Power equipment state three-dimensional visual diagnosis method, system, equipment and medium
By generating three-dimensional thermal voxel clouds and defect evidence volumes, the problems of infrared thermal imaging data mapping distortion and the inability to upgrade defect information to higher dimensions are solved, realizing high-precision three-dimensional visualization diagnosis of power equipment status, and improving operation and maintenance response efficiency and equipment safety.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-31
- Publication Date
- 2026-04-10
AI Technical Summary
In existing power equipment condition diagnosis technologies, infrared thermal imaging data is not corrected for physical factors when mapped to three-dimensional space, resulting in distorted temperature field distribution and affecting the accuracy of thermal anomaly identification. Alarm information from intelligent defect inspection systems cannot be effectively upgraded to three-dimensional space, making it difficult to coordinate analysis with temperature field data. Furthermore, the diagnostic results cannot be associated with digital twin models, making it difficult for maintenance personnel to accurately locate potential equipment hazards.
By simultaneously acquiring 3D point cloud data, visible light image data, and defect alarm information streams, a 3D thermal voxel cloud and defect evidence body are generated. Physical corrections and data registration are performed, an evidence theory weighted model is constructed, a confidence voxel distribution is generated, and it is associated with the digital twin model of the power equipment to achieve 3D visualization.
It effectively eliminates physical interference in infrared thermal imaging data, realizes the dimensionality and quantification of defect information, improves the accuracy of collaborative diagnosis of thermal anomalies and structural defects in power equipment, enhances the interpretability of diagnostic results, and improves operation and maintenance response efficiency and equipment operation safety.
Smart Images
Figure CN121837859A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and in particular to a method, system, device and medium for three-dimensional visualization diagnosis of the status of power equipment. Background Technology
[0002] With the deep integration of smart grids and digital twin technologies, intelligent operation and maintenance of key equipment such as transformers and circuit breakers in power stations has become a core requirement for ensuring the safe and stable operation of the power grid. To comprehensively perceive equipment status, modern substations have widely deployed visible light cameras, infrared thermal imagers, millimeter-wave radar, and image recognition-based intelligent defect inspection systems, forming a multimodal perception system. These sensors can acquire equipment information from multiple dimensions, including geometry, vision, temperature, and defects, forming a multi-dimensional, multimodal perception data system aimed at achieving comprehensive monitoring of the status of power equipment through diverse sensing methods.
[0003] However, current technologies for condition diagnosis of power equipment suffer from several drawbacks. Firstly, the mapping of infrared thermal imaging data to three-dimensional space fails to correct for interfering physical factors, easily leading to distorted temperature field distribution. The resulting thermal field data cannot accurately reflect the actual thermal state of the power equipment, severely impacting the accuracy of thermal anomaly detection. Secondly, alarm information generated by intelligent defect inspection systems is largely limited to two-dimensional image coordinates and text descriptions, lacking effective dimensionality enhancement and quantification mechanisms. This makes it difficult to accurately map confidence levels and spatial locations to three-dimensional geometric space, hindering collaborative analysis with temperature field and other data. Furthermore, the outputs of existing power equipment diagnostic technologies cannot be effectively correlated with the component-level semantic information of the power equipment's digital twin model. This prevents the accurate presentation of thermal anomalies and structural defects in the digital twin model in a three-dimensional visual format, making it difficult for maintenance personnel to intuitively and efficiently pinpoint the specific location and severity of potential equipment hazards.
[0004] The aforementioned technical problems are particularly pronounced in substation scenarios with complex electromagnetic environments and dense equipment layouts, severely restricting the accuracy of collaborative diagnosis of thermal anomalies and structural defects in power equipment, the interpretability of diagnostic results, and the efficiency of operation and maintenance response, thus failing to ensure the timely operation safety of power equipment. Summary of the Invention
[0005] The purpose of this invention is to provide a three-dimensional visualization diagnostic method, system, device, and medium for power equipment status, thereby improving the diagnostic accuracy of power equipment status, the operation and maintenance response efficiency of power equipment, and operational safety.
[0006] To achieve the above objectives, the present invention provides a three-dimensional visualization diagnostic method for the status of power equipment, comprising: Simultaneously collect multimodal sensing data of the target power equipment area; wherein, the data includes three-dimensional point cloud data, visible light image data, infrared thermal imaging data, and defect alarm information stream; A three-dimensional thermal voxel cloud is generated based on the three-dimensional point cloud data and the infrared thermal imaging data; A three-dimensional defect evidence body is generated based on the three-dimensional point cloud data, the visible light image data, and the defect alarm information stream. Based on the three-dimensional thermal voxel cloud and the three-dimensional defect evidence body, an evidence theory weighted model is constructed, and reasoning is performed on the evidence theory weighted model to generate a confidence voxel distribution; The confidence voxel distribution is associated with the component-level semantics of the digital twin model of the power equipment to generate a three-dimensional visualization heat map and defect distribution map of the power equipment components.
[0007] Optionally, before generating the three-dimensional thermal voxel cloud based on the three-dimensional point cloud data and the infrared thermal imaging data, the method further includes: Using the spatial coordinate system constructed from the three-dimensional point cloud data as a reference, the external parameter matrices of the visible light image acquisition device and the infrared thermal imaging acquisition device relative to the reference are calculated. Based on the extrinsic parameter matrix, the visible light image data and infrared thermal imaging data corresponding to the synchronization timestamp are mapped onto the three-dimensional point cloud surface to complete the data registration.
[0008] Optionally, generating a three-dimensional thermal voxel cloud based on the three-dimensional point cloud data and the infrared thermal imaging data includes: Physical corrections are performed on the registered infrared thermal imaging data to generate a three-dimensional thermal voxel cloud. The physical corrections include: The infrared thermometric values are corrected based on the emissivity of the materials of the power equipment components and the law of radiation. Invalid temperature data caused by obstruction between components of electrical equipment is detected and compensated using the ray projection method. High reflectivity surface regions in a 3D point cloud are identified, and the amount of environmental heat source reflection interference is estimated based on the high reflectivity surface regions to correct the infrared temperature measurement values.
[0009] Optionally, generating a three-dimensional defect evidence body based on the three-dimensional point cloud data, the visible light image data, and the defect alarm information stream includes: Based on the registration relationship between the visible light image data and the three-dimensional point cloud data, the two-dimensional image coordinates in the defect alarm information stream are mapped to the spatial coordinate system constructed by the three-dimensional point cloud data to determine the three-dimensional spatial location of the defect. Based on the three-dimensional spatial location of the defect, a three-dimensional spatial evidence unit associated with the defect is constructed; The defect category and confidence information in the defect alarm information stream are assigned to the three-dimensional spatial evidence unit to generate a three-dimensional defect evidence body.
[0010] Optionally, the weighted model of evidence theory is defined as follows: ; in, The credibility weight of the i-th information source is obtained based on the historical accuracy statistics of the evidence from the i-th information source. The spatial consistency weight is calculated based on the dispersion of evidence from spatial neighboring voxels of each voxel, which is a preset number of such voxels. The temporal consistency weight is calculated based on the smoothness of evidence changes for each voxel over multiple consecutive time frames. The identification framework is defined as the set of candidate propositions for the state of electrical equipment; A and B are... A subset of propositions; Indicates the i-th information source for the recognition framework The basic confidence assignment function for the subset A of propositions.
[0011] Optionally, the step of reasoning on the weighted model of evidence theory to generate a confidence voxel distribution includes: The evidence corresponding to the infrared heat source and the evidence corresponding to the defect source, which have been processed by the evidence theory weighted model, are fused and calculated according to the preset evidence fusion rules to obtain the comprehensive basic reliability allocation. During the fusion calculation process, the conflict coefficient between evidence from different sources is evaluated. When the conflict coefficient reaches a preset conflict threshold, the overall basic reliability allocation is adjusted using conflict resolution rules. Based on the adjusted comprehensive basic confidence distribution, the degree of support and non-opposition of each voxel for each candidate power equipment state proposition is calculated. Combined with the three-dimensional spatial location of each voxel, a confidence voxel distribution containing the distribution of the degree of support for each voxel's equipment state is generated.
[0012] Optionally, after associating the confidence voxel distribution with the component-level semantics of the digital twin model of the power equipment to generate a three-dimensional visualized heat map and defect distribution map of the power equipment components, the method further includes: Based on the confidence voxel distribution, the operation and maintenance priority index of power equipment components is evaluated, and diagnostic decision information is generated.
[0013] To achieve the above objectives, the present invention also provides a three-dimensional visualization diagnostic system for the status of power equipment, comprising: The multi-source data acquisition and synchronization module is used to synchronously acquire multimodal sensing data of the target power equipment area; wherein, the data includes three-dimensional point cloud data, visible light image data, infrared thermal imaging data, and defect alarm information stream; A three-dimensional thermal field voxelization module is used to generate a three-dimensional thermal voxel cloud based on the three-dimensional point cloud data and the infrared thermal imaging data. The defect evidence 3D mapping module is used to generate a 3D defect evidence body based on the 3D point cloud data, the visible light image data, and the defect alarm information stream; The multi-source evidence fusion module is used to construct an evidence theory weighted model based on the three-dimensional thermal voxel cloud and the three-dimensional defect evidence body, and to infer the evidence theory weighted model to generate a confidence voxel distribution. The 3D visualization module is used to associate the confidence voxel distribution with the component-level semantics of the digital twin model of the power equipment, and generate a 3D visualized heat distribution map and defect distribution map of the power equipment components.
[0014] To achieve the above objectives, the present invention also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the three-dimensional visualization diagnostic method for power equipment status as described above.
[0015] To achieve the above objectives, the present invention also provides a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the three-dimensional visualization diagnostic method for the status of power equipment as described above.
[0016] Compared with existing technologies, this invention provides a three-dimensional visualization diagnostic method, system, equipment, and medium for power equipment status. Firstly, it generates a three-dimensional thermal voxel cloud based on three-dimensional point cloud data and infrared thermal imaging data, physically correcting the infrared thermal imaging data to effectively eliminate physical interference factors, avoid temperature field distribution distortion, and ensure that the thermal field data accurately reflects the actual thermal state of the equipment. Simultaneously, it maps two-dimensional defect alarm information to three-dimensional space and generates a three-dimensional defect evidence body containing confidence quantification information, achieving dimensionality enhancement and standardized representation of defect information, laying the foundation for co-dimensional collaborative analysis with thermal field data. Finally, by associating the confidence voxel distribution with the digital twin model of the power equipment, it generates a three-dimensional visualized thermal distribution map and defect distribution map, allowing thermal anomalies and structural defect states to be intuitively presented in the digital twin model, facilitating maintenance personnel to accurately locate the location and extent of potential hazards. This invention can effectively improve the accuracy of collaborative diagnosis of thermal anomalies and structural defects in power equipment, enhance the interpretability of diagnostic results, thereby improving maintenance response efficiency and providing reliable protection for the safe operation of power equipment in complex substation scenarios. Attached Figure Description
[0017] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a flowchart of a three-dimensional visualization diagnostic method for the status of power equipment provided in an embodiment of the present invention; Figure 2 This is a structural block diagram of a three-dimensional visualization diagnostic system for the status of power equipment provided in an embodiment of the present invention; Figure 3 This is a structural block diagram of a terminal device provided in an embodiment of the present invention. Detailed Implementation
[0019] The technical solutions of 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.
[0020] See Figure 1 , Figure 1 This is a flowchart of a three-dimensional visualization diagnostic method for the status of power equipment provided in an embodiment of the present invention. The three-dimensional visualization diagnostic method for the status of power equipment includes steps S1 to S5: Step S1: Synchronously collect multimodal sensing data of the target power equipment area; wherein, the data includes three-dimensional point cloud data, visible light image data, infrared thermal imaging data, and defect alarm information stream; Specifically, the system simultaneously acquires 3D point cloud data from millimeter-wave radar, 2D image sequences from visible light cameras, temperature field image sequences from infrared thermal imagers, and alarm information streams from the intelligent defect inspection system. It should be noted that the intelligent defect inspection system is an intelligent monitoring system deployed within power stations that can automatically identify structural defects in power equipment (such as rust, loosening, and dirt) and output structured alarm information. The output alarm information stream includes the defect category, 2D image coordinates, confidence score, and timestamp.
[0021] For example, in a substation inspection scenario, a millimeter-wave radar sensor at a fixed monitoring point can be used to collect three-dimensional point cloud data of the target power equipment area at a rate of 10 frames per second. The point cloud density is no less than 1000 points per cubic meter, and each point contains three-dimensional coordinates and reflection intensity information. Simultaneously, a visible light camera rigidly connected to the radar sensor or fixed through a known pose relationship acquires a sequence of two-dimensional color images of the same area at a resolution of 1920×1080 and a rate of 30 frames per second as visible light image data. Similarly, an infrared thermal imager, maintaining a fixed relative pose to the radar sensor, acquires a sequence of temperature field images at a resolution of 640×512 and a rate of 25 frames per second as infrared thermal imaging data, with each pixel containing a 14-bit integer value representing the apparent temperature. In addition, alarm information streams from existing intelligent defect inspection systems within the substation can be received in real time via a network interface. The information stream consists of JSON format data packets, with each data packet corresponding to a defect identification event. The data packets include fields such as: defect category, such as "insulator contamination", "fitting corrosion", and "loose connection point"; the alarm information stream also includes the two-dimensional pixel coordinates of the defect in the visible light image that triggered the alarm and the confidence score (ranging from 0 to 1) given by the identification algorithm.
[0022] It is important to note that these four types of data must be strictly aligned in time. A hardware-triggered synchronization mechanism can be used, where a unified clock source sends synchronization trigger pulses to the radar, visible light camera, and infrared thermal imager, ensuring that the start time deviation of each set of point cloud frames, visible light image frames, and infrared image frames is less than 1 millisecond. For the defect alarm information stream, software timestamp alignment can be used, matching the alarm timestamp with the image frame timestamp to associate the alarm information with the specific image frame. All synchronized data is tagged with a unified timestamp and cached in a circular data buffer, awaiting extraction and processing by subsequent modules according to time slices.
[0023] It is worth noting that the purpose of simultaneously collecting multimodal sensing data of the target power equipment area is to break the isolation barrier of multi-source sensing data in traditional inspections. By ensuring strict synchronization of various types of data in the time dimension and correlation in the spatial dimension, it provides a consistent basic data input for subsequent operations.
[0024] Step S2: Generate a three-dimensional thermal voxel cloud based on the three-dimensional point cloud data and the infrared thermal imaging data; In one alternative embodiment, prior to step S2, the method further includes: Using the spatial coordinate system constructed from the three-dimensional point cloud data as a reference, the external parameter matrices of the visible light image acquisition device and the infrared thermal imaging acquisition device relative to the reference are calculated. Based on the extrinsic parameter matrix, the visible light image data and infrared thermal imaging data corresponding to the synchronization timestamp are mapped onto the three-dimensional point cloud surface to complete the data registration.
[0025] Specifically, the collected 3D point cloud data is first preprocessed. The preprocessing process includes: applying a plane fitting algorithm based on random sampling consistency to segment and remove ground point clouds from the original point cloud; using a statistical outlier removal algorithm to remove discrete noise points whose average distance to their K nearest neighbors exceeds three times the standard deviation; and finally, using a clustering algorithm based on Euclidean distance to segment the remaining point cloud into several point cloud clusters, each cluster roughly corresponding to a physical equipment component, such as a transformer tank, high-voltage bushing, or heat sink, to obtain equipment structure point cloud clusters.
[0026] After point cloud preprocessing is completed, the extrinsic parameter matrix of the visible light acquisition device camera is calculated. Scale-invariant image feature points are extracted from the synchronized visible light image. Simultaneously, in the preprocessed device structure point cloud cluster, a set of three-dimensional feature points corresponding to the image feature points are selected manually or through automated algorithms. Using this set of three-dimensional to two-dimensional point correspondences, combined with the known camera intrinsic parameter matrix, the perspective N-point algorithm is used for iterative optimization to solve for the camera rotation matrix and translation vector that minimizes the reprojection error, which is the extrinsic parameter matrix of the visible light acquisition device camera.
[0027] It should be noted that since infrared thermal imagers and visible light cameras are usually installed with a rigid connection, their relative poses are obtained through calibration during installation and stored as known parameters.
[0028] Therefore, after obtaining the extrinsic parameter matrix of the visible light camera, the extrinsic parameter matrix of the infrared thermal imager relative to the point cloud world coordinate system, i.e., the radar world coordinate system, can be directly calculated and derived based on the known rigid transformation relationship.
[0029] Finally, based on the two extrinsic parameter matrices mentioned above, the visible light image data and infrared thermal imaging data corresponding to the synchronization timestamps are mapped onto the 3D point cloud surface, completing the spatiotemporal registration of the data, that is, unifying the spatial reference of all sensor inputs. For example, point cloud frames, visible light image frames, and infrared image frames under the same time tag are extracted from the buffer to form a multimodal data packet that is aligned in both time and space for output.
[0030] Further, step S2 includes: Physical corrections are performed on the registered infrared thermal imaging data to generate a three-dimensional thermal voxel cloud. The physical corrections include: The infrared thermometric values are corrected based on the emissivity of the materials of the power equipment components and the law of radiation. Invalid temperature data caused by obstruction between components of electrical equipment is detected and compensated using the ray projection method. High reflectivity surface regions in a 3D point cloud are identified, and the amount of environmental heat source reflection interference is estimated based on the high reflectivity surface regions to correct the infrared temperature measurement values.
[0031] It should be noted that the core function of step S2 is to accurately map and correct the two-dimensional infrared temperature field into thermal property information attached to the surface of the three-dimensional device, and further voxelize it into regular three-dimensional mesh data.
[0032] For example, temperature projection is first performed. Using the obtained extrinsic and intrinsic parameter matrices of the infrared thermal imager, the pixel coordinates corresponding to each surface point in the point cloud on the infrared image plane are calculated. The apparent temperature value of the pixel is obtained from the infrared image through bilinear interpolation, and this apparent temperature value is used as the initial temperature attribute of the three-dimensional point.
[0033] However, the temperature values obtained by direct projection suffer from various physical distortions. Therefore, this embodiment of the invention constructs a thermal field correction model comprising three sub-models for physical correction.
[0034] One is the material emissivity correction sub-model, which mainly relies on the semantic information of equipment components provided by the predefined digital twin model of power equipment, such as "transformer tank steel plate," "porcelain insulator," and "copper conductive rod." It queries the pre-generated material emissivity library for the standard emissivity ε of the corresponding material and applies it to the measured apparent temperature. Based on the approximate linearization of the Stefan-Boltzmann law, its true surface temperature is calculated. The calculation formula is: For example, this correction will significantly improve the temperature reading for smooth metal surfaces with low emissivity.
[0035] Secondly, there is the occlusion detection sub-model. For a target point cloud surface point, the occlusion detection sub-model emits a ray from its 3D position towards the optical center of the infrared thermal imager. Using an accelerated bounding box hierarchy algorithm, it quickly detects whether this ray intersects with other point cloud clusters in the scene before reaching the thermal imager. If an intersection exists, the target point is determined to be occluded by other components, and its projected temperature data is deemed invalid; the target point is then considered an invalid point. For invalid points, an unoccluded visible point is searched within its spatial neighborhood, and its temperature value is estimated using inverse distance weighted interpolation for compensation.
[0036] Thirdly, there is the environmental reflection interference suppression sub-model. This sub-model identifies regions in the point cloud that belong to high-reflectivity metal surfaces, and estimates the impact of reflected radiation on the temperature measurement of this region based on the known locations of strong heat sources in the digital twin model of power equipment, such as the solar azimuth angle and nearby heat-generating equipment. And deduct it from the infrared measurement value.
[0037] After completing the three physical correction steps described above, a 3D point cloud voxelization with accurate temperature attributes is obtained. This divides the 3D space containing the power equipment into a cubic grid with sides of 0.01 meters. For each voxel, the average value of all corrected temperature points falling within it is calculated as the temperature value of that voxel. If a voxel contains no points, it is marked as empty. Finally, a 3D thermal voxel cloud containing temperature attributes is generated, where each valid 3D thermal voxel stores a floating-point temperature value.
[0038] It is worth noting that the embodiments of the present invention are based on three-dimensional point cloud data and infrared thermal imaging data. After registration, the infrared temperature measurement values are subjected to triple physical correction to finally generate a three-dimensional thermal pixel cloud. This can effectively solve the problems of temperature measurement distortion and incomplete data caused by differences in material properties, obstruction between components and environmental reflection interference in traditional infrared temperature measurement. The three-dimensional thermal pixel cloud can accurately carry the true temperature attributes of the equipment, and rely on the unified spatial reference of the three-dimensional point cloud to realize the binding of temperature information with the three-dimensional spatial position of the equipment. This provides real, complete and highly consistent thermal anomaly data support for three-dimensional visualization, which can improve the accuracy and reliability of thermal condition diagnosis of power equipment.
[0039] Step S3: Generate a three-dimensional defect evidence body based on the three-dimensional point cloud data, the visible light image data, and the defect alarm information stream; In an optional embodiment, step S3 includes steps S301 to S303: Step S301: Based on the registration relationship between the visible light image data and the three-dimensional point cloud data, map the two-dimensional image coordinates in the defect alarm information stream to the spatial coordinate system constructed by the three-dimensional point cloud data to determine the three-dimensional spatial location of the defect. Step S302: Based on the three-dimensional spatial location of the defect, construct a three-dimensional spatial evidence unit associated with the defect; Step S303: Assign the defect category and confidence information in the defect alarm information stream to the three-dimensional spatial evidence unit to generate a three-dimensional defect evidence body.
[0040] For example, firstly, using the intrinsic and extrinsic parameter matrices of the visible light camera, the two-dimensional pixel coordinates provided in the defect alarm information stream are back-projected into three-dimensional space, forming a ray originating from the camera's optical center and passing through that pixel. This ray represents an infinite number of possible three-dimensional locations where the defect may exist.
[0041] Next, the intersection point between this ray and the implicit device surface model constructed from the registered point cloud is calculated. Specifically, sampling is performed along the ray direction in steps of 0.001 meters. The point cloud density within a certain radius around the sampling point is checked. If the density exceeds a threshold, the location is determined to be part of the device surface, and this sampling point is used as the precise 3D location of the defect. If a precise intersection point cannot be found due to sparse point cloud or noise, a search is performed within the neighborhood of the point cloud near the ray. A spherical search domain with a radius of 0.05 meters is defined, and all point clouds within the domain are collected. The weighted centroid of these points is calculated as the approximate 3D location of the defect, with the weights determined by the point cloud reflection intensity or the distance from the ray. Thus, the location of the defect in 3D space is determined.
[0042] Then, a three-dimensional evidence body is constructed for the defect. This evidence body is a small cubic region with a side length of 0.02 meters, centered on the three-dimensional location of the defect. It includes spatial location coordinates, defect category label, and initial evidence quality. The initial evidence quality is the confidence score in the alarm information stream, denoted as m_i(Defect), representing the confidence level supporting the proposition "this type of defect exists." Simultaneously, the identification framework Θ (the set of propositions for all possible states defined by power equipment condition diagnosis) is defined to include two basic propositions: defect exists and normal state. Therefore, the initial confidence level of the evidence body for the "normal state" proposition is m_i(Normal) = 0, and the total uncertainty is m_i(Θ) = 1 - m_i(Defect).
[0043] Through the above steps, this invention can instantiate each two-dimensional alarm into a three-dimensional defect evidence body with clear spatial attributes and uncertainty measurement. This enables the structured transformation of defect information from the image domain to the three-dimensional geometric domain, allowing the defect to not only have a clear spatial location but also form a complete evidence carrier containing location, category, and confidence level. This ensures the collaborative analysis conditions of defect information and other multi-source data such as three-dimensional thermal voxel clouds in a unified dimension, and provides standardized and quantifiable structural defect input for subsequent multi-source evidence fusion, thereby improving the spatial accuracy and interpretability of defect diagnosis.
[0044] Step S4: Based on the three-dimensional thermal voxel cloud and the three-dimensional defect evidence body, construct an evidence theory weighted model, perform reasoning on the evidence theory weighted model, and generate a confidence voxel distribution; It should be noted that the three-dimensional thermal voxel cloud itself is considered a source of evidence. The temperature value of each voxel is compared with the historical normal temperature baseline and transformed into evidence for the proposition of "thermal anomaly". For example, if the temperature of a voxel exceeds the baseline threshold, evidence supporting the proposition of "overheating" is generated. The defect evidence body is the source of evidence for the proposition of "structural defect". Therefore, the goal of step S4 is to fuse these two types of heterogeneous evidence, the three-dimensional thermal voxel cloud and the defect evidence body, within a unified three-dimensional voxel space, and generate a complete confidence distribution for each voxel regarding the device status (such as normal, overheating, structural defect, and combined anomaly). At the same time, the uncertainty of the fusion result is quantified. To achieve this goal, this embodiment of the invention constructs an evidence weighting model that integrates source confidence, spatial consistency, and temporal consistency.
[0045] In one alternative embodiment, the weighted model of evidence theory is defined as follows: ; in, The credibility weight of the i-th information source is obtained based on the historical accuracy statistics of the evidence from the i-th information source, such as the historical temperature measurement accuracy of an infrared thermal imager or the historical identification accuracy of a defect inspection system. In this embodiment of the invention, when i is a source related to an infrared thermal imager, the information source is a three-dimensional thermal voxel cloud, providing evidence related to "thermal anomalies"; when i is a source related to a defect inspection system, the information source is a three-dimensional defect evidence body, providing evidence related to "structural defects".
[0046] The spatial consistency weight is calculated based on the dispersion of evidence from a predetermined number of spatial neighboring voxels for each voxel. For example, it is calculated based on the dispersion of evidence from 50 spatial neighboring voxels of voxel v, calculating the variance of the confidence scores of these 50 neighboring voxels for the same proposition. A smaller variance indicates higher spatial consistency. The larger the value, the lower the value. The smaller the value.
[0047] The temporal consistency weight is calculated based on the smoothness of evidence change for each voxel over multiple consecutive time frames. For example, the temporal consistency weight is calculated based on the smoothness of evidence change for voxel v over the most recent 10 consecutive time frames, specifically by calculating the sum of the absolute values of the first-order differences of the confidence scores over these 10 time frames. A smaller sum indicates a smoother temporal evolution. The larger the value, the lower the value. The smaller the value.
[0048] The identification framework refers to the set of propositions defining all possible states in the condition diagnosis of power equipment; A and B are... A subset of propositions; Indicates the i-th information source for the recognition framework The basic probability assignment (BPA, also known as the mass function value) of the subset A of propositions is the mass function of the original evidence.
[0049] It is worth noting that evidence can be weighted using the evidence theory weighting model in the embodiments of the present invention, thereby adaptively enhancing the weight of evidence that is consistent in space and time and has a reliable source, and suppressing the influence of noise and isolated outliers.
[0050] Further, the step of reasoning on the weighted model of the evidence theory to generate a confidence voxel distribution includes steps a to c: Step a. The evidence corresponding to the infrared heat source and the evidence corresponding to the defect source, after being processed by the weighted model of the evidence theory, are fused and calculated according to the preset evidence fusion rule to obtain the comprehensive basic reliability allocation; wherein, the preset evidence fusion rule is the Dempster combination rule, which is an important component of Dempster-Shafer's theory (evidence theory) and is used to merge evidence from different sources to deal with uncertainty and ambiguity.
[0051] Step b. During the fusion calculation, the conflict coefficient between evidence from different sources is evaluated. When the conflict coefficient reaches a preset conflict threshold, the overall basic reliability allocation is adjusted using a conflict resolution rule. Preferably, the conflict resolution rule is either the Yager correction rule or the discount factor decay rule.
[0052] Step c. Based on the adjusted comprehensive basic confidence distribution, calculate the degree of support and non-opposition of each voxel for each candidate power equipment state proposition, and generate a confidence voxel distribution containing the distribution of the degree of support of each voxel for the equipment state proposition, in combination with the three-dimensional spatial location of each voxel.
[0053] It should be noted that the Dempster combination rule is used to fuse the evidence corresponding to the infrared heat source after processing by the aforementioned evidence theory weighted model with the evidence corresponding to the defect source. This rule can combine the information from two independent evidence sources to generate a new, comprehensive evidence mass function. During the fusion calculation, when the two evidence sources have high conflict, that is, when their state judgments on the same voxel are diametrically opposed and both have high confidence levels, the classic Dempster rule may produce counterintuitive results.
[0054] To address this, the module integrates a conflict resolution mechanism, employing the Yager correction rule. The Yager correction rule is a classic correction rule for handling highly conflicting evidence fusion under the Dempster rule. Its core principle is that when the judgments of different evidence sources contradict each other and both have high confidence levels, the confidence level of the conflict is no longer forcibly assigned to specific propositions. Instead, the confidence quality generated by all conflicts is uniformly assigned to propositions representing global uncertainty in the recognition framework. This avoids the counterintuitive fusion results under high conflict under the Dempster rule and ensures the rationality of evidence fusion.
[0055] For example, when the conflict coefficient during combination exceeds the threshold of 0.8, the system automatically switches to Yager's correction rule, assigning conflict quality to the Θ proposition representing global uncertainty, or introducing a discount factor to attenuate conflicting evidence before recombining. The fusion ultimately outputs two key metrics for each voxel v: the confidence function value Bel(A) for each proposition A, representing the sum of all evidence supporting proposition A; and the plausibility function value Pl(A), representing the sum of evidence not opposing proposition A. The difference between the two, Pl(A) - Bel(A), is the uncertainty interval of that proposition. Thus, each voxel not only has the most probable state judgment but also a clear measure of uncertainty.
[0056] It is worth noting that the embodiments of the present invention can effectively avoid the problem of distorted fusion results in the case of high conflict of multi-source evidence through the above operations, thereby ensuring the rationality and reliability of evidence fusion reasoning. At the same time, it can realize the precise binding of device status confidence information and three-dimensional spatial position, so that the generated confidence voxel distribution can completely and accurately represent the status judgment results of each spatial position of the device, and greatly improve the accuracy and comprehensiveness of device status diagnosis.
[0057] Step S5: Associate the confidence voxel distribution with the component-level semantics of the digital twin model of the power equipment to generate a three-dimensional visualization heat map and defect distribution map of the power equipment components.
[0058] It should be noted that the core input of step S5 is the confidence voxel distribution after fusion and uncertainty quantification. This requires pre-storing a refined digital twin 3D model of the target substation. This model not only has a geometric shape but also includes semantic information at the component level. For example, the entire transformer model can be decomposed into independently addressable component objects such as "oil tank," "high-voltage bushing A phase," "low-voltage bushing B phase," "radiator assembly," and "oil conservator."
[0059] For example, firstly, based on the component hierarchy of the digital twin model, thermal voxels and defect evidence in 3D space are associated with their corresponding components. Specifically, through spatial location queries, all voxels falling within the 3D bounding box of a component are marked as auxiliary data of that component. For instance, the temperature, defect confidence, and other information of all voxels located inside the geometric model of "high-voltage bushing A phase" are bound to that bushing component. After binding, 3D visualization rendering is performed, and the graphics engine is driven to overlay rendering on the digital twin 3D scene. For thermal distribution, a gradient color spectrum from blue to red is used to render the temperature field on the equipment surface in real time, with abnormally high temperature areas marked with a striking flashing red. For defect distribution, markers with specific icons and colors are rendered at the corresponding 3D spatial locations, such as using a yellow exclamation mark to indicate "fitting corrosion." Maintenance personnel can observe the overall status panorama of the equipment from any angle through interactive devices.
[0060] In one alternative embodiment, after step S5, the method further includes: Based on the confidence voxel distribution, the operation and maintenance priority index of power equipment components is evaluated, and diagnostic decision information is generated.
[0061] For example, during 3D visualization, for each component, the fused confidence and uncertainty data of all voxels within it are aggregated. Based on this data, the component's maintenance priority index is calculated. This index can be a score from 0 to 100, calculating the highest confidence level for various anomalies on the component, the average uncertainty of the anomalies, and the component's criticality level in the power grid. A higher index indicates a greater urgency for intervention. Finally, a structured diagnostic report containing diagnostic decision information is automatically generated based on the maintenance priority index. For example, the report content may include: a list of abnormal components and their 3D location screenshots, anomaly types and confidence levels, a quantified maintenance priority index ranking, analysis of major uncertainty sources, and preliminary handling recommendations. Furthermore, the report can be pushed to maintenance personnel via workstation interfaces, mobile terminals, or augmented reality devices, providing comprehensive support from global situational awareness to specific location-based decision-making.
[0062] In summary, the three-dimensional visualization diagnostic method for power equipment provided by this invention generates a three-dimensional thermal voxel cloud based on three-dimensional point cloud data and infrared thermal imaging data. This process physically corrects the infrared thermal imaging data, effectively eliminating physical interference factors such as material emissivity and shading effects, avoiding distortion of the temperature field distribution, and ensuring that the thermal field data accurately reflects the actual thermal state of the equipment. Relying on the registration relationship between the three-dimensional point cloud and visible light image data, two-dimensional defect alarm information is mapped to three-dimensional space, generating a three-dimensional defect evidence body containing confidence quantification information. This achieves dimensional upgrading and standardized representation of defect information, laying the foundation for co-dimensional collaborative analysis with thermal field data. By associating the confidence voxel distribution with the component-level semantics of the power equipment's digital twin model, a three-dimensional visualized thermal distribution map and defect distribution map are generated. This allows thermal anomalies and structural defect states to be intuitively presented in the digital twin model, facilitating accurate location and severity of potential hazards for maintenance personnel. This invention can effectively improve the accuracy of collaborative diagnosis of thermal anomalies and structural defects in power equipment, enhance the interpretability of diagnostic results, thereby improving maintenance response efficiency and providing reliable protection for the safe operation of power equipment in complex substation scenarios.
[0063] Based on the above method items, the present invention provides corresponding system items embodiments.
[0064] See Figure 2 , Figure 2 This is a structural block diagram of a three-dimensional visualization diagnostic system for power equipment status provided in an embodiment of the present invention. The three-dimensional visualization diagnostic system for power equipment status includes: The multi-source data acquisition and synchronization module 21 is used to synchronously acquire multimodal sensing data of the target power equipment area; wherein, the data includes three-dimensional point cloud data, visible light image data, infrared thermal imaging data, and defect alarm information stream; The three-dimensional thermal field voxelization module 22 is used to generate a three-dimensional thermal voxel cloud based on the three-dimensional point cloud data and the infrared thermal imaging data. The defect evidence 3D mapping module 23 is used to generate a 3D defect evidence body based on the 3D point cloud data, the visible light image data and the defect alarm information stream; The multi-source evidence fusion module 24 is used to construct an evidence theory weighted model based on the three-dimensional thermal voxel cloud and the three-dimensional defect evidence body, and to reason about the evidence theory weighted model to generate a confidence voxel distribution. The 3D visualization module 25 is used to associate the confidence voxel distribution with the component-level semantics of the digital twin model of the power equipment to generate a 3D visualization heat map and defect distribution map of the power equipment components.
[0065] In one optional embodiment, the three-dimensional visualization diagnostic system for power equipment status further includes a data registration module, used for: Using the spatial coordinate system constructed from the three-dimensional point cloud data as a reference, the external parameter matrices of the visible light image acquisition device and the infrared thermal imaging acquisition device relative to the reference are calculated. Based on the extrinsic parameter matrix, the visible light image data and infrared thermal imaging data corresponding to the synchronization timestamp are mapped onto the three-dimensional point cloud surface to complete the data registration.
[0066] In one alternative embodiment, the three-dimensional thermal field voxelization module 22 is used for: Physical corrections are performed on the registered infrared thermal imaging data to generate a three-dimensional thermal voxel cloud. The physical corrections include: The infrared thermometric values are corrected based on the emissivity of the materials of the power equipment components and the law of radiation. Invalid temperature data caused by obstruction between components of electrical equipment is detected and compensated using the ray projection method. High reflectivity surface regions in a 3D point cloud are identified, and the amount of environmental heat source reflection interference is estimated based on the high reflectivity surface regions to correct the infrared temperature measurement values.
[0067] In one optional embodiment, the defect evidence three-dimensional mapping module 23 is used for: Based on the registration relationship between the visible light image data and the three-dimensional point cloud data, the two-dimensional image coordinates in the defect alarm information stream are mapped to the spatial coordinate system constructed by the three-dimensional point cloud data to determine the three-dimensional spatial location of the defect. Based on the three-dimensional spatial location of the defect, a three-dimensional spatial evidence unit associated with the defect is constructed; The defect category and confidence information in the defect alarm information stream are assigned to the three-dimensional spatial evidence unit to generate a three-dimensional defect evidence body.
[0068] In one alternative embodiment, the weighted model of evidence theory is defined as follows: ; in, The credibility weight of the i-th information source is obtained based on the historical accuracy statistics of the evidence from the i-th information source. The spatial consistency weight is calculated based on the dispersion of evidence from spatial neighboring voxels of each voxel, which is a preset number of such voxels. The temporal consistency weight is calculated based on the smoothness of evidence changes for each voxel over multiple consecutive time frames. The identification framework is defined as the set of candidate propositions for the state of electrical equipment; A and B are... A subset of propositions; Indicates the i-th information source for the recognition framework The basic confidence assignment function for the subset A of propositions.
[0069] In one optional embodiment, the multi-source evidence fusion module 24 is configured to: The evidence corresponding to the infrared heat source and the evidence corresponding to the defect source, which have been processed by the evidence theory weighted model, are fused and calculated according to the preset evidence fusion rules to obtain the comprehensive basic reliability allocation. During the fusion calculation process, the conflict coefficient between evidence from different sources is evaluated. When the conflict coefficient reaches a preset conflict threshold, the overall basic reliability allocation is adjusted using conflict resolution rules. Based on the adjusted comprehensive basic confidence distribution, the degree of support and non-opposition of each voxel for each candidate power equipment state proposition is calculated. Combined with the three-dimensional spatial location of each voxel, a confidence voxel distribution containing the distribution of the degree of support for each voxel's equipment state is generated.
[0070] In one optional embodiment, the three-dimensional visualization diagnostic system for power equipment status further includes a diagnostic decision information output module, used for: Based on the confidence voxel distribution, the operation and maintenance priority index of power equipment components is evaluated, and diagnostic decision information is generated.
[0071] It should be noted that the three-dimensional visualization diagnostic system for power equipment status provided in this embodiment of the invention is used to execute all the process steps of the three-dimensional visualization diagnostic method for power equipment status in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.
[0072] This invention also provides a terminal device, such as... Figure 3The diagram shown is a structural block diagram of a preferred embodiment of a terminal device provided by the present invention. The terminal device includes a processor 31, a memory 32, and a computer program stored in the memory 32 and configured to be executed by the processor 31. When the processor 31 executes the computer program, it implements the three-dimensional visualization diagnostic method for power equipment status as described in any of the above embodiments.
[0073] In addition, embodiments of the present invention also provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the three-dimensional visualization diagnostic method for power equipment status as described in any of the above embodiments.
[0074] When the processor 31 executes the computer program, it implements the steps in the above-described embodiment of the three-dimensional visualization diagnostic method for the status of power equipment, for example... Figure 1 The method for three-dimensional visualization diagnosis of power equipment status, as shown, includes all steps. Alternatively, when the processor 31 executes the computer program, it implements the functions of each module in the above-described embodiment of the three-dimensional visualization diagnosis system for power equipment status, for example... Figure 2 The diagram illustrates the functions of each module in the 3D visualization diagnostic system for power equipment status.
[0075] Preferably, the computer program can be divided into one or more modules / units, which are stored in the memory 32 and executed by the processor 31 to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the terminal device.
[0076] The processor 31 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or the processor 31 can be any conventional processor. The processor 31 is the control center of the terminal device, connecting various parts of the terminal device through various interfaces and lines.
[0077] The memory 32 mainly includes a program storage area and a data storage area. The program storage area can store the operating system, applications required for at least one function, etc., and the data storage area can store related data, etc. In addition, the memory 32 can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, and a flash card, etc., or the memory 32 can also be other volatile solid-state storage devices.
[0078] It should be noted that the aforementioned terminal devices may include, but are not limited to, processors and memory, as will be understood by those skilled in the art. Figure 3 The structural block diagram shown is merely a structural example of the terminal device described above and does not constitute a limitation on the structure of the terminal device. The terminal device may include more or fewer components than shown, or combine certain components, or use different components.
[0079] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.
Claims
1. A three-dimensional visualization diagnostic method for the status of power equipment, characterized in that, include: Simultaneously collect multimodal sensing data of the target power equipment area; wherein, the data includes three-dimensional point cloud data, visible light image data, infrared thermal imaging data, and defect alarm information stream; A three-dimensional thermal voxel cloud is generated based on the three-dimensional point cloud data and the infrared thermal imaging data; A three-dimensional defect evidence body is generated based on the three-dimensional point cloud data, the visible light image data, and the defect alarm information stream. Based on the three-dimensional thermal voxel cloud and the three-dimensional defect evidence body, an evidence theory weighted model is constructed, and reasoning is performed on the evidence theory weighted model to generate a confidence voxel distribution; The confidence voxel distribution is associated with the component-level semantics of the digital twin model of the power equipment to generate a three-dimensional visualization heat map and defect distribution map of the power equipment components.
2. The three-dimensional visualization diagnostic method for the status of power equipment as described in claim 1, characterized in that, Before generating the three-dimensional thermal voxel cloud based on the three-dimensional point cloud data and the infrared thermal imaging data, the method further includes: Using the spatial coordinate system constructed from the three-dimensional point cloud data as a reference, the external parameter matrices of the visible light image acquisition device and the infrared thermal imaging acquisition device relative to the reference are calculated. Based on the extrinsic parameter matrix, the visible light image data and infrared thermal imaging data corresponding to the synchronization timestamp are mapped onto the three-dimensional point cloud surface to complete the data registration.
3. The three-dimensional visualization diagnostic method for the status of power equipment as described in claim 2, characterized in that, The generation of a three-dimensional thermal voxel cloud based on the three-dimensional point cloud data and the infrared thermal imaging data includes: Physical corrections are performed on the registered infrared thermal imaging data to generate a three-dimensional thermal voxel cloud. The physical corrections include: The infrared thermometric values are corrected based on the emissivity of the materials of the power equipment components and the law of radiation. Invalid temperature data caused by obstruction between components of electrical equipment is detected and compensated using the ray projection method. High reflectivity surface regions in a 3D point cloud are identified, and the amount of environmental heat source reflection interference is estimated based on the high reflectivity surface regions to correct the infrared temperature measurement values.
4. The three-dimensional visualization diagnostic method for the status of power equipment as described in claim 1, characterized in that, The step of generating a three-dimensional defect evidence body based on the three-dimensional point cloud data, the visible light image data, and the defect alarm information stream includes: Based on the registration relationship between the visible light image data and the three-dimensional point cloud data, the two-dimensional image coordinates in the defect alarm information stream are mapped to the spatial coordinate system constructed by the three-dimensional point cloud data to determine the three-dimensional spatial location of the defect. Based on the three-dimensional spatial location of the defect, a three-dimensional spatial evidence unit associated with the defect is constructed; The defect category and confidence information in the defect alarm information stream are assigned to the three-dimensional spatial evidence unit to generate a three-dimensional defect evidence body.
5. The three-dimensional visualization diagnostic method for the status of power equipment as described in claim 1, characterized in that, The weighted model of evidence theory is defined as follows: ； in, The credibility weight of the i-th information source is obtained based on the historical accuracy statistics of the evidence from the i-th information source. The spatial consistency weight is calculated based on the dispersion of evidence from spatial neighboring voxels of each voxel, which is a preset number of such voxels. The temporal consistency weight is calculated based on the smoothness of evidence changes for each voxel over multiple consecutive time frames. The identification framework is defined as the set of candidate propositions for the state of electrical equipment; A and B are... A subset of propositions; Indicates the i-th information source for the recognition framework The basic confidence assignment function for the subset A of propositions.
6. The three-dimensional visualization diagnostic method for the status of power equipment as described in claim 5, characterized in that, The reasoning process for the weighted model of the evidence theory to generate a confidence voxel distribution includes: The evidence corresponding to the infrared heat source and the evidence corresponding to the defect source, which have been processed by the evidence theory weighted model, are fused and calculated according to the preset evidence fusion rules to obtain the comprehensive basic reliability allocation. During the fusion calculation process, the conflict coefficient between evidence from different sources is evaluated. When the conflict coefficient reaches a preset conflict threshold, the overall basic reliability allocation is adjusted using conflict resolution rules. Based on the adjusted comprehensive basic confidence distribution, the degree of support and non-opposition of each voxel for each candidate power equipment state proposition is calculated. Combined with the three-dimensional spatial location of each voxel, a confidence voxel distribution containing the distribution of the degree of support for each voxel's equipment state is generated.
7. The three-dimensional visualization diagnostic method for the status of power equipment as described in claim 1, characterized in that, After associating the confidence voxel distribution with the component-level semantics of the digital twin model of the power equipment to generate a three-dimensional visualized heat map and defect distribution map of the power equipment components, the method further includes: Based on the confidence voxel distribution, the operation and maintenance priority index of power equipment components is evaluated, and diagnostic decision information is generated.
8. A three-dimensional visualization diagnostic system for the status of power equipment, characterized in that, include: The multi-source data acquisition and synchronization module is used to synchronously acquire multimodal sensing data of the target power equipment area; wherein, the data includes three-dimensional point cloud data, visible light image data, infrared thermal imaging data, and defect alarm information stream; A three-dimensional thermal field voxelization module is used to generate a three-dimensional thermal voxel cloud based on the three-dimensional point cloud data and the infrared thermal imaging data. The defect evidence 3D mapping module is used to generate a 3D defect evidence body based on the 3D point cloud data, the visible light image data, and the defect alarm information stream; The multi-source evidence fusion module is used to construct an evidence theory weighted model based on the three-dimensional thermal voxel cloud and the three-dimensional defect evidence body, and to infer the evidence theory weighted model to generate a confidence voxel distribution. The 3D visualization module is used to associate the confidence voxel distribution with the component-level semantics of the digital twin model of the power equipment, and generate a 3D visualized heat distribution map and defect distribution map of the power equipment components.
9. A terminal device, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the three-dimensional visualization diagnostic method for the status of power equipment as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the three-dimensional visualization diagnostic method for the status of power equipment as described in any one of claims 1 to 7.