A method and system for real-time correction of accuracy drift in current transformers

Abstract

This invention discloses a real-time correction method and system for the accuracy drift of current transformers, solving the problem that existing technologies struggle to accurately correct error drift in current transformers, leading to out-of-tolerance operation. The method includes: real-time acquisition of parameters from the operating current transformer to generate a time-series monitoring sequence with a unified time scale; calculation of characteristic parameters representing the dynamic changes in error based on the time-series monitoring sequence; selection of candidate models from a pre-set mathematical model library based on the characteristic parameters, and fitting the models using the least squares method to automatically select a current error correction curve describing the current error state; comparison of the current error correction curve with the original error curve calibrated at the factory of the current transformer, and graded correction judgment based on the comparison results. This achieves accurate identification, adaptive correction, and state assessment of current transformer error drift, extending the effective lifespan of the current transformer.

Description

Technical Field

[0001] This invention relates to the field of power measurement technology, and in particular to a method and system for real-time correction of accuracy drift of current transformers. Background Technology

[0002] Current transformers are key equipment in electricity metering, and their long-term accuracy directly affects trade settlement and power grid monitoring. Current transformers are calibrated at the factory according to standards, with a "factory error curve" and corresponding accuracy class. Error testing of current transformers determines whether they meet the accuracy class requirements, thus ensuring the accuracy of the testing.

[0003] CN113030832A describes a device and method for detecting composite errors in current transformers, comprising: a detection platform, a support plate, a primary winding power distributor, a secondary winding detector, and a control device. The detection platform has a detection slot, on which detection components are mounted. Each detection component includes a power terminal and a detection terminal, both of which are retractable. The support plate is slidably connected to the detection slot. A current transformer carrying unit is mounted on the support plate, each carrying unit supporting one current transformer under test. The primary winding power distributor is electrically connected to the power terminals of each detection component. The secondary winding detector is electrically connected to the detection terminals of each detection component via an adjustable load. The control device is electrically connected to both the primary winding power distributor and the secondary winding detector. This method allows for simultaneous composite error detection of multiple current transformers, saving time and effort. However, this method only performs factory testing on current transformers. During long-term operation, due to factors such as insulation aging, core degradation, load and temperature changes, the error characteristics of current transformers will slowly "drift" and gradually deviate from the original curve, which may lead to out-of-tolerance operation, resulting in measurement inaccuracies and safety risks. Summary of the Invention

[0004] The purpose of this invention is to solve the problem that it is difficult to accurately correct the error drift of current transformers in the prior art, which leads to the operation of current transformers beyond tolerance, measurement inaccuracy and safety risks. The invention provides a method and system for real-time correction of the accuracy drift of current transformers, which realizes accurate identification, adaptive correction and condition assessment of the error drift of current transformers, and extends the effective life of current transformers.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A method for real-time correction of accuracy drift in a current transformer includes the following steps: The parameters of the current transformer in operation are collected in real time, and a time-series monitoring sequence with a unified time scale is generated. Based on the time-series monitoring sequence, characteristic parameters representing the dynamic changes in error are calculated; Based on the feature parameters, candidate models are selected from a preset mathematical model library, and the least squares method is used to fit the model to automatically select the optimal fitting curve that describes the current error state as the current error correction curve. The current error correction curve is compared with the original error curve calibrated at the factory of the current transformer, and a graded correction judgment is made based on the comparison results.

[0006] The method executes automatically in a fixed cycle and dynamically updates the current error correction curve and correction judgment result using the latest operating data. It can run independently on resource-constrained embedded monitoring terminals, realizing localized, real-time intelligent analysis from data acquisition and curve reconstruction to correction decision-making.

[0007] Preferably, the graded correction judgment includes: if the current error correction curve is within the allowable error range corresponding to the factory accuracy level of the current transformer, the current accuracy level of the current transformer remains unchanged; if the current error correction curve exceeds the allowable error range, but can be mapped to the shape of the original error curve through a preset mathematical transformation rule, then a corresponding digital compensation parameter is generated; if the current error correction curve exceeds the allowable error range and cannot be associated with the original curve through a mathematical transformation rule, then the current transformer is abnormal.

[0008] As a preferred method, a sliding window statistical method based on a circular buffer is used to identify and remove outliers from the time-series monitoring sequence: a fixed-length memory buffer is allocated for each monitoring parameter, and the mean μ and standard deviation σ within the window are dynamically updated as new data is continuously written. Based on the preset discrimination criterion 3σ criterion, outliers caused by transient interference are automatically identified and removed to obtain a preprocessed data sequence.

[0009] A fixed-length data window is maintained for each monitoring parameter. The mean and standard deviation of the data within the window are dynamically calculated, and data points that exceed the reasonable range are identified and removed based on preset statistical criteria.

[0010] Preferably, the parameters of the operating current transformer collected include the three-phase ratio difference, the three-phase phase angle difference, and the temperature of the current transformer.

[0011] Preferably, S3 includes: obtaining the hourly average sequence of the difference over a past time period t; using recursive least squares or simplified linear regression to calculate the linear regression slope of the hourly average sequence of the difference over the observation period as the trend strength; calculating the root mean square value of the second difference of the hourly average of the difference or the quadratic coefficient of the fitted quadratic curve as the curvature strength; and calculating the correlation coefficient between the hourly average sequence of the difference and the temperature sequence as the temperature-error correlation feature.

[0012] Trend strength is obtained by calculating the slope of the linear regression of the error data sequence; curvature strength is obtained by calculating the coefficient of the quadratic term of the second difference of the sequence or the fitted quadratic curve; temperature-error correlation characteristics are obtained by calculating the correlation coefficient between the error sequence and the temperature sequence within a sliding window.

[0013] As a preferred option, when selecting the current error correction curve, the minimum sum of squared residuals and the highest coefficient of determination are used as the joint criteria.

[0014] Preferably, the preset mathematical model library includes linear models, exponential models, and polynomial models.

[0015] Preferably, the mathematical transformation rules include translation transformation, scaling transformation, and combinations thereof.

[0016] A real-time accuracy drift correction system for a current transformer includes: Embedded terminal and device, wherein the embedded terminal includes: The data acquisition module collects parameters of the current transformer in operation in real time; The data preprocessing module uses a sliding window statistical method based on a circular buffer to clean the collected real-time parameters online. The feature analysis module extracts error change features based on the cleaned parameters; The model fitting module establishes the current error correction curve based on the least squares method. The correction judgment module performs hierarchical processing based on the correlation between the current error correction curve and the original error curve.

[0017] Preferably, the data acquisition module includes an error sensor that is non-contactly mounted on the core of the current transformer under test, and a temperature sensor mounted on the primary conductor output terminal of the current transformer.

[0018] Therefore, the present invention has the following beneficial effects: by collecting the current signal and temperature signal of the secondary side of the current transformer, the ratio difference, angle difference and temperature monitoring sequence are generated in real time and preprocessed. Feature extraction is performed on the preprocessed data, and least squares fitting is performed based on the extracted features to obtain the current error correction curve. The current error correction curve is intelligently compared and correlated with the original error curve calibrated by the current transformer at the factory, and hierarchical judgment is performed to generate digital compensation parameters. This realizes online correction of recoverable errors and extends the effective life of the current transformer. It also provides accurate early warning of irreversible degradation, avoids misjudgment, and improves the scientificity and accuracy of condition assessment. Attached Figure Description

[0019] Figure 1 This is a flowchart of the real-time correction method in this invention.

[0020] Figure 2 This is a schematic diagram of the data preprocessing process in this invention.

[0021] Figure 3 This is a schematic diagram of the feature parameter extraction process in this invention.

[0022] Figure 4 This is a schematic diagram of the architecture of the real-time correction system in this invention.

[0023] In the diagram: 1. Embedded terminal; 2. Device end; 3. Data acquisition module; 4. Data preprocessing module; 5. Feature analysis module; 6. Model fitting module; 7. Correction and judgment module. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this application will be further described in detail below through embodiments and in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only one preferred embodiment of this application and are only used to explain this application. They do not limit the scope of protection of this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0025] Over long-term operation, the error characteristics of current transformers can shift, leading to metering inaccuracies and potential safety risks. Currently, error status management mainly relies on two methods: one is periodic offline verification, which requires power outages and cannot reflect the true state of continuous operation, resulting in monitoring blind spots; the other is online threshold monitoring, which can be performed without power outages, but is usually limited to over-limit alarms and lacks the ability to deeply analyze and model drift trends. It cannot accurately describe the drift curve, determine whether it is regular or compensable, or predict the remaining lifespan.

[0026] Therefore, existing technologies lack a complete online solution that automatically completes the entire process of "error tracking - curve reconstruction - state judgment - adaptive correction".

[0027] With the development of intelligent sensing and edge computing technologies, real-time intelligent analysis at the equipment site has become possible. Therefore, there is an urgent need for a new method that can be deployed on field equipment to dynamically reconstruct the current "calibration curve" through continuous analysis of operating data and intelligently compare it with the "factory output curve".

[0028] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0029] Example 1: This embodiment provides a method for real-time correction of accuracy drift in current transformers, such as... Figure 1As shown, the process includes the following steps: Step 1, real-time acquisition of parameters of the operating current transformer to generate a time-series monitoring sequence with a unified time scale; Step 2, calculation of characteristic parameters representing the dynamic changes in error based on the time-series monitoring sequence; Step 3, selection of candidate models from a preset mathematical model library based on the characteristic parameters, and fitting of the models using the least squares method to automatically select the current error correction curve describing the current error state; Step 4, comparison of the current error correction curve with the original error curve calibrated at the factory of the current transformer, and graded correction judgment based on the comparison results.

[0030] The real-time accuracy drift correction method for current transformers provided in this embodiment aims to overcome the shortcomings of existing technologies, such as monitoring blind spots caused by reliance on periodic offline verification, lack of adaptive modeling capability for long-term error evolution trends in online monitoring, and difficulty in intelligently correcting and classifying error drift. It can be deployed on field equipment and is suitable for devices with power metering and edge computing capabilities, such as smart fusion terminals, smart meters, and smart measurement switches. Through continuous analysis of operating data, it dynamically reconstructs the current "calibration curve" and intelligently compares it with the "factory curve." This method can not only assess in real time whether the original accuracy level is met, but also distinguish the nature of the drift: for systematic drift that is regular and can be mapped to the factory curve through mathematical transformation, accuracy should be restored through digital compensation; for irregular and uncontrollable degradation, it can accurately identify and warn of replacement, thereby promoting the transformation of the operation and maintenance mode from "periodic verification / out-of-tolerance replacement" to a more precise direction of "state perception, intelligent correction, and predictive maintenance."

[0031] The real-time accuracy drift correction method for current transformers provided in this embodiment is specially designed to run independently on resource-constrained embedded monitoring terminals, realizing localized and real-time intelligent analysis from data acquisition and curve reconstruction to correction decision-making.

[0032] The following examples and specific application scenarios further illustrate the technical solution and effects of the present invention. The following examples are explanations of the present invention, but the present invention is not limited to the following examples.

[0033] The first step is to collect the parameters of the current transformer in operation in real time and generate a time-series monitoring sequence with a unified time scale.

[0034] The current signal and temperature sensor data on the secondary side of the current transformer are collected synchronously, and the ratio difference, angle difference and temperature monitoring sequence are calculated and generated in real time.

[0035] In this embodiment, the obtained time-series detection sequence is further subjected to data cleaning and preprocessing using an embedded sliding window method. The specific process is as follows: Figure 2As shown, a sliding window statistical method based on a circular buffer is used for real-time data filtering. A fixed-length memory buffer is allocated for each monitoring parameter, and the mean μ and standard deviation σ within the window are dynamically updated as new data is continuously written. Outliers caused by transient interference are automatically identified and removed according to the preset discrimination criterion 3σ, and data continuity is achieved through previous value preservation or linear interpolation, forming a high-quality time-series dataset suitable for embedded environments.

[0036] The second step involves calculating characteristic parameters that characterize the dynamic changes in error, based on the time-series monitoring sequence obtained in the first step.

[0037] Efficient feature extraction is performed on the preprocessed time-series monitoring sequences. For example... Figure 3 As shown, the extracted feature parameters include: Trend strength: The slope of the error sequence over the observation period is calculated using recursive least squares or simplified linear regression.

[0038] Curvature intensity: The acceleration / deceleration characteristics of the changing trend are approximated by calculating the second-order difference of the sequence.

[0039] Correlation characteristics: Analyze the correlation between temperature series and error series to quantify the degree of influence of temperature on error drift.

[0040] The third step involves selecting candidate models from a pre-defined mathematical model library based on the feature parameters obtained in the second step, fitting the models using the least squares method, and automatically selecting the current error correction curve that describes the current error state.

[0041] The specific process includes: pre-setting a simplified mathematical model library containing basic function forms such as linear functions, exponential functions, and polynomial functions; based on the extracted long-term feature sequences, using the least squares method for fitting to solve for the model parameters; and minimizing the sum of squared residuals (RSS) and the coefficient of determination (R²) for fitting. 2 The highest criterion is the joint criterion, which automatically selects the optimal mathematical model that best describes the current error evolution pattern as the current error correction curve.

[0042] The fourth step is to compare the current error correction curve with the original error curve calibrated at the factory of the current transformer, and to make a graded correction judgment based on the comparison results.

[0043] The specific process is as follows: Figure 1As shown, the process includes: intelligently comparing and analyzing the correlation between the current error correction curve and the original error curve calibrated at the factory of the current transformer, and performing a graded judgment: if the current error correction curve is within the allowable error range corresponding to the factory accuracy level of the current transformer, then the current accuracy level of the current transformer is determined to remain unchanged, and the current transformer is in normal condition; if the current error correction curve exceeds the allowable error range, but can be mapped to the shape of the original error curve through preset mathematical transformation rules (such as translation, scaling, or a combination thereof), then corresponding digital compensation parameters are generated to compensate the transformer's real-time output to restore it to the factory accuracy level; if the current error correction curve exceeds the allowable error range and cannot be correlated with the original curve through mathematical transformation rules, then the current transformer is determined to be abnormal and a replacement prompt is generated.

[0044] The real-time accuracy drift correction method for current transformers provided in this embodiment has the following beneficial effects: (1) It realizes online tracking and edge intelligent correction of error drift, and relies on the traditional mode of offline verification and simple over-limit alarm. It can deeply identify and classify the nature of error drift, which significantly improves the accuracy of state perception and the autonomy of decision-making.

[0045] (2) An adaptive correction mechanism based on curve correlation was constructed, and an innovative hierarchical judgment strategy of "comparison of correlation between correction curve and factory curve" was proposed. Systematic and regular drifts were identified through mathematical transformation rules, and digital compensation parameters were generated, realizing online software correction of recoverable errors and extending the effective life of equipment; accurate early warning of irreversible degradation was provided, avoiding misjudgment and improving the scientificity and accuracy of condition assessment.

[0046] (3) It has promoted the transformation of operation and maintenance model towards status-based and precise operation and maintenance: the real-time accuracy status judgment and compensability conclusions provided provide direct and quantitative basis for operation and maintenance decision-making. It has promoted the transformation of strategy from "periodic verification" and "replace when out of tolerance" to "status perception, intelligent correction and predictive maintenance", which helps to achieve lean management of the entire life cycle of assets, reduce operation and maintenance costs, and ensure measurement fairness.

[0047] Example 2: Based on Embodiment 1, this embodiment provides a real-time accuracy drift correction system for current transformers, which is used to implement the real-time accuracy drift correction method for current transformers in Embodiment 1.

[0048] Specifically, a real-time accuracy drift correction system for current transformers, such as... Figure 4As shown, it includes an embedded terminal 1 and a device 2. The embedded terminal can run a complete closed-loop analysis process from data perception to intelligent correction. The embedded terminal includes a data acquisition module 3, a data preprocessing module 4, a feature analysis module 5, a model fitting module 6, and a correction judgment module 7. Each module adopts a lightweight algorithm design and works together to directly output the current error correction curve and accuracy status classification judgment results on the device, thereby providing real-time and autonomous decision support for on-site status maintenance and precise intervention.

[0049] The embedded terminal utilizes its internal high-precision analog and data acquisition modules to achieve synchronous acquisition and localized processing of multi-source data. Through its built-in high-precision ADC module and digital interface, it synchronously and in real-time acquires current signals from the secondary side of the transformer and ambient temperature data. Within the embedded terminal, it calculates and generates ratio difference, phase difference, and temperature monitoring sequences in real time. All data is time-aligned and formatted for storage on the device, providing a reliable local data source for subsequent analysis.

[0050] The embedded terminal utilizes its internal data preprocessing module to perform embedded sliding window data cleaning and preprocessing. It employs a sliding window statistical method based on a circular buffer to perform real-time data filtering and online cleaning of the real-time data stream. By dynamically calculating statistical features and applying discrimination criteria, it automatically removes outliers while ensuring data continuity, resulting in a high-quality input data sequence.

[0051] The main features are as follows: a fixed-length memory buffer is allocated for each monitoring parameter, and the mean μ and standard deviation σ within the window are dynamically updated as new data is continuously written. Based on the preset discrimination criterion 3σ, outliers caused by transient interference are automatically identified and eliminated, and data continuity is achieved through previous value preservation or linear interpolation, forming a high-quality time-series dataset suitable for embedded environments.

[0052] The embedded terminal integrates a feature analysis module to achieve lightweight feature parameter extraction: Tailored to the computational characteristics of embedded systems, lightweight feature extraction is performed on preprocessed time-series data. Optimized algorithms suitable for embedded processors are employed to calculate key parameters, including trend strength, curvature strength, and temperature-error correlation features, to quantitatively describe the dynamic evolution of the error. All calculations are accelerated using fixed-point optimization or lightweight floating-point libraries.

[0053] A simplified mathematical model library containing basic forms such as linear, exponential, logarithmic, and polynomial models is embedded in the Flash memory of the embedded terminal, and a model fitting module is configured. This model fitting module is used to implement embedded adaptive model fitting and correction curve construction: based on the extracted long-term historical feature sequence, a least-squares fitting program optimized for embedded processors is called to quickly solve the model parameters and automatically select the optimal mathematical model that best describes the current error evolution law as the current error correction curve.

[0054] The embedded terminal integrates a calibration and judgment module to achieve graded calibration and judgment of the device's accuracy status. It intelligently compares and analyzes the correlation between the current error correction curve and the original error curve calibrated at the factory of the current transformer, and completes the status classification (normal / compensable / abnormal) judgment according to preset rules, generating corresponding digital compensation parameters or replacement prompts. The judgment results and compensation parameters are stored locally on the terminal and can be reported.

[0055] The embedded terminal stores the correction judgment results, compensation parameters and abnormal prompt information in local non-volatile memory, and can upload them to the background system or make local instructions through its communication interface, thus completing the closed loop from edge perception to intelligent correction.

[0056] The above architecture supports the real-time correction method for current transformer accuracy drift provided in Embodiment 1, which is executed automatically and cyclically at a set period. It uses the latest continuously collected operating data to continuously update the correction curve and judgment results, thereby realizing continuous online monitoring and adaptive correction management of the accuracy status of the current transformer.

[0057] The real-time accuracy drift correction system for current transformers provided in this embodiment has the following beneficial effects: (1) Online tracking and edge intelligent correction of error drift were realized: By deploying the complete “monitoring-modeling-judgment-correction” process on the embedded terminal, real-time and continuous tracking and local intelligent analysis of the error characteristics of the current transformer were realized. This completely changed the traditional mode of relying on offline verification and simple over-limit alarm, and enabled in-depth identification and classification of the nature of error drift, significantly improving the accuracy of state perception and the autonomy of decision-making.

[0058] (2) Excellent engineering practicality and deployment flexibility: The overall solution fully considers the resource constraints of embedded devices. Through the design of sliding window statistics, lightweight feature extraction, and optimized fitting algorithms, it ensures efficient and stable operation under limited resources. The system has a high degree of automation, can work independently or in collaboration with the backend, and is very suitable for large-scale deployment in smart substations, distribution areas and other scenarios, providing an efficient and reliable technical tool for the intelligent operation and maintenance of power metering equipment.

[0059] Example 3: This embodiment provides a real-time accuracy drift correction system for current transformers. Based on Embodiment 2, the modules inside the embedded terminal are further described in detail.

[0060] This embodiment implements a real-time accuracy drift correction method for current transformers using an embedded intelligent monitoring terminal deployed at the target current transformer site. The embedded intelligent monitoring terminal uses an industrial-grade microcontroller as its core platform. It is an independent edge hardware platform that performs data acquisition, processing, analysis, and correction judgment, integrating a data acquisition module, a data preprocessing module, a feature analysis module, a model fitting module, and a correction judgment module.

[0061] The data acquisition unit includes an error acquisition unit and a temperature acquisition unit.

[0062] The error acquisition unit employs a miniature error sensor based on the zero flux compensation principle, which is non-contactly mounted on the core of the transformer under test to sense the error magnetic field. Its output ratio difference (ε%) and angle difference (δ) analog current signals are sampled once per second by the terminal's built-in 24-bit high-precision Σ-Δ ADC.

[0063] The temperature acquisition unit uses a digital temperature sensor mounted close to the primary conductor output terminal of the current transformer to acquire temperature data via I... 2 The C interface connects to the microcontroller to acquire temperature (T) data at the same frequency.

[0064] All sampled data collected by the data acquisition unit is tagged with the same precise time stamp inside the embedded terminal to form a time-series data packet (t, ε%, δ, T), and temporarily stored in the local buffer.

[0065] The embedded intelligent monitoring terminal also includes peripheral units, including a real-time clock (RTC), a ferroelectric memory (FRAM) for non-volatile data storage, and a 4G communication module for data uploading.

[0066] During operation, the embedded intelligent monitoring terminal performs synchronous data acquisition and localized processing. The error sensor senses the error magnetic field of the transformer core in a non-contact manner, working synchronously with the temperature sensor. The microcontroller samples and calculates the signals, generating real-time data on the ratio difference (ε%), angle difference (δ), and temperature per minute. This data is then appended with a precise UTC time stamp generated by the RTC, forming a complete record {time, ε%, δ, T}, which is temporarily stored in a circular data queue allocated in SRAM.

[0067] The embedded intelligent monitoring terminal's preprocessing module performs real-time cleaning of the data stream. The system creates circular buffers of length 10080 (7 days' worth of data) for each of the three parameters: specificity difference, angle difference, and temperature. As new data is continuously written, the mean (μ) and standard deviation (σ) of the window are dynamically updated recursively. Outliers are identified and removed in real-time according to the 3σ criterion, and continuity is ensured using the previous value preservation method. Finally, a high-quality hourly data sequence is output for subsequent analysis.

[0068] Every day at dawn, the system calculates the hourly average of the cleaned and valid minute data from the past 24 hours to form a more stable hourly data sequence for subsequent feature analysis, thereby reducing data fluctuations.

[0069] The embedded intelligent monitoring terminal initiates lightweight feature extraction on a weekly basis. The system retrieves the hourly average sequence of the ratio difference over the past four weeks (a total of 672 data points), and uses a recursive least squares method optimized for embedded environments to linearly fit the four-week sequence, calculating the linear regression slope as the trend strength. Assuming the fitted straight line equation is the ratio difference equal to the product of 0.0012 and time t minus 0.035, then the slope of 0.0012% / hour represents the trend strength for this week, indicating an average error drift of 0.0012% per hour. The root mean square value of the second difference of the sequence is calculated as the curvature strength.

[0070] Assuming a calculated result of 0.0003, this value is used as the curvature intensity. A small value indicates a relatively flat recent error trend, without significant accelerated degradation. The correlation coefficient between the ratio difference and the temperature sequence, i.e., the Pearson correlation coefficient, is calculated as a temperature-error association feature. Assuming a correlation coefficient of 0.65, this indicates a positive correlation between temperature and error; rising temperature may cause a positive shift in error. Every Sunday at midnight, the system initiates a feature extraction task, using the above three feature parameters to form a feature vector describing the dynamic evolution of the error.

[0071] The embedded intelligent monitoring terminal performs adaptive model fitting based on long-term historical characteristics. The system calls upon the trend intensity sequence (26 data points) from the past six months (26 weeks), utilizing a library of linear (i.e., linear function), exponential (i.e., exponential function), and polynomial (i.e., polynomial function) models stored in Flash memory, and performs fitting using a least-squares program optimized for embedded systems. The minimum sum of squared residuals (RSS) and coefficient of determination (R²) are used as the benchmarks. 2 The highest criterion is used to select the optimal model.

[0072] Calculations showed that the exponential model met the requirements, so the system automatically selected it as the optimal model and fitted the parameters. The parameter 'a' of the exponential model was 0.0010, and the parameter 'b' was five times the natural constant e to the power of negative six. This exponential model is defined as the current error correction curve (here, the trend evolution curve) describing the long-term drift trend of the current error.

[0073] The embedded intelligent monitoring terminal performs accuracy status classification correction judgment on the device side. It reads the original error curve data of the current transformer calibrated at the factory from the FRAM of the embedded intelligent monitoring terminal (the ratio difference limit range is -0.35% to +0.35% under 5%~120% rated current).

[0074] The system compares the current measured and corrected error value (based on the latest monitoring data) with the factory limit. It finds that at 50% rated load, the current error is 0.42%, exceeding the allowable range of level 0.2 (+0.35%). The system then performs a morphological comparison between the current error correction curve (exponential model) and the factory error curve (a set of discrete limit points). The algorithm attempts to find a set of transformation parameters (translation Δ, scaling factor k) such that, after scaling and translation, the current error correction curve falls within the envelope of the factory curve to the greatest extent possible.

[0075] Calculations revealed a set of parameters with a scaling factor of 0.95 and a shift of -0.05, ensuring that the majority of points on the transformed current error correction curve fall within the factory limit range. Therefore, the system determines that the current error drift is systematic and regular, and its status is "compensable." The system uses the aforementioned transformation parameters (scaling factor of 0.95 and shift of -0.05) as digital compensation parameters. During subsequent real-time measurement, the invasive intelligent monitoring terminal will automatically apply compensation to the ratio difference calculated from the original sampling: the compensated ratio difference is the product of the original sampled ratio difference and the scaling factor, minus the shift, thereby restoring the output value to within the 0.2 level accuracy range.

[0076] The system packages and stores data such as the status judgment result ("compensable"), compensation parameters, and the time when the extrapolated prediction (based on the current correction curve) error may reach the uncompensable critical point (e.g., expected in 180 days) in FRAM, and uploads it to the backend operation and maintenance platform via the 4G module. The local LED indicator on the terminal displays yellow (indicating a corrected status).

[0077] Finally, the above process of "data cleaning and preprocessing - feature parameter extraction - adaptive model fitting and calibration curve construction - accuracy status classification and correction judgment" is executed automatically once a week. The latest week's data will be included in the calculations for the following week, and the features, model, and judgment results will be updated continuously. If, in a future cycle, the system finds that it cannot find effective mathematical transformation parameters to map the calibration curve to the factory curve, the status will be judged as "abnormal," a red alarm will be generated, and a replacement will be prompted; the LED indicator will turn red. This achieves continuous online monitoring, intelligent correction, and proactive management of the current transformer's accuracy status.

[0078] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Other variations and modifications are possible without departing from the technical solutions described in the claims.

Claims

1. A method for real-time correction of current transformer accuracy drift, characterized in that, include: The parameters of the current transformer in operation are collected in real time, and a time-series monitoring sequence with a unified time scale is generated. Based on the time-series monitoring sequence, characteristic parameters representing the dynamic changes in error are calculated; Based on the aforementioned feature parameters, candidate models are selected from a preset mathematical model library, and the least squares method is used to fit the model to automatically select the current error correction curve. The current error correction curve is compared with the original error curve calibrated at the factory of the current transformer, and a graded correction judgment is made based on the comparison results.

2. The method of claim 1, wherein, The graded correction judgment includes: if the current error correction curve is within the allowable error range corresponding to the factory accuracy level of the current transformer, the current accuracy level of the current transformer remains unchanged; if the current error correction curve exceeds the allowable error range, but can be mapped to the shape of the original error curve through a preset mathematical transformation rule, then a corresponding digital compensation parameter is generated; if the current error correction curve exceeds the allowable error range and cannot be associated with the original curve through a mathematical transformation rule, then the current transformer is abnormal.

3. The method of claim 1, wherein, An outlier identification and removal method based on a circular buffer sliding window is used to identify and remove outliers from the time-series monitoring sequence: a fixed-length memory buffer is allocated to each monitoring parameter, and the mean μ and standard deviation σ within the window are dynamically updated as new data is continuously written. Based on the preset discrimination criterion 3σ criterion, outliers caused by transient interference are automatically identified and removed to obtain a preprocessed data sequence.

4. The method for real-time correction of accuracy drift of a current transformer according to claim 1 or 2 or 3, characterized in that, The parameters collected from the operating current transformers include the three-phase ratio difference, the three-phase phase angle difference, and the temperature of the current transformers.

5. The method of claim 4, wherein, S3 includes: obtaining the hourly average sequence of the ratio difference over a past time period t; using recursive least squares or simplified linear regression to calculate the linear regression slope of the hourly average sequence of the ratio difference within the observation period as the trend strength; calculating the root mean square value of the second difference of the hourly average of the ratio difference or the quadratic coefficient of the fitted quadratic curve as the curvature strength; and calculating the correlation coefficient between the hourly average sequence of the ratio difference and the temperature sequence as the temperature-error correlation feature.

6. The method for real-time correction of accuracy drift of a current transformer according to claim 1 or 2 or 3, characterized in that, When selecting the current error correction curve, the minimum sum of squared residuals and the highest coefficient of determination are used as the joint criteria.

7. The method for real-time correction of accuracy drift of a current transformer according to claim 1 or 2 or 3, characterized in that, The preset mathematical model library includes linear models, exponential models, and polynomial models.

8. The method of claim 2, wherein, The mathematical transformation rules include translation transformation, scaling transformation, and combinations of both.

9. A current transformer accuracy drift real-time correction system, adopting the current transformer accuracy drift real-time correction method of any one of claims 1-8, characterized in that, Includes an embedded terminal and a device, wherein the embedded terminal includes: The data acquisition module collects parameters of the current transformer in operation in real time; The data preprocessing module uses a sliding window statistical method based on a circular buffer to clean the collected real-time parameters online. The feature analysis module extracts error change features based on the cleaned parameters; The model fitting module establishes the current error correction curve based on the least squares method. The correction judgment module performs hierarchical processing based on the correlation between the current error correction curve and the original error curve.

10. The real-time correction system for current transformer accuracy drift according to claim 9, wherein, The data acquisition module includes an error sensor that is non-contactly mounted on the core of the current transformer under test, and a temperature sensor installed at the output end of the primary conductor of the current transformer.

Images