Low-impedance voltage transformer evaluation method based on real-time data adaptive driving

Abstract

The invention relates to the technical field of electrical variable evaluation, in particular to a low-impedance voltage transformer evaluation method based on instant data adaptive driving. The method comprises the following steps: generating a mutual inductor instant data set by collecting a mutual inductor primary side electrical signal and a mutual inductor secondary side electrical signal corresponding to a low-impedance voltage mutual inductor in real time and synchronously collecting corresponding temperature and humidity data and vibration data; performing dynamic quality evaluation screening and variational modal feature extraction on the real-time data set of the mutual inductor to obtain a dynamic feature vector of the mutual inductor; constructing a mutual inductor self-adaptive evaluation model based on the mutual inductor dynamic feature vector, and evaluating and generating mutual inductor real-time evaluation state data; and obtaining historical fault sample data corresponding to the low-impedance voltage transformer and carrying out real-time comparison fault early warning to generate a real-time state fault early warning signal corresponding to the low-impedance voltage transformer. The evaluation efficiency of the operation state of the low-impedance voltage transformer can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electric variable evaluation, and in particular to a low-impedance voltage transformer evaluation method based on real-time data adaptive driving. Background Art

[0002] Low-impedance voltage transformers (VTs) are 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 of power systems, which places high demands on equipment, the accuracy, stability, and reliability of voltage transformers are crucial. Currently, power systems have increasingly stringent performance requirements for low-impedance voltage transformers, and traditional testing methods are unable to comprehensively and accurately evaluate their operating performance, especially their dynamic response capabilities under different operating conditions. However, existing low-impedance voltage transformer evaluation methods typically use static testing methods to evaluate the accuracy and stability of the voltage transformer by measuring the ratio of its output signal to its input signal. While this method can provide certain performance data, it fails to account for various dynamic factors in the power system, such as voltage fluctuations and load changes. As a result, the evaluation results cannot fully and timely reflect the voltage transformer's performance under different dynamic conditions, thereby reducing the efficiency of voltage transformer evaluation. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide a low-impedance voltage transformer evaluation method based on real-time data adaptive driving to solve at least one of the above technical problems.

[0004] To achieve the above object, a low-impedance voltage transformer evaluation method based on real-time data adaptive driving includes the following steps: Step S1: The primary side electrical signal and the secondary side electrical signal of the low-impedance voltage transformer corresponding to the low-impedance voltage transformer are collected in real time through the Rogowski coil and the resistor divider, and the temperature and humidity sensor and the vibration sensor are simultaneously connected to collect the corresponding temperature and humidity data and vibration data to generate a real-time transformer data set; Step S2: Performing dynamic quality assessment and screening on the transformer real-time data set to generate a quality-screened valid real-time data set; performing variational modal feature extraction on the valid real-time data set to obtain a transformer dynamic feature vector corresponding to the amplitude error, phase offset, and harmonic distortion; Step S3: constructing a transformer adaptive evaluation model based on the transformer dynamic feature vector, and using the effective real-time data set to drive the iterative update of the corresponding transformer adaptive evaluation model parameters, and evaluating and generating transformer real-time evaluation status data; Step S4: Obtain historical fault sample data corresponding to the low-impedance voltage transformer, and perform instant fault warning based on the real-time evaluation status data of the transformer and the historical fault sample data to generate an instant status fault warning signal corresponding to the low-impedance voltage transformer.

[0005] Furthermore, step S1 includes the following steps: Step S11: collecting the primary side electrical signal of the transformer corresponding to the low-impedance voltage transformer in real time through the differential sampling structure corresponding to the Rogowski coil; Step S12: constructing a four-level resistor divider network by four resistor dividers to collect the secondary side electrical signal of the transformer corresponding to the low-impedance voltage transformer in real time; Step S13: collecting temperature, humidity and vibration data corresponding to the surrounding environment of the low-impedance voltage transformer by synchronously accessing the temperature and humidity sensor and the vibration sensor; Step S14: The primary side electrical signal of the transformer, the secondary side electrical signal of the transformer, the temperature and humidity data, and the vibration data corresponding to the low-impedance voltage transformer are clock-synchronized through a field programmable gate array and merged into the same data set to generate a transformer real-time data set.

[0006] Furthermore, step S11 includes the following steps: Finite element simulation was used to optimize the spacing between the Rogowski coils and the thickness of the shielding layer above them. Based on the optimized spacing and the thickness of the shielding layer on the Rogowski coil, the leakage magnetic interference suppression evaluation is performed to obtain the primary side leakage magnetic interference suppression ratio corresponding to the Rogowski coil; The electrical sampling accuracy corresponding to the low-impedance working condition is determined based on the primary-side leakage magnetic interference suppression ratio corresponding to the Rogowski coil, and the primary-side electrical signal of the transformer corresponding to the low-impedance voltage transformer is collected in real time through the differential sampling structure corresponding to the Rogowski coil based on the electrical sampling accuracy.

[0007] Furthermore, step S2 includes the following steps: Step S21: performing soft threshold denoising processing on the electrical signal in the transformer real-time data set to generate a transformer denoised electrical signal; Step S22: Calculate the cross-correlation coefficient between the voltage and current signals of the transformer de-noised electrical signal. When the cross-correlation coefficient is less than 0.95, trigger the signal re-sampling mechanism and iterate the de-noising process to generate a transformer re-correlated electrical signal. Step S23: Calculate the corresponding environmental change rate based on the environmental parameters corresponding to temperature, humidity and vibration in the real-time data set of the mutual inductor, and perform abnormal environment elimination processing on the environmental parameters corresponding to temperature, humidity and vibration based on the environmental change rate to obtain the mutual inductor filtered abnormal environment data; Step S24: dynamically evaluating and screening the transformer's re-correlated electrical signals and transformer's filtered environmental data based on the data integrity and accuracy scores to generate a quality-screened valid real-time data set; Step S25: performing variational modal feature extraction on the valid real-time data set to obtain a mutual inductor dynamic feature vector corresponding to the amplitude error, phase offset, and harmonic distortion.

[0008] Furthermore, step S21 includes the following steps: The electrical signal in the transformer real-time data set is decomposed into five layers using db10 wavelet packets to generate five layers of transformer real-time electrical signal frequency sub-bands. Obtain corresponding transformer electrical signal sub-band kurtosis through the 5-layer transformer instantaneous electrical signal frequency sub-band, and perform noisy sub-band identification on the 5-layer transformer instantaneous electrical signal frequency sub-band based on the transformer electrical signal sub-band kurtosis to generate transformer electrical signal noisy frequency sub-band; Soft threshold denoising is performed based on the noisy frequency sub-band of the transformer electrical signal, and the electrical signal is reconstructed with the other transformer real-time electrical signal frequency sub-bands to generate a transformer denoised electrical signal.

[0009] Furthermore, step S25 includes the following steps: Step S251: performing variational modal decomposition on the electrical signal in the valid instantaneous data set to decompose the electrical signal into 8-12 eigenmodes, and screening 3-5 principal components related to low impedance characteristics through correlation coefficients; Step S252: Obtain the signal amplitude distribution and signal phase distribution corresponding to each principal component, and perform amplitude error and phase offset statistics based on the signal amplitude distribution and signal phase distribution corresponding to each principal component to obtain the amplitude error and phase offset corresponding to each principal component; Step S253: obtaining corresponding harmonic components from the electrical signal in the valid real-time data set, and performing harmonic distortion evaluation on the harmonic components corresponding to the electrical signal based on the environmental data in the valid real-time data set to obtain harmonic distortion corresponding to each harmonic component; Step S254: combining the amplitude error and phase offset corresponding to each main component and the harmonic distortion corresponding to each subharmonic component into the same eigenvector to obtain a mutual inductor dynamic eigenvector including the amplitude error, phase offset and harmonic distortion.

[0010] Furthermore, the amplitude error and phase offset statistics according to the signal amplitude distribution and signal phase distribution corresponding to each principal component described in step S252 include the following steps: Perform time-frequency synchronization on the signal amplitude distribution and signal phase distribution corresponding to each principal component to obtain the amplitude distribution and phase distribution corresponding to each principal component at the same time-frequency; Perform amplitude error statistics between the amplitude distributions corresponding to each principal component at the same time and frequency to obtain the amplitude error corresponding to each principal component; The phase differences between the main components are obtained by the phase distributions corresponding to the main components at the same time-frequency. Based on the phase differences between the main components, the phase distributions corresponding to the main components at the same time-frequency are statistically analyzed to obtain the phase shifts corresponding to the main components.

[0011] Furthermore, the harmonic distortion evaluation of each harmonic component corresponding to the electrical signal based on the environmental data in the valid real-time data set in step S253 includes the following steps: Based on the environmental data in the valid real-time data set, an environmental-electrical offset coupling analysis is performed on each harmonic component corresponding to the electrical signal to analyze the gradient and vibration frequency corresponding to temperature and humidity. Based on the gradient and vibration frequency corresponding to temperature and humidity, an offset coupling evaluation is performed between the amplitude or phase corresponding to each harmonic component to obtain the offset coupling coefficient between each temperature, humidity and vibration component and each harmonic amplitude or phase; Based on the environmental data in the effective real-time data set, the environmental-electrical mutual information of each harmonic component corresponding to the electrical signal is calculated to obtain the mutual information value between each component of temperature, humidity and vibration and the amplitude or phase of each harmonic; Based on the offset coupling coefficient and mutual information value between the temperature, humidity, and vibration components and the amplitude or phase of each harmonic, the harmonic distortion of each harmonic component corresponding to the electrical signal is evaluated to obtain the harmonic distortion corresponding to each harmonic component, including the amplitude deviation rate, phase drift, and distortion energy ratio under the influence of the environment.

[0012] Furthermore, step S3 includes the following steps: Step S31: constructing a transformer adaptive evaluation model using a Kalman filter framework, taking the transformer dynamic eigenvector as a state variable, and using the electrical and environmental quantities in the valid real-time data set as input variables to drive the update of the state transfer matrix and observation matrix corresponding to the transformer adaptive evaluation model, while achieving recursive estimation and prediction of the state variables; Step S32: Acquire real-time status data corresponding to the low-impedance voltage transformer, and construct a parameter adaptation mechanism based on gradient descent. When the deviation between the real-time status data and the recursive estimation prediction exceeds a preset threshold, trigger an iterative update of the Kalman gain parameters of the transformer adaptive evaluation model to re-evaluate and generate the corresponding transformer real-time evaluation result. Step S33: quantifying the uncertainty state corresponding to the real-time evaluation result of the transformer by using a Bayesian inference method to evaluate and generate real-time evaluation state data of the transformer including corresponding confidence levels.

[0013] Furthermore, step S4 includes the following steps: Step S41: Acquire historical fault sample data corresponding to the low-impedance voltage transformer; Step S42: performing a comparative analysis of any two instantaneous errors between the real-time evaluation state data of the transformer and the historical fault sample data to obtain the evaluation state errors between any two real-time states of the transformer and the states corresponding to the historical faults; Step S43: Calculate the corresponding transformer evaluation state error change rate based on the evaluation state error between any two real-time states of the transformer and the states corresponding to historical faults, and perform an error fault warning on the low-impedance voltage transformer based on the transformer evaluation state error change rate. When the transformer evaluation state error change rate is greater than 15%, an abnormal fault warning is triggered to generate an immediate state fault warning signal corresponding to the low-impedance voltage transformer.

[0014] Beneficial effects of the present invention: The low-impedance voltage transformer evaluation method based on real-time data adaptive drive proposed in the present invention has the beneficial effect of acquiring real-time data of the low-impedance voltage transformer through the Rogowski coil and the resistor divider compared with the prior art, and acquiring the electrical signals of the primary and secondary sides thereof. These signals are crucial for analyzing the working status of the transformer. The Rogowski coil helps to reflect the electrical characteristics of the transformer by accurately sensing the current signal, while the resistor divider can effectively reduce the voltage signal appropriately so that it can be processed and analyzed by the subsequent system. The real-time acquisition of the primary and secondary side electrical signals can accurately reflect the fluctuations of the transformer's current, voltage and other parameters, and help evaluate its working status. At the same time, the acquisition functions of temperature, humidity, and vibration sensors further enhance the comprehensive perception of the transformer environment. Temperature and humidity data can reflect the impact of environmental factors on the performance of the transformer. For example, high temperature or humidity may cause the insulation performance of the transformer to deteriorate, thereby affecting the stability of the electrical signal. Vibration data can be used to monitor the mechanical operating status of the transformer, such as poor electrical contact, equipment aging, or abnormal vibration. The collection of this data can provide a basis for subsequent dynamic evaluation and fault diagnosis, ensuring that the status information of the transformer is more comprehensive and accurate. Secondly, by dynamically evaluating and screening the quality of the transformer's real-time data set, we ensure that only high-quality valid data is retained. This process is a key link in data preprocessing. Through quality screening, abnormal, noisy, and inaccurate data are eliminated, providing accurate input for subsequent analysis. After data screening, variational modal feature extraction is performed to further extract the dynamic characteristics of the transformer. Variational modal analysis (VMD) can perform multi-scale decomposition of the signal and extract features closely related to the transformer's operating state, such as amplitude error, phase offset, and harmonic distortion. These features reflect the transformer's performance under different operating conditions. For example, harmonic distortion represents the transformer's electrical noise during operation, while amplitude error and phase offset are early signs of equipment failure. This can transform complex electrical signals into more representative dynamic feature vectors, providing key information for the subsequent adaptive evaluation model. Then, based on the previously extracted transformer dynamic feature vectors, an adaptive evaluation model is constructed. This model can evaluate the transformer's real-time status and provide the current health status of the device. By introducing an adaptive algorithm, the model can dynamically adjust the evaluation criteria according to different operating conditions to adapt to possible changes in the transformer's state. This feature makes the model highly real-time and adaptable, and can fully and comprehensively reflect the performance of the voltage transformer under different dynamic conditions. As valid real-time data sets are continuously input, the model parameters are updated through an iterative process, thereby achieving self-optimization. This adaptive update mechanism not only improves the model's prediction accuracy, but also automatically adjusts its evaluation strategy when the transformer status changes, ensuring more accurate evaluation results, thereby improving the evaluation efficiency of voltage transformers.Finally, by comparing and analyzing the transformer's real-time evaluation status data with historical fault sample data, immediate fault warnings can be achieved. By comparing with historical fault samples, potential fault signs in the real-time data can be identified. For example, if the current transformer's dynamic characteristics are similar to a characteristic in a historical fault sample, it indicates that the transformer is at risk of failure. In this case, a fault warning signal will be triggered, prompting operators to conduct further inspection and repair. This process not only improves the accuracy of fault diagnosis but also enables real-time monitoring of equipment health, thereby realizing intelligent diagnosis of transformers, greatly improving maintenance efficiency and reducing equipment downtime. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] 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: Figure 1 Schematic diagram of the steps of the low-impedance voltage transformer evaluation method based on real-time data adaptive driving of the present invention; Figure 2 for Figure 1 Detailed step flow diagram of step S1; Figure 3 for Figure 1 Detailed step flow chart of step S2 in FIG. DETAILED DESCRIPTION

[0016] 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.

[0017] 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.

[0018] 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.

[0019] To achieve this, please refer to Figures 1 to 3 The present invention provides a low-impedance voltage transformer evaluation method based on real-time data adaptive driving, the method comprising the following steps: Step S1: The primary side electrical signal and the secondary side electrical signal of the low-impedance voltage transformer corresponding to the low-impedance voltage transformer are collected in real time through the Rogowski coil and the resistor divider, and the temperature and humidity sensor and the vibration sensor are simultaneously connected to collect the corresponding temperature and humidity data and vibration data to generate a real-time transformer data set; Step S2: Performing dynamic quality assessment and screening on the transformer real-time data set to generate a quality-screened valid real-time data set; performing variational modal feature extraction on the valid real-time data set to obtain a transformer dynamic feature vector corresponding to the amplitude error, phase offset, and harmonic distortion; Step S3: constructing a transformer adaptive evaluation model based on the transformer dynamic feature vector, and using the effective real-time data set to drive the iterative update of the corresponding transformer adaptive evaluation model parameters, and evaluating and generating transformer real-time evaluation status data; Step S4: Obtain historical fault sample data corresponding to the low-impedance voltage transformer, and perform instant fault warning based on the real-time evaluation status data of the transformer and the historical fault sample data to generate an instant status fault warning signal corresponding to the low-impedance voltage transformer.

[0020] In the embodiment of the present invention, please refer to Figure 1 FIG. 1 is a flow chart of the steps of a low-impedance voltage transformer evaluation method based on real-time data adaptive driving according to the present invention. In this example, the low-impedance voltage transformer evaluation method based on real-time data adaptive driving includes the following steps: Step S1: The primary side electrical signal and the secondary side electrical signal of the low-impedance voltage transformer corresponding to the low-impedance voltage transformer are collected in real time through the Rogowski coil and the resistor divider, and the temperature and humidity sensor and the vibration sensor are simultaneously connected to collect the corresponding temperature and humidity data and vibration data to generate a real-time transformer data set; In an embodiment of the present invention, a dedicated data acquisition system is used to collect low-impedance voltage transformer data in real time. The Rogowski coil has a measurement range of 0.1A-1000A and a frequency response of 0.1Hz-1MHz, which is used to collect the primary current signal of the transformer. The resistor divider has a voltage divider ratio of 1000:1 and an accuracy of ±0.05%, which is used to collect the primary voltage signal of the transformer. The secondary electrical signal is collected using a high-precision digital multimeter with a sampling rate of 100kHz and a resolution of 6.5 bits. The temperature and humidity sensor has a measurement range of -55℃-125℃, 0-100% RH, and an accuracy of ±0.3℃ and ±2%. RH; The vibration sensor has a measurement range of ±16g, a resolution of 10μg / LSB, and a sampling rate of 4kHz. All sensors are connected to the data acquisition instrument through a dedicated interface module, which supports synchronous sampling. The sampling frequency is uniformly set to 10kHz. The acquisition instrument has a built-in GPS module to ensure that all signal sampling times are synchronized with a time accuracy of ±1μs. During the acquisition process, a set of data is collected every 10ms. Each set of data includes primary side voltage and current, secondary side voltage and current, temperature and humidity, and vibration data. The data is collected continuously for 10 seconds to form a data frame, which is stored in the SD card built into the acquisition instrument to generate a real-time data set for the mutual inductor.

[0021] Step S2: Performing dynamic quality assessment and screening on the transformer real-time data set to generate a quality-screened valid real-time data set; performing variational modal feature extraction on the valid real-time data set to obtain a transformer dynamic feature vector corresponding to the amplitude error, phase offset, and harmonic distortion; In this embodiment of the present invention, a dynamic quality assessment and screening of a real-time transformer data set is performed. First, a data integrity check is performed. A dedicated algorithm is used to calculate the proportion of missing values in each data frame. If a data frame contains more than 5% missing values, it is marked as invalid. Next, an accuracy assessment is performed by comparing the secondary-side electrical signal with a reference signal generated by a standard signal source (Fluke 5720A, accuracy of ±0.002%). The amplitude and phase errors are calculated. If the error exceeds ±0.5%, the data frame is deemed inaccurate. A de-noised electrical signal is generated using soft thresholding. A five-layer decomposition is performed using the db10 wavelet, and the median absolute deviation method is used to calculate the threshold. The cross-correlation coefficient of the denoised voltage and current signals was calculated. If it was less than 0.95, the signal resampling mechanism was triggered, and data was recollected using the backup acquisition channel. The environmental change rate was calculated for the temperature, humidity, and vibration data, and abnormal data with change rates exceeding the thresholds (temperature ±0.1°C / s, humidity ±5% RH / s, and vibration ±0.5g / s) were eliminated. Finally, data with a score higher than 80 was screened out using the data integrity and accuracy scoring system (integrity weight 0.4, accuracy weight 0.6) to generate a valid real-time data set. Variational modal feature extraction was performed on the valid real-time data set, and the electrical signal was decomposed into 10 intrinsic mode functions using the VMD algorithm. The number of decomposed modes was set to 10, and the balance parameter α was set to 2000. The correlation coefficient of each intrinsic mode with the low-impedance reference signal was calculated to screen out three principal components. The amplitude and phase distributions of the principal components were obtained by FFT transformation. The amplitude error and phase offset were calculated by comparing them with the ideal state. At the same time, the 1st to 15th harmonic components are extracted and combined with environmental data to evaluate harmonic distortion. The amplitude deviation rate, phase drift, and distortion energy ratio are calculated and finally merged to generate a dynamic feature vector of the mutual inductor containing 11 dimensions.

[0022] Step S3: constructing a transformer adaptive evaluation model based on the transformer dynamic feature vector, and using the effective real-time data set to drive the iterative update of the corresponding transformer adaptive evaluation model parameters, and evaluating and generating transformer real-time evaluation status data; In an embodiment of the present invention, a Kalman filter adaptive evaluation model is constructed based on the dynamic eigenvector of the mutual inductor. The dynamic eigenvector is used as the state variable, and electrical and environmental quantities are used as input and observation variables. The initial state transfer matrix is set to the identity matrix, and the observation matrix is determined based on the physical relationship between the electrical quantity and the state variable. The process noise covariance matrix is initialized to 0.01 times the identity matrix, and the measurement noise covariance matrix is initialized to 0.05 times the identity matrix. Model parameters are iteratively updated through recursive calculation. At each sampling moment, the state is first predicted based on the current input and observation data, and then the state is updated using the Kalman gain. The state transfer matrix and the observation matrix are also updated simultaneously. When the deviation between the real-time state data and the recursive estimate exceeds 0.1 (normalized norm), a gradient descent-based parameter adaptation mechanism is triggered. The learning rate is set to 0.01, and the deviation is reduced to below 0.08 after 5 iterations. A Bayesian inference method is used to quantify the uncertainty of the evaluation results. The real-time evaluation results are treated as Gaussian distributed random variables, the Kalman filter prediction value is used as the prior probability, and the real-time state data is used as the likelihood function. Using the Markov Chain Monte Carlo method, 1000 sample points are generated, and the mean and covariance of the posterior distribution are calculated to obtain a 95% confidence interval. For example, the 95% confidence interval for a certain performance indicator is [0.78, 0.92]. The confidence information is combined with the evaluation results to generate real-time evaluation status data for the transformer, including the confidence level.

[0023] Step S4: Obtain historical fault sample data corresponding to the low-impedance voltage transformer, and perform instant fault warning based on the real-time evaluation status data of the transformer and the historical fault sample data to generate an instant status fault warning signal corresponding to the low-impedance voltage transformer.

[0024] In an embodiment of the present invention, historical fault sample data of a low-impedance voltage transformer is obtained from a historical database that stores 1,000 fault samples over the past three years, including information such as the dynamic feature vector at the time of the fault, the fault type, and environmental parameters. The real-time evaluation status data is compared with each historical fault sample for instant error analysis, the difference in 11 dimensions is calculated, and the Euclidean distance formula is used to obtain the comprehensive evaluation status error. The evaluation status error change rate of two adjacent comparisons is calculated, and the threshold is set to 15%. If the error change rate calculated at a certain moment is 16.7%, which exceeds the threshold, the abnormal fault warning mechanism is triggered. An audible and visual alarm signal is emitted through a dedicated alarm device, and an instant status fault warning signal containing the warning time, warning type, and real-time evaluation status data is generated at the same time. The signal is sent to the monitoring center server via a 4G communication module. The server records the warning information in the database and pushes it to the maintenance personnel's handheld terminal device to remind them to conduct equipment inspection and maintenance in a timely manner.

[0025] Furthermore, step S1 includes the following steps: Step S11: collecting the primary side electrical signal of the transformer corresponding to the low-impedance voltage transformer in real time through the differential sampling structure corresponding to the Rogowski coil; Step S12: constructing a four-level resistor divider network by four resistor dividers to collect the secondary side electrical signal of the transformer corresponding to the low-impedance voltage transformer in real time; Step S13: collecting temperature, humidity and vibration data corresponding to the surrounding environment of the low-impedance voltage transformer by synchronously accessing the temperature and humidity sensor and the vibration sensor; Step S14: The primary side electrical signal of the transformer, the secondary side electrical signal of the transformer, the temperature and humidity data, and the vibration data corresponding to the low-impedance voltage transformer are clock-synchronized through a field programmable gate array and merged into the same data set to generate a transformer real-time data set.

[0026] 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: Step S11: collecting the primary side electrical signal of the transformer corresponding to the low-impedance voltage transformer in real time through the differential sampling structure corresponding to the Rogowski coil; In an embodiment of the present invention, a Rogowski coil differential sampling structure is used to collect the primary side electrical signal of a low-impedance voltage transformer. The Rogowski coil is designed with a circular cross-section, a cross-sectional radius of 5 mm, an average radius of 20 cm, 1000 turns, a turn spacing of 0.8 mm, and is covered with a 0.12 mm thick copper foil shielding layer. The differential sampling structure consists of two completely symmetrical windings, which are respectively wound on the positive and negative sides of the conductor to be measured. The output signals of the two windings are connected to a low-noise differential amplifier (model AD8422, CMRR ≥ 100 dB, gain bandwidth product ≥ 10 MHz). After the primary side current signal is converted into an induced electromotive force by the Rogowski coil, it is first passed through a 10 A π-type filter network consisting of a nF polypropylene film capacitor and a 100μH ferrite core inductor is used for anti-aliasing filtering. The signal then enters a second-order Butterworth low-pass filter (cutoff frequency 20kHz) for anti-aliasing. The filtered signal is converted into a digital signal by a 16-bit ADC (model ADS8320, sampling rate 100kHz, ENOB ≥ 14 bits) and transmitted to a field-programmable gate array via an SPI interface (clock frequency 10MHz). At a certain test moment, when the primary side current was 100A, the corresponding voltage value of the collected digital signal was 2.503V. After calibration and conversion, the current value was 100.12A, with a measurement error of ±0.12%.

[0027] Step S12: constructing a four-level resistor divider network by four resistor dividers to collect the secondary side electrical signal of the transformer corresponding to the low-impedance voltage transformer in real time; In this embodiment of the present invention, a four-stage resistor divider network is constructed using four resistor dividers to collect secondary-side electrical signals. Each resistor divider consists of 10 metal film resistors connected in series with a nominal resistance of 100 kΩ, an accuracy of ±0.01%, and a temperature coefficient of 5 ppm / °C, forming a high-precision voltage divider resistor of 1000 kΩ. The voltage divider ratios of the four voltage dividers are 100:1, 200:1, 500:1, and 1000:1, respectively, and are switched by a high-precision analog switch (model ADG1608, on-resistance ≤4 Ω). The secondary-side voltage signal first passes through a π-type filter structure consisting of four 10nF polypropylene film capacitors and two 100μH ferrite-core inductors, and then passes through a first-order RC low-pass filter circuit (1kΩ resistor and 10μF capacitor, cutoff frequency 16Hz). The filtered signal is connected to a 16-bit ADC (model ADS1255, sampling rate 30kHz, ENOB ≥ 15 bits) for digital conversion. When the secondary-side voltage is 100V, the system automatically selects the 1000:1 voltage divider ratio channel. The corresponding voltage value of the collected digital signal is 0.10005V. After conversion, the secondary-side voltage value is 100.05V, with a measurement error of ±0.05%.

[0028] Step S13: collecting temperature, humidity and vibration data corresponding to the surrounding environment of the low-impedance voltage transformer by synchronously accessing the temperature and humidity sensor and the vibration sensor; In an embodiment of the present invention, environmental parameters are collected by using special sensors: a temperature sensor (measuring range of -55°C to +125°C, accuracy of ±0.5°C) is installed in a special protective box 5 cm away from the transformer housing; a humidity sensor (measuring range of 0% to 100% RH, accuracy of ±3% RH) is installed in a well-ventilated location; and a vibration sensor (measuring range of ±16g, resolution of 0.003g / LSB) is fixed to the transformer base with M3 stainless steel screws. All three sensors are connected to the field programmable gate array via an SPI interface, and the data acquisition frequency is 10Hz. During a certain test period, the temperature sensor measured an ambient temperature range of 23.5°C to 25.8°C, the humidity sensor measured a relative humidity range of 45% RH to 52% RH, and the vibration sensor measured vibration frequencies mainly concentrated at two frequency points of 50Hz and 100Hz, with amplitudes of 0.05g and 0.03g, respectively. All data are recorded in the form of timestamps to ensure temporal correlation with the electrical signal data.

[0029] Step S14: The primary side electrical signal of the transformer, the secondary side electrical signal of the transformer, the temperature and humidity data, and the vibration data corresponding to the low-impedance voltage transformer are clock-synchronized through a field programmable gate array and merged into the same data set to generate a transformer real-time data set.

[0030] In the embodiment of the present invention, the clock synchronization and merging of multi-source data are achieved by using a field programmable gate array. First, a 20MHz system clock is generated by a phase-locked loop (PLL) as the sampling clock source for all ADCs and sensors. The primary side electrical signal ADC (ADS8320) and the secondary side electrical signal ADC (ADS1255) both adopt a synchronous trigger mode. The trigger signal is output by the same GPIO pin of the field programmable gate array to ensure that the two ADCs start sampling at the same time. The data collection of the temperature and humidity sensor and the vibration sensor adopts a query mode. The field programmable gate array sends a read instruction to the sensor every 100ms and records it. The reading time is recorded. All collected data are accompanied by a 32-bit timestamp (resolution 1μs). The timestamp count starts from system startup. The FIFO buffer area (depth 4096) inside the field programmable gate array is used to temporarily store data from various data sources. When the data volume in the buffer area reaches 2048, all data are sorted by timestamp and merged into a data set containing primary-side electrical signals, secondary-side electrical signals, temperature and humidity data, and vibration data. The merged data set is sent to the data processing center via the Gigabit Ethernet interface (using the TCP / IP protocol) at a frequency of 100Hz, realizing real-time synchronous acquisition and transmission of multi-source data of low-impedance voltage transformers.

[0031] Furthermore, step S11 includes the following steps: Finite element simulation was used to optimize the spacing between the Rogowski coils and the thickness of the shielding layer above them. In the embodiment of the present invention, finite element modeling of the Rogowski coil is performed by using ANSYS Maxwell electromagnetic field simulation software, the simulation frequency range is set to 50 Hz to 10 kHz, the excitation current amplitude is 100 A, the Rogowski coil adopts a circular cross-section design with a cross-sectional radius of 5 mm, an average radius of 20 cm, 1000 turns, and an initial setting of the coil turn spacing of 0.5 mm. The shielding layer adopts a copper foil with a thickness of 0.1 mm, a relative magnetic permeability of 1, and an electrical conductivity of 5.8×10 7 S / m. During the simulation process, keeping other parameters unchanged, the turn spacing was gradually adjusted from 0.3mm to 1.0mm in 0.1mm steps. At the same time, the shielding layer thickness was gradually adjusted from 0.05mm to 0.2mm in 0.05mm steps. For each parameter combination, the internal magnetic field intensity distribution, induced electromotive force, and external leakage magnetic flux were calculated. For example, when the turn spacing was 0.7mm and the shielding layer thickness was 0.15mm, the uniformity of the magnetic field intensity inside the coil at 50Hz reached 98.5%, and the induced electromotive force amplitude fluctuation at 10kHz was less than ±0.5%. By comparing the simulation results under different parameter combinations, the optimal parameters were ultimately determined to be a turn spacing of 0.8mm and a shielding layer thickness of 0.12mm. At this time, the external leakage magnetic flux was reduced to 15% of the initial value across the entire frequency range.

[0032] Preferably, a leakage magnetic interference suppression evaluation is performed on the Rogowski coil based on the optimized spacing and the thickness of the shielding layer thereon, so as to obtain a primary side leakage magnetic interference suppression ratio corresponding to the Rogowski coil; In the embodiment of the present invention, a Rogowski coil was made based on the previously determined optimal parameters (turn spacing 0.8mm, shielding layer thickness 0.12mm), and a leakage magnetic interference suppression evaluation platform was built. A standard current source (output range 0-500A, frequency range 20Hz-20kHz, accuracy ±0.01%) was used to apply currents of different frequencies and amplitudes to the test wire. A magnetic field probe (measuring range 0-10mT, frequency response DC-100kHz, accuracy ±0.5%) was placed 10cm away from the Rogowski coil to detect leakage magnetic signals. The three conditions were tested: no shielding layer, initial shielding layer (0.1mm copper foil), and optimized shielding layer (0.12mm copper foil). The test was carried out under the following conditions. When the standard current source outputs 50Hz and 100A, the leakage magnetic signal detected by the magnetic field probe under the condition of no shielding layer is 2.5mT. The leakage magnetic signal under the initial shielding layer condition is reduced to 0.8mT. The leakage magnetic signal under the optimized shielding layer condition is further reduced to 0.3mT. The primary side leakage magnetic interference suppression ratio is calculated by the formula: suppression ratio = leakage magnetic signal amplitude without shielding layer / leakage magnetic signal amplitude after optimization shielding layer. For 50Hz signal, the suppression ratio is 2.5 / 0.3≈8.33; for 1kHz signal, the suppression ratio is 3.2 / 0.45≈7.11. In the entire test frequency range, the leakage magnetic interference suppression ratio of the optimized Rogowski coil reaches an average of more than 7.5.

[0033] Preferably, the electrical sampling accuracy corresponding to the low-impedance working condition is determined based on the primary-side leakage magnetic interference suppression ratio corresponding to the Rogowski coil, and based on the electrical sampling accuracy, the primary-side electrical signal of the transformer corresponding to the low-impedance voltage transformer is collected in real time through the differential sampling structure corresponding to the Rogowski coil.

[0034] In an embodiment of the present invention, the electrical sampling accuracy of the Rogowski coil under low-impedance conditions is evaluated based on the previously obtained primary-side leakage magnetic interference suppression ratio. On a low-impedance voltage transformer test platform, three typical low-impedance conditions with load impedances of 1Ω, 5Ω, and 10Ω are set. A standard voltage source is used to apply a test voltage of 50Hz and 100V. The electrical sampling accuracy is calculated by comparing the primary-side current signal collected by the Rogowski coil with the measurement results of a standard current transformer (accuracy level 0.01). Under the condition of a load impedance of 1Ω, when the primary-side current is 100A, the Rogowski coil measurement value is 99.95A, with an absolute error of 0.05A and a relative error of 0.05%; when the current drops to 10A, the measurement value is 9.992A, with a relative error of 0.08%. Under the condition of a load impedance of 10Ω, the 100A current is The measured value was 99.98A, with a relative error of 0.02%. Combining the test results under different working conditions, the electrical sampling accuracy of the Rogowski coil under low-impedance conditions was determined to be ±0.1%. Based on this accuracy, the Rogowski coil's differential sampling structure was used to acquire the primary-side electrical signals of the low-impedance voltage transformer in real time. The differential sampling structure consists of two completely symmetrical Rogowski coil windings, one wound around the positive and the other on the negative sides of the conductor under test. The output signals of the two windings are processed by a differential amplifier (CMRR ≥ 100dB, gain-bandwidth product ≥ 10MHz) to effectively suppress common-mode interference. The collected signals are first converted to digital signals by a 16-bit ADC (sampling rate 100kHz, ENOB ≥ 14 bits) and then transmitted to the data processing unit via the SPI interface, achieving high-precision real-time acquisition of primary-side electrical signals under low-impedance conditions.

[0035] Furthermore, step S2 includes the following steps: Step S21: performing soft threshold denoising processing on the electrical signal in the transformer real-time data set to generate a transformer denoised electrical signal; Step S22: Calculate the cross-correlation coefficient between the voltage and current signals of the transformer de-noised electrical signal. When the cross-correlation coefficient is less than 0.95, trigger the signal re-sampling mechanism and iterate the de-noising process to generate a transformer re-correlated electrical signal. Step S23: Calculate the corresponding environmental change rate based on the environmental parameters corresponding to temperature, humidity and vibration in the real-time data set of the mutual inductor, and perform abnormal environment elimination processing on the environmental parameters corresponding to temperature, humidity and vibration based on the environmental change rate to obtain the mutual inductor filtered abnormal environment data; Step S24: dynamically evaluating and screening the transformer's re-correlated electrical signals and transformer's filtered environmental data based on the data integrity and accuracy scores to generate a quality-screened valid real-time data set; Step S25: performing variational modal feature extraction on the valid real-time data set to obtain a mutual inductor dynamic feature vector corresponding to the amplitude error, phase offset, and harmonic distortion.

[0036] As an embodiment of the present invention, refer to Figure 3 As shown, Figure 1 Detailed step flow diagram of step S2 in the embodiment, step S2 includes the following steps: Step S21: performing soft threshold denoising processing on the electrical signal in the transformer real-time data set to generate a transformer denoised electrical signal; In an embodiment of the present invention, a dedicated signal processing device is used to perform soft threshold denoising on the electrical signal in the real-time data set of the mutual inductor. It is assumed that the electrical signal is collected by a device with a sampling rate of 100kHz, lasts for 1 second, and contains a discrete sequence of 100,000 sampling points. The device has a built-in wavelet transform module, and the db10 wavelet is used to perform a 5-layer decomposition of the electrical signal, decomposing the signal into multiple sub-bands of different frequencies. In each sub-band after decomposition, the median of the signal coefficient is calculated and multiplied by 1.4826 as the threshold. For the signal coefficient in each sub-band, If its absolute value is less than the threshold, the coefficient is set to 0; if it is greater than the threshold, the value obtained by subtracting the threshold multiplied by its sign (positive or negative) from the coefficient is used. For example, if the coefficient value of a signal in a sub-band is 5 and the calculated threshold is 3, the coefficient value becomes 2 after processing. After the above soft threshold processing is performed on the coefficients of all sub-bands, the processed sub-band signal is reconstructed using the inverse wavelet transform to generate the mutual inductor denoising electrical signal. In this way, the noise in the electrical signal can be effectively removed. For example, the noise fluctuation with a peak value of 8 in the original signal can be reduced to below 1, thereby improving the signal quality.

[0037] Step S22: Calculate the cross-correlation coefficient between the voltage and current signals of the transformer de-noised electrical signal. When the cross-correlation coefficient is less than 0.95, trigger the signal re-sampling mechanism and iterate the de-noising process to generate a transformer re-correlated electrical signal. In an embodiment of the present invention, the electrical signal of the mutual inductor is denoised, and the cross-correlation coefficient between the voltage and current signals therein is calculated using a dedicated signal analysis instrument. The instrument has a built-in correlation operation module, which synchronously aligns the denoised voltage signal and current signal, and calculates the cross-correlation coefficient of the voltage and current signals in each data segment, with 100 sampling points as a group of data segments. Assuming that in a certain set of data segments, the calculated mutual correlation coefficient is 0.93, which is less than the set threshold of 0.95, the signal re-sampling mechanism is triggered, and the backup high-precision data acquisition equipment is started. The sampling rate of this equipment is 200kHz, and the accuracy is one level higher than that of the original equipment. The voltage and current signals of the transformer are re-collected, and the acquisition time is still 1 second to obtain new electrical signal data. The newly collected signal is again subjected to the corresponding soft threshold denoising processing. The db10 wavelet is also selected for 5-layer decomposition. According to the same soft threshold calculation and processing method, and wavelet inverse transform reconstruction, the transformer re-correlated electrical signal is generated. Through continuous iteration until the calculated mutual correlation coefficient is greater than or equal to 0.95, it is ensured that there is a high correlation between the voltage and current signals, providing a reliable data basis for subsequent analysis.

[0038] Step S23: Calculate the corresponding environmental change rate based on the environmental parameters corresponding to temperature, humidity and vibration in the real-time data set of the mutual inductor, and perform abnormal environment elimination processing on the environmental parameters corresponding to temperature, humidity and vibration based on the environmental change rate to obtain the mutual inductor filtered abnormal environment data; In an embodiment of the present invention, environmental parameters corresponding to temperature, humidity and vibration are extracted from the real-time data set of the mutual inductor. The temperature data is collected by a temperature sensor with a measurement range of -55°C to +125°C and an accuracy of ±0.5°C; the humidity data is collected by a humidity sensor with a measurement range of 0% to 100% RH and an accuracy of ±3% RH; the vibration data is collected by a vibration sensor with a measurement range of ±16g and a resolution of 0.003g / LSB. The environmental change rate at adjacent moments is calculated with a time interval of 1 minute. For temperature, if the temperatures at adjacent moments are 23°C and 23.5°C respectively, and the time interval is 60 seconds, the temperature change rate is (23.5-23) / 60=0.0083°C / s; the change rate of humidity and vibration data is calculated in the same way. The abnormal environment threshold is set, the temperature change rate threshold is 0.1°C / s, and the humidity change rate threshold is 5%. RH / min, the vibration acceleration change rate threshold is 0.5g / s. When the environmental change rate at a certain moment exceeds the corresponding threshold, the environmental parameters at that moment are considered abnormal data and are discarded. For example, if the temperature change rate at a certain moment reaches 0.15℃ / s, exceeding the temperature change rate threshold, all temperature, humidity, and vibration data at that moment and within 10 seconds before and after are discarded to obtain the transformer-filtered abnormal environmental data, ensuring the accuracy and reliability of the environmental parameters and preventing abnormal environmental data from interfering with subsequent analysis.

[0039] Step S24: dynamically evaluating and screening the transformer's re-correlated electrical signals and transformer's filtered environmental data based on the data integrity and accuracy scores to generate a quality-screened valid real-time data set; In an embodiment of the present invention, a data integrity and accuracy scoring system is established to dynamically evaluate and screen the quality of the transformer's heavily correlated electrical signals and the transformer's filtered environmental data. For electrical signals, the proportion of missing data in the signal is checked. If the missing data exceeds 5%, a lower integrity score is given. By comparing with a reference signal generated by a high-precision standard signal source, the errors in signal amplitude and phase are calculated. The larger the error, the lower the accuracy score. For environmental data, check the continuity of the data. If there is a data interruption, reduce the integrity score. Compare the data collected by multiple sensors in similar positions. If the difference is too large, it will affect the accuracy score. Set the weights of data integrity and accuracy to 0.4 and 0.6 respectively, and calculate the comprehensive score. Assuming that the integrity score of the transformer's re-correlated electrical signal is 80 points and the accuracy score is 85 points, its comprehensive score is 80×0.4+85×0.6=83 points; the integrity score of the transformer's filtered environmental data is 85 points and the accuracy score is 82 points. Its comprehensive score is 85×0.4+82×0.6=83.2 points. Set the comprehensive score threshold to 80 points, filter out data with scores above the threshold, and generate a valid real-time data set that has been quality-screened to ensure that the data used for subsequent analysis is of high quality.

[0040] Step S25: performing variational modal feature extraction on the valid real-time data set to obtain a mutual inductor dynamic feature vector corresponding to the amplitude error, phase offset, and harmonic distortion.

[0041] In an embodiment of the present invention, variational modal feature extraction is performed on an effective real-time data set, electrical signal data is first separated from the data set, and then processed using a variational mode decomposition (VMD) algorithm. The electrical signal is input into a dedicated signal processing device with a built-in VMD operation module. During the VMD decomposition process, the number of decomposition modes is set to 10, and the balance parameter α is set to 2000. Through iterative calculation, the electrical signal is decomposed into 10 intrinsic mode functions (IMFs). The correlation coefficient between each intrinsic mode and a reference signal under a typical low-impedance operating condition of a low-impedance voltage transformer is calculated, and 3-5 principal components related to the low-impedance characteristic are screened out. For these principal components, a fast Fourier transform (FFT) algorithm is used to obtain their signal amplitude distribution and signal phase distribution, which are compared with the amplitude and phase under ideal conditions to calculate the amplitude error and phase offset. At the same time, the FFT algorithm is used to extract the harmonic components of the electrical signal. Combined with the environmental data in the data set, the harmonic distortion is evaluated according to the established method, and the amplitude deviation rate, phase drift and distortion energy ratio under the influence of the environment are calculated. Finally, the amplitude error and phase offset corresponding to each main component are combined with the harmonic distortion data corresponding to each harmonic component to form a transformer dynamic feature vector containing amplitude error, phase offset and harmonic distortion, providing key feature data for the performance evaluation of low-impedance voltage transformers.

[0042] Furthermore, step S21 includes the following steps: The electrical signal in the transformer real-time data set is decomposed into five layers using db10 wavelet packets to generate five layers of transformer real-time electrical signal frequency sub-bands. In the embodiment of the present invention, the electrical signal in the real-time data set of the mutual inductor is decomposed by selecting the db10 wavelet packet. Taking the primary side electrical signal as an example, the original signal is a discrete sequence containing 10,000 sampling points with a sampling frequency of 100 kHz. The original signal is input into a dedicated signal processing unit, which has a built-in wavelet packet decomposition algorithm module. When the 5-layer wavelet packet decomposition is performed, the first layer decomposes the original signal into high-frequency sub-bands and low-frequency sub-bands. The second layer decomposes the high-frequency sub-bands and low-frequency sub-bands obtained in the first layer again to generate four Sub-bands, and so on. After 5 layers of decomposition, a total of 32 frequency sub-bands are generated. Each frequency sub-band corresponds to a different frequency range. For example, the first layer of low-frequency sub-band covers the frequency range of 0-50kHz, and the high-frequency sub-band covers the frequency range of 50-100kHz. As the number of decomposition layers increases, the frequency band division becomes more refined. During the decomposition process, the calculation is strictly performed in accordance with the basis function characteristics of the db10 wavelet packet to ensure the accurate energy distribution of the signal in each frequency sub-band. Finally, 5 layers of mutual inductor real-time electrical signal frequency sub-bands are generated, providing multi-scale signal feature data for subsequent analysis.

[0043] Preferably, the corresponding transformer electrical signal sub-band kurtosis is obtained through the 5-layer transformer instantaneous electrical signal frequency sub-band, and the noisy sub-band is identified for the 5-layer transformer instantaneous electrical signal frequency sub-band based on the transformer electrical signal sub-band kurtosis to generate the transformer electrical signal noisy frequency sub-band; In an embodiment of the present invention, the kurtosis of the mutual inductor electrical signal sub-band corresponding to each frequency sub-band is calculated for the 5-layer mutual inductor instantaneous electrical signal frequency sub-band. The kurtosis is calculated using the formula of the ratio of the fourth-order center distance to the square of the variance. Taking the 100 sampling point data contained in a certain frequency sub-band as an example, the mean of the frequency sub-band data is first calculated, and then the sum of the fourth power of the difference between each sampling point and the mean is calculated, and divided by the number of data to obtain the fourth-order center distance; at the same time, the variance is calculated and squared, and the fourth-order center distance is divided by the square of the variance to obtain the value of the frequency sub-band. Kurtosis value: Set the kurtosis threshold to 3 (normal signal kurtosis is usually close to 3). When the kurtosis value of a frequency sub-band is greater than 3, the frequency sub-band is identified as a noisy sub-band. For example, among the 32 frequency sub-bands, the kurtosis values of the 5th, 12th, and 20th frequency sub-bands are 3.8, 4.2, and 3.6, respectively, all greater than the threshold of 3. These three frequency sub-bands are identified as noisy frequency sub-bands of the transformer electrical signal. In this way, the kurtosis of all frequency sub-bands is comprehensively analyzed, and the frequency sub-bands affected by noise are accurately identified, providing a clear target for subsequent denoising.

[0044] Preferably, soft threshold denoising is performed based on the noisy frequency sub-band of the mutual inductor electrical signal, and electrical signal reconstruction is performed with other mutual inductor instantaneous electrical signal frequency sub-bands to generate a mutual inductor denoised electrical signal.

[0045] In an embodiment of the present invention, a soft threshold denoising process is performed based on the identified noisy frequency sub-band of the mutual inductor electrical signal. For each data point in the noisy frequency sub-band, a threshold is set to the median of the absolute value of the frequency sub-band data multiplied by 1.4826 (for estimating the noise standard deviation). If the absolute value of the data point is less than the threshold, the data point is set to 0; if it is greater than the threshold, the data point value is subtracted from the threshold and multiplied by the data point sign (positive or negative). For example, if the value of a data point in the noisy frequency sub-band is 5 and the threshold is 3, the value of the data point becomes 2 after processing. After this operation is performed on the noisy frequency sub-band data, the denoised noisy frequency sub-band and the remaining transformer real-time electrical signal frequency sub-bands that are not affected by noise are reconstructed using the inverse wavelet packet transform. During the reconstruction process, the frequency sub-band data are merged strictly according to the frequency band correspondence and weights during decomposition to restore the transformer denoised electrical signal. After this processing, the noise in the original electrical signal is effectively removed. For example, the noise fluctuation with a peak value of 8 in the original signal is reduced to below 1, providing high-quality signal data for the subsequent accurate evaluation of the low-impedance voltage transformer performance.

[0046] Furthermore, step S25 includes the following steps: Step S251: performing variational modal decomposition on the electrical signal in the valid instantaneous data set to decompose the electrical signal into 8-12 eigenmodes, and screening 3-5 principal components related to low impedance characteristics through correlation coefficients; In an embodiment of the present invention, a variational mode decomposition (VMD) algorithm is used to process the electrical signal within a valid real-time data set. Assuming that the electrical signal is acquired by a device with a sampling rate of 100 kHz, lasts for 1 second, and contains a discrete sequence of 100,000 sampling points, the electrical signal is input into a dedicated signal processing device with a built-in VMD operation module. During the VMD decomposition process, the number of decomposition modes is set to 10, and the balance parameter α is set to 2000. Through iterative calculation, the electrical signal is decomposed into 10 intrinsic mode functions (IMFs). Each intrinsic mode function represents a component with different frequency characteristics in the electrical signal. For example, the first eigenmode primarily contains low-frequency components, and subsequent eigenmodes successively cover signal characteristics in higher frequency ranges. To screen the principal components related to low-impedance characteristics, the correlation coefficient of each eigenmode is calculated with a reference signal under typical low-impedance operating conditions of a low-impedance voltage transformer. The reference signal is obtained through statistical analysis of a large amount of low-impedance operating condition test data. For example, after calculation, the correlation coefficients between the 3rd, 5th, and 7th eigenmodes and the reference signal are 0.85, 0.88, and 0.82, respectively, which are higher than those of other eigenmodes. These three eigenmodes are determined as principal components. These principal components can effectively reflect the signal characteristics of the low-impedance voltage transformer that are closely related to the low-impedance characteristics during operation, providing key data for subsequent analysis.

[0047] Step S252: Obtain the signal amplitude distribution and signal phase distribution corresponding to each principal component, and perform amplitude error and phase offset statistics based on the signal amplitude distribution and signal phase distribution corresponding to each principal component to obtain the amplitude error and phase offset corresponding to each principal component; In an embodiment of the present invention, by obtaining the corresponding signal amplitude distribution and signal phase distribution for the three main components previously screened out, a 1024-point FFT transform is performed on each main component using a fast Fourier transform (FFT) algorithm, and the time domain signal is converted into a frequency domain signal, thereby obtaining the amplitude and phase information of each main component in the frequency range of 0-50kHz. Taking the third main component as an example, at a frequency of 50Hz, the amplitude is 12V and the phase is 12°; at a frequency of 100Hz, the amplitude is 8V and the phase is 8°. The amplitude and phase under ideal conditions are used as reference standards (which can be obtained through mutual inductor design parameters or high-precision calibration tests). At a frequency of 50Hz, assuming that the ideal amplitude is 10V and the ideal phase is 10°, the calculated amplitude error is (12-10) / 10×100%=20%, the phase difference is 12°-10°=2°, and the corresponding phase offset calculation formula is: ,in is the phase difference, The corresponding phase offset is calculated for the frequency corresponding to the main component. Such amplitude error and phase offset calculations are performed for each frequency point of all main components in the frequency range of 0-50kHz to obtain the complete amplitude error distribution and phase offset distribution data of each main component. Through these data, we can clearly understand the deviation of each main component from the ideal state at different frequencies, which provides an important basis for evaluating the performance of low-impedance voltage transformers.

[0048] Step S253: obtaining corresponding harmonic components from the electrical signal in the valid real-time data set, and performing harmonic distortion evaluation on the harmonic components corresponding to the electrical signal based on the environmental data in the valid real-time data set to obtain harmonic distortion corresponding to each harmonic component; In an embodiment of the present invention, each harmonic component of the electrical signal is extracted from the valid real-time data set, and the electrical signal is converted into the frequency domain with the help of the FFT algorithm to obtain the amplitude and phase information of the 1-15 harmonic components. At the same time, the environmental data in the data set is extracted, the temperature data is collected by the temperature sensor, the humidity data is collected by the humidity sensor, and the vibration data is collected by the vibration sensor. The harmonic distortion is evaluated according to the subsequent method, and the gradients corresponding to the temperature and humidity are calculated. For example, if the temperature at adjacent moments is 23°C and 23.5°C at a time interval of 1 minute, and the time interval is 60 seconds, then the temperature gradient is (23.5-23) / 60=0.0083°C / s; the humidity gradient is calculated in the same way. The main vibration frequency components are obtained by performing FFT on the vibration data. Assuming that 50Hz and 100Hz are detected as the main vibrations, the main frequency components are obtained. The dynamic frequency is obtained by fitting the least squares method to establish a linear relationship model between the harmonic amplitude or phase and the environmental variables, and the offset coupling coefficient between each environmental component and the amplitude or phase of each harmonic is calculated; then, by dividing the data interval, the joint probability distribution matrix and the marginal probability distribution matrix are constructed, and the environmental-electrical mutual information value is calculated. Finally, the offset coupling coefficient weight is set to 0.6, and the mutual information value weight is set to 0.4. The comprehensive influencing factor of each harmonic component is comprehensively calculated, and then the amplitude deviation rate, phase drift and distortion energy ratio under the influence of the environment are obtained, and the harmonic distortion of each harmonic component is comprehensively evaluated.

[0049] Step S254: combining the amplitude error and phase offset corresponding to each main component and the harmonic distortion corresponding to each subharmonic component into the same eigenvector to obtain a mutual inductor dynamic eigenvector including the amplitude error, phase offset and harmonic distortion.

[0050] In an embodiment of the present invention, the previously obtained amplitude error and phase offset data corresponding to each principal component are combined with the obtained harmonic distortion data corresponding to each subharmonic component to set the format of the feature vector. The first three columns store the amplitude error data of the three principal components, the middle three columns store the phase offset data of the three principal components, and the last 15 columns store the amplitude deviation rate, phase drift, and distortion energy percentage data of the 1st to 15th harmonic components. For example, the amplitude error of the third principal component at a frequency of 50Hz is 20%, and the phase offset is 2°; the amplitude deviation rate of the third harmonic under environmental influence is 15%, the phase drift is 3°, and the distortion energy percentage is 8%. These data are sequentially filled into the corresponding positions of the feature vector to form a complete transformer dynamic feature vector. This feature vector integrates key information related to amplitude, phase, and harmonic distortion of the low-impedance voltage transformer during operation, providing a unified and comprehensive data foundation for subsequent operations such as transformer performance evaluation and fault diagnosis based on these features. Furthermore, the amplitude error and phase offset statistics according to the signal amplitude distribution and signal phase distribution corresponding to each principal component described in step S252 include the following steps: Perform time-frequency synchronization on the signal amplitude distribution and signal phase distribution corresponding to each principal component to obtain the amplitude distribution and phase distribution corresponding to each principal component at the same time-frequency; In the embodiment of the present invention, a time-frequency analysis method based on fast Fourier transform (FFT) is used to process the principal components obtained after principal component analysis of the transformer de-noising electrical signal. Assume that the principal component analysis of a low-impedance voltage transformer de-noising electrical signal obtains 5 principal components, each of which is a discrete sequence containing 10,000 sampling points and a sampling frequency of 100 kHz. The 5 principal component signals are respectively input into a dedicated signal processing device. The device has a built-in FFT operation module, which performs a 1024-point FFT transformation on each principal component signal to convert the time domain signal into a time domain signal. The frequency domain signal is obtained by analyzing the spectral distribution of each principal component within the frequency range of 0-50kHz. Because each principal component may experience time delay or frequency offset during acquisition, resulting in time-frequency asynchrony, a phase synchronization algorithm is used to adjust these principal components. For example, after FFT transformation, the amplitude of principal component 1 at 50Hz is 10V and the phase is 15°; the amplitude of principal component 2 at 50Hz is 8V and the phase is 10°. By calculating the phase difference between the two principal components at each frequency point, time delay compensation is performed on the phase-lagging principal component 2. Specifically, the number of sample points required for delay is calculated based on the sampling frequency and phase difference, and all sample points of principal component 2 are shifted accordingly. Frequency offset is also corrected by adjusting the reference frequency of the FFT transformation to align the two principal components at the same frequency point. This time-frequency synchronization is repeated for all five principal components, ultimately obtaining the amplitude and phase distributions of each principal component at the same time-frequency point, providing a unified benchmark for subsequent analysis.

[0051] Preferably, amplitude error statistics are performed between the amplitude distributions corresponding to the respective principal components at the same time-frequency to obtain the amplitude errors corresponding to the respective principal components; In an embodiment of the present invention, amplitude error statistics are performed based on the amplitude distribution corresponding to each principal component at the same time and frequency obtained previously. Taking 5 principal components as an example, the amplitudes of the 5 principal components at each frequency point are extracted in turn within the frequency range of 0-50kHz and with a frequency interval of 1Hz. For example, at the frequency point of 50Hz, the amplitude of main component 1 is 10V, the amplitude of main component 2 is 8V, the amplitude of main component 3 is 9V, the amplitude of main component 4 is 7V, and the amplitude of main component 5 is 6V. Taking the amplitude of main component 1 as the reference standard, the amplitude errors of other main components and main component 1 are calculated. The amplitude error of main component 2 = (8-10) / 10×100% = -20%; the amplitude error of main component 3 = (9-10) / 10×100% = -10%; and so on, the amplitude errors of main components 4 and 5 are calculated. The above amplitude error calculation is performed for all frequency points in the frequency range of 0-50kHz to obtain the amplitude error data of each main component at each frequency point relative to the reference main component. These data are organized into a data table according to the order of main components and frequency order to clearly show the amplitude error of each main component at different frequencies. By analyzing these amplitude error data, we can find that some main components have amplitude attenuation or gain abnormalities at specific frequencies. For example, the amplitude error of main component 4 reaches -30% at 100Hz, which provides an important basis for evaluating the performance of low-impedance voltage transformers.

[0052] Preferably, the corresponding phase differences between the main components are obtained by the phase distributions corresponding to the main components at the same time-frequency, and the phase shift statistics of the phase distributions corresponding to the main components at the same time-frequency are performed based on the corresponding phase differences between the main components to obtain the phase shifts corresponding to the main components.

[0053] In an embodiment of the present invention, the phase difference between each main component is obtained based on the previously obtained phase distribution corresponding to each main component at the same time-frequency. Taking 5 main components as an example, in the frequency range of 0-50kHz, with a frequency interval of 1Hz, the phase values of the 5 main components at each frequency point are extracted. At the frequency point of 50Hz, the phase of main component 1 is 15°, and the phase of main component 2 is 10°. Then, the phase difference between main component 1 and main component 2 is 15°-10°=5°; the phase of main component 3 is 12°, and the phase difference between main component 1 and main component 3 is 15°-12°=3°. The phase differences between all main components at the frequency point are calculated in turn. Based on the corresponding phase differences between each main component, the phase distribution corresponding to each main component at the same time-frequency is subjected to phase offset statistics. For example, the corresponding phase offset calculation formula is: ,in is the phase difference, For the frequency corresponding to the main component, such phase offset statistics are performed on all frequency points and all main components in the frequency range of 0-50kHz, and the phase offset data of each main component relative to the reference main component at each frequency point are obtained. These data are organized into curve charts to intuitively display the phase offset of each main component at different frequencies. By analyzing the phase offset data, the changes in the phase relationship between the main components of the low-impedance voltage transformer at different frequencies can be understood. For example, it is found that the phase offset of main component 3 suddenly increases in the frequency range of 200Hz-300Hz, which is of great significance for evaluating the phase characteristics and performance stability of the transformer.

[0054] Furthermore, the harmonic distortion evaluation of each harmonic component corresponding to the electrical signal based on the environmental data in the valid real-time data set in step S253 includes the following steps: Based on the environmental data in the valid real-time data set, an environmental-electrical offset coupling analysis is performed on each harmonic component corresponding to the electrical signal to analyze the gradient and vibration frequency corresponding to temperature and humidity. Based on the gradient and vibration frequency corresponding to temperature and humidity, an offset coupling evaluation is performed between the amplitude or phase corresponding to each harmonic component to obtain the offset coupling coefficient between each temperature, humidity and vibration component and each harmonic amplitude or phase; In an embodiment of the present invention, environmental data (temperature and humidity data, vibration data) and harmonic component data corresponding to the electrical signal are extracted from the valid real-time data set. The temperature data is collected by a temperature sensor with a measurement range of -55°C to +125°C and an accuracy of ±0.5°C; the humidity data is collected by a humidity sensor with a measurement range of 0% to 100% RH and an accuracy of ±3%RH; the vibration data is collected by a vibration sensor with a measurement range of ±16g and a resolution of 0.003g / LSB. The harmonic component data of the electrical signal are derived from the result of frequency domain analysis of the denoised electrical signal of the mutual inductor. Assuming that harmonics 1-15 are covered, the gradient corresponding to the temperature and humidity is calculated. With a time interval of 1 minute, the ratio of the temperature difference between adjacent moments to the time interval is calculated to obtain the temperature gradient. For example, if the temperatures at adjacent moments are 23°C and 23.5°C, respectively, and the time interval is 60 seconds, the temperature gradient is (23.5-23) / 60= 0.0083℃ / s; humidity gradient calculation is similar. The vibration frequency is obtained by performing fast Fourier transform (FFT) on the vibration data to obtain its spectrum distribution and determine the main vibration frequency components. Assuming that 50Hz and 100Hz are detected as the main vibration frequencies, for each harmonic component, taking the 5th harmonic as an example, analyze the relationship between its amplitude or phase and the temperature and humidity gradient, vibration frequency, and use the least squares fitting method to establish a linear relationship model between the harmonic amplitude or phase and the environmental variables. If there is a linear relationship between the 5th harmonic amplitude and the temperature gradient, the equation is obtained by fitting. ( is the 5th harmonic amplitude, is the temperature gradient, 、 is the fitting coefficient). For example, the offset coupling coefficient between the temperature gradient and the 5th harmonic amplitude is calculated to be 0.2 according to the model. Similarly, the offset coupling coefficients between humidity, vibration frequency and each harmonic amplitude or phase are calculated to comprehensively quantify the degree of correlation between each environmental component and the harmonic amplitude or phase.

[0055] Preferably, the environmental-electrical mutual information calculation is performed on each harmonic component corresponding to the electrical signal based on the environmental data in the valid real-time data set to obtain the mutual information value between each component of temperature, humidity and vibration and the amplitude or phase of each harmonic; In an embodiment of the present invention, the environmental-electrical mutual information is calculated based on the environmental data and the harmonic component data of the electrical signal in the valid real-time data set, the temperature data is divided into 10 intervals (such as -55℃~-45℃, -45℃~-35℃, etc.), the humidity data is divided into 8 intervals (such as 0%~10% RH, 10%~20% RH, etc.), the vibration frequency is classified according to the main frequency component (50Hz, 100Hz, etc.), and the amplitude and phase of each harmonic are also divided into corresponding intervals. Taking the temperature and the third harmonic amplitude as an example, the number of data occurrences when the temperature is in a certain interval and the third harmonic amplitude is in a certain interval is counted, and a joint probability distribution matrix is constructed. ; At the same time, the probability of temperature and third harmonic amplitude being in different intervals is counted respectively to obtain the marginal probability distribution matrix and , mutual information calculation formula The mutual information value between temperature and the third harmonic amplitude is calculated to be 0.3 bit. Similar mutual information calculations are performed on humidity, vibration frequency and each harmonic amplitude or phase. For example, the mutual information value between humidity and the seventh harmonic phase is 0.25 bit, and the mutual information value between the vibration frequency of 50Hz and the 11th harmonic amplitude is 0.18 bit. These mutual information values reflect the degree of information correlation between each environmental component and each harmonic amplitude or phase. The larger the value, the stronger the dependence between the two.

[0056] Preferably, the harmonic distortion of each harmonic component corresponding to the electrical signal is evaluated based on the offset coupling coefficient and mutual information value between the temperature, humidity and vibration components and the amplitude or phase of each harmonic to obtain the harmonic distortion corresponding to each harmonic component, including the corresponding amplitude deviation rate, phase drift and distortion energy ratio under environmental influence.

[0057] In an embodiment of the present invention, the harmonic distortion of each harmonic component of the electrical signal is evaluated based on the previously obtained offset coupling coefficient and mutual information value, and the weight coefficient is set. The offset coupling coefficient weight is 0.6, and the mutual information value weight is 0.4. Taking the 8th harmonic as an example, it is known that the offset coupling coefficient of temperature and the 8th harmonic amplitude is 0.15, and the mutual information value is 0.2 bit; the offset coupling coefficient of humidity and the 8th harmonic amplitude is 0.12, and the mutual information value is 0.18 bit; the offset coupling coefficient of the vibration frequency 50Hz and the 8th harmonic amplitude is 0.1, and the mutual information value is 0.15 To calculate the comprehensive impact factor of the eighth harmonic amplitude under environmental influences, first calculate the weighted value of each environmental component. For temperature, the weighted value is 0.6 × 0.15 + 0.4 × 0.2 = 0.09 + 0.08 = 0.17; for humidity, the weighted value is 0.6 × 0.12 + 0.4 × 0.18 = 0.072 + 0.072 = 0.144; and for a vibration frequency of 50 Hz, the weighted value is 0.6 × 0.1 + 0.4 × 0.15 = 0.06 + 0.06 = 0.12. Then, the weighted values of each environmental component are added together to obtain the comprehensive impact factor, which is 0.17 + 0.144 + 0.12 = 0.434. Based on the comprehensive impact factor, the amplitude deviation rate is calculated. Assuming that the original 8th harmonic amplitude is 2V, the amplitude after considering the environmental impact becomes 2×(1+0.434) = 2.868V, and the amplitude deviation rate is (2.868-2) / 2×100% = 43.4%. Similarly, the phase drift is calculated. Assuming that the original phase is 10°, it becomes 15° after calculation, and the phase drift is 5°. The distortion energy ratio is calculated according to the harmonic energy formula. ( The energy change of the 8th harmonic before and after considering environmental influences is calculated. Combined with the total harmonic energy, it is concluded that the 8th harmonic distortion energy accounts for 12%. This calculation is performed for all subharmonic components to obtain the corresponding harmonic distortion of each harmonic component, including the amplitude deviation rate, phase shift, and distortion energy ratio under environmental influences. This comprehensively assesses the impact of environmental factors on the electrical signal harmonics of low-impedance voltage transformers.

[0058] Furthermore, step S3 includes the following steps: Step S31: constructing a transformer adaptive evaluation model using a Kalman filter framework, taking the transformer dynamic eigenvector as a state variable, and using the electrical and environmental quantities in the valid real-time data set as input variables to drive the update of the state transfer matrix and observation matrix corresponding to the transformer adaptive evaluation model, while achieving recursive estimation and prediction of the state variables; In the embodiment of the present invention, the Kalman filter framework is used to construct a transformer adaptive evaluation model, and the previously obtained transformer dynamic feature vector is used as the state variable The vector contains data in 11 dimensions, including amplitude error, phase offset, and harmonic distortion, with electrical quantities (voltage, current fundamental wave and each harmonic amplitude, phase) and environmental quantities (temperature, humidity, vibration frequency and amplitude) in the effective real-time data set as input variables. and observed variables , set the state transfer matrix It is an 11×11 square matrix, and its initial value is the unit matrix; the observation matrix It is a 7×11 matrix. The initial value is determined according to the physical relationship between electrical quantities and state variables. The process noise covariance matrix Initialized to 0.01 times the identity matrix, measurement noise covariance matrix Initialized to 0.05 times the unit matrix, at time t=1, the electrical and environmental quantities in the effective real-time data set are obtained as input and observation , combined with the initial state estimate (obtained from historical data statistics), the status is predicted and updated using the following formula: ; ; ; ; ;in is an 11×7 control input matrix. The initial value is determined according to the system characteristics. Through the above recursive calculation, the state estimation value at time t=1 is obtained and the error covariance matrix As time goes by, the above process is repeated with new input and observation data to achieve recursive estimation and prediction of state variables and update the state transfer matrix and the observation matrix , so that it gradually approaches the real characteristics of the system.

[0059] Step S32: Acquire real-time status data corresponding to the low-impedance voltage transformer, and construct a parameter adaptation mechanism based on gradient descent. When the deviation between the real-time status data and the recursive estimation prediction exceeds a preset threshold, trigger an iterative update of the Kalman gain parameters of the transformer adaptive evaluation model to re-evaluate and generate the corresponding transformer real-time evaluation result. In an embodiment of the present invention, by obtaining real-time status data corresponding to a low-impedance voltage transformer, which contains information such as amplitude error, phase offset, and harmonic distortion of the same dimension as the dynamic eigenvector of the transformer, the real-time status data is compared with the recursive estimation prediction value of the previous Kalman filter model, and the deviation between the two is calculated. The preset threshold of the deviation is set to 0.1 (expressed in the normalized eigenvector norm). When the deviation calculated at a certain moment is 0.15, which exceeds the preset threshold, a parameter adaptation mechanism based on gradient descent is triggered. The mechanism updates the Kalman gain parameter through the following steps: defining a loss function ,in is the sample size, is the observed value, For the predicted value, calculate the loss function with respect to the Kalman gain Gradient , according to the gradient descent formula Update the Kalman gain, where the learning rate Set it to 0.01, and repeat the above steps until the deviation between the real-time status data and the recursive estimation prediction is less than or equal to the preset threshold. For example, after 5 iterations, the deviation drops to 0.08. At this time, stop the iteration and rerun the Kalman filter model with the updated Kalman gain parameter to generate the corresponding real-time evaluation result of the transformer, which contains information such as the health status and performance indicators of the transformer.

[0060] Step S33: quantifying the uncertainty state corresponding to the real-time evaluation result of the transformer by using a Bayesian inference method to evaluate and generate real-time evaluation state data of the transformer including corresponding confidence levels.

[0061] In the embodiment of the present invention, the uncertainty state corresponding to the real-time evaluation result of the mutual inductor is quantified by the Bayesian inference method, and the real-time evaluation result of the mutual inductor obtained previously is regarded as a random variable, assuming that it obeys the Gaussian distribution. ,in is the mean vector of the evaluation results, is the covariance matrix. According to Bayes’ theorem, the posterior probability ,in Indicates the true state of the transformer, Represents the observed data, and uses the predicted value of the Kalman filter model as the prior probability , taking real-time state data as likelihood function When calculating the likelihood function, it is assumed that the observation noise follows a Gaussian distribution , then the likelihood function can be expressed as: ,in is the observation data dimension, The posterior distribution is sampled by the Markov Chain Monte Carlo (MCMC) method to generate 1000 sample points, and the mean and covariance of these sample points are calculated to obtain the parameters of the posterior distribution. and Based on the posterior distribution, the 95% confidence interval is calculated. For example, for a certain performance indicator of the transformer, its 95% confidence interval is [0.78, 0.92], which means that there is a 95% probability that the indicator falls within this interval. These confidence intervals are used as confidence information and merged with the real-time evaluation results of the transformer to generate real-time evaluation status data of the transformer corresponding to the confidence level, providing a reliable basis for the operation and maintenance of low-impedance voltage transformers.

[0062] Furthermore, step S4 includes the following steps: Step S41: Acquire historical fault sample data corresponding to the low-impedance voltage transformer; In an embodiment of the present invention, historical fault sample data corresponding to a low-impedance voltage transformer is obtained from a dedicated historical data storage device. This storage device uses a solid-state drive (SSD) with a capacity of 1TB and is capable of fast data access. The historical fault sample data covers various fault conditions that have occurred in low-impedance voltage transformers over the past three years and includes 1,000 fault samples. Each sample records in detail the transformer's dynamic feature vector data at the time of the fault, including data in 11 dimensions such as amplitude error, phase offset, and harmonic distortion. It also records information such as the fault type, time of occurrence, and environmental parameters at the time. For example, one fault sample shows that at 2:20 pm on May 10, 2022, the transformer experienced a core saturation fault. At this time, the amplitude error at the fundamental frequency reached 25%, the phase offset was 18 degrees, the third harmonic distortion energy accounted for 15%, and the ambient temperature was 32°C and the humidity was 60% RH. These historical fault sample data provide a rich reference basis for subsequent comparative analysis.

[0063] Step S42: performing a comparative analysis of any two instantaneous errors between the real-time evaluation state data of the transformer and the historical fault sample data to obtain the evaluation state errors between any two real-time states of the transformer and the states corresponding to the historical faults; In an embodiment of the present invention, an instant error comparison analysis is performed on the previously obtained real-time evaluation status data of the transformer and the acquired historical fault sample data. The real-time evaluation status data includes information on 11 dimensions, such as the amplitude error, phase offset, and harmonic distortion of the transformer at the current moment, and carries corresponding confidence data. A sample in the historical fault sample data is used as a comparison object. Assuming that the core saturation fault sample on May 10, 2022 is selected, for the amplitude error dimension in the real-time evaluation status data, if the amplitude error at the current fundamental frequency is 10%, and the amplitude error at the frequency in the fault sample is 25%, then the error of this dimension is |10%-25%|=15%; for the phase offset dimension, if the current The phase offset is 5°, and it is 18° in the fault sample, so the error in this dimension is |5°-18°|=13°; the difference calculation is performed on the data of 11 dimensions in turn, and then the Euclidean distance formula is used to calculate the comprehensive evaluation state error. The square root of the sum of the squares of the errors in the 11 dimensions is taken to obtain the evaluation state error between the real-time state and the state corresponding to the historical fault. For example, after calculation, the evaluation state error of this comparison is 18.6%. The above comparison calculation is performed on each sample in the historical fault sample data with the real-time evaluation state data to obtain multiple evaluation state errors between any two real-time states of the transformer and the state corresponding to the historical fault, which fully reflects the degree of difference between the current state and the historical fault state.

[0064] Step S43: Calculate the corresponding transformer evaluation state error change rate based on the evaluation state error between any two real-time states of the transformer and the states corresponding to historical faults, and perform an error fault warning on the low-impedance voltage transformer based on the transformer evaluation state error change rate. When the transformer evaluation state error change rate is greater than 15%, an abnormal fault warning is triggered to generate an immediate state fault warning signal corresponding to the low-impedance voltage transformer.

[0065] In an embodiment of the present invention, the rate of change of the transformer evaluation state error is calculated based on multiple evaluation state errors obtained previously. Taking the evaluation state errors obtained by comparing two adjacent comparisons with historical fault sample data as an example, assuming that the evaluation state error obtained by the first comparison is 12%, and the evaluation state error obtained by the second comparison is 14%, the error change rate calculation formula is (14%-12%) / 12%×100%≈16.7%. Such calculation is performed on all two adjacent evaluation state errors to obtain a series of transformer evaluation state error change rates. The error change rate threshold is set to 15%. When a calculated error change rate is greater than 15%, such as 16.7% calculated above, the abnormal fault early warning mechanism is triggered. At this time, a warning signal is issued by a dedicated alarm device. The alarm device is equipped with an audio-visual alarm function, which emits a red flashing light and plays a buzzing sound of a specific frequency. At the same time, an immediate status fault warning signal corresponding to the low-impedance voltage transformer is generated. The signal contains information such as warning time, warning type (based on the error change rate exceeding the limit), current real-time assessment status data, etc., and is sent to the server of the monitoring center through a wireless network (such as a 4G communication module). The server records the warning information in the database and pushes it to the handheld terminal devices of relevant maintenance personnel so that the low-impedance voltage transformer can be inspected and maintained in time to prevent the occurrence or expansion of the fault.

[0066] 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.

[0067] 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 low-impedance voltage transformer evaluation method based on real-time data adaptive driving, characterized in that: The following steps are involved: Step S1: The primary side electrical signal and the secondary side electrical signal of the low-impedance voltage transformer corresponding to the low-impedance voltage transformer are collected in real time through the Rogowski coil and the resistor divider, and the temperature and humidity sensor and the vibration sensor are simultaneously connected to collect the corresponding temperature and humidity data and vibration data to generate a real-time transformer data set; Step S2: Performing dynamic quality assessment and screening on the transformer real-time data set to generate a quality-screened valid real-time data set; performing variational modal feature extraction on the valid real-time data set to obtain a transformer dynamic feature vector corresponding to the amplitude error, phase offset, and harmonic distortion; Step S3: constructing a transformer adaptive evaluation model based on the transformer dynamic feature vector, and using the effective real-time data set to drive the iterative update of the corresponding transformer adaptive evaluation model parameters, and evaluating and generating transformer real-time evaluation status data; Step S4: Obtain historical fault sample data corresponding to the low-impedance voltage transformer, and perform instant fault warning based on the real-time evaluation status data of the transformer and the historical fault sample data to generate an instant status fault warning signal corresponding to the low-impedance voltage transformer.

2. The low-impedance voltage transformer evaluation method based on real-time data adaptive driving according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: collecting the primary side electrical signal of the transformer corresponding to the low-impedance voltage transformer in real time through the differential sampling structure corresponding to the Rogowski coil; Step S12: constructing a four-level resistor divider network by four resistor dividers to collect the secondary side electrical signal of the transformer corresponding to the low-impedance voltage transformer in real time; Step S13: collecting temperature, humidity and vibration data corresponding to the surrounding environment of the low-impedance voltage transformer by synchronously accessing the temperature and humidity sensor and the vibration sensor; Step S14: The primary side electrical signal of the transformer, the secondary side electrical signal of the transformer, the temperature and humidity data, and the vibration data corresponding to the low-impedance voltage transformer are clock-synchronized through a field programmable gate array and merged into the same data set to generate a transformer real-time data set.

3. The low-impedance voltage transformer evaluation method based on real-time data adaptive driving according to claim 2 is characterized in that: Step S11 includes the following steps: Finite element simulation was used to optimize the spacing between the Rogowski coils and the thickness of the shielding layer above them. Based on the optimized spacing and the thickness of the shielding layer on the Rogowski coil, the leakage magnetic interference suppression evaluation is performed to obtain the primary side leakage magnetic interference suppression ratio corresponding to the Rogowski coil; The electrical sampling accuracy corresponding to the low-impedance working condition is determined based on the primary-side leakage magnetic interference suppression ratio corresponding to the Rogowski coil, and the primary-side electrical signal of the transformer corresponding to the low-impedance voltage transformer is collected in real time through the differential sampling structure corresponding to the Rogowski coil based on the electrical sampling accuracy.

4. The low-impedance voltage transformer evaluation method based on real-time data adaptive driving according to claim 1 is characterized in that: Step S2 includes the following steps: Step S21: performing soft threshold denoising processing on the electrical signal in the transformer real-time data set to generate a transformer denoised electrical signal; Step S22: Calculate the cross-correlation coefficient between the voltage and current signals of the transformer de-noised electrical signal. When the cross-correlation coefficient is less than 0.95, trigger the signal re-sampling mechanism and iterate the de-noising process to generate a transformer re-correlated electrical signal. Step S23: Calculate the corresponding environmental change rate based on the environmental parameters corresponding to temperature, humidity and vibration in the real-time data set of the mutual inductor, and perform abnormal environment elimination processing on the environmental parameters corresponding to temperature, humidity and vibration based on the environmental change rate to obtain the mutual inductor filtered abnormal environment data; Step S24: dynamically evaluating and screening the transformer's re-correlated electrical signals and transformer's filtered environmental data based on the data integrity and accuracy scores to generate a quality-screened valid real-time data set; Step S25: performing variational modal feature extraction on the valid real-time data set to obtain a mutual inductor dynamic feature vector corresponding to the amplitude error, phase offset, and harmonic distortion.

5. The low-impedance voltage transformer evaluation method based on real-time data adaptive driving according to claim 4 is characterized in that: Step S21 includes the following steps: The electrical signal in the transformer real-time data set is decomposed into five layers using db10 wavelet packets to generate five layers of transformer real-time electrical signal frequency sub-bands. Obtain corresponding transformer electrical signal sub-band kurtosis through the 5-layer transformer instantaneous electrical signal frequency sub-band, and perform noisy sub-band identification on the 5-layer transformer instantaneous electrical signal frequency sub-band based on the transformer electrical signal sub-band kurtosis to generate transformer electrical signal noisy frequency sub-band; Soft threshold denoising is performed based on the noisy frequency sub-band of the transformer electrical signal, and the electrical signal is reconstructed with the other transformer real-time electrical signal frequency sub-bands to generate a transformer denoised electrical signal.

6. The low-impedance voltage transformer evaluation method based on real-time data adaptive driving according to claim 4 is characterized in that: Step S25 includes the following steps: Step S251: performing variational modal decomposition on the electrical signal in the valid instantaneous data set to decompose the electrical signal into 8-12 eigenmodes, and screening 3-5 principal components related to low impedance characteristics through correlation coefficients; Step S252: Obtain the signal amplitude distribution and signal phase distribution corresponding to each principal component, and perform amplitude error and phase offset statistics based on the signal amplitude distribution and signal phase distribution corresponding to each principal component to obtain the amplitude error and phase offset corresponding to each principal component; Step S253: obtaining corresponding harmonic components from the electrical signal in the valid real-time data set, and performing harmonic distortion evaluation on the harmonic components corresponding to the electrical signal based on the environmental data in the valid real-time data set to obtain harmonic distortion corresponding to each harmonic component; Step S254: combining the amplitude error and phase offset corresponding to each main component and the harmonic distortion corresponding to each subharmonic component into the same eigenvector to obtain a mutual inductor dynamic eigenvector including the amplitude error, phase offset and harmonic distortion.

7. The low-impedance voltage transformer evaluation method based on real-time data adaptive driving according to claim 6 is characterized in that: The step S252 of performing amplitude error and phase offset statistics based on the signal amplitude distribution and signal phase distribution corresponding to each principal component includes the following steps: Perform time-frequency synchronization on the signal amplitude distribution and signal phase distribution corresponding to each principal component to obtain the amplitude distribution and phase distribution corresponding to each principal component at the same time-frequency; Perform amplitude error statistics between the amplitude distributions corresponding to each principal component at the same time and frequency to obtain the amplitude error corresponding to each principal component; The phase differences between the main components are obtained by the phase distributions corresponding to the main components at the same time-frequency. Based on the phase differences between the main components, the phase distributions corresponding to the main components at the same time-frequency are statistically analyzed to obtain the phase shifts corresponding to the main components.

8. The low-impedance voltage transformer evaluation method based on real-time data adaptive driving according to claim 6, characterized in that: The harmonic distortion evaluation of each harmonic component corresponding to the electrical signal based on the environmental data in the valid real-time data set in step S253 includes the following steps: Based on the environmental data in the valid real-time data set, an environmental-electrical offset coupling analysis is performed on each harmonic component corresponding to the electrical signal to analyze the gradient and vibration frequency corresponding to temperature and humidity. Based on the gradient and vibration frequency corresponding to temperature and humidity, an offset coupling evaluation is performed between the amplitude or phase corresponding to each harmonic component to obtain the offset coupling coefficient between each temperature, humidity and vibration component and each harmonic amplitude or phase; Based on the environmental data in the effective real-time data set, the environmental-electrical mutual information of each harmonic component corresponding to the electrical signal is calculated to obtain the mutual information value between each component of temperature, humidity and vibration and the amplitude or phase of each harmonic; Based on the offset coupling coefficient and mutual information value between the temperature, humidity, and vibration components and the amplitude or phase of each harmonic, the harmonic distortion of each harmonic component corresponding to the electrical signal is evaluated to obtain the harmonic distortion corresponding to each harmonic component, including the amplitude deviation rate, phase drift, and distortion energy ratio under the influence of the environment.

9. The low-impedance voltage transformer evaluation method based on real-time data adaptive driving according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: constructing a transformer adaptive evaluation model using a Kalman filter framework, taking the transformer dynamic eigenvector as a state variable, and using the electrical and environmental quantities in the valid real-time data set as input variables to drive the update of the state transfer matrix and observation matrix corresponding to the transformer adaptive evaluation model, while achieving recursive estimation and prediction of the state variables; Step S32: Acquire real-time status data corresponding to the low-impedance voltage transformer, and construct a parameter adaptation mechanism based on gradient descent. When the deviation between the real-time status data and the recursive estimation prediction exceeds a preset threshold, trigger an iterative update of the Kalman gain parameters of the transformer adaptive evaluation model to re-evaluate and generate the corresponding transformer real-time evaluation result. Step S33: quantifying the uncertainty state corresponding to the real-time evaluation result of the transformer by using a Bayesian inference method to evaluate and generate real-time evaluation state data of the transformer including corresponding confidence levels.

10. The low-impedance voltage transformer evaluation method based on real-time data adaptive driving according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: Acquire historical fault sample data corresponding to the low-impedance voltage transformer; Step S42: performing a comparative analysis of any two instantaneous errors between the real-time evaluation state data of the transformer and the historical fault sample data to obtain the evaluation state errors between any two real-time states of the transformer and the states corresponding to the historical faults; Step S43: Calculate the corresponding transformer evaluation state error change rate based on the evaluation state error between any two real-time states of the transformer and the states corresponding to historical faults, and perform an error fault warning on the low-impedance voltage transformer based on the transformer evaluation state error change rate. When the transformer evaluation state error change rate is greater than 15%, an abnormal fault warning is triggered to generate an immediate state fault warning signal corresponding to the low-impedance voltage transformer.

Images