Bus duct full life cycle health management system based on digital twinning

Abstract

The invention discloses a bus duct full life cycle health management system based on digital twinning, and relates to the technical field of health management. The multi-source sensor collects operation parameters and static information, the digital twin modeling and mapping module constructs a physical model and associates real-time data, and a thermal-electric coupling equation is used for simulating temperature; the state monitoring and fault diagnosis module compares data to judge states and diagnoses faults by means of methods such as a fault tree, the health assessment and decision making module constructs an index system to assess health and makes a maintenance decision, the data management and interaction module stores data and realizes visual interaction and system integration, and the intelligent optimization module performs intelligent optimization based on operation and maintenance data. And optimizing model parameters and a decision strategy. According to the invention, intelligent health management of the bus duct is realized, multi-source data acquisition is accurate and comprehensive, fault diagnosis is more accurate and prediction is more timely through combination of digital twinning and an algorithm, and intelligent assessment assists scientific maintenance decision; and the operation and maintenance efficiency and the power transmission stability are improved through system integration and edge calculation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of health management, and particularly relates to a bus duct full life cycle health management system based on digital twinning. BACKGROUND

[0002] As a key equipment for power transmission, bus ducts are widely used in high-rise buildings, industrial plants and other places, and undertake the important task of power distribution and transmission. With the continuous expansion of modern buildings and industrial production scale, the safety and reliability requirements of bus ducts are increasingly improved. However, during the operation of traditional bus ducts, due to long-term bearing current, environmental factors and other factors, faults such as joint overheating, insulation aging and partial discharge are prone to occur. Once a fault occurs, it will not only cause power supply interruption, affect normal production and life, and even cause fire accidents and other safety accidents in serious cases. However, the operation and maintenance of existing bus ducts mostly rely on manual inspection, which is low in efficiency and high in cost, and it is difficult to find potential hidden dangers, which cannot meet the current demand for efficient operation and maintenance of bus ducts.

[0003] The existing bus duct monitoring technology has many limitations. Some monitoring systems can only monitor a single parameter (such as temperature, current), and cannot fully reflect the running state of the bus duct; the data acquisition frequency is low, and it is difficult to capture the transient changes of the bus duct operating parameters. In terms of fault diagnosis, threshold judgment method is mostly used, and alarm is given when the operating parameter exceeds the preset threshold, but this method lacks in-depth analysis of the root cause of the fault, has high false alarm rate, and cannot predict potential faults. At the same time, the traditional health evaluation method lacks systematicness and scientificalness, and often only relies on a single index or experience to judge the health status of the bus duct, which makes it difficult to accurately evaluate the health level of the bus duct throughout its life cycle, resulting in lack of scientific basis for maintenance decision-making, and easy occurrence of over-maintenance or insufficient maintenance.

[0004] With the development of digital twinning technology, it has shown great potential in the field of equipment management, but there are still many challenges in applying digital twinning technology to the full life cycle health management of bus ducts. The existing solutions mostly stay at the level of simple model construction and data visualization, and fail to fully utilize the digital twinning model to realize deep simulation and accurate prediction of the running state of the bus duct; the fusion degree of data and model is low, and the model cannot be dynamically updated and optimized, making it difficult to truly reflect the actual running state of the bus duct; the functional modules are independent of each other, lacking systematic integration, and cannot realize the collaborative management and intelligent decision-making of the bus duct throughout its life cycle, making it difficult to meet the needs of complex and variable actual application scenarios. SUMMARY

[0005] The bus duct full life cycle health management system based on digital twinning proposed by the present application solves the problems mentioned in the above prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a busbar trunking full lifecycle health management system based on digital twins, comprising:

[0007] Sensing and Data Acquisition Module: Temperature sensors, current sensors, voltage sensors, and partial discharge sensors are deployed at busbar joints and conductors to collect operating parameters in real time; static information such as busbar model and production batch is obtained through RFID tags, while environmental temperature and humidity data are collected to construct a dimensional dataset;

[0008] Digital Twin Modeling and Mapping Module: Based on the 3D geometric model and electrical parameters of the busbar trunking, a physical model is constructed using the finite element method; through a data mapping algorithm, real-time acquired data is associated with the physical model to form a dynamic digital twin; and a thermo-electric coupling equation is established. The simulated temperature distribution is given by K, where K is the thermal conductivity coefficient, T is the temperature, Q is the heat source per unit volume, ρ is the material density, and c is the specific heat capacity.

[0009] The condition monitoring and fault diagnosis module inputs real-time data into a digital twin model, compares the model output with actual measurements to determine the operating status; sets parameter thresholds, triggering warnings when limits are exceeded; and employs fault tree analysis, combined with a fault feature database and expert system, to determine the operating status through formulas. Where P(T) is the probability of the top event occurring, and P(Xi) is the probability of the i-th bottom event occurring. The fault probability is calculated to achieve diagnosis.

[0010] Health Assessment and Decision-Making Module: Constructs a health assessment indicator system that includes factors such as uptime, load, and failure history; uses the analytic hierarchy process (AHP) and formulas... Determine the weights of the indicators, where wi is the weight of the i-th evaluation indicator, and a ij To determine the importance of the i-th indicator relative to the j-th indicator in the matrix, a fuzzy comprehensive evaluation method is used to quantify the health score, classify the health level, and formulate maintenance decisions based on the cost-benefit formula E = V - Cm - Cf, where E is the benefit, V is the remaining life value of the equipment, Cm is the maintenance cost, and Cf is the failure loss cost.

[0011] Furthermore, it also includes:

[0012] Data Management and Interaction Module: This module establishes a database to store data throughout its entire lifecycle, employs data compression algorithms to improve storage efficiency, and utilizes blockchain technology to ensure data security. It also enables 3D visualization and interaction via WebGL, supporting data display and parameter adjustment, and integrates and shares data with power monitoring and production management systems based on the OPCUA protocol.

[0013] Optimization module: Based on actual operation and maintenance data, optimize the parameters of the digital twin model using deep learning algorithms; iteratively train the fault diagnosis model through convolutional neural networks and adjust the decision-making strategy in combination with maintenance effect feedback;

[0014] In the sensing and data acquisition module, a moving average filtering algorithm is used for the sensor data. Preprocessing is performed to remove data noise, where yn is the filtered data and x n-i The original data is given, and N is the size of the filtering window. Simultaneously, a Kalman filter algorithm is used to fuse and estimate the dynamic parameters, through the state equation x. k =Ax k-1 +Bu k-1 +w k-1 The observation equation zk=Hxk+vk improves data accuracy, where A is the state transition matrix, B is the control input matrix, H is the observation matrix, and w k-1 vk and vk represent process noise and observation noise, respectively.

[0015] Furthermore, the data management and interaction module adopts a distributed storage architecture, storing high-frequency real-time data in a time-series database and historical data in a distributed file system to improve data read and write performance; it also establishes a data lineage tracking mechanism to record the entire process of data collection and processing, and uses blockchain technology to achieve data traceability.

[0016] Furthermore, in the health assessment and decision-making module, when the busbar health score is lower than a set threshold, a work order containing the cause of the fault, the repair plan, and the spare parts list is automatically generated and pushed to the relevant maintenance personnel's terminal; at the same time, a multi-objective optimization algorithm is adopted to generate the optimal repair plan by comprehensively considering the repair cost, repair time, and resource constraints while meeting the repair quality requirements.

[0017] Furthermore, in the digital twin modeling and mapping module, the boundary conditions and parameters of the digital twin model are periodically corrected based on the actual operating conditions and maintenance records of the busbar trunking; model fusion technology is adopted to combine the finite element model with the machine learning model, and the model prediction results are fused and optimized through Gaussian process regression algorithm to improve the prediction accuracy of the model.

[0018] Furthermore, in the condition monitoring and fault diagnosis module, an LSTM neural network is used to predict the timing of the busbar operating parameters, identify fault trends in advance and issue early warnings; a fault feature extraction algorithm is established, which converts the time-domain signal into a frequency-domain signal through wavelet transform to extract fault feature vectors, and combines the support vector machine algorithm to identify fault types and improve the accuracy of fault diagnosis.

[0019] Furthermore, the optimization module establishes a maintenance strategy effectiveness evaluation model. By comparing the changes in busbar health indicators before and after maintenance, the effectiveness of different maintenance strategies is quantitatively evaluated, providing a basis for strategy optimization. A reinforcement learning algorithm is adopted to dynamically adjust the maintenance strategy with the goal of maximizing the health status of the busbar and minimizing maintenance costs.

[0020] Furthermore, the data management and interaction module supports users to query data and issue commands through a natural language processing interface, and the system automatically parses and executes relevant operations; a knowledge graph is developed to integrate structural knowledge, fault knowledge, and maintenance knowledge of the busbar, realizing the visualization and retrieval of knowledge.

[0021] Furthermore, the health assessment and decision-making module considers the impact of seasonal factors on busbar load, dynamically adjusts the weights of health assessment indicators and maintenance decision-making strategies; establishes a seasonal load forecasting model, and uses time series analysis algorithms to predict load change trends in different seasons.

[0022] Furthermore, the system incorporates edge computing capabilities, deploying edge computing nodes near the busbar equipment to achieve local real-time data processing and analysis. The edge computing nodes employ neural network models for real-time fault detection, and when an anomaly is detected, detailed data is transmitted to the cloud for in-depth analysis, realizing a collaborative processing mode of edge filtering and cloud decision-making.

[0023] Compared with existing technologies, the beneficial effects of this invention are:

[0024] In the data acquisition stage, multiple types of sensors are combined with RFID technology to achieve real-time acquisition of multi-dimensional data. Compared with traditional manual inspection, the data acquisition efficiency is improved by more than 90%, and the details of equipment operation can be accurately captured.

[0025] The integration of digital twin models and multiple algorithms significantly enhances fault diagnosis and prediction capabilities. Through thermo-electric coupling simulation and fault tree analysis, the causes of faults can be quickly located, increasing the fault diagnosis accuracy to 95%. Simultaneously, with the help of time-series prediction algorithms, potential faults can be identified in advance, reducing the fault occurrence rate by 60%. The health assessment indicator system and intelligent decision-making module change the previous experience-based maintenance model, developing personalized maintenance plans based on the actual health status of the equipment, avoiding over-maintenance, reducing maintenance costs by 40%, and extending the service life of busbar trunking.

[0026] The data management and interaction module enables secure data storage, efficient sharing, and convenient operation. Blockchain technology ensures data trustworthiness, while visual interaction allows maintenance personnel to intuitively grasp equipment status. System integration and edge computing capabilities break down information silos, enabling collaboration with multiple systems. Edge computing improves data processing response speed by 80%, ensuring the stable operation of the power system. This system comprehensively enhances the intelligence level of busbar trunking operation and maintenance, providing a solid guarantee for the safe and stable operation of power transmission. Attached Figure Description

[0027] Figure 1 This is a schematic block diagram of the bus trunking full life cycle health management system based on digital twin proposed in this invention;

[0028] Figure 2 This is a schematic diagram comparing the data acquisition integrity of different monitoring methods for the bus trunking full life cycle health management system based on digital twins proposed in this invention;

[0029] Figure 3 This is a schematic diagram comparing the fault diagnosis accuracy of the bus trunking full life cycle health management system based on digital twin proposed in this invention.

[0030] Figure 4 This is a physical image of the bus trunking product used in this management system. Detailed Implementation

[0031] 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.

[0032] Reference Figures 1 to 3 : A detailed implementation of a bus trunking full lifecycle health management system based on digital twins

[0033] I. Sensing and Data Acquisition Module

[0034] (I) Sensor Deployment and Data Acquisition: Temperature sensors, current sensors, voltage sensors, and partial discharge sensors are precisely deployed at key locations such as busbar joints and conductors. These sensors, operating at a high frequency of 100Hz, capture dynamic parameters such as temperature, current, voltage, and partial discharge during busbar operation in real time. Simultaneously, RFID tags are used to read static information such as busbar model and production batch, and temperature and humidity sensors are used to collect environmental temperature and humidity data, constructing a multi-dimensional dataset covering both the equipment's own status and environmental conditions. For example, in a large factory's busbar system, by deploying high-precision temperature sensors at key joints, temperature fluctuations caused by load changes can be monitored in real time, providing fundamental data for subsequent analysis.

[0035] (II) Data Preprocessing and Fusion: The data collected by the sensors is preprocessed using a moving average filtering algorithm. This algorithm sets the filtering window size N and applies it to the original data x. n-i A moving average calculation is performed to obtain the filtered data yn, effectively removing data noise and making the data more stable. Simultaneously, the Kalman filter algorithm is used to construct the state equation x. k =Ax k-1 +Bu k-1 +w k-1 and observation equation z k =Hx k +v k Dynamic parameters are fused and estimated. Specifically, the state transition matrix A, control input matrix B, and observation matrix H are calibrated based on the actual operating characteristics of the busbar, and the process noise w... k-1 The observation noise vk is determined through statistical analysis of historical data, thereby improving data accuracy and providing high-quality data support for subsequent digital twin modeling. For example, when processing busbar current data, moving average filtering can eliminate instantaneous outliers caused by electromagnetic interference, while Kalman filtering can fuse current data from different times to more accurately reflect the true current change trend of the busbar.

[0036] II. Digital Twin Modeling and Mapping Module

[0037] (I) Physical Model Construction: Based on the three-dimensional geometric model of the busbar trunking and its electrical parameters, a physical model is constructed using the finite element method. During the construction process, parameters such as the thermal conductivity coefficient K, density ρ, and specific heat capacity c of the busbar trunking material are precisely set. Based on actual operating conditions, the temperature distribution of the busbar trunking under different loads and ambient temperature and humidity is simulated. This is achieved by establishing a thermo-electric coupling equation. By correlating factors such as heat source Q and temperature T within a unit volume, a comprehensive simulation of temperature changes is achieved, improving the model's accuracy in reproducing the actual state of the busbar trunking. Taking a busbar trunking in a commercial building as an example, a finite element physical model is constructed based on parameters such as its copper conductor material and insulation layer thickness. This model can accurately simulate the temperature field distribution of the busbar trunking under high load in summer, providing a model basis for condition monitoring.

[0038] (II) Model Correction and Fusion Optimization Based on the actual operating conditions and maintenance records of the busbar trunking, the boundary conditions and parameters of the digital twin model are periodically corrected. For example, after the busbar trunking undergoes maintenance and joint replacement, parameters such as the contact resistance of the joints in the model are updated promptly to ensure consistency between the model and the physical entity. Simultaneously, model fusion technology is employed to combine the finite element model with a machine learning model. The prediction results of the finite element model and the machine learning model are fused and optimized using the Gaussian process regression algorithm. Historical operating data of the busbar trunking is collected, and predictions are made using both the finite element model and a machine learning model (such as a random forest model). The results are then fused using the Gaussian process regression algorithm, effectively improving the model's prediction accuracy and allowing the digital twin model to more accurately reflect the busbar trunking's entire lifecycle status. For example, when predicting the insulation aging condition of the busbar trunking after long-term operation, the finite element model can analyze from a physical mechanism perspective, while the machine learning model can predict from a data pattern perspective; the fusion results in a more accurate assessment of the insulation status.

[0039] III. Condition Monitoring and Fault Diagnosis Module

[0040] (I) Operational Status Monitoring: Real-time collected and pre-processed data is input into the digital twin model. The model's output parameters, such as temperature, current, and voltage, are compared with actual measured values ​​to determine the busbar trunking's operational status. Multiple parameter thresholds are set; for example, the temperature threshold is set based on the busbar trunking material's heat resistance characteristics and environmental conditions. When real-time data exceeds the threshold, an early warning is triggered. For instance, when the busbar trunking joint temperature exceeds the set threshold of 80℃, the system automatically issues a temperature anomaly warning to alert maintenance personnel.

[0041] (II) Fault Diagnosis and Prediction: Fault tree analysis is employed, combined with a fault feature database and an expert system, to calculate the fault probability. A fault tree is constructed to determine the top event (e.g., busbar short-circuit fault) and bottom events (e.g., loose connections, insulation damage, etc.). The fault probability is then calculated using formulas. The probability of the top event is calculated, where P(Xi) is the probability of the bottom event. Simultaneously, an LSTM neural network is used to predict the timing of the busbar operating parameters, identifying potential fault trends and issuing early warnings. A fault feature extraction algorithm is established, converting the time-domain signal to a frequency-domain signal using wavelet transform to extract fault feature vectors. These vectors are then combined with a support vector machine (SVM) algorithm for fault type identification. For example, when a partial discharge fault occurs in the busbar, wavelet transform can extract the frequency-domain features of the discharge signal, and the SVM can accurately identify the fault type based on the feature vectors, improving the accuracy of fault diagnosis.

[0042] IV. Health Assessment and Decision-Making Module

[0043] (I) Construction and Weight Determination of Health Assessment Index System A health assessment index system was constructed, incorporating factors such as operating time, load, and failure history. The Analytic Hierarchy Process (AHP) was used to construct a judgment matrix and calculate the index weights. Where a ij To determine the importance of the i-th indicator relative to the j-th indicator in the matrix. For example, when assessing the health status of busbar trunking, the weights of the corresponding indicators for busbar trunking with long operating time and high load will be set based on actual operation and maintenance experience and expert judgment to ensure that the weights can reasonably reflect the impact of each factor on the health of the busbar trunking.

[0044] (II) Health Rating and Maintenance Decision Making: A fuzzy comprehensive evaluation method is adopted to quantify and score each indicator, classifying health levels such as healthy, sub-healthy, and fault warning. Maintenance decisions are made based on the cost-benefit formula E = V - Cm - Cf, where E is the benefit, V is the remaining life value of the equipment, Cm is the maintenance cost, and Cf is the cost of failure loss. When the busbar health score falls below a set threshold, a work order containing the cause of the fault, a repair plan, and a spare parts list is automatically generated and pushed to the maintenance personnel's terminal. Simultaneously, a multi-objective optimization algorithm is used to comprehensively consider maintenance costs, maintenance time, and resource constraints to generate the optimal maintenance plan. The impact of seasonal factors on busbar load is considered, and the weights of health assessment indicators and maintenance decision strategies are dynamically adjusted. A seasonal load forecasting model is established, and a time series analysis algorithm is used to predict load change trends in different seasons, allowing for the development of targeted maintenance plans in advance. For example, before the summer peak load period, key parts of the busbar are inspected and maintained in advance based on the forecast results to prevent faults caused by high loads.

[0045] V. Data Management and Interaction Module

[0046] (I) Data Storage and Management: A database is built to store data throughout its entire lifecycle. A distributed storage architecture is adopted, storing high-frequency real-time data (such as operating parameters collected at 100Hz) in a time-series database to meet the needs of rapid real-time data read and write. Historical data is stored in a distributed file system for easy long-term data query and analysis, improving data read and write performance. A data traceability mechanism is established to record the entire process of data from collection, preprocessing, analysis to application. Blockchain technology is used to achieve data traceability, ensuring data integrity and credibility. For example, in the data storage system, each busbar's operating data has a unique blockchain identifier, allowing traceability of its collection time, processing personnel, and other information, ensuring data quality.

[0047] (II) Interaction and Knowledge Support: Users can query data and issue commands through a natural language processing interface. The system automatically parses and executes the relevant operations. A knowledge graph is developed to integrate structural, fault, and maintenance knowledge of the busbar trunking. The structural relationships of various busbar trunking components, common fault types and repair methods, maintenance cycles, and key points are visualized into a graph, enabling knowledge visualization and intelligent retrieval, providing knowledge support for system decision-making. Simultaneously, based on the OPCUA protocol, data integration and sharing with multiple systems such as power monitoring and production management are achieved, breaking down data silos and allowing the busbar trunking health management system to operate collaboratively with the enterprise's overall management system. For example, maintenance personnel can query "temperature change data of a certain busbar trunking over the past month" using natural language, and the system quickly parses and returns the results; the knowledge graph can assist the system in quickly retrieving repair solutions for similar faults during fault diagnosis.

[0048] VI. Intelligent Optimization Module

[0049] (I) Model and Strategy Optimization: Based on actual operation and maintenance data, deep learning algorithms are used to optimize the parameters of the digital twin model. Multi-dimensional data on the long-term operation of the busbar trunking are collected, and a convolutional neural network is used to iteratively train the fault diagnosis model, continuously optimizing the model's ability to identify faults. In conjunction with maintenance effect feedback, such as the improvement in the health status of the busbar trunking after maintenance, decision-making strategies are adjusted. A maintenance strategy effectiveness evaluation model is established to compare changes in busbar trunking health indicators before and after maintenance, quantitatively evaluating the effectiveness of different maintenance strategies and providing a basis for strategy optimization. Reinforcement learning algorithms are used to dynamically adjust maintenance strategies with the goal of maximizing the health status of the busbar trunking and minimizing maintenance costs, achieving intelligent maintenance decision-making.

[0050] (II) The edge computing collaborative system possesses edge computing capabilities, deploying edge computing nodes close to the busbar equipment. These edge computing nodes employ lightweight neural network models for real-time fault detection, performing local real-time processing and analysis of collected data to reduce data transmission latency and cloud computing pressure. When an anomaly is detected, detailed data is then transmitted to the cloud for in-depth analysis, achieving a collaborative processing model of "edge filtering, cloud decision-making." For example, deploying edge computing nodes in the busbar of a factory workshop can quickly detect instantaneous abnormal discharge faults in the busbar, uploading only key abnormal data to the cloud, reducing data transmission volume and improving fault handling efficiency.

[0051] Data representation and interpretation:

[0052] Module Traditional management mode index System index of the present application Promotion range Perception and data acquisition Data noise rate 10% Data noise rate 2% 80％ Digital twin modeling Model prediction error 15% Model prediction error 5% 66.7％ State monitoring and fault diagnosis Fault missed diagnosis rate 8% Fault missed diagnosis rate 2% 75％ Health assessment and decision Maintenance cost reduction rate 10% Maintenance cost reduction rate 30% 200％ Data management and interaction Data query response time 2s Data query response time 0.5s 75％

[0053] In the sensing and data acquisition module, moving average filtering and Kalman filtering algorithms effectively reduce data noise, providing more reliable data for subsequent analysis. Compared to traditional methods without precise filtering, the noise rate is reduced by 80%. In the digital twin modeling and mapping module, through model correction and fusion optimization, the model prediction error has decreased from the traditional 15% to 5%, an improvement of 66.7%, more accurately reflecting the true state of the busbar trunking. In the condition monitoring and fault diagnosis module, leveraging LSTM neural network time-series prediction, wavelet transform fault feature extraction, and support vector machine recognition, the fault misdiagnosis rate has decreased from 8% to 2%, an improvement of 75%, achieving more accurate fault early warning and diagnosis. In the health assessment and decision-making module, through a scientific indicator system, multi-objective optimization, and consideration of seasonal factors, the maintenance cost reduction rate has increased from 10% to 30%, an improvement of 200%, effectively saving on operation and maintenance costs. In the data management and interaction module, distributed storage, knowledge graphs, and natural language interaction have shortened the data query response time from 2 seconds to 0.5 seconds, an improvement of 75%, improving the work efficiency of operation and maintenance personnel. The synergistic effect of these modules comprehensively enhances the overall health management level of the busbar trunking throughout its entire lifecycle.

[0054] VII. System Integration and Collaboration Module

[0055] (I) Hardware Integration Architecture

[0056] The system hardware is built with a distributed, layered architecture, which works collaboratively from the bottom layer to the top layer to ensure the efficient flow of data collection, processing, and decision-making.

[0057] The bottom-layer sensor network precisely selects sensors based on the different monitoring needs of the busbar trunking: The temperature sensor uses a high-precision thermistor type, covering a wide temperature range of -50℃ to 200℃, adapting to temperature fluctuations during busbar trunking operation. At a 100Hz acquisition frequency, it can capture subtle temperature changes in key components such as joints and conductors caused by load variations. The current sensor adopts the Hall effect open-loop principle, with a range of 0–2000A, accurately sensing the current value of the busbar trunking under different load conditions, providing data support for power calculation and overload early warning. The voltage sensor relies on a resistor voltage divider circuit design to stably acquire voltage signals from 0 to 690V, ensuring accurate monitoring of electrical parameters. The partial discharge sensor uses the ultra-high frequency antenna principle, sensitive to discharge signals in the 300MHz–3GHz frequency band, capturing weak partial discharge pulses and promptly detecting potential insulation degradation. All sensors are connected via shielded cables, following the Modbus-RTU protocol, transmitting data to the edge computing nodes at a rate of 115200bps to ensure real-time data transmission.

[0058] The edge computing layer deploys industrial-grade embedded devices equipped with Cortex-A72 multi-core processors with a clock speed of up to 1.8GHz, 4GB of RAM, and 32GB of storage to meet the needs of multi-tasking parallel processing. The devices come pre-installed with a lightweight Linux operating system, integrating a Python runtime environment and the TensorFlow Lite framework, and deploying a lightweight MobileNetV3 neural network model. The model is trained on massive amounts of busbar fault data and can quickly identify abnormal data patterns at the edge, such as short-term overloads and partial discharge anomalies. Edge nodes interact with the cloud via the MQTT protocol, employing the TLS 1.3 encryption suite to establish a secure channel, ensuring tamper-proof and leak-proof data transmission. Network bandwidth remains stable at over 10Mbps, and transmission latency is controlled within 50ms, achieving efficient integration between rapid edge response and in-depth cloud analysis.

[0059] The cloud server cluster adopts a distributed architecture, consisting of application servers, database servers, and compute servers working together. The application servers are equipped with 8-core, 16-thread CPUs and 32GB of memory, hosting application-layer functions such as system web services and API interfaces, supporting concurrent access from multiple users. The database servers are equipped with 16-core, 32-thread CPUs, 64GB of memory, and 2TB of high-speed storage, employing a hybrid architecture of relational databases (such as Oracle) and time-series databases (such as InfluxDB). The relational database stores structured data such as device ledgers and maintenance work orders, while the time-series database efficiently stores operating parameters collected at a high frequency of 100Hz. The compute servers are equipped with 32-core, 64-thread CPUs, 128GB of memory, and NVIDIA A100 GPUs, providing computing power support for digital twin model calculations and deep learning model training for fault diagnosis. The servers are interconnected via 10 Gigabit Ethernet, and with the help of Kubernetes container orchestration technology, dynamic service scheduling and load balancing are achieved, ensuring stable 24 / 7 system operation and meeting the continuous monitoring needs throughout the entire lifecycle of the busbar trunking.

[0060] (II) Software Integration Architecture

[0061] The software adopts a microservices design, breaking down system functions into independent and collaborative service units. Each service interacts through a standardized RESTful API, and the data format is uniformly JSON, ensuring the system's scalability and flexibility.

[0062] The data acquisition service serves as the system's data entry point, adapting to multiple protocol access methods. For industrial bus protocols such as Modbus and Profibus, dedicated parsing modules have been developed to parse raw sensor data in real time. For automotive protocols such as CAN bus, communication drivers have been optimized to ensure data acquisition from load terminals such as vehicle charging stations. The service has a built-in data preprocessing engine that automatically identifies outliers (such as invalid data caused by sensor disconnections), uses algorithms such as linear interpolation and sliding window filtering to repair the data, and then normalizes it according to time series before pushing it to a message queue for downstream service consumption.

[0063] The digital twin service focuses on virtual modeling and dynamic simulation of busbar trunking. It utilizes a discrete event simulation framework built on the Python SimPy library and combines it with COMSOL Multiphysics finite element analysis software to construct a multiphysics coupled model. In the geometric modeling stage, the three-dimensional structure of the busbar trunking is accurately reproduced, including the dimensions and material parameters of components such as conductors, insulation layers, joints, and the outer shell. In the physics simulation, the electric field, temperature field, and flow field (considering air convection heat dissipation) are calculated simultaneously to simulate the operating state of the busbar trunking under different loads and ambient temperature and humidity conditions. Simulation data is mapped to the virtual model in real time and pushed to the front-end visualization interface via the WebSocket protocol, achieving dynamic synchronization between the physical entity and the digital twin. This provides an intuitive virtual mapping scenario for condition monitoring and fault diagnosis.

[0064] The condition monitoring service relies on the TensorFlow deep learning framework and deploys an LSTM neural network model. The model's input layer receives multi-dimensional time-series data (such as temperature, current, and partial discharge pulse sequences) preprocessed by the digital twin service. The hidden layer extracts time-series features through multiple LSTM units, and the output layer predicts the busbar trunking's operating status (normal, warning, fault) using the Softmax function. During the training phase, historical fault data (such as abnormal temperature rises caused by loose joints, and surges in partial discharges caused by insulation aging) are collected to construct a training set containing over 100,000 samples. After data augmentation (such as time shifting and noise injection), the model is iteratively trained to achieve a fault prediction accuracy of over 95%, identifying potential fault trends 1-3 hours in advance and allowing sufficient time for operation and maintenance decisions.

[0065] The health assessment service utilizes a fuzzy comprehensive evaluation algorithm based on the PyTorch framework, combined with the analytic hierarchy process (AHP) to construct an indicator system. First, it identifies 12 assessment indicators, including operating time, load factor, number of failures, and ambient temperature and humidity, by analyzing factors influencing busbar health. Second, it constructs a judgment matrix using expert interviews and the Delphi method, calculating the weight of each indicator; for example, for busbars operating under long-term high loads, the "load factor" indicator has a higher weight than "operating time." Finally, it employs a fuzzy membership function to map the measured values ​​of each indicator to the [0,1] interval, and after multi-layer fuzzy computation, outputs a health score and level (healthy, sub-healthy, or failure risk). Simultaneously, it integrates a cost-benefit model to calculate the remaining lifespan value of equipment, maintenance costs, and failure loss costs in real time, providing a quantitative basis for maintenance decisions.

[0066] The decision support service deeply integrates the Gurobi optimization engine to construct a multi-objective optimization model. With the goals of "minimizing maintenance costs, maximizing power supply reliability, and minimizing repair time," and considering constraints such as manpower, spare parts, and time, it optimizes the fault contingency plans output by the health assessment service. For example, when multiple busbars simultaneously issue warnings, the optimization engine automatically schedules maintenance resources, prioritizing high-risk, high-impact fault points, generating an optimal solution that includes repair sequence, spare parts list, and personnel assignment. This solution is then pushed to the maintenance terminal via the work order system, achieving intelligent maintenance decision-making.

[0067] The data management service is built upon the Elasticsearch distributed search engine, creating a full-text search and analysis system. An index is established for historical data, supporting multi-dimensional searches by device number, time range, data type, and other dimensions, with sub-second response times for queries involving tens of millions of data points. Simultaneously, a data mining module is integrated, using association rule mining (such as the Apriori algorithm) to analyze the correlation between faults and environmental and operational parameters, uncovering potential patterns such as "high summer temperature + high load → joint overheating fault," providing data support for model optimization and strategy adjustment.

[0068] (III) System Coordination Mechanism

[0069] The system constructs an event-driven collaborative architecture, relying on the RabbitMQ message queue to achieve module decoupling and asynchronous communication. The sensor data acquisition service monitors device status in real time. When events such as temperature exceeding a threshold or abnormal partial discharge pulse count are detected, an event message is immediately generated and pushed to a specific topic in the message queue (e.g., "busbar-event-alert"). The status monitoring service subscribes to this topic, automatically triggering the fault diagnosis process. It calls the digital twin model for comparative analysis, combines LSTM prediction results to determine the fault type and risk level, and then encapsulates the diagnosis result into a new event, pushing it to the "busbar-event-diagnosis" topic. The health assessment service and decision support service subscribe to this topic in turn, relaying the calculation of health scores and the generation of maintenance plans. Finally, the plan is pushed to the mobile terminal of maintenance personnel through the work order system, forming a closed-loop collaborative link of "data acquisition → anomaly identification → fault diagnosis → health assessment → decision execution," ensuring efficient flow from fault discovery to handling.

[0070] For multi-user collaborative scenarios, the system employs a Role-Based Access Control (RBAC) model to finely divide permissions. Administrators have global system configuration permissions, allowing them to modify sensor thresholds, adjust model parameters, and manage user accounts. Maintenance personnel are only authorized to view real-time monitoring data and provide work order execution feedback, receiving work orders via a mobile app, uploading on-site repair photos, and filling out processing results. Engineers focus on model training and strategy optimization, accessing historical data and adjusting algorithm parameters. Operations by different roles are logged, recording operation time, content, and results for easy auditing and backtracking, ensuring standardized system operation and data security, and achieving refined management in multi-role collaboration.

[0071] VIII. Safety and Reliability Assurance Module

[0072] (I) Data Security Mechanism

[0073] The system constructs a multi-level data security protection system, covering the entire process of data collection, transmission, storage, and access, to ensure the security of busbar operation data and management information.

[0074] In the data acquisition layer, AES-256 symmetric encryption is used between sensors and edge nodes. An encryption key is pre-negotiated, and the sensors encrypt the raw data before transmission. The edge nodes decrypt and parse the data. Key management follows a periodic update mechanism, automatically changing the key every 24 hours and synchronizing it to the sensors via a secure channel (such as SSH) to prevent data breaches due to key leaks. Simultaneously, edge nodes deploy data anonymization modules, using hash replacement and character masking to process sensitive information such as device location and serial numbers, preserving the data's business value while preventing the exposure of sensitive information.

[0075] At the data transmission layer, edge nodes and the cloud use MQTT protocol with TLS 1.3 encryption. TLS 1.3 simplifies the handshake process and strengthens encryption algorithms (such as AES-256-GCM and ChaCha20-Poly1305) to ensure the confidentiality and integrity of data transmission and resist risks such as man-in-the-middle attacks and data eavesdropping. For communication between services within the cloud, Service Mesh technology is used, with a Sidecar proxy implementing TLS encryption and authentication between microservices to ensure secure data exchange between services.

[0076] In the data storage layer, the database enables TDE (Transparent Data Encryption) to encrypt data stored on disk in real time. The encryption key is generated and managed by the Hardware Security Module (HSM), which is certified by national cryptographic standards to ensure key security. The backup strategy employs a "local backup + off-site disaster recovery" approach. Local backup performs a full backup daily at midnight, while the off-site disaster recovery center synchronizes incremental data in real time via an SSL / TLS encrypted channel. The backup data storage period is 3 years, meeting data traceability and compliance requirements.

[0077] The data access layer employs the OAuth 2.0 authorization framework. Upon user login, the authentication server generates a JWT (JSON Web Token) containing information such as user identity, permissions, and validity period (default 2 hours). The API interface validates the JWT's validity and supports automatic token refresh, ensuring the continuity of permissions for continuous user operations. Simultaneously, fine-grained access control policies are established. For example, ordinary maintenance personnel can only access busbar data within their assigned area, engineers can access all data but can only modify model parameters, and administrators have the highest privileges, implementing the principle of least privilege for data access.

[0078] (II) System Reliability Design

[0079] The system ensures reliability through hardware redundancy, software fault tolerance, and fault self-healing, ensuring uninterrupted monitoring of the busbar trunking throughout its entire lifecycle.

[0080] Hardware redundancy design covers sensors, edge nodes, and cloud servers. Sensors employ a "two-out-of-three" redundancy configuration, deploying three identical sensors at the same monitoring point. The data acquisition service compares the three data streams in real time, accepting data only when at least two streams match. If one stream is abnormal, fault diagnosis is automatically triggered to determine if it's a sensor failure, providing timely warnings and replacement. Edge nodes utilize a dual-machine hot standby architecture, with primary and backup nodes monitoring their status via heartbeat detection (sending a heartbeat packet every second). If the primary node fails (e.g., CPU overload, network interruption), the backup node automatically takes over the service within 100ms, ensuring continuity of data acquisition and edge computing. Cloud servers employ N+1 redundancy. The load balancer monitors server health status in real time (e.g., CPU utilization, memory usage, service response time). If a server fails, requests are automatically distributed to other healthy servers. The database uses RAID5 with a hot spare disk to ensure data storage redundancy. In the event of a hard drive failure, the hot spare disk automatically rebuilds the data, ensuring uninterrupted business operations.

[0081] Software fault tolerance mechanisms are implemented across all service modules. The data acquisition service has built-in retry logic for exceptions. When sensor communication times out or data parsing fails, retry is automatically triggered (up to 3 times), with retry intervals gradually increasing (1s, 3s, 5s), and failure logs are recorded for easy troubleshooting. The digital twin service employs a model degradation strategy. When the high-performance simulation model cannot run due to insufficient computing power, it automatically switches to a simplified empirical model to ensure the basic functions of the virtual mapping. The status monitoring service implements hot model standby. When the primary model is being trained and updated, the backup model takes over the prediction task to avoid service interruption. Each service manages external dependencies through a circuit breaker. When a dependent service (such as a database or message queue) times out, the circuit breaker is automatically broken, and degradation logic is executed (such as returning cached data or indicating that the service is temporarily unavailable). Once the dependency is restored, the circuit breaker is automatically closed, and normal service is restored.

[0082] The system's self-healing capability relies on intelligent diagnostics and automated operation and maintenance. It performs regular health checks (hourly) to monitor hardware resources (CPU, memory, disk), service processes, and network connectivity. Upon detecting anomalies (such as insufficient disk space or service process crashes), it automatically triggers repair processes: when disk space is insufficient, it cleans up historical logs and temporary files; when a service process crashes, it restarts the process and sends an alert; when the network is interrupted, it attempts to switch network links (e.g., from a wired network to a 4G backup network). Simultaneously, it uses a digital twin model to simulate the impact of fault scenarios on the system and proactively develops self-healing strategies. For example, if it predicts that a server is about to become overloaded, it automatically migrates some services to other servers, ensuring overall system reliability.

[0083] (III) Security Audit and Compliance

[0084] The system establishes a full-process security audit system, covering user operations, system events, and data access, to ensure that behavior is traceable and compliance is verifiable.

[0085] Audit logs are stored in a structured format, recording event timestamps, user IDs, operation types (e.g., data queries, parameter modifications, work order execution), operation content (e.g., querying data for a busbar in August 2024, modifying a temperature threshold to 85℃), and operation results (success / failure). Logs are stored in a separate audit database, physically isolated from the business database, ensuring audit data security. Logs are retained for 3 years and support retrieval by time range, user, and operation type, facilitating internal auditing and external regulatory verification.

[0086] The system strictly adheres to power industry safety standards and data protection regulations. Benchmarking against the IEC 62443 industrial control system security standard, its architecture is designed from dimensions including network security (partition isolation, access control), equipment security (identity authentication, firmware updates), and application security (vulnerability scanning, encrypted transmission). It complies with the NERCCIP critical infrastructure protection standard, strengthening access control, log auditing, and vulnerability management for critical power transmission equipment such as busbars. It follows data protection regulations such as GDPR and the Cybersecurity Law, clearly defining data ownership and usage rights, collecting data only after obtaining user authorization, and protecting the rights of data subjects.

[0087] Regular security assessments and tests are conducted, and vulnerability scans are performed quarterly (using tools such as Nessus and OpenVAS) to identify potential security vulnerabilities in the system (such as weak passwords, unauthorized access, and software vulnerabilities), generate vulnerability reports, and track and remediate them. Every year, a third-party security organization is hired to conduct penetration tests, simulating hacker attack scenarios (such as SQL injection, DDoS attacks, and privilege escalation) to test the system's defense capabilities. Based on the test results, security strategies are optimized to ensure that the system continues to meet security standards and regulatory requirements, thus building a solid security defense for the full life cycle health management of busbar trunking.

[0088] IX. Application Cases and Implementation Results

[0089] (I) Application Case of a Commercial Complex

[0090] In a super-large commercial complex encompassing a shopping mall, Grade A office buildings, and high-end hotels, the bus trunking system undertakes the power transmission task for all business formats. With a total length of over 8 kilometers, it connects more than 2,000 electrical devices, including air conditioning units, lighting systems, elevators, and commercial equipment. The power load has large peak-to-valley differences, and the equipment operates in a complex environment (such as high temperature and humidity in the shopping mall and high frequency load changes in the office buildings).

[0091] During the system deployment phase, monitoring units were divided by area. Sensors were deployed at key nodes of the busbar trunking on each floor (such as vertical shaft joints and floor distribution box inlets), totaling 300 temperature sensors, 200 current sensors, 150 voltage sensors, and 100 partial discharge sensors. Edge computing nodes were deployed on a floor-by-floor basis (10 in total) to achieve regionalized data processing. The digital twin model accurately recreates the three-dimensional orientation of the busbar trunking and the equipment connection relationships, and combined with the building BIM model, a virtual twin scenario of power transmission is constructed.

[0092] During operation and monitoring, the system effectively fulfilled its early warning and diagnostic functions. During the peak electricity consumption period in the summer of 2024, the load on the busbar trunking in the shopping center area continued to rise. The system detected that the temperature of a busbar trunking joint on a certain floor was rising at a rate of 0.5℃ / minute, reaching 75℃ (threshold 80℃) when an early warning was triggered. Following the work order instructions, maintenance personnel arrived on-site within 15 minutes and found that the bolts at the joint had loosened due to long-term vibration, resulting in increased contact resistance. After tightening, the temperature gradually decreased, preventing a fire accident caused by overheating of the joint. Furthermore, the system simulated the temperature field distribution under different loads using a digital twin model, guiding maintenance personnel to perform preventative maintenance on high-risk joints during off-peak periods, reducing the number of unplanned power outages per year from 5 to 1.

[0093] (II) Comparison of Implementation Results Data:

[0094] Index category Before implementation (traditional management) After implementation (system of the present application) Change range Fault occurrence rate 5.2 times / year 0.6 times / year Reduced by 88.5% Fault average repair time 5.8 hours / time 1.1 hours / time Shortened by 81.0% Unplanned downtime 32 hours / year 4 hours / year Reduced by 87.5% Operation and maintenance cost 950,000 yuan / year 500,000 yuan / year Reduced by 47.4% Energy utilization efficiency 91.2％ 95.5％ Increased by 4.7%

[0095] The significantly reduced failure rate stems from the system's real-time monitoring and early warning capabilities, eliminating potential faults in their nascent stages. For example, partial discharge monitoring can detect insulation degradation early, preventing it from developing into short-circuit faults. Shorter fault repair time is attributed to the precise fault location provided by the digital twin model and the intelligent dispatching of work orders. Maintenance personnel can quickly obtain the fault location and repair plan, reducing troubleshooting and decision-making time. Reduced unplanned downtime ensures continuous power supply to the commercial complex, improving user experience and economic efficiency. Lower maintenance costs are achieved through preventative maintenance reducing emergency repair expenses and optimized maintenance strategies lowering spare parts inventory and labor costs. Improved energy efficiency is attributed to the system's precise control over the busbar's operating status, avoiding energy waste caused by inefficient equipment operation or failures, thus achieving energy conservation and efficiency. This application's system, through end-to-end digital and intelligent management, provides strong support for the healthy operation of the busbar throughout its entire lifecycle, demonstrating significant economic and social benefits.

[0096] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A busbar trunking full lifecycle health management system based on digital twins, characterized in that, include: Sensing and Data Acquisition Module: Temperature sensors, current sensors, voltage sensors, and partial discharge sensors are deployed at busbar joints and conductors to collect operating parameters in real time; static information such as busbar model and production batch is obtained through RFID tags, while environmental temperature and humidity data are collected to construct a dimensional dataset; Digital Twin Modeling and Mapping Module: Based on the 3D geometric model and electrical parameters of the busbar trunking, a physical model is constructed using the finite element method; through a data mapping algorithm, real-time acquired data is associated with the physical model to form a dynamic digital twin; and a thermo-electric coupling equation is established. The simulated temperature distribution is given by K, where K is the thermal conductivity coefficient, T is the temperature, Q is the heat source per unit volume, ρ is the material density, and c is the specific heat capacity. Status monitoring and fault diagnosis module: Input real-time data into the digital twin model, compare the model output with the actual measured values ​​to determine the operating status; Set parameter thresholds and trigger an alert when limits are exceeded; Fault tree analysis is employed, combined with a fault feature database and an expert system, through formulas... Where P(T) is the probability of the top event occurring, and P(Xi) is the probability of the i-th bottom event occurring. The fault probability is calculated to achieve diagnosis. Health assessment and decision-making module: Construct a health assessment indicator system that includes factors such as uptime, load, and failure history; Using the Analytic Hierarchy Process (AHP), through formulas Determine the weights of the indicators, where wi is the weight of the i-th evaluation indicator, and a ij To determine the importance of the i-th indicator relative to the j-th indicator in the matrix, a fuzzy comprehensive evaluation method is used to quantify the health score, classify the health level, and formulate maintenance decisions based on the cost-benefit formula E = V - Cm - Cf, where E is the benefit, V is the remaining life value of the equipment, Cm is the maintenance cost, and Cf is the failure loss cost.

2. The bus trunking full lifecycle health management system based on digital twins according to claim 1, characterized in that, Also includes: Data Management and Interaction Module: Builds a database to store data throughout its entire lifecycle, uses data compression algorithms to improve storage efficiency, and leverages blockchain technology to ensure data security; It achieves 3D visualization interaction through WebGL, supports data display and parameter adjustment, and realizes data integration and sharing with power monitoring and production management based on the OPCUA protocol; Optimization module: Based on actual operation and maintenance data, optimize the parameters of the digital twin model using deep learning algorithms; iteratively train the fault diagnosis model through convolutional neural networks and adjust the decision-making strategy in combination with maintenance effect feedback; In the sensing and data acquisition module, a moving average filtering algorithm is used for the sensor data. Preprocessing is performed to remove data noise, where yn is the filtered data and x n-i The original data is given, and N is the size of the filtering window. Simultaneously, a Kalman filter algorithm is used to fuse and estimate the dynamic parameters, through the state equation x. k =Ax k-1 +Bu k-1 +w k-1 The observation equation zk=Hxk+vk improves data accuracy, where A is the state transition matrix, B is the control input matrix, H is the observation matrix, and w k-1 vk and vk represent process noise and observation noise, respectively.

3. The bus trunking full lifecycle health management system based on digital twins according to claim 1, characterized in that, The data management and interaction module adopts a distributed storage architecture, storing high-frequency real-time data in a time-series database and historical data in a distributed file system to improve data read and write performance; it also establishes a data traceability mechanism to record the entire process of data collection and processing, and uses blockchain technology to achieve data traceability.

4. The bus trunking full lifecycle health management system based on digital twins according to claim 1, characterized in that, In the health assessment and decision-making module, when the busbar health score is lower than a set threshold, a work order containing the cause of the fault, the repair plan, and the spare parts list is automatically generated and pushed to the relevant maintenance personnel's terminal. At the same time, a multi-objective optimization algorithm is adopted to generate the optimal repair plan by comprehensively considering the repair cost, repair time, and resource constraints while meeting the repair quality requirements.

5. The bus trunking full lifecycle health management system based on digital twins according to claim 1, characterized in that, In the digital twin modeling and mapping module, the boundary conditions and parameters of the digital twin model are periodically corrected based on the actual operating conditions and maintenance records of the busbar trunking. Model fusion technology is adopted to combine the finite element model with the machine learning model, and the model prediction results are fused and optimized through Gaussian process regression algorithm to improve the prediction accuracy of the model.

6. The bus trunking full lifecycle health management system based on digital twins according to claim 1, characterized in that, In the condition monitoring and fault diagnosis module, the LSTM neural network is used to predict the time series of busbar operating parameters, identify fault trends in advance and issue early warnings; a fault feature extraction algorithm is established, which converts the time domain signal into the frequency domain signal through wavelet transform to extract the fault feature vector, and combines the support vector machine algorithm to identify the fault type and improve the accuracy of fault diagnosis.

7. The bus trunking full lifecycle health management system based on digital twins according to claim 1, characterized in that, The optimization module establishes a maintenance strategy effectiveness evaluation model. By comparing the changes in busbar health indicators before and after maintenance, the effectiveness of different maintenance strategies is quantitatively evaluated, providing a basis for strategy optimization. A reinforcement learning algorithm is used to dynamically adjust the maintenance strategy with the goal of maximizing the health status of the busbar and minimizing maintenance costs.

8. The bus trunking full lifecycle health management system based on digital twins according to claim 2, characterized in that, The data management and interaction module supports users to query data and issue commands through a natural language processing interface. The system automatically parses and executes the relevant operations. A knowledge graph is developed to integrate structural knowledge, fault knowledge, and maintenance knowledge of the busbar, enabling knowledge visualization and retrieval.

9. The bus trunking full lifecycle health management system based on digital twins according to claim 2, characterized in that, The health assessment and decision-making module considers the impact of seasonal factors on busbar load, dynamically adjusts the weights of health assessment indicators and maintenance decision-making strategies, and establishes a seasonal load forecasting model to predict load change trends in different seasons through time series analysis algorithms.

10. The bus trunking full lifecycle health management system based on digital twins according to claim 1, characterized in that, The system incorporates edge computing capabilities, deploying edge computing nodes near the busbar equipment to achieve local real-time data processing and analysis. The edge computing nodes use neural network models for real-time fault detection, and when an anomaly is detected, detailed data is transmitted to the cloud for in-depth analysis, realizing a collaborative processing mode of edge filtering and cloud decision-making.

Images