Power transformer operation state evaluation method, system and equipment based on micro-service architecture and medium

Abstract

The invention discloses a micro-service architecture-based power transformer operation state evaluation method, system and equipment and a medium, and relates to the technical field of power equipment state monitoring and intelligent diagnosis, and the method comprises the steps: collecting multi-source operation data of a power transformer, carrying out the unified processing of the multi-source operation data, constructing a standardized feature vector, and carrying out the calculation of the multi-source operation data; based on multi-physics field coupling simulation, calculating field distribution and mechanical response of the power transformer, inputting standardized feature vectors into an intelligent classification model, classifying winding deformation faults, outputting fault types and confidence coefficients, performing interactive visual rendering on simulation results and intelligent classification results, and performing fault curve comparison display. And distributed service management is carried out by using a containerized micro-service mode. According to the invention, deep fusion and collaborative optimization are carried out under the micro-service architecture, so that full-chain intelligent improvement of transformer faults from mechanism revealing and accurate identification to visual decision support is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power equipment state monitoring and intelligent diagnosis, in particular to a power transformer operation state evaluation method, system, device and medium based on a micro-service architecture. BACKGROUND

[0002] As the core equipment in the power grid, the operation state of the power transformer is directly related to the safety and stability of the entire power system. Winding deformation is one of the most important mechanical faults of the transformer, caused by short-circuit current impact, leading to insulation damage and even transformer burnout. Therefore, it is of great significance to accurately and efficiently evaluate and diagnose the state of the transformer winding.

[0003] Currently, the evaluation of the state of the transformer winding mainly relies on various offline tests and online monitoring data. The existing technical solutions generally have the following limitations: first, the traditional centralized architecture system has low integration, poor stability, difficult maintenance, low efficiency, inflexibility, and serious information island phenomenon. The state evaluation relies on multiple independent systems, the system data formats are different, the interfaces are heterogeneous, which leads to difficult data fusion and makes it difficult to form a unified evaluation view, restricting the diagnosis efficiency and accuracy. Second, the multi-physical field coupling simulation requires high computing resources, which is in conflict with the real-time diagnosis demand. Precise evaluation of the winding state requires the establishment of an electromagnetic-thermal-mechanical multi-physical field coupling model. The forward simulation process is computationally intensive, difficult to expand in a single application, and unable to adapt to the rapid simulation demand under varying working conditions, making it difficult to effectively integrate with the online diagnosis system.

[0004] Finally, the existing diagnosis methods lack intelligence and scalability. Although machine learning algorithms have been applied to winding fault classification, the model training, data preprocessing, and feature extraction steps are usually tightly coupled with the business logic, making it difficult for the system to iteratively upgrade with new data and new algorithms. Model management is chaotic, and the diagnosis service has poor flexibility in high-concurrency scenarios. SUMMARY

[0005] In view of the above existing problems, the present application provides a power transformer operation state evaluation method, system, device and medium based on a micro-service architecture, to solve the problem of energy waste caused by selecting the wrong energy storage method in the prior art.

[0006] To solve the above technical problems, a power transformer operation state evaluation method based on a micro-service architecture is proposed, comprising, Multi-source operation data of a power transformer is collected, and the multi-source operation data is uniformly processed to construct a standardized feature vector; field distribution and mechanical response of the power transformer are calculated based on multi-physical field coupling simulation, the standardized feature vector is input into an intelligent classification model, winding deformation fault is classified, and fault type and confidence are output; simulation results and intelligent classification results are interactively visualized and rendered, and fault curve comparison and display are performed, and distributed service management is performed by using containerized micro-service mode.

[0007] As a preferred scheme of the power transformer operation state evaluation method based on the micro-service architecture, the uniform processing includes selecting a public frequency grid, frequency aligning frequency domain observation data, and mapping frequency points of different samples to a uniform grid. Missing points in the frequency response data are detected, and interpolation methods are used to complete the missing points to obtain continuous frequency sequences, and the amplitude features after completion are standardized.

[0008] As a preferred scheme of the power transformer operation state evaluation method based on the micro-service architecture, the multi-physical field coupling simulation includes calculating magnetic induction intensity distribution, loss density and electromagnetic force through electromagnetic field simulation, calculating steady-state temperature distribution through thermal field simulation, and calculating displacement and stress response under short-circuit impact through mechanical simulation.

[0009] As a preferred scheme of the power transformer operation state evaluation method based on the micro-service architecture, the intelligent classification model includes inputting the feature vector into a multi-class classifier, calculating the discriminant function output of each class, and distinguishing different fault types. The discriminant output is converted into the confidence of the fault category by a probability mapping function, and an interpretable diagnostic basis is provided.

[0010] As a preferred scheme of the power transformer operation state evaluation method based on the micro-service architecture, the calculation of the field distribution and the mechanical response of the power transformer includes solving the electromagnetic field by using the finite element method, dividing the calculation domain into elements by establishing a magnetic vector potential control equation, selecting a shape function to approximate the magnetic vector potential on each element, substituting the control equation to execute the weighted residual method to obtain the finite element equation, solving the sparse linear equation set to obtain the node magnetic vector potential value after applying the boundary condition, and calculating the loss density and electromagnetic force. The magnetic vector potential control equation is expressed as: wherein, is the magnetic vector potential, is the magnetic permeability, is the current density, is the curl operator, i.e. the curl of a vector field, is the shape function of the i-th node, is the node index in the finite element mesh, is the magnetic scalar potential of the i-th node; The finite element equation is expressed as: where, is the stiffness matrix, i.e. the coefficient matrix composed of shape functions and material parameters, is the node magnetic scalar potential vector, i.e. the column vector composed of all node magnetic scalar potential values, is the load vector, i.e. the vector composed of current density excitation, is the volume of the calculation domain, i.e. the spatial region covered by the finite element model, is the volume infinitesimal, is the shape function of the j-th node, and is the node index in the finite element mesh, is the transpose operation; The formula for calculating the loss density and electromagnetic force is expressed as: where, is the loss density, is the electrical conductivity, is the electric field intensity, is the modulus of the magnetic induction intensity, is the divergence operator, i.e. the divergence of a tensor field, is the Maxwell stress tensor, is the magnetic induction intensity, is the magnetic field intensity, is the unit matrix, is the transpose operation, is the electromagnetic force density; The thermal field is solved by combining the thermal circuit method and the finite element method, the heat conduction equation, the heat convection equation and the heat radiation equation are solved, and the steady-state temperature distribution is obtained; wherein, the heat conduction equation is based on the Fourier law, the heat convection equation uses the Newton cooling law, and the heat radiation equation considers the Stefan-Boltzmann law; The mechanical response is solved based on a mass-spring-damper equivalent model, the radial and axial electromagnetic forces of the winding wire cake are calculated according to the short-circuit impulse current, the equivalent dynamic model is constructed in combination with material properties and the partition of the wire cake, and the dynamic response equation is solved to obtain displacement and stress indexes.

[0011] As a preferred scheme of the power transformer operation state evaluation method based on the micro-service architecture, the classification of winding deformation faults includes using a random forest model, dividing samples into a training set and a validation set in a training process, constructing each decision tree by minimizing the Gini coefficient, setting tree parameters, fusing each tree result by using a Bagging strategy, outputting a fault category as a voting result, and setting a confidence level as an average of all tree output probabilities. The training data includes frequency response curves collected through simulation and experiments, covering normal, axial displacement, radial deformation, inter-winding spacing change and inter-winding short-circuit fault types.

[0012] As a preferred scheme of the power transformer operation state evaluation method based on the micro-service architecture, the interactive visualization rendering includes three-dimensional scalar field and vector field drawing of electromagnetic, thermal and mechanical results, temperature field display by using a mixed mode of surface rendering, section rendering and isosurface rendering, color mapping and labeling according to a numerical range, support for superimposed comparison of normal curves and fault curves in frequency response analysis, curve difference display through a graphical interface, and generation of a statistical table for comparison of key indicators. In the rendering process, multi-level detail technology is used to perform block processing on large-scale grid data, and the detail level is dynamically adjusted according to the distance of the viewpoint. The interactive visualization rendering further includes using off-screen frame buffering and texture reuse mechanisms to directly reuse the last frame rendering result when the camera view changes are less than a set threshold, and storing common scene data through a rendering cache. The distributed service management includes providing a unified service entrance through an API gateway, performing transport layer security communication and authentication and access control based on a JSON Web Token, verifying user identity and permissions, and performing asynchronous communication between services through a persistent message queue, with an idempotent identifier, a version number and a time sequence stamp carried in the message body, service-to-service calls decoupled through the message queue for fault recovery and retry.

[0013] The preferred technical scheme has the beneficial effects that the discriminant function of the multi-class classifier realizes accurate differentiation of various fault types such as axial displacement and radial deformation, the probability mapping function converts the discriminant output into a confidence level, providing a quantitative basis for operation and maintenance decisions, enhancing the credibility of the diagnosis result, and providing an interpretable diagnosis basis, enabling professionals to understand the judgment logic of the model.

[0014] As a preferred embodiment of the power transformer operation status evaluation system based on microservice architecture described in this invention, it is characterized by including a data access and preprocessing module, a forward simulation module, an inverse diagnosis module, and a visualization and reporting module.

[0015] The data access and preprocessing module is used to align frequency response data by establishing a common frequency grid, fill in missing data points using interpolation methods, and perform feature standardization operations to transform raw data from different sources and with different dimensions into standardized feature data with a unified format.

[0016] The forward modeling module is used to solve the electromagnetic field distribution, temperature field distribution, and mechanical response of the transformer based on preprocessed data through microservice-based finite element calculation, thermal circuit analysis, and equivalent dynamic model, and to perform coupled simulation of electromagnetic-thermal-mechanical multiphysics fields.

[0017] The inversion diagnosis module is used to input standardized feature vectors into a trained machine learning classification model, and intelligently identify and classify the winding state through an integrated decision tree, outputting the fault type and confidence level.

[0018] The visualization and reporting module is used to perform three-dimensional visualization of multiphysics results using a hybrid rendering strategy, and to use multi-level detail technology and rendering caching mechanism for data interaction, transforming simulation and diagnostic results into intuitive graphics and reports.

[0019] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of a method for evaluating the operating status of a power transformer based on a microservices architecture.

[0020] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for evaluating the operating status of a power transformer based on a microservices architecture.

[0021] The beneficial effects of this invention are as follows: This invention transforms multi-source heterogeneous information into high-quality, standardized feature vectors through unified data processing, laying a data foundation for the generalization ability and diagnostic reliability of intelligent models; it also reveals the physical mechanism of faults through refined multi-physics coupling simulation based on the finite element method, enabling predictive maintenance and providing simulation training data for intelligent models; the intelligent classification model constructed using algorithms such as random forests achieves automated, high-precision classification and confidence quantification of faults, improving diagnostic efficiency and objectivity; and through interactive visualization using optimization techniques such as multi-level details and rendering caching, the field distribution and diagnostic results are presented in a graphical interface, lowering the analysis threshold and improving the efficiency and accuracy of operation and maintenance decisions. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 The overall flowchart of a power transformer operation status evaluation method based on microservice architecture provided in one embodiment of the present invention is shown.

[0024] Figure 2 The flowchart of a system scheme for a power transformer operation status evaluation system based on a microservice architecture is provided as an embodiment of the present invention. Detailed Implementation

[0025] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0026] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for evaluating the operating status of power transformers based on a microservice architecture is provided, including: S100: Collects multi-source operation data of power transformers, performs unified processing on the multi-source operation data, and constructs standardized feature vectors.

[0027] S200: Based on multiphysics coupling simulation, it calculates the field distribution and mechanical response of power transformers, inputs standardized feature vectors into the intelligent classification model, classifies winding deformation faults, and outputs fault type and confidence level.

[0028] S300: Interactively visualizes and renders simulation results and intelligent classification results, and displays fault curves for comparison. It also uses containerized microservices for distributed service management.

[0029] It should be noted that by preprocessing multi-source data and constructing reliable standard inputs, the accuracy and efficiency of the analysis are improved; multi-physics coupling simulation based on the finite element method, thermal circuit method and equivalent dynamic model realizes a comprehensive physical mechanism assessment of the transformer's operating state; the use of intelligent classification models such as random forest combined with probability mapping enhances the reliability and interpretability of the diagnostic results; and interactive visualization is performed to transform the data into intuitive graphics, and rendering performance is optimized by combining off-screen frame buffering and texture reuse mechanisms.

[0030] Example 2, refer to Figure 1 This is a second embodiment of the present invention, which provides a method for evaluating the operating status of power transformers based on a microservice architecture, including: In step S100, the multi-source operating data of the power transformer collected includes transformer structural parameters, operating condition parameters, and winding deformation diagnostic datasets; wherein, transformer structural parameters and operating condition parameters include rated capacity, wiring method, material parameters, winding geometry, oil passage layout, and environmental and load conditions; the winding deformation diagnostic datasets include frequency domain and time domain observations collected by simulation or experiment, and the source type identifier is recorded in the sample layer.

[0031] In this embodiment of the application, the standardization process in step S100 includes steps S101 to S103: S101: Perform data alignment on the winding deformation diagnosis dataset, define a common frequency grid based on the full sample frequency range, and ensure that the frequency points of different samples are mapped to a unified grid. The common frequency grid formula is expressed as: in, A common frequency grid is a sequence of frequency points uniformly divided on a logarithmic scale. The minimum frequency of the entire sample is the lowest frequency value among all frequency response curves in the dataset. The highest frequency across the entire sample is the highest frequency value among all frequency response curves in the dataset. For the index of the common frequency grid, The total number of points in the common frequency grid. It is a logarithmic function with base 10.

[0032] S102: Perform interpolation to complete missing data points, generating supplementary data at the missing locations using interpolation methods; in, For the m-th target frequency point in the common frequency grid F, and For the original data and Two adjacent frequency points, and To the original frequency point and The amplitude obtained from the simulation, The value at the target frequency point is obtained through linear interpolation. The amplitude estimate at that location, and For the index of the common frequency grid.

[0033] S103: Perform feature normalization to scale the completed data, standardize the feature value distribution, and output preprocessed data suitable for subsequent simulation and diagnosis; The normalization formula is expressed as: in, For the r-th amplitude feature before standardization, The sample mean of the amplitude feature. The sample standard deviation of the amplitude characteristic. This represents the r-th amplitude feature after standardization.

[0034] In an optional implementation, in step S100, the unification process further includes: fitting known frequency points by constructing a cubic spline function to generate a smooth interpolation curve in the missing interval, and using Min-Max normalization to map the amplitude to the [0,1] interval.

[0035] In another optional implementation, in step S100, the unification process may further include finding the k most similar complete samples in the feature space, calculating the missing values ​​by weighted average, and standardizing by decimal scaling, and normalizing by moving the decimal point position of the amplitude data.

[0036] In this embodiment of the application, in step S200, the multiphysics coupling simulation includes steps S201 to S203: S201: Solve the electromagnetic field distribution through numerical simulation, and calculate the electromagnetic force and loss density based on the input parameters; The electromagnetic field is solved using the finite element method. By establishing the magnetic vector potential control equation, the computational domain is divided into tetrahedral or hexahedral elements. A shape function is selected to approximate the magnetic vector potential on each element. After substituting into the control equation, the weighted residual method is executed to obtain the finite element equation. After applying boundary conditions, the sparse linear equation system is solved to obtain the nodal magnetic vector potential values, and the loss density and electromagnetic force are calculated. The governing equations of the magnetic vector potential are expressed as follows: in, The magnetic vector potential, Permeability, For current density, The curl operator is used to calculate the curl of a vector field. Let be the shape function of the i-th node. For node indexes in a finite element mesh, Let be the magnetic vector potential of the i-th node; The finite element equation is expressed as: in, The stiffness matrix is ​​a coefficient matrix composed of shape functions and material parameters. The nodal magnetic potential vector is a column vector composed of all nodal magnetic potential values. The load vector is the vector composed of current density excitation. The computational domain volume is the spatial region covered by the finite element model. It is a volume infinitesimal element. Let j be the shape function of the j-th node. and For node indexes in a finite element mesh, This is a transpose operation; The formulas for calculating loss density and electromagnetic force are expressed as follows: in, For loss density, For electrical conductivity, For electric field strength, The modulus of magnetic flux density. The divergence operator is used to calculate the divergence of a tensor field. For Maxwell's stress tensor, Magnetic flux density The magnetic field strength, It is the identity matrix. For transpose operation, It represents the electromagnetic force density.

[0037] S202: Solve the temperature distribution through the heat transfer equation and obtain the steady-state thermal field by combining the mechanisms of heat conduction, heat convection and heat radiation; The thermal field solution combines the thermal circuit method and the finite element method to solve the heat conduction equation, heat convection equation and heat radiation equation to obtain the steady-state temperature distribution. Among them, the heat conduction equation is based on Fourier's law, the heat convection equation uses Newton's law of cooling, and the heat radiation equation considers Stefan-Boltzmann's law. The heat conduction equation is expressed as: in, For temperature, , and Let be the thermal conductivity of the material in the x, y, and z directions. Internal heat source intensity refers to the heat generated per unit volume. For material density, Specific heat capacity is the amount of heat required to raise the temperature by one unit per unit mass of a substance. For time, The rate of change of temperature over time. , and This represents the rate of change of temperature along the x, y, and z directions; The heat convection equation is expressed as: in, For convective heat transfer, heat flux density is the amount of heat transferred per unit time through a unit area. The surface convective heat transfer coefficient is... For the temperature of the hot fluid, This refers to the temperature of the cold surface. The thermal radiation equation is expressed as: in, Radiative heat flow is the heat flow transferred through radiation. The Stefan-Boltzmann constant is... and The temperatures of the two radiating surfaces, and Let the emissivity of the two surfaces be... and Let be the radii of the two surfaces.

[0038] S203: Calculates mechanical response through equivalent dynamic model, solves displacement and stress distribution based on external force input, and outputs multi-physics coupling results; The mechanical response solution is based on the mass-spring-damping equivalent model. The radial and axial electromagnetic forces on the winding disc are calculated according to the short-circuit impact current. An equivalent dynamic model is constructed by combining material properties and the partitioning of the disc. The dynamic response equation is solved to obtain displacement and stress indices, thereby realizing the prediction of short-circuit deformation of the winding. The mechanical response calculated using the equivalent dynamic model is expressed as follows: in, The system's quality matrix, The damping matrix of the system refers to the various energy dissipation mechanisms present in the transformer structure. The system's stiffness matrix is ​​composed of the equivalent spring stiffness between the winding coils (such as spacers or insulating paperboard) and the supporting structure. Let be the system's displacement vector. Let u be the first derivative of the displacement vector u with respect to time (velocity vector). Let u be the second derivative of the displacement vector u with respect to time (acceleration vector). Let be the vector of external forces acting on the system.

[0039] In an optional implementation, in step S200, the multiphysics coupling simulation further includes dividing the transformer region into a regular grid, directly solving the coupled electromagnetic-thermal-mechanical partial differential equations through differential approximation, and using implicit iteration in alternating directions to ensure numerical stability.

[0040] In another optional implementation, in step S200, the multiphysics coupling simulation may further include performing physical field calculations based on nodes rather than meshes, constructing shape functions by moving least squares, and optimizing multiphysics coupling parameter matching using particle swarm optimization.

[0041] Furthermore, in this embodiment of the application, in step S200, the intelligent classification model includes steps S211~S213: S211: Construct 200 decision trees with a maximum depth of 10, and ensemble them using a Bagging strategy. For each type, calculate the discriminant function output, expressed as: in, The output value of the discriminant function for the l-th fault category. Index for fault categories, The input feature vector is the preprocessed amplitude-frequency curve data. This is the weight vector corresponding to the l-th fault category. For transpose operation, For kernel function mapping, This is the bias term for the l-th fault category.

[0042] S212: Minimize the Gini coefficient during training to split nodes, and use an 8:2 ratio to divide the training set and validation set.

[0043] S213: Input a standardized frequency response feature vector, output the probability distribution of four fault types: axial displacement, radial deformation, inter-panel spacing variation, and inter-panel short circuit. A voting mechanism is used to determine the final fault category and calculate the confidence level, expressed by the following formula: in, Let be the predicted probability that data o belongs to the l-th fault category. It is a natural exponential function. The slope calibration parameter for the l-th fault category, The output value of the discriminant function for the l-th fault category. The location calibration parameters are for the l-th fault category.

[0044] In an optional implementation, in step S200, the intelligent classification model further includes training a binary classifier for every two fault categories, mapping features to a high-dimensional space using a radial basis function to find the optimal hyperplane, and using the Platt scaling method to convert the classification results into probability outputs.

[0045] In another optional implementation, in step S200, the intelligent classification model may further include constructing a network structure containing convolutional layers and fully connected layers, automatically extracting local features of the frequency response curve through the convolutional layers, implementing fault classification through the fully connected layers, using Dropout to prevent overfitting, and outputting the probability distribution using Softmax.

[0046] In this embodiment of the application, step S300, the interactive visualization rendering includes steps S301 to S303: S301: Electromagnetic, thermal, and mechanical results are used to create three-dimensional scalar and vector fields. The temperature field is displayed using a hybrid method of surface rendering, section rendering, and isosurface rendering, and color mapping and labeling are performed according to the numerical range.

[0047] S302: Displays the fault categories and confidence levels of the inversion diagnosis output using curves and tables, and supports the comparison of normal and fault curves of FRA; The rendering process employs multi-level detail technology, which divides large-scale mesh data into blocks and dynamically adjusts the level of detail based on the viewpoint distance.

[0048] S303: An acceleration strategy based on chunked / pyramid multi-level detail (LOD) and rendering cache to ensure frame rate and rendering performance of large-scale meshes, and enables off-screen frame buffering and texture reuse to reuse the previous frame's results to reduce GPU load when the camera field of view changes less than a threshold.

[0049] In an optional implementation, in step S300, the interactive visualization rendering further includes using the three-dimensional physical field data as volume data, directly generating the visualization result of the internal structure through a ray casting algorithm, and using a transfer function to highlight key numerical ranges.

[0050] In another alternative implementation, in step S300, the interactive visualization rendering may further include using a large number of particles to trace the distribution of electromagnetic force and thermal convection, characterizing the field strength and direction through particle trajectories, colors and lengths, and combining them with time-series animation to display the dynamic process.

[0051] Furthermore, the distributed service management includes steps S311-S313: S311: Provides a unified service entry point through the API gateway, performs transport layer secure communication and authentication and access control based on JSON WebToken, and verifies user identity and permissions.

[0052] S312: Asynchronous communication between services is achieved through persistent message queues. The message body carries an idempotent identifier, version number, and time stamp to ensure idempotency of message processing and end-to-end traceability.

[0053] S313: Service calls are decoupled through message queues, supporting fault recovery and retry mechanisms.

[0054] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0055] Example 3, referring to Figure 2 This is the third embodiment of the present invention. This embodiment provides a power transformer operation status evaluation system based on a microservice architecture, including a data access and preprocessing module, a forward simulation module, an inverse diagnosis module, and a visualization and reporting module.

[0056] The data access and preprocessing module is used to align frequency response data by establishing a common frequency grid, fill in missing data points using interpolation methods, and perform feature standardization operations to transform raw data from different sources and with different dimensions into standardized feature data with a unified format.

[0057] The forward modeling module is used to solve the electromagnetic field distribution, temperature field distribution, and mechanical response of the transformer based on preprocessed data through microservice-based finite element calculation, thermal circuit analysis, and equivalent dynamic model, and to perform coupled simulation of electromagnetic-thermal-mechanical multiphysics fields.

[0058] The inversion diagnosis module is used to input standardized feature vectors into a trained machine learning classification model, and intelligently identify and classify the winding state through an integrated decision tree, outputting the fault type and confidence level.

[0059] The visualization and reporting module is used to perform three-dimensional visualization of multiphysics results using a hybrid rendering strategy, and to use multi-level detail technology and rendering caching mechanism for data interaction, transforming simulation and diagnostic results into intuitive graphics and reports.

[0060] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0061] Example 4, the fourth embodiment of the present invention, differs from the previous three embodiments in that: If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0062] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0063] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0064] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

Claims

1. A method for evaluating the operating status of power transformers based on a microservice architecture, characterized by: include, Collect multi-source operation data of power transformers, perform unified processing on the multi-source operation data, and construct standardized feature vectors; Based on multiphysics coupling simulation, the field distribution and mechanical response of power transformers are calculated. The standardized feature vectors are input into the intelligent classification model to classify winding deformation faults and output the fault type and confidence level. Interactive visualization rendering of simulation results and intelligent classification results is performed, and fault curves are compared and displayed. Distributed service management is carried out using containerized microservices.

2. The method for evaluating the operating status of power transformers based on a microservice architecture as described in claim 1, characterized in that: The unification process includes selecting a common frequency grid, aligning the frequency domain observation data with frequencies, and mapping the frequency points of different samples to a unified grid. Missing points in the frequency response data are detected and filled using interpolation methods to obtain a continuous frequency sequence. The amplitude characteristics of the filled data are then standardized.

3. The method for evaluating the operating status of power transformers based on a microservice architecture as described in claim 2, characterized in that: The multiphysics coupling simulation includes calculating the magnetic induction intensity distribution, loss density, and electromagnetic force through electromagnetic field simulation, calculating the steady-state temperature distribution through thermal field simulation, and calculating the displacement and stress response under short-circuit impact through mechanical simulation.

4. The method for evaluating the operating status of power transformers based on a microservice architecture as described in claim 3, characterized in that: The intelligent classification model includes inputting feature vectors into a multi-class classifier, calculating the discriminant function output for each class, and distinguishing different fault types. The discrimination output is converted into a confidence level of the fault category through a probability mapping function, and an interpretable diagnostic basis is provided.

5. The power transformer operation status evaluation method based on microservice architecture as described in claim 4, characterized in that: The calculation of the field distribution and mechanical response of the power transformer includes solving the electromagnetic field using the finite element method, dividing the computational domain into elements by establishing the magnetic vector potential control equation, selecting the shape function as the magnetic vector potential in each element, substituting it into the control equation, executing the weighted residual method to obtain the finite element equation, applying boundary conditions, solving the sparse linear equation system to obtain the nodal magnetic vector potential value, and calculating the loss density and electromagnetic force. The governing equations of the magnetic vector potential are expressed as follows: in, The magnetic vector potential, Permeability, For current density, The curl operator is used to calculate the curl of a vector field. Let be the shape function of the i-th node. For node indexes in a finite element mesh, Let be the magnetic vector potential of the i-th node; The finite element equation is expressed as: in, The stiffness matrix is ​​a coefficient matrix composed of shape functions and material parameters. The nodal magnetic potential vector is a column vector composed of all nodal magnetic potential values. The load vector is the vector composed of current density excitation. The computational domain volume is the spatial region covered by the finite element model. It is a volume infinitesimal element. Let j be the shape function of the j-th node. and For node indexes in a finite element mesh, This is a transpose operation; The formulas for calculating loss density and electromagnetic force are expressed as follows: in, For loss density, For electrical conductivity, For electric field strength, The modulus of magnetic flux density. The divergence operator is used to calculate the divergence of a tensor field. For Maxwell's stress tensor, Magnetic flux density The magnetic field strength, It is the identity matrix. For transpose operation, Electromagnetic force density; The thermal field is solved by combining the thermal circuit method and the finite element method, and the heat conduction equation, heat convection equation and heat radiation equation are solved to obtain the steady-state temperature distribution. Among them, the heat conduction equation is based on Fourier's law, the heat convection equation uses Newton's law of cooling, and the heat radiation equation considers Stefan-Boltzmann's law. The mechanical response is solved based on the mass-spring-damping equivalent model. The radial and axial electromagnetic forces on the winding disc are calculated according to the short-circuit impact current. An equivalent dynamic model is constructed by combining material properties and the partitioning of the disc. The displacement and stress indices are obtained by solving the dynamic response equation.

6. The method for evaluating the operating status of power transformers based on a microservice architecture as described in claim 5, characterized in that: The classification of winding deformation faults includes using a random forest model, dividing the samples into training and validation sets proportionally during training, constructing decision trees by minimizing the Gini coefficient, setting tree parameters, using a Bagging strategy to fuse the results of each tree, outputting the fault category as the voting result, and the confidence level as the average of the output probabilities of all trees. The training data includes frequency response curves acquired through simulation and experiments, covering normal, axial displacement, radial deformation, inter-pie spacing variation, and inter-pie short-circuit fault types.

7. The method for evaluating the operating status of power transformers based on a microservice architecture as described in claim 6, characterized in that: The interactive visualization rendering includes drawing three-dimensional scalar and vector fields of electromagnetic, thermal, and mechanical results; displaying the temperature field using a mixed method of surface rendering, section rendering, and isosurface rendering; color mapping and labeling according to numerical range; supporting the superposition and comparison of normal and fault curves in frequency response analysis; displaying curve differences through a graphical interface; and generating statistical tables to compare key indicators. The rendering process employs multi-level detail technology, which divides large-scale mesh data into blocks and dynamically adjusts the level of detail based on the viewpoint distance. The interactive visualization rendering also includes using an off-screen frame buffer and texture reuse mechanism to directly reuse the rendering result of the previous frame when the change in the camera field of view is less than a set threshold, and storing scene data through a rendering cache. The distributed service management includes providing a unified service entry point through an API gateway, performing transport layer secure communication and JSON Web Token-based authentication and access control, verifying user identity and permissions, and conducting asynchronous inter-service communication through a persistent message queue. The message body carries an idempotent identifier, version number, and time stamp. Inter-service calls are decoupled through the message queue for fault recovery and retries.

8. A power transformer operation status evaluation system based on microservice architecture, employing the power transformer operation status evaluation method based on microservice architecture as described in any one of claims 1 to 7, characterized in that, It includes a data access and preprocessing module, a forward simulation module, an inversion diagnosis module, and a visualization and reporting module; The data access and preprocessing module is used to align frequency response data by establishing a common frequency grid, fill in missing data points using interpolation methods, and perform feature standardization operations to transform raw data from different sources and with different dimensions into standardized feature data with a unified format. The forward modeling module is used to solve the electromagnetic field distribution, temperature field distribution and mechanical response of the transformer based on the preprocessed data through microservice-based finite element calculation, thermal circuit analysis and equivalent dynamic model, and to perform coupled simulation of electromagnetic-thermal-mechanical multiphysics fields. The inversion diagnosis module is used to input standardized feature vectors into a trained machine learning classification model, and intelligently identify and classify the winding state through an integrated decision tree, outputting the fault type and confidence level. The visualization and reporting module is used to perform three-dimensional visualization of multiphysics results using a hybrid rendering strategy, and to use multi-level detail technology and rendering caching mechanism for data interaction, transforming simulation and diagnostic results into intuitive graphics and reports.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the power transformer operation status evaluation method based on microservice architecture as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the power transformer operation status evaluation method based on microservice architecture as described in any one of claims 1 to 7.

Images