A gantry state intelligent recognition and three-dimensional visualization linkage method

Abstract

This invention relates to the field of computer graphics and 3D model processing, and discloses a method for intelligent identification and 3D visualization linkage of transformer substation status. The method includes the following steps: constructing an initial 3D model and establishing a spatial location mapping relationship; acquiring multi-source monitoring data and dynamically scheduling transmission channels; performing multi-dimensional correlation analysis on the monitoring data, extracting rendering features, and outputting fault classification and location results; converting the features into dynamic rendering attributes, driving the generation of a 3D linkage model and embedding it into a GIS map; and triggering multi-level early warning visualization based on risk assessment, and outputting closed-loop linkage control signals. This invention deeply integrates AI algorithms with the graphics rendering pipeline, achieving precise mapping and dynamic manipulation of physical entities and digital bases, effectively eliminating false alarms due to environmental interference, and realizing a seamless closed loop from high-dimensional visualization to physical hard isolation tripping, significantly improving the proactive safety defense and intelligent operation and maintenance capabilities of transformer substations.

Description

Technical Field

[0001] This invention relates to the field of computer graphics and 3D model processing technology, specifically to a method for intelligent recognition and 3D visualization linkage of bench changing states. Background Technology

[0002] As a crucial power hub node at the end of the distribution network, the operating status of the benchtop transformer directly affects the reliability of regional power supply. Currently, the most similar existing implementation to this invention is a "benchtop transformer status monitoring device based on single image recognition." This device mainly collects images of the equipment's appearance by installing a fixed camera near the benchtop transformer, and uses basic image recognition algorithms such as edge detection and template matching to identify the status of key components such as the transformer body and terminals in the images to determine whether there are any problems such as missing components or obvious damage. The identification results are then uploaded to the backend management terminal in the form of simple image annotations.

[0003] While the existing solutions mentioned above reduce reliance on manual inspections to some extent, their core flaws lie in their extremely limited perception dimensions and simplistic recognition algorithms. Relying solely on basic image recognition algorithms, this solution struggles to meet the recognition needs of complex scenarios. For example, its accuracy drops significantly in low-light conditions at night or in adverse weather conditions such as fog or rain, where image blurring causes interference. Furthermore, this solution is completely incapable of detecting non-visual, hidden defects such as abnormal internal temperatures or partial discharges, failing to comprehensively reflect the true operating status of the transformer. More critically, this solution lacks data fusion and intelligent deep analysis capabilities. Because it does not integrate equipment operating parameters obtained from underlying sensors such as temperature, current, and vibration, the system cannot achieve multi-dimensional cross-modal data correlation analysis of "image + physical parameters." This data silo phenomenon makes it difficult for the system to accurately determine the physical cause of faults. For example, simply detecting discoloration on the terminal surface through an image cannot determine whether it's due to system overload causing overall overheating or localized poor contact leading to heat generation. This deficiency means the device can only passively identify faults after they occur, completely losing its proactive fault warning function.

[0004] Furthermore, the existing device suffers from severe visualization deficiencies in human-computer interaction, resulting in a significant lack of ease of operation and maintenance. Its backend management interface can only present a flattened view using two-dimensional static images overlaid with simple text descriptions, completely lacking high-fidelity 3D model spatial mapping, cross-dimensional operational data time-series trend charts, and macro-level fault location maps, among other high-dimensional visualization methods. This extremely low level of visualization and dimensional fragmentation makes it difficult for operation and maintenance personnel to intuitively and quickly grasp the overall 3D operational status of the transformer in the digital space, leading to extremely low efficiency in fault tracing, location, and on-site handling. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for intelligent identification and three-dimensional visualization linkage of test bench status. This method solves the technical problems of existing test bench monitoring technologies, which are limited by a single perception dimension, lack of in-depth fusion analysis of multi-source heterogeneous data, and limitation by two-dimensional static display, thus failing to achieve high-precision fault early warning and location and three-dimensional spatial dynamic closed-loop linkage control in complex environments.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of the present invention provides a method for intelligent identification and three-dimensional visualization linkage of test bench changing states, comprising: Computer-aided design technology is used to construct an initial three-dimensional model of the bench transformer, and the three-dimensional spatial position mapping relationship of each three-dimensional component in the initial three-dimensional model is established. The multi-source monitoring data of the transformer platform is used as the data source for updating the three-dimensional scene. The multi-source monitoring data includes image data containing visible light images and infrared images, electrical parameter data, and environmental data acquired by a dual-spectrum camera. Intelligent identification and correlation analysis are performed on the multi-source monitoring data to extract three-dimensional rendering features and generate fault classification and location results of the transformer and temperature distribution heat map of the physical components of the transformer. The multi-source monitoring data, the temperature distribution heat map of the physical components of the transformer, and the fault classification and location results are transformed into three-dimensional dynamic rendering attributes. Based on the three-dimensional spatial position mapping relationship, they are fused and mapped to the corresponding three-dimensional spatial position of the initial three-dimensional model, driving the initial three-dimensional model to perform visual dynamic overlay rendering to generate a three-dimensional linkage model of the transformer. Based on the preset three-dimensional spatial early warning triggering conditions in the aforementioned three-dimensional linkage model of the test bench, a three-dimensional visualization of a multi-level early warning mechanism is executed, and the corresponding closed-loop linkage control signal is output according to the fault classification and location results.

[0007] In one specific implementation, the step of converting the multi-source monitoring data, the temperature distribution heat map of the physical components of the transformer, and the fault classification and location results into three-dimensional dynamic rendering attributes, fusing and mapping them to the corresponding three-dimensional spatial positions of the initial three-dimensional model based on the three-dimensional spatial position mapping relationship, and driving the initial three-dimensional model to perform visual dynamic overlay rendering to generate a three-dimensional linkage model of the transformer includes: The initial 3D model is rendered using a 3D rendering engine at a preset ratio, and the 3D viewpoint of the initial 3D model is adjusted to the target angle in response to interactive commands. At the three-dimensional spatial location of the initial three-dimensional model, the corresponding electrical parameter data is superimposed in real time as a three-dimensional floating label, and the temperature distribution heat map of the physical components of the transformer is mapped onto the surface of the corresponding three-dimensional component as a material texture; according to the fault classification and location results, the location of the faulty three-dimensional component is highlighted in the initial three-dimensional model, and the faulty transformer three-dimensional linkage model is displayed with alarm flashing on the panoramic three-dimensional map in conjunction with the geographic information system.

[0008] In one specific implementation, the step of intelligently identifying and correlating the multi-source monitoring data, extracting three-dimensional rendering features, and generating fault classification and location results for the transformer and a temperature distribution heatmap of the physical components of the transformer includes: The visible light image in the image data is analyzed using a deep learning object detection algorithm to identify abnormal appearances of the physical components of the transformer. Temperature field extraction is performed on the infrared image in the image data to calculate the temperature values ​​of the physical components of the transformer and generate a temperature distribution heat map of the physical components. Specifically, the radiant energy on the equipment surface is obtained using the principle of infrared radiation thermometry and converted into an actual temperature field based on calibration parameters. The transformation relationship satisfies: in, This represents the total radiant energy received by the infrared sensor. The emissivity of the surface of the component being tested. Atmospheric transmittance, For environmental background radiation, Here, represents the Stefan-Boltzmann constant. Trend analysis is performed on the electrical parameter data based on a Long Short-Term Memory (LSTM) network time-series prediction model to identify abnormal states of the electrical parameters. Specifically, the time-series electrical parameters are processed using the cell state update mechanism of the LTM network, with its forgetting gate vector... The calculation model is as follows: in, It is the Sigmoid activation function. Here is the weight matrix for the forget gate. This is the hidden state from the previous moment. Input the electrical parameters for the current moment. This is the bias term. A multi-factor logistic regression model is used to perform correlation analysis between the environmental data and the electrical parameter data to determine the impact of environmental factors on the operation of the transformer. Specifically, let the feature vector formed by the combination of environmental data and electrical data be . The probability of environmental factors causing operational abnormalities satisfy: in, The intercept is... For corresponding feature components The regression coefficients.

[0009] Based on the decision tree algorithm, the system integrates the abnormal appearance of the physical components of the transformer, the temperature distribution heatmap of the physical components, the abnormal electrical parameters, and the impact of environmental factors on the operation of the transformer to automatically classify fault types. Combined with the location coordinates of the data acquisition equipment collecting the multi-source monitoring data, the fault classification and location results are calculated. Specifically, the decision tree performs feature selection based on information gain or the Gini coefficient when splitting nodes. Let the dataset... The empirical entropy is : in, For the data belonging to the first The model determines the final classification leaf node by maximizing the decrease in information entropy based on the proportion of samples of the fault type.

[0010] In one specific implementation, the three-dimensional visualization presentation of the multi-level early warning mechanism based on the preset three-dimensional spatial early warning triggering conditions in the three-dimensional linkage model of the test bench, and the output of the corresponding closed-loop linkage control signal according to the fault classification and location results, includes: Based on the preset risk level corresponding to the fault classification and location results, the warning level is divided into general warning, important warning, and emergency warning, and each warning level is presented in different colors or animation effects in the three-dimensional linkage model of the transformer platform. When the emergency warning is triggered, a closed-loop linkage control signal is generated and output to automatically link the preset protection device in the transformer platform to cut off the faulty line. The operation and maintenance work order containing the three-dimensional spatial coordinates of the fault location, the fault type, and the handling suggestions is automatically generated through the mobile communication network and dispatched to the mobile terminal of the operation and maintenance personnel.

[0011] In one specific implementation, the acquisition of multi-source monitoring data from the transformer substation is used as a data source for updating the three-dimensional scene. The multi-source monitoring data includes image data (containing visible light and infrared images), electrical parameter data, and environmental data acquired via a dual-spectrum camera. The specific acquisition process includes: The image data is continuously acquired at a preset cycle by the dual-spectrum camera with gimbal control function. The current, voltage, and power parameters of the high-voltage side and low-voltage side of the transformer are obtained as the electrical parameter data; The temperature, humidity, wind speed, rainfall, and smoke concentration of the environment in which the transformer is located are obtained as environmental data.

[0012] In one specific implementation, the multi-source monitoring data is acquired using a dual-mode data exchange system, employing both wireless network transmission and fiber optic wired transmission. Furthermore, when the network is interrupted, the multi-source monitoring data is temporarily stored locally using edge computing nodes, and the interrupted data is retransmitted after the network is restored to ensure the data integrity for 3D visualization rendering.

[0013] A second aspect of the present invention provides a linkage device for intelligent identification and three-dimensional visualization of changing states of a test bench, comprising: The three-dimensional construction module is used to construct the initial three-dimensional model of the bench transformer using computer-aided design technology, and to establish the three-dimensional spatial position mapping relationship of each three-dimensional component in the initial three-dimensional model. The data acquisition module is used to acquire multi-source monitoring data of the transformer as a data source for updating the three-dimensional scene. The multi-source monitoring data includes image data containing visible light and infrared images, electrical parameter data, and environmental data acquired by a dual-spectrum camera. The collaborative analysis module is used to intelligently identify and correlate the multi-source monitoring data, extract three-dimensional rendering features, and generate the fault classification and location results of the transformer and the temperature distribution heat map of the physical components of the transformer. The 3D rendering manipulation module is used to convert the multi-source monitoring data, the temperature distribution heat map of the physical components of the transformer, and the fault classification and location results into 3D dynamic rendering attributes, and to fuse and map them to the corresponding 3D spatial position of the initial 3D model based on the 3D spatial position mapping relationship, thereby driving the initial 3D model to perform visual dynamic overlay rendering and generating a 3D linkage model of the transformer. The early warning linkage module is used to perform a three-dimensional visualization of a multi-level early warning mechanism based on the preset three-dimensional spatial early warning trigger conditions in the three-dimensional linkage model of the test bench, and output the corresponding closed-loop linkage control signal according to the fault classification and location results.

[0014] In one specific implementation, the 3D rendering manipulation module is specifically used to: render the initial 3D model using a 3D rendering engine at a preset ratio, and adjust the 3D perspective of the initial 3D model at a target angle in response to interactive commands; At the three-dimensional spatial location of the initial three-dimensional model, the corresponding electrical parameter data is superimposed in real time in the form of three-dimensional floating labels, and the temperature distribution heat map of the physical components of the transformer is mapped onto the surface of the corresponding three-dimensional component as a material texture. Based on the fault classification and location results, the location of the faulty three-dimensional component is highlighted in the initial three-dimensional model, and the faulty test bench three-dimensional linkage model is displayed with alarm flashing on the panoramic three-dimensional map in conjunction with the geographic information system.

[0015] A third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a bench state change intelligent recognition and three-dimensional visualization linkage method as described in any of the above specific embodiments.

[0016] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the intelligent identification and three-dimensional visualization linkage method for bench state change as described in any of the above specific embodiments.

[0017] This invention provides a method for intelligent identification and 3D visualization linkage of test bench changing states. It has the following beneficial effects: 1. Constructing a multi-dimensional perception and holographic state characterization mechanism to eliminate monitoring blind spots. Breaking away from the limitations of traditional benchtop transformer monitoring methods that rely on a single sensor, this solution deeply integrates visible light and infrared thermal imaging images from dual-spectrum cameras, electrical operating parameters, and on-site climate indicators to construct a multi-source heterogeneous underlying perception network. This mechanism can not only keenly detect physical damage or foreign object intrusion on the equipment surface, but also penetrate the surface to observe abnormal fluctuations in the internal temperature field and current load, and trace the causes of harsh environments. Cross-validation of multi-dimensional data completely eliminates the risks of missed or false judgments that are easily caused by single-dimensional monitoring, achieving a comprehensive and holographic understanding of the equipment's operational health status.

[0018] 2. Proactive defense and early warning based on deep temporal evolution, reshaping operation and maintenance strategies. Unlike the passive response mode of post-event alarms, this method cleverly introduces temporal prediction models such as Long Short-Term Memory (LSTM) networks and decision tree classification architectures, endowing the system with the ability to uncover hidden degradation trends from massive historical and real-time streaming data. By predicting and integrating key indicators such as abnormal terminal temperature rise and load shift in advance, the system can accurately pinpoint weak points and present tiered alarms with visual animations before substantial equipment downtime or damage occurs. This technology significantly advances the defense line of power grid operation and maintenance, truly promoting a substantial leap in management from "locking the stable door after the horse has bolted" to "prevention before the event."

[0019] 3. Create an intuitive and efficient 3D digital twin interactive experience, eliminating information silos. Addressing the pain points of traditional maintenance backend interfaces—fragmented data and heavy reliance on manual information piecing together to "fill in the gaps" in the field—this technology, relying on computer-aided design and a 3D rendering engine, reconstructs a digital twin of the equipment in virtual space. By transforming the originally tedious current and voltage readings into spatially floating labels, dynamically mapping the hidden thermodynamic distribution to the material texture of the equipment surface, and supplementing this with precise highlighting of fault nodes, maintenance personnel gain "X-ray vision." This method of instantly reading cross-dimensional high-level information on a panoramic 3D canvas greatly reduces the preparation cycle for manual judgment and decision-making.

[0020] 4. Establish a seamless hardware and software closed-loop scheduling mechanism to accelerate fault response. To eliminate the operational gap between data analysis and on-site handling, this invention connects the intelligent status recognition to equipment action execution. Once the spatial warning trigger condition is determined to be an emergency risk, the system will not only immediately link the underlying protection device to issue a hard isolation command to cut off the faulty line, but also simultaneously utilize the mobile communication network to send a customized work order containing precise three-dimensional spatial coordinates, confirmed fault type, and repair strategy directly to the mobile terminals of frontline personnel. This coherent action of "algorithm decision-making - automatic isolation - precise work order dispatch" significantly reduces downtime and operational gaps caused by power grid anomalies.

[0021] 5. Provides a robust network and flexible evolution base for extreme operating conditions, ensuring long-term reliable operation. Considering that outdoor test benches are increasingly deployed in remote areas susceptible to severe weather or signal attenuation, this device features an adaptive communication routing system with both fiber optic and wireless links at the underlying architecture. Furthermore, it incorporates local storage and breakpoint resumption mechanisms at the front-end edge computing nodes. This ensures that even under extreme conditions such as communication interruptions or strong interference, the data link for updating the 3D scene remains absolutely intact. In addition, the decoupled software architecture allows the system to seamlessly iterate on core fault identification algorithms via the cloud, enabling it to readily accommodate the access and diagnostic needs of future new power components. Attached Figure Description

[0022] Figure 1 This is an overall structural block diagram of the device of the present invention; Figure 2 This is a schematic diagram of the method flow of the present invention; Figure 3 This is a schematic diagram of the spatial deployment of the multi-source data acquisition module of the present invention; Figure 4 This is a flowchart of the multidimensional intelligent analysis algorithm logic of the core control module of the present invention; Figure 5 This is a schematic diagram of the three-dimensional dynamic rendering pipeline and visualization interface of the present invention; Figure 6 This is a timing diagram for the fault warning and closed-loop linkage control of the present invention; Figure 7 This is a block diagram of the underlying hardware architecture of the electronic device of the present invention. Detailed Implementation

[0023] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Please see the appendix Figure 1 - Appendix Figure 7 This invention provides a method for intelligent identification and 3D visualization linkage of transformer status, comprising the following steps: constructing an initial 3D model of the transformer using computer-aided design technology, and establishing the 3D spatial position mapping relationship of each 3D component in the initial 3D model. When constructing a multi-dimensional digital mapping system, a precise association between the physical equipment topology and the virtual scene space is established through homogeneous coordinate transformation. A calculation model for the transformation from the local coordinate system to the world coordinate system is established, and its mathematical expression is: in, Let be the homogeneous spatial position vector of the 3D component in a unified world coordinate system. Let be the local spatial coordinate vector of the 3D component in its independently modeled state. This is a 3D affine transformation matrix containing spatial translation, rotation, and scaling parameters. This mapping provides a geometric reference for subsequent spatial anchoring of multi-source data.

[0025] Multi-source monitoring data from the transformer substation is used as the data source for updating the 3D scene. This multi-source monitoring data includes image data (containing visible light and infrared images) acquired through a dual-spectrum camera, electrical parameter data, and environmental data. This process, through the array deployment of a multimodal sensor network, overcomes the limitations of single-sensor operation, enabling the capture of holographic information from equipment appearance and internal operating mechanisms to external climatic interference factors.

[0026] Intelligent identification and correlation analysis are performed on multi-source monitoring data to extract 3D rendering features and generate fault classification and location results for the transformer, as well as temperature distribution heat maps of the transformer's physical components. The collected heterogeneous data sources are processed through a built-in deep network and predictive classification algorithm for high-dimensional feature extraction and cross-comparison, transforming the chaotic sensor signals at the bottom layer into structured fault coordinate vectors and continuous thermal radiation matrices, thus establishing the 3D spatial attributes of the graphic to be rendered.

[0027] Multi-source monitoring data, temperature distribution heatmaps of the transformer's physical components, and fault classification and location results are transformed into 3D dynamic rendering attributes. Based on 3D spatial location mapping relationships, these attributes are fused and mapped to the corresponding 3D spatial locations of the initial 3D model, driving the initial 3D model to perform visual dynamic overlay rendering and generating a 3D linked model of the transformer. Utilizing a 3D graphics rendering pipeline engine, abstract operation and maintenance parameters are reconstructed into material texture maps and spatial lighting effects, achieving a high-fidelity projection of the equipment's physical degradation state into a computer digital space image.

[0028] Based on the preset 3D spatial early warning trigger conditions in the 3D linkage model of the test bench, the system implements a multi-level early warning mechanism through 3D visualization, and outputs corresponding closed-loop linkage control signals according to the fault classification and location results. Within the 3D visualization management platform, the system visualizes potential hazards in the form of animated evolution, and simultaneously issues hard isolation commands to the site according to the risk level and triggers work order flow, completing a closed-loop proactive prevention process from state perception and digital twin decision-making to physical action intervention.

[0029] This embodiment details the construction logic of the digital base and the highly reliable data acquisition mechanism of the intelligent identification and 3D visualization linkage device for transformer status. To achieve a holographic presentation of the transformer's operating status, an initial 3D model of the transformer is first constructed using computer-aided design technology. The construction process strictly adheres to the engineering drawing dimensions of the physical entity. A high-fidelity digital model is generated by finely dividing key electrical components such as the transformer body, bushings, terminals, and circuit breakers into polygonal meshes. Based on this, a 3D spatial mapping relationship is established for each 3D component in the initial 3D model. The polygonal surface topology of each component is associated and bound with a unique device identifier, thereby constructing a virtual carrier addressable via a coordinate system, providing geometric anchor points for the accurate alignment of subsequent real-time monitoring data.

[0030] In this invention, multi-source monitoring data of the transformer substation is used as the data source for updating the 3D scene. This acquisition process relies on a sensor matrix deployed in an array on-site. Specifically, image data is continuously acquired at a preset cycle using a dual-spectrum camera with pan-tilt control functionality. The dual-spectrum camera integrates visible light and infrared detection channels. The visible light lens is responsible for capturing physical morphological changes such as damage, corrosion, and foreign object obstruction on the surface of components; the infrared thermal imaging lens penetrates the surface and simultaneously acquires the thermal radiation energy distribution on the equipment surface, providing a basic image matrix for quantitative inversion of the temperature field. The pan-tilt control mechanism is controlled by backend commands, achieving comprehensive visual coverage of the transformer substation in all directions.

[0031] In terms of electrical operation monitoring, the system acquires current, voltage, and power parameters from the high-voltage and low-voltage sides of the transformer as electrical parameter data. Underlying electrical sensors are directly coupled to the transformer's input and output cables, enabling high-frequency sampling of electrical parameters to accurately reflect the actual load level of the equipment and the power quality of the grid. Simultaneously, the system acquires environmental data such as temperature, humidity, wind speed, rainfall, and smoke concentration in the environment surrounding the transformer. This meteorological and disaster precursor information establishes necessary boundary conditions for subsequently filtering out environmental interference and determining the actual causes of faults, ensuring the robustness of the analysis model.

[0032] Considering that outdoor monitoring stations are increasingly deployed in areas with strong electromagnetic interference, harsh weather conditions, and weak network coverage, multi-source monitoring data is exchanged using a dual-mode approach: wireless network transmission and fiber optic wired transmission, during acquisition and uploading. The system's underlying communication gateway continuously assesses the overall quality of the current network link and dynamically schedules transmission channels. Current network quality index. The computational model satisfies: in, This represents the packet loss rate of the current communication link. This is the maximum packet loss rate threshold that the system can tolerate. Currently available bandwidth, Minimum bandwidth limit required for concurrent transmission of multimodal data; For end-to-end latency of network transmission, This is the delay penalty coefficient; , , These are the weighting coefficients of packet loss rate, available bandwidth, and transmission delay in the comprehensive evaluation model, respectively. When the link maintenance threshold is lower than the preset baseline, the communication gateway seamlessly switches to fiber optic wired transmission mode to ensure stable data output.

[0033] Furthermore, in situations where extreme weather causes both network paths to be interrupted, the system utilizes front-end edge computing nodes to temporarily store multi-source monitoring data locally. These edge computing nodes employ a ring-shaped buffer queue mechanism to manage high-frequency time-series data, prioritizing the retention of critical data frames carrying abnormal fluctuation characteristics using an importance assessment algorithm. Once network connectivity is restored, the edge computing nodes immediately initiate a breakpoint retransmission procedure, pushing the temporarily stored data packets to the backend cloud in chronological order based on internally generated timestamp sequences and data checksums. This mechanism completely eliminates data gaps caused by communication blind spots, ensuring the integrity and continuity of data for 3D visualization rendering, and enabling the operations and maintenance management center to trace the historical degradation trajectory of transformers without loss.

[0034] This embodiment details the in-depth analysis logic for intelligent identification and correlation analysis of multi-source monitoring data, extraction of 3D rendering features, and final generation of fault classification and location results. First, a deep learning object detection algorithm is used to analyze visible light images in the image data to identify abnormal appearances of the transformer's physical components. A convolutional neural network is then used to extract high-dimensional geometric features such as edges and textures from the visible light images, outputting bounding boxes and classification confidence scores for abnormal phenomena such as bushing damage, loose terminals, and foreign object obstruction, thus establishing a baseline assessment of the physical component's appearance health status.

[0035] Furthermore, this invention extracts the temperature field from the infrared image data, calculates the temperature values ​​of the physical components of the transformer, and generates a thermal map of the temperature distribution of the physical components. Specifically, it obtains the radiant energy of the equipment surface in the infrared band using the principle of infrared radiation thermometry, and converts it into an actual two-dimensional temperature field based on calibration parameters. The mathematical model of its transformation relationship satisfies: in, Spatial coordinates received by the infrared sensor target surface Total radiant energy at the location, The target emissivity of the surface of the physical component being measured. The atmospheric transmittance between the infrared measurement window and the target. Background radiation energy, This is the Stefan-Boltzmann constant. Using this model, the system filters out ambient thermal background noise and accurately inverts the high-temperature accumulation regions at key nodes.

[0036] In the electrical operation dimension, the system uses a Long Short-Term Memory (LSTM) network time-series prediction model to perform trend analysis on electrical parameter data and identify abnormal states of electrical parameters. The LTM network processes current, voltage, and power fluctuation data with long-sequence dependencies through its internal gating mechanism. When processing data at the current moment, it uses a forget gate to determine the degree of retention of historical state information; the forget gate vector... The calculation model is as follows: in, It is the Sigmoid activation function. The weight matrix set for the forget gate. Let be the hidden state vector output by the network at the previous time step. The sequence of multidimensional electrical parameters input at the current moment. This is the bias term. The model compares the residuals of the output predicted parameter trajectory with the actual collected values ​​to determine whether there are any abnormal states in the electrical parameters that deviate from the normal trajectory.

[0037] To isolate the non-fault-related interferences of meteorological conditions on equipment operation, this invention employs a multi-factor logistic regression model to correlate environmental data with electrical parameter data, determining the impact of environmental factors on the operation of the transformer test bench. The real-time collected environmental data, such as temperature, humidity, and wind speed, combined with electrical parameters, constitute a joint feature vector. The probability that environmental factors are the primary cause of equipment operating state deviation. satisfy: in, Characterizing the occurrence of environmental disturbance events, The intercept term of the regression model, For the corresponding joint feature components The regression coefficients are used to quantify the contribution of external climate to transformer oil temperature rise or insulation performance fluctuations.

[0038] Ultimately, the system, based on a decision tree algorithm, integrates the abnormal appearance of the physical components of the transformer, temperature distribution heatmaps, abnormal electrical parameters, and the impact of environmental factors on the transformer's operation to automatically classify fault types. During the construction and node splitting of the decision tree, core features are selected based on an information gain mechanism. Let's assume a multi-dimensional feature fusion dataset. The empirical entropy is : in, The total number of predefined benchtop transformer fault types in the system. For the data belonging to the first The algorithm determines the proportion of samples belonging to each fault type by maximizing the decrease in empirical entropy, until the leaf node outputs the determined fault type. Subsequently, by combining the spatial coordinates of the sensor devices collected from multi-source monitoring data, high-precision fault classification and localization results are output using spatial geometric calculations, providing accurate data anchors and attribute inputs for subsequent 3D spatial dynamic rendering.

[0039] This embodiment details the specific operational logic for transforming multi-source monitoring data, temperature distribution heatmaps of the transformer's physical components, and fault classification and location results into 3D dynamic rendering attributes, and driving the generation of a 3D linkage model of the transformer. First, relying on the built-in 3D graphics rendering pipeline engine, an initial 3D model containing polygonal meshes and basic materials is loaded. To achieve free interaction and precise observation from the user's perspective, a virtual camera's observation matrix and projection matrix are constructed, projecting the 3D component positions in the world coordinate system onto the 2D screen viewport. The homogeneous transformation calculation model of its vertex coordinates satisfies: in, These are the clipping space coordinates before rasterization by the graphics pipeline. To define the orthogonal or perspective projection matrix of the field of view and the depth of field frustum, The view matrix is ​​determined by the spatial position and viewing orientation of the virtual camera. The transformation matrix of the model. These are the local coordinates of a 3D component. Through this rendering pipeline, the system attaches abstract low-level data to specific spatial geometric anchor points.

[0040] In this invention, the system converts the acquired high-frequency real-time electrical parameter data, such as current and voltage, into three-dimensional floating label attributes. By reading the mapping relationship between the device's unique identifier and its three-dimensional spatial position, a text rendering component is instantiated on top of the corresponding three-dimensional component model's bounding box. The three-dimensional spatial coordinates of the floating label maintain the dynamic bulletin board orientation as the virtual camera's viewpoint moves, ensuring that maintenance personnel can intuitively read the real-time operating status data closely bound to the three-dimensional component under any zoom and rotation operations, achieving cross-dimensional digital overlay display.

[0041] Furthermore, based on the extracted temperature distribution heatmap, the system employs material texture mapping technology to seamlessly wrap the two-dimensional infrared temperature field features onto the mesh surface of the three-dimensional component. The texture parameterized coordinates of the model surface are set as follows: The system uses spatial interpolation and a color gradient mapping function to convert the temperature matrix into a diffuse color map that the rendering engine can recognize. This involves rendering the texture color values ​​of local surfaces. The mathematical expression is: in, To map to the current texture coordinates The actual inversion temperature value at that location, and These are the preset minimum and maximum temperature reference thresholds for global calibration, respectively. This is a nonlinear color transfer function that transforms a normalized temperature scalar into an RGB color vector. Through this mapping logic, the hidden thermodynamic distribution is directly visualized as a dynamic material texture on the device surface.

[0042] After receiving the fault classification and location results from the decision tree algorithm, the system initiates the highlight shader rendering program for the faulty node. For 3D components marked as abnormal, the rendering engine overwrites their original lighting model and material properties, injecting dynamically alternating self-illumination parameters over time to create a high-intensity visual warning effect. Dynamic self-luminescence intensity The following flicker control function must be satisfied: in, This is the fundamental luminescence constant of the material of this component. The luminescence amplitude coefficient is determined by the severity level of the fault. This is a preset warning flashing frequency. This mechanism makes the faulty part stand out in a complex 3D scene.

[0043] Finally, the 3D linkage model of the transformer platform, with its underlying attribute overlay and dynamic material replacement completed, is nested and integrated into the macroscopic panoramic map of the geographic information system. The system instantiates this linkage model in a 3D geographic coordinate system based on the GPS latitude and longitude coordinates of the transformer platform's field acquisition equipment. When a local 3D component triggers highlight rendering, the system synchronously drives the corresponding transformer platform primitive on the macroscopic map to generate a linked flashing indicator. Through the hierarchical nesting of microscopic device twins and macroscopic spatial geography, monitoring data silos are completely eliminated, significantly reducing the preparation cycle for maintenance personnel to locate and analyze complex power distribution networks.

[0044] This embodiment details the complete business logic of executing a multi-level early warning mechanism based on the preset three-dimensional spatial early warning trigger conditions in the three-dimensional linkage model of the test bench transformer, providing three-dimensional visualization, and outputting closed-loop linkage control signals according to fault classification and location results. To achieve precise hierarchical control of the test bench transformer's deterioration state, the system constructs a comprehensive risk assessment model based on multi-dimensional feature fusion. This model quantifies and weights the previously extracted equipment temperature deviation, electrical parameter anomaly probability, and appearance defect severity. A comprehensive risk index for specific physical components is also provided. The computational model satisfies: in, The current real-time temperature of the component is obtained from infrared thermal imaging. This is the reference normal temperature for this component under current load and environmental conditions. This is the preset limit safety temperature threshold; The probability of abnormal electrical parameters is output jointly by logistic regression and time series model. The severity score of physical defects output by the target detection algorithm; , , These represent the weighting coefficients for temperature characteristics, electrical characteristics, and appearance defect characteristics in the comprehensive risk assessment. This formula transforms multi-source heterogeneous degradation symptoms into comparable values ​​for a single dimension.

[0045] In this invention, the system initiates a multi-level early warning mechanism based on the calculated comprehensive risk index, and maps it in real time to a 3D visualization management platform for graphical presentation. The visual rendering driver function for the early warning status... Construct a piecewise discrete mapping relationship: in, , , The risk assessment thresholds are arranged in a tiered manner, from general warning to important warning to emergency warning. When under... When in this state, the rendering engine applies a yellow outer glow outline effect to the target 3D part and displays a pop-up notification in the platform sidebar indicating a localized temperature rise or slight deviation; when it rises to... When in a certain state, the system switches to an orange flashing effect and automatically pushes alarm information to the mobile terminals of maintenance personnel via the communication interface; once the limit is exceeded, it triggers... In this state, faulty nodes in the 3D digital twin will display a high-intensity red breathing flashing effect, visually forcibly capturing the attention of dispatchers.

[0046] Upon triggering an emergency warning, the system seamlessly connects the underlying hardware and software communication links, automatically issuing physical-level closed-loop linkage control signals. In response to thermodynamic emergencies such as persistently high transformer oil temperature or severe overload, the control center sends a high-level start command to the field auxiliary equipment controller, forcibly activating cooling fans or cooling devices for physical cooling. If the overall assessment indicates a highly dangerous fault such as a line short circuit, insulation breakdown, or ignition of an open flame, the system uses hardwired connections or a low-latency industrial bus to directly issue a tripping command to the relay protection device at the transformer site. This action bypasses the manual approval process, achieving millisecond-level physical isolation of the faulty line and effectively preventing further expansion of equipment damage.

[0047] As on-site control actions are executed, the system simultaneously initiates an intelligent scheduling program for emergency repair resources in the background. The information processing engine extracts high-precision fault coordinates, diagnosed fault types, and associated historical maintenance knowledge base data, automatically packaging and generating standardized electronic maintenance work orders. This work order data package not only includes a text description and a list of spare parts and tools to be replaced, but also carries a specific-view 3D screenshot of the fault node and a navigation path. Subsequently, relying on the wireless communication network, the system dispatches the work order to the smart terminal of the repair personnel closest to the faulty transformer. Through this series of sequential actions—from risk assessment, digital spatial alarms, physical equipment intervention to work order dispatch—a complete business loop of "monitoring and perception - algorithm decision-making - automatic isolation - precise scheduling" is established, transforming traditional passive and delayed emergency repairs into a highly automated proactive defense system.

[0048] In this embodiment, a corresponding virtual device logical architecture is constructed based on the aforementioned intelligent identification and 3D visualization linkage method for changing test bench states. The device specifically includes a 3D base construction module, a multi-source data acquisition module, a multi-dimensional collaborative analysis module, a rendering attribute-driven module, and an early warning and closed-loop linkage module. The 3D base construction module is responsible for establishing the initial mapping relationship between the physical device spatial topology and the virtual geometric mesh; its output is passed to subsequent functional levels as a spatial rendering benchmark. The multi-source data acquisition module uses the underlying driver interface to poll the field sensor array, continuously injecting cross-dimensional time-series monitoring sequences into the system. The multi-dimensional collaborative analysis module encapsulates a composite algorithm cluster including a target detection network, a time-series prediction model, and a decision tree, performing dimensionality reduction and high-order feature extraction on the injected data stream, outputting structured degradation judgment conclusions. Based on the above diagnostic conclusions, the rendering attribute-driven module and the early warning and closed-loop linkage module respectively perform dynamic material replacement in the digital twin space and trigger relay protection hard isolation actions in the physical space, thereby forming a complete software functional chain from underlying data perception and central logic operation to physical entity control.

[0049] This invention provides a low-level hardware architecture for an electronic device to support the highly coupled algorithm matrix and 3D graphics rendering pipeline. The electronic device includes a central processing unit (CPU), non-volatile memory (NDRAM), random access memory (RAM), and a multi-mode network communication interface, all interconnected via a high-speed system bus. The NDRAM stores the basic instruction set, including the operating system kernel, deep learning inference framework, and 3D graphics library, as well as the pre-trained weight networks for each algorithm. Upon power-on reset, the CPU loads the core computer program from the NDRAM into the RAM for high-speed execution, sequentially scheduling the execution logic for multi-modal data alignment, feature vector cross-analysis, and 3D viewport refresh.

[0050] When the central processing unit (CPU) of an electronic device performs high-concurrency image detection, temporal prediction, and 3D rendering tasks, a hardware-level task execution latency constraint model is established at the underlying level to ensure the real-time response characteristics of 3D visualization linkage and closed-loop control commands. This model defines the total computational latency required for the CPU to process a complete multidimensional recognition and rendering cycle. The following evaluation formula must be satisfied: in, This represents the total number of computational subtasks that the system schedules in parallel within a single processing cycle. For the first The total number of machine instructions generated after compiling each computational subtask This refers to the pipelined instruction throughput of the central processing unit. For single-byte addressing and physical read / write latency of random access memory, This represents the total number of bytes of memory used during the execution of this subtask. This refers to the inherent electrical transmission delay of the system bus during high-speed heterogeneous data transfer. Computational latency is strictly constrained through hardware-level system clock allocation and direct memory access scheduling. It is within the preset system safety response time window, thus ensuring millisecond-level precise synchronization between fault linkage signals and 3D alarm screens.

[0051] Furthermore, a computer-readable storage medium is provided, which contains computer program instructions burned into it, which can be read and executed by the central processing unit of the aforementioned electronic device. When the computer program is interpreted and compiled, it enables the acquisition of multi-source heterogeneous data from the test bench, the calculation of the three-dimensional spatial mapping relationship of the device, the forward propagation of parameters for the multimodal intelligent recognition algorithm, the invocation of the drive engine for the three-dimensional visualization dynamic rendering pipeline, and the generation of multi-level early warning and hard isolation linkage control signals. This computer-readable storage medium encompasses magnetic storage devices, optical storage disks, and various semiconductor solid-state storage chips. As a non-volatile physical entity carrying the instruction sequence of the core technical solution of this invention, it ensures the reliability and persistent operation of the test bench proactive defense and holographic visualization operation and maintenance logic across different edge computing nodes and cloud servers.

[0052] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent recognition and three-dimensional visualization linkage of test bench changing states, characterized in that, include: Computer-aided design technology is used to construct an initial three-dimensional model of the bench transformer, and the three-dimensional spatial position mapping relationship of each three-dimensional component in the initial three-dimensional model is established. The multi-source monitoring data of the transformer platform is used as the data source for updating the three-dimensional scene. The multi-source monitoring data includes image data containing visible light images and infrared images, electrical parameter data, and environmental data acquired by a dual-spectrum camera. Intelligent identification and correlation analysis are performed on the multi-source monitoring data to extract three-dimensional rendering features and generate fault classification and location results of the transformer and temperature distribution heat map of the physical components of the transformer. The multi-source monitoring data, the temperature distribution heat map of the physical components of the transformer, and the fault classification and location results are transformed into three-dimensional dynamic rendering attributes. Based on the three-dimensional spatial position mapping relationship, they are fused and mapped to the corresponding three-dimensional spatial position of the initial three-dimensional model, driving the initial three-dimensional model to perform visual dynamic overlay rendering to generate a three-dimensional linkage model of the transformer. Based on the preset three-dimensional spatial early warning triggering conditions in the aforementioned three-dimensional linkage model of the test bench, a three-dimensional visualization of a multi-level early warning mechanism is executed, and the corresponding closed-loop linkage control signal is output according to the fault classification and location results.

2. The method for intelligent identification and three-dimensional visualization linkage of test bench changing states according to claim 1, characterized in that, The process of converting the multi-source monitoring data, the temperature distribution heatmap of the physical components of the transformer, and the fault classification and location results into three-dimensional dynamic rendering attributes, and fusing them onto the corresponding three-dimensional spatial positions of the initial three-dimensional model based on the three-dimensional spatial position mapping relationship, drives the initial three-dimensional model to perform visual dynamic overlay rendering, generating a three-dimensional linkage model of the transformer, including: The initial 3D model is rendered using a 3D rendering engine at a preset ratio, and the 3D viewpoint of the initial 3D model is adjusted to the target angle in response to interactive commands. At the three-dimensional spatial location of the initial three-dimensional model, the corresponding electrical parameter data is superimposed in real time in the form of three-dimensional floating labels, and the temperature distribution heat map of the physical components of the transformer is mapped onto the surface of the corresponding three-dimensional component as a material texture. Based on the fault classification and location results, the location of the faulty three-dimensional component is highlighted in the initial three-dimensional model, and the faulty test bench three-dimensional linkage model is displayed with alarm flashing on the panoramic three-dimensional map in conjunction with the geographic information system.

3. The method for intelligent identification and three-dimensional visualization linkage of test bench changing states according to claim 1, characterized in that, The process of intelligently identifying and correlating the multi-source monitoring data, extracting 3D rendering features, and generating fault classification and location results for the transformer and a temperature distribution heatmap of the transformer's physical components includes: The visible light image in the image data is analyzed based on a deep learning object detection algorithm to identify abnormal appearances of the physical components of the test bench. Temperature field is extracted from the infrared image in the image data, the temperature values ​​of the physical components of the transformer are calculated, and a temperature distribution heat map of the physical components of the transformer is generated. The electrical parameter data is analyzed for trends based on a long short-term memory network time-series prediction model to identify abnormal electrical parameter states. A multi-factor logistic regression model is used to correlate the environmental data with the electrical parameter data to determine the impact of environmental factors on the operation of the transformer. Based on the decision tree algorithm, the abnormal appearance of the physical components of the transformer, the temperature distribution heat map of the physical components of the transformer, the abnormal electrical parameters, and the impact of environmental factors on the operation of the transformer are integrated to automatically classify the fault types. The fault classification and location results are calculated by combining the location coordinates of the acquisition equipment that collects the multi-source monitoring data.

4. The method for intelligent identification and three-dimensional visualization linkage of test bench changing states according to claim 1, characterized in that, The three-dimensional visualization presentation of the multi-level early warning mechanism based on the preset three-dimensional spatial early warning triggering conditions in the three-dimensional linkage model of the test bench, and the output of corresponding closed-loop linkage control signals according to the fault classification and location results, includes: Based on the preset risk level corresponding to the fault classification and location results, the warning level is divided into general warning, important warning and emergency warning, and each warning level is presented in different colors or animation effects in the three-dimensional linkage model of the test bench. When the emergency warning is triggered, a closed-loop linkage control signal is generated and output to automatically link the preset protection device in the transformer to cut off the faulty line. The operation and maintenance work order containing the three-dimensional spatial coordinates of the fault location, the fault type, and the handling suggestions is automatically generated through the mobile communication network and dispatched to the mobile terminal of the operation and maintenance personnel.

5. The method for intelligent identification and three-dimensional visualization linkage of test bench changing states according to claim 1, characterized in that, The acquisition of multi-source monitoring data from the transformer platform is used as the data source for updating the 3D scene. This multi-source monitoring data includes image data (containing visible light and infrared images), electrical parameter data, and environmental data acquired via a dual-spectrum camera. The specific acquisition process includes: The image data is continuously acquired at a preset cycle by the dual-spectrum camera with gimbal control function. The current, voltage, and power parameters of the high-voltage side and low-voltage side of the transformer are obtained as the electrical parameter data; The temperature, humidity, wind speed, rainfall, and smoke concentration of the environment in which the transformer is located are obtained as environmental data.

6. The method for intelligent identification and three-dimensional visualization linkage of test bench changing states according to claim 1, characterized in that, During the acquisition process, the multi-source monitoring data is transmitted in two modes: wireless network transmission and fiber optic wired transmission. When the network is interrupted, the multi-source monitoring data is temporarily stored locally using edge computing nodes. Once the network is restored, the interrupted data is retransmitted to ensure the data integrity of the 3D visualization rendering.

7. A linkage device for intelligent recognition and three-dimensional visualization of changing states of a test bench, characterized in that, include: The three-dimensional construction module is used to construct the initial three-dimensional model of the bench transformer using computer-aided design technology, and to establish the three-dimensional spatial position mapping relationship of each three-dimensional component in the initial three-dimensional model. The data acquisition module is used to acquire multi-source monitoring data of the transformer as a data source for updating the three-dimensional scene. The multi-source monitoring data includes image data containing visible light and infrared images, electrical parameter data, and environmental data acquired by a dual-spectrum camera. The collaborative analysis module is used to intelligently identify and correlate the multi-source monitoring data, extract three-dimensional rendering features, and generate the fault classification and location results of the transformer and the temperature distribution heat map of the physical components of the transformer. The 3D rendering manipulation module is used to convert the multi-source monitoring data, the temperature distribution heat map of the physical components of the transformer, and the fault classification and location results into 3D dynamic rendering attributes, and to fuse and map them to the corresponding 3D spatial position of the initial 3D model based on the 3D spatial position mapping relationship, thereby driving the initial 3D model to perform visual dynamic overlay rendering and generating a 3D linkage model of the transformer. The early warning linkage module is used to perform a three-dimensional visualization of a multi-level early warning mechanism based on the preset three-dimensional spatial early warning trigger conditions in the three-dimensional linkage model of the test bench, and output the corresponding closed-loop linkage control signal according to the fault classification and location results.

8. The intelligent identification and three-dimensional visualization linkage device for changing states of a test bench as described in claim 7, characterized in that, The 3D rendering manipulation module is specifically used for: The initial 3D model is rendered using a 3D rendering engine at a preset ratio, and the 3D viewpoint of the initial 3D model is adjusted to the target angle in response to interactive commands. At the three-dimensional spatial location of the initial three-dimensional model, the corresponding electrical parameter data is superimposed in real time in the form of three-dimensional floating labels, and the temperature distribution heat map of the physical components of the transformer is mapped onto the surface of the corresponding three-dimensional component as a material texture. Based on the fault classification and location results, the location of the faulty three-dimensional component is highlighted in the initial three-dimensional model, and the faulty test bench three-dimensional linkage model is displayed with alarm flashing on the panoramic three-dimensional map in conjunction with the geographic information system.

9. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a method for intelligent identification and three-dimensional visualization linkage of bench changing state as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements a method for intelligent identification and three-dimensional visualization linkage of bench state changes as described in any one of claims 1 to 6.

Images