A method and system for hot spot temperature inversion in power transformers

Abstract

This invention belongs to the field of power transformer fault diagnosis technology, and relates to a method and system for hot spot temperature inversion in power transformers. The system uses a data acquisition module to collect real-time operating status data of the power transformer, generating multi-source data. A data processing module processes the multi-source data to generate multi-source feature data. A high-fidelity multiphysics simulation model of the power transformer is constructed to generate a reliable sample dataset. This reliable sample dataset is used to train an artificial neural network model until convergence, resulting in a hot spot temperature inversion model. The multi-source feature data is input into the trained hot spot temperature inversion model for real-time inference calculations, outputting the current hot spot temperature value of the internal windings of the power transformer and the corresponding operating status assessment result. In the human-computer interaction visualization interface on the host computer, the hot spot temperature value, temperature distribution cloud map, and operating status assessment result are rendered and displayed in real-time based on the 3D transformer model.

Description

Technical Field

[0001] This invention relates to the field of power transformer fault diagnosis technology, specifically to a hot spot temperature inversion method and system applied to power transformers. Background Technology

[0002] As a core device for energy transmission and transformation in power grids, the internal hot spot temperature of power transformers is a key indicator reflecting the insulation aging and operational health of the equipment. Real-time temperature monitoring of transformers faces severe challenges, especially in remote and complex environments such as mines, isolated islands, and mountainous areas. Current technologies typically rely on manual periodic inspections to read instrument data or wired sensor networks for data acquisition. While these methods can obtain some operational status information, manual methods suffer from data lag and poor real-time performance. Wired methods are difficult to implement in complex environments, have high maintenance costs, and are susceptible to environmental interference, making it difficult to achieve continuous and stable uploading of operational status information. Furthermore, some methods propose combining physical field simulation with neural network algorithms. By establishing a high-fidelity multiphysics simulation model of the transformer to obtain reliable samples under different operating conditions, and using a backpropagation neural network to construct a hot spot temperature inversion model for dynamic calculation of hot spot temperature, this approach achieves higher computational accuracy compared to traditional thermal models and thermal circuit models. The above scheme still has the following problems: When constructing the hot spot temperature inversion model, the existing methods focus on conventional thermal and electrical parameters such as ambient temperature, load factor, and tank surface temperature, without considering the power transformer oil chromatographic data parameters that can reflect internal latent faults and overheating conditions.

[0003] In view of this, it is very necessary to provide a hot spot temperature inversion method and system for power transformers to solve the above-mentioned defects in the prior art. Summary of the Invention

[0004] The purpose of this invention is to solve the technical problem that existing hot spot temperature inversion models do not consider the parameters of power transformer oil chromatography data that can reflect internal latent faults and overheating states. The invention provides a hot spot temperature inversion method and system for power transformers to solve the technical problems existing in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, this application provides a hot spot temperature inversion method for power transformers, comprising the following steps: Step S1: Construct a wireless communication network to collect real-time transformer operating status data and generate multi-source data by deploying sensor nodes and environmental monitoring nodes at key locations of the power transformer. Step S2: The collected multi-source data is transmitted to the host computer in real time through the wireless communication network. The host computer processes the received multi-source data to obtain multi-source feature data. Step S3: Construct a high-fidelity multiphysics simulation model of a power transformer. Simulate the model by changing the boundary conditions and load conditions to generate simulation sample data covering different operating conditions, which serves as a reliable sample dataset. Use the reliable sample dataset to train the artificial neural network model offline to obtain a hotspot temperature inversion model. Step S4: Input the multi-source feature data as the input feature vector into the hot spot temperature inversion model, perform real-time inference calculation, and invert the current hot spot temperature value of the internal winding of the power transformer and the corresponding operating status evaluation result. Step S5: In the human-computer interaction visualization interface of the host computer, the hot spot temperature value, temperature distribution cloud map and operation status evaluation results calculated by inversion are rendered and displayed in real time based on the three-dimensional transformer model, and an early warning prompt is triggered for abnormal status.

[0006] Furthermore, the present invention also provides a hot spot temperature inversion system for power transformers, comprising: The data acquisition module contains: A wireless communication network is constructed, and various sensor nodes and detection devices are deployed in key parts of the power transformer and its surrounding environment to collect real-time power transformer operating status data and generate multi-source data. The data processing module contains: The system receives multi-source data collected by the data acquisition module through a wireless communication network, processes the multi-source data, and generates multi-source feature data. The model training module contains: A high-fidelity multiphysics simulation model of a power transformer is constructed. A reliable sample dataset is generated based on the high-fidelity multiphysics simulation model of the power transformer. The artificial neural network model is trained using the reliable sample dataset until convergence, and a hotspot temperature inversion model is obtained. The hotspot temperature inversion module contains: Multi-source feature data is used as input feature vectors to input the trained hot spot temperature inversion model, and real-time inference calculation is performed to output the current hot spot temperature value of the internal winding of the power transformer and the corresponding operating status evaluation result. The visual interactive display module includes: In the human-computer interaction visualization interface of the host computer, the hot spot temperature value, temperature distribution cloud map and operation status evaluation results are rendered and displayed in real time based on the three-dimensional transformer model, and an early warning prompt is triggered when the operation status is abnormal.

[0007] The beneficial effects of this invention are as follows: This invention achieves real-time acquisition and stable transmission of transformer operating status data by deploying sensor nodes and environmental monitoring nodes at key locations of power transformers and constructing a data acquisition and transmission link in conjunction with a wireless communication network. This solves the problems of unstable data transmission and difficulty in real-time, automatic, and stable transmission to the host computer in remote and complex environments, as well as the complex wiring and maintenance difficulties inherent in traditional wired networks. It is suitable for complex environments or scenarios where wiring is inconvenient, such as mines, isolated islands, and mountainous areas. By incorporating power transformer oil chromatography data parameters into the hotspot temperature inversion model, it can reflect latent faults and overheating conditions inside the transformer, improving the sensitivity and accuracy of the hotspot temperature inversion model to abnormal states. This solves the problems of hotspot temperature prediction deviations and insufficient fault identification capabilities caused by the failure to consider power transformer oil chromatography data in existing technologies.

[0008] This invention transmits and processes collected multi-source data to generate multi-source feature data, which is then input into a trained hotspot temperature inversion model for real-time inference calculation. This enables rapid inversion of hotspot temperature values ​​and operating status assessment results of internal windings of power transformers, improving the real-time performance and accuracy of hotspot temperature calculation and status assessment, and solving the problem of the inability to perform hotspot temperature inversion and fault diagnosis in a timely manner.

[0009] This invention improves the reliability and generalization ability of the hotspot temperature inversion model by constructing a high-fidelity multiphysics simulation model of a power transformer to generate a reliable sample dataset and training the artificial neural network model offline. Furthermore, this invention provides real-time rendering and display of hotspot temperature values, temperature distribution cloud maps, and operational status assessment results based on a 3D transformer model within a host computer human-computer interactive visualization interface. It also triggers early warning prompts when abnormal operational status occurs, enhancing the visualization capability and anomaly identification efficiency of the transformer's operational status, and improving the timeliness and intuitiveness of operational monitoring and fault early warning.

[0010] Therefore, it is evident that the present invention has outstanding substantive features and significant progress compared with the prior art, and the beneficial effects of its implementation are also obvious. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0012] Figure 1 This is a flowchart of a hotspot temperature inversion method applied to power transformers; Figure 2 This is a schematic diagram of a hot spot temperature inversion system applied to power transformers; Figure 3 This is a flowchart of a hotspot temperature inversion method applied to power transformers.

[0013] The module consists of 1-data acquisition module, 2-data processing module, 3-model training module, 4-hotspot temperature inversion module, and 5-visual interactive display module. Detailed Implementation

[0014] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The following embodiments are explanations of the present invention, but the present invention is not limited to the following implementation methods.

[0015] Example 1: like Figure 1 and Figure 3 As shown, this embodiment provides a hotspot temperature inversion method for power transformers. This method can collect and transmit multi-source data on the operating status of power transformers. Through temperature sensor nodes, ambient temperature and humidity detection nodes, current sensors, voltage sensors, and transformer oil chromatography analysis equipment deployed on the power transformer body and its surrounding environment, multi-source data including the temperature of characteristic points on the power transformer surface, ambient temperature and humidity, winding current, terminal voltage, and the composition and content of dissolved gases in the oil are acquired. The multi-source data is then transmitted to a host computer for processing and analysis via a wireless communication network. The hotspot temperature inversion method for power transformers includes the following steps: Step S1: Construct a wireless communication network to collect real-time transformer operating status data and generate multi-source data by deploying sensor nodes and environmental monitoring nodes at key locations of the power transformer. Step S1 specifically includes: S11. Connect the host computer and the relay router via wired or wireless means. The relay router serves as the root node of the wireless communication network. Configure the network identification information of the router. Computer terminals or mobile terminals access the established wireless communication network through the router to perform data interaction between external terminals and the network. Deploy network nodes and sensor nodes at all levels. Data collected by the sensor nodes is relayed through network nodes at each level, forwarded by the relay router, and uploaded to the host computer.

[0016] The wireless communication network mentioned in step S11 includes a system management host, i.e., a host computer, which is used to centrally manage the wireless communication network and serve as the central node for data processing and analysis; the relay routing device is a relay node device set up in the wireless communication network, connected to the host computer, and performs data forwarding and communication path relay between network nodes at all levels, so that the data collected by each sensor node can be transmitted to the host computer level by level. The router device mentioned in step S11 is an access configuration device for a wireless communication network, used to configure the identification information of the wireless communication network, control the computer terminal or mobile terminal to access the wireless communication network, and realize data access and communication connection between the external terminal and the wireless communication network; the computer terminal or mobile terminal is used to ultimately acquire, process, and analyze the data transmitted in the wireless communication network. The wireless communication network described in step S11 also includes network nodes at various levels, which are used to receive power transformer operating status data collected by each sensor node and upload it to the host computer through relay routing equipment.

[0017] Wireless communication networks have self-organizing and self-healing capabilities. When a node in the network goes offline or the signal is interfered with, the communication path can be automatically reconstructed to ensure the stability and continuity of data transmission, thereby enabling real-time and reliable transmission of power transformer operation data.

[0018] S12. Temperature measurement node equipment, environmental information detection node equipment, and electrical information detection node equipment are deployed at sensor nodes and environmental monitoring nodes. The key parts of the power transformer are the winding surface, the core surface, and the heat dissipation area of ​​the oil tank shell. The surrounding environment is the air environment around the power transformer installation area. The power transformer operating status data is collected in real time through each sensor node, and after edge computing processing, multi-source data is generated. The multi-source data includes environmental temperature and humidity data, temperature data of characteristic points on the surface of the power transformer, electrical quantity data of the power transformer, and oil chromatography data of the power transformer. The key components of the power transformer mentioned in step S12 are used to characterize the internal temperature rise distribution of the power transformer; the surrounding environment is used to characterize the external heat dissipation conditions of the power transformer. It should be noted that by constructing a wireless communication network and deploying various sensor nodes at key locations of power transformers, real-time acquisition of multi-source data on the operating status of power transformers can be achieved, improving the comprehensiveness and timeliness of data acquisition. By adopting a collaborative deployment approach for multiple types of sensors, the operating status of power transformers can be reflected from multiple dimensions such as temperature, environment, electrical, and oil chromatography, improving the accuracy of status characterization. In addition, the construction of the wireless communication network reduces the complexity of traditional wired deployment and enhances adaptability and deployment flexibility in complex environments.

[0019] Step S2: The collected multi-source data is transmitted to the host computer in real time through the wireless communication network. The host computer processes the received multi-source data to obtain multi-source feature data. Step S2 specifically includes: S21. The multi-source data is converted into a data format recognizable by the chip device through a data conversion device, the data is encapsulated by the chip device, and transmitted by the antenna to the sub-nodes in the wireless communication network; the data is forwarded step by step through network nodes at each level, and the communication path is relayed through relay routing devices to transmit the multi-source data to the host computer. S22. Perform data processing on the multi-source data to generate multi-source feature data; the data processing operations include: The multi-source data received by the host computer is sorted according to the acquisition time, and resampled at a uniform time interval to perform time alignment of the multi-source data. The time-aligned multi-source data is then subjected to anomaly identification and removal, and missing data is filled in. Normalization transformation is performed to convert different physical quantities to a uniform numerical range and eliminate the influence of dimensions. Based on the normalized multi-source data of the power transformer's operating status, temperature characteristic parameters, environmental characteristic parameters, electrical characteristic parameters, and oil chromatography characteristic parameters are extracted to construct multi-source feature data.

[0020] It should be noted that the real-time transmission and centralized processing of multi-source data on the operating status of power transformers through wireless communication networks improves the timeliness and completeness of data acquisition and avoids the lag problems caused by manual collection. At the same time, data processing of multi-source data improves the quality and consistency of multi-source data, providing a reliable data foundation for subsequent hotspot temperature inversion. By extracting temperature characteristic parameters, environmental characteristic parameters, electrical characteristic parameters, and oil chromatographic characteristic parameters, multi-source characteristic data is constructed, enhancing the ability to characterize the operating status of power transformers and improving the accuracy and stability of hotspot temperature inversion results.

[0021] Step S3: Construct a high-fidelity multiphysics simulation model of a power transformer. Simulate the model by changing the boundary conditions and load conditions to generate simulation sample data covering different operating conditions, which serves as a reliable sample dataset. Use the reliable sample dataset to train the artificial neural network model offline to obtain a hotspot temperature inversion model. Step S3 specifically includes: S31. Based on the structural parameters of the power transformer, structural modeling of the winding, core and tank is performed to generate a high-fidelity multiphysics simulation model of the power transformer. Step S31 specifically includes: S311. An electromagnetic field calculation model is established based on Maxwell's equations to describe the current distribution in the windings of a power transformer and the electromagnetic response process of the terminal voltage.

[0022]

[0023] Where E is the electric field strength, H is the magnetic field strength, J is the current density, and B is the magnetic flux density; By solving the above equations using the finite element method, an electromagnetic field distribution model for a power transformer is constructed. S312. A temperature field model is constructed based on Fourier's law of heat conduction to characterize the internal heat diffusion process of a power transformer:

[0024] Where T is temperature, ρ is density, c is specific heat capacity, k is thermal conductivity, and Q is the heat source term; it is used to characterize the temperature change process of the winding, core, and tank structure of a power transformer.

[0025] S313. Construct an insulating oil fluid heat transfer model based on the Navier-Stokes equations to describe the oil circulation and convective heat transfer processes:

[0026]

[0027] Where v is the flow rate, p is the pressure, and μ is the dynamic viscosity; it is used to characterize the circulating heat dissipation process of insulating oil inside the tank.

[0028] S314. Couple the electromagnetic field calculation model, the heat conduction calculation model, and the fluid heat transfer calculation model to construct a multiphysics coupling relationship:

[0029]

[0030]

[0031] Among them, electromagnetic loss generates a heat source term Q, and the temperature field response T is affected by both the heat source and the fluid velocity field. A high-fidelity multiphysics simulation model of a power transformer is generated by performing coupled calculations of the electromagnetic field, temperature field, and fluid heat transfer field through iterative solution.

[0032] By constructing a high-fidelity multiphysics simulation model of a power transformer and performing multiphysics coupling calculations based on electromagnetic field, temperature field, and fluid heat transfer field, a unified characterization of the electromagnetic energy conversion and heat transfer process inside the power transformer is achieved. This provides a high-fidelity physical basis for the subsequent generation of simulation data and improves the physical consistency and calculation accuracy of the simulation data.

[0033] In existing technologies, heat transfer within oil-immersed transformers primarily relies on the natural convection of oil, transferring heat from the windings and core surfaces to the tank surface. The oil flow distribution must strictly satisfy the Navier-Stokes equations for a complete flow field solution, accurately characterizing the velocity, pressure, and temperature distribution of the oil flow. While the Navier-Stokes equations are complex and computationally intensive, accurately reflecting the fluid heat transfer field, they are difficult to directly generate multi-condition simulation data suitable for neural network training. This invention addresses this by constructing a simplified Navier-Stokes equations model for insulating oil fluid heat transfer. Simultaneously, it combines electromagnetic-thermal coupling calculations, generating multi-condition simulation data by adjusting boundary conditions and load conditions. Compared to existing technologies, this invention reduces the number of formulas and simplifies their expression, effectively reducing simulation computational complexity and accelerating the generation of multi-condition samples. While ensuring physical realism, it facilitates offline training with neural networks to generate reliable samples. By combining multi-physics coupling and intelligent model training, it enables rapid inversion of transformer hotspot temperatures, providing a foundation for operational status assessment and visualization.

[0034] S32. Based on the high-fidelity multiphysics simulation model of power transformer, multiphysics simulation calculations are performed on power transformer by changing boundary conditions such as ambient temperature and humidity and load conditions. Based on the coupled solution results of electromagnetic field, temperature field and fluid heat transfer field, simulation sample data under different operating conditions are obtained and used as a reliable sample dataset. Step S32 specifically includes: S321. Apply environmental temperature boundary conditions and environmental humidity boundary conditions to the high-fidelity multiphysics simulation model of the power transformer, so that they act on the temperature field model and the insulating oil fluid heat transfer model and are transferred to the coupled field solution process through the boundary conditions; by solving the temperature field model and the insulating oil fluid heat transfer model, the response of the external environment temperature and humidity of the power transformer is obtained, and environmental temperature and humidity data are generated. S322. In the high-fidelity multiphysics simulation model of power transformer, a three-dimensional structural model of winding, core and tank is established. Coupled simulation is carried out through multiphysics coupling relationship to obtain the surface heat distribution results of power transformer and extract the temperature data of characteristic points on the surface of power transformer. S323. Apply different load rise and fall conditions parameters to the high-fidelity multiphysics simulation model of the power transformer, change the load current and terminal voltage conditions parameters, and simulate to obtain the winding current distribution and terminal voltage change results, and generate power transformer electrical quantity data. S324. Based on the temperature distribution results output by the temperature field calculation model in the high-fidelity multiphysics simulation model of power transformer, and combined with the pyrolysis characteristic parameters of oil-paper insulation and insulating oil, the simulation feature mapping calculation of the insulating oil pyrolysis process is performed to obtain the evolution characteristics of dissolved gas in the oil and generate power transformer oil chromatographic data. S325. The environmental temperature and humidity data, power transformer surface feature point temperature data, power transformer electrical quantity data, and power transformer oil chromatography data obtained in steps S321 to S324 are summarized and integrated to generate a simulation sample dataset covering different operating conditions, and this dataset is used as a reliable sample dataset.

[0035] By applying environmental temperature, humidity, and load increase / decrease parameters to a high-fidelity multiphysics simulation model, a full-condition simulation of the state change process of a power transformer under multiple operating conditions is achieved, generating multi-dimensional simulation sample data containing temperature and humidity characteristics, electrical quantity characteristics, and oil chromatographic characteristics, thereby improving the richness and reliability of the simulation sample data.

[0036] The existing technology uses high-fidelity finite element simulation of single-phase oil-immersed self-cooled transformers. Through fine geometric modeling, multi-layer insulation materials, structured and unstructured mesh generation, and structural features such as oil channel supports and heat sinks, combined with transient multiphysics field solving, the dynamic inversion of hot spot temperature is achieved. The simulation results are close to the actual heat distribution, but the calculation is complex, the mesh is fine, and it takes a long time to generate training samples covering multiple loads and environmental conditions.

[0037] Compared with existing technologies, this invention, based on existing technologies, employs environmental temperature and humidity boundary conditions to couple environmental changes with load conditions for unified solution, enabling rapid simulation of transformer states under multiple operating conditions. The introduction of environmental temperature and humidity boundaries is because transformer tank temperature and hot spot temperature are affected by external environmental temperature and humidity; considering these conditions improves the realism and applicability of the simulation samples. This invention can quickly generate multi-dimensional simulation datasets covering all operating conditions. Compared with existing technologies, this invention improves the efficiency of simulation sample data generation, reduces computational costs, and simultaneously ensures the physical authenticity of the simulation data and the integrity of hot spot temperature information, providing a reliable data foundation for offline neural network training, real-time hot spot temperature inversion, and operational status monitoring and risk warning.

[0038] S33. Construct an artificial neural network, train the artificial neural network using a reliable sample dataset, and iteratively optimize the network parameters through backpropagation and gradient descent methods to enable the model to gradually learn the nonlinear mapping relationship between the running state characteristics and hotspot temperatures. During the training process, the model training state is judged by setting the convergence threshold of the loss function and the error threshold of the validation set. After the convergence condition is met, the trained hotspot temperature inversion model is obtained. Step S33 specifically includes: S331. Construct an artificial neural network model for hot spot temperature inversion of power transformers. The artificial neural network model adopts a multi-layer feedforward neural network structure, including an input layer, a hidden layer, and an output layer. The input layer receives feature data from a trusted sample dataset, including ambient temperature and humidity data, surface feature point temperature data of the power transformer, electrical quantity data of the power transformer, and oil chromatography data of the power transformer. The hidden layer adopts a fully connected network structure and introduces a nonlinear activation function to enhance the fitting ability of complex nonlinear mapping relationships, and is used for feature fusion and high-dimensional mapping of the feature data. The output layer outputs the predicted hot spot temperature of the power transformer. This multi-layer feedforward neural network structure is used to model the nonlinear mapping between the operating state characteristics of power transformers and hot spot temperatures, so as to meet the computational needs of rapid hot spot temperature inversion under multiple operating conditions.

[0039] S332. Divide the credible sample dataset into a training set and a validation set, input the training set into the artificial neural network model, and use the hotspot temperature output by the high-fidelity multiphysics simulation model of the power transformer as the supervised learning label; A loss function is defined to quantify the model output error; the loss function is the mean squared error function. This is used to measure the deviation between the predicted hotspot temperature and the simulated hotspot temperature. This represents the predicted hotspot temperature output by the artificial neural network. The value represents the hotspot temperature calculated by the high-fidelity multiphysics simulation model of the power transformer, and N represents the number of samples.

[0040] The gradient of the loss function with respect to the network weights is calculated by backpropagation algorithm, and the network parameters are iteratively updated by gradient descent optimization method, so that the artificial neural network model can gradually learn the mapping relationship between the hot spot temperature of the power transformer and various simulation feature data under simulation conditions. During the training process, reliable simulation sample data under different operating conditions are continuously input to enhance the model's ability to fit the coupling relationship of multiple physics fields and the changing conditions of multiple operating conditions. Through multiple iterations of training, the artificial neural network model gradually learns the hot spot temperature variation patterns of power transformers under different operating conditions. S333. Evaluate the trained artificial neural network model based on the validation set, and stop training when any of the following convergence conditions are met: (1) The loss function converges to the preset threshold ε, that is, the change of the loss function is less than ε in several consecutive iterations; (2) The mean square error on the validation set is less than the preset error threshold δ; (3) The rate of decrease in model prediction error tends to stabilize and no longer improves significantly.

[0041] The preset error threshold δ is determined based on the statistical range of the output error of the simulation model and the historical modeling accuracy requirements; Once the convergence condition is met, the trained artificial neural network model will be used as the hot spot temperature inversion model for power transformers, and will be used for real-time hot spot temperature calculation and inversion analysis under the operating conditions of power transformers.

[0042] By supervising the training of credible sample data generated based on multiphysics simulation, the artificial neural network model is enabled to autonomously learn the temperature change patterns of power transformer hotspots, achieving an effective conversion from physical simulation to data-driven model and improving the real-time performance and accuracy of hotspot temperature inversion.

[0043] Compared to existing technologies, which primarily rely on high-fidelity multiphysics simulation samples to train neural networks for dynamic hotspot temperature inversion, the loss function can lead to gradient amplification when the sample size is large. This makes it difficult for the weight update magnitude to match the sample size during training, thus affecting convergence speed and training stability. This invention addresses this issue by averaging the total error, ensuring a balanced contribution of each sample's error to weight updates. This reduces the impact of sample size variations on the training process, improving gradient stability and convergence efficiency.

[0044] It should be noted that by constructing a high-fidelity multiphysics simulation model of power transformers and introducing a coupling calculation mechanism of electromagnetic, temperature, and fluid fields, the simulation process can realistically reproduce the thermal-electric-current coupling evolution law inside the transformer under different operating conditions, obtaining simulation sample data with high consistency and high reliability. Based on this simulation sample data, an artificial neural network is trained offline, improving the learning and generalization ability of the artificial neural network model to complex nonlinear mapping relationships, thereby enhancing the reliability of online monitoring and fault early warning of power transformers.

[0045] Step S4: Input the multi-source feature data as the input feature vector into the hot spot temperature inversion model, perform real-time inference calculation, and invert the current hot spot temperature value of the internal winding of the power transformer and the corresponding operating status evaluation result. Step S4 specifically includes: S41. Input the multi-source feature data as the input feature vector into the trained hot spot temperature inversion model, and perform layer-by-layer mapping and feature fusion on the input feature vector through forward propagation calculation to output the current hot spot temperature value of the internal winding of the power transformer. Step S41 in the forward propagation calculation process is as follows: after the input feature vector is mapped to the hidden layer through the input layer, it is linearly transformed by the weight matrix and the bias parameter, and after being processed by the nonlinear activation function, it is passed to the output layer layer by layer to obtain the hotspot temperature value.

[0046] By inputting multi-source feature data into a trained hotspot temperature inversion model and performing forward propagation calculations, a layer-by-layer nonlinear mapping and feature fusion of transformer operating state characteristics is achieved, thereby enabling rapid output of power transformer hotspot temperature values ​​and improving the real-time performance and computational efficiency of hotspot temperature inversion.

[0047] S42. Compare the output hot spot temperature value with the preset operating threshold, and obtain the operating status assessment result according to the classification rules; output the hot spot temperature value and the corresponding operating status assessment result to the host computer for power transformer operating status monitoring, load assessment and fault early warning decision support.

[0048] The grading criteria for the operational status assessment results in step S42 are as follows: When the hotspot temperature value is lower than the first preset threshold, it is determined to be in normal operating condition; When the hot spot temperature value is between the first preset threshold and the second preset threshold, it is determined to be a slight overload state; When the hot spot temperature value is higher than the second preset threshold but lower than the ultimate safety threshold, it is determined to be a severe overload state. When the temperature value of a hot spot reaches or exceeds the limit safety threshold, it is determined to be an abnormal warning state; The first preset threshold in step S42 The temperature range of 90℃ to 105℃ is used to characterize the upper limit of the temperature at which the transformer can operate stably for a long period of time. Second preset threshold The temperature is set to 110℃~130℃ to characterize the risk threshold temperature for transformer overload operation. Limit safety threshold The temperature is set to 140℃~160℃ to characterize the critical temperature for thermal aging and failure risk of insulating materials.

[0049] By comparing hot spot temperature values ​​with preset operating thresholds in a tiered manner, the operating status of power transformers can be quantitatively assessed and graded, thereby improving the ability to identify and warn of operational risks.

[0050] It should be noted that by inputting multi-source feature data into the hotspot temperature inversion model and performing real-time inference calculations, the hotspot temperature of the power transformer can be inverted, accurately characterizing the thermal operating state of its internal windings. Combined with a threshold-based hierarchical evaluation mechanism, real-time monitoring and intelligent early warning of the transformer's operating status are achieved, improving the accuracy and applicability of power transformer operating status assessment.

[0051] Step S5: In the human-computer interaction visualization interface of the host computer, the hot spot temperature value, temperature distribution cloud map and operation status evaluation results calculated by inversion are rendered and displayed in real time based on the three-dimensional transformer model, and an early warning prompt is triggered for abnormal status.

[0052] Step S5 specifically includes: S51. A three-dimensional transformer model is established based on the structural parameters of the power transformer to reflect the spatial structural relationship between the transformer windings, core and tank, serving as a visualization carrier for operating status data; the three-dimensional transformer model is a display model, used only for visualization and not involved in the simulation calculation of electromagnetic field, temperature field and fluid field; In the 3D transformer model, a display mapping area corresponding to the actual temperature measurement nodes and key structural locations is preset to carry the dynamic rendering of hot spot temperature values ​​and temperature distribution data. By establishing a three-dimensional transformer model that corresponds to the actual structure of the transformer and pre-setting the mapping relationship between temperature measurement nodes and key structures, the hot spot temperature and spatial location can be accurately correlated, improving the intuitiveness and positioning accuracy of temperature visualization.

[0053] S52. Input the obtained hotspot temperature values ​​and operating status evaluation results into the human-computer interaction visualization interface, and map them onto the 3D transformer model for real-time rendering and display, specifically including: The hot spot temperature values ​​are mapped to the corresponding winding hot spot areas in the 3D transformer model and displayed in the form of numerical labels; Based on the hot spot temperature values, a temperature distribution cloud map is generated on the surface of the three-dimensional transformer model, and the temperature field distribution is reflected by the color gradient change. The operational status assessment results are mapped to corresponding status identifiers, and the overall status is displayed in the 3D transformer model using color differentiation. Normal operation is indicated by a green display; slight overload is indicated by a yellow display; severe overload is indicated by an orange display; and abnormal warning is indicated by a red display. When the operational status assessment result reaches the heavy overload state or abnormal warning state, a warning prompt is triggered in the human-machine interaction visualization interface. The warning prompt includes interface highlighting, pop-up prompts and alarm information output, which are used to prompt technicians to take corresponding measures in a timely manner. Simultaneously, the hot spot temperature values ​​and operating status assessment results are output to the host computer monitoring interface to support power transformer operating status monitoring, load regulation, and fault early warning decision-making.

[0054] By mapping hotspot temperature values ​​to operational status assessment results in a hierarchical manner and rendering them with color gradients, the thermal distribution and operational risk status of transformers can be presented intuitively, improving the efficiency of anomaly identification and the perception capabilities of maintenance personnel.

[0055] It should be noted that by constructing a real-time visualization mechanism based on a 3D transformer model, a multi-dimensional integrated presentation of hotspot temperature, temperature distribution, and operating status is achieved, making the expression of operating status more intuitive and clear. Combined with dynamic rendering and a hierarchical early warning mechanism, alarm information can be triggered in a timely manner when abnormal conditions occur, improving fault response speed and enhancing the visualization level and intelligent decision support capabilities of power transformer operation monitoring.

[0056] Example 2: like Figure 2 As shown in the figure, this embodiment provides a hot spot temperature inversion system for power transformers, comprising: Data acquisition module 1, in which: A wireless communication network is constructed, and various sensor nodes and detection devices are deployed in key parts of the power transformer and its surrounding environment to collect real-time power transformer operating status data and generate multi-source data.

[0057] The wireless communication network includes a host computer, relay routing devices, router devices, and network nodes at various levels. The host computer and relay routing devices are connected via wired or wireless means. The relay routing devices serve as the root node of the wireless communication network. The network identification information of the router devices is configured. Computer terminals or mobile terminals are connected to the wireless communication network through the router devices. Network nodes and sensor nodes at various levels are deployed. Data collected by the sensor nodes is relayed level by level through the network nodes, forwarded by the relay routing devices, and uploaded to the host computer. In the data acquisition module 1, sensor node devices are deployed at key locations of the power transformer and in its surrounding environment. The sensor node devices include: Temperature measurement node equipment, environmental information detection node equipment, electrical information detection node equipment, and transformer oil chromatography analysis equipment; The key components of the power transformer include: the winding surface, the core surface, and the heat dissipation area of ​​the tank casing; the surrounding environment refers to the air environment around the power transformer installation area. The temperature measurement node device includes an infrared temperature sensor for collecting temperature data of feature points on the surface of the power transformer; The environmental information detection node device includes a temperature and humidity sensor for collecting ambient temperature and humidity data; The electrical information detection node equipment includes a current sensor and a voltage sensor, which are used to collect power transformer winding current data and terminal voltage data, respectively. The transformer oil chromatography analysis equipment is used to obtain the content and ratio parameters of dissolved gas components in transformer oil, and generate power transformer oil chromatography data to characterize the internal fault characteristics of power transformers. Each sensor node is a sub-node edge device in a wireless communication network. The sub-node edge device includes a housing for protecting internal components, a battery for power supply, a data conversion device for data format conversion, a chip device for processing data and transmitting signals, and an antenna for enhancing signal transmission capability. It can be used to convert the collected data into a transmittable format and send it to the wireless communication network.

[0058] The multi-source data collected by the data acquisition module 1 includes: Environmental temperature and humidity data, temperature data of characteristic points on the surface of the power transformer, electrical quantity data of the power transformer, and oil chromatography data of the power transformer.

[0059] Data processing module 2, in which: The system receives multi-source data collected by data acquisition module 1 through a wireless communication network, processes the multi-source data, and generates multi-source feature data. The data processing module 2 includes a data transmission process and a data processing process, wherein... During data transmission: Multi-source data is converted into a data format recognizable by the chip device via a data conversion device, the data is encapsulated by the chip device, and then transmitted to the sub-nodes in the wireless communication network by the antenna; Data is forwarded step by step through network nodes at each level, and relayed through relay routing equipment to transmit multi-source data to the host computer. During data processing: The multi-source data received by the host computer is sorted according to the acquisition time and resampled at a uniform time interval to achieve time alignment of the multi-source data; Anomalies are identified and removed from time-aligned multi-source data, and missing data is filled in. The processed multi-source data is normalized and transformed to map different physical quantities to a unified numerical range; Based on the normalized multi-source data, temperature characteristic parameters, environmental characteristic parameters, electrical characteristic parameters, and transformer oil chromatographic characteristic parameters were extracted to construct multi-source characteristic data.

[0060] Model training module 3, in which: A high-fidelity multiphysics simulation model of a power transformer is constructed. A reliable sample dataset is generated based on the high-fidelity multiphysics simulation model of the power transformer. The artificial neural network model is trained using the reliable sample dataset until convergence, and a hotspot temperature inversion model is obtained.

[0061] The model training module 3 includes a simulation modeling process, a reliable sample dataset generation process, and an artificial neural network model training process.

[0062] The simulation modeling in the model training module 3 specifically includes: Based on the structural parameters of the power transformer, structural modeling of the winding, core and tank is performed to generate a high-fidelity multiphysics simulation model of the power transformer. By establishing electromagnetic field models, temperature field models, and insulating oil fluid heat transfer models, and performing coupled calculations, a unified characterization of the internal thermal-electrical-current coupling process of power transformers is achieved. The sample generation in the model training module 3 specifically includes: By changing the boundary conditions and load conditions of the high-fidelity multiphysics simulation model, the high-fidelity multiphysics simulation model of the power transformer is simulated, generating simulation sample data covering different operating conditions, which is then used as a reliable sample dataset. The simulation sample data includes ambient temperature and humidity data, temperature data of characteristic points on the surface of the power transformer, electrical quantity data of the power transformer, and oil chromatography data of the power transformer. The artificial neural network model training process in the model training module 3 specifically includes: An artificial neural network model for hot spot temperature inversion of power transformers is constructed, which adopts a multi-layer feedforward neural network structure, including an input layer, a hidden layer and an output layer. A reliable sample dataset is input into the artificial neural network model, and the hotspot temperature obtained from simulation calculation is used as the supervised learning label. The network parameters are iteratively optimized through backpropagation and gradient descent methods. Through multiple iterations of training, the model gradually learns the nonlinear mapping relationship between the operating state characteristics of power transformers and hot spot temperatures. When the preset convergence condition is met, the trained hotspot temperature inversion model is output.

[0063] Hotspot temperature inversion module 4, in which: Multi-source feature data is used as input feature vectors to input into the trained hot spot temperature inversion model, which performs real-time inference calculations and outputs the current hot spot temperature value of the internal winding of the power transformer and the corresponding operating status evaluation results.

[0064] The real-time inference calculation in the hotspot temperature inversion module 4 specifically includes: Multi-source feature data is used as input feature vectors and input into a trained hot spot temperature inversion model. The input feature vectors are mapped and fused layer by layer through forward propagation calculation to obtain the current hot spot temperature value of the winding inside the power transformer. The forward propagation calculation process is as follows: the input feature vector is mapped to the hidden layer through the input layer, linearly transformed by the weight matrix and bias parameters, and then passed to the output layer layer by layer after being processed by the nonlinear activation function, and the current hotspot temperature value is output. Through the inference process of this model, nonlinear mapping and feature fusion of multi-source feature data are realized, thereby quickly obtaining the current hot spot temperature value of the power transformer and improving the efficiency and response speed of real-time inversion calculation. The operational status evaluation process in the hotspot temperature inversion module 4 specifically includes: The output hotspot temperature value is compared with the preset operating threshold, and an operating status evaluation result is generated according to the grading rules. At the same time, the hotspot temperature value and the operating status evaluation result are output to the host computer.

[0065] Visual interactive display module 5, in which: In the human-computer interaction visualization interface of the host computer, the hot spot temperature value, temperature distribution cloud map and operation status evaluation results are rendered and displayed in real time based on the three-dimensional transformer model, and an early warning prompt is triggered when the operation status is abnormal.

[0066] The process of constructing the 3D transformer model in the visualization and interactive display module 5 specifically includes: A three-dimensional transformer model is established based on the structural parameters of the power transformer to reflect the spatial structural relationship between the transformer windings, core, and tank, serving as a visualization carrier for operating status data. The three-dimensional transformer model is a display model, used only for visualization and display, and does not participate in the calculation of electromagnetic field, temperature field, and fluid field. In the 3D transformer model, a display mapping area corresponding to the temperature measurement node and key structural location is preset to carry the dynamic rendering of hot spot temperature values ​​and temperature distribution data.

[0067] The real-time rendering in the visual interactive display module 5 specifically includes: The hot spot temperature values ​​are mapped to the corresponding winding hot spot areas in the 3D transformer model and displayed in a numerical annotation manner; A temperature distribution cloud map is generated on the surface of a 3D transformer model based on hot spot temperature values, and the temperature distribution state is characterized by color gradient changes. The operational status assessment results are mapped to corresponding status identifiers, and the overall status is displayed using color differentiation. When the operation status assessment result reaches the heavy overload state or abnormal warning state, a warning prompt is triggered in the human-computer interaction visualization interface. The warning prompt includes interface highlighting, pop-up prompt and alarm information output. Simultaneously, the hot spot temperature values ​​and operational status assessment results are output to the host computer monitoring interface to support power transformer operational status monitoring and operational control decisions.

[0068] The above-disclosed embodiments are merely preferred embodiments of the present invention, but the present invention is not limited thereto. Any non-creative variations that can be conceived by those skilled in the art, as well as any improvements and modifications made without departing from the principles of the present invention, should fall within the protection scope of the present invention.

Claims

1. A method for hot spot temperature inversion applied to power transformers, characterized in that, Includes the following steps: Step S1: Construct a wireless communication network to collect real-time transformer operating status data and generate multi-source data by deploying sensor nodes and environmental monitoring nodes at key locations of the power transformer. Step S2: The collected multi-source data is transmitted to the host computer in real time through the wireless communication network. The host computer processes the received multi-source data to obtain multi-source feature data. Step S3: Construct a high-fidelity multiphysics simulation model of a power transformer. Simulate the model by changing the boundary conditions and load conditions, and generate simulation sample data covering different operating conditions as a reliable sample dataset. The artificial neural network model was trained offline using the aforementioned reliable sample dataset to obtain a hotspot temperature inversion model; Step S4: Input the multi-source feature data as the input feature vector into the hot spot temperature inversion model, perform real-time inference calculation, and invert the current hot spot temperature value of the internal winding of the power transformer and the corresponding operating status evaluation result. Step S5: In the human-computer interaction visualization interface of the host computer, the hot spot temperature value, temperature distribution cloud map and operation status evaluation results calculated by inversion are rendered and displayed in real time based on the three-dimensional transformer model, and an early warning prompt is triggered for abnormal status.

2. The hot spot temperature inversion method applied to power transformers according to claim 1, characterized in that, Step S1 specifically includes: S11. Connect the host computer and the relay router via wired or wireless means. The relay router is used as the root node of the wireless communication network. Configure the network identification information of the router. Connect the computer terminal or mobile terminal to the wireless communication network through the router. Deploy network nodes and sensor nodes at all levels. The data collected by the sensor nodes is relayed through network nodes at all levels, forwarded by the relay router, and uploaded to the host computer. S12. Temperature measurement node equipment, environmental information detection node equipment, and electrical information detection node equipment are deployed at sensor nodes and environmental monitoring nodes. The key parts of the power transformer are the winding surface, the core surface, and the heat dissipation area of ​​the oil tank shell. The surrounding environment is the air environment around the power transformer installation area. The power transformer operating status data is collected in real time through each sensor node, and after edge computing processing, multi-source data is generated. The multi-source data includes environmental temperature and humidity data, temperature data of characteristic points on the surface of the power transformer, electrical quantity data of the power transformer, and oil chromatography data of the power transformer.

3. The hot spot temperature inversion method applied to power transformers according to claim 2, characterized in that, Step S3 specifically includes: S31. Based on the structural parameters of the power transformer, structural modeling of the winding, core and tank is performed to generate a high-fidelity multiphysics simulation model of the power transformer. S32. Based on the high-fidelity multiphysics simulation model of power transformer, multiphysics simulation calculations are performed on power transformer through boundary conditions and load conditions. Based on the coupled solution results of electromagnetic field, temperature field and fluid heat transfer field, simulation sample data under different operating conditions are obtained and used as a reliable sample dataset. S33. Construct an artificial neural network, train the artificial neural network using a reliable sample dataset, and iteratively optimize the network parameters through backpropagation and gradient descent methods to enable the model to gradually learn the nonlinear mapping relationship between the operating state characteristics and hotspot temperatures. During the training process, the model training state is judged by setting the convergence threshold of the loss function and the error threshold of the validation set. After the convergence condition is met, the trained hotspot temperature inversion model is obtained.

4. The hot spot temperature inversion method for power transformers according to claim 3, characterized in that, Step S31 specifically includes: S311. An electromagnetic field calculation model is established based on Maxwell's equations to describe the current distribution in the windings of a power transformer and the electromagnetic response process of the terminal voltage. Where E is the electric field strength, H is the magnetic field strength, J is the current density, and B is the magnetic flux density; By solving the above equations using the finite element method, an electromagnetic field distribution model for a power transformer is constructed. S312. A temperature field model is constructed based on Fourier's law of heat conduction to characterize the internal heat diffusion process of a power transformer: Where T is temperature, ρ is density, c is specific heat capacity, k is thermal conductivity, and Q is the heat source term; used to characterize the temperature change process of the winding, core, and tank structure of a power transformer. S313. Construct an insulating oil fluid heat transfer model based on the Navier-Stokes equations to describe the oil circulation and convective heat transfer processes: Where v is the flow rate, p is the pressure, and μ is the dynamic viscosity; used to characterize the heat dissipation process of insulating oil circulating inside the tank; S314. Couple the electromagnetic field calculation model, the heat conduction calculation model, and the fluid heat transfer calculation model to construct a multiphysics coupling relationship: Among them, electromagnetic loss generates a heat source term Q, and the temperature field response T is affected by both the heat source and the fluid velocity field. A high-fidelity multiphysics simulation model of a power transformer is generated by performing coupled calculations of the electromagnetic field, temperature field, and fluid heat transfer field through iterative solution.

5. The hot spot temperature inversion method for power transformers according to claim 4, characterized in that, Step S32 specifically includes: S321. Apply environmental temperature boundary conditions and environmental humidity boundary conditions to the high-fidelity multiphysics simulation model of the power transformer, so that they act on the temperature field model and the insulating oil fluid heat transfer model and are transferred to the coupled field solution process through the boundary conditions; by solving the temperature field model and the insulating oil fluid heat transfer model, the response of the external environment temperature and humidity of the power transformer is obtained, and environmental temperature and humidity data are generated. S322. In the high-fidelity multiphysics simulation model of power transformer, a three-dimensional structural model of winding, core and tank is established. Coupled simulation is carried out through multiphysics coupling relationship to obtain the surface heat distribution results of power transformer and extract the temperature data of characteristic points on the surface of power transformer. S323. Apply different load rise and fall conditions parameters to the high-fidelity multiphysics simulation model of the power transformer, change the load current and terminal voltage conditions parameters, and simulate to obtain the winding current distribution and terminal voltage change results, and generate power transformer electrical quantity data. S324. Based on the temperature distribution results output by the temperature field calculation model in the high-fidelity multiphysics simulation model of power transformer, and combined with the pyrolysis characteristic parameters of oil-paper insulation and insulating oil, the simulation feature mapping calculation of the insulating oil pyrolysis process is performed to obtain the evolution characteristics of dissolved gas in the oil and generate power transformer oil chromatographic data. S325. The ambient temperature and humidity data, the temperature data of the surface feature points of the power transformer, the electrical quantity data of the power transformer, and the oil chromatographic data of the power transformer are summarized and integrated to generate a simulation sample dataset covering different operating conditions, and this dataset is used as a reliable sample dataset.

6. The hot spot temperature inversion method for power transformers according to claim 5, characterized in that, Step S33 specifically includes: S331. Construct an artificial neural network model for hot spot temperature inversion of power transformers. The artificial neural network model adopts a multi-layer feedforward neural network structure, including an input layer, a hidden layer, and an output layer. The input layer receives feature data from a trusted sample dataset, including ambient temperature and humidity data, surface feature point temperature data of the power transformer, electrical quantity data of the power transformer, and oil chromatography data of the power transformer. The hidden layer adopts a fully connected network structure and introduces a nonlinear activation function to enhance the fitting ability of complex nonlinear mapping relationships, and is used for feature fusion and high-dimensional mapping of the feature data. The output layer outputs the predicted hot spot temperature of the power transformer. S332. Divide the credible sample dataset into a training set and a validation set, input the training set into the artificial neural network model, and use the hotspot temperature output by the high-fidelity multiphysics simulation model of the power transformer as the supervised learning label; A loss function is defined to quantify the model output error; the loss function is the mean squared error function. This is used to measure the deviation between the predicted hotspot temperature and the simulated hotspot temperature. This represents the predicted hotspot temperature output by the artificial neural network. N represents the hot spot temperature calculated by the high-fidelity multiphysics simulation model of the power transformer, where N represents the number of samples. The gradient of the loss function with respect to the network weights is calculated by backpropagation algorithm, and the network parameters are iteratively updated by gradient descent optimization method, so that the artificial neural network model can gradually learn the mapping relationship between the hot spot temperature of the power transformer and various simulation feature data under simulation conditions. During the training process, continuous input of reliable simulation sample data under different operating conditions enhances the model's ability to fit the coupling relationship of multiple physics fields and the changing conditions of multiple operating conditions. Perform multiple iterations of training; S333. Evaluate the trained artificial neural network model based on the validation set, and stop training when any of the following convergence conditions are met: (1) The loss function converges to the preset threshold ε, that is, the change of the loss function is less than ε in several consecutive iterations; (2) The mean square error on the validation set is less than the preset error threshold δ; (3) The rate of decrease in model prediction error tends to stabilize and no longer improves significantly; Once the convergence condition is met, the hotspot temperature inversion model is obtained.

7. The hot spot temperature inversion method for power transformers according to claim 6, characterized in that, Step S4 specifically includes: S41. Input the multi-source feature data as the input feature vector into the trained hot spot temperature inversion model, and perform layer-by-layer mapping and feature fusion on the input feature vector through forward propagation calculation to output the current hot spot temperature value of the internal winding of the power transformer. S42. Compare the output hot spot temperature value with the preset operating threshold, and obtain the operating status assessment result according to the classification rules; output the hot spot temperature value and the corresponding operating status assessment result to the host computer for power transformer operating status monitoring, load assessment and fault early warning decision support.

8. The hot spot temperature inversion method applied to power transformers according to claim 7, characterized in that, Step S5 specifically includes: S51. A three-dimensional transformer model is established based on the structural parameters of the power transformer to reflect the spatial structural relationship between the transformer windings, core and tank, and to serve as a visualization carrier for operating status data. In the 3D transformer model, a display mapping area corresponding to the actual temperature measurement nodes and key structural locations is preset to carry the dynamic rendering of hot spot temperature values ​​and temperature distribution data. S52. Input the obtained hotspot temperature values ​​and operating status evaluation results into the human-computer interaction visualization interface, and map them onto the 3D transformer model for real-time rendering and display, specifically including: The hot spot temperature values ​​are mapped to the corresponding winding hot spot areas in the 3D transformer model and displayed in the form of numerical labels; Based on the hot spot temperature values, a temperature distribution cloud map is generated on the surface of the three-dimensional transformer model, and the temperature field distribution is reflected by the color gradient change. The operational status assessment results are mapped to corresponding status identifiers, and the overall status is displayed in the 3D transformer model using color differentiation. When the operation status assessment results reach the state of severe overload or abnormal warning, an early warning prompt is triggered in the human-machine interaction visualization interface. At the same time, the hot spot temperature value and the operation status assessment results are synchronously output to the host computer monitoring interface.

9. A hot spot temperature inversion system applied to power transformers, characterized in that, include: The data acquisition module contains: A wireless communication network is constructed, and various sensor nodes and detection devices are deployed in key parts of the power transformer and its surrounding environment to collect real-time power transformer operating status data and generate multi-source data. The data processing module contains: The system receives multi-source data collected by the data acquisition module through a wireless communication network, processes the multi-source data, and generates multi-source feature data. The model training module contains: A high-fidelity multiphysics simulation model of a power transformer is constructed. A reliable sample dataset is generated based on the high-fidelity multiphysics simulation model of the power transformer. The artificial neural network model is trained using the reliable sample dataset until it converges, and a hotspot temperature inversion model is obtained. The hotspot temperature inversion module contains: Multi-source feature data is used as input feature vectors to input the trained hot spot temperature inversion model, and real-time inference calculation is performed to output the current hot spot temperature value of the internal winding of the power transformer and the corresponding operating status evaluation result. The visual interactive display module includes: In the human-computer interaction visualization interface of the host computer, the hot spot temperature value, temperature distribution cloud map and operation status evaluation results are rendered and displayed in real time based on the three-dimensional transformer model, and an early warning prompt is triggered when the operation status is abnormal.

10. A hotspot temperature inversion system for power transformers according to claim 9, characterized in that, In the data acquisition module, sensor node devices are deployed at key parts of the power transformer and in the surrounding environment. The sensor node devices include: temperature measurement node devices, environmental information detection node devices, electrical information detection node devices, and transformer oil chromatography analysis equipment. The key components of the power transformer include the winding surface, the core surface, and the heat dissipation area of ​​the tank casing; the surrounding environment refers to the air environment around the power transformer installation area. The temperature measurement node device includes an infrared temperature sensor for collecting temperature data of feature points on the surface of the power transformer; The environmental information detection node device includes a temperature and humidity sensor for collecting ambient temperature and humidity data; The electrical information detection node equipment includes a current sensor and a voltage sensor, which are used to collect power transformer winding current data and terminal voltage data, respectively. The transformer oil chromatography analysis equipment is used to obtain the content and ratio parameters of dissolved gas components in transformer oil, generate power transformer oil chromatography data, and characterize the internal fault characteristics of power transformers. Each sensor node is a sub-node edge device in a wireless communication network. The sub-node edge device includes a housing for protecting internal components, a battery for power supply, a data conversion device for data format conversion, a chip device for processing data and transmitting signals, and an antenna for enhancing signal transmission capability.

Images