High-voltage switchgear temperature field monitoring method and system based on digital twinning

Abstract

The application belongs to the technical field of high-voltage switchgear state monitoring, and particularly relates to a high-voltage switchgear temperature field monitoring method and system based on digital twinning, comprising: obtaining a three-dimensional model and real-time operation data of the high-voltage switchgear based on digital twinning; obtaining a temperature field prediction result of the high-voltage switchgear according to the obtained real-time operation data and a deep learning reduced-order model; wherein the deep learning reduced-order model comprises a deep feedforward neural network for mapping the obtained real-time operation data from a working condition parameter space to a low-dimensional feature space and a convolutional autoencoder for reconstructing a temperature field from low-dimensional feature vector data output by the deep feedforward neural network; associating the obtained temperature field prediction result with the three-dimensional model, performing visual interactive display of the high-voltage switchgear temperature field, and completing monitoring of the high-voltage switchgear temperature field based on digital twinning.

Description

Technical Field

[0001] This invention belongs to the field of high-voltage switchgear condition monitoring technology, specifically relating to a method and system for monitoring the temperature field of high-voltage switchgear based on digital twins. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] High-voltage switchgear is a core component of the power grid. During the interruption of large currents and long-term operation, it generates strong coupling effects from multiple physical fields, including electromagnetic, thermal, and hydrodynamic fields. For example, short-circuit currents generate intense Joule heating at conductors and contacts, causing overheating and deformation of conductive components; the arcing and extinguishing processes are accompanied by enormous thermal shocks and gas flows; and the thermal expansion of the equipment structure affects parameters such as contact resistance. To ensure the safe and reliable operation of high-voltage switchgear, it is necessary to accurately obtain the temperature field distribution of the switchgear under various operating conditions, and to use this temperature field distribution for fault warning and lifespan assessment.

[0004] Currently, obtaining the temperature field of high-voltage switchgear mainly relies on full-scale numerical simulation calculations, using the finite element / finite volume method to solve multi-field coupled partial differential equations. However, for 550kV GIS equipment with complex structures and huge scale, it is necessary to divide the grid into millions or even tens of millions of levels in the process of constructing a full-order simulation model and solve highly nonlinear control equations. The simulation calculation for a single operating condition often takes tens of hours, which is extremely inefficient.

[0005] Based on existing research, digital twin technology can be used to obtain the temperature field of high-voltage switchgear. While existing digital twin technologies can achieve high-fidelity rendering of 3D models of high-voltage switchgear, assembly and maintenance animation simulation, and VR interaction, they still rely on traditional numerical calculation methods or classical reduced-order models (such as POD reduction) to predict the temperature field. This results in low computational efficiency and the inability to update in real time. Especially when encountering scenarios with strong nonlinear boundary conditions or material properties, the traditional POD-Galerkin reduced-order model relies on a linear trial function space, making it difficult to accurately characterize nonlinear effects, leading to a significant decrease in computational accuracy. For example, in GIS switches, the contact resistance of the fingers changes nonlinearly with temperature; the error of the traditional POD method increases significantly under these conditions.

[0006] Existing deep learning reduced-order ROM (DL-ROM) models are mostly geared towards general-purpose equipment such as aerospace blades and turbine blades. They lack specific design for the unique operating conditions of high-voltage switchgear, such as strong electromagnetic-thermal-structural coupling and short-term high-current impacts, and also fail to consider the integration requirements for real-time data interaction with digital twin technology. Therefore, given the significant shortcomings of existing technologies in real-time prediction of temperature fields in GIS equipment, there is an urgent need for a customized and rapid monitoring method tailored to the specific needs of this field. Summary of the Invention

[0007] To address the aforementioned issues, this invention proposes a temperature field monitoring method and system for high-voltage switchgear based on digital twins. It employs a temperature field calculation method combining a deep learning reduced-order model with digital twin technology to achieve real-time and rapid prediction of the temperature field of high-voltage switchgear under any time and operating condition. Furthermore, by combining digital twins with real-time data interaction of the high-voltage switchgear, it enables real-time status monitoring of the high-voltage switchgear, thereby improving the safety and reliability of its operation.

[0008] According to some embodiments, the first aspect of the present invention provides a method for monitoring the temperature field of high-voltage switchgear based on digital twins, employing the following technical solution:

[0009] A method for monitoring the temperature field of high-voltage switchgear based on digital twins, comprising:

[0010] Based on digital twins, obtain 3D models and real-time operating data of high-voltage switchgear;

[0011] Based on the acquired real-time operating data and the deep learning reduction model, the temperature field prediction result of the high-voltage switchgear is obtained; wherein, the deep learning reduction model includes a deep feedforward neural network for mapping the acquired real-time operating data from the operating parameter space to a low-dimensional feature space and a convolutional autoencoder for reconstructing the temperature field from the low-dimensional feature vector data output by the deep feedforward neural network.

[0012] The obtained temperature field prediction results are correlated with the three-dimensional model to perform interactive visualization of the temperature field of high-voltage switchgear, thus completing the monitoring of the temperature field of high-voltage switchgear based on digital twin.

[0013] As a further technical limitation, the deep feedforward neural network is used to learn the nonlinear mapping relationship from the wide operating condition parameter space to the low-dimensional feature space of the high-voltage switchgear. The operating condition parameter vector in the acquired real-time operating data is input, and the input information is mapped to the low-dimensional feature space through the hidden layer of the network with multiple fully connected layers to obtain the low-dimensional feature representation of the real-time operating data. The convolutional autoencoder includes an encoder convolutional neural network and a decoder convolutional neural network. The low-dimensional feature representation vector of the obtained real-time operating data is transformed into dimensionality-reduced features under the action of the encoder convolutional neural network and the decoder convolutional neural network, completing the prediction from temperature field parameters to temperature field and realizing temperature field reconstruction.

[0014] As a further technical limitation, in the process of obtaining the three-dimensional model of the high-voltage switchgear, a three-dimensional geometric model of the high-voltage switchgear is constructed based on its structure. The three-dimensional geometric model is then meshed according to the structural type of the high-voltage switchgear. Based on the meshed geometric model, the coupling control equations of different physical fields are obtained, and a multi-physics coupling model of the high-voltage switchgear is constructed. Based on the constructed multi-physics coupling model, full-order simulation calculations of the high-voltage switchgear are performed, and a multi-condition full-order simulation dataset of the high-voltage switchgear is constructed to obtain the three-dimensional model of the high-voltage switchgear.

[0015] Furthermore, the loss function of the deep learning order reduction model is a weighted sum of the global error function and the key region error function of the temperature field reconstruction; wherein, the global error function is the mean square error between the reconstructed temperature field and the original simulation field, the reconstructed temperature field is the temperature field prediction result; the original simulation field is multi-condition full-order simulation data; the key region error function is the error term of the key parts in the high-voltage switchgear; the network parameters of the deep feedforward neural network and the convolutional autoencoder are adjusted by the backpropagation algorithm to reduce the loss function of the deep learning order reduction model; when the loss function of the deep learning order reduction model reaches a preset loss threshold, the training of the deep learning order reduction model is completed, and the mapping from real-time operating data condition parameters to the temperature field is completed.

[0016] As a further technical limitation, after the temperature field prediction results are correlated with the three-dimensional model, a three-dimensional geometric model is obtained that displays the temperature distribution of various parts of the high-voltage switchgear in real time as a cloud map. Combined with the digital twin display of temperature data of key parts of the high-voltage switchgear and the change curve of hot spot temperature, the temperature field evolution of the high-voltage switchgear can be viewed, and the visualization interaction of the high-voltage switchgear can be completed.

[0017] As a further technical limitation, when the temperature of the high-voltage switchgear rises abnormally, a temperature anomaly warning is issued based on the visualized interactive results of the high-voltage switchgear and combined with digital twin. The cause of the abnormal temperature rise is analyzed based on the temperature field prediction results, thus completing the monitoring of the temperature field of the high-voltage switchgear.

[0018] According to some embodiments, a second aspect of the present invention provides a high-voltage switchgear temperature field monitoring system based on digital twins, employing the following technical solution:

[0019] A temperature field monitoring system for high-voltage switchgear based on digital twins, comprising:

[0020] The acquisition module is configured to acquire a 3D model and real-time operating data of the high-voltage switchgear based on digital twins.

[0021] The prediction module is configured to obtain the temperature field prediction result of the high-voltage switchgear based on the acquired real-time operating data and the deep learning reduction model; wherein the deep learning reduction model includes a deep feedforward neural network for mapping the acquired real-time operating data from the operating parameter space to a low-dimensional feature space and a convolutional autoencoder for reconstructing the temperature field from the low-dimensional feature vector data output by the deep feedforward neural network.

[0022] The monitoring module is configured to associate the obtained temperature field prediction results with the three-dimensional model, perform interactive visualization of the temperature field of the high-voltage switchgear, and complete the monitoring of the temperature field of the high-voltage switchgear based on digital twin.

[0023] According to some embodiments, a third aspect of the present invention provides a computer-readable storage medium, employing the following technical solution:

[0024] A computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the high-voltage switchgear temperature field monitoring method based on digital twin as described in the first aspect of the present invention.

[0025] According to some embodiments, the fourth aspect of the present invention provides an electronic device, which adopts the following technical solution:

[0026] An electronic device includes a memory, a processor, and a program stored in the memory and running on the processor, wherein the processor executes the program to implement the steps in the high-voltage switchgear temperature field monitoring method based on digital twin as described in the first aspect of the present invention.

[0027] According to some embodiments, the fifth aspect of the present invention provides a computer program product, which adopts the following technical solution:

[0028] A computer program product includes software code, wherein the program in the software code performs the steps of the high-voltage switchgear temperature field monitoring method based on digital twin as described in the first aspect of the present invention.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0030] This invention employs a temperature field calculation method combining a deep learning-based reduced-order model and digital twin technology. It extracts low-dimensional features from high-dimensional operating parameters of high-voltage switchgear using a deep feedforward neural network, and then maps these low-dimensional features to the temperature field using a convolutional autoencoder. Based on the temperature field distribution results obtained from the deep learning-based reduced-order model and the temperature field prediction results obtained from the digital twin, real-time simulation and prediction of the high-voltage switchgear temperature field are achieved. This enables real-time calculation of the multi-physics temperature field of high-voltage switchgear based on digital twin technology, integrating temperature field "calculation-simulation-diagnosis-decision-making" into a single system. This allows for real-time status monitoring of high-voltage switchgear, improving the safety and reliability of its operation. Attached Figure Description

[0031] The accompanying drawings, which form part of this embodiment, are used to provide a further understanding of this embodiment. The illustrative embodiments and their descriptions are used to explain this embodiment and do not constitute an improper limitation of this embodiment.

[0032] Figure 1 This is a flowchart of the high-voltage switchgear temperature field monitoring method based on digital twin in Embodiment 1 of the present invention;

[0033] Figure 2 This is an overall block diagram of the high-voltage switchgear temperature field monitoring method based on digital twin in Embodiment 1 of the present invention;

[0034] Figure 3 This is a structural block diagram of the high-voltage switchgear temperature field monitoring system based on digital twins in Embodiment 2 of the present invention. Detailed Implementation

[0035] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0036] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0037] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0038] In this invention, terms such as "upper," "lower," "left," "right," "front," "back," "vertical," "horizontal," "side," and "bottom" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are only used to facilitate the description of the structural relationships of the various components or elements of this invention and do not specifically refer to any component or element in this invention. They should not be construed as limiting the invention.

[0039] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0040] Example 1

[0041] Embodiment 1 of this invention introduces a method for monitoring the temperature field of high-voltage switchgear based on digital twins.

[0042] like Figure 1 The method for monitoring the temperature field of high-voltage switchgear based on digital twins, as shown, includes:

[0043] Based on digital twins, obtain 3D models and real-time operating data of high-voltage switchgear;

[0044] Based on the acquired real-time operating data and the deep learning reduction model, the temperature field prediction result of the high-voltage switchgear is obtained; wherein, the deep learning reduction model includes a deep feedforward neural network for mapping the acquired real-time operating data from the operating parameter space to a low-dimensional feature space and a convolutional autoencoder for reconstructing the temperature field from the low-dimensional feature vector data output by the deep feedforward neural network.

[0045] The obtained temperature field prediction results are correlated with the three-dimensional model to perform interactive visualization of the temperature field of high-voltage switchgear, thus completing the monitoring of the temperature field of high-voltage switchgear based on digital twin.

[0046] This embodiment combines a deep learning reduced-order model (DL-ROM) with a 550kV GIS high-voltage switchgear digital twin platform to rapidly calculate the temperature field of high-voltage switchgear. Specifically, a deep neural network model is constructed to replace traditional numerical simulation, enabling second-level prediction of the multi-physics temperature field of GIS equipment. Based on the high-fidelity 3D model rendering, equipment assembly and maintenance animation, and VR interactive functions provided by the existing digital twin platform, a DL-ROM algorithm module is integrated to achieve real-time and rapid prediction of the temperature field of GIS equipment at any time and under any operating condition, enhancing the completeness and intelligence of the platform's functions.

[0047] like Figure 2 As shown, the high-voltage switchgear temperature field monitoring method based on digital twins in this embodiment is divided into an offline training stage and an online prediction stage, and mainly includes the following steps:

[0048] Step 1: Construction of a multiphysics simulation model for a 550kV GIS high-voltage switch;

[0049] In the offline phase, a high-precision, full-size simulation model of the GIS equipment is established to generate a large amount of temperature field data required for neural network training. Specifically:

[0050] Based on the design drawings and physical structure of the 550kV GIS equipment, a 3D geometric model of the main components of the GIS equipment was constructed and meshed. To balance accuracy and computational cost, complex structures were appropriately simplified in the modeling. For example, the conductive circuit model included conductive moving contacts, contact seats, spring contacts, petal-shaped stationary contacts, and conductive rods; the heat dissipation structure included basin-type insulators, insulating tie rods, and aluminum alloy shells; and the fluid region included... The gas arc-extinguishing chamber, etc. During the mesh generation process, locally refined tetrahedral meshes are used for key heat-generating areas such as conductive contacts to obtain higher resolution temperature gradients; boundary layer meshes are generated at the interface between the gas chamber and the metal wall to more accurately capture the fluid heat transfer boundary layer effect.

[0051] After meshing the 3D geometric model, the coupled control equations for the electromagnetic field, temperature field, and flow field inside the GIS equipment are defined; specifically:

[0052] The electromagnetic field adopts Maxwell's equations under the quasi-static approximation, and the electromagnetic field state is represented by the magnetic vector potential A, which satisfies the electromagnetic field governing equations. Under the quasi-static approximation, the magnetic vector potential A satisfies: ;in, Permeability, For electrical conductivity, denoted as current density.

[0053] electric field strength Derived from the magnetic vector potential, i.e. .

[0054] The governing equations for heat conduction and convection are: ;in, For density, For specific heat capacity, Thermal conductivity, For fluid velocity, This is the Joule heat source term.

[0055] For the fluid dynamics governing equations (incompressible Navier-Stokes), the fluid velocity and pressure satisfy the continuity equation and momentum equation, which can be expressed as:

[0056] ;

[0057] ;

[0058] in, For dynamic viscosity, Volume forces, including electromagnetic forces.

[0059] Based on the aforementioned electromagnetic, thermal, and flow field equations, a set of control equations for multi-physics coupling within high-voltage switchgear is obtained. This set is used for high-fidelity simulation calculations of the temperature field distribution of the equipment under different operating conditions. Through this multi-physics coupling model, the physical responses of GIS equipment during processes such as current flow, heat generation, gas flow, and structural stress can be comprehensively simulated.

[0060] Based on the obtained multiphysics coupling model, full-order simulation calculations were performed on typical operating conditions to obtain rich temperature field distribution data. The design of typical operating conditions should cover the main operating range and extreme cases of GIS equipment. For example, the current level should cover 500A under light load, 5000A under rated load, 6000A under overload, and 63kA for breaking short circuit current; the time nodes should include 0.01s for short circuit instant, 1s for breaking process, 100s for temperature rise process, and 10000s for long-term steady state; the ambient temperature should consider low winter temperature of -20℃, normal temperature of 25℃, and high summer temperature of 40℃.

[0061] It should be noted that the full-order simulation calculation in this embodiment uses existing finite element numerical calculation technology to solve the simulation model of boundary conditions under multiple working conditions in order to create a training dataset.

[0062] By combining different current amplitudes, durations, and environmental conditions, simulation schemes can be formed for 48 typical operating conditions (4 types of current × 4 types of time × 3 types of environment). For each operating condition, transient coupled electromagnetic-thermal-fluid simulation is performed on a full-size model (i.e., a pre-processed simplified geometric model simulating the size of real GIS equipment for finite element calculations) to obtain the temperature field distribution from the initial time to the target time. Since each set of operating conditions requires solving a large number of nonlinear equations, it may still take tens of hours on a high-performance computing platform. Based on the 48 operating conditions, a total of approximately 48 × 1.5 million node scale temperature field snapshot data is generated; this high-precision, high-resolution temperature field data will serve as the basic dataset for training the deep learning reduced-order model.

[0063] Step 2: DL-ROM Model Construction (DFNN+CAE Architecture)

[0064] After obtaining sufficient simulation data, a reduced-order model of the temperature field is performed using deep learning methods. In this embodiment, the DL-ROM model consists of two sub-networks: a deep feedforward neural network (DFNN) and a convolutional autoencoder (CAE) decoder. These sub-networks are responsible for mapping input parameters to low-dimensional features and reconstructing the full-field temperature from low-dimensional features, respectively. Specifically, this includes:

[0065] (1) Deep Feedforward Neural Network (DFNN)

[0066] DFNN is used to learn the nonlinear mapping relationship from a wide operating condition parameter space to a low-dimensional feature space of GIS equipment. The input of DFNN is a vector containing the main operating condition parameters; for example, three dimensions could correspond to the operating current. I (e.g., 0-63kA), ambient temperature (e.g., -20 to +40℃) and operating time t (Simulation evolution time).

[0067] This embodiment employs four hidden layers (multiple fully connected layers), with the number of neurons being 256, 128, 64, and 32 respectively. The ReLU activation function (parameter α=0.01) is used in each layer to enhance non-linear expressive power and avoid gradient vanishing. The output layer of the DFNN has a dimension of [missing information]. The low-dimensional feature vectors are used to characterize the main feature patterns of the temperature field under corresponding operating conditions. For example, different dimensions of the feature vectors can represent the temperature rise characteristics of different parts / components in GIS equipment (such as average contact temperature, insulator temperature gradient, average conductor rod temperature, shell temperature rise rate, etc.). Through DFNN pre-training, it can effectively compress the information on the influence of a wide range of operating conditions (hundreds of thousands of possible combinations) on the temperature field, providing concise feature inputs for subsequent CAE decoding.

[0068] The training process of DFNN (Deep Feedforward Neural Network Supervised Learning) is as follows: Establish operating parameters such as current... I Ambient temperature T amb Running time t Dimensionality reduction characteristics of temperature field The mapping relationship enables "parameter input" Rapid calculation of "temperature field prediction".

[0069] Let DFNN be a multi-layer fully connected network, with the mapping function being... ,but:

[0070]

[0071] in, For the DFNN layer weights and biases, This is the activation function.

[0072] Dimensionality reduction features generated by CAE To calculate the mean square error, we have the following: Historical temperature field data obtained through CAE Encoding The corresponding parameters Constitute training pairs ( (2) Convolutional Autoencoder (CAE)

[0073] CAE is responsible for reconstructing the temperature field of GIS from a low-dimensional feature representation to a full three-dimensional temperature field distribution. CAE consists of two convolutional neural networks, an encoder and a decoder, with the bottleneck layer in the middle connected to the feature vector output by the DFNN.

[0074] CAE unsupervised pre-training (temperature field data) will generate a high-dimensional temperature field data matrix. ,in N This represents the number of samples (snapshots of the temperature field under different parameters). D Using spatial dimensions (number of grid nodes), the encoder reconstructs the data through self-learning. Approximating the original data This allows us to learn low-dimensional feature representations of the data.

[0075] The bottleneck layer is set at the output of the CAE encoder and the input of the decoder. It is the intermediate layer with the lowest feature dimension in the network. The bottleneck layer is a low-dimensional feature interaction node. On the one hand, it receives the device parameter feature vector output by DFNN. On the other hand, it drives the CAE decoder to reconstruct the temperature field, realizing the closed loop of the mapping 'parameter → feature physical field'.

[0076] The bottleneck layer, serving as the core hub connecting the parameter space and the physical field space, overcomes the limitations of traditional CAE through multi-source feature fusion and dynamic mapping mechanisms; its core function is to integrate the temperature field spatial pattern features extracted by the CAE encoder. and Generated parameter association features The contribution of dynamic balancing of physical priors and parameter-driven approaches is discussed.

[0077]

[0078] in, For activation function, As weight, This represents element-wise multiplication. It represents the spatial distribution pattern of the temperature field captured by the encoder (such as the location of contact hotspots and the temperature gradient of the heat dissipation channel). Indicates current load I Ambient temperature T env Running time t The low-dimensional vector mapped to.

[0079] The CAE encoder takes high-fidelity temperature field data as input and compresses and extracts it into low-dimensional features. For example, for each temperature field snapshot (about 1.5 million grid nodes), interpolation or downsampling can be used to normalize it into a three-dimensional tensor of about 200×200×3 (equivalent to 120,000 effective nodes), which is then input into several layers of three-dimensional convolution for encoding.

[0080] In this embodiment, the encoder uses 3 convolutional layers, each with a kernel size of 3×3 and a stride of 2. A batch normalization layer is added after the convolution to accelerate convergence. The encoder finally outputs an 8-dimensional feature vector to characterize the main pattern features of the input temperature field.

[0081] The encoder compresses the high-dimensional temperature field into a low-dimensional feature vector. (Dimensionality reduction space); Specifically; Suppose the encoder consists of multiple convolutional layers, and the mapping function is... ,but:

[0082]

[0083] in, For the first Layer weights and biases For activation function, To reduce dimensionality.

[0084] The CAE decoder performs the opposite operation to the encoder. Specifically, it takes the aforementioned 8-dimensional feature vector as input and gradually "amplifies" the features through three layers of transposed convolutions to reconstruct the original temperature field distribution. The decoder uses a 4×4 convolution kernel with a stride of 2 and introduces skip connections into the network to fuse intermediate features from the encoding stage with the corresponding decoding layer, thereby recovering temperature distribution details in complex structures such as contact details and narrow channels. Simultaneously, the CAE encoder incorporates the fusion of geometric topological features from GIS equipment, such as using the geometric identifiers or coordinates of key equipment components as additional input, allowing the network to retain topological information while compressing features. Through customized structural optimization, the CAE can more accurately reconstruct the temperature field of regions with complex geometric structures, significantly reducing the order reduction reconstruction error.

[0085] The decoder will convert the low-dimensional feature vector Reconstructing high-dimensional temperature field data The decoder is the inverse process of the encoder, and the mapping function is... ,but:

[0086]

[0087] in, For decoder number Layer weights and biases For activation function, To predict the target, .

[0088] It should be noted that in the CAE decoder, skip connections, through a cross-layer feature reuse mechanism, directly transmit the detailed temperature field features captured by each level of the encoder (such as contact edge gradients, geometric contours of narrow channels, etc.) to the corresponding level of the decoder, and then concatenate or weightedly fuse them with the upsampled feature maps by channel. This process preserves high-frequency spatial information from the encoding stage, avoiding the loss of details caused by downsampling, thereby accurately restoring the temperature distribution at complex structures in GIS equipment during temperature field reconstruction (such as local temperature rise spikes caused by contact resistance of the contacts, strong gradient distribution of SF6 gas in narrow channels), and further enhances physical realism through geometric topological feature embedding, significantly reducing reconstruction errors.

[0089] (3) Combine and train the two sub-networks DFNN and CAE to obtain an end-to-end reduced-order mapping model. During the combined training process, the parameters of DFNN and CAE are jointly optimized using a large amount of multi-condition temperature field snapshot data generated in the previous steps.

[0090] Update the encoder and decoder parameters through backpropagation to minimize This makes the low-dimensional feature vector It contains key features of the temperature field.

[0091] The reconstruction error is represented by the mean square error. .

[0092] This embodiment employs a joint training strategy, defining the loss function as a weighted sum of the global error and the critical area error in temperature field reconstruction. The global error function uses the mean square error (MSE) between the reconstructed temperature field (temperature field prediction result) and the original simulation field (multi-condition full-order simulation data). The critical area error function defines higher-weighted error terms for key components in the GIS equipment, such as heat-prone points and insulator interfaces, ensuring the prediction accuracy of these areas. The network parameters of the DFNN and CAE are adjusted using a backpropagation algorithm to continuously reduce the training error. Once a predetermined accuracy threshold is reached on the validation set, the training process ends, resulting in a trained DL-ROM reduced-order model. This model essentially contains a fast mapping from operating parameters to the temperature field, capable of instantaneously outputting the corresponding three-dimensional temperature field distribution given input conditions.

[0093] It should be noted that the backpropagation algorithm in this embodiment is a mature existing technology in the field of deep learning. The backpropagation algorithm adjusts the parameters of DFNN and CAE to reduce training error through the following mechanism: First, the input data is forward propagated through DFNN and CAE to generate feature vectors. After dynamic fusion through the bottleneck layer, the temperature field is reconstructed by the CAE decoder, and the loss between the reconstructed value and the true value (such as MSE) is calculated. Then, the gradient of the loss with respect to the parameters of each layer is derived in reverse using the chain rule. The gradient directly optimizes the local feature extraction capability of the CAE encoder through skip connections and is passed to the output features of DFNN through the weights of the fusion layer, thereby updating the internal weights of DFNN in reverse. At the same time, the transposed convolution parameters of CAE decoder and the convolution kernel parameters of encoder are also adjusted synchronously through the reconstruction loss gradient. The optimizer (such as Adam) updates all parameters according to the gradient direction, making the feature mapping of DFNN more in line with physical laws and the encoding and decoding of CAE more accurately capture spatial patterns. Through iterative training, a closed loop of "loss driving gradient → gradient updating parameters → parameter optimizing features → feature reducing loss" is formed, ultimately achieving a continuous decrease in training error under dual-network collaborative optimization.

[0094] Step 3: Integration of Digital Twin Platform and Real-time Temperature Field Calculation

[0095] The DL-ROM model trained in step two is deployed and integrated into the digital twin platform of the GIS equipment to achieve real-time temperature field prediction and visualization; specifically:

[0096] On the backend server or cloud of the digital twin platform, the trained DL-ROM model is loaded and deployed, and then connected to the real-time data stream through the platform's API interface. During platform operation, the platform periodically or on demand acquires the operating parameters of the GIS equipment (such as real-time current, switch opening and closing status, etc.). (Gas pressure, ambient temperature, etc.) calls the deployed DL-ROM model interface to pass these input parameters to the model in real time for calculation.

[0097] When the digital twin platform receives new operating parameters, the DL-ROM model immediately performs predictive calculations. The calculation results of the DL-ROM reduced-order model are mapped to a new color JSON file through the backend server, updating the old file. After the update is completed, it is transmitted to the digital twin model via POST communication. The twin model reads the updated JSON file and simultaneously updates the variable colors for rendering. Specifically, the platform feeds the input multi-dimensional operating parameters into the DFNN sub-network, quickly mapping them to obtain low-dimensional feature vectors. This feature is then input into the CAE decoder network to reconstruct a complete three-dimensional temperature field. Because the computational load of neural network inference is very small (compared to solving complex PDEs), the entire prediction process only takes milliseconds to output the temperature field results. This second-level response capability enables the platform to achieve real-time updates of the temperature field of GIS equipment. Whether it is daily rated operation, temporary overload, or the temperature rise during a fault short circuit, it can quickly predict and promptly reflect the current thermal distribution of the equipment.

[0098] The digital twin platform correlates the temperature field results output from the DL-ROM model with the 3D model of the equipment for visualization. For example, the temperature distribution of various parts is displayed as a cloud map on the 3D geometric model of the GIS equipment, with high-temperature areas marked as red hotspots. Simultaneously, the platform's UI interface displays the temperature values ​​of key measuring points and the curves of hotspot temperatures over time. Maintenance personnel wearing VR devices or on the monitoring interface can intuitively view the evolution of the internal temperature field of the equipment, obtaining comprehensive thermal information complementary to the actual sensor temperature measurement points. When the temperature rises abnormally, the platform can issue an early warning and analyze possible causes of the fault based on simulation results. When on-site online monitoring current load or ambient temperature information is input into the reduced-order model, it quickly calculates and, based on the temperature field chromatogram rendered by the digital twin model, can determine whether the temperature inside the GIS is excessively high in a localized cloud map or generally high within the air chamber, thus identifying simulated localized overheating or poor air chamber heat dissipation. For example, localized overheating of a contact may indicate poor contact or excessive resistance; generally high temperatures within the air chamber may be related to poor heat dissipation or overload.

[0099] This embodiment integrates the electromagnetic, thermal, and fluid coupling effects acting on GIS equipment at the data level, and simultaneously collects temperature field snapshots under the influence of multiple factors such as current, time, and ambient temperature. It directly trains the neural network to learn the nonlinear mapping relationship of "multi-field input → temperature field output". This end-to-end data-driven modeling breaks through the limitations of previous models that only relied on physical simplification assumptions, and greatly improves the model's ability to cope with complex coupled working conditions.

[0100] To address the wide operating range and complex geometric topology of 550kV GIS switchgear, this embodiment employs a dual-branch deep neural network structure. A deep feedforward neural network (DFNN) extracts low-dimensional feature representations of high-dimensional operating parameters (e.g., current ranging from 500A to 5000A, time from milliseconds to hours, and ambient temperature from -20℃ to 40℃), enabling efficient learning of the relationship between wide operating parameters and low-dimensional features. A convolutional autoencoder (CAE) maps these low-dimensional features back to the complete temperature field, achieving the reconstruction of the full-field temperature field from low-dimensional features. By jointly optimizing the composite loss function of reconstruction error and key area error, the prediction accuracy for fault-prone areas is significantly improved, ensuring sufficient accuracy of the reduced-order model in the area of ​​interest.

[0101] The DL-ROM reduced-order model in this embodiment can be seamlessly embedded into existing digital twin platforms, directly utilizing the platform's real-time data stream for online calculations. Through the digital twin platform's interactive interface, users can input or retrieve operational data of the GIS equipment at any time (such as the instantaneous current value during a short-circuit fault or the steady-state load after long-term operation). The DL-ROM model will instantly output the corresponding temperature field distribution results, displayed as a 3D cloud map bound to the equipment's 3D model. Maintenance personnel can intuitively observe the evolution of the internal temperature field, hotspot locations, and thermal stress trends through the platform's UI or VR devices, achieving a closed-loop "simulation-diagnosis" process and enabling timely maintenance decisions upon detecting anomalies. This domain-customized integration of deep learning rapid simulation technology significantly enhances the practicality and intelligent decision support capabilities of the digital twin system, representing an innovative fusion at the application level rather than a simple application of the basic deep learning algorithm itself.

[0102] This embodiment employs a temperature field calculation method combining a deep learning-based reduced-order model and digital twin technology. It extracts low-dimensional features from the high-dimensional operating parameters of the high-voltage switchgear using a deep feedforward neural network, and then maps these low-dimensional features to the temperature field using a convolutional autoencoder. Based on the temperature field distribution results obtained from the deep learning-based reduced-order model and the temperature field prediction results obtained from the digital twin, real-time simulation and prediction of the high-voltage switchgear temperature field are achieved. This enables real-time calculation of the multi-physics temperature field of the high-voltage switchgear based on digital twin technology, integrating temperature field "calculation-simulation-diagnosis-decision-making" into a single process. This allows for real-time status monitoring of the high-voltage switchgear, improving its operational safety and reliability.

[0103] Example 2

[0104] Embodiment 2 of the present invention introduces a temperature field monitoring system for high-voltage switchgear based on digital twins.

[0105] like Figure 3 The temperature field monitoring system for high-voltage switchgear based on digital twins, as shown, includes:

[0106] The acquisition module is configured to acquire a 3D model and real-time operating data of the high-voltage switchgear based on digital twins.

[0107] The prediction module is configured to obtain the temperature field prediction result of the high-voltage switchgear based on the acquired real-time operating data and the deep learning reduction model; wherein the deep learning reduction model includes a deep feedforward neural network for mapping the acquired real-time operating data from the operating parameter space to a low-dimensional feature space and a convolutional autoencoder for reconstructing the temperature field from the low-dimensional feature vector data output by the deep feedforward neural network.

[0108] The monitoring module is configured to associate the obtained temperature field prediction results with the three-dimensional model, perform interactive visualization of the temperature field of the high-voltage switchgear, and complete the monitoring of the temperature field of the high-voltage switchgear based on digital twin.

[0109] The detailed steps are the same as those of the high-voltage switchgear temperature field monitoring method based on digital twin provided in Example 1, and will not be repeated here.

[0110] Example 3

[0111] Embodiment 3 of the present invention provides a computer-readable storage medium.

[0112] A computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the high-voltage switchgear temperature field monitoring method based on digital twin as described in Embodiment 1 of the present invention.

[0113] The detailed steps are the same as those of the high-voltage switchgear temperature field monitoring method based on digital twin provided in Example 1, and will not be repeated here.

[0114] Example 4

[0115] Embodiment 4 of the present invention provides an electronic device.

[0116] An electronic device includes a memory, a processor, and a program stored in the memory and running on the processor. When the processor executes the program, it implements the steps in the high-voltage switchgear temperature field monitoring method based on digital twin as described in Embodiment 1 of the present invention.

[0117] The detailed steps are the same as those of the high-voltage switchgear temperature field monitoring method based on digital twin provided in Example 1, and will not be repeated here.

[0118] Example 5

[0119] Embodiment 5 of the present invention provides a computer program product.

[0120] A computer program product includes software code, wherein the program in the software code performs the steps of the high-voltage switchgear temperature field monitoring method based on digital twin as described in Embodiment 1 of the present invention.

[0121] The detailed steps are the same as those of the high-voltage switchgear temperature field monitoring method based on digital twin provided in Example 1, and will not be repeated here.

[0122] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0123] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0124] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0125] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0126] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0127] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0128] The above description is merely a preferred embodiment of this practice and is not intended to limit the scope of this practice. Various modifications and variations can be made to this practice by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this practice should be included within the protection scope of this practice.

Claims

1. A high-voltage switchgear temperature field monitoring method based on digital twinning, characterized in that, The method comprises the following steps: obtaining a three-dimensional model and real-time operation data of the high-voltage switching device based on digital twinning; in the process of obtaining the three-dimensional model of the high-voltage switching device, a three-dimensional geometric model of the high-voltage switching device is constructed according to the structure of the high-voltage switching device, mesh division of the three-dimensional geometric model is performed according to the structure type of the high-voltage switching device, coupling control equations of different physical fields are obtained based on the geometric model after mesh division, and a multi-physical field coupling model of the high-voltage switching device is constructed; full-order simulation calculation of the high-voltage switching device is performed according to the constructed multi-physical field coupling model, a multi-working-condition full-order simulation data set of the high-voltage switching device is constructed, and a three-dimensional model of the high-voltage switching device is obtained; obtaining a temperature field prediction result of the high-voltage switching device according to the obtained real-time operation data and a deep learning reduced-order model; wherein the deep learning reduced-order model comprises a deep feedforward neural network for mapping the obtained real-time operation data from a working condition parameter space to a low-dimensional feature space and a convolutional autoencoder for reconstructing a temperature field from low-dimensional feature vector data output by the deep feedforward neural network; the deep feedforward neural network is used to learn a nonlinear mapping relationship from a wide working condition parameter space of the high-voltage switching device to a low-dimensional feature space, input a working condition parameter vector in the obtained real-time operation data, and map the real-time operation data to the low-dimensional feature space through a network hidden layer adopting multiple full connection layers to obtain a low-dimensional feature representation of the real-time operation data; the convolutional autoencoder comprises an encoder convolutional neural network and a decoder convolutional neural network, and the obtained low-dimensional feature representation vector of the real-time operation data is converted into a reduced dimension feature under the action of the encoder convolutional neural network and the decoder convolutional neural network, a prediction from a temperature field parameter to a temperature field is completed, and temperature field reconstruction is realized; associating the obtained temperature field prediction result with the three-dimensional model, performing visual interaction of the temperature field of the high-voltage switching device, and completing monitoring of the temperature field of the high-voltage switching device based on digital twinning.

2. The high-voltage switchgear temperature field monitoring method based on digital twinning according to claim 1, characterized in that, The loss function of the deep learning reduced-order model is a weighted sum of a global error function of temperature field reconstruction and a key area error function; wherein the global error function adopts a mean square error of a reconstructed temperature field and an original simulation field, the reconstructed temperature field is a temperature field prediction result, the original simulation field is a multi-working-condition full-order simulation data, and the key area error function is an error term of a key part of the high-voltage switching device; the network parameters of the deep feedforward neural network and the convolutional autoencoder are adjusted through a back propagation algorithm to reduce the loss function of the deep learning reduced-order model; when the loss function of the deep learning reduced-order model reaches a preset loss threshold, the training of the deep learning reduced-order model is completed, and the mapping from the real-time operation data working condition parameter to the temperature field is completed.

3. The high-voltage switchgear temperature field monitoring method based on digital twinning according to claim 1, characterized in that, After the temperature field prediction result is associated with the three-dimensional model, a three-dimensional geometric model displaying temperature distribution of each part of the high-voltage switching device in real time is obtained in a cloud chart, the change curve of temperature data and hot spot temperature of a key part of the high-voltage switching device is displayed in combination with digital twinning, the temperature field evolution of the high-voltage switching device is viewed, and visual interaction of the high-voltage switching device is completed.

4. The high-voltage switchgear temperature field monitoring method based on digital twinning of claim 1, wherein, When the temperature of the high-voltage switching device abnormally rises, a temperature abnormality early warning is issued according to the visual interaction result of the high-voltage switching device in combination with the digital twin, the cause of the abnormal temperature rise is analyzed according to the temperature field prediction result, and the monitoring of the temperature field of the high-voltage switching device is completed.

5. A high-voltage switchgear temperature field monitoring system based on digital twinning, characterized in that, Comprise: An acquisition module configured to acquire a three-dimensional model and real-time operation data of a high-voltage switching device based on a digital twin; in the process of acquiring the three-dimensional model of the high-voltage switching device, a three-dimensional geometric model of the high-voltage switching device is constructed according to the structure of the high-voltage switching device, grid division of the three-dimensional geometric model is performed according to the structure type of the high-voltage switching device, coupling control equations of different physical fields are acquired based on the geometric model after grid division, and a multi-physical field coupling model of the high-voltage switching device is constructed; full-order simulation calculation of the high-voltage switching device is performed according to the constructed multi-physical field coupling model, a multi-working-condition full-order simulation data set of the high-voltage switching device is constructed, and the three-dimensional model of the high-voltage switching device is obtained; A prediction module configured to obtain a temperature field prediction result of the high-voltage switching device according to the acquired real-time operation data and a deep learning reduced-order model; wherein the deep learning reduced-order model comprises a deep feedforward neural network for mapping the acquired real-time operation data from a working condition parameter space to a low-dimensional feature space and a convolutional autoencoder for reconstructing a temperature field from a low-dimensional feature vector data output by the deep feedforward neural network; the deep feedforward neural network is used to learn a nonlinear mapping relationship from a wide working condition parameter space of the high-voltage switching device to a low-dimensional feature space, input a working condition parameter vector in the acquired real-time operation data, and map the real-time operation data to the low-dimensional feature space through a network hidden layer adopting multiple full connection layers to obtain a low-dimensional feature representation of the real-time operation data; the convolutional autoencoder comprises an encoder convolutional neural network and a decoder convolutional neural network, and the obtained low-dimensional feature representation vector of the real-time operation data is converted into a reduced dimension feature under the action of the encoder convolutional neural network and the decoder convolutional neural network, the prediction from a temperature field parameter to a temperature field is completed, and the temperature field reconstruction is realized; A monitoring module configured to associate the obtained temperature field prediction result with the three-dimensional model, perform visual interaction of the temperature field of the high-voltage switching device, and complete the monitoring of the temperature field of the high-voltage switching device based on the digital twin.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to realize the steps of the high-voltage switching device temperature field monitoring method based on the digital twin in any one of claims 1-4.

7. An electronic device comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, The processor executes the program to realize the steps of the high-voltage switching device temperature field monitoring method based on the digital twin in any one of claims 1-4.

8. A computer program product comprising software code, characterized in that, The program in the software code executes the steps of the high-voltage switching device temperature field monitoring method based on the digital twin in any one of claims 1-4.

Images