Fault prediction method and system for power sensing equipment, and storage medium

Abstract

The invention relates to the technical field of power sensing equipment fault prediction, in particular to a power sensing equipment fault prediction method and system and a storage medium. The method comprises the following steps of performing geometric model construction on power sensing equipment to obtain grid division model data; carrying out material attribute endowing based on association of microscopic characteristics and macroscopic attributes on the grid division model data to obtain material attribute grid model data; performing model parameter calibration on the material attribute grid model data to obtain a calibrated equipment twinborn body; carrying out environmental parameter loading on the calibrated twin of the equipment, and carrying out multi-physical field coupling simulation in the equipment to obtain stress field distribution data in the equipment; key point identification is carried out on the stress field distribution data in the equipment, and key point positions and stress data are obtained. The reliability of the power sensing equipment is improved through the power sensing equipment fault prediction technology, and the operation and maintenance cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of power sensing device fault prediction, and in particular to a power sensing device fault prediction method, system and storage medium. Background Art

[0002] Power sensing devices, such as sensors, transformer oil chromatographs, and online monitoring devices, serve as the "eyes" and "nerve endings" of modern smart grids. They collect operational status data from power equipment, providing critical information for condition assessment, fault diagnosis, and predictive maintenance. Effectively leveraging this sensing data for fault prediction can significantly improve the reliability, safety, and economic efficiency of power systems, and is a key technology for achieving intelligent operation and maintenance. Early approaches to fault prediction for power sensing devices primarily employed experience-based maintenance strategies, such as scheduled maintenance and post-event maintenance. This has evolved to condition-based maintenance strategies, such as vibration analysis, oil quality analysis, and partial discharge detection.

[0003] However, existing fault prediction technologies still suffer from some key drawbacks. For example, existing methods often assume that devices operate in constant or ideal conditions, ignoring the impact of dynamic environmental factors on device stress and performance degradation. Furthermore, insufficient characterization of material degradation mechanisms makes detailed predictions difficult. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, system and storage medium for predicting faults of power sensing equipment to solve at least one of the above technical problems.

[0005] To achieve the above objectives, a method for predicting power sensing device faults includes the following steps:

[0006] Step S1: construct a geometric model of the power sensing device to obtain mesh model data; assign material properties based on the association between microscopic characteristics and macroscopic properties to the mesh model data to obtain material property mesh model data; calibrate model parameters of the material property mesh model data to obtain a calibrated device twin;

[0007] Step S2: Loading environmental parameters to the calibrated device twin and performing multi-physics field coupling simulation inside the device to obtain stress field distribution data inside the device; identifying key points in the stress field distribution data inside the device to obtain key point positions and stress data;

[0008] Step S3: Determine degradation model parameters based on key point positions and stress data to obtain degradation model parameters; perform macroscopic performance degradation simulation based on microscopic crack propagation on key point positions and stress data based on the degradation model and random factor input data to obtain macroscopic performance degradation data; construct a performance degradation trajectory for the key point performance degradation data to obtain a performance degradation trajectory;

[0009] Step S4: fitting the failure time distribution based on the failure threshold according to the performance degradation trajectory to obtain the failure time probability density function parameters; predicting the failure probability of the power sensing device according to the failure time probability density function parameters to obtain the failure probability curve.

[0010] The present invention uses CAD modeling, meshing, micro-macro property association and multi-physics coupling modeling, and combines real-time sensor data for model calibration. Step S1 constructs a high-fidelity digital twin of the power sensing device. The digital twin not only accurately reflects the geometric structure and material properties of the device, but also takes into account the influence of the material's microscopic properties on the macroscopic performance, and can dynamically adapt to the actual operating state of the device through real-time data calibration. Compared with traditional fault prediction methods based on experience or simplified models, the method based on digital twins can more accurately and comprehensively simulate the operating behavior of the device, providing a solid foundation for subsequent stress analysis, degradation simulation and fault prediction, and significantly improving the accuracy and reliability of the prediction. By loading the real-time monitored environmental parameters (temperature, humidity, electromagnetic interference, etc.) onto the calibrated device twin and performing multi-physics coupling simulation, step S2 can accurately calculate the stress distribution of the device under the actual operating environment. Compared with traditional stress analysis methods based on constant environment assumptions or empirical formulas, this method can more realistically reflect the impact of environmental factors on device stress and improve the accuracy of stress analysis. In addition, by identifying the key points of stress concentration, the parts of the equipment that are most prone to failure can be accurately located, providing a focus for subsequent degradation simulation and fault prediction, and improving the efficiency and pertinence of the prediction. By selecting a degradation model that matches the key point material and failure mode and introducing random factors, step S3 can simulate the microscopic crack propagation process of key components of the equipment under actual operating stress, and then predict the changes in macroscopic performance parameters (such as dielectric strength and resistivity) over time. Compared with traditional degradation prediction methods based on empirical life models or statistical models, this degradation simulation method based on physical failure mechanisms can more accurately reflect the degradation process of the equipment and improve the reliability of the prediction. In addition, through Monte Carlo simulation and performance degradation trajectory construction, the uncertainty of the degradation process can be quantified, providing more comprehensive information for failure probability assessment. By comparing the performance degradation trajectory with the failure threshold and fitting the failure time distribution, step S4 can convert the performance degradation prediction of the equipment into a failure probability prediction. Compared with traditional failure prediction methods based on a single threshold, this probability distribution-based method can more comprehensively evaluate the reliability of the equipment and quantify the uncertainty of failure occurrence. Furthermore, Bayesian updating enables dynamic adjustment of failure probabilities using real-time monitoring data, improving the timeliness and accuracy of predictions. The resulting failure probability curve intuitively displays the probability of equipment failure at different time periods in the future, providing critical decision support for predictive maintenance.Therefore, the present invention provides a method for predicting faults of power sensing equipment. By constructing a digital twin model and combining multi-physics field coupling simulation, microscopic characteristic considerations, multi-scale degradation modeling, and Bayesian update technology, it solves the shortcomings of traditional methods in terms of the impact of environmental factors and dynamic stress, and the characterization of material degradation mechanisms, thereby improving the reliability of power sensing equipment and reducing operation and maintenance costs.

[0011] Preferably, step S1 includes the following steps:

[0012] Step S11: creating a geometric model of the power sensing device to obtain three-dimensional geometric model data;

[0013] Step S12: Meshing the three-dimensional geometric model data to obtain meshed model data;

[0014] Step S13: assigning material properties based on the association between microscopic properties and macroscopic properties to the mesh model data to obtain material property mesh model data;

[0015] Step S14: establishing a multi-physics field coupling model according to the material property grid model data to obtain a multi-physics field coupling model;

[0016] Step S15: collecting sensor data from the power sensing device in real time to obtain real-time sensor data;

[0017] Step S16: Input the real-time sensor data into the multi-physics field coupling model to calibrate the model parameters to obtain a calibrated device twin.

[0018] The present invention creates an accurate three-dimensional geometric model through CAD tools, which lays an accurate geometric foundation for subsequent finite element analysis, ensures that the simulation can fully consider the real impact of the device structure, avoids the error caused by geometric simplification, and improves the reliability of the simulation results. Grid division is performed using professional pre-processing tools to discretize the continuous geometric model, realizing the necessary conditions for finite element analysis. At the same time, by controlling the grid size and type, both calculation accuracy and calculation efficiency are taken into account, and the effectiveness of the analysis is ensured by grid quality inspection. By establishing a micro-macro property association model (such as the Hall-Petch formula), the microstructural characteristics of the material (such as grain size) are linked to the macro properties (such as conductivity), achieving a more refined and accurate description of material properties than the traditional method that only relies on macro properties, thereby improving the physical authenticity of the simulation model. By establishing a multi-physical field coupling model, multiple physical fields such as the electromagnetic field, temperature field, and their interactions involved in the operation of the power sensing device are incorporated into a unified simulation framework, achieving a more comprehensive and realistic simulation of the device operation status, and improving the reliability and predictive ability of the simulation results. By collecting real-time equipment operating data, we obtain information about the equipment's status under real-world operating conditions. This provides a critical, timely basis for model calibration and fault prediction, avoiding the potential biases that can arise from relying solely on historical or empirical data. Parameter calibration of the multi-physics coupling model using real-time sensor data effectively reduces model errors, ensuring a high degree of consistency between model output and actual measurement data. This creates a digital twin that accurately reflects the equipment's current status and provides a reliable virtual platform for subsequent analysis.

[0019] Preferably, step S13 includes the following steps:

[0020] Step S131: identifying the microscopic characteristics of key components of the mesh model data according to a preset material property database to obtain a list of materials for key components;

[0021] Step S132: searching for macro material property parameters based on the key component material list to obtain a macro property parameter table; acquiring micro material property data based on the key component material list to obtain micro property data;

[0022] Step S133: determining the association between micro-properties and macro-properties based on the micro-property data and the macro-property parameter table to obtain an associated property list;

[0023] Step S134: performing conductivity-grain size correlation for conductor materials, performing dielectric strength-lattice defect density correlation for insulating materials, and performing elastic modulus-grain size correlation for metal materials according to the correlation attribute list to obtain a constitutive model library;

[0024] Step S135: assigning and verifying model parameters of the constitutive model library to obtain a constitutive model parameter set;

[0025] Step S136: assigning mesh unit material properties to the meshing model data according to the constitutive model parameter set, the macroscopic property parameter table, and the microscopic property data to obtain material property mesh model data.

[0026] The present invention realizes the automatic identification of the key component materials of the power sensing device by matching the preset material property database with the grid model data, avoids the errors and omissions that may occur in manual identification, improves the efficiency and accuracy of the material property assignment, and lays the foundation for subsequent analysis. Based on the list of key component materials, the material property data of both macroscopic and microscopic aspects are systematically obtained from the material property database, and a comprehensive material property parameter table and microscopic property data table are constructed, which provide complete and reliable data support for the subsequent establishment of micro-macro correlation model and material property assignment. By analyzing the intrinsic connection between the microscopic properties of key component materials and their macroscopic properties, it is clarified which microscopic properties have a significant impact on which macroscopic properties, and a list of associated properties is established, which points the way for constructing an accurate micro-macro constitutive model and avoids blind modeling or omitting key factors. According to the key properties of different materials, specific micro-macro constitutive models (such as Hall-Petch formula, empirical formula, etc.) are selected and established, and the microstructural characteristics of the material are quantitatively linked to its macroscopic properties to form a constitutive model library, which provides a key mathematical tool for realizing the refined simulation of material properties. By consulting literature, experimental data, or databases, specific values are assigned to the parameters in the constitutive model. Comparison and verification with known experimental data or publicly available data ensure the accuracy and reliability of the model, forming a constitutive model parameter set that can be directly used in simulation. By combining the verified constitutive model parameters, macroscopic property parameters, and microscopic characteristic data with the mesh model, each mesh element is assigned accurate material properties that reflect the microstructural characteristics. This results in material property mesh model data containing rich material information, providing high-quality input for subsequent multi-physics field coupling simulations.

[0027] Preferably, step S2 includes the following steps:

[0028] Step S21: Acquire environmental parameter data; apply the environmental parameter data as boundary conditions to the calibrated device twin to obtain the device twin after loading the environmental parameters;

[0029] Step S22: performing multi-physics field coupling simulation on the device twin after loading the environmental parameters to obtain simulation result data of each physical field;

[0030] Step S23: performing stress extraction and post-processing on the simulation result data of each physical field to obtain stress field distribution data inside the device;

[0031] Step S24: performing key point identification on the stress field distribution data inside the device to obtain key point positions and stress data.

[0032] The present invention monitors or obtains environmental parameter data (temperature, humidity, electromagnetic interference, etc.) in real time and accurately applies it as boundary conditions to the calibrated device twin model, so that the simulation model can reflect the real environment in which the device is located, provides input that conforms to the actual working conditions for subsequent simulations, and improves the reliability of the simulation results. Through multi-physical field coupling simulation, the interaction between multiple physical fields such as electromagnetic field, temperature field, stress field (and possible flow field) is fully considered, and detailed distribution data of each physical field (temperature, electromagnetic field, stress, etc.) of the device after loading environmental parameters are obtained, providing comprehensive data support for in-depth analysis of the operating status and performance of the device. Key stress data is extracted from the simulation results, and necessary post-processing is performed (such as calculating principal stress, von Mises stress, stress concentration factor, fatigue damage, etc.), and detailed stress field distribution data inside the device that can be used for further analysis is obtained, providing key information for identifying weak links in the equipment and evaluating its structural reliability. By using the stress threshold method, stress gradient method, or combining structural characteristics, key points of stress concentration or high stress levels can be accurately identified from detailed stress field distribution data, and the positions and stress data of these key points can be extracted, providing key focus points for subsequent degradation simulation and fault prediction, thereby improving the efficiency and pertinence of the analysis.

[0033] Preferably, step S3 includes the following steps:

[0034] Step S31: selecting a degradation model and determining degradation model parameters based on key point positions and stress data and a preset material property database to obtain a degradation model and degradation model parameters;

[0035] Step S32: introducing random factors into degradation model parameters to obtain random factor input data;

[0036] Step S33: performing macro-performance degradation simulation based on micro-crack propagation on key point positions and stress data according to the degradation model and random factor input data to obtain macro-performance degradation data;

[0037] Step S34: constructing a performance degradation trajectory for the key point performance degradation data to obtain a performance degradation trajectory.

[0038] The present invention selects the most appropriate degradation model (such as Arrhenius, IPL, Coffin-Manson, etc.) for each key point by combining the key point stress data and the material property database, and accurately determines the model parameters. The advantage of doing so is that personalized modeling can be performed for different materials, different failure modes and different working conditions, avoiding the errors caused by using a general model, significantly improving the accuracy and reliability of subsequent simulation results, and laying a solid foundation for subsequent performance degradation analysis. By considering the randomness of aspects such as material properties, environmental conditions and manufacturing processes, and using probability distribution (such as normal distribution, Weibull distribution) and Monte Carlo sampling methods, these uncertainty factors are introduced into the degradation model parameters. The advantage of doing so is that it can more realistically reflect the dispersion of equipment performance degradation in the real world, avoid the problem of over-idealization of the prediction results of a single deterministic model, make the simulation results more robust, and provide more comprehensive information for subsequent reliability evaluation. By establishing a link between microscopic crack growth models (such as the Paris Law) and macroscopic performance degradation models (such as the relationship between dielectric strength and crack length), and combining multiscale simulation methods (micro, meso, and macro), this approach enables simulation of the entire process from microscopic defect evolution to macroscopic performance degradation. This approach offers the advantage of explaining the root causes of device performance degradation from a physical perspective, avoiding the limitations of empirical models and improving prediction accuracy and reliability, particularly for complex failure mechanisms. It also accurately outputs time-varying macroscopic performance parameters. By organizing and visualizing macroscopic performance degradation data obtained from multiple Monte Carlo simulations, a set of performance degradation trajectories is generated, which can be further statistically analyzed (such as calculating average trajectories and confidence intervals). This approach offers the advantage of visually demonstrating the device performance degradation process over time and quantifying the uncertainty in the degradation process. This provides key input for subsequent failure time prediction, failure probability calculation, and maintenance decision-making, making the overall prediction results more practical.

[0039] Preferably, step S331 includes the following steps:

[0040] Step S331: performing micro-scale crack initiation simulation based on key point positions and stress data, degradation model, and random factor input data to obtain micro-crack initiation data;

[0041] Step S332: constructing a mesoscale crack propagation model based on the microcrack initiation data and key point positions and stress data to obtain a mesoscale crack propagation model;

[0042] Step S333: applying a mesoscale stress field to the mesoscale crack growth model to obtain a mesoscale stress field;

[0043] Step S334: performing crack growth simulation according to the mesoscopic stress field, and recording the crack growth trajectory and rate to obtain mesoscopic crack growth data;

[0044] Step S335: performing microcrack density statistics in the key point area based on the mesoscopic crack growth data, analyzing the time variation pattern, and obtaining microcrack density data;

[0045] Step S336: selecting a macro insulation performance parameter degradation model based on the microcrack density data, and calculating the time-varying macro insulation performance parameters to obtain macro performance degradation data.

[0046] The present invention simulates the initiation process of micro cracks by establishing a model containing defects at the micro scale (such as molecular dynamics or micro finite element) and applying stress / strain history at key points, combining degradation models and random factors. The advantage of doing so is that it can reveal the root cause of crack generation (such as lattice defects, stress concentration, etc.) based on the microstructure of the material, and link the macro stress with the evolution of micro defects, providing accurate initial conditions and physical basis for subsequent meso and macro scale simulations, and improving the accuracy of overall prediction. Based on the micro crack initiation data, a model containing initial cracks is established at the meso scale (such as millimeter-level finite element), and a suitable crack extension model (such as ParisLaw or cohesion model) is selected. The advantage of doing so is that multiple small cracks generated at the micro scale can be introduced into the meso scale model as initial conditions, simulating the expansion behavior of cracks more realistically, while avoiding the computational difficulties caused by directly processing a large number of micro cracks in the macro model, improving computational efficiency, and providing the necessary model for subsequent crack extension simulation. Based on the key point locations and stress data, the mesoscale stress field was recalculated to account for the presence of cracks and applied as boundary conditions to the mesoscale model. This approach accurately reflects the stress concentration effect at the crack tip, a key driver of crack growth, avoiding errors introduced by using a stress field that does not account for cracks and ensuring the accuracy of subsequent crack growth simulations. Based on the mesoscale crack growth model and stress field, the crack propagation process was simulated using methods such as the extended finite element method (XFEM) or the virtual crack closure technique (VCCT), and the position and growth rate of the crack tip were recorded over time. This approach provides detailed information on crack propagation, including the crack path, length, and velocity. This information is crucial for understanding failure mechanisms and predicting remaining life, and provides a data foundation for subsequent microcrack density statistics. By counting the number of cracks and calculating the microcrack density within the key point region, the temporal variation of the microcrack density was analyzed. The advantage of doing this is that the crack extension information at the mesoscopic scale can be converted into a macroscopically usable parameter - microcrack density. This parameter can reflect the overall degree of damage to the material and provide a key input for the subsequent macroscopic performance degradation model. Based on the microcrack density data, the macroscopic insulation performance parameter degradation model (such as the resistivity, dielectric strength, dielectric loss tangent, and partial discharge inception voltage models) is selected and applied, and the changes of these parameters over time are calculated. The advantage of doing this is that the micro / mesoscopic scale crack evolution information is ultimately linked to the macroscopic measurable performance parameters, realizing a complete prediction chain from microscopic damage to macroscopic failure, and obtaining time-varying macroscopic performance degradation data that can be directly used for equipment status assessment and life prediction. This step also provides degradation data for a variety of macroscopic performance parameters, making the prediction results more comprehensive.

[0047] Preferably, step S336 is specifically as follows:

[0048] Determine a coupled degradation model based on performance parameters according to microcrack density data to obtain a coupled degradation model formula, wherein the coupled degradation model formula includes a resistivity degradation model formula, a dielectric strength degradation model formula, a dielectric loss tangent degradation model formula, and a partial discharge inception voltage degradation model formula;

[0049] Acquire simulation time step sequence data; acquire temperature values and electric field strength values corresponding to the time step according to the simulation time step sequence data to obtain temperature data and electric field strength data;

[0050] The time-varying resistivity is calculated based on the microcrack density data, resistivity degradation model formula, temperature data and electric field intensity data to obtain the time-varying resistivity data;

[0051] The time-varying dielectric strength data is obtained by calculating the time-varying dielectric strength based on the microcrack density data, the dielectric strength degradation model formula, the temperature data and the electric field strength data;

[0052] The time-varying dielectric loss tangent is calculated based on the microcrack density data, dielectric loss tangent degradation model formula, temperature data, and electric field strength data to obtain the time-varying dielectric loss data;

[0053] The time-varying partial discharge inception voltage is calculated based on the microcrack density data, the dielectric strength degradation model formula, the temperature data, and the electric field strength data to obtain the time-varying partial discharge voltage data.

[0054] The macro performance degradation data is integrated by integrating the time-varying resistivity data, time-varying dielectric strength data, time-varying dielectric loss data and time-varying partial discharge voltage data to obtain the macro performance degradation data.

[0055] The present invention establishes a coupled degradation model formula between microcrack density and other key macroscopic performance parameters (resistivity, dielectric strength, dielectric loss tangent, partial discharge inception voltage). The advantage of this is that it can quantitatively describe how microscopic damage (cracks) affects macroscopic performance, and takes into account the combined effects of multiple physical field factors such as temperature and electric field intensity, rather than just the influence of a single factor. This coupled model is more accurate than the degradation model that considers each parameter separately, better reflects the actual situation, and provides a reliable mathematical basis for subsequent time-varying performance parameter calculations. The time step sequence and the temperature and electric field intensity data at the key points corresponding to each time step are extracted from the previous multi-physics field simulation results. The advantage of this is that it provides the necessary time reference and environmental condition input for subsequent time-varying performance parameter calculations. Because the degradation model is usually a function of time, temperature and electric field intensity, the accurate acquisition of this data is a prerequisite for ensuring the accuracy of subsequent calculations. This also reflects the advantage of multi-physics field coupled simulation, integrating information from different physical fields. The microcrack density, temperature, electric field intensity and resistivity degradation model formula are combined to calculate the change of resistivity over time. The benefit of this approach is that a specific, time-varying macroscopic performance parameter (resistivity) is obtained, which can be directly used to assess the insulation condition of the equipment. This parameter is calculated based on microscopic damage evolution and multi-physics field coupling effects, thus providing high reliability. The temporal variation of dielectric strength was calculated. Dielectric strength is a key performance indicator of insulation materials, and obtaining its time-varying data is crucial for assessing the insulation strength and remaining life of the equipment. The temporal variation of the dielectric loss tangent was calculated. The dielectric loss tangent reflects the energy loss of the insulation material under the action of an electric field, and its changes can indicate the degree of insulation aging. By employing a partial discharge inception voltage degradation model, the occurrence of partial discharge (PD) can be predicted. PD is a key precursor to insulation failure. A decrease in its inception voltage indicates that the insulation is more susceptible to PD, which accelerates the failure process. The multiple time-varying macroscopic performance parameters (resistivity, dielectric strength, dielectric loss, and PD voltage) calculated previously are integrated to form a complete macroscopic performance degradation dataset. This approach provides comprehensive information on the insulation condition of the equipment, allowing for a multi-faceted assessment of the equipment's health and remaining life. This avoids the potential bias of single-parameter evaluation and improves the reliability and accuracy of the assessment.

[0056] Preferably, step S4 includes the following steps:

[0057] Step S41: Obtaining equipment technical specifications and historical operating data; determining a failure threshold based on the equipment technical specifications, historical operating data, and performance degradation trajectory to obtain a failure threshold;

[0058] Step S42: fitting the failure time distribution according to the failure threshold and the performance degradation trajectory to obtain the failure time probability density function parameters;

[0059] Step S43: Calculate the failure probability of the failure time probability density function parameters to obtain failure probability data for different time periods;

[0060] Step S44: acquiring real-time monitoring data; performing Bayesian update on the failure probability data of different time periods using the real-time monitoring data to obtain updated failure probability data;

[0061] Step S45: Draw a failure probability curve for the updated failure probability data to obtain a failure probability curve.

[0062] The present invention determines the failure thresholds of key performance parameters by comprehensively analyzing the equipment's technical specifications (such as factory specifications and industry standards), historical operating data (such as online monitoring, offline testing, and fault records), and previously obtained performance degradation trajectories. This approach has the advantage of combining theoretical standards, practical experience, and simulation results, avoiding the limitations of single-source information. This makes the determined failure thresholds closer to reality and more representative, providing a reliable basis for subsequent failure time prediction and failure probability calculation. Based on the failure thresholds and performance degradation trajectories, the failure time of each simulation is determined, and through statistical analysis, these failure time data are fitted into a probability distribution function (such as a Weibull distribution, a lognormal distribution, etc.), and the parameters of the distribution function are obtained. This approach has the advantage of converting discrete failure time data into a continuous probability model, facilitating subsequent failure probability calculations and reflecting the randomness and uncertainty of failure times, making the prediction results more statistically significant. Using the failure time probability density function, the probability of equipment failure in different future time periods is calculated. This approach provides quantified, time-varying failure probability data. This data can be directly used to assess equipment reliability and provide a critical basis for decision-making in maintenance plans (such as preventive maintenance and condition-based maintenance), eliminating the subjectivity and blindness of empirical decisions. By acquiring real-time monitoring data from the equipment (such as temperature, current, and partial discharge), and using the Bayesian updating method, this real-time information is combined with the previous failure probability prediction (prior probability) to obtain an updated failure probability (posterior probability). This approach allows for dynamic adjustment of the failure probability prediction based on the actual operating status of the equipment, making the prediction more accurate and timely. This is crucial for timely detecting potential equipment failures and avoiding unplanned downtime. Plotting the Bayesian-updated failure probability data as a time-varying curve provides a visual representation of the equipment's failure risk over time, enabling maintenance personnel to clearly understand the equipment's health status. This provides a crucial reference for taking timely measures to prevent accidents, and improves the scientific and effective nature of equipment management.

[0063] Preferably, the present invention further provides a power-aware device fault prediction system for executing the power-aware device fault prediction method described above, the power-aware device fault prediction system comprising:

[0064] The twin modeling module is used to construct a geometric model of the power sensing device to obtain mesh model data; assign material properties to the mesh model data based on the association between microscopic characteristics and macroscopic properties to obtain material property mesh model data; and calibrate model parameters of the material property mesh model data to obtain a calibrated device twin;

[0065] The stress mapping module is used to load environmental parameters on the calibrated device twin and perform multi-physics field coupling simulation inside the device to obtain stress field distribution data inside the device; it also identifies key points in the stress field distribution data inside the device to obtain key point locations and stress data;

[0066] The degradation simulation module is used to determine the degradation model parameters based on the key point positions and stress data to obtain the degradation model parameters; perform macro-performance degradation simulation based on micro-crack propagation on the key point positions and stress data according to the degradation model and random factor input data to obtain macro-performance degradation data; and construct a performance degradation trajectory for the key point performance degradation data to obtain the performance degradation trajectory;

[0067] The failure probability assessment module is used to fit the failure time distribution based on the failure threshold according to the performance degradation trajectory to obtain the failure time probability density function parameters; and to predict the failure probability of the power sensing device based on the failure time probability density function parameters to obtain the failure probability curve.

[0068] Preferably, when the computer program is executed, the power-aware device fault prediction method as described above is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 A flowchart illustrating the steps of a method for predicting faults of power-aware devices.

[0070] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0071] 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.

[0072] 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.

[0073] 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.

[0074] To achieve this, please refer to Figure 1 , a method for predicting faults of power-aware equipment, comprising the following steps:

[0075] Step S1: construct a geometric model of the power sensing device to obtain mesh model data; assign material properties based on the association between microscopic characteristics and macroscopic properties to the mesh model data to obtain material property mesh model data; calibrate model parameters of the material property mesh model data to obtain a calibrated device twin;

[0076] Step S2: Loading environmental parameters to the calibrated device twin and performing multi-physics field coupling simulation inside the device to obtain stress field distribution data inside the device; identifying key points in the stress field distribution data inside the device to obtain key point positions and stress data;

[0077] Step S3: Determine degradation model parameters based on key point positions and stress data to obtain degradation model parameters; perform macroscopic performance degradation simulation based on microscopic crack propagation on key point positions and stress data based on the degradation model and random factor input data to obtain macroscopic performance degradation data; construct a performance degradation trajectory for the key point performance degradation data to obtain a performance degradation trajectory;

[0078] Step S4: fitting the failure time distribution based on the failure threshold according to the performance degradation trajectory to obtain the failure time probability density function parameters; predicting the failure probability of the power sensing device according to the failure time probability density function parameters to obtain the failure probability curve.

[0079] In an embodiment of the present invention, the power sensing device fault prediction method includes the following steps:

[0080] Step S1: construct a geometric model of the power sensing device to obtain mesh model data; assign material properties based on the association between microscopic characteristics and macroscopic properties to the mesh model data to obtain material property mesh model data; calibrate model parameters of the material property mesh model data to obtain a calibrated device twin;

[0081] In an embodiment of the present invention, first, a CAD tool such as AutoCAD is used to create an accurate three-dimensional geometric model of the device and its components (windings, insulation layers, cores, etc.) based on the design drawings of the power sensing device (such as a current transformer), and export it to a common format. Then, a pre-processing tool such as HyperMesh is used to perform geometric cleaning and meshing on the three-dimensional model, set different mesh sizes according to component characteristics and simulation accuracy requirements, generate a tetrahedral mesh, and export mesh data. Next, a database containing the macroscopic properties and microscopic properties of the material is constructed, and a micro-macroscopic correlation model is established based on the Hall-Petch formula, etc. In HyperMesh, a material is specified for each component of the mesh model, and the macroscopic property value is calculated based on the correlation model and the microscopic property data, and the property is assigned to the mesh unit to obtain a mesh model containing material properties. Finally, in simulation tools such as COMSOL, the material property mesh model is imported to establish a coupling model of the electromagnetic field and the heat transfer field. The model parameters (such as thermal conductivity and magnetic permeability) are calibrated using sensor data (current, temperature, etc.) in actual operation to obtain a calibrated device twin that can accurately reflect the device status.

[0082] Step S2: Loading environmental parameters to the calibrated device twin and performing multi-physics field coupling simulation inside the device to obtain stress field distribution data inside the device; identifying key points in the stress field distribution data inside the device to obtain key point positions and stress data;

[0083] In an embodiment of the present invention, the environmental parameter data (temperature, humidity, electromagnetic interference, etc.) monitored or obtained in real time are used as boundary conditions and loaded onto the calibrated device twin model obtained in step S1 (such as setting temperature boundaries and magnetic field boundaries in COMSOL). Then, multi-physics field coupling simulation (steady state or transient) is performed using simulation tools such as COMSOL to obtain simulation result data such as the temperature field and electromagnetic field inside the device. Then, the stress data of each component of the device (including stress components, principal stresses, von Mises stresses, etc.) are extracted from the simulation results, and post-processing is performed (such as calculating stress concentration coefficients and fatigue damage). Finally, by setting stress thresholds, finding the area with the maximum stress gradient, or combining structural characteristics, the key points of stress concentration inside the device are identified, and the positions and stress data of these key points are recorded.

[0084] Step S3: Determine degradation model parameters based on key point positions and stress data to obtain degradation model parameters; perform macroscopic performance degradation simulation based on microscopic crack propagation on key point positions and stress data based on the degradation model and random factor input data to obtain macroscopic performance degradation data; construct a performance degradation trajectory for the key point performance degradation data to obtain a performance degradation trajectory;

[0085] In the embodiment of the present invention, according to the key point position and the stress data obtained in step S2, in conjunction with the material property database, determine the material and its main failure mode (such as thermal aging, electrical aging, fatigue, etc.) to which each key point belongs, and select corresponding degradation model (such as Arrhenius model, IPL model, Coffin-Manson model, etc.), and consult literature or experiment to determine model parameters. Then, considering the randomness of material properties, environmental conditions and manufacturing process, random factors (such as normal distribution) are introduced for model parameters, and Monte Carlo method is used for sampling. Then, based on micro crack propagation model (such as Paris formula) and macro performance degradation model, degradation simulation is carried out to key point, the initiation and expansion of simulated cracks are simulated, and the variation of macro performance parameters (such as dielectric strength, resistivity) over time is calculated. Finally, the variation data of macro performance parameters obtained by multiple Monte Carlo simulations over time are plotted into performance degradation trajectory.

[0086] Step S4: fitting the failure time distribution based on the failure threshold according to the performance degradation trajectory to obtain the failure time probability density function parameters; predicting the failure probability of the power sensing device according to the failure time probability density function parameters to obtain a failure probability curve;

[0087] In an embodiment of the present invention, first, according to the technical specifications (such as product manuals, industry standards) and historical operating data (such as online monitoring data, fault records) of the power sensing device, the failure threshold of key performance parameters (such as insulation resistance, dielectric strength, partial discharge level) is determined. Then, the performance degradation trajectory obtained in step S3 is compared with the failure threshold, the time when each trajectory reaches the failure threshold is extracted, and the probability distribution function (such as Weibull distribution, lognormal distribution) is fitted using these failure time data to obtain the failure time probability density function parameters. Then, by integrating the probability density function, the probability of failure of the device in different time periods is calculated. Finally, the likelihood function is constructed using real-time monitoring data (such as temperature, current, partial discharge amount), and combined with the Bayesian formula, the failure probability is updated, and a failure probability curve is drawn to reflect the change of the device failure probability over time.

[0088] Preferably, step S1 includes the following steps:

[0089] Step S11: creating a geometric model of the power sensing device to obtain three-dimensional geometric model data;

[0090] Step S12: Meshing the three-dimensional geometric model data to obtain meshed model data;

[0091] Step S13: assigning material properties based on the association between microscopic properties and macroscopic properties to the mesh model data to obtain material property mesh model data;

[0092] Step S14: establishing a multi-physics field coupling model according to the material property grid model data to obtain a multi-physics field coupling model;

[0093] Step S15: collecting sensor data from the power sensing device in real time to obtain real-time sensor data;

[0094] Step S16: Input the real-time sensor data into the multi-physics field coupling model to calibrate the model parameters to obtain a calibrated device twin.

[0095] In an embodiment of the present invention, a current transformer is taken as an example, and its core components include a primary winding (conductor), a secondary winding (conductor), an epoxy resin insulation layer, an iron core, and a casing. First, obtain the CAD design drawing of the current transformer. The drawing should contain the precise dimensions and assembly relationships of all components. Use a computer-aided design (CAD) tool, such as AutoCAD Mechanical 2023 version, to open and read the CAD drawing file (.dwg or .dxf format). In AutoCAD, build a three-dimensional solid model of each component one by one according to the drawing. The primary winding and secondary winding are created using the spiral tool, and the diameter and number of turns of the wire are set. The epoxy resin insulation layer is created using the stretch or rotate command according to the inner and outer diameters and height given in the drawing. The iron core adopts a silicon steel sheet laminated structure. First, create the two-dimensional outline of a single silicon steel sheet, and then generate a three-dimensional model using the array and stretch commands. The casing is created according to the drawing dimensions using commands such as stretch and shell extraction. After all components are created, use the "assemble" function in AutoCAD according to the assembly relationship to assemble the components into a complete three-dimensional model of the current transformer. Finally, export the model to a common format, such as a 3D geometric model data file in STEP (.stp) or IGES (.igs) format.

[0096] Use a finite element pre-processing tool, such as HyperMesh 2022, to import the STEP-format 3D geometry model data file generated in step S11. In HyperMesh, first perform geometry cleanup to check and repair any defects in the model, such as overlapping surfaces, gaps, and sharp corners. Then, select a tetrahedral meshing algorithm, as tetrahedral meshes are more adaptable to complex geometries. Set different mesh sizes based on the characteristics of the current transformer components and the accuracy requirements of the subsequent simulation. For example, for the primary and secondary windings, a smaller mesh size, such as 0.5 mm, is used because accurate current density distribution is required. For the epoxy insulation layer, a slightly larger mesh size, such as 1 mm, is used because the electric field strength and temperature distribution need to be calculated. For the core and casing, a larger mesh size, such as 2 mm, is used because their mechanical properties are of primary concern. In HyperMesh, use the "Tetramesh" function to perform meshing. After meshing, check the mesh quality to ensure that metrics such as the aspect ratio and Jacobian meet the requirements. Finally, export the mesh model to a common format, such as Abaqus's .inp format or ANSYS's .cdb format as a meshing model data file.

[0097] Build a material property database and establish a correlation model between microscopic characteristics and macroscopic properties.

[0098] 1. Material Property Database Construction: Collect and organize macroscopic property data for the materials used in each component of the current transformer, including electrical conductivity, thermal conductivity, relative dielectric constant, magnetic permeability, Young's modulus, Poisson's ratio, and so on. This data can be obtained from technical manuals provided by material suppliers, public material databases (such as MatWeb), or experimental measurements. Also, collect microscopic property data for each material, such as the grain size of conductors and the lattice defect density of insulating materials. This data can be obtained through experimental methods such as metallographic microscopy and X-ray diffraction analysis. Organize this data into a table to construct the material property database.

[0099] 2. Establishment of micro-macro correlation model: For conductors (such as copper), establish the correlation between conductivity and grain size. For example, use the modified form of the Hall-Petch formula: σ = σ0 + k * d^ (-1 / 2), where σ is conductivity, σ0 is the conductivity when the grain size is infinite, k is the Hall-Petch coefficient, and d is the average grain size. For insulating materials (such as epoxy resin), establish the correlation between dielectric strength and lattice defect density. For example, use the empirical formula: Eb = Eb0-a * N^b, where Eb is dielectric strength, Eb0 is the dielectric strength when there are no defects, N is the lattice defect density, and a and b are empirical coefficients. For iron cores (such as silicon steel sheets), establish the correlation between Young's modulus and grain size, and the modified form of the Hall-Petch formula can also be used.

[0100] 3. Material property assignment: In the finite element pre-processing tool (such as HyperMesh), read the meshing model data file generated in step S12. Assign the corresponding material name to each component (primary winding, secondary winding, epoxy resin insulation layer, iron core, shell). According to the data in the material property database and the established micro-macro correlation model, calculate the macro property value of each component under specific microscopic characteristics. For example, if the average grain size of the primary winding is 10μm, its conductivity is calculated according to the Hall-Petch formula. Assign these macro property values to the corresponding mesh cells. Finally, export the mesh model data file containing material property information, such as Abaqus's .inp format.

[0101] Use a finite element simulation tool, such as COMSOL Multiphysics version 6.0, to import the mesh model data file containing material properties generated in step S13. In COMSOL, select the AC / DC Module and the Heat Transfer Module. In the AC / DC Module, add the Current physics interface to simulate the electromagnetic field distribution in the current transformer. Set the input current for the primary winding and specify the load resistance for the secondary winding. In the Heat Transfer Module, add the Heat Transfer in Solids physics interface to simulate the temperature distribution in the current transformer. Set the ambient temperature and consider Joule heating (heat generated by current). Add the Electromagnetic Heat Source coupling interface under the Multiphysics node to achieve bidirectional coupling between the electromagnetic and temperature fields. This means that Joule heating generated by the electromagnetic field affects the temperature distribution, and temperature changes in turn affect the material's conductivity, which in turn affects the electromagnetic field distribution. In COMSOL, set the solver type to Steady State or Transient, depending on your specific analysis needs. Set solution parameters, such as solution accuracy and maximum number of iterations. Save the COMSOL model file (.mph), which is a multi-physics coupling model.

[0102] Install sensors on an actual current transformer. For example, install current sensors on the primary and secondary windings to measure current; install temperature sensors on the epoxy insulation and core to measure temperature; and install a vibration sensor on the casing to measure vibration. Use a data acquisition system, such as the NI CompactDAQ system, to connect these sensors. Configure the data acquisition system's sampling frequency and time. For example, set the current sensor sampling frequency to 1 kHz and the temperature sensor sampling frequency to 1 Hz, and collect data continuously for 24 hours. The data acquisition system transmits the collected sensor data to a computer in real time. On the computer, use data acquisition software, such as LabVIEW 2022, to receive and store this data. Save the data as a text file (.txt) or Excel file (.xlsx) to obtain real-time sensor data.

[0103] In COMSOL Multiphysics, open the multiphysics coupling model established in step S14. Import the real-time sensor data collected in step S15. Use COMSOL's "Parameter Estimation" function. Select the model parameters that need to be calibrated, such as the thermal conductivity of the epoxy insulation layer and the magnetic permeability of the iron core. Set the initial values and value ranges of these parameters. Use the measured values of current, temperature, etc. in the real-time sensor data as "experimental data." Use the calculated values of current, temperature, etc. at the corresponding positions calculated by the COMSOL model as "simulation data." Select an optimization algorithm, such as the Levenberg-Marquardt algorithm. Set optimization parameters, such as the maximum number of iterations and the convergence tolerance. Run the "Parameter Estimation" function, and COMSOL will automatically adjust the model parameters to minimize the error between the simulation data and the experimental data. When the error reaches the set requirement or the maximum number of iterations is reached, the optimization ends. At this point, the model parameters are the calibrated parameters. Save the COMSOL model file containing the calibrated parameters, which is the calibrated device twin.

[0104] Preferably, step S13 includes the following steps:

[0105] Step S131: identifying the microscopic characteristics of key components of the mesh model data according to a preset material property database to obtain a list of materials for key components;

[0106] Step S132: searching for macro material property parameters based on the key component material list to obtain a macro property parameter table; acquiring micro material property data based on the key component material list to obtain micro property data;

[0107] Step S133: determining the association between micro-properties and macro-properties based on the micro-property data and the macro-property parameter table to obtain an associated property list;

[0108] Step S134: performing conductivity-grain size correlation for conductor materials, performing dielectric strength-lattice defect density correlation for insulating materials, and performing elastic modulus-grain size correlation for metal materials according to the correlation attribute list to obtain a constitutive model library;

[0109] Step S135: assigning and verifying model parameters of the constitutive model library to obtain a constitutive model parameter set;

[0110] Step S136: assigning mesh unit material properties to the meshing model data according to the constitutive model parameter set, the macroscopic property parameter table, and the microscopic property data to obtain material property mesh model data.

[0111] In an embodiment of the present invention, a preset material property database is established to contain information on commonly used materials for power sensing devices. The database is stored in a tabular format, with each row representing a material and each column representing a property. Property columns include: material name, material category (conductor, insulator, magnetic material, etc.), main components, typical applications, macroscopic properties (such as electrical conductivity, thermal conductivity, dielectric constant, magnetic permeability, Young's modulus, Poisson's ratio, etc.), and microscopic properties (such as grain size range, lattice defect density range, etc.). For example, for a current transformer, the database should contain information on materials such as copper (conductor), epoxy resin (insulator), and silicon steel sheet (magnetic material). Data sources include technical manuals provided by material suppliers, public material databases (such as MatWeb), and published scientific papers. Key component identification: The mesh model data file generated in step S12 (e.g., in .inp format) is read. This file contains geometric information and component information for the mesh cells. By parsing this file, the names of the various components in the current transformer, such as "primary winding," "secondary winding," "epoxy resin insulation layer," "iron core," and "housing," are extracted. Bill of Materials Generation: Extracted component names are matched against "Material Name" or "Typical Application" in the material properties database. For example, "Primary Winding" and "Secondary Winding" are matched to "Copper," "Epoxy Insulation" is matched to "Epoxy Resin," and "Iron Core" is matched to "Silicon Steel Sheet." The matched material names are listed to form a bill of materials for key components. This list is presented in a table format with two columns: "Component Name" and "Corresponding Material."

[0112] Macro-property parameter search: Based on the key component materials list generated in step S131, the macro-property parameters of the corresponding materials are searched one by one in the pre-set material properties database. For example, for "copper" in the list, its electrical conductivity, thermal conductivity, density, specific heat capacity, and other parameter values are searched in the database. The found parameter values are recorded to form a macro-property parameter table. This table is presented in tabular format, with each row representing a component and each column representing a macro-property. Micro-property data acquisition: Also based on the key component materials list, the micro-property data of the corresponding materials is searched in the material properties database. For example, for "copper," its grain size range is searched; for "epoxy resin," its lattice defect density range is searched. It is important to note that micro-property data is typically a range, not a fixed value. This is because the microstructure of actual materials has a certain degree of randomness. The found micro-property data is recorded to form a micro-property data table. This table is presented in tabular format, with each row representing a component and each column representing a micro-property.

[0113] Correlation analysis: Based on material science theory and existing research results, the correlation between the microscopic properties of key component materials and their macroscopic properties is analyzed. For example, for conductors, their electrical conductivity is mainly affected by grain size; for insulators, their dielectric strength is mainly affected by lattice defect density; for magnetic materials, their magnetic permeability is affected by grain size and crystal orientation. Establishment of a list of associated properties: The correlation relationships obtained from the analysis are organized into a list of associated properties. The list is presented in a table format, with each row representing a component, listing the main microscopic properties that affect its key macroscopic properties. For example, the row corresponding to "primary winding (copper)" in the list lists "conductivity" and "grain size"; the row corresponding to "epoxy resin insulation layer" lists "dielectric strength" and "lattice defect density".

[0114] Conductivity-Grain Size Correlation (Conductors): A modified form of the Hall-Petch equation is used: σ = σ0 + k*d^(-1 / 2), where σ is the conductivity, σ0 is the conductivity at infinite grain size, k is the Hall-Petch coefficient, and d is the average grain size. σ0 and k can be determined by consulting the literature or experimentally. Dielectric Strength-Lattice Defect Density Correlation (Insulators): An empirical formula is used: Eb = Eb0 - a*N^b, where Eb is the dielectric strength, Eb0 is the dielectric strength at no defects, N is the lattice defect density, and a and b are empirical coefficients. Eb0, a, and b can be determined by consulting the literature or experimentally. Elastic Modulus-Grain Size Correlation (Metallic Materials): A modified form of the Hall-Petch equation is used: E = E0 + ke*d^(-1 / 2), where E is the elastic modulus, E0 is the elastic modulus at infinite grain size, ke is the Hall-Petch coefficient, and d is the average grain size. E0 and ke can be obtained by consulting literature or experimental determination. Building a constitutive model library: Store the above formulas in text form and establish a correspondence with the material names in the bill of materials for key components. This creates a constitutive model library containing multiple constitutive model formulas.

[0115] Parameter assignment: For each formula in the constitutive model library, determine the values of each parameter in the formula based on the data in the material properties database, published literature, or experimental measurement results. For example, for the Hall-Petch formula for copper, determine the values of σ0 and k. Model verification: Use existing experimental data or publicly available material data to verify the assigned constitutive model. For example, for the Hall-Petch formula for copper, input different grain sizes d, calculate the conductivity σ, and compare the calculated results with the experimental measurements. If the error is large, it is necessary to adjust the model parameters or consider a more complex model. Repeated adjustments and verifications are performed until the model calculation results are in good agreement with the experimental data. Constitutive model parameter set generation: The verified constitutive model and its determined parameter values are organized into a table to form a constitutive model parameter set. The table contains information such as the material name, constitutive model formula, parameter name, and parameter value.

[0116] Read the meshing model data file (such as .inp format) generated in step S12. This file contains the geometric information and component information of the mesh unit. Traverse each component and determine its corresponding material based on the key component material list generated in step S131. For each component, obtain its micro-characteristic data (such as grain size, lattice defect density, etc.) from the micro-characteristic data table. Note that it is necessary to sample the micro-characteristic data according to the actual situation here because the micro-characteristic data is usually a range. According to the constitutive model parameter set, select the corresponding constitutive model formula, and substitute the micro-characteristic data into the formula to calculate the corresponding macro-property values (such as conductivity, dielectric strength, elastic modulus, etc.). The calculated macro-property values, as well as the macro-property values that do not need to be calculated in the macro-property parameter table (such as density, specific heat capacity, etc.), are assigned to each mesh unit of the component. After the mesh unit properties of all components are assigned, save them as a new mesh model data file (such as .inp format), which is the material property mesh model data. This file contains the geometric information, component information and material property information of each mesh unit.

[0117] Preferably, step S2 includes the following steps:

[0118] Step S21: Acquire environmental parameter data; apply the environmental parameter data as boundary conditions to the calibrated device twin to obtain the device twin after loading the environmental parameters;

[0119] Step S22: performing multi-physics field coupling simulation on the device twin after loading the environmental parameters to obtain simulation result data of each physical field;

[0120] Step S23: performing stress extraction and post-processing on the simulation result data of each physical field to obtain stress field distribution data inside the device;

[0121] Step S24: performing key point identification on the stress field distribution data inside the device to obtain key point positions and stress data.

[0122] In the embodiment of the present invention, environmental parameter data is acquired: environmental parameter data includes temperature, humidity, electromagnetic interference intensity, mechanical vibration, etc.

[0123] Temperature and humidity: Install a temperature and humidity sensor, such as the SHT31, near power sensing devices (current transformers, for example) to monitor ambient temperature and humidity in real time. Set the sensor's data acquisition frequency, such as once per minute. Transfer the collected data to a computer using a data acquisition system (such as NI CompactDAQ) and save it as a text file or Excel file.

[0124] Electromagnetic interference intensity: Use an electromagnetic field intensity meter, such as the TES-92, to measure the electromagnetic interference intensity at various locations around the current transformer. Record the measurement locations and the corresponding electromagnetic field intensity values.

[0125] Mechanical vibration: If necessary, a vibration sensor can be installed on the housing of the current transformer to measure the vibration.

[0126] Applying boundary conditions: Open the calibrated device twin model obtained in step S16 (COMSOL Multiphysics model file .mph).

[0127] Temperature Boundary Condition: In COMSOL, select the Heat Transfer physics interface and add a Temperature boundary under Boundary Conditions. Select all exterior surfaces of the current transformer housing to designate them as temperature boundaries. Enter the acquired ambient temperature data (for example, from a text file) as the temperature value. If the ambient temperature varies with time, enter a time-varying temperature function.

[0128] Humidity boundary condition: If you need to consider the effect of humidity on the performance of insulation materials, you can add the "Humidity Transport" physics interface in COMSOL (requires the corresponding module) and add a "Humidity" boundary in the "Boundary Conditions".

[0129] Electromagnetic Interference Boundary Condition: In COMSOL, select the AC / DC physics interface and add a Magnetic Field boundary under Boundary Conditions. Create an air domain around the current transformer, enclosing the entire device. Designate the outer surface of the air domain as a magnetic field boundary. Input the electromagnetic interference intensity as the magnetic field strength value based on the measurement results of the electromagnetic field intensity meter.

[0130] Device twin after loading environmental parameters: After completing the above boundary condition settings, save the COMSOL model file.

[0131] This model file is the device twin after the environment parameters are loaded.

[0132] Simulation type setting: Open the device twin model (COMSOLMultiphysics model file .mph) after loading the environmental parameters obtained in step S21. Set the simulation type according to the analysis requirements. For example, if you are concerned about the temperature distribution and electromagnetic field distribution in the steady state, select the "steady-state" study; if you are concerned about the transient changes of the device during startup or failure, select the "transient" study. Solver settings: In COMSOL, select "Solver Configuration" under the "Study" node. Select the solver type according to the model size and complexity. For linear problems, direct solvers (such as MUMPS) are generally more effective; for nonlinear problems, iterative solvers (such as GMRES) are generally more stable. Set the relative tolerance and absolute tolerance of the solver to control the solution accuracy. Simulation run: Click the "Calculate" button in COMSOL to start the simulation. COMSOL will perform multi-physics coupling calculations based on the set physical field interface, boundary conditions and solver parameters. During the calculation process, COMSOL will display the calculation progress and residual curve.

[0133] Simulation result data: After the simulation is completed, COMSOL will automatically generate various result data, including:

[0134] Temperature field distribution: temperature distribution cloud diagram and values of each component of the current transformer.

[0135] Electromagnetic field distribution: distribution cloud diagrams and values of electric field strength, magnetic field strength, current density, etc. inside and around the current transformer.

[0136] Stress field distribution: stress distribution cloud diagram and value of each component of the current transformer (to be processed in detail in step S23).

[0137] Flow field distribution: If there is fluid in the model, there will be flow field results.

[0138] Export these result data as text files, table files, or image files as the simulation result data of each physical field.

[0139] Stress Data Extraction: Open the simulation results (COMSOL Multiphysics model file .mph) obtained in step S22. Under the "Results" node in COMSOL, find the "Stress (solid)" dataset. This dataset contains the stress tensor data for each component of the current transformer. You can observe the stress distribution by plotting the stress components (such as von Mises stress and principal stress). You can also use the "Derived Values" function to calculate stress values at specific locations or components. Export this stress data as a text file or a table file.

[0140] Stress data post-processing:

[0141] Stress Tensor Calculation: Stress data exported from COMSOL typically contains the components of the stress tensor (σx, σy, σz, τxy, τyz, τzx). Principal stresses (σ1, σ2, σ3) and von Mises equivalent stresses can be calculated as needed.

[0142] Stress concentration factor calculation: Determine stress concentration areas (such as sharp corners, hole edges, etc.) and calculate the stress concentration factor (the ratio of maximum stress to nominal stress).

[0143] Fatigue analysis: If you are concerned about the fatigue life of a device, you can combine the material's fatigue properties (such as the SN curve) with stress data to calculate fatigue damage accumulation. For example, you can use Miner's linear damage accumulation rule.

[0144] Data visualization: Generate stress distribution cloud maps, contour maps, etc.

[0145] Stress field distribution data inside the equipment: The post-processed stress data (including stress components, principal stresses, von Mises stresses, stress concentration factors, fatigue damage, etc.) are organized into tables or text files as stress field distribution data inside the equipment.

[0146] Key point identification method:

[0147] Stress threshold method: Set a stress threshold, such as the yield strength or fatigue limit of the material, and identify the regions or nodes in the stress field distribution data that exceed the threshold as key points.

[0148] Stress Gradient Method: Calculates the gradient of the stress field and identifies regions or nodes with large gradient values as key points. Large stress gradients indicate dramatic stress changes and are prone to stress concentration.

[0149] Combined with structural features: Combined with the structural features of the current transformer, such as the contact surface between the winding and the insulation layer, the sharp corners of the iron core, etc., these parts are identified as key points.

[0150] Determine the key point locations: Based on the selected key point identification method, find nodes or regions that meet the conditions in the stress field distribution data. Record the coordinates of these nodes or regions as the key point locations.

[0151] Key point stress data extraction: Extract stress values (including stress components, principal stress, von Mises stress, etc.) at key points in stress field distribution data.

[0152] Key point locations and stress data: Organize key point locations and corresponding stress data into a table or text file as key point location and stress data. The table contains key point number, coordinates, stress components, principal stresses, von Mises stresses, and other information.

[0153] Preferably, step S3 includes the following steps:

[0154] Step S31: selecting a degradation model and determining degradation model parameters based on key point positions and stress data and a preset material property database to obtain a degradation model and degradation model parameters;

[0155] Step S32: introducing random factors into degradation model parameters to obtain random factor input data;

[0156] Step S33: performing macro-performance degradation simulation based on micro-crack propagation on key point positions and stress data according to the degradation model and random factor input data to obtain macro-performance degradation data;

[0157] Step S34: constructing a performance degradation trajectory for the key point performance degradation data to obtain a performance degradation trajectory.

[0158] In the embodiment of step S31 of the present invention:

[0159] Degradation model selection: Based on the key point location and stress data obtained in step S24 and the material property database preset in step S131, determine the material to which each key point belongs and its main failure mode. For example, for a current transformer:

[0160] Primary winding and secondary winding (copper): The main failure modes are electrochemical corrosion and fatigue. Epoxy resin insulation layer:

[0161] The main failure modes are thermal aging, electrical aging, and breakdown caused by partial discharge. Iron core (silicon steel sheet): The main failure modes are magnetic aging and fatigue.

[0162] For different failure modes, select the corresponding degradation model. For example:

[0163] For thermal aging of epoxy resins, the Arrhenius model is used: k = A*exp(-Ea / (R*T)), where k is the reaction rate constant, A is the pre-exponential factor, Ea is the activation energy, R is the gas constant, and T is the absolute temperature. For electrical aging of epoxy resins, the Inverse Power Law (IPL) model is used: t = C*E^(-n), where t is the lifespan, C and n are material-dependent constants, and E is the electric field strength. For metal fatigue, the Coffin-Manson model is used for high stress levels; the Basquin model is used for low stress levels.

[0164] Degradation model parameter determination: For the selected degradation model, determine the parameter values in the model.

[0165] Obtain reference values for model parameters by consulting material property databases, literature, or conducting experiments. For example, for the Arrhenius model, consult the activation energy Ea and pre-exponential factor A of epoxy resin. Use stress data (such as temperature and electric field strength) at key locations to adjust model parameters. For example, if the temperature at the key point is higher, the Arrhenius model will calculate a higher reaction rate constant k. Organize the finalized model parameters into a table.

[0166] In the embodiment of step S32:

[0167] Identify random factors: Consider randomness in material properties, environmental conditions, and manufacturing processes. For example:

[0168] Randomness of material properties: Parameters such as the activation energy Ea and pre-exponential factor A of epoxy resin are not actually fixed values, but fluctuate within a certain range. Randomness of environmental conditions: Parameters such as ambient temperature and humidity will fluctuate with time and space. Randomness of manufacturing process: The turn spacing of the winding, the thickness of the insulation layer, etc.

[0169] There will be certain deviations in the manufacturing process.

[0170] Probability distribution of random factors: For each random factor, select a probability distribution to describe its fluctuation range. Common probability distributions include normal distribution, lognormal distribution, and Weibull distribution. Based on existing data or experience, determine the parameters of the probability distribution (such as mean and standard deviation). For example, assume that the activation energy Ea of epoxy resin follows a normal distribution with a mean of 1.2 eV and a standard deviation of 0.05 eV.

[0171] Random sampling: Using the Monte Carlo method, random sampling is performed for each random factor. The number of samplings is determined based on the accuracy requirements; for example, 1000 samplings are performed. For each sampling, a set of random numbers is generated, representing the specific value of each random factor. For example, in the first sampling, Ea = 1.22 eV; in the second sampling, Ea = 1.18 eV; and so on.

[0172] Random factor input data: The random numbers obtained from each sampling are substituted into the degradation model determined in step S31 as model parameters. This results in a different set of model parameters for each sampling. These model parameters are organized into a table or array and used as random factor input data.

[0173] In the embodiment of step S33:

[0174] Microcrack Growth Model Establishment: For insulating materials (such as epoxy resin), a microcrack growth model is established. For example, a modified form of the Paris equation is used: da / dN = C(ΔK)^m, where da / dN is the crack growth rate, C and m are material constants, and ΔK is the stress intensity factor range. ΔK can be calculated based on stress data at key points.

[0175] Macro-performance degradation modeling: Establishing the relationship between micro-crack growth and macro-performance degradation. For example, for epoxy resin, the dielectric strength Eb can be related to the crack length a: Eb = Eb0 - f(a), where Eb0 is the initial dielectric strength and f(a) is a function related to the crack length.

[0176] Degradation simulation process:

[0177] For each set of random factor input data obtained in step S32 (i.e., each set of degradation model parameters):

[0178] Initialization: Set the initial time t=0 and the initial crack length a=a0 (a0 can be a very small value, representing the initial defect).

[0179] cycle:

[0180] According to the current time t and the stress data (such as temperature and electric field strength) at the key points, the parameter values in the degradation model (such as k in the Arrhenius model, C and n in the IPL model) are calculated.

[0181] According to the micro crack growth model (such as the Paris formula), the crack growth amount Δa within the current time step Δt is calculated.

[0182] Update crack length: a=a+Δa.

[0183] According to the macro performance degradation model, the macro performance parameter value (such as dielectric strength Eb) corresponding to the current crack length a is calculated.

[0184] Update time: t=t+Δt.

[0185] The above cycle is repeated until the set simulation time is reached or the macro performance parameter value is lower than the set failure threshold.

[0186] Record the macro performance parameter values at each time step.

[0187] Macro performance degradation data: All randomly sampled macro performance parameter values ​​changing over time are organized into tables or arrays as macro performance degradation data.

[0188] In the embodiment of step S34:

[0189] Data organization: The macro-performance degradation data obtained in step S33 is organized into a table. Each row of the table represents the result of a Monte Carlo simulation (corresponding to a set of random factor input data), each column represents a time point, and the value in the cell represents the macro-performance parameter value (such as dielectric strength) at that time point and in that simulation.

[0190] Trajectory plotting: For each row of data in the table, plot it as a curve, with time on the horizontal axis and the macro-performance parameter value on the vertical axis. This creates a set of performance degradation trajectories. Each trajectory represents the performance degradation of the device over time under a possible combination of random factors.

[0191] Statistical analysis (optional):

[0192] The average of all trajectories at each time point is calculated to obtain the average performance degradation trajectory.

[0193] Calculate the standard deviation or confidence interval of all trajectories at each time point to reflect the uncertainty of performance degradation.

[0194] Performance degradation trajectory: The drawn performance degradation trajectory (including single simulation trajectory, average trajectory, confidence interval, etc.) is output in the form of a graph or table as the performance degradation trajectory.

[0195] Preferably, step S331 includes the following steps:

[0196] Step S331: performing micro-scale crack initiation simulation based on key point positions and stress data, degradation model, and random factor input data to obtain micro-crack initiation data;

[0197] Step S332: constructing a mesoscale crack propagation model based on the microcrack initiation data and key point positions and stress data to obtain a mesoscale crack propagation model;

[0198] Step S333: applying a mesoscale stress field to the mesoscale crack growth model to obtain a mesoscale stress field;

[0199] Step S334: performing crack growth simulation according to the mesoscopic stress field, and recording the crack growth trajectory and rate to obtain mesoscopic crack growth data;

[0200] Step S335: performing microcrack density statistics in the key point area based on the mesoscopic crack growth data, analyzing the time variation pattern, and obtaining microcrack density data;

[0201] Step S336: selecting a macro insulation performance parameter degradation model based on the microcrack density data, and calculating the time-varying macro insulation performance parameters to obtain macro performance degradation data.

[0202] In the embodiment of step S331 of the present invention:

[0203] Microscale Modeling: For insulating materials (such as epoxy resin), create a 2D or 3D model at the microscale (e.g., micrometer level) that includes lattice defects. Molecular dynamics simulation tools such as LAMMPS or finite element analysis tools such as Abaqus can be used. The model geometry should represent the material's microstructure.

[0204] Stress loading: Based on the keypoint locations and stress data obtained in step S24, the stress (or stress history) at the keypoints is applied as a load to the microscale model. For example, in Abaqus, displacement or force boundary conditions corresponding to the stress state at the keypoints can be applied to the model boundaries.

[0205] Degradation model introduction: The degradation model selected in step S31 (such as the Arrhenius model or the IPL model) is combined with the microscale model. For example, in a molecular dynamics simulation, the atomic diffusion rate at different temperatures can be calculated according to the Arrhenius model and used as a simulation parameter.

[0206] Consideration of random factors: The random factor input data obtained in step S32 is introduced into the microscale model. For example, in Abaqus, random assignment of material properties (such as Young's modulus and fracture toughness) can be achieved by writing Python scripts.

[0207] Crack initiation criterion: Set the crack initiation criterion. For example, when the stress or strain in a certain area exceeds the microscopic strength of the material, crack initiation is considered to have occurred there. The microscopic strength can be obtained through experimental measurement or molecular dynamics simulation.

[0208] Simulation process: Run the microscale model. During the simulation, monitor the stress, strain, and atomic / element state of each region in the model. Determine whether new microcracks have formed based on the crack initiation criteria. Record the time, location, and initial size of new cracks.

[0209] Microcrack initiation data: Organize the microcrack initiation information obtained from the simulation (including the number, location, and size of cracks over time) into a table or array as microcrack initiation data. For example, at time t1, n1 microcracks were generated, with locations (x1, y1), (x2, y2), ..., and initial sizes a1, a2, ..., respectively.

[0210] In the embodiment of step S332:

[0211] Mesoscale modeling: Create a 2D or 3D model containing the initial crack at the mesoscale (e.g., millimeter level). Finite element analysis tools such as Abaqus can be used. The model geometry should be representative of the material region near the critical point.

[0212] Initial Crack Introduction: Based on the microcrack initiation data obtained in step S331, the initiated microcracks are introduced as initial cracks into the mesoscale model. For example, in Abaqus, the "Seam Crack" function can be used to create a crack of a specified size at a specified location.

[0213] Crack Growth Model Selection: Select an appropriate crack growth model. For example, use the Paris formula: da / dN = C(ΔK)^m, where da / dN is the crack growth rate, C and m are material constants, and ΔK is the stress intensity factor range. C and m can be determined experimentally or obtained from a material property database.

[0214] Cohesive Zone Model (optional): To more accurately simulate the fracture process at the crack tip, the Cohesive Zone Model (CZM) can be used. The CZM describes the fracture behavior of the material by introducing a cohesive force-displacement relationship at the crack surface.

[0215] Mesoscale crack growth model: The above model parameters (such as C and m in the Paris formula, or the parameters in CZM) and the initial crack information are integrated into the finite element model to form a mesoscale crack growth model.

[0216] In the embodiment of step S333:

[0217] Stress field calculation: Based on the key point locations and stress data obtained in step S24, the stress field distribution near the key points is calculated. Since the stress field obtained in step S2 is based on the continuum mechanics assumption, and the mesoscale model takes into account the presence of cracks, the stress field needs to be recalculated.

[0218] Applying boundary conditions: Apply the calculated stress field as a boundary condition to the mesoscale model. For example, in Abaqus, you can apply displacement or force boundary conditions corresponding to the stress field to the model's boundaries. If the stress field varies with time, you will need to apply time-varying boundary conditions.

[0219] Mesoscale stress field: After applying boundary conditions, run a finite element model (such as Abaqus) to calculate the mesoscale stress field that takes the presence of cracks into account. This stress field will be used to drive crack growth.

[0220] In the embodiment of step S334:

[0221] Crack propagation simulation: Crack propagation simulation is performed based on the mesoscale crack propagation model established in step S332 and the mesoscale stress field obtained in step S333. For example, in Abaqus, the extended finite element method (XFEM) or virtual crack closure technology (VCCT) can be used to simulate crack propagation. Crack propagation trajectory recording: During the simulation process, the changes in the position of the crack tip over time are recorded. For example, in Abaqus, the coordinates of the crack tip can be output using the "History Output" function. Crack propagation rate calculation: The crack propagation rate is calculated based on the changes in the crack tip position over time. For example, the difference method can be used to calculate da / dt. Mesoscopic crack propagation data: The crack propagation trajectory (the changes in the crack tip position over time) and the crack propagation rate data are organized into tables or arrays as mesoscopic crack propagation data.

[0222] In the embodiment of step S335:

[0223] Definition of key point area: Based on the key point position obtained in step S24, determine an area containing the key point. The size of this area should be larger than the size of the mesoscopic scale model so as to include all possible crack propagation paths. Crack counting: Based on the mesoscopic crack propagation data obtained in step S334, count the number of cracks existing in the key point area at different times. Microcrack density calculation: Divide the number of cracks by the area (or volume) of the key point area to obtain the microcrack density. Time variation law analysis: Draw a curve of microcrack density changing with time. Analyze the trend of the curve, for example, does the microcrack density increase with time? How does the rate of increase change? Microcrack density data: Organize the data of microcrack density changing with time into a table or array as microcrack density data.

[0224] In the embodiment of step S336:

[0225] Selection of macro-insulation performance parameter degradation model: Based on material science principles and experimental data, select a model that can reflect the impact of microcrack density on macro-insulation performance.

[0226] Calculation of Time-Varying Macroscopic Insulation Performance Parameters: Substitute the microcrack density data D(t) obtained in step S335 into the above model to calculate the time-varying macroscopic insulation performance parameters such as resistivity ρ(t), dielectric strength Eb(t), and dielectric loss tangent tanδ(t). More specifically, the calculation process is explained using the dielectric strength degradation model as an example:

[0227] Get simulation time step sequence data: Extract time step sequence data from the solution results of simulation software (such as COMSOL or Abaqus). For example, if the total simulation duration is 1000 hours and the time step is 1 hour, the time series is t = [0, 1, 2, ..., 1000] hours.

[0228] Obtaining temperature and electric field strength data: For each time step, extract the temperature T(t) and electric field strength E(t) data at key points from the simulation results. These data have been obtained from the previous multiphysics coupling simulation.

[0229] Calculate the time-varying dielectric strength: For each time step t, substitute the microcrack density D(t), temperature T(t), and electric field strength E(t) into the dielectric strength degradation model Eb(t) = Eb0*f2(D(t), T(t), E(t)) to calculate the dielectric strength Eb(t) at that moment. Note that the f2 function here may depend on the microcrack density, temperature, and electric field strength simultaneously, and its specific form needs to be determined based on material properties and experimental data.

[0230] Similarly, other macroscopic insulation performance parameters such as time-varying resistivity ρ(t) and dielectric loss tangent tanδ(t) can be calculated.

[0231] Macro-performance degradation data: The calculated time-varying macro-insulation performance parameters (such as Eb(t), ρ(t), and tanδ(t)) are organized into tables or arrays as macro-performance degradation data. This data reflects the time-varying macro-performance of the insulation material under the influence of microcrack propagation.

[0232] Preferably, step S336 is specifically as follows:

[0233] Determine a coupled degradation model based on performance parameters according to microcrack density data to obtain a coupled degradation model formula, wherein the coupled degradation model formula includes a resistivity degradation model formula, a dielectric strength degradation model formula, a dielectric loss tangent degradation model formula, and a partial discharge inception voltage degradation model formula;

[0234] Acquire simulation time step sequence data; acquire temperature values and electric field strength values corresponding to the time step according to the simulation time step sequence data to obtain temperature data and electric field strength data;

[0235] The time-varying resistivity is calculated based on the microcrack density data, resistivity degradation model formula, temperature data and electric field intensity data to obtain the time-varying resistivity data;

[0236] The time-varying dielectric strength data is obtained by calculating the time-varying dielectric strength based on the microcrack density data, the dielectric strength degradation model formula, the temperature data and the electric field strength data;

[0237] The time-varying dielectric loss tangent is calculated based on the microcrack density data, dielectric loss tangent degradation model formula, temperature data, and electric field strength data to obtain the time-varying dielectric loss data;

[0238] The time-varying partial discharge inception voltage is calculated based on the microcrack density data, the dielectric strength degradation model formula, the temperature data, and the electric field strength data to obtain the time-varying partial discharge voltage data.

[0239] The macro performance degradation data is integrated by integrating the time-varying resistivity data, time-varying dielectric strength data, time-varying dielectric loss data and time-varying partial discharge voltage data to obtain the macro performance degradation data.

[0240] In the embodiments of the present invention, coupling relationship models between microcrack density and other macroscopic performance parameters are established based on material science theory, experimental data, and existing research results. These models should be able to reflect the comprehensive impact of factors such as microcrack density, temperature, and electric field strength on material properties.

[0241] Resistivity degradation model formula: The following model can be used: ρ(t) = ρ0*[1+α*D(t)^β]*exp(γ*(T(t)-T0)); where ρ(t) is the resistivity at time t, ρ0 is the initial resistivity (resistivity when there are no defects), D(t) is the microcrack density at time t, T(t) is the temperature at time t, T0 is the reference temperature, and α, β, and γ are material constants obtained by fitting experimental data.

[0242] Dielectric strength degradation model formula: The following model can be used: Eb(t) = Eb0*[1-a*D(t)^b]*exp(-c*(T(t)-T0))*(E0 / E(t))^n; where Eb(t) is the dielectric strength at time t, Eb0 is the initial dielectric strength, D(t) is the microcrack density at time t, T(t) is the temperature at time t, E(t) is the electric field strength at time t, T0 is the reference temperature, E0 is the reference electric field strength, and a, b, c, and n are material constants obtained by fitting experimental data.

[0243] Dielectric loss tangent degradation model formula: The following model can be used: tanδ(t)=tanδ0*[1+p*D(t)^q]*exp(r*(T(t)-T0)); where tanδ(t) is the dielectric loss tangent at time t, tanδ0 is the initial dielectric loss tangent, and p, q, and r are constants related to the material.

[0244] Partial discharge inception voltage degradation model formula: The following model can be used: Vpd(t) = Vpd0*[1-x*D(t)^y]*exp(-z*(T(t)-T0))*(E0 / E(t))^m; where Vpd(t) is the partial discharge inception voltage at time t, Vpd0 is the initial partial discharge inception voltage, D(t) is the microcrack density at time t, and x, y, z, and m are material-related constants.

[0245] The above model formulas are stored in text or code form to form a coupled degradation model formula library.

[0246] Extract time-step sequence data from the solution results of finite element simulation software (such as COMSOL and Abaqus). For example, if the total simulation duration is 1000 hours and the time step is 1 hour, the time series is t = [0, 1, 2, ..., 1000] hours. Save the time series data as a text file or array. Open the result file of a previously performed multiphysics coupled simulation (such as a .mph file from COMSOL or a .odb file from Abaqus). Find the output variables for the temperature field and electric field intensity in the results. For each time step, extract the temperature and electric field intensity values at the key points. In COMSOL, use the "Derived Values" -> "Point Calculation" function, select the key points, and specify the variables to be extracted (temperature or electric field intensity). In Abaqus, use the "Probe Values" tool, select the key points, and specify the variables to be extracted. Save the extracted temperature and electric field intensity values in chronological order as a text file or array to form the temperature data T(t) and the electric field intensity data E(t).

[0247] For each time step t: Obtain the current microcrack density value from D(t). Obtain the current temperature value from T(t). Substitute these values into the resistivity degradation model formula to calculate the current resistivity ρ(t). Save the calculated resistivity ρ(t) in chronological order as a text file or array to form time-varying resistivity data.

[0248] For each time step t: Obtain the current microcrack density value from D(t). Obtain the current temperature value from T(t). Obtain the current electric field strength value from E(t). Substitute these values into the dielectric strength degradation model formula to calculate the current dielectric strength Eb(t). Save the calculated dielectric strength Eb(t) in chronological order as a text file or array to form time-varying dielectric strength data.

[0249] For each time step, the microcrack density and temperature values are substituted into the dielectric loss tangent degradation model formula to calculate the time-varying dielectric loss tangent and obtain the time-varying dielectric loss data.

[0250] Substituting the data into the partial discharge inception voltage degradation model formula, the time-varying partial discharge voltage data is obtained.

[0251] The time-varying resistivity data, time-varying dielectric strength data, time-varying dielectric loss data, and time-varying partial discharge voltage data obtained in the previous steps are integrated into a table or data structure.

[0252] Preferably, step S4 includes the following steps:

[0253] Step S41: Obtaining equipment technical specifications and historical operating data; determining a failure threshold based on the equipment technical specifications, historical operating data, and performance degradation trajectory to obtain a failure threshold;

[0254] Step S42: fitting the failure time distribution according to the failure threshold and the performance degradation trajectory to obtain the failure time probability density function parameters;

[0255] Step S43: Calculate the failure probability of the failure time probability density function parameters to obtain failure probability data for different time periods;

[0256] Step S44: acquiring real-time monitoring data; performing Bayesian update on the failure probability data of different time periods using the real-time monitoring data to obtain updated failure probability data;

[0257] Step S45: Draw a failure probability curve for the updated failure probability data to obtain a failure probability curve.

[0258] In an embodiment of the present invention, technical specification documents of power sensing devices (taking current transformers as an example) are collected, such as product manuals, factory inspection reports, industry standards, etc. The rated values and allowable ranges of key performance parameters are extracted from these documents. For example, for current transformers, the following parameters are of interest: rated current ratio, accuracy level, insulation resistance, dielectric strength, and partial discharge level; historical operating data of this model of current transformer or similar equipment is collected, including: online monitoring data: such as temperature, current, partial discharge amount, gas content in oil (if applicable), etc. Offline detection data: such as insulation resistance, dielectric strength, dielectric loss tangent, etc. Fault records: such as fault occurrence time, fault type, fault cause, etc. Maintenance records: such as maintenance time, maintenance content, replacement parts, etc. Failure thresholds are set based on the allowable ranges of performance parameters given in the technical specifications. For example, if the technical specifications stipulate that the insulation resistance must not be lower than 1000MΩ, 1000MΩ can be used as the failure threshold of the insulation resistance. Analyze historical operating data, especially fault data, to find out the changing trends and critical values of performance parameters before the fault occurs. For example, if it's found that most failures occur when the dielectric strength is below 60% of the initial value, 60% of the initial value can be used as the dielectric strength failure threshold. Combined with the performance degradation trajectories obtained in step S34, observe when different trajectories reach critical values determined by technical specifications or historical data. By comprehensively considering the conditions of multiple trajectories, set the failure threshold. Organize the finalized failure thresholds into a table, including parameter names and thresholds.

[0259] For each performance degradation trajectory, find the time point at which its key performance parameters (such as dielectric strength and insulation resistance) drop to the failure threshold. This is used as the failure time for that simulation. If a trajectory does not reach the failure threshold within the simulation time, its failure time is recorded as infinity or a large value. Statistical analysis is performed on all simulated failure time data to fit a probability distribution function. The fitted probability distribution function and its parameters are recorded as the parameters of the failure time probability density function.

[0260] Calculate the probability density function \(f(t)\) based on the failure time probability density function parameters obtained in step S42. For different time periods \([t_1, t_2]\), calculate the probability that the device fails within this time period. This can be obtained by integrating the probability density function \(f(t)\) over the interval \([t_1, t_2]\): \(P(t_1 < T < t_2)=\int_{t_1}^{t_2}f(t)dt\); where \(T\) is the failure time (random variable). Numerical integration methods (such as the trapezoidal rule, Simpson's rule) can be used to calculate the integral. For example, in MATLAB, the integral function can be used for numerical integration. Calculate the failure probabilities for different time periods such as the next 1 month, 3 months, 6 months, 1 year, 2 years, etc. Organize the calculated failure probability data into a table, including the time period and the corresponding failure probability.

[0261] Install sensors on the power perception device to monitor the operating status of the device in real time. For example, for current transformers, monitor temperature, current, partial discharge amount, gas content in oil (if applicable), etc. Use a data acquisition system to transmit the sensor data to a computer in real time. Save the data as a text file, database or data stream. Use the failure probability data for different time periods obtained in step S43 as the prior probability \(P(H)\). Based on the real-time monitoring data, construct the likelihood function \(P(D|H)\). The likelihood function represents the probability of observing the current monitoring data under the assumption that the device has failed. For example, if the monitored partial discharge amount increases abnormally, the value of the likelihood function will be larger. The specific form of the likelihood function needs to be determined according to the device type, monitoring parameters and failure mode. It can be constructed based on expert experience, fault tree analysis or machine learning methods.

[0262] According to Bayes' formula, calculate the posterior probability \(P(H|D)\): \(P(H|D)=\frac{P(D|H)\times P(H)}{P(D)}\); where: \(P(H|D)\) is the updated failure probability (posterior probability); \(P(D|H)\) is the likelihood function; \(P(H)\) is the prior probability; \(P(D)\) is the evidence factor, which can be obtained by integrating or summing over all possible \(H\). Numerical calculation methods or Monte Carlo methods can be used for Bayesian updating. Use the calculated posterior probability \(P(H|D)\) as the updated failure probability data.

[0263] Take time as the abscissa and failure probability as the ordinate to draw a curve graph.绘图软件(如MATLAB,Excel,Python的matplotlib库) can be used for drawing. In the graph, data labels, legends, etc. can be added to make the curve easier to understand. Save the drawn curve graph as a picture file or vector graph file as the failure probability curve. This curve intuitively shows the trend of the probability of the device failing in different future time periods.

[0264] It should be noted that there is an incomplete expression in the translation of item where "绘图软件(如MATLAB,Excel,Python的matplotlib库)" is not in English. It should be something like "Drawing software (such as MATLAB, Excel, the matplotlib library in Python)". Please check and correct it according to the actual situation.Preferably, the present invention further provides a power-aware device fault prediction system for executing the power-aware device fault prediction method described above, the power-aware device fault prediction system comprising:

[0265] The twin modeling module is used to construct a geometric model of the power sensing device to obtain mesh model data; assign material properties to the mesh model data based on the association between microscopic characteristics and macroscopic properties to obtain material property mesh model data; and calibrate model parameters of the material property mesh model data to obtain a calibrated device twin;

[0266] The stress mapping module is used to load environmental parameters on the calibrated device twin and perform multi-physics field coupling simulation inside the device to obtain stress field distribution data inside the device; it also identifies key points in the stress field distribution data inside the device to obtain key point locations and stress data;

[0267] The degradation simulation module is used to determine the degradation model parameters based on the key point positions and stress data to obtain the degradation model parameters; perform macro-performance degradation simulation based on micro-crack propagation on the key point positions and stress data according to the degradation model and random factor input data to obtain macro-performance degradation data; and construct a performance degradation trajectory for the key point performance degradation data to obtain the performance degradation trajectory;

[0268] The failure probability assessment module is used to fit the failure time distribution based on the failure threshold according to the performance degradation trajectory to obtain the failure time probability density function parameters; and to predict the failure probability of the power sensing device based on the failure time probability density function parameters to obtain the failure probability curve.

[0269] Preferably, when the computer program is executed, the power-aware device fault prediction method as described above is implemented.

[0270] 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.

[0271] 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 method for predicting faults of power sensing equipment, characterized in that: The following steps are involved: Step S1: construct a geometric model of the power sensing device to obtain mesh model data; assign material properties based on the association between microscopic characteristics and macroscopic properties to the mesh model data to obtain material property mesh model data; calibrate model parameters of the material property mesh model data to obtain a calibrated device twin; Step S2: Loading environmental parameters to the calibrated device twin and performing multi-physics field coupling simulation inside the device to obtain stress field distribution data inside the device; identifying key points in the stress field distribution data inside the device to obtain key point positions and stress data; Step S3: Determine degradation model parameters based on key point positions and stress data to obtain degradation model parameters; perform macro-performance degradation simulation based on micro-crack propagation on key point positions and stress data based on the degradation model and random factor input data to obtain macro-performance degradation data; Construct a performance degradation trajectory for key point performance degradation data to obtain a performance degradation trajectory; Step S4: fitting the failure time distribution based on the failure threshold according to the performance degradation trajectory to obtain the failure time probability density function parameters; predicting the failure probability of the power sensing device according to the failure time probability density function parameters to obtain the failure probability curve.

2. The power sensing device fault prediction method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: creating a geometric model of the power sensing device to obtain three-dimensional geometric model data; Step S12: Meshing the three-dimensional geometric model data to obtain meshed model data; Step S13: assigning material properties based on the association between microscopic properties and macroscopic properties to the mesh model data to obtain material property mesh model data; Step S14: establishing a multi-physics field coupling model according to the material property grid model data to obtain a multi-physics field coupling model; Step S15: collecting sensor data from the power sensing device in real time to obtain real-time sensor data; Step S16: Input the real-time sensor data into the multi-physics field coupling model to calibrate the model parameters to obtain a calibrated device twin.

3. The power sensing device fault prediction method according to claim 2, characterized in that: Step S13 includes the following steps: Step S131: identifying the microscopic characteristics of key components of the mesh model data according to a preset material property database to obtain a list of materials for key components; Step S132: searching for macro material property parameters based on the key component material list to obtain a macro property parameter table; acquiring micro material property data based on the key component material list to obtain micro property data; Step S133: determining the association between micro-properties and macro-properties based on the micro-property data and the macro-property parameter table to obtain an associated property list; Step S134: performing conductivity-grain size correlation for conductor materials, performing dielectric strength-lattice defect density correlation for insulating materials, and performing elastic modulus-grain size correlation for metal materials according to the correlation attribute list to obtain a constitutive model library; Step S135: assigning and verifying model parameters of the constitutive model library to obtain a constitutive model parameter set; Step S136: assigning mesh unit material properties to the meshing model data according to the constitutive model parameter set, the macroscopic property parameter table, and the microscopic property data to obtain material property mesh model data.

4. The method for predicting faults of power sensing equipment according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: Acquire environmental parameter data; apply the environmental parameter data as boundary conditions to the calibrated device twin to obtain the device twin after loading the environmental parameters; Step S22: performing multi-physics field coupling simulation on the device twin after loading the environmental parameters to obtain simulation result data of each physical field; Step S23: performing stress extraction and post-processing on the simulation result data of each physical field to obtain stress field distribution data inside the device; Step S24: performing key point identification on the stress field distribution data inside the device to obtain key point positions and stress data.

5. The power sensing device fault prediction method according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: selecting a degradation model and determining degradation model parameters based on key point positions and stress data and a preset material property database to obtain a degradation model and degradation model parameters; Step S32: introducing random factors into degradation model parameters to obtain random factor input data; Step S33: performing macro-performance degradation simulation based on micro-crack propagation on key point positions and stress data according to the degradation model and random factor input data to obtain macro-performance degradation data; Step S34: constructing a performance degradation trajectory for the key point performance degradation data to obtain a performance degradation trajectory.

6. The method for predicting faults of power sensing equipment according to claim 5, characterized in that: Step S331 includes the following steps: Step S331: performing micro-scale crack initiation simulation based on key point positions and stress data, degradation model, and random factor input data to obtain micro-crack initiation data; Step S332: constructing a mesoscale crack propagation model based on the microcrack initiation data and key point positions and stress data to obtain a mesoscale crack propagation model; Step S333: applying a mesoscale stress field to the mesoscale crack growth model to obtain a mesoscale stress field; Step S334: performing crack growth simulation according to the mesoscopic stress field, and recording the crack growth trajectory and rate to obtain mesoscopic crack growth data; Step S335: performing microcrack density statistics in the key point area based on the mesoscopic crack growth data, analyzing the time variation pattern, and obtaining microcrack density data; Step S336: selecting a macro insulation performance parameter degradation model based on the microcrack density data, and calculating the time-varying macro insulation performance parameters to obtain macro performance degradation data.

7. The power sensing device fault prediction method according to claim 6, characterized in that: Step S336 is specifically as follows: Determine a coupled degradation model based on performance parameters according to microcrack density data to obtain a coupled degradation model formula, wherein the coupled degradation model formula includes a resistivity degradation model formula, a dielectric strength degradation model formula, a dielectric loss tangent degradation model formula, and a partial discharge inception voltage degradation model formula; Acquire simulation time step sequence data; acquire temperature values and electric field strength values corresponding to the time step according to the simulation time step sequence data to obtain temperature data and electric field strength data; The time-varying resistivity is calculated based on the microcrack density data, resistivity degradation model formula, temperature data and electric field intensity data to obtain the time-varying resistivity data; The time-varying dielectric strength data is obtained by calculating the time-varying dielectric strength based on the microcrack density data, the dielectric strength degradation model formula, the temperature data and the electric field strength data; The time-varying dielectric loss tangent is calculated based on the microcrack density data, dielectric loss tangent degradation model formula, temperature data, and electric field strength data to obtain the time-varying dielectric loss data; The time-varying partial discharge inception voltage is calculated based on the microcrack density data, the dielectric strength degradation model formula, the temperature data, and the electric field strength data to obtain the time-varying partial discharge voltage data. The macro performance degradation data is integrated by integrating the time-varying resistivity data, time-varying dielectric strength data, time-varying dielectric loss data and time-varying partial discharge voltage data to obtain the macro performance degradation data.

8. The method for predicting faults of power sensing equipment according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: Obtaining equipment technical specifications and historical operating data; determining a failure threshold based on the equipment technical specifications, historical operating data, and performance degradation trajectory to obtain a failure threshold; Step S42: fitting the failure time distribution according to the failure threshold and the performance degradation trajectory to obtain the failure time probability density function parameters; Step S43: Calculate the failure probability of the failure time probability density function parameters to obtain failure probability data for different time periods; Step S44: acquiring real-time monitoring data; performing Bayesian update on the failure probability data of different time periods using the real-time monitoring data to obtain updated failure probability data; Step S45: Draw a failure probability curve for the updated failure probability data to obtain a failure probability curve.

9. A power sensing device fault prediction system, characterized in that: For executing the power-aware device fault prediction method according to claim 1, the power-aware device fault prediction system comprises: The twin modeling module is used to construct a geometric model of the power sensing device to obtain mesh model data; assign material properties to the mesh model data based on the association between microscopic characteristics and macroscopic properties to obtain material property mesh model data; and calibrate model parameters of the material property mesh model data to obtain a calibrated device twin; The stress mapping module is used to load environmental parameters on the calibrated device twin and perform multi-physics field coupling simulation inside the device to obtain stress field distribution data inside the device; it also identifies key points in the stress field distribution data inside the device to obtain key point locations and stress data; The degradation simulation module is used to determine the degradation model parameters based on the key point positions and stress data to obtain the degradation model parameters; perform macro-performance degradation simulation based on micro-crack propagation on the key point positions and stress data according to the degradation model and random factor input data to obtain macro-performance degradation data; and construct a performance degradation trajectory for the key point performance degradation data to obtain the performance degradation trajectory; The failure probability assessment module is used to fit the failure time distribution based on the failure threshold according to the performance degradation trajectory to obtain the failure time probability density function parameters; and to predict the failure probability of the power sensing device based on the failure time probability density function parameters to obtain the failure probability curve.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed, the power-aware device fault prediction method according to any one of claims 1 to 8 is implemented.

Images