Bypass load transfer box car change

Abstract

The application provides a bypass load transfer box transformer vehicle, which comprises a vehicle body, a high-voltage bypass module, a voltage transformation module, a low-voltage power distribution module and a control unit; the control unit is integrated with a monitoring and control system based on digital twin driving; in the monitoring and control system, an edge computing layer dynamically reconstructs the electric field intensity distribution, thermal stress distribution and local discharge characteristic quantity inside an insulating component through multi-physical field coupling simulation, generates virtual perception data, and fuses the virtual perception data with physical sensor measured data to output an integrated evaluation index of the insulating state; a cloud platform layer is used for iterative optimization of the twin model; and a control layer automatically generates and executes optimal control actions according to a preset multi-objective self-healing strategy library. The application solves the problems of inaccurate state perception, artificial control and slow fault response of the existing box transformer vehicle by integrating the monitoring and control system based on digital twin driving in the control unit of the bypass load transfer box transformer vehicle.

Description

Technical Field

[0001] This invention belongs to the technical field of emergency power supply equipment, and in particular relates to a bypass load transfer transformer. Background Technology

[0002] Bypass load transfer transformers are the main mobile equipment for emergency repairs and load transfers in power distribution networks. They are widely used in scenarios such as power distribution network upgrades and fault repairs. By integrating high-voltage bypass, transformer, and distribution modules, they can achieve rapid bypass switching of power supply lines and ensure power supply continuity.

[0003] Existing monitoring and control systems for bypass load transfer transformers mostly collect only single physical quantities such as current and voltage, and can only achieve basic threshold alarms. They cannot effectively perceive the electric field, thermal stress, and other states inside the insulation components. The insulation status assessment is limited in scope and has low accuracy. At the same time, their control actions mostly rely on manual judgment and execution, lacking real-time status analysis and automatic control capabilities. Fault response is lagging, and it is difficult to predict the risk of insulation degradation in advance. Power outages are easily caused by local insulation faults, and after a fault occurs, the optimal self-healing strategy cannot be generated quickly. The load interruption time is long, and the equipment operation reliability and intelligence level are insufficient. Summary of the Invention

[0004] The purpose of this invention is to provide a bypass load transfer transformer substation, which aims to solve the technical problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution.

[0006] According to one embodiment of the present invention, a bypass load transfer transformer is provided. A bypass load transfer transformer includes a vehicle body, a high-voltage bypass module, a transformer module, a low-voltage power distribution module, and a control unit; The control unit integrates a monitoring and control system based on digital twin technology, which includes: A multimodal sensing layer is configured at the insulating components of the high-voltage bypass module and the transformer module to collect multi-dimensional physical quantities; The edge computing layer includes a PLC controller, which has a built-in processor and memory. The memory stores a twin model that is mapped to the transformer vehicle in real time and an edge inference program. When the processor executes the edge inference program, it receives real-time data from the multimodal sensing layer, dynamically reconstructs the electric field intensity distribution, thermal stress distribution and partial discharge characteristics inside the insulation component through multi-physics coupling simulation, generates virtual sensing data, and fuses it with the measured data from physical sensors to output a fused comprehensive evaluation index of insulation status. The cloud platform layer connects to the edge computing layer via a wireless communication module, stores historical operational data, and deploys digital twin models for iterative optimization of the twin models. The control layer, integrated into the edge computing layer, automatically generates and executes the optimal control action based on the preset multi-objective self-healing strategy library when the comprehensive evaluation index of the fused insulation state exceeds the preset safety threshold. This includes one or more combinations of switching to the backup circuit, adjusting the load distribution, and initiating local cooling.

[0007] Furthermore, the multimodal sensing layer includes a current sensor, a voltage sensor, a temperature sensor, a humidity sensor, a partial discharge sensor, and a vibration sensor.

[0008] Furthermore, the twin model is constructed using a hybrid-driven adaptive hierarchical order reduction modeling method, including: The full-order snapshot generation unit, configured in the cloud platform layer, generates a high-fidelity full-order snapshot dataset through finite element simulation for the three-dimensional electromagnetic field, thermal field, and force field coupling behavior of insulating components under different combinations of operating parameters, and constructs a sample space for multi-physics field coupling response. The hierarchical reduction unit is used to compress the full-order model of the cloud platform layer into a lightweight model; The error compensation unit, deployed in the edge computing layer, constructs a statistical model of the prediction error of the reduced-order model based on Gaussian process regression. When the PLC controller receives real-time sensor data, it uses the measured data to perform online error compensation on the output of the reduced-order model, thereby improving the accuracy of the virtual sensing data. The adaptive update unit for operating conditions triggers incremental model updates when it detects that the current operating condition exceeds the coverage of the snapshot sample space, or when the prediction error of the downgraded model continues to exceed a preset threshold.

[0009] Furthermore, the triggering of incremental model updates includes: the edge side uploading the measured data under the new operating conditions to the cloud platform; the cloud platform supplementing the full-order simulation snapshot through an adaptive sampling augmentation strategy; updating the modal basis using the incremental POD algorithm; and distributing the updated reduced-order model parameters to the edge side to achieve online evolution of the lightweight model.

[0010] Furthermore, the hierarchical order reduction unit includes: The first-level order reduction module is configured to use intrinsic orthogonal decomposition to perform global mode extraction on full-order snapshot data, and obtain the principal mode basis and corresponding order reduction coefficients of electromagnetic field, temperature field and stress field, respectively. The second-level order reduction module is configured to introduce discrete empirical interpolation to process the nonlinear term, select the optimal interpolation point set, and compress the computational dimension of the nonlinear term. The third-layer order reduction module is configured to use a radial basis function neural network to construct a nonlinear mapping relationship between the order reduction coefficients and the operating parameters, which is used to predict the physical field distribution under the new operating conditions.

[0011] Furthermore, in the edge computing layer, the virtual sensing data includes at least the location of the electric field distortion point inside the insulating material, the maximum electric field strength value, and the partial discharge initiation voltage, which cannot be directly measured. The multiphysics coupling simulation includes: A fast electric field reconstruction algorithm based on the finite element method calculates the electric field intensity distribution at various points inside the insulating component in real time based on measured voltage, current and geometric parameters. The temperature field reconstruction algorithm based on the thermal network model calculates the internal hot spot temperature in real time according to the measured temperature, load current and heat dissipation conditions.

[0012] Furthermore, the monitoring and control system also includes an analysis layer for topological association of the insulation system. This layer uses each insulation component in the high-voltage bypass module, transformer module, and low-voltage distribution module as a graph node, and electrical connection relationships and physical proximity relationships as graph edges to construct an insulation system topology graph. A graph neural network is then used to analyze the topology graph, identify the propagation path of insulation degradation between nodes, predict the risk of cascade failure, and output a warning of weak links.

[0013] Furthermore, the graph neural network is a graph attention network, which assigns different attention weights to different neighboring nodes and dynamically learns the coupling influence strength between the insulation states of each component; when an abnormal decrease in the insulation index of any node is detected, the degree of impact on its neighboring nodes is automatically analyzed, and the monitoring frequency of neighboring nodes is increased in advance.

[0014] Furthermore, in the control layer, the multi-objective self-healing strategy library contains a variety of preset strategies, and a reinforcement learning algorithm is used to optimize the strategy selection online; The control layer takes minimizing load interruption time, maximizing remaining lifetime, and minimizing energy consumption as its objective functions. Based on the comprehensive evaluation index of the current insulation status, the load importance level, and environmental conditions, it makes real-time decisions on the optimal self-healing action through a Q-learning network.

[0015] Furthermore, the self-healing actions performed by the control layer include: When the detected partial discharge characteristic exceeds the first threshold, the backup insulation circuit is activated and transfer control is executed. When the temperature of a hot spot exceeds the second threshold, the automatic temperature control system is activated and the load on that branch is dynamically reduced. When multiple components are detected to be at risk of cascading failure, active isolation and power supply topology reconstruction are performed.

[0016] Compared with the prior art, the beneficial effects of the bypass load transfer transformer of the present invention are: This invention integrates a digital twin-driven monitoring and control system into the control unit of a bypass load transfer transformer substation. By combining multimodal sensing, cloud-edge collaborative modeling, data fusion evaluation, and automatic self-healing control, it specifically addresses the technical problems of existing transformer substations, such as single monitoring dimensions, inaccurate state perception, reliance on manual control, and delayed fault response. Specifically, the edge computing layer incorporates a built-in twin model, which, combined with multiphysics coupling simulation, dynamically reconstructs the electric field, thermal stress, and partial discharge characteristics inside insulating components that cannot be directly measured. This generates virtual sensing data and fuses it with measured data, significantly improving the accuracy of state judgment. The edge computing layer and cloud platform... The edge layer, composed of multiple layers, enables real-time data processing and local control. The cloud platform layer completes historical data storage and model iteration optimization, realizing the iteration of the twin model and ensuring that the model always adapts to the actual operating conditions of the transformer substation. The control layer is integrated into the edge computing layer and has a pre-built multi-objective self-healing strategy library. When the insulation status index exceeds the threshold, it automatically generates and executes the optimal control action without manual intervention, realizing rapid automatic response and self-healing control of faults. This solves the problems of existing technologies such as reliance on manual control, delayed fault response, and lack of optimal self-healing strategies, significantly shortening the load interruption time and improving the continuity and operational reliability of emergency power supply for transformer substations. Attached Figure Description

[0017] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0018] In the attached diagram: Figure 1 This is a schematic diagram of the structure of a bypass load transfer transformer vehicle provided in an embodiment of the present invention; Figure 2 This is a structural block diagram of the monitoring and control system provided in an embodiment of the present invention; Figure 3 This is a structural block diagram of the twin model provided in an embodiment of the present invention.

[0019] The attached figures are labeled as follows: 100. Vehicle body; 101. High-voltage bypass module; 102. Transformer module; 103. Low-voltage power distribution module; 104. Control unit; 1041. Multimodal Perception Layer; 1042. Edge Computing Layer; 1043. Cloud Platform Layer; 1044. Control Layer; 10421, Full-order snapshot generation unit; 10422, Hierarchical order reduction unit; 10423, Error compensation unit; 10424, Operating condition adaptive update unit; 104221, First-level order reduction module; 104222, Second-level order reduction module; 104223, Third-level order reduction module. Detailed Implementation

[0020] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0022] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0023] Please refer to Figure 1 In one embodiment of the present invention, a bypass load transfer box-type vehicle is provided, including a vehicle body 100 and a high-voltage bypass module 101, a transformer module 102 and a low-voltage power distribution module 103 installed on the vehicle body 100. In this embodiment of the invention, the high-voltage bypass module 101 is a high-voltage power supply line bypass and switching device for the transformer substation vehicle. It can provide a high-voltage side circuit path and switching capability for emergency repair of the distribution network and load transfer. Its insulating component is the installation position of the multi-modal sensing layer. The insulation status under high-voltage conditions is monitored in real time through the multi-modal sensing layer. In this embodiment of the invention, the transformer module 102 is a voltage conversion device for the transformer car, used to realize the voltage transformation function from high voltage to low voltage. Its internal insulation components are subject to electromagnetic and thermal stress, and insulation status monitoring is required. In this embodiment of the invention, the low-voltage power distribution module 103 is a low-voltage side power distribution device for the transformer substation vehicle. It is used to distribute the transformed low-voltage electrical energy to different load terminals and together with the high-voltage bypass module 101 and the transformer module 102, it constitutes the power supply link of the transformer substation vehicle in this embodiment of the invention.

[0024] Please continue to refer to Figure 1 and Figure 2 The present invention also includes a control unit 104, which integrates a monitoring and control system. This monitoring and control system is built based on digital twin drive. By constructing a real-time mapping with the physical transformer vehicle, it can realize the data fusion of virtual perception and physical measurement, thereby improving the accuracy of the monitoring and control system.

[0025] Specifically, such as Figure 3 As shown, the monitoring and control system provided in this embodiment of the invention includes a multimodal sensing layer 1041, an edge computing layer 1042, a cloud platform layer 1043, and a control layer 1044. In one embodiment of the present invention, a multimodal sensing layer 1041 is provided, in which the sensing layer is disposed at the insulating components of the high-voltage bypass module 101 and the transformer module 102, for collecting multi-dimensional physical quantities; Preferred, The multimodal sensing layer 1041 includes a current sensor, a voltage sensor, a temperature sensor, a humidity sensor, a partial discharge sensor, and a vibration sensor. The current sensor is installed at the high voltage inlet and outlet terminals and on both sides of the bypass switch of the high voltage bypass module 101. The current sensor is also installed at the high voltage side inlet, low voltage side outlet terminals and winding insulation leads of the transformer module 102. These are the connection points between the insulating components and the conductive components, so that real-time current data around the insulating components can be collected, providing electrical parameters for the edge computing layer 1042 to reconstruct the electric field intensity distribution and determine whether there is leakage current or insulation degradation caused by overcurrent in the insulating components. Accordingly, the voltage sensor in this embodiment of the invention is installed in conjunction with the current sensor. That is, it is installed at both ends of the high-voltage bypass busbar and the insulating bushing of the high-voltage bypass module 101, and at the high-voltage winding insulation end and the tap changer insulation part of the transformer module 102. The voltage sensor in this embodiment is used to collect the real-time voltage value at both ends of the insulating component, and combined with the current data, it provides voltage parameters for the reconstruction of the electric field. It can be used to calculate the electric field intensity distribution inside the insulating component and determine whether there is electric field distortion. Furthermore, the temperature sensor provided by the present invention is installed in the insulating bushing of the high voltage bypass module 101, the insulating support of the bypass switch, the winding insulation layer of the transformer module 102, the iron core insulation component, the insulation bonding point of the tank wall, and other locations where the insulating components are prone to heat generation, in order to collect the real-time temperature of the insulating component body and its surroundings. Accordingly, the humidity sensor in this embodiment of the invention is installed inside the closed cabinet of the high-voltage bypass module 101 and the transformer module 102, and in the moisture-proof sealed cavity of the insulating component. It is preferentially deployed at the ventilation opening and the corner of the insulating component where condensation is easy to occur, and is used to collect humidity data of the environment in which the insulating component is located. In this embodiment of the invention, the partial discharge sensor is installed in locations where partial discharge is prone to occur, such as the high-voltage insulating bushing and the insulation of the bypass cable head of the high-voltage bypass module 101, and the winding insulating shell and the bushing flange insulation of the transformer module 102. The vibration sensor of the present invention is installed on the transformer body insulation base and winding insulation fixing parts of the transformer module 102, the bypass switch insulation operating mechanism and bus insulation support of the high voltage bypass module 101, and other parts that are prone to loosening or wear of insulation components due to vibration; it can collect vibration frequency and amplitude data of insulation components and related structures to determine whether there is physical damage to insulation components caused by mechanical vibration.

[0026] Please continue to refer to Figure 3 The monitoring and control system provided in this embodiment of the invention also includes an edge computing layer 1042, which includes a PLC controller. The PLC controller has a built-in processor and a memory. The memory stores a twin model that is mapped to the transformer vehicle in real time and an edge inference program. When the processor executes the edge inference program, it can receive real-time data from the multimodal sensing layer 1041 and dynamically reconstruct the electric field intensity distribution, thermal stress distribution and partial discharge characteristics inside the insulation component through multi-physics coupling simulation to generate virtual sensing data. The virtual sensing data is then fused with the measured data from the physical sensors to output a fused comprehensive evaluation index of the insulation status. Specifically, in this embodiment of the invention, the twin model stored in the memory is constructed using a hybrid-driven adaptive hierarchical reduction modeling method; specifically, the twin model includes a full-order snapshot generation unit 10421, a hierarchical reduction unit 10422, an error compensation unit 10423, and a working condition adaptive update unit 10424. In this embodiment of the invention, the full-order snapshot generation unit 10421 is configured in the cloud platform layer 1043. For the three-dimensional electromagnetic field, thermal field and force field coupling behavior of the insulating component under different working condition parameter combinations, a high-fidelity full-order snapshot dataset is generated through finite element simulation to construct a sample space for multi-physics field coupling response. The hierarchical reduction unit 10422 is used to compress the full-order model of the cloud platform layer 1043 into a lightweight model; Please continue to refer to Figure 3 In this embodiment of the invention, the hierarchical reduction unit 10422 includes a first-layer reduction module 104221, a second-layer reduction module 10422, and a third-layer reduction module 104223. In one implementation, the first-layer reduction module 104221 of the present invention is configured to perform global mode extraction on the full-order snapshot data using intrinsic orthogonal decomposition to obtain the principal mode basis and corresponding reduction coefficients for the electromagnetic field, temperature field, and stress field, respectively. The first-layer reduction module 104221 takes the high-fidelity full-order snapshot data of multi-physics fields (electromagnetic field, temperature field, and stress field) generated by the cloud platform layer 1043 as input, and uses the intrinsic orthogonal decomposition POD data dimensionality reduction method to perform global mode extraction on the full-order snapshot data. Through this process, the principal mode basis that can characterize the core features of each physical field is selected, and the reduction coefficients corresponding to each principal mode basis are solved, thereby achieving preliminary dimensionality compression of the full-order snapshot data and extracting the key feature information of each physical field. In one implementation, the second-layer order reduction module 104222 of the present invention is configured to introduce discrete empirical interpolation to process the nonlinear term, select the optimal interpolation point set, and compress the computational dimension of the nonlinear term. The second-layer order reduction module 104222 first analyzes the high-dimensional feature space of the nonlinear term based on the discrete empirical interpolation method. For example, by extracting the basis vectors obtained by empirical orthogonal decomposition (POD), it selects the optimal interpolation point set that can best represent the characteristics of the nonlinear term. The calculation of the original high-dimensional nonlinear term is mapped to the low-dimensional subspace corresponding to the optimal interpolation point set. Only the nonlinear term value at the interpolation point is calculated. Then, the approximate overall nonlinear term result is restored through interpolation reconstruction, thereby greatly compressing the computational dimension of the nonlinear term and significantly reducing the overall computational load while ensuring computational accuracy. In one implementation, the third-layer order reduction module 104223 of the present invention is configured to use a radial basis function neural network (RBFNN) to construct a nonlinear mapping relationship between the order reduction coefficients and the operating parameters, which is used to predict the physical field distribution under the new operating conditions. Compared with directly fitting the high-dimensional physical field, the physical field is first reduced in dimension by using the order reduction coefficients, and then the nonlinear relationship between the order reduction coefficients and the operating parameters is fitted by RBFNN. This reduces the computational complexity and ensures the accuracy of the nonlinear mapping by utilizing the local approximation characteristics of RBFNN.

[0027] The error compensation unit 10423 of this embodiment of the invention is deployed in the edge computing layer 1042. It constructs a statistical model of the prediction error of the reduced-order model based on Gaussian process regression. When the PLC controller receives real-time sensor data, it uses the measured data to perform online error compensation on the output of the reduced-order model, thereby improving the accuracy of the virtual sensing data. In the adaptive update unit 10424 provided in this embodiment of the invention, when it is detected that the current working condition exceeds the coverage of the snapshot sample space, or the prediction error of the reduced-order model continues to exceed a preset threshold, an incremental model update is triggered, including: the edge side uploads the measured data under the new working condition to the cloud platform, the cloud platform supplements the full-order simulation snapshot through an adaptive sampling augmentation strategy, updates the modal basis using an incremental POD algorithm, and sends the updated reduced-order model parameters to the edge side to realize the online evolution of the lightweight model.

[0028] In the edge computing layer 1042, the virtual sensing data includes at least the location of the electric field distortion point inside the insulating material, the maximum electric field strength value, and the partial discharge initiation voltage, which cannot be directly measured. In multiphysics coupling simulation, the present invention uses a fast electric field reconstruction algorithm based on the finite element method to calculate the electric field intensity distribution at each point inside the insulating component in real time based on the measured voltage, current and geometric parameters. In the electric field reconstruction algorithm, the measured electrical parameters and the geometric parameters of the insulating components are used as inputs. Through field discretization, equation construction and numerical solution, the electric field intensity distribution of each point inside the insulating component is reconstructed in real time, and key indicators such as electric field distortion points and maximum electric field intensity values ​​are identified. Among them, the measured electrical parameters are the real-time voltage value across the insulating component collected by the voltage sensor and the real-time current value around the insulating component collected by the current sensor; Geometric parameters are pre-stored in the edge computing layer memory as 3D geometric model parameters of the insulating components; Auxiliary parameters include the inherent electromagnetic properties of the insulating material, such as its dielectric constant and conductivity, as well as the power grid frequency under the current operating conditions; In the field discretization, the physical field of the insulating component is discretized by mesh. For example, dense mesh is used for key areas where the electric field of the insulating component changes drastically, and sparse mesh is used for areas where the electric field is evenly distributed. In the equation construction, the governing equations of the electric field distribution are constructed by combining the operating state of the insulating components, including the Poisson equation / Laplace equation. The dielectric constant of the insulating material and the measured electric / current boundary conditions are substituted to determine the electric field constraint relationship of each discrete grid node in the field. In the numerical solution, the finite element numerical solution method is used to discretize the governing equations, transforming the continuous electric field distribution problem into a system of algebraic equations with discrete grid nodes. At the same time, for the computational optimization algorithm of the edge computing layer, fast iterative methods such as the preprocessing conjugate gradient method are used to reduce the computation time and achieve the solution of electric field intensity values ​​in milliseconds. Finally, after obtaining the electric field strength values ​​of all grid nodes in this embodiment, the electric field strength data of any point in the field is completed by interpolation algorithm. Finally, the three-dimensional electric field strength distribution cloud map and indicators inside the insulating component (such as the specific value of electric field strength at each point, the coordinate position of the electric field distortion point, the maximum electric field strength value and its location) are output, thus completing the electric field reconstruction.

[0029] In multiphysics coupling simulation, this invention also uses a temperature field reconstruction algorithm based on a thermal network model to calculate the internal hot spot temperature in real time according to the measured temperature, load current and heat dissipation conditions. The thermal network model provided in this embodiment is used to set the insulating components and surrounding heat dissipation structures as thermal nodes, and the heat transfer paths between components as thermal branches. By establishing thermal balance equations to solve the temperature of each thermal node, the internal temperature field of the insulating components can be reconstructed in real time. In this invention, the characteristics of the insulation components of the high-voltage bypass module 101 and transformer module 102 of the bypass load transfer transformer substation are combined. The thermal network is constructed in a hierarchical and regional manner. The temperature field of the insulation components is reconstructed into three levels: the core heat-generating area, the insulation conduction area, and the heat dissipation boundary area. Each level is abstracted into several thermal nodes, which are connected by thermal branches. The thermal branches are the heat transfer channels between the thermal nodes. The thermal resistance of the branches is calculated according to the heat transfer type. When implemented, only the main heat transfer mode of the transformer substation insulation components is considered. Ultimately, this invention constructs a visualized thermal network topology in the edge computing layer using hot nodes as vertices and hot branches as edges. Based on the constructed thermal network topology, a node thermal balance equation is established, and the temperature value of each unknown hot node is obtained through numerical solution, thereby realizing the real-time calculation of the hot spot temperature inside the insulating component.

[0030] As can be seen, this invention, by embedding a twin model in the edge computing layer and combining it with multiphysics coupling simulation, can dynamically reconstruct electric fields, thermal stresses, and partial discharge characteristics inside insulating components that cannot be directly measured, generating virtual sensing data and fusing it with measured data, thus significantly improving the accuracy of state judgment. Please continue to refer to Figure 3 The monitoring and control system provided in this embodiment of the invention also includes a cloud platform layer 1043, which is connected to the edge computing layer 1042 through a wireless communication module, stores historical operating data and deploys a digital twin model for iterative optimization of the twin model; The edge computing layer and cloud platform layer of this invention form an edge side that enables real-time data processing and local control. The cloud platform layer completes historical data storage and model iteration optimization, realizing the iteration of the twin model and ensuring that the model always adapts to the actual operating conditions of the transformer substation. Please continue to refer to Figure 3 The monitoring and control system provided in this embodiment of the invention also includes a control layer 1044, which is integrated into the edge computing layer 1042. When the integrated insulation state comprehensive evaluation index exceeds the preset safety threshold, the optimal control action is automatically generated and executed according to the preset multi-objective self-healing strategy library. The control action includes one or more combinations of switching the backup circuit, adjusting the load distribution, and starting local cooling. In this embodiment of the invention, the control layer is integrated into the edge computing layer and has a pre-built multi-objective self-healing strategy library. When the insulation status index exceeds the threshold, the optimal control action is automatically generated and executed without manual intervention. This achieves rapid automatic response and self-healing control of faults, solving the problems of existing technologies such as manual control, delayed fault response, and lack of optimal self-healing strategies. It can effectively shorten the load interruption time and improve the continuity and operational reliability of emergency power supply for transformer substations.

[0031] Please continue to refer to Figure 3In one implementation of the present invention, the monitoring and control system further includes an analysis layer 1045 for topological association of the insulation system, which is used to construct an insulation system topology graph by taking each insulation component in the high-voltage bypass module 101, transformer module 102, and low-voltage power distribution module 103 as graph nodes, and electrical connection relationships and physical proximity relationships as graph edges; and further uses a graph neural network to analyze the topology graph, identify the propagation path of insulation degradation between nodes, predict the risk of cascade failure, and output a warning of weak links; The graph neural network provided in this embodiment of the invention is a graph attention network, which assigns different attention weights to different neighboring nodes and dynamically learns the coupling influence strength between the insulation states of each component; when an abnormal decrease in the insulation index of any node is detected, the degree of impact on its neighboring nodes is automatically analyzed, and the monitoring frequency of neighboring nodes is increased in advance.

[0032] Furthermore, in the control layer 1044, the multi-objective self-healing strategy library contains a variety of preset strategies, and the strategy selection is optimized online using a reinforcement learning algorithm; the control layer takes minimizing load interruption time, maximizing remaining lifetime, and minimizing energy consumption as objective functions, and makes the optimal self-healing action in real time through a Q-learning network based on the current insulation status comprehensive evaluation index, load importance level, and environmental conditions. Preferably, the self-healing actions performed by the control layer in this embodiment of the invention include, but are not limited to: When the partial discharge characteristic quantity is detected to exceed the first threshold, the backup insulation circuit is activated and the transfer control is executed. The first threshold is used to distinguish between normal insulation fluctuations and initial deterioration. First, the partial discharge background value of healthy equipment after the transformer substation is put into operation / maintenance is collected. Combined with the full life cycle operation data of the same type of equipment, the first threshold is determined based on the background value and statistical confidence interval, combined with fault tests and industry standards. When the hot spot temperature exceeds the second threshold, the automatic temperature control system is activated and the load on the branch is dynamically reduced. The second threshold is used to match the material's heat resistance limit and operating margin. The second threshold is obtained by combining the material grade and specification limit with dynamic correction of the operating conditions. When multiple components are detected to be at risk of cascading failure, active isolation and power supply topology reconstruction are performed.

[0033] This invention integrates a digital twin-driven monitoring and control system into the control unit of a bypass load transfer transformer substation. By combining multimodal perception, cloud-edge collaborative modeling, data fusion evaluation, and automatic self-healing control, it specifically addresses the technical problems of existing transformer substations, such as single monitoring dimensions, inaccurate state perception, reliance on manual control, and delayed fault response. The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0034] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0035] In the above embodiments of this application, the information collected is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data all comply with relevant laws, regulations and standards, take necessary protective measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0036] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0037] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0038] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0039] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A bypass load transfer box transformer, comprising a vehicle body (100), a high-voltage bypass module (101), a transformer module (102), a low-voltage power distribution module (103), and a control unit (104). Its features are: The control unit (104) integrates a monitoring and control system based on digital twin technology, the system including: A multimodal sensing layer (1041) is configured at the insulating components of the high-voltage bypass module (101) and the transformer module (102) for collecting multi-dimensional physical quantities; The edge computing layer (1042) includes a PLC controller, which has a built-in processor and memory. The memory stores a twin model that is mapped to the transformer vehicle in real time and an edge inference program. When the processor executes the edge inference program, it receives real-time data from the multimodal sensing layer (1041), dynamically reconstructs the electric field intensity distribution, thermal stress distribution and partial discharge characteristics inside the insulation component through multi-physics coupling simulation, generates virtual sensing data, and fuses it with the measured data of physical sensors to output a fused comprehensive evaluation index of insulation status. The cloud platform layer (1043) is connected to the edge computing layer (1042) through a wireless communication module, stores historical operation data and deploys digital twin models for iterative optimization of the twin models; The control layer (1044), integrated into the edge computing layer (1042), automatically generates and executes the optimal control action based on the preset multi-objective self-healing strategy library when the integrated insulation state comprehensive evaluation index exceeds the preset safety threshold. This includes one or more combinations of switching the backup circuit, adjusting the load distribution, and starting local cooling.

2. The bypass load transfer transformer substation according to claim 1, characterized in that, The multimodal sensing layer (1041) includes a current sensor, a voltage sensor, a temperature sensor, a humidity sensor, a partial discharge sensor, and a vibration sensor.

3. The bypass load transfer transformer substation according to claim 2, characterized in that, The twin model is constructed using a hybrid-driven adaptive hierarchical reduction modeling method, including: The full-order snapshot generation unit (10421) is configured in the cloud platform layer (1043). It generates a high-fidelity full-order snapshot dataset through finite element simulation to construct a sample space for multi-physics coupling response, targeting the three-dimensional electromagnetic field, thermal field, and force field coupling behavior of insulating components under different combinations of operating parameters. The hierarchical reduction unit (10422) is used to compress the full-order model of the cloud platform layer (1043) into a lightweight model; The error compensation unit (10423) is deployed in the edge computing layer (1042). It constructs a statistical model of the prediction error of the reduced-order model based on Gaussian process regression. When the PLC controller receives real-time sensor data, it uses the measured data to perform online error compensation on the output of the reduced-order model. The adaptive update unit (10424) for operating conditions triggers incremental model updates when it detects that the current operating condition exceeds the coverage of the snapshot sample space, or when the prediction error of the downgraded model continues to exceed a preset threshold.

4. The bypass load transfer transformer substation according to claim 3, characterized in that, The triggering of incremental model updates includes: the edge side uploading measured data under new operating conditions to the cloud platform; the cloud platform supplementing full-order simulation snapshots through an adaptive sampling augmentation strategy; updating the modal basis using an incremental POD algorithm; and distributing the updated reduced-order model parameters to the edge side to achieve online evolution of the lightweight model.

5. The bypass load transfer transformer substation according to claim 4, characterized in that, The hierarchical order reduction unit (10422) includes: The first-level order reduction module (104221) is configured to use intrinsic orthogonal decomposition to perform global mode extraction on full-order snapshot data to obtain the principal mode basis and corresponding order reduction coefficients of electromagnetic field, temperature field and stress field, respectively. The second-level order reduction module (104222) is configured to introduce discrete empirical interpolation to process the nonlinear term, select the optimal interpolation point set, and compress the computational dimension of the nonlinear term; The third-layer order reduction module (104223) is configured to use a radial basis function neural network to construct a nonlinear mapping relationship between the order reduction coefficients and the operating parameters, which is used to predict the physical field distribution under the new operating conditions.

6. The bypass load transfer transformer substation according to claim 5, characterized in that, In the edge computing layer (1042), the virtual sensing data includes at least the location of the electric field distortion point inside the insulating material, the maximum electric field strength value, and the partial discharge initiation voltage, which cannot be directly measured. The multiphysics coupling simulation includes: A fast electric field reconstruction algorithm based on the finite element method calculates the electric field intensity distribution at various points inside the insulating component in real time based on measured voltage, current and geometric parameters. The temperature field reconstruction algorithm based on the thermal network model calculates the internal hot spot temperature in real time according to the measured temperature, load current and heat dissipation conditions.

7. The bypass load transfer transformer substation according to any one of claims 2 to 6, characterized in that, The monitoring and control system also includes an analysis layer (1045) for topological association of the insulation system, which is used to construct an insulation system topology graph by taking each insulation component in the high voltage bypass module (101), transformer module (102), and low voltage power distribution module (103) as graph nodes, and electrical connection relationships and physical proximity relationships as graph edges. The graph neural network is used to analyze the topology, identify the propagation path of insulation degradation between nodes, predict the risk of cascade failure, and output early warning of weak links.

8. The bypass load transfer transformer substation according to claim 7, characterized in that, The graph neural network is a graph attention network, which assigns different attention weights to different neighboring nodes and dynamically learns the coupling influence strength between the insulation states of each component. When an abnormal decrease in the insulation index of any node is detected, the degree of impact on its neighboring nodes is automatically analyzed, and the monitoring frequency of neighboring nodes is increased in advance.

9. The bypass load transfer transformer substation according to claim 8, characterized in that, In the control layer (1044), the multi-objective self-healing strategy library contains a variety of preset strategies, and the strategy selection is optimized online using a reinforcement learning algorithm; The control layer takes minimizing load interruption time, maximizing remaining lifetime, and minimizing energy consumption as its objective functions. Based on the comprehensive evaluation index of the current insulation status, the load importance level, and environmental conditions, it makes real-time decisions on the optimal self-healing action through a Q-learning network.

10. The bypass load transfer transformer substation according to claim 9, characterized in that, The self-healing actions performed by the control layer include: When the detected partial discharge characteristic exceeds the first threshold, the backup insulation circuit is activated and transfer control is executed. When the temperature of a hot spot exceeds the second threshold, the automatic temperature control system is activated and the load on that branch is dynamically reduced. When multiple components are detected to be at risk of cascading failure, active isolation and power supply topology reconstruction are performed.

Images