogowski coil's mutual inductance coefficient, the impedance frequency response matrix of the resistor divider, and the electromagnetic interference modal decomposition vector field. For symmetrical Rogowski coils, based on their spatial distribution field of mutual inductance coefficient, it is found that the coil's mutual inductance coefficient is most symmetrical along its axis, resulting in the strongest suppression capability for common-mode interference. Therefore, the symmetrical Rogowski coil layout is adjusted so that its axis is parallel to the primary conductor of the transformer, leveraging this symmetrical structure to enhance the suppression capability for common-mode interference. Furthermore, for differential-mode interference, the suppression of differential-mode interference is further optimized by connecting a balancing resistor in series at the output of the Rogowski coil. According to the impedance frequency response matrix of the resistor divider, within the frequency range of 100Hz-1kHz, the imaginary part of the input impedance of the resistor divider is small and stable, and the real part of the output impedance is relatively low. To match the differential sampling, a resistor with an input impedance of 1kΩ and an output impedance of 100Ω is selected as the matching resistor of the differential sampling circuit to ensure that the signal is not distorted during transmission. A filter circuit is designed for common-mode interference and differential-mode interference of different directions and frequencies in the electromagnetic interference modal decomposition vector field. For high-frequency common-mode interference, a π-type filter circuit consisting of two 10nF capacitors and a 100Ω inductor is used; for low-frequency differential-mode interference, a second-order low-pass filter circuit is used with a cutoff frequency set to 10Hz to effectively suppress different types of electromagnetic interference.

[0102] Step S145: Connect the differential sampling and common-mode suppression network circuit to the sampling loops corresponding to the primary-side voltage and the secondary-side voltage of the transformer to perform spatial electromagnetic elimination on their output signals, so as to extract the effective signal components corresponding to the primary-side voltage and the secondary-side voltage of the transformer through differential sampling, and use the common-mode suppression network to suppress the common-mode noise and differential-mode noise generated by spatial electromagnetic interference. At the same time, the processed signal is filtered and amplified and modulated to remove residual high-frequency noise and low-frequency drift, and generate the original transformer test signal, which includes the primary-side signal and the secondary-side signal.

[0103] In an embodiment of the present invention, a designed differential sampling and common-mode suppression network circuit is connected to the sampling loops corresponding to the primary voltage and the secondary voltage of the transformer. Taking a low-impedance voltage transformer with a rated voltage ratio of 10,000 V / 100 V as an example, the actual measured value of the primary voltage is 9,800 V, and the actual measured value of the secondary voltage is 98 V. In the sampling loop, the differential sampling circuit collects the primary and secondary voltage signals of the transformer through two symmetrical sampling resistors, respectively. The symmetrical Rogowski coil and matching resistor are used to extract the effective signal components corresponding to the primary and secondary voltages. At the same time, the common-mode suppression network effectively suppresses the common-mode noise and differential-mode noise generated by spatial electromagnetic interference. For example, in an environment with common-mode noise of 10V amplitude and 500Hz frequency and differential-mode noise of 5V amplitude and 200Hz frequency, after processing through the common-mode suppression network, the common-mode noise is reduced to 0.1V and the differential-mode noise is reduced to 0.05V. The processed signals are then filtered, amplified, and modulated using a fourth-order Butterworth low-pass filter with a cutoff frequency of 1kHz to remove residual high-frequency noise. A non-inverting proportional amplifier circuit with an amplification factor of 10 is used to adjust the signal amplitude to an appropriate range. A DC bias voltage of 2V is added through an adder circuit to eliminate low-frequency drift. The resulting original transformer test signal is 980.0V on the primary side and 9.8V on the secondary side, meeting the signal requirements for high-precision performance testing.

[0104] Furthermore, step S2 includes the following steps:

[0105] Step S21: performing spectrum transformation and decomposition on the original test signal of the mutual inductor to generate amplitude distribution and phase distribution corresponding to the fundamental wave and each harmonic component in the primary and secondary side signals;

[0106] In an embodiment of the present invention, a digital oscilloscope with a sampling rate of 100kHz and a resolution of 16 bits is selected to collect the original test signal of the mutual inductor, and a discrete signal sequence with a duration of 1 second and 100,000 sampling points is obtained. The sequence includes primary and secondary side signals. The collected signal sequence is input into a dedicated signal processing device. The device has a built-in fast Fourier transform (FFT) module, which performs a 1024-point FFT transformation on the signal. Taking the primary side signal as an example, after the FFT transformation, in the frequency domain, the amplitude at the fundamental frequency (assuming it is 50Hz) is 98V and the phase is 10°; the amplitude at the third harmonic (150Hz) is 5V and the phase is 20°; the amplitude at the fifth harmonic (250Hz) is 3V and the phase is 30°. The secondary side signal is also subjected to FFT transformation to obtain the amplitude and phase corresponding to its fundamental wave and each harmonic component. For example, the amplitude of the fundamental wave is 0.097V and the phase is 8°. , in order of frequency, the amplitude distribution and phase distribution corresponding to the fundamental wave and each harmonic component in the primary and secondary side signals are sorted into a data table, which clearly presents the amplitude and phase information of the signal at different frequency components, providing basic data for subsequent analysis.

[0107] Step S22: performing amplitude and phase error analysis on the amplitude distribution and phase distribution corresponding to the fundamental wave components in the primary and secondary side signals to obtain frequency domain characteristics of the mutual inductor signal corresponding to the amplitude error and phase error;

[0108] In an embodiment of the present invention, amplitude and phase error analysis is performed based on the amplitude distribution and phase distribution corresponding to the fundamental wave components in the primary and secondary side signals obtained previously. It is known that the rated voltage ratio of the transformer is 1000:1. At the fundamental frequency of 50 Hz, the primary side signal amplitude is 98 V, and the secondary side signal amplitude is 0.097 V. Theoretically, the secondary side fundamental wave amplitude should be 0.098 V. The amplitude error is calculated according to the formula: Amplitude Error = (Actual Secondary Side Amplitude - Theoretical Secondary Side Amplitude) / Theoretical Secondary Side Amplitude × 100%, that is, (0.097-0.098) / 0.098×100%≈-1.02%. For the phase error, the primary side fundamental wave phase is 10°, and the secondary side fundamental wave phase is 8°. The phase error = primary side phase - secondary side phase, that is, 10°-8°=2°. These amplitude error and phase error data are recorded, and their variation patterns under different test conditions are analyzed. For example, when changing the input voltage or frequency, observe the fluctuations of the amplitude error and phase error, and summarize the frequency domain characteristics of the transformer signal corresponding to the amplitude error and phase error. For example, at certain specific frequency points, the amplitude error increases and the phase error suddenly changes, so as to evaluate the accuracy performance of the transformer in fundamental wave signal transmission.

[0109] Step S23: performing harmonic distortion evaluation analysis on the amplitude distribution and phase distribution corresponding to each harmonic component in the primary and secondary side signals to obtain the frequency domain characteristics of the mutual inductor signal corresponding to the harmonic distortion;

[0110] In the embodiment of the present invention, the harmonic distortion evaluation analysis is performed based on the amplitude distribution and phase distribution corresponding to each harmonic component in the primary and secondary side signals previously obtained. The total distortion rate of each harmonic component is first calculated using the formula: Total distortion rate = × 100% (where for Subharmonic amplitude, is the fundamental amplitude), taking the primary side signal as an example, the fundamental amplitude =98V, 3rd harmonic amplitude =5V, 5th harmonic amplitude =3V, then the total distortion rate = ×100%≈5.2%. At the same time, the relative proportions between the various harmonic components are analyzed, such as the proportion of the third harmonic = (5 / 98)×100%≈5.1%, and the proportion of the fifth harmonic = (3 / 98)×100%≈3.1%. Combined with the total distortion rate and the proportion of each harmonic, they are compared with the standard requirements or normal operating indicators. If the total distortion rate exceeds the specified threshold (assuming it is 5%) and the proportion of some harmonics is abnormal, it is determined that a harmonic distortion problem exists. The frequency domain characteristics of the transformer signal corresponding to the harmonic distortion are summarized, such as excessively high amplitude at specific harmonic frequencies and uneven harmonic energy distribution. This is used to evaluate the performance of the transformer in a harmonic environment.

[0111] Step S24: calculating the instantaneous offset of the original test signal of the transformer to obtain the instantaneous offset of the transformer signal;

[0112] In an embodiment of the present invention, a signal mutation region frame is determined by performing signal curvature analysis on the original test signal of the mutual inductor. After the signal is collected by a digital oscilloscope, the signal distribution curvature at each spatial point is calculated by using a second-order difference approximation method, and a curvature tensor field is constructed. The signal mutation region frame is determined by setting a threshold (such as the mean plus 3 times the standard deviation, assuming the threshold is 0.01). Assuming that a determined signal mutation region frame is the 4990th-5010th sampling point, at the 5000th sampling point, the primary side signal amplitude is =100V, Phase =15°; Secondary side signal amplitude =0.099V, phase =12°, the calculation of the instantaneous offset of the signal takes into account the amplitude and phase differences. First calculate the amplitude offset ,get =99.901V; then calculate the phase offset ,get =3°, and finally, the offset is calculated by the formula (in is the rated voltage of the transformer, assuming =100V) to calculate the instantaneous offset of the transformer signal, which is 99.902V. This calculation is performed for all key sampling points in the signal mutation area frame to obtain complete transformer signal instantaneous offset data, reflecting the difference between the primary and secondary sides of the signal at the moment of mutation.

[0113] Step S25: constructing a mutual inductor signal characteristic matrix according to the mutual inductor signal frequency domain characteristics and the mutual inductor signal instantaneous offset.

[0114] In an embodiment of the present invention, a mutual inductor signal feature matrix is ​​constructed based on the previously obtained frequency domain characteristics of the mutual inductor signal corresponding to the amplitude error and phase error, the frequency domain characteristics of the mutual inductor signal corresponding to the harmonic distortion, and the obtained instantaneous offset of the mutual inductor signal. The feature matrix is ​​set as a 4-row and multi-column matrix, where the first row records the amplitude error data, such as the amplitude error values ​​corresponding to frequency points such as 50 Hz, 100 Hz, and 150 Hz, arranged in sequence; the second row records the phase error data; the third row records the harmonic distortion-related data, including the total distortion rate and the proportion of each harmonic; the fourth row records the instantaneous offset data of the mutual inductor signal, corresponding to the calculation results of each signal mutation area frame. The mutual inductor signal feature matrix comprehensively integrates the 25-dimensional key feature information of the mutual inductor signal in the frequency domain and time domain, providing a unified data format and analysis basis for subsequent high-precision performance evaluation and fault diagnosis of the mutual inductor based on these features.

[0115] Furthermore, step S22 includes the following steps:

[0116] Performing time-frequency synchronization processing on the amplitude distribution and phase distribution corresponding to the fundamental wave components in the primary and secondary side signals to obtain the amplitude distribution and phase distribution corresponding to the primary and secondary side signal components in the same time-frequency range;

[0117] In an embodiment of the present invention, a time-frequency analysis method based on fast Fourier transform (FFT) is adopted to process the primary and secondary side signals in the generated original test signal of the mutual inductor. The primary and secondary side signals are collected by using a digital oscilloscope with a sampling rate of 100 kHz and a 16-bit resolution, and stored as discrete signal sequences with a duration of 1 second. Each signal sequence contains 100,000 sampling points. The collected primary and secondary side signal sequences are input into a dedicated signal processing device. The device has a built-in FFT operation module, which performs a 1024-point FFT transform on the signal, converts the time domain signal into a frequency domain signal, and obtains the signal's frequency spectrum distribution in the frequency range of 0-50 kHz. In the frequency domain, the fundamental frequency (assuming it's 50 Hz) and the frequency locations of its harmonic components are identified. Taking the primary signal as an example, after FFT transformation, the amplitude of the fundamental component at 50 Hz is found to be 98 V, with a phase of 10°. The amplitude of the fundamental component of the secondary signal at 50 Hz is 0.097 V, with a phase of 8°. Because the sampling start times of the two signals may differ, resulting in time-frequency asynchrony, a phase synchronization algorithm is used to adjust the two signals. By calculating the phase difference between the fundamental components of the two signals (here, 2°), phase compensation is performed on the secondary signal, adjusting all sampling points of the secondary signal by a 2° phase difference. Simultaneously, the signals are time-aligned based on their sampling frequencies and timestamps to ensure they are on the same time axis. After processing, the amplitude and phase distributions of the primary and secondary signal components within the same time-frequency range are obtained, providing a unified benchmark for subsequent error analysis.

[0118] Preferably, amplitude error statistics are performed between the amplitude distributions corresponding to the primary and secondary side signal components in the same time-frequency range to obtain frequency domain characteristics of the mutual inductor signal corresponding to the amplitude error between the primary and secondary side signals;

[0119] In this embodiment of the present invention, amplitude error statistics are calculated based on the previously obtained amplitude distribution of the primary and secondary signal components within the same time-frequency range. Within the 0-50 kHz frequency range, the amplitudes of the primary and secondary signals are sequentially extracted at 1 Hz intervals. For example, at 50 Hz, the primary signal amplitude is 98 V and the secondary signal amplitude is 0.097 V. Based on the transformer's rated voltage ratio (assuming a 1000:1 ratio), the theoretical secondary signal amplitude should be 0.098 V. The amplitude error is calculated using the formula: Amplitude Error = (Actual Secondary Amplitude - Theoretical Secondary Amplitude) / Theoretical Secondary Amplitude × 100%, i.e., (0.097 - 0.098) / 0.098 × 100% ≈ -1.02%. This amplitude error calculation is performed for all frequency points within the 0-50 kHz frequency range, resulting in a series of amplitude error data. This data is arranged in frequency order, and a curve showing the amplitude error as a function of frequency is plotted. By analyzing the curve characteristics, such as the sudden increase in the amplitude error to -2% at 100Hz and the recovery to around -1% at 200Hz, the frequency domain characteristics of the transformer signal corresponding to the amplitude error between the primary and secondary side signals are summarized. For example, the existence of amplitude attenuation or gain anomalies at certain specific frequency points provides a basis for evaluating the frequency response characteristics of the transformer.

[0120] Preferably, the phase error is calculated based on the phase distributions corresponding to the primary and secondary side signal components in the same time-frequency range to obtain the mutual inductor signal frequency domain characteristics corresponding to the phase error between the primary and secondary side signals.

[0121] In an embodiment of the present invention, the phase error is calculated based on the phase distribution corresponding to the primary and secondary signal components in the same time-frequency range previously obtained. Similarly, within the frequency range of 0-50kHz, the phase values ​​of the primary and secondary signals at each frequency point are extracted with a frequency interval of 1Hz. At the 50Hz frequency point, the primary signal phase is 10° and the secondary signal phase is 8°. The phase error is calculated using the formula: phase error = primary phase - secondary phase, i.e. 10° - 8° = 2°. The phase error is calculated for all frequency points within the frequency range in turn to obtain the phase error data for each frequency point. These phase error data are sorted in frequency order to draw a curve showing the phase error changing with frequency. Observing the curve, it is found that in the frequency range of 50Hz to 100Hz, the phase error gradually increases from 2° to 5°, and then slowly decreases in the range of 100Hz to 200Hz. By analyzing the entire curve, the frequency domain characteristics of the transformer signal corresponding to the phase error between the primary and secondary side signals are summarized. For example, the phase error is small and changes smoothly in the low-frequency band, while the phase error fluctuates significantly in the medium and high-frequency bands. These characteristics can reflect the phase transmission characteristics of the transformer at different frequencies and provide important frequency domain information for comprehensively evaluating the high-precision performance of low-impedance voltage transformers.

[0122] Furthermore, step S23 includes the following steps:

[0123] Step S231: performing harmonic proportion statistics calculation based on the amplitude distribution and phase distribution corresponding to each harmonic component in the primary and secondary side signals to obtain the proportion between each harmonic component of the primary and secondary side signals;

[0124] In an embodiment of the present invention, by utilizing the frequency domain data of the primary and secondary side signals that have been previously processed by time-frequency synchronization, the amplitude corresponding to each harmonic component is extracted within the frequency range of 0-50kHz with a frequency interval of 1Hz. It is assumed that the amplitude of the primary side signal at the fundamental frequency of 50Hz is 98V, and the amplitude of the secondary side signal at the fundamental frequency of 50Hz is 0.097V; the amplitude of the primary side signal at the third harmonic (150Hz) is 5V, and the amplitude of the secondary side signal at the third harmonic is 0.005V; the amplitude of the primary side signal at the fifth harmonic (250Hz) is 3V, and the amplitude of the secondary side signal at the fifth harmonic is 0.003V. The formula is used to calculate the proportion of each harmonic component: harmonic proportion = (harmonic component amplitude / fundamental component amplitude) × 100%. For the third harmonic of the primary side signal, the proportion = (5 / 98) × 100% ≈ 5.1% The proportion of the third harmonic in the secondary side signal is (0.005 / 0.097) × 100% ≈ 5.2%. Similarly, the proportion of each harmonic in the primary and secondary side signals is calculated. For example, the proportion of the fifth harmonic in the primary side signal is (3 / 98) × 100% ≈ 3.1%, and the proportion of the fifth harmonic in the secondary side signal is (0.003 / 0.097) × 100% ≈ 3.1%. The proportion of each harmonic component in the primary and secondary side signals is recorded separately to form a complete harmonic proportion data list, which clearly shows the relative proportion of each harmonic in the signal.

[0125] Step S232: constructing a vector field for each harmonic component vector based on the coupling relationship between adjacent harmonic component vectors by treating the amplitude distribution and phase distribution corresponding to each harmonic component in the primary and secondary side signals as harmonic component vectors having corresponding amplitude and phase attributes, and calculating the coupling relationship between adjacent harmonic component vectors, thereby generating harmonic component vector fields corresponding to the primary and secondary side signals;

[0126] In the embodiment of the present invention, the amplitude and phase corresponding to each harmonic component in the primary and secondary side signals are used as attributes to construct a harmonic component vector. Taking the primary side signal as an example, the amplitude of the third harmonic component is 5V and the phase is 20°. The constructed vector is ; The amplitude of the 5th harmonic component is 3V and the phase is 30°, and the constructed vector is The secondary side signal is constructed in the same way. For example, the amplitude of the third harmonic component is 0.005V, the phase is 22°, and the vector is , calculate the coupling relationship between adjacent harmonic component vectors, using the vector dot product formula: coupling coefficient = ,in 、 For adjacent harmonic component vectors, for example, to calculate the coupling coefficient of the primary side signal's 3rd harmonic and 5th harmonic vectors, first calculate the dot product , and then calculate the two vector moduli 、 , substitute into the formula to obtain the coupling coefficient. According to this method, the coupling coefficient between all adjacent harmonic component vectors is calculated, and a vector field is constructed in the form of a spatial grid. Each harmonic component vector is placed at the corresponding frequency coordinate point. The connection strength and direction between the vectors are determined according to the coupling coefficient, and the harmonic component vector field corresponding to the primary and secondary side signals is generated. This vector field intuitively displays the mutual relationship and spatial distribution characteristics between the harmonic components. For example, in some frequency ranges, the harmonic component vectors are closely coupled, indicating that these harmonics have a large impact on each other.

[0127] Step S233: Calculate the harmonic energy proportion and harmonic energy transfer path of each harmonic component in the entire signal cycle according to the harmonic component vector field corresponding to the primary and secondary side signals;

[0128] In the embodiment of the present invention, the harmonic energy corresponding to each harmonic component in the entire signal cycle is calculated based on the previously generated harmonic component vector field, using the formula: Harmonic energy = (harmonic component amplitude 2 / 2). Taking the third harmonic of the primary signal as an example, with an amplitude of 5V, the harmonic energy = (5² / 2) = 12.5W; the third harmonic of the secondary signal has an amplitude of 0.005V, and the harmonic energy = (0.005² / 2) = 1.25×0.00001W. Calculate the harmonic energy percentage: Harmonic energy percentage = (harmonic energy / total harmonic energy) × 100%. The total harmonic energy is the sum of the energies of all harmonics. For example, if the total harmonic energy of the primary signal is 12.5 + (3² / 2) + …(other harmonic energies), the third harmonic energy percentage = (12.5 / total harmonic energy) × 100%. To determine the harmonic energy transfer path, analyze the connection direction and coupling strength of the vectors in the harmonic component vector field. If the coupling coefficient of adjacent harmonic component vectors is high and points to a certain harmonic, it is determined that there is a path for energy transfer to that harmonic. For example, in the vector field, it is found that the third harmonic vector of the primary side signal is tightly coupled with the fifth harmonic vector and points to the fifth harmonic, indicating that part of the third harmonic energy is transferred to the fifth harmonic. Based on this, the energy transfer path between the harmonics in the entire signal cycle is sorted out, clearly showing the flow trend of harmonic energy in the signal.

[0129] Step S234: quantifying the amplitude-phase coupling distortion rate of the amplitude distribution and phase distribution corresponding to each harmonic component in the primary and secondary side signals based on the harmonic energy proportion corresponding to each harmonic component in the entire signal cycle and the harmonic energy transfer path, to obtain the total distortion rate of the harmonic components corresponding to the primary and secondary side signals;

[0130] In an embodiment of the present invention, the amplitude-phase coupling distortion rate corresponding to each harmonic component in the primary and secondary side signals is quantified based on the harmonic energy proportion of each harmonic component in the entire signal cycle and the harmonic energy transfer path, and the distortion rate calculation formula is introduced: Amplitude-Phase Coupling Distortion Rate = (harmonic energy ratio deviation) + (correlation between phase deviation and energy transfer), where 、 is the weight coefficient (assuming =0.6, =0.4), the harmonic energy share deviation is calculated as the absolute value of the difference between the actual harmonic energy share and the ideal state (no distortion). The correlation between phase deviation and energy transfer is comprehensively evaluated based on the phase difference and coupling strength of adjacent harmonics in the harmonic energy transfer path. Taking the third harmonic of the primary side signal as an example, the ideal energy share is 3%, the actual energy share is 5.1%, and the energy share deviation is |5.1%-3%|=2.1%. If the third harmonic transfers energy to the fifth harmonic, the correlation between the phase difference and coupling strength is 0.7, then the amplitude-phase coupling distortion rate = 0.6×2.1%+0.4×0.7=0.2926. The above calculation is performed for each harmonic of the primary and secondary side signals to obtain the amplitude-phase coupling distortion rate of each harmonic. The weighted sum of all harmonic distortion rates is used to obtain the total distortion rate of the harmonic components corresponding to the primary and secondary side signals. For example, the weighted sum of the distortion rates of each harmonic of the primary side signal is 8%. , the secondary side signal is 8.2%. The total distortion rate quantifies the degree of amplitude and phase distortion caused by harmonics in the signal.

[0131] Step S235: performing harmonic distortion evaluation analysis based on the proportion of each harmonic component of the primary and secondary side signals and the corresponding total distortion rate of the harmonic components to obtain the frequency domain characteristics of the mutual inductor signal corresponding to the harmonic distortion.

[0132] In the embodiment of the present invention, the harmonic distortion evaluation and analysis is performed based on the previously obtained ratio of each harmonic component of the primary and secondary side signals and the obtained total distortion rate of the harmonic components, and the evaluation standard is set. When the total distortion rate is less than 5% and the ratio of each harmonic is within a reasonable range (such as the odd harmonic ratio does not exceed 10%, and the even harmonic ratio does not exceed 5%), the harmonic distortion is considered to be small; otherwise, the harmonic distortion is large. Taking the data of 8% total distortion rate of the primary side signal, 5.1% of the third harmonic ratio, and 3.1% of the fifth harmonic ratio as an example, due to the total distortion rate If the harmonic distortion exceeds 5% and the proportion of some harmonics exceeds the reasonable range, it is determined that the primary-side signal has large harmonic distortion. The secondary-side signal is evaluated in the same way. The frequency domain characteristics of the transformer signal corresponding to the harmonic distortion are summarized based on the evaluation results. For example, when harmonic distortion exists in the transformer, it is manifested in the frequency domain as an abnormal increase in the amplitude at a specific harmonic frequency, disordered coupling relationships in the harmonic component vector field, and complex and irregular energy transfer paths. These characteristics provide an important reference for evaluating the high-precision performance of low-impedance voltage transformers and help determine whether the working status of the transformer in a harmonic environment meets the requirements.

[0133] Furthermore, step S24 includes the following steps:

[0134] Step S241: obtaining the mutual inductor signal distribution corresponding to each spatial point through the original test signal of the mutual inductor, and calculating the signal distribution curvature corresponding to each spatial point according to the mutual inductor signal distribution corresponding to each spatial point;

[0135] In the embodiment of the present invention, the original test signal of the mutual inductor is collected by using a digital oscilloscope at a sampling rate of 100kHz and converted into a sequence of 100,000 discrete sampling points. Each sampling point corresponds to a "space point" in the time dimension, and the signal amplitude of the sampling point is used as the signal strength attribute of the point. For example, at the 1000th sampling point, the signal amplitude is 98V, and at the 1001st sampling point, the signal amplitude becomes 98.5V. In order to calculate the signal distribution curvature corresponding to each spatial point, the second-order difference approximation method is used. For the discrete signal sequence ( Indicates the sampling point number), in the The curvature of the signal distribution at a spatial point The calculation formula is: , taking the 1000th sampling point as an example, it is known that , , , substituting into the formula we can get: ≈0.003, and all 100,000 sampling points are calculated in turn to obtain the signal distribution curvature corresponding to each spatial point, forming a complete curvature data sequence, which reflects the change in the curvature degree of the signal in the time dimension.

[0136] Step S242: constructing a curvature tensor field of the transformer original test signal based on the corresponding signal distribution curvature at each spatial point to generate a transformer test signal curvature tensor field;

[0137] In the embodiment of the present invention, a curvature tensor field is constructed based on the previously obtained signal distribution curvature corresponding to each spatial point. The time dimension of the signal is regarded as a one-dimensional space, and each sampling point is regarded as a node in the space. For each node (i.e., sampling point), a second-order tensor is constructed. The tensor contains the curvature information of the signal at that point in the time dimension and the curvature correlation information with the adjacent nodes. The specific construction method is: let The curvature of the node is , define the curvature tensor ( =1, because it is a one-dimensional space), where = , while considering the influence of adjacent nodes, introduce the cross term = = (reflecting the changing trend of curvature between adjacent nodes), It can be set as a parameter related to the curvature change rate according to actual needs. Here, =0, taking the 1000th sampling point as an example, it is known that , , ,but =0.003, = =0.001, =0, perform the above tensor construction operation on all sampling points, arrange these tensors in the time sequence of the sampling points, and generate the mutual inductor test signal curvature tensor field, which comprehensively describes the distribution and change characteristics of the signal curvature in the time dimension.

[0138] Step S243: determining a signal curvature mutation region of the original transformer test signal based on the transformer test signal curvature tensor field to obtain a corresponding signal mutation region frame on the transformer test signal;

[0139] In an embodiment of the present invention, based on the generated mutual inductor test signal curvature tensor field, the signal curvature mutation area is determined, and a curvature mutation threshold is set. This threshold is obtained by analyzing the curvature tensor field data of a large number of normal mutual inductor test signals and taking the mean plus 3 times the standard deviation. Assuming that after statistical analysis, the curvature mutation threshold is determined to be 0.01, each tensor node (i.e., sampling point) in the curvature tensor field is traversed, and the Frobenius norm (a measure of tensor size) of the tensor at each node is calculated. The formula is , taking the tensor of the 1000th sampling point For example, ≈0.0032. When the Frobenius norm of the tensor at a node exceeds the set curvature mutation threshold, 10 sampling points are extended forward and backward with the node as the center to define a signal mutation region frame. For example, at the 5000th sampling point, the calculated Frobenius norm of the tensor is 0.012, which exceeds the threshold of 0.01. Then the 4990th-5010th sampling points are determined to constitute a signal mutation region frame. By traversing and analyzing the entire curvature tensor field, multiple signal mutation region frames are obtained. These region frames accurately identify the locations where the signal curvature changes dramatically.

[0140] Step S244: Calculate the instantaneous offset of the signal according to the signal mutation region frame corresponding to the mutual inductor test signal to obtain the instantaneous offset of the mutual inductor signal.

[0141] In the embodiment of the present invention, the instantaneous signal offset calculation is performed based on the previously determined signal mutation region frame. For each signal mutation region frame, the amplitude and phase information of the primary side signal and the secondary side signal in the frame are extracted respectively. Assuming that in a certain signal mutation region frame (the 4990th to 5010th sampling points), at the 5000th sampling point, the amplitude of the primary side signal is =100V, Phase =15°; Secondary side signal amplitude =0.099V, phase =12°, the calculation of the instantaneous offset of the signal takes into account the amplitude and phase differences. First calculate the amplitude offset ,get =99.901V; then calculate the phase offset ,get =3°, and finally, the offset is calculated by the formula (in is the rated voltage of the transformer, assuming =100V) to calculate the instantaneous offset of the transformer signal, which is 99.902V. This calculation is performed for all key sampling points within the signal mutation region frame to obtain complete transformer signal instantaneous offset data. This data can accurately reflect the difference between the primary and secondary side signals at the moment of mutation, providing an important basis for evaluating the high-precision performance of low-impedance voltage transformers.

[0142] Furthermore, step S244 includes the following steps:

[0143] Calculate the signal energy flow density distribution corresponding to the mutual inductor test signal at each time point through the signal mutation region frame corresponding to the mutual inductor test signal;

[0144] In an embodiment of the present invention, the threshold detection method is first used to identify the signal mutation region frame of the transformer test signal obtained by the previous steps. The amplitude change threshold is set to 20% of the fundamental amplitude. When the change in the signal amplitude at a certain sampling point exceeds the threshold, the area of ​​10 sampling points before and after the point is defined as the signal mutation region frame. Assume that in a certain transformer test signal, the signal amplitude suddenly jumps from 98V of the fundamental amplitude to 120V at the 1000th sampling point, exceeding the threshold, so that the 990th to 1010th sampling points constitute a signal mutation region frame. At each time point (i.e., each sampling point), the signal energy flow density distribution is calculated based on the Poynting vector theory. For the voltage signal and the corresponding current signal (obtained through Rogowski coil and other equipment), at each sampling point, the signal energy flow density The calculation formula is In the above signal mutation region frame, the voltage at the 990th sampling point is 98V, and the corresponding current is 1A after conversion. The signal energy flow density at this point is At the 991st sampling point, the voltage becomes 99V, the current is 1.01A, and the energy flow density is According to this method, the signal energy flow density corresponding to each time point in the signal mutation area frame is calculated, and the energy flow density distribution data in the area frame is obtained, which clearly shows the distribution state of energy at different times when the signal mutation occurs.

[0145] Preferably, an energy divergence operation is performed according to the signal energy flow density distribution corresponding to the mutual inductor test signal at each time point to obtain the signal energy divergence value corresponding to the mutual inductor test signal at each time point;

[0146] In the embodiment of the present invention, the signal energy flow density distribution corresponding to each time point is obtained. , perform energy divergence calculation, use numerical difference method to approximate energy divergence, for discrete signal energy flow density sequence, at a certain time point At the signal energy dispersion value The calculation formula is ,in is the sampling time interval (assuming the sampling frequency is 100kHz, then s), within the above signal mutation region frame, for the 991st sampling point, , , then the signal energy divergence value at this point =100000W / (m 2 s), according to this calculation method, energy divergence calculation is performed on each time point in the signal mutation region frame to obtain the signal energy divergence value sequence corresponding to each time point. These values ​​reflect the degree of change of signal energy in the time dimension. For example, at the key time point of signal mutation, the energy divergence value will have an obvious peak, which provides a quantitative basis for the subsequent determination of the instantaneous offset position of the signal.

[0147] Preferably, the signal instantaneous offset position of the signal mutation region frame corresponding to the mutual inductor test signal is determined based on the signal energy divergence value corresponding to the mutual inductor test signal at each time point, so as to calibrate the position of the signal energy flow drastically changing determined by the signal energy divergence value as the instantaneous offset position of the corresponding signal;

[0148] In the embodiment of the present invention, the instantaneous offset position of the signal is determined for the signal mutation region frame on the mutual inductor test signal based on the previously obtained signal energy dispersion value sequence, and the energy dispersion threshold is set to 3 times the average energy dispersion value (determined by calculating the average value of all energy dispersion values ​​in the signal mutation region frame). When the signal energy dispersion value at a certain time point exceeds the threshold, the position of the point is determined to be the position where the signal energy flow changes drastically, and it is calibrated as the instantaneous offset position of the corresponding signal. Assuming that in the above-mentioned signal mutation region frame, the average energy dispersion value calculated is 10000W / (m 2 s), the threshold is 30000W / (m 2 s), the energy divergence value of the 995th sampling point reached 35000 W / (m 2 s), exceeds the threshold, the position of the 995th sampling point is calibrated as the instantaneous offset position of the signal. By traversing the energy divergence values ​​of all time points in the signal mutation region frame, multiple possible instantaneous offset positions of the signal are accurately determined. These positions clearly indicate the moments when the signal energy changes rapidly during the mutation process, which helps to deeply analyze the dynamic characteristics of the signal.

[0149] Preferably, based on the instantaneous offset position of the corresponding signal, the signal instantaneous offset calculation is performed between the primary side and secondary side signal distributions at the same position on the original test signal of the mutual inductor to obtain the instantaneous offset of the mutual inductor signal.

[0150] In the embodiment of the present invention, based on the instantaneous offset position of the corresponding signal previously determined, the signal instantaneous offset calculation is performed on the primary and secondary side signal distributions at the same position on the original test signal of the mutual inductor, and at the instantaneous offset position, the primary side signal amplitude is obtained respectively. and the secondary side signal amplitude , and their corresponding phases and , the calculation of the instantaneous offset of the signal takes into account the amplitude and phase differences, and first calculates the amplitude offset , and then calculate the phase offset , and finally through the formula (in The instantaneous offset of the transformer signal is calculated by calculating the rated voltage of the transformer. For example, at a certain instantaneous offset position, the amplitude of the primary side signal is =100V, secondary side signal amplitude =0.099V, primary side signal phase =15°, secondary side signal phase =12°, rated voltage of transformer =100V, then the amplitude offset =99.901V, phase offset =3°, the instantaneous offset of the signal is ≈99.902V. By performing the above calculations on all instantaneous offset positions, the complete instantaneous offset data of the transformer signal is finally obtained. This data can accurately reflect the difference between the primary and secondary side signals at the moment of mutation, and provides a key indicator for evaluating the high-precision performance of low-impedance voltage transformers.

[0151] Furthermore, step S3 includes the following steps:

[0152] Step S31: Acquire environmental parameters corresponding to the low-impedance voltage transformer, including temperature, humidity, and vibration frequency;

[0153] In this embodiment of the present invention, an environmental parameter acquisition system is constructed using temperature, humidity, and vibration sensors. The temperature sensor has a measurement range of -55°C to +125°C with an accuracy of ±0.5°C; the humidity sensor has a measurement range of 0% to 100% RH with an accuracy of ±3% RH; and the vibration sensor has a measurement range of ±16g with a resolution of 0.003g / LSB. All three sensors are connected to a data acquisition instrument via an SPI interface, which simultaneously collects the three environmental parameters at a 10Hz sampling rate. During the low-impedance voltage transformer test, the temperature sensor was installed on the transformer housing, the humidity sensor was installed in a dry and ventilated location 1 meter away from the transformer, and the vibration sensor was fixed to the transformer base. During a one-hour test period, a total of 3600 sets of environmental parameter data were collected. For example, in a certain test, the recorded temperature data ranged from 23.5°C to 25.8°C, the humidity data ranged from 45% RH to 52% RH, and the vibration frequency was mainly concentrated at two frequency points, 50Hz and 100Hz, with amplitudes of 0.05g and 0.03g, respectively. The collected environmental parameter data was synchronized and aligned with the previously constructed mutual inductor signal feature matrix according to the timestamp, forming a joint data set containing environmental parameters and signal characteristics.

[0154] Step S32: by designing a three-layer fully connected neural network, including an input layer with 25 dimensions, a hidden layer with 128 nodes, and an output layer with 2 dimensions;

[0155] In an embodiment of the present invention, a three-layer fully connected neural network is designed and a dedicated neural network hardware accelerator is used to implement the network architecture. The input layer is set to 25 dimensions, corresponding to the 25 feature dimensions of the previously constructed transformer signal feature matrix (such as amplitude error, phase error, harmonic proportion, etc. at different frequency points). The hidden layer uses 128 neuron nodes, and each neuron uses a ReLU activation function (f (x) = max (0, x)) for nonlinear transformation. The output layer is set to 2 dimensions, corresponding to the two performance parameters of the frequency response and load characteristics of the low-impedance voltage transformer, respectively. The network training adopts a stochastic gradient descent optimization algorithm, the initial learning rate is set to 0.001, the batch size is set to 32, the number of training rounds is 1000, and L2 regularization is used to prevent overfitting. The regularization coefficient is initially set to 0.0001. The network weight initialization adopts the He normal distribution initialization method to ensure that the network has a suitable weight distribution at the beginning of training. During the training process, the network performance is monitored using a validation set. When the root mean square error of the validation set no longer decreases after 50 consecutive rounds, the early stopping mechanism is triggered to terminate the training.

[0156] Step S33: By using the Bayesian optimization algorithm to search for the optimal hyperparameters corresponding to the three-layer fully connected neural network, including the learning rate, regularization coefficient and the number of hidden layer nodes, and constructing the neural network model based on Bayesian optimization with the root mean square error as the optimization target, the mutual inductor signal feature matrix and environmental parameters are input at the same time and the corresponding mutual inductor performance joint prediction model is obtained through 5-fold cross-validation training. The attention mechanism is also introduced to give higher weights to the performance parameters corresponding to the frequency response and load characteristics, so that the output layer predicts and outputs a set of performance parameters corresponding to the frequency response and load characteristics.

[0157] In an embodiment of the present invention, the hyperparameters of a three-layer fully connected neural network are optimized using a Bayesian optimization algorithm. The optimization goal is to minimize the root mean square error (RMSE) of the model on the validation set. The hyperparameter search space is set as follows: the learning rate range is 0.0001 to 0.01 (logarithmic scale), the regularization coefficient range is 0.00001 to 0.001 (logarithmic scale), and the number of hidden layer nodes ranges from 64 to 256 (integer). During the Bayesian optimization process, a Gaussian process is used as a surrogate model, and the expected improvement (EI) is selected as the acquisition function. Initially, 10 sets of hyperparameters are randomly sampled, and then 50 iterations are performed. In each iteration, the Bayesian optimization algorithm predicts the next set of most promising hyperparameter combinations based on historical hyperparameter configurations and their corresponding RMSE values. For example, in one iteration, the algorithm recommends a learning rate of 0.0008, a regularization coefficient of 0.00005, and 144 hidden layer nodes. The neural network is trained using this set of hyperparameters, and the performance is evaluated in a 5-fold cross-validation. The dataset is randomly divided into 5 subsets. Four subsets are used to train the model each time, and one subset is used for validation. The average RMSE is calculated after repeating 5 times. To introduce the attention mechanism, an attention layer is added between the hidden layer and the output layer. This layer weights the input features and gives higher weights to features related to frequency response and load characteristics. The specific implementation is as follows: the correlation coefficient between each input feature and the target performance parameter is calculated, the correlation coefficient is used as the attention weight, and the hidden layer output is weighted summed. For example, analysis found that the phase error at the 50Hz frequency point is highly correlated with the frequency response performance. This feature is given a larger weight in the attention layer. After 50 rounds of Bayesian optimization iterations, the optimal hyperparameter combination was obtained: learning rate 0.00075, regularization coefficient 0.00006, and number of hidden layer nodes 156. By using this set of hyperparameters and the attention mechanism to train the final model, a prediction accuracy of RMS=0.023 was achieved on the test set. A joint prediction model for transformer performance based on Bayesian optimization and attention mechanism was successfully constructed. This model can accurately predict the frequency response and load characteristics of low-impedance voltage transformers, providing data support for transformer performance evaluation.

[0158] Furthermore, the present invention also provides a high-precision performance test processing system for a low-impedance voltage transformer, which is used to execute the high-precision performance test processing method for a low-impedance voltage transformer as described above. The high-precision performance test processing system for a low-impedance voltage transformer includes:

[0159] The transformer test signal excitation module is used to generate a corresponding controllable test excitation signal through a tunable standard voltage source and a wide-band load network, and apply the controllable test excitation signal to the low-impedance voltage transformer to synchronously collect the corresponding transformer primary and secondary voltages; perform spatial electromagnetic cancellation based on the transformer primary and secondary voltages to generate the transformer original test signal;

[0160] The transformer characteristic matrix construction module is used to perform frequency domain feature decomposition on the original transformer test signal to obtain the transformer signal frequency domain features corresponding to the amplitude error, phase error, and harmonic distortion; calculate the instantaneous offset of the original transformer test signal to obtain the instantaneous offset of the transformer signal; and construct the transformer signal characteristic matrix based on the transformer signal frequency domain features and the transformer signal instantaneous offset;

[0161] The transformer performance prediction module is used to obtain the environmental parameters corresponding to the low-impedance voltage transformer. A neural network model based on Bayesian optimization is constructed, and the transformer signal feature matrix and environmental parameters are input to train the corresponding transformer performance joint prediction model, which predicts and outputs a set of performance parameters corresponding to the frequency response and load characteristics.

[0162] The performance calibration test evaluation module is used to generate a calibration compensation coefficient by comparing the performance parameter set with the preset standard threshold value, and inject the calibration compensation coefficient into the transformer performance joint prediction model to iteratively optimize the test until the error converges and test and evaluate, thereby generating a high-precision performance test report corresponding to the low-impedance voltage transformer.

[0163] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0164] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A high-precision performance test processing method for a low-impedance voltage transformer, characterized in that: The following steps are involved: Step S1: Generate a corresponding controllable test excitation signal through a tunable standard voltage source and a wide-band load network, and apply the controllable test excitation signal to a low-impedance voltage transformer to synchronously collect the corresponding transformer primary voltage and transformer secondary voltage; perform spatial electromagnetic cancellation based on the transformer primary voltage and transformer secondary voltage to generate the transformer original test signal; Step S2: performing frequency domain feature decomposition on the original test signal of the mutual inductor to obtain the frequency domain features of the mutual inductor signal corresponding to the amplitude error, phase error and harmonic distortion; Calculate the instantaneous offset of the original test signal of the transformer to obtain the instantaneous offset of the transformer signal; construct a transformer signal feature matrix based on the frequency domain characteristics of the transformer signal and the instantaneous offset of the transformer signal; Step S3: Obtaining environmental parameters corresponding to the low-impedance voltage transformer; constructing a neural network model based on Bayesian optimization, and inputting the transformer signal feature matrix and environmental parameters to train a corresponding transformer performance joint prediction model to predict and output a performance parameter set corresponding to the frequency response and load characteristics; Step S4: Generate a calibration compensation coefficient by comparing the performance parameter set with the preset standard threshold value, and inject the calibration compensation coefficient into the transformer performance joint prediction model to iteratively optimize the test until the error converges and test and evaluate, and generate a high-precision performance test report corresponding to the low-impedance voltage transformer.

2. The high-precision performance test processing method for a low-impedance voltage transformer according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Generate a sinusoidal excitation signal with an adjustable fundamental frequency by using a tunable standard voltage source to cover the rated operating frequency band corresponding to the low-impedance voltage transformer; Step S12: Building a broadband load network including resistive, inductive, and capacitive components, and configuring a load impedance adjustment for a sinusoidal excitation signal corresponding to an adjustable fundamental frequency based on the broadband load network to generate a corresponding controllable test excitation signal; Step S13: applying a controllable test excitation signal to the low-impedance voltage transformer, and synchronously collecting the corresponding transformer primary side voltage and transformer secondary side voltage using a symmetrical Rogowski coil and a high-precision resistor divider; Step S14: Design a differential sampling and common-mode suppression network circuit based on a symmetrical Rogowski coil and a high-precision resistor divider to perform spatial electromagnetic elimination on the primary side voltage and the secondary side voltage of the transformer to generate an original transformer test signal.

3. The high-precision performance test processing method for a low-impedance voltage transformer according to claim 2, characterized in that: Step S14 includes the following steps: Step S141: simulating the magnetic field distribution generated by the primary side voltage of the transformer and the corresponding coupling condition of the symmetrical Rogowski coil according to the law of electromagnetic induction for the symmetrical Rogowski coil, and calculating the mutual inductance coefficient of the symmetrical Rogowski coil in all directions to generate a spatial distribution field of the mutual inductance coefficient of the Rogowski coil; Step S142: calculating the input impedance and output impedance of the high-precision resistor divider at different frequencies based on the high-precision resistor divider and taking into account the parasitic parameters of the resistor elements and the frequency variation, and constructing a corresponding resistor divider impedance frequency response matrix according to the different frequencies; Step S143: Decomposing the interference electromagnetic field corresponding to the low-impedance voltage transformer into common-mode interference and differential-mode interference, and determining the distribution of the interference electromagnetic field in the space surrounding the low-impedance voltage transformer to generate an electromagnetic interference modal decomposition vector field, wherein the vector of each spatial point in the field includes the magnitude and direction of the common-mode interference and the differential-mode interference; Step S144: Designing a differential sampling and common-mode suppression network circuit based on the spatial distribution field of the Rogowski coil mutual inductance coefficient and the impedance frequency response matrix of the resistor divider in combination with the electromagnetic interference modal decomposition vector field, thereby optimizing the symmetrical Rogowski coil layout by utilizing the symmetrical structure corresponding to the symmetrical Rogowski coil to suppress common-mode interference and differential-mode interference, selecting appropriate impedance parameters based on the impedance frequency response corresponding to the high-precision resistor divider to match the differential sampling, and designing a corresponding filter circuit for the electromagnetic interference mode; Step S145: Connect the differential sampling and common-mode suppression network circuit to the sampling loops corresponding to the primary-side voltage and the secondary-side voltage of the transformer to perform spatial electromagnetic elimination on their output signals, so as to extract the effective signal components corresponding to the primary-side voltage and the secondary-side voltage of the transformer through differential sampling, and use the common-mode suppression network to suppress the common-mode noise and differential-mode noise generated by spatial electromagnetic interference. At the same time, the processed signal is filtered and amplified and modulated to remove residual high-frequency noise and low-frequency drift, and generate the original transformer test signal, which includes the primary-side signal and the secondary-side signal.

4. The high-precision performance test processing method for a low-impedance voltage transformer according to claim 3, characterized in that: Step S2 includes the following steps: Step S21: performing spectrum transformation and decomposition on the original test signal of the mutual inductor to generate amplitude distribution and phase distribution corresponding to the fundamental wave and each harmonic component in the primary and secondary side signals; Step S22: performing amplitude and phase error analysis on the amplitude distribution and phase distribution corresponding to the fundamental wave components in the primary and secondary side signals to obtain frequency domain characteristics of the mutual inductor signal corresponding to the amplitude error and phase error; Step S23: performing harmonic distortion evaluation analysis on the amplitude distribution and phase distribution corresponding to each harmonic component in the primary and secondary side signals to obtain the frequency domain characteristics of the mutual inductor signal corresponding to the harmonic distortion; Step S24: calculating the instantaneous offset of the original test signal of the transformer to obtain the instantaneous offset of the transformer signal; Step S25: constructing a mutual inductor signal characteristic matrix according to the mutual inductor signal frequency domain characteristics and the mutual inductor signal instantaneous offset.

5. The high-precision performance test processing method for a low-impedance voltage transformer according to claim 4, characterized in that: Step S22 includes the following steps: Performing time-frequency synchronization processing on the amplitude distribution and phase distribution corresponding to the fundamental wave components in the primary and secondary side signals to obtain the amplitude distribution and phase distribution corresponding to the primary and secondary side signal components in the same time-frequency range; Amplitude error statistics are performed between the amplitude distributions corresponding to the primary and secondary side signal components in the same time-frequency range to obtain the frequency domain characteristics of the mutual inductor signal corresponding to the amplitude error between the primary and secondary side signals; The phase error is calculated based on the phase distributions corresponding to the primary and secondary side signal components in the same time-frequency range to obtain the frequency domain characteristics of the mutual inductor signal corresponding to the phase error between the primary and secondary side signals.

6. The high-precision performance test processing method for a low-impedance voltage transformer according to claim 4, characterized in that: Step S23 includes the following steps: Step S231: performing harmonic proportion statistics calculation based on the amplitude distribution and phase distribution corresponding to each harmonic component in the primary and secondary side signals to obtain the proportion between each harmonic component of the primary and secondary side signals; Step S232: constructing a vector field for each harmonic component vector based on the coupling relationship between adjacent harmonic component vectors by treating the amplitude distribution and phase distribution corresponding to each harmonic component in the primary and secondary side signals as harmonic component vectors having corresponding amplitude and phase attributes, and calculating the coupling relationship between adjacent harmonic component vectors, thereby generating harmonic component vector fields corresponding to the primary and secondary side signals; Step S233: Calculate the harmonic energy proportion and harmonic energy transfer path of each harmonic component in the entire signal cycle according to the harmonic component vector field corresponding to the primary and secondary side signals; Step S234: quantifying the amplitude-phase coupling distortion rate of the amplitude distribution and phase distribution corresponding to each harmonic component in the primary and secondary side signals based on the harmonic energy proportion corresponding to each harmonic component in the entire signal cycle and the harmonic energy transfer path, to obtain the total distortion rate of the harmonic components corresponding to the primary and secondary side signals; Step S235: performing harmonic distortion evaluation analysis based on the proportion of each harmonic component of the primary and secondary side signals and the corresponding total distortion rate of the harmonic components to obtain the frequency domain characteristics of the mutual inductor signal corresponding to the harmonic distortion.

7. The high-precision performance test processing method for a low-impedance voltage transformer according to claim 4, characterized in that: Step S24 includes the following steps: Step S241: obtaining the mutual inductor signal distribution corresponding to each spatial point through the original test signal of the mutual inductor, and calculating the signal distribution curvature corresponding to each spatial point according to the mutual inductor signal distribution corresponding to each spatial point; Step S242: constructing a curvature tensor field of the transformer original test signal based on the corresponding signal distribution curvature at each spatial point to generate a transformer test signal curvature tensor field; Step S243: determining a signal curvature mutation region of the original transformer test signal based on the transformer test signal curvature tensor field to obtain a corresponding signal mutation region frame on the transformer test signal; Step S244: Calculate the instantaneous offset of the signal according to the signal mutation region frame corresponding to the mutual inductor test signal to obtain the instantaneous offset of the mutual inductor signal.

8. The high-precision performance test processing method for a low-impedance voltage transformer according to claim 7, characterized in that: Step S244 includes the following steps: Calculate the signal energy flow density distribution corresponding to the mutual inductor test signal at each time point through the signal mutation region frame corresponding to the mutual inductor test signal; Performing energy divergence calculation according to the signal energy flow density distribution corresponding to the mutual inductor test signal at each time point to obtain the signal energy divergence value corresponding to the mutual inductor test signal at each time point; Determine the instantaneous offset position of the signal on the signal mutation region frame corresponding to the mutual inductor test signal based on the signal energy divergence value corresponding to each time point of the mutual inductor test signal, so as to calibrate the position of the drastic change of the signal energy flow determined by the signal energy divergence value as the instantaneous offset position of the corresponding signal; Based on the instantaneous offset position of the corresponding signal, the signal instantaneous offset between the primary side and secondary side signal distributions at the same position on the original test signal of the mutual inductor is calculated to obtain the instantaneous offset of the mutual inductor signal.

9. The high-precision performance test processing method for a low-impedance voltage transformer according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: Acquire environmental parameters corresponding to the low-impedance voltage transformer, including temperature, humidity, and vibration frequency; Step S32: by designing a three-layer fully connected neural network, including an input layer with 25 dimensions, a hidden layer with 128 nodes, and an output layer with 2 dimensions; Step S33: By using the Bayesian optimization algorithm to search for the optimal hyperparameters corresponding to the three-layer fully connected neural network, including the learning rate, regularization coefficient and the number of hidden layer nodes, and constructing the neural network model based on Bayesian optimization with the root mean square error as the optimization target, the mutual inductor signal feature matrix and environmental parameters are input at the same time and the corresponding mutual inductor performance joint prediction model is obtained through 5-fold cross-validation training. The attention mechanism is also introduced to give higher weights to the performance parameters corresponding to the frequency response and load characteristics, so that the output layer predicts and outputs a set of performance parameters corresponding to the frequency response and load characteristics.

10. A high-precision performance test processing system for a low-impedance voltage transformer, characterized in that: For executing the high-precision performance test processing method of a low-impedance voltage transformer according to claim 1, the high-precision performance test processing system of the low-impedance voltage transformer comprises: The transformer test signal excitation module is used to generate a corresponding controllable test excitation signal through a tunable standard voltage source and a wide-band load network, and apply the controllable test excitation signal to the low-impedance voltage transformer to synchronously collect the corresponding transformer primary and secondary voltages; perform spatial electromagnetic cancellation based on the transformer primary and secondary voltages to generate the transformer original test signal; The transformer characteristic matrix construction module is used to perform frequency domain feature decomposition on the original transformer test signal to obtain the transformer signal frequency domain features corresponding to the amplitude error, phase error, and harmonic distortion; calculate the instantaneous offset of the original transformer test signal to obtain the instantaneous offset of the transformer signal; and construct the transformer signal characteristic matrix based on the transformer signal frequency domain features and the transformer signal instantaneous offset; The transformer performance prediction module is used to obtain the environmental parameters corresponding to the low-impedance voltage transformer. A neural network model based on Bayesian optimization is constructed, and the transformer signal feature matrix and environmental parameters are input to train the corresponding transformer performance joint prediction model, which predicts and outputs a set of performance parameters corresponding to the frequency response and load characteristics. The performance calibration test evaluation module is used to generate a calibration compensation coefficient by comparing the performance parameter set with the preset standard threshold value, and inject the calibration compensation coefficient into the transformer performance joint prediction model to iteratively optimize the test until the error converges and test and evaluate, thereby generating a high-precision performance test report corresponding to the low-impedance voltage transformer.

Images