Electromagnetic transformer state tracking and prospective verification method and system based on digital twinning

Abstract

The invention provides an electromagnetic transformer state tracking and foresight verification method and system based on digital twinning. The system comprises an on-site lightweight acquisition module, a digital twin construction and calibration module, an intelligent analysis and prediction module and a visual application module, wherein the on-site lightweight acquisition module is responsible for acquiring on-site lightweight state data of a mutual inductor; the digital twin construction and calibration module constructs and dynamically updates a high-fidelity virtual model consistent with a physical entity based on a multi-physical field model and a parameter identification algorithm; the intelligent analysis and prediction module performs all-condition simulation and trend prediction by using the calibrated virtual model; and the visual application module is used for generating a diagnosis report and early warning information and providing decision support for the user. The invention aims to solve the problems of low efficiency and lack of deep diagnosis and prospective prediction capabilities of the existing electromagnetic transformer verification method, and realizes accurate mapping of the internal state and full-working-condition performance simulation and prediction of the electromagnetic transformer verification method.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power equipment monitoring, and particularly relates to an electromagnetic mutual inductor state tracking and prospective calibration method and system based on digital twinning. BACKGROUND

[0002] In the power system measurement and protection system, the precision and reliability of the electromagnetic mutual inductor are crucial, and its periodic calibration is a core link to ensure the safe and stable operation of the power grid. At present, the electromagnetic mutual inductor state calibration generally adopts comparison method and indirect measurement method, which all need the operation and maintenance personnel to carry equipment to the site for testing, and there are significant bottlenecks in the calibration efficiency and the prediction of the equipment state. The existing technical system is difficult to realize the operation and maintenance mode upgrade from "periodic maintenance" to "accurate early warning" when facing a large number of mutual inductor assets distributed in a wide range.

[0003] Therefore, it is imminent to develop a new calibration method deeply integrating physical mechanism and digital technology. The method needs to be able to construct a virtual mirror of the mutual inductor, and to complete the accurate calibration of the model through the minimum on-site test data, and then to realize the full-condition virtual calibration, internal state deep diagnosis and accurate prediction of performance degradation trend, forming a systematic solution that can accurately map, deeply trace and prospectively predict the overall health status of the electromagnetic mutual inductor, and providing core technical support for the transformation of the power system operation and maintenance strategy from "periodic prevention" to "prospective maintenance". SUMMARY

[0004] The purpose of the application is to provide an electromagnetic mutual inductor state tracking and prospective calibration method and system based on digital twinning, aiming to solve the problems of low efficiency, lack of deep diagnosis and prospective prediction ability of the existing electromagnetic mutual inductor calibration method, and to realize the accurate mapping of the internal state, full-condition performance simulation and prediction.

[0005] To achieve the above purpose, the technical scheme of the application is as follows:

[0006] An electromagnetic mutual inductor state tracking and prospective calibration method based on digital twinning, specifically comprising the following steps:

[0007] S1, on-site light data acquisition: acquiring the key state data of the mutual inductor to be tested by a portable intelligent calibrator, including the excitation characteristic, basic error and winding direct current resistance, and uploading the encrypted data to a digital twinning platform after encryption and packaging;

[0008] S2, digital twin construction and calibration: the digital twin platform constructs an initial multi-physical field model based on the design parameters of the transformer and physical laws; and uses a parameter identification algorithm to fit the field data collected in step S1 with the model output, dynamically adjusts the internal parameters of the model, so that the digital twin accurately reflects the current real state of the physical transformer;

[0009] S3, intelligent analysis and prediction: using the calibrated high-fidelity digital twin to simulate various extreme working conditions and load conditions in virtual space, calculating the full-range error characteristics of the transformer, generating a virtual calibration report; locating the defect root cause according to the calibration data and historical calibration data and generating a health diagnosis report; based on the analysis of the performance degradation trend of the historical calibration data, predicting the remaining life and failure risk of the transformer;

[0010] S4, visualization application and decision: the platform automatically generates a comprehensive report including the calibration report, health diagnosis report and predictive maintenance suggestion, and visualizes the report through the front-end interface; automatically triggers an early warning for the identified high-risk equipment, and guides the maintenance personnel to perform accurate maintenance.

[0011] Preferably, the encrypted data is transmitted to the cloud digital twin platform through a wireless network, including 4G, 5G or Internet of Things network.

[0012] Preferably, step S2 is as follows:

[0013] S21, initial model construction: the platform constructs an initial multi-physical field model integrating electromagnetic field, circuit and thermodynamic effects according to the design drawings, material parameters and rated parameters of the transformer;

[0014] S22, parameter identification optimization: an optimization algorithm is used to minimize the difference between the model output and the field measured data as the objective function, and the optimal internal parameter set of the model is solved in reverse.

[0015] Preferably, the parameter identification optimization process uses CMA-ES algorithm, which includes the following steps:

[0016] S221, initialization: set the initial search distribution as a multivariate normal distribution N(m (o) , σ (o)2 C (o) ), wherein m (o) is the mean of the initial solution of the parameter vector, σ (o) is the initial step size, and C (o) is the initial covariance matrix;

[0017] S222, sampling and evaluation: the gth generation is sampled from the current distribution N(m (g) , σ (g)2 C (g)) independently sampled to generate λ offspring P k (g) , k = 1, …, λ, for each offspring individual, run the digital twin model, calculate its objective function value Loss(P k (g) );

[0018] S223, selection and recombination: sort the λ offspring according to the loss value, select μ elite individuals with the smallest loss value, and use the weighted sum of the selected elite individuals to update the mean value m of the search distribution to determine the center position of the next generation search distribution; the weight of the elite individual is determined according to the loss value ranking, and the individual with smaller loss value is given higher weight;

[0019] S224, covariance matrix adaptation: update the covariance matrix C using the information of the current generation of elite individuals and historical information;

[0020] S225, step size control: by comparing the actual length of the conjugate evolution path with its expected length under random selection, it is judged whether the step size σ needs to be increased or decreased; if the actual length of the conjugate evolution path exceeds the expected length, the step size is increased to speed up the convergence; if the actual length of the conjugate evolution path is less than the expected length, the step size is reduced to perform more detailed local search;

[0021] S226, iteration and termination: let g = g + 1, repeat steps S222 to S225 until the preset termination condition is met; finally output the optimal mean value m as the optimal parameter vector P identified.

[0022] Preferably, step S3 is specifically as follows:

[0023] S31, full working condition virtual verification: drive the digital twin M to simulate running in the cloud under complex working conditions including overload, extreme temperature and harmonic, calculate its full range error curve including the ratio error curve and phase error curve of the transformer, and generate a virtual verification report;

[0024] S32, deep health diagnosis: compare the current calibration parameter P with the historical parameter P _historical , quantitatively analyze the parameter change, locate the defect source, and generate a health diagnosis report;

[0025] S33, performance trend prediction: based on the sequence of previous calibration parameters, adopt a time series prediction model or a machine learning algorithm, fit the performance degradation trajectory, predict the remaining useful life RUL and recommend the personalized next verification time T _next .

[0026] Preferably, step S33 is specifically as follows:

[0027] S331, define health index and failure criteria:

[0028] From the calibration results, a key parameter is selected as the health index, which monotonically changes over time and reflects the device aging;

[0029] Set the failure threshold of the health index;

[0030] S332, build a prediction model:

[0031] The historical calibration time points t1, t2,..., t n And the corresponding health index values HI1, HI2,..., HI n As training data; and train the model based on time series prediction algorithm to learn the trend of health index degradation over time;

[0032] Step 333, predict the remaining useful life RUL:

[0033] Using the trained model, predict the future trend curve of the health index;

[0034] The calculation method of the remaining useful life RUL is: from the current time point, to the future time point when the predicted health index curve first touches or exceeds the failure threshold;

[0035] Step 334, recommend the next calibration time T_next according to the predicted remaining useful life RUL.

[0036] Preferably, the time series prediction algorithm includes linear regression, support vector regression or neural network.

[0037] Preferably, the next calibration time T_next is recommended based on the safety factor, and the specific calculation is as follows:

[0038] T_next = current time + RUL × k

[0039] Wherein, k is a safety factor less than 1.

[0040] Preferably, step S4 is as follows:

[0041] S41, automatic report generation: the system automatically integrates virtual calibration report, health diagnosis report and trend prediction result to generate a structured device status comprehensive report;

[0042] S42, multi-terminal visual warning: through the front-end interface including Web interface and mobile App, the real-time state, risk level and prediction trend of the device are displayed in various forms including charts and dashboards, and the high-risk device is automatically notified of the warning;

[0043] S43, decision support and closed loop: specific maintenance recommendations are included in the report and early warning information, and the maintenance personnel perform on-site maintenance according to the maintenance recommendations, and the new data generated after maintenance are used as the input of the next calibration, thereby forming a closed loop management of continuous optimization.

[0044] An electromagnetic mutual inductor state tracking and forward-looking calibration system based on digital twinning, which is realized by any of the above-mentioned electromagnetic mutual inductor state tracking and forward-looking calibration methods based on digital twinning, comprising a field lightweight acquisition module, a digital twin construction and calibration module, an intelligent analysis and prediction module, and a visualization application module:

[0045] The field lightweight acquisition module is responsible for acquiring lightweight state data of the mutual inductor in the field;

[0046] The digital twin construction and calibration module constructs and dynamically updates a high-fidelity virtual model consistent with the physical entity based on a multi-physical field model and a parameter identification algorithm;

[0047] The intelligent analysis and prediction module uses the calibrated virtual model to perform full-condition simulation and trend prediction;

[0048] The visualization application module generates a diagnosis report and early warning information to provide decision support for users.

[0049] Compared with the prior art, the present application has the following beneficial effects:

[0050] (1) The present application adopts a deep state tracking technology based on digital twinning and parameter identification algorithm, realizing a fundamental leap from "black box" calibration to "white box" diagnosis of electromagnetic mutual inductors. Traditional calibration methods can only measure the overall output error of the mutual inductor, and cannot understand the specific reasons for the internal state change, while the present method can accurately quantify the small changes of internal key parameters such as winding resistance and core magnetic properties by constructing a high-fidelity multi-physical field model and using field lightweight test data to calibrate the model parameters, thereby realizing accurate positioning of the fault root cause and deep health assessment. This method not only greatly improves the depth and insight of the calibration, but also reveals the internal relationship between performance degradation and internal parameter changes, which is difficult to achieve by traditional methods. In addition, the parameter identification algorithm can adaptively adjust the model to match the true state of the physical entity, which means that the system can continuously maintain an accurate mapping of the current health status of the mutual inductor, providing a solid data foundation for subsequent predictive maintenance. This feature makes the present method more accurate and reliable in device state management and fault warning, and can continuously provide accurate state assessment in complex and variable operating environments.

[0051] (2) This invention innovatively introduces performance trend prediction and full-condition cloud simulation functions based on machine learning, which is revolutionary in the field of electromagnetic transformer calibration. This intelligent prediction module can not only drive machine learning models based on historical calibration data to accurately predict the performance degradation trajectory and remaining life of the transformer, realizing the strategic shift from "regular preventive maintenance" to "proactive condition-based maintenance", but also use the calibrated digital twin to complete performance simulation under various extreme and dangerous conditions in virtual space, replacing the heavy and risky physical tests. This method greatly improves the safety and efficiency of calibration, enabling operation and maintenance personnel to see potential fault risks weeks or even months in advance and formulate the optimal maintenance strategy and resource allocation plan. In addition, the system can also dynamically and personalized recommend the best calibration cycle according to the unique performance degradation mode of each device, thereby maximizing equipment utilization and reducing the total life cycle operation and maintenance cost while ensuring safety. Through intelligent trend prediction and comprehensive virtual simulation, this method not only greatly improves the foresight and economy of calibration, but also enhances the reliability of power grid operation and further promotes the intelligent upgrade of power system operation and maintenance mode. Attached Figure Description

[0052] Figure 1 This is a system architecture diagram of the electromagnetic current transformer state depth tracking and prospective verification system based on digital twins, as described in this invention. Detailed Implementation

[0053] The following is in conjunction with the appendix Figure 1 The technical solution of the present invention will be described in detail below.

[0054] This invention proposes a method for state tracking and prospective verification of electromagnetic current transformers based on digital twins, specifically including the following steps:

[0055] S1. Lightweight on-site data acquisition: Maintenance personnel use a portable intelligent calibrator to quickly test the current transformer on-site, collect key status data of the current transformer under test, including excitation characteristics, basic error, and winding DC resistance, and upload the encrypted data to the digital twin platform after encryption and packaging.

[0056] S2. Digital Twin Construction and Calibration: The digital twin platform constructs an initial multiphysics model based on the design parameters and physical laws of the transformer; and uses a parameter identification algorithm to fit the field data collected in step S1 with the model output, dynamically adjusting the internal parameters of the model so that the digital twin accurately reflects the current real state of the physical transformer.

[0057] S3. Intelligent Analysis and Prediction: Using a calibrated high-fidelity digital twin, various extreme operating conditions and load conditions are simulated in virtual space to calculate the full-range error characteristics of the instrument transformer and generate a virtual verification report; based on the current calibration data and historical calibration data, the root cause of defects is located and a health diagnosis report is generated; based on historical calibration data, the performance degradation trend is analyzed and the remaining life and failure risk of the instrument transformer are predicted.

[0058] S4. Visual Application and Decision Making: The platform automatically generates a comprehensive report that includes calibration reports, health diagnosis reports, and predictive maintenance recommendations, and displays it visually through the front-end interface; it automatically triggers early warnings for identified high-risk equipment, guiding maintenance personnel to perform precise maintenance.

[0059] In this embodiment, step S1 is as follows:

[0060] S11. Test connection: Reliably connect the portable intelligent calibrator to the secondary circuit of the current transformer under test;

[0061] S12. Perform lightweight tests: Perform simplified test sequences, such as recording the excitation current at several key points, measuring the ratio difference (ε%) and phase difference (δ) under specific loads, and the winding resistance (R), etc.

[0062] S13. Data Encapsulation and Upload: The calibrator encrypts and packages the above test data and transmits it to the cloud-based digital twin platform via a wireless network, including 4G, 5G, or IoT networks.

[0063] In this embodiment, step S2 is as follows:

[0064] S21. Initial Model Construction: Based on the design drawings, material parameters, and rated parameters of the current transformer, the platform constructs a multi-physics initial model integrating electromagnetic fields, circuits, and thermodynamic effects.

[0065] S22. Parameter identification and optimization: An optimization algorithm (such as a genetic algorithm) is used to minimize the difference between the model output and the field measured data as the objective function, and the optimal internal parameter set of the model (such as winding resistance and core magnetization curve parameters) is solved in reverse.

[0066] The parameter identification process can be described as the following optimization problem:

[0067] Find P

[0068] to Minimize Loss = || F(M(P, X _input )) - S ||

[0069] Subject to P _min < P < P _max

[0070] Where P is the vector of internal parameters to be identified (such as winding resistance, leakage inductance, core saturation flux density, etc.), X _input F is the input stimulus of the model, S is the model output function, S is the field measurement data uploaded in step S1, and Loss is the loss function. By solving this optimization problem, the parameter set P that best matches the output of the digital twin with the physical entity's measured data can be obtained, thus obtaining the calibrated model M.

[0071] S23. Model Validation and Update: Use the calibrated model to simulate other known working conditions to verify its accuracy, and continuously update the model with each new data input.

[0072] In this embodiment, the parameter identification and optimization process adopts the CMA-ES algorithm based on the covariance matrix adaptive evolution strategy, which specifically includes the following steps:

[0073] S221. Initialization: Set the initial search distribution to a multivariate normal distribution N(m (o) , σ (o)2 C (o) ), where m (o) The mean of the initial solution for the parameter vector is usually randomly set within its boundary constraints, σ. (o) Let C be the initial step size. (o) The initial covariance matrix is ​​usually set as the identity matrix I;

[0074] S222, Sampling and Evaluation: In the g-th generation, from the current distribution N(m)... (g) , σ (g)2 C (g) ) generates λ offspring P representing candidate parameter vectors through independent sampling. k (g) For each offspring individual, k=1,...,λ, run the digital twin model and calculate its objective function value (loss value) Loss(P). k (g) );

[0075] S223. Selection and Reorganization: The λ offspring are sorted according to their loss values. The μ superior individuals with the smallest loss values ​​are selected. The mean m of the search distribution is updated using the weighted sum of the selected superior individuals to determine the central position of the next generation search distribution. The weight of the superior individuals indicates their importance. The weight of the superior individuals is determined according to their loss value ranking. The higher the ranking (the smaller the loss value), the higher the weight is given, so that it plays a greater role in updating the mean and guides the search direction to move towards the region with better performance.

[0076] S224, Covariance Matrix Adaptation: The covariance matrix C is updated using information from the current generation of selected individuals (evolutionary path) and historical information; this is equivalent to adaptively "learning" the local shape of the objective function in the parameter space (such as the major axis direction of the ellipsoid), so that the search distribution can more effectively move along the valley of error descent, greatly accelerating the convergence speed;

[0077] S225, Step Size Control: By comparing the actual length of the conjugate evolution path with its expected length under random selection, it is determined whether the step size σ needs to be increased or decreased. The path records historical information of the search direction. If the actual length exceeds the expected length, it indicates that the movement direction has been consistent and effective for several consecutive generations, so the step size is increased to accelerate convergence. Conversely, if the path length is short, it indicates that the search direction is unstable, and the step size needs to be decreased to perform a more refined local search.

[0078] S226. Iteration and Termination: Let g = g + 1, repeat steps S222 to S225 until the preset termination condition is met (such as reaching the maximum number of iterations, or the quality of the solution no longer significantly improves in consecutive iterations); finally, output the optimal mean m as the identified optimal parameter vector P.

[0079] In this embodiment, step S3 is as follows:

[0080] S31. Full-condition virtual verification: Drive the digital twin M to simulate complex operating conditions including overload, extreme temperature, and harmonics in the cloud, calculate its full-range error curve, including the ratio error curve and phase error curve of the current transformer, and generate a virtual verification report.

[0081] S32. In-depth health diagnosis: Compare the current calibration parameter P with the historical parameter P. _historical (e.g., winding resistance, core magnetization curve parameters), quantitatively analyze parameter changes (e.g., "winding resistance increased by 5%", "core permeability decreased by 2%", etc.), locate the root cause of defects, and generate a health diagnosis report;

[0082] S33. Performance Trend Prediction: Based on the parameter sequences from previous calibrations, a time series prediction model or machine learning algorithm is used to fit the performance degradation trajectory, predict the remaining lifetime (RUL), and recommend a personalized next calibration time (T). _next .

[0083] In this embodiment, step S33 is as follows:

[0084] S331. Define health indicators and failure criteria:

[0085] From the results of previous calibrations, a key parameter is selected as a health indicator. This key parameter (e.g., winding resistance value or ratio error under a specific load) changes monotonically over time and reflects the aging of the equipment.

[0086] Failure thresholds for health indicators are set according to technical specifications or expert experience; when the health indicators deteriorate beyond the threshold, the equipment is considered to have reached the end of its lifespan.

[0087] S332. Constructing a prediction model:

[0088] The historical verification time points t1, t2, ..., t n And the corresponding health indicator values ​​HI1, HI2, ..., HI n Used as training data; and used to train models based on time series prediction algorithms (such as linear regression, support vector regression, or neural networks) to learn the trend of health indicators deteriorating over time;

[0089] Step 333: Predict Remaining Lifetime (RUL):

[0090] Using a trained model, predictions are made for future points in time to obtain future trend curves of health indicators (including the best predicted value and possible fluctuation range, i.e., confidence interval).

[0091] The remaining useful life (RUL) is calculated as the length of time from the current point in time until the predicted health indicator curve first touches or exceeds the failure threshold in the future.

[0092] Step 334: Recommend the next verification time T_next based on the predicted remaining useful life (RUL):

[0093] To ensure safety and allow for a maintenance window, the next verification time, T_next, is determined based on a safety factor and is calculated as follows: (The calculation method is not repeated in the original text.)

[0094] T_next = current time + RUL × k

[0095] Where k is a safety factor less than 1 (e.g., 0.7, meaning that the test is performed at 70% of the remaining lifespan).

[0096] In this embodiment, step S4 is as follows:

[0097] S41. Automatic Report Generation: The system automatically integrates virtual verification reports, health diagnosis reports, and trend prediction results to generate a structured comprehensive equipment status report;

[0098] S42. Multi-terminal visual early warning: Through front-end interfaces including web interface and mobile app, the real-time status, risk level and predicted trend of the device are displayed in various forms including charts and dashboards, and early warning notifications are automatically issued for high-risk devices.

[0099] S43. Decision Support and Closed Loop: The reports and early warning information include specific maintenance recommendations (such as "tighten primary terminals" and "re-inspection recommended in 6 months"). Maintenance personnel perform on-site maintenance according to the maintenance recommendations, and the new data generated after maintenance serves as input for the next calibration, thus forming a closed-loop management for continuous optimization.

[0100] After the above four steps, a state-deep tracking and forward-looking verification system for electromagnetic current transformers based on digital twins is formed, thereby realizing full life-cycle management of equipment from deep perception of the current state to accurate early warning of future risks.

[0101] In summary, this invention innovatively utilizes digital twin technology, combining lightweight field testing with deep cloud simulation, to achieve a transformation of electromagnetic instrument transformers from "periodic verification" to "deep state tracking" and "proactive management," significantly improving the depth, efficiency, and intelligence level of electromagnetic instrument transformer verification.

[0102] This invention also proposes a state tracking and prospective verification system for electromagnetic instrument transformers based on digital twins. The system is implemented using any of the aforementioned methods for state tracking and prospective verification of electromagnetic instrument transformers based on digital twins, and includes a lightweight field acquisition module, a digital twin construction and calibration module, an intelligent analysis and prediction module, and a visualization application module.

[0103] The lightweight field acquisition module is responsible for acquiring lightweight status data of the current transformer at the field, such as excitation characteristics and basic errors, and uploading the encrypted data to the cloud platform.

[0104] The digital twin construction and calibration module receives this data and, based on a multiphysics model and parameter identification algorithm, constructs and dynamically updates a high-fidelity virtual model consistent with the physical entity. The parameter identification algorithm continuously adjusts the internal parameters of the digital model to ensure that the output of the virtual model accurately matches the on-site measured data, thereby completing the dynamic calibration of the digital twin and making the virtual model a precise mirror image of the physical entity.

[0105] The intelligent analysis and prediction module uses the calibrated virtual model to perform full-condition simulation and trend prediction in virtual space;

[0106] The visualization application module will display the generated comprehensive reports and early warning information on multiple platforms and provide specific decision support for operation and maintenance personnel.

[0107] In summary, this invention introduces an innovative paradigm for instrument transformer verification. This system achieves digitalization, intelligence, and forward-looking verification by constructing a digital twin synchronized with the physical entity. The parameter identification algorithm plays a crucial role in the calibration phase, solving optimization problems to ensure the virtual model accurately maps the real state of the physical entity, providing a unique and reliable data source for subsequent in-depth diagnostics and trend prediction. Furthermore, the system combines machine learning and big data technologies to predict performance degradation and generate precise maintenance strategies, ultimately realizing a strategic shift from "periodic maintenance" to "condition-based maintenance," demonstrating the enormous potential of digital twin technology in the field of intelligent operation and maintenance of power equipment.

[0108] The above are preferred embodiments of the present invention. Any changes made to the technical solution of the present invention that do not exceed the scope of the technical solution of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for state tracking and prospective verification of electromagnetic current transformers based on digital twins, characterized in that, Specifically, the following steps are included: S1. Lightweight Data Acquisition on Site: Key status data of the transformer under test are collected using a portable intelligent calibrator, including excitation characteristics, basic error, and winding DC resistance. The encrypted data is then encrypted, packaged, and uploaded to the digital twin platform. S2. Digital Twin Construction and Calibration: The digital twin platform constructs an initial multiphysics model based on the design parameters and physical laws of the transformer; and uses a parameter identification algorithm to fit the field data collected in step S1 with the model output, dynamically adjusting the internal parameters of the model so that the digital twin accurately reflects the current real state of the physical transformer. S3. Intelligent Analysis and Prediction: Using a calibrated high-fidelity digital twin, various extreme operating conditions and load conditions are simulated in virtual space to calculate the full-range error characteristics of the instrument transformer and generate a virtual verification report. Based on the current calibration data and historical calibration data, the root cause of the defect is located and a health diagnosis report is generated. Based on historical calibration data analysis, performance degradation trends are analyzed to predict the remaining lifespan and failure risk of the instrument transformer; S4. Visual Application and Decision Making: The platform automatically generates a comprehensive report that includes calibration reports, health diagnosis reports, and predictive maintenance recommendations, and displays it visually through the front-end interface; it automatically triggers early warnings for identified high-risk equipment, guiding maintenance personnel to perform precise maintenance.

2. The method for state tracking and prospective verification of electromagnetic current transformers based on digital twins according to claim 1, characterized in that, Encrypted data is transmitted to a cloud-based digital twin platform via a wireless network, including 4G, 5G, or IoT networks.

3. The method for state tracking and prospective verification of electromagnetic current transformers based on digital twins according to claim 1, characterized in that, Step S2 is as follows: S21. Initial Model Construction: Based on the design drawings, material parameters, and rated parameters of the current transformer, the platform constructs a multi-physics initial model integrating electromagnetic fields, circuits, and thermodynamic effects. S22. Parameter identification and optimization: An optimization algorithm is used to solve for the optimal internal parameter set of the model by minimizing the difference between the model output and the field measured data as the objective function.

4. The method for state tracking and prospective verification of electromagnetic current transformers based on digital twins according to claim 3, characterized in that, The parameter identification and optimization process uses the CMA-ES algorithm, which includes the following steps: S221. Initialization: Set the initial search distribution to a multivariate normal distribution N(m (o) , σ (o)2 C (o) ), where m (o) Let σ be the mean of the initial solution for the parameter vector. (o) Let C be the initial step size. (o) This is the initial covariance matrix; S222, Sampling and Evaluation: In the g-th generation, from the current distribution N(m)... (g) , σ (g)2 C (g) ) generates λ offspring P representing candidate parameter vectors through independent sampling. k (g) For each offspring individual, k=1,...,λ, run the digital twin model and calculate its objective function value Loss(P). k (g) ); S223. Selection and Reorganization: Rank the λ offspring according to their loss values, select the μ superior individuals with the smallest loss values, and use the weighted sum of the selected superior individuals to update the mean m of the search distribution, thus determining the central position of the next generation search distribution; the weight of the superior individuals is determined according to the loss value ranking, and the individual with the smaller the loss value is given a higher weight. S224, Covariance Matrix Adaptation: Update the covariance matrix C using information from the current generation of selected individuals and historical information; S225, Step Size Control: By comparing the actual length of the conjugate evolution path with its expected length under random selection, it is determined whether the step size σ needs to be increased or decreased. If the actual length of the conjugate evolution path exceeds the expected length, the step size is increased to accelerate convergence; if the actual length of the conjugate evolution path is less than the expected length, the step size is decreased to perform a more refined local search. S226. Iteration and Termination: Let g = g + 1, repeat steps S222 to S225 until the preset termination condition is met; finally, output the optimal mean m as the identified optimal parameter vector P.

5. The method for state tracking and prospective verification of electromagnetic current transformers based on digital twins according to claim 1, characterized in that, Step S3 is as follows: S31. Full-condition virtual verification: Drive the digital twin M to simulate complex operating conditions including overload, extreme temperature, and harmonics in the cloud, calculate its full-range error curve, including the ratio error curve and phase error curve of the current transformer, and generate a virtual verification report. S32. In-depth health diagnosis: Compare the current calibration parameter P with the historical parameter P. _historical Quantitative analysis of parameter changes, pinpointing the root cause of defects, and generating a health diagnosis report; S33. Performance Trend Prediction: Based on the parameter sequences from previous calibrations, a time series prediction model or machine learning algorithm is used to fit the performance degradation trajectory, predict the remaining lifetime (RUL), and recommend a personalized next calibration time (T). _next .

6. The method for state tracking and prospective verification of electromagnetic current transformers based on digital twins according to claim 1, characterized in that, Step S33 is as follows: S331. Define health indicators and failure criteria: From the results of previous calibrations, a key parameter is selected as a health indicator. This key parameter changes monotonically over time and reflects the aging of the equipment. Set failure thresholds for health indicators; S332. Constructing a prediction model: The historical verification time points t1, t2, ..., t n And the corresponding health indicator values ​​HI1, HI2, ..., HI n As training data; and to train a model based on a time series prediction algorithm to learn the trend of health indicators deteriorating over time; Step 333: Predict Remaining Lifetime (RUL): Using the trained model, we can predict future time points and obtain the future trend curves of health indicators. The remaining useful life (RUL) is calculated as the length of time from the current point in time until the predicted health indicator curve first touches or exceeds the failure threshold in the future. Step 334: Recommend the next verification time T_next based on the predicted remaining useful life (RUL).

7. The method for state tracking and prospective verification of electromagnetic current transformers based on digital twins according to claim 6, characterized in that, The time series prediction algorithm includes linear regression, support vector regression, or neural networks.

8. The method for state tracking and prospective verification of electromagnetic current transformers based on digital twins according to claim 6, characterized in that, The recommended next verification time T_next is based on a safety factor, calculated as follows: T_next = current time + RUL × k Where k is a safety factor less than 1.

9. The method for state tracking and prospective verification of electromagnetic current transformers based on digital twins according to claim 1, characterized in that, Step S4 is as follows: S41. Automatic Report Generation: The system automatically integrates virtual verification reports, health diagnosis reports, and trend prediction results to generate a structured comprehensive equipment status report; S42. Multi-terminal visual early warning: Through front-end interfaces including web interface and mobile app, the real-time status, risk level and predicted trend of the device are displayed in various forms including charts and dashboards, and early warning notifications are automatically issued for high-risk devices. S43. Decision Support and Closed Loop: The reports and early warning information contain specific maintenance recommendations. Maintenance personnel perform on-site maintenance based on the maintenance recommendations. The new data generated after maintenance serves as the input for the next calibration, thus forming a closed-loop management system for continuous optimization.

10. A state tracking and prospective verification system for electromagnetic current transformers based on digital twins, characterized in that, The system is implemented using the electromagnetic current transformer state tracking and prospective verification method based on digital twins as described in any one of claims 1-9, and includes a lightweight on-site acquisition module, a digital twin construction and calibration module, an intelligent analysis and prediction module, and a visualization application module: The lightweight field acquisition module is responsible for acquiring lightweight status data of the current transformer at the field. The digital twin construction and calibration module is based on a multiphysics model and parameter identification algorithm to construct and dynamically update a high-fidelity virtual model that is consistent with the physical entity. The intelligent analysis and prediction module uses the calibrated virtual model to perform full-condition simulation and trend prediction. The visualization application module generates diagnostic reports and early warning information to provide decision support for users.

Images