Lightweight digital twin modeling method and system for power equipment

Abstract

The invention provides a lightweight digital twinborn modeling method and system for power equipment, and relates to the technical field of glass assembly and conveying, and the method comprises the following steps: S1, dynamic sensing: synchronously collecting equipment multi-physical field data through a vibration sensor, an infrared thermal imager and a partial discharge detector; and S2, edge processing: executing in the embedded circuit board. According to the method and the system, vibration, temperature and partial discharge data are synchronously acquired, and a sampling strategy is dynamically matched in combination with an equipment type, so that the early fault detection coverage rate is improved, the omission ratio is reduced, the model volume is reduced based on curvature-driven thinning and octree subdivision, the memory occupation of finite element calculation is reduced, and meanwhile, the precision of a key region is kept; real-time data and historical baselines are fused through dynamic HI values, and health degree assessment errors are stable for a long time through dual-channel verification and deviation triggering correction.

Description

Technical Field

[0001] The present invention relates to the field of glass assembly and transmission technology, and specifically to a lightweight digital twin modeling method and system for power equipment. Background Art

[0002] With the rapid development of smart grids, digital twin technology has become a core tool for power equipment condition monitoring and fault warning. Traditional methods collect data from single sensors, such as vibration, temperature, or partial discharge, and combine it with finite element simulation to build virtual models of the equipment, enabling offline analysis of some conditions. Recent advances in edge computing and multi-physics coupling algorithms have made real-time twin modeling possible, providing the technical foundation for the transition of equipment health management from "post-repair" to "predictive maintenance."

[0003] However, existing technologies still have significant flaws. First, a single sensor cannot fully reflect the mechanical, thermal, and electrical multi-field coupling status of the equipment, resulting in a high rate of missed fault detection. Second, the traditional health index (HI) uses a fixed weight formula and does not consider parameter drift caused by equipment aging. Long-term error accumulation leads to lower model accuracy. Third, traditional twin modeling adopts high data recording accuracy, but too much redundant data leads to overly large modeling specifications, which makes it easy to make errors and has low model simulation prediction efficiency. Summary of the Invention

[0004] The present invention aims to solve the shortcomings existing in the background technology, provide a lightweight digital twin modeling method and system for power equipment, and improve the defects of traditional model sensors using a single sensor, large health deviation, and modeling specifications that are too large to be convenient for calculation and installation.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a lightweight digital twin modeling method for power equipment, comprising the following steps:

[0006] Step S1: Dynamic sensing: synchronously collect multi-physical field data of the equipment through vibration sensors, infrared thermal imagers, and partial discharge detectors;

[0007] Step S2: Edge processing: Execute in the embedded circuit board: vibration data wavelet denoising (threshold λ = 3σ ± 20%), temperature data environmental gradient compensation (compensation coefficient α = 0.8-1.2), dynamic data compression (critical data lossless / non-critical data compression ratio ≥ 5:1);

[0008] Step S3: Lightweight modeling: Simplified modeling is performed based on the data obtained in step S1-2;

[0009] Step S4: Closed-loop verification: input the real-time monitoring data into the finite element simulation model, and trigger the iterative correction of material parameters (step size ±3%) when the mechanical stress deviation is greater than 5%;

[0010] Step S5: Generate three-level warnings based on the model health (HI value) and trigger them in a linked manner:

[0011] -HI drops 5%: Increase sampling rate to 2 times the baseline value

[0012] -HI drops by 10%: Start collaborative analysis of adjacent devices

[0013] -HI drops 15%: dispatch the nearest repair resource.

[0014] Furthermore, the dynamic data compression in step S2 intelligently switches modes based on spectral characteristics:

[0015] Fundamental frequency vibration (<100Hz) uses lossy compression (DCT+entropy coding);

[0016] High-frequency resonance (>500Hz) switches to lossless mode (FLAC encoding);

[0017] Blockchain evidence records are generated synchronously when the mode is switched.

[0018] Furthermore, step S3 also includes:

[0019] Step S31: Automatically identifying the conductive connection portion and the stress concentration area based on the device CAD model;

[0020] Step S32: Maintain 1mm point cloud accuracy for key areas, and use curvature-driven thinning (retention rate 40%-60%) for non-key areas;

[0021] Step S33: Outputting a lightweight three-dimensional model.

[0022] Furthermore, when the curvature-driven thinning in step S32 is implemented: the planar area (curvature K < 0.01 / mm) is thinned to 40%; the arc transition area (0.01 ≤ K < 0.1 / mm) retains 60% of the point cloud; and the edge features (K ≥ 0.1 / mm) maintain the complete topological structure.

[0023] Furthermore, the closed-loop verification in step S4 includes a dual-channel mechanism:

[0024] Channel 1: Predicting device state evolution based on LSTM;

[0025] Channel 2: Apply material mechanics formulas to verify rationality;

[0026] When the dual-channel deviation is greater than 8%, manual review is triggered.

[0027] Furthermore, the model health in step S5 is a HI value, where the HI value is generated by weighted fusion, where

[0028] A lightweight digital twin modeling system for power equipment, characterized by comprising:

[0029] Sensing layer: Contains MEMS vibration array (range ±50g), fiber optic temperature measurement network (accuracy ±0.5°C), and ultra-high frequency partial discharge detection unit (300MHz-1.5GHz);

[0030] Edge layer: equipped with a multi-core processor, built-in adaptive compression engine, and device feature extraction library;

[0031] Twin layer: integrated lightweight 3D model (1mm accuracy in key areas) and multi-physics coupling simulator;

[0032] Decision-making layer: Provides blockchain evidence chain (full life cycle data), three-level response instruction generation and spare parts supply chain interface.

[0033] Furthermore, the lightweight modeling module of the twin layer adopts the following technologies:

[0034] A. Octree subdivision of key areas (minimum voxel 0.5mm);

[0035] B. Simplify the Delaunay triangulation in non-critical areas (reducing the number of faces by 60%);

[0036] C. Topology optimization maintains mechanical properties.

[0037] Further,.

[0038] Furthermore, the system has a closed-loop data flow connecting multiple levels:

[0039] Upstream: Perception layer → Edge layer (raw data, bandwidth priority);

[0040] Downstream: Decision layer → Perception layer (sampling strategy, latency sensitivity);

[0041] Verification flow: Twin layer → Edge layer (calibration instructions, QoS guarantee);

[0042] Each data stream uses IEEE 1588 protocol time synchronization.

[0043] The present invention provides a lightweight digital twin modeling method and system for power equipment, which has the following features:

[0044] Beneficial effects:

[0045] The advantage of the present invention is that it synchronously collects vibration, temperature and partial discharge data, and dynamically matches the sampling strategy with the equipment type, thereby improving the early fault detection coverage and reducing the missed detection rate.

[0046] Secondly, based on curvature-driven thinning and octree subdivision, the model volume is reduced and the memory usage of finite element calculations is reduced, while the accuracy of key areas is maintained;

[0047] Finally, by integrating real-time data and historical baselines with dynamic HI values, and through dual-channel verification and deviation-triggered correction, the health assessment error is stabilized in the long term. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 Schematic diagram of the method structure of the present invention.

[0049] Figure 2 Schematic diagram of the system structure of the present invention.

[0050] Figure 3 Schematic diagram of closed-loop verification of the model of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the embodiments described are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0052] The disclosure below provides many different embodiments or examples for realizing different structures of the present application. In order to simplify the disclosure of the present application, the components and settings of specific examples are described below. Of course, they are merely examples and are not intended to limit the present application. In addition, the present application may repeat reference numbers and / or reference letters in different examples, and such repetition is for the purpose of simplicity and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed. In addition, the present application provides examples of various specific processes and materials, but those of ordinary skill in the art will appreciate the application of other processes and / or the use of other materials.

[0053] The embodiment of the present application provides a lightweight digital twin modeling method and system for power equipment. The lightweight digital twin modeling method and system for power equipment can realize the synchronous collection of vibration, temperature, and partial discharge data, and combine the equipment type dynamic matching sampling strategy to improve the early fault detection coverage and reduce the missed detection rate. Based on curvature-driven thinning and octree subdivision, the model volume is reduced, and the memory usage of finite element calculation is reduced. At the same time, the accuracy of key areas is maintained. The real-time data and historical baseline are integrated through dynamic HI value, and the health assessment error is stabilized in the long term through dual-channel verification and deviation trigger correction. The lightweight digital twin modeling method and system for power equipment are described in detail below. It should be noted that the description order of the following embodiments does not limit the preferred order of the embodiments.

[0054] The following describes the present application in detail with reference to the accompanying drawings and specific implementation methods. Figure 1-3 In this embodiment, the embodiment 1 provided

[0055] A lightweight digital twin modeling method for power equipment includes the following steps:

[0056] Step S1: Dynamic sensing: synchronously collect multi-physical field data of the equipment through vibration sensors, infrared thermal imagers, and partial discharge detectors;

[0057] Step S2: Edge processing: Execute in the embedded circuit board: vibration data wavelet denoising (threshold λ = 3σ ± 20%), temperature data environmental gradient compensation (compensation coefficient α = 0.8-1.2), dynamic data compression (critical data lossless / non-critical data compression ratio ≥ 5:1);

[0058] Step S3: Lightweight modeling: Simplified modeling is performed based on the data obtained in step S1-2;

[0059] Step S4: Closed-loop verification: input the real-time monitoring data into the finite element simulation model, and trigger the iterative correction of material parameters (step size ±3%) when the mechanical stress deviation is greater than 5%;

[0060] Step S5: Generate three-level warnings based on the model health (HI value) and trigger them in a linked manner:

[0061] -HI drops 5%: Increase sampling rate to 2 times the baseline value

[0062] -HI drops by 10%: Start collaborative analysis of adjacent devices

[0063] -HI drops 15%: dispatch the nearest repair resource.

[0064] During the modeling process, traditional single sensors cannot fully reflect the multi-physical field status of equipment. Instead, deploying vibration sensors (10Hz-10kHz) to capture mechanical anomalies, infrared thermal imagers (≤0.05°C) to monitor temperature distribution, and partial discharge detectors (1MHz-30MHz) to capture insulation defects. Automatically matching the sampling frequency to the equipment type (for example, transformers focus on low frequencies, while GIS equipment focuses on high frequencies) can improve multi-dimensional data fusion and enhance fault detection coverage.

[0065] Furthermore, the dynamic data compression in step S2 intelligently switches modes based on spectral characteristics: fundamental frequency vibration (<100Hz) uses lossy compression (DCT+entropy coding); high-frequency resonance (>500Hz) switches to lossless mode (FLAC encoding); blockchain evidence records are generated synchronously when switching modes;

[0066] Because the raw data contains noise and occupies a large amount of bandwidth, direct transmission to the cloud is inefficient. Therefore, a dynamic threshold (λ = 3σ ± 20%) is used to filter out vibration signal noise. A three-dimensional gradient model is constructed using 8-12 environmental sensors to correct the measurement values. At the same time, the fundamental frequency vibration is encoded using DCT + entropy (compression ratio ≥ 5:1), and high-frequency resonance is switched to FLAC lossless mode to improve data processing.

[0067] Furthermore, step S3 also includes:

[0068] Step S31: Automatically identifying the conductive connection portion and the stress concentration area based on the device CAD model;

[0069] Step S32: Maintain 1mm point cloud accuracy for key areas, and use curvature-driven thinning (retention rate 40%-60%) for non-key areas;

[0070] Step S33: outputting a lightweight three-dimensional model;

[0071] Furthermore, when the curvature-driven thinning in step S32 is implemented: the planar area (curvature K < 0.01 / mm) is thinned to 40%; the arc transition area (0.01 ≤ K < 0.1 / mm) retains 60% of the point cloud; the edge features (K ≥ 0.1 / mm) maintain the complete topological structure;

[0072] Since traditional full-precision modeling requires too much data and cannot interact in real time, the CAD model is used to automatically identify conductive connections (such as sleeve joints) and stress concentration areas (such as flange edges), and non-critical areas are thinned out according to the curvature (flat surfaces are thinned to 40%, and curved surfaces retain 60%) to reduce the model volume and the loss of finite elements.

[0073] Furthermore, the closed-loop verification in step S4 includes a dual-channel mechanism:

[0074] Channel 1: Predicting device state evolution based on LSTM;

[0075] Channel 2: Apply material mechanics formulas to verify rationality;

[0076] When the dual-channel deviation is greater than 8%, manual review is triggered.

[0077] Equipment materials will produce comparative deviations during long-term use. At this time, the model needs to be dynamically corrected. By inputting real-time stress data into the ANSYS finite element model, the deviation between the simulation value and the measured value is calculated. When the deviation is greater than 5%, the elastic modulus is automatically adjusted, and the historical version is retained for trend analysis.

[0078] Furthermore, the model health in step S5 is a HI value, where the HI value is generated by weighted fusion, where

[0079] Its simplified description is: HI = (current abnormality level of the three major indicators ÷ safety line) × dynamic weight + historical health change × attenuation weight. The HI value can be used to determine the overall health of the model, and the health value of the power equipment can be determined through the overall health. Then, twin modeling is used to model and protect the power equipment. The model simulates the health value of the power equipment and determines its working status, making it easier for staff to capture equipment abnormalities and repair the equipment in a timely manner.

[0080] Example 2

[0081] Based on the first embodiment, a lightweight digital twin modeling system for power equipment is provided, characterized by comprising:

[0082] Sensing layer: Contains MEMS vibration array (range ±50g), fiber optic temperature measurement network (accuracy ±0.5°C), and ultra-high frequency partial discharge detection unit (300MHz-1.5GHz);

[0083] Edge layer: equipped with a multi-core processor, built-in adaptive compression engine, and device feature extraction library;

[0084] Twin layer: integrated lightweight 3D model (1mm accuracy in key areas) and multi-physics coupling simulator;

[0085] Decision-making layer: Provides blockchain evidence chain (full life cycle data), three-level response instruction generation and spare parts supply chain interface;

[0086] Furthermore, the lightweight modeling module of the twin layer adopts the following technologies:

[0087] A. Key area octree subdivision (minimum voxel 0.5mm)

[0088] B. Delaunay triangulation simplification in non-critical areas (reducing the number of faces by 60%)

[0089] C. Topology optimization maintains mechanical properties.

[0090] The perception layer uses a MEMS vibration array (±50g range) to capture the equipment's mechanical vibration spectrum in real time. A fiber-optic temperature measurement network constructs a three-dimensional temperature field with ±0.5°C accuracy. A UHF partial discharge detection unit (300MHz-1.5GHz) precisely captures electromagnetic signals from insulation defects. These three elements, triggered synchronously at the microsecond level, enable holographic multi-physics field perception. Their high-density data captures mechanical, thermal, and electrical anomaly characteristics, improving early fault detection and providing a comprehensive raw data pool for subsequent analysis.

[0091] At the edge layer, multi-core processors process raw data from the perception layer in parallel. An adaptive compression engine dynamically switches between DCT and FLAC encoding, compressing data bandwidth to less than one-fifth. A temperature gradient compensation model corrects thermal imaging deviations based on 8-12 environmental reference points. A feature extraction library extracts impact components from vibration signals through wavelet packet decomposition (10 layers). Real-time processing at the edge effectively filters out invalid data, reducing cloud load while retaining key fault characteristics and providing a lightweight data stream for the twin layer.

[0092] Twin layer: Based on the 1mm precision point cloud model of the edge layer preprocessing, combined with the multi-physics field coupling simulator (mechanical-thermal-electrical coupling algorithm), the device status is dynamically mapped: the stress concentration area is predicted by ANSYS transient analysis, the temperature field iteratively corrects the material thermal conductivity, and the partial discharge signal drives the insulation aging model.

[0093] The decision-making layer relies on a blockchain evidence chain to consolidate full lifecycle data (3-5 pieces of evidence per second). When the health index (HI) drops to a threshold, a three-level response engine automatically executes the strategy: a 5% deviation triggers high-frequency monitoring, a 10% deviation calls for collaborative verification with adjacent equipment, and a 15% deviation leverages the supply chain interface to dispatch the nearest spare part. Through trusted data traceability and smart contract execution, the efficiency of emergency response to major failures is improved and the rate of mis-purchased spare parts is reduced.

[0094] Furthermore, the system has a closed-loop data flow connecting multiple levels:

[0095] Upstream: Perception layer → Edge layer (raw data, bandwidth priority);

[0096] Downstream: Decision layer → Perception layer (sampling strategy, latency sensitivity);

[0097] Verification flow: Twin layer → Edge layer (calibration instructions, QoS guarantee);

[0098] Each data stream uses IEEE 1588 protocol time synchronization;

[0099] During the modeling process, one-way data flow cannot support real-time feedback. Therefore, upstream flow, downstream flow and verification flow are used in combination to reduce data transmission delay and avoid excessive network bandwidth utilization during data transmission.

[0100] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0101] The above is a detailed introduction to a lightweight digital twin modeling method and system for electric power equipment provided in the embodiments of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the technical solutions and core ideas of the present application; ordinary technicians in this field should understand that they can still modify the technical solutions recorded in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A lightweight digital twin modeling method for power equipment, characterized in that: The steps include: Step S1: Dynamic sensing: synchronously collect multi-physical field data of the equipment through vibration sensors, infrared thermal imagers, and partial discharge detectors; Step S2: Edge processing: Execute in the embedded circuit board: vibration data wavelet denoising (threshold λ = 3σ ± 20%), temperature data environmental gradient compensation (compensation coefficient α = 0.8-1.2), dynamic data compression (critical data lossless / non-critical data compression ratio ≥ 5:1); Step S3: Lightweight modeling: Simplified modeling is performed based on the data obtained in step S1-2; Step S4: Closed-loop verification: input the real-time monitoring data into the finite element simulation model, and trigger the iterative correction of material parameters (step size ±3%) when the mechanical stress deviation is greater than 5%; Step S5: Generate three-level warnings based on the model health and trigger them in a coordinated manner: -HI drops by 5%: Increase the sampling rate to 2 times the baseline value; -HI drops by 10%: start collaborative analysis of adjacent devices; -HI drops 15%: dispatch the nearest repair resource.

2. The lightweight digital twin modeling method for power equipment according to claim 1 is characterized in that: The dynamic data compression in step S2 intelligently switches modes through spectrum characteristics: Fundamental frequency vibration (<100Hz) uses lossy compression (DCT+entropy coding); High-frequency resonance (>500Hz) switches to lossless mode (FLAC encoding); Blockchain evidence records are generated synchronously when the mode is switched.

3. The lightweight digital twin modeling method for power equipment according to claim 1 is characterized in that: The step S3 further includes: Step S31: Automatically identifying the conductive connection portion and the stress concentration area based on the device CAD model; Step S32: Maintain 1mm point cloud accuracy for key areas, and use curvature-driven thinning (retention rate 40%-60%) for non-key areas; Step S33: Outputting a lightweight three-dimensional model.

4. The lightweight digital twin modeling method for power equipment according to claim 3 is characterized in that: When the curvature-driven thinning in step S32 is implemented: the planar area (curvature K < 0.01 / mm) is thinned to 40%; the arc transition area (0.01≤K < 0.1 / mm) retains 60% of the point cloud; the edge features (K≥0.1 / mm) maintain the complete topological structure.

5. The lightweight digital twin modeling method for power equipment according to claim 1 is characterized in that: The closed-loop verification of step S4 includes a dual-channel mechanism: Channel 1: Predicting device state evolution based on LSTM; Channel 2: Apply material mechanics formulas to verify rationality; When the dual-channel deviation is greater than 8%, manual review is triggered.

6. The lightweight digital twin modeling method for power equipment according to claim 1, characterized in that: The model health in step S5 is an HI value, wherein the HI value is generated by weighted fusion, wherein 7. A lightweight digital twin modeling system for power equipment applicable to the method of claims 1-5, characterized in that: include: Sensing layer: Contains MEMS vibration array (range ±50g), fiber optic temperature measurement network (accuracy ±0.5°C), and ultra-high frequency partial discharge detection unit (300MHz-1.5GHz); Edge layer: equipped with a multi-core processor, built-in adaptive compression engine, and device feature extraction library; Twin layer: integrated lightweight 3D model (1mm accuracy in key areas) and multi-physics coupling simulator; Decision-making layer: Provides blockchain evidence chain (full life cycle data), three-level response instruction generation and spare parts supply chain interface.

8. The lightweight digital twin modeling system for power equipment according to claim 6, characterized in that: The lightweight modeling module of the twin layer adopts the following technology: A. Octree subdivision of key areas (minimum voxel 0.5mm); B. Simplify the Delaunay triangulation in non-critical areas (reducing the number of faces by 60%); C. Topology optimization maintains mechanical properties.

9. The lightweight digital twin modeling system for power equipment according to claim 1, characterized in that: The system has a closed-loop data flow connecting multiple levels: Upstream: Perception layer → Edge layer (raw data, bandwidth priority); Downstream: Decision layer → Perception layer (sampling strategy, latency sensitivity); Verification flow: Twin layer → Edge layer (calibration instructions, QoS guarantee); Each data stream uses IEEE 1588 protocol time synchronization.

Images