Cloud edge collaboration method and system for AI intelligent Internet of Things equipment data processing

Abstract

The invention discloses a cloud edge cooperation method for AI intelligent Internet of Things equipment data processing, and relates to the technical field of data processing, and the method comprises the steps: S1, intelligent data collection, S2, edge side AI preprocessing, S3, edge-cloud end cooperation reasoning, S4, intelligent data transmission, S5, cloud end AI big data analysis, S6, real-time feedback and self-optimization, S7, adaptive resource scheduling, and S8, full-link visualization. Through AI-driven dynamic sampling and multi-modal data fusion, the efficiency and precision of data acquisition are optimized, redundancy or omission caused by fixed sampling is avoided, meanwhile, the transmission load is reduced, layered task dynamic unloading and intelligent transmission protocol optimization are achieved, the flexibility and stability of cloud edge collaboration are improved, and the cloud edge collaboration efficiency is improved. Manual intervention is reduced through a real-time feedback and self-optimization mechanism, the autonomy of the system is enhanced, and the interpretability and fault diagnosis capability of the system are remarkably improved through a visual panel and a causal reasoning model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a cloud-edge collaborative method and system for AI intelligent Internet of Things device data processing. BACKGROUND

[0002] Device data processing involves collecting data through sensors, performing real-time analysis and anomaly detection using edge computing, and then uploading key data to the cloud for deep analysis and model training. This process ensures data accuracy and timeliness, supports intelligent decision-making and optimization, and improves the efficiency and reliability of the Internet of Things system.

[0003] Existing device data processing cloud-edge collaborative methods have many defects: traditional data collection relies on fixed sampling frequency or manual configuration, resulting in low efficiency and insufficient accuracy; edge device computing power is limited, making it difficult to handle complex tasks or high-dimensional data; cloud-edge collaborative task allocation is rigid and cannot dynamically adapt to network or task changes; centralized cloud analysis has high latency, and raw data transmission can easily cause privacy risks; the system lacks self-optimization capabilities and relies on manual intervention to adjust parameters or models, with slow response; and traditional methods have poor result interpretability, making fault root cause analysis time-consuming and laborious. SUMMARY

[0004] The present application aims to provide a cloud-edge collaborative method and system for AI intelligent Internet of Things device data processing to solve the problems raised in the background.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solution: a cloud-edge collaborative method for AI intelligent Internet of Things device data processing, comprising the following steps:

[0006] S1: Intelligent data collection, dynamically adjusting sensor sampling frequency using AI and integrating multi-source data;

[0007] S2: Edge-side AI preprocessing, deploying lightweight AI models to detect anomalies and compress data in real time;

[0008] S3: Edge-cloud collaborative reasoning, edge processing low-latency tasks and cloud processing complex analysis, dynamically allocating computing resources;

[0009] S4: Intelligent data transmission, optimizing transmission protocols and paths using AI;

[0010] S5: Cloud AI big data analysis, aggregating multi-node data to train global models, predicting trends and generating digital twins;

[0011] S6: Real-time feedback and self-optimization, issuing optimization parameters to edge devices and dynamically updating models;

[0012] S7: Adaptive resource scheduling, AI-based intelligent allocation of edge and cloud resources, on-demand adjustment of computing tasks;

[0013] S8: Full-link visualization, locate fault causes, generate visual reports.

[0014] Preferably, the step S1 comprises the following steps:

[0015] S11: Use reinforcement learning AI algorithm to dynamically adjust sensor sampling frequency according to data change rate;

[0016] S12: Integrate visual, sound, and vibration multi-source data, and perform preliminary fusion processing through CNN and RNN multi-model;

[0017] S13: Real-time detection of data integrity and consistency by AI model, automatic elimination of abnormal and noise data;

[0018] S14: Trigger high-priority data collection in specific scenarios based on preset rules and AI prediction;

[0019] S15: Temporarily store high-frequency and redundant data on the edge side, upload to the cloud on demand, and reduce bandwidth occupancy.

[0020] Preferably, the step S2 comprises the following steps:

[0021] S21: Lightweight AI model deployment, deploy lightweight AI model on edge device, and run on resource-constrained device;

[0022] S22: Real-time anomaly detection, use time series model to detect data anomalies in real time, and trigger alarm and repair mechanism;

[0023] S23: Data dimensionality reduction and feature extraction, reduce dimensionality of high-dimensional data through PCA technology, and extract key features using CNN;

[0024] S24: Adaptive data compression, use dynamic compression algorithm to adjust compression rate according to network bandwidth and task demand;

[0025] S25: Local cache and priority management, cache preprocessed data according to priority, and upload critical data first.

[0026] Preferably, the step S3 comprises the following steps:

[0027] S31: According to task complexity, divide levels, handle low-latency tasks on edge side, and handle compute-intensive tasks on cloud side;

[0028] S32: Dynamic computing offloading, based on network status and edge device resource load, dynamically decide to keep part of model computation on edge execution;

[0029] S33: Model slicing and collaborative execution, large AI models are split into sub-modules, lightweight parts are run on the edge side, and the remaining calculations are completed in the cloud, with collaborative interaction through intermediate results;

[0030] S34: Resource-aware scheduling, real-time monitoring of edge and cloud resources, and optimization of task allocation through scheduling algorithms;

[0031] S35: Data synchronization and consistency guarantee, using incremental synchronization and differential update mechanism to ensure data consistency between edge and cloud.

[0032] Preferably, the step S4 comprises the following steps:

[0033] S41: Based on network status and data type, using reinforcement learning to dynamically select the optimal protocol;

[0034] S42: Using graph neural network to analyze network topology, predicting congested nodes and dynamically adjusting routing paths;

[0035] S43: Using LSTM to predict future network load and adjust transmission strategy in advance;

[0036] S44: Classifying data into urgent data and ordinary data through classification model, transmitting urgent data in real time and allowing to preempt bandwidth, and transmitting ordinary data after batch aggregation;

[0037] S45: The cloud predicts the type of data that the edge node needs to upload in the future by analyzing historical data, and issues cache instructions to the edge node in advance.

[0038] Preferably, the classification model uses a logistic regression model, and the specific formula is:

[0039] f(X) = σ(W T X + b)

[0040] Where X is the input feature vector, W is the weight vector, b is the bias term, σ(W T X + b) output range is between (0, 1), if σ(W T X + b) ≥ 0.5, it is classified as urgent data, otherwise as ordinary data.

[0041] Preferably, the step S5 comprises the following steps:

[0042] S51: Federated learning and data aggregation, aggregating data from multiple edge nodes, training a global model through federated learning, and ensuring data privacy;

[0043] S52: Time series prediction and digital twin modeling, using time series model to predict device life and energy consumption, and building digital twin to simulate running state;

[0044] S53: Multimodal analysis and anomaly detection, fusion of multi-source data, detection of anomalies and localization of root causes through deep learning;

[0045] S54: Model optimization and report generation, dynamic updating of cloud models, generation of interactive reports combined with visualization tools, and support for decision optimization.

[0046] Preferably, the step S6 comprises the following steps:

[0047] S61: After the global data of the cloud AI analysis is optimized, the optimized parameters are generated and sent to the edge device to adjust its operation strategy;

[0048] S62: Lightweight model incremental update, model incremental update is pushed through OTA technology, only the changed part is transmitted, reducing bandwidth occupation and speeding up the update speed;

[0049] S63: Dynamic performance monitoring and adjustment, real-time monitoring of edge device execution effect, dynamic adjustment of optimization strategy combined with feedback data, forming a closed loop control;

[0050] S64: Fault recovery and rollback mechanism, if the update and parameter adjustment cause abnormality, automatically trigger rollback to stable version.

[0051] Preferably, the step S7 comprises the following steps:

[0052] S71: Resource demand assessment, real-time assessment of computing resource demand of edge and cloud devices through AI algorithm, generation of scheduling demand combined with task priority;

[0053] S72: Dynamic task allocation, based on resource manager, dynamically allocating tasks to edge and cloud;

[0054] S73: Load balancing and elastic scaling, dynamically adjusting resource allocation according to device load state;

[0055] S74: Energy efficiency optimization, optimizing resource scheduling strategy combined with device energy consumption data.

[0056] The application also provides an AI intelligent Internet of Things device data processing cloud-edge collaborative system, comprising:

[0057] Intelligent data acquisition module, dynamically adjusting sensor sampling frequency by AI, and integrating multi-source data;

[0058] Edge-side AI preprocessing module, deploying lightweight AI model, real-time detection of anomalies and data compression;

[0059] Edge-cloud collaborative inference module, edge processing low-latency tasks, cloud processing complex analysis, and dynamic allocation of computing resources;

[0060] Intelligent data transmission module, which optimizes transmission protocols and paths through AI;

[0061] Cloud-based AI big data analysis module aggregates multi-node data to train global models, predict trends, and generate digital twins;

[0062] Real-time feedback and self-optimization module sends optimization parameters to edge devices and dynamically updates the model;

[0063] Adaptive resource scheduling module, which intelligently allocates edge and cloud resources based on AI and adjusts computing tasks on demand;

[0064] The full-link visualization module locates the cause of the fault and generates a visualization report.

[0065] The technical effects and advantages of the present invention are as follows:

[0066] (1) The present invention optimizes the efficiency and accuracy of data collection through AI-driven dynamic sampling and multimodal data fusion, avoiding redundancy or omissions caused by fixed sampling. The lightweight edge AI model and adaptive data compression technology alleviate the bottleneck of insufficient edge device resources and reduce the transmission load. The dynamic offloading of layered tasks and intelligent transmission protocol optimization improve the flexibility and stability of cloud-edge collaboration. Federated learning and digital twin technology take into account both data privacy and cloud analysis efficiency. Real-time feedback and self-optimization mechanisms reduce manual intervention and enhance system autonomy. The visualization panel and causal reasoning model significantly improve the system's interpretability and fault diagnosis capabilities.

[0067] (2) The present invention further reduces system operating costs by utilizing dynamic resource scheduling and energy efficiency optimization. The incremental update mechanism extends the equipment life cycle. The continuous evolution capability of the global AI model provides more accurate support for business decision-making. Full-link visualization not only facilitates operation and maintenance, but also provides a unified data view for cross-departmental collaboration, making it easy to use. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 This is a flow chart of the cloud-edge collaborative method for AI smart IoT device data processing in the present invention.

[0069] Figure 2 This is a flow chart of step S1 of the present invention.

[0070] Figure 3 This is a flow chart of step S2 of the present invention.

[0071] Figure 4 This is a flow chart of step S3 of the present invention.

[0072] Figure 5 This is a flow chart of step S4 of the present invention.

[0073] Figure 6 For the flowchart of step S5 of the present application.

[0074] Figure 7 For the flowchart of step S6 of the present application.

[0075] Figure 8 For the flowchart of step S7 of the present application. DETAILED DESCRIPTION

[0076] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0077] The present application provides an AI intelligent Internet of Things device data processing cloud-edge collaboration method as shown in Figures 1-8 The present application provides an AI intelligent Internet of Things device data processing cloud-edge collaboration method as shown in

[0078] S1: intelligent data acquisition, using AI to dynamically adjust the sampling frequency of the sensor, and integrating multi-source data, according to different application scenarios and data characteristics, AI can accurately determine the optimal sampling frequency, and effectively integrate the data of multiple data sources;

[0079] S2: edge-side AI preprocessing, deploying a lightweight AI model, real-time detection of abnormalities and compression of data, the lightweight model can run quickly on the edge side, timely discovery of abnormal conditions in the data, and efficient compression of data;

[0080] S3: edge-cloud collaborative reasoning, edge processing low-latency tasks, cloud processing complex analysis, dynamic allocation of computing resources, edge and cloud cooperate with each other, flexibly allocate resources according to task characteristics, and ensure efficient and stable operation of the system;

[0081] S4: intelligent data transmission, AI optimizes transmission protocols and paths, AI can analyze network conditions, select the best transmission protocol and path, reduce data transmission delay, and ensure efficient and accurate transmission of data;

[0082] S5: cloud AI big data analysis, aggregating multi-node data to train global models, predicting trends and generating digital twins, deeply mining multi-node data to train more accurate global models, and providing strong support for business decision-making;

[0083] S6: real-time feedback and self-optimization, issuing optimization parameters to edge devices, dynamically updating models, and issuing optimization parameters in a timely manner according to analysis results, so that the edge device model can be updated in real time, and the overall system performance is improved;

[0084] S7: Adaptive resource scheduling, AI-based intelligent allocation of edge and cloud resources, on-demand adjustment of computing tasks, AI dynamically allocates resources based on system load and task demand, ensuring full utilization of edge and cloud resources;

[0085] S8: Full-link visualization, locate fault causes, generate visual reports, the entire data processing flow is fully presented on the display, not only can show the flow of data from sensors through edge side to cloud and then feedback to edge devices, distinguish data types and flow sizes by different colors and line thicknesses, but also can present the running status and resource usage of each node such as sensors, edge devices, cloud servers, etc., and can also show the business process and execution time and results of each step in the form of a flowchart. In terms of fault cause positioning, through data anomaly analysis, real-time monitoring of collected data, combined with historical data and rule library, it is determined whether the anomaly is caused by sensor failure, network problem or algorithm error. With the help of device fault diagnosis, according to the device operation log and performance index, combined with the fault mode library to determine whether it is a hardware or software failure; Use link tracking to add a unique identifier to the data to trace the data source and processing process, accurately locate the fault position. When generating a visual report, templates can be customized according to different user needs, the system automatically extracts relevant data from the database to fill in the template, and generates various charts such as column charts and line charts to visually display the data. Finally, the report is exported in PDF, Excel, etc.

[0086] Step S1 includes the following steps:

[0087] S11: Use reinforcement learning AI algorithm to dynamically adjust sensor sampling frequency according to data change rate, which can accurately match data changes, avoid over-sampling and waste resources, improve sampling efficiency, and ensure data effectiveness;

[0088] S12: Integrate visual, sound, and vibration multi-source data, and perform preliminary fusion processing through CNN and RNN multi-model to fully leverage the advantages of different models, more comprehensively capture data features, and improve the accuracy and reliability of subsequent data analysis;

[0089] S13: Real-time detection of data integrity and consistency by AI model, automatic elimination of abnormal and noise data, which can ensure the quality of data entering the system, reduce the interference of error data on the analysis results, and improve the stability of the overall system;

[0090] S14: Based on preset rules and AI prediction, trigger high-priority data collection in specific scenarios, which can timely obtain key data and meet the needs of specific scenarios, providing more targeted information support for business decision-making;

[0091] S15: Temporarily store high-frequency and redundant data on the edge side and upload to the cloud on demand to reduce bandwidth occupation, effectively reduce network transmission pressure, avoid bandwidth waste, and ensure timely and accurate uploading of important data to the cloud.

[0092] Step S2 includes the following steps:

[0093] S21: Lightweight AI model deployment, deploy lightweight AI models on edge devices and run on resource-constrained devices, fully utilize edge device resources, reduce dependence on the cloud, and realize local fast processing of data;

[0094] S22: Real-time anomaly detection, use time series models to detect data anomalies in real time and trigger alarm and repair mechanisms, can timely discover abnormal conditions in data, quickly respond and solve problems, and ensure stable operation of the system;

[0095] S23: Data dimensionality reduction and feature extraction, reduce the dimensionality of high-dimensional data through PCA technology, and extract key features using CNN, which can reduce data processing complexity, focus on key information, and improve the efficiency and accuracy of subsequent analysis;

[0096] S24: Adaptive data compression, use dynamic compression algorithm, adjust compression rate according to network bandwidth and task demand, can flexibly cope with different network conditions, save transmission bandwidth while ensuring data quality;

[0097] S25: Local cache and priority management, cache preprocessed data according to priority, upload key data first, can ensure timely uploading of important data, avoid loss of key information due to network problems, and improve the effectiveness of data transmission.

[0098] Step S3 includes the following steps:

[0099] S31: According to the task complexity, divide the level, edge side handles low delay task, cloud end handles calculation intensive task, can give full play to the advantages of edge and cloud, reduce delay, improve calculation efficiency, meet the performance requirements of different tasks;

[0100] S32: Dynamic calculation offloading, based on network status and edge device resource load, dynamically decide to retain part of the model calculation in edge execution, can flexibly adapt to network and device state, avoid resource waste, and improve the overall performance of the system;

[0101] S33: Model slicing and collaborative execution, split large AI models into sub-modules, run lightweight parts on the edge side, and complete the remaining calculation on the cloud, realize collaboration through intermediate results interaction, can reduce the burden of edge devices, and at the same time, utilize the powerful computing power of the cloud to speed up model operation;

[0102] S34: Resource-aware scheduling, real-time monitoring of edge and cloud resources, optimizing task allocation through scheduling algorithms, reasonable allocation of tasks according to resource usage, improving resource utilization, ensuring stable operation of the system;

[0103] S35: Data synchronization and consistency guarantee, using incremental synchronization and differential update mechanism to ensure data consistency between edge and cloud, reducing data transmission volume, improving synchronization efficiency, ensuring data accuracy and consistency, providing reliable support for analysis and decision-making.

[0104] Step S4 includes the following steps:

[0105] S41: Based on network status and data type, dynamically select the optimal protocol using reinforcement learning, accurately match the protocol according to real-time conditions, improve the adaptability of data transmission, and ensure efficient and stable transmission of different types of data;

[0106] S42: Use graph neural network to analyze network topology, predict congestion nodes and dynamically adjust routing path, detect network congestion risk in advance, bypass congestion nodes in time, ensure smooth data transmission path, and reduce transmission delay;

[0107] S43: Use LSTM to predict future network load, adjust transmission strategy in advance, enable the system to actively adapt to network changes, reasonably plan data transmission, and avoid data transmission interruption or delay due to network overload;

[0108] S44: Classify data into urgent data and ordinary data through classification model, transmit urgent data in real time, allow to preempt bandwidth, transmit ordinary data in batches after aggregation, ensure the timeliness of urgent data, and improve the transmission efficiency of ordinary data, optimize bandwidth usage;

[0109] S45: The cloud analyzes historical data to predict the type of data that the edge node needs to upload in the future, and issues cache instructions to the edge node in advance, so that the edge node can prepare in advance, reduce data waiting time, and improve overall transmission efficiency.

[0110] The classification model uses a logistic regression model, and the specific formula is:

[0111] f(X) = σ(W T X + b)

[0112] Where X is the input feature vector, W is the weight vector, b is the bias term, σ(W T X + b) outputs range between (0, 1), if σ(W T X + b) ≥ 0.5, it is classified as urgent data, otherwise as ordinary data.

[0113] The calculation complexity of the logistic regression model is low, and the training and prediction speed is fast. In the scene that needs to process a large amount of data in real time, the data classification task can be quickly completed by importing the data into the above model, ensuring that emergency data can be identified in time and transmitted preferentially. For resource-constrained devices in edge computing environments, the normal operation of the system can be ensured under limited computing resources.

[0114] Step S5 includes the following steps:

[0115] S51: Federated learning and data aggregation, aggregate data of multiple edge nodes, train global model through federated learning, ensure data privacy, integrate data to train strong global model under the premise of protecting the privacy of each node data, and improve overall performance;

[0116] S52: Time series prediction and digital twin modeling, use time series model to predict device life and energy consumption, and build digital twin simulation running state, which can master device condition in advance and maintain in time. Simulation operation can find potential problems and reduce operation and maintenance cost;

[0117] S53: Multi-modal analysis and anomaly detection, fuse multi-source data, detect anomalies and locate root causes through deep learning, and more accurately find anomalies and quickly locate causes to ensure stable and efficient continuous operation of the system;

[0118] S54: Model optimization and report generation, dynamically update cloud model, generate interactive report combined with visualization tool, support decision optimization, real-time update model to improve performance, and interactive report intuitive display provides strong basis for decision making.

[0119] Step S6 includes the following steps:

[0120] S61: Parameter optimization and delivery After cloud AI analyzes global data, generate optimized parameters and deliver to edge devices to adjust their running strategies, which can make edge device running strategy more optimal and improve overall system performance and efficiency;

[0121] S62: Lightweight model incremental update, push model incremental update through OTA technology, only transmit the changed part to reduce bandwidth occupation and speed up the update speed;

[0122] S63: Dynamic performance monitoring and adjustment, real-time monitor edge device execution effect, dynamically adjust optimization strategy combined with feedback data, form closed-loop control, ensure system always in optimal running state, improve stability;

[0123] S64: Fault recovery and rollback mechanism, if update and parameter adjustment cause abnormality, automatically trigger rollback to stable version, which can ensure continuous and reliable operation of the system and reduce potential risks caused by update.

[0124] Step S7 includes the following steps:

[0125] S71: Resource demand assessment, real-time assessment of edge and cloud device computing resource demand through AI algorithm, combined with task priority to generate scheduling demand, can accurately grasp device resource demand, provide scientific basis for subsequent task allocation, and improve scheduling rationality;

[0126] S72: Dynamic task allocation, based on resource manager, dynamically allocate tasks to edge and cloud, can fully utilize the advantages of different devices, improve task processing efficiency, and avoid resource idling or overload;

[0127] S73: Load balancing and elastic scaling, dynamically adjust resource allocation according to device load status, can keep the system stable running, meet the task demand at the same time, and flexibly respond to business volume changes, improve overall performance;

[0128] S74: Energy efficiency optimization, combined with device energy consumption data, optimize resource scheduling strategy, can reduce energy consumption on the basis of ensuring system performance, realize the win-win of resource utilization and energy saving, and reduce operating costs.

[0129] An AI intelligent Internet of Things device data processing cloud-edge collaborative system, comprising:

[0130] Intelligent data acquisition module, dynamically adjust sensor sampling frequency with AI, and integrate multi-source data;

[0131] Edge-side AI preprocessing module, deploy lightweight AI model, real-time detect abnormal and compress data;

[0132] Edge-cloud collaborative inference module, edge handles low-latency tasks, cloud handles complex analysis, dynamically allocates computing resources;

[0133] Intelligent data transmission module, optimize transmission protocol and path through AI;

[0134] Cloud AI big data analysis module, aggregate multi-node data to train global model, predict trends and generate digital twin;

[0135] Real-time feedback and self-optimization module, issue optimization parameters to edge devices, dynamically update models;

[0136] Adaptive resource scheduling module, based on AI intelligent allocation of edge and cloud resources, adjust computing tasks on demand;

[0137] Full-link visualization module, locate fault causes, generate visual reports.

[0138] The intelligent data acquisition module uses AI to dynamically adjust the sampling frequency and integrate multi-source data, can flexibly collect and obtain comprehensive information, the edge side AI preprocessing module deploys a lightweight model to detect abnormalities and compress data in real time, can discover problems in time and reduce transmission cost. The edge-cloud collaborative reasoning module makes the edge handle low-latency tasks and the cloud handle complex analysis, dynamically allocates resources, ensures real-time response and improves resource utilization. The intelligent data transmission module uses AI to optimize transmission protocols and paths to ensure stable data transmission. The cloud AI big data analysis module aggregates data to train global models, predicts trends and generates digital twins, improves model accuracy and generalization ability, and the real-time feedback and self-optimization module issues parameters and updates models to continuously optimize the system. The adaptive resource scheduling module intelligently allocates resources and adjusts tasks to achieve optimal configuration. The full-link visualization module locates faults and generates reports to facilitate maintenance personnel to solve problems and improve system maintainability and reliability.

[0139] Finally, it should be noted that: the above only for the preferred embodiments of the present application, and not for limiting the present application, although the foregoing detailed description of the present application is made with reference to the foregoing embodiments, for those skilled in the art, it still can be modified to the technical solutions recorded in the foregoing embodiments, or equivalent replacement of some technical features, any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application, should be included in the protection scope of the present application.

Claims

1. A cloud-edge collaborative method for AI intelligent IoT device data processing, characterized in that: The following steps are involved: S1: Intelligent data acquisition, using AI to dynamically adjust sensor sampling frequency and integrate multi-source data; S2: Edge-side AI preprocessing, deploying lightweight AI models, detecting anomalies in real time, and compressing data; S3: Edge-cloud collaborative reasoning, processing low-latency tasks at the edge and complex analysis in the cloud, dynamically allocating computing resources; S4: Intelligent data transmission, using AI to optimize transmission protocols and paths; S5: Cloud-based AI big data analysis, aggregating multi-node data to train global models, predict trends, and generate digital twins; S6: Real-time feedback and self-optimization, sending optimization parameters to edge devices and dynamically updating the model; S7: Adaptive resource scheduling, using AI to intelligently allocate edge and cloud resources and adjust computing tasks on demand; S8: Full-link visualization, fault cause location, and visualization report generation.

2. The cloud-edge collaborative method for AI intelligent IoT device data processing according to claim 1 is characterized in that: The step S1 comprises the following steps: S11: Uses reinforcement learning AI algorithms to dynamically adjust sensor sampling frequency based on data change rates; S12: Integrate visual, sound, and vibration multi-source data and perform preliminary fusion processing through CNN and RNN multi-models; S13: Use AI models to detect the integrity and consistency of collected data in real time, and automatically remove abnormal and noisy data; S14: Triggering high-priority data collection for specific scenarios based on preset rules and AI predictions; S15: Temporarily stores high-frequency and redundant data on the edge and uploads it to the cloud on demand to reduce bandwidth usage.

3. The cloud-edge collaborative method for AI intelligent IoT device data processing according to claim 1 is characterized in that: The step S2 comprises the following steps: S21: Lightweight AI model deployment, deploying lightweight AI models on edge devices and running them on resource-constrained devices; S22: Real-time anomaly detection, using time series models to detect data anomalies in real time and trigger alarms and repair mechanisms; S23: Data dimensionality reduction and feature extraction, using PCA technology to reduce the dimensionality of high-dimensional data and using CNN to extract key features; S24: Adaptive data compression, using a dynamic compression algorithm to adjust the compression rate based on network bandwidth and task requirements; S25: Local caching and priority management, caching the pre-processed data according to priority, and uploading key data first.

4. The cloud-edge collaborative method for AI intelligent IoT device data processing according to claim 1 is characterized in that: The step S3 comprises the following steps: S31: Divide the tasks into different levels according to their complexity, with low-latency tasks handled at the edge and computationally intensive tasks handled in the cloud. S32: Dynamic computing offloading, based on network status and edge device resource load, dynamically decides to retain part of the model calculation for execution at the edge; S33: Model Sharding and Collaborative Execution: This splits large AI models into submodules, with lightweight components running on the edge and the remaining computations performed in the cloud. Collaboration is achieved through the interaction of intermediate results. S34: Resource-aware scheduling, real-time monitoring of edge and cloud resources, and optimization of task allocation through scheduling algorithms; S35: Data synchronization and consistency assurance, using incremental synchronization and differential update mechanisms to ensure data consistency between the edge and the cloud.

5. The cloud-edge collaborative method for AI intelligent IoT device data processing according to claim 1 is characterized in that: The step S4 comprises the following steps: S41: Dynamically select the optimal protocol based on network status and data type using reinforcement learning; S42: Use graph neural networks to analyze network topology, predict congested nodes, and dynamically adjust routing paths; S43: Use LSTM to predict future network load and adjust transmission strategies in advance; S44: Data is divided into urgent data and ordinary data through a classification model. Urgent data is transmitted in real time, allowing bandwidth to be preempted. Ordinary data is aggregated and transmitted in batches. S45: The cloud analyzes historical data to predict the type of data that the edge node needs to upload in the future, and sends cache instructions to the edge node in advance.

6. The cloud-edge collaborative method for AI intelligent IoT device data processing according to claim 5, characterized in that: The classification model adopts a logistic regression model, and the specific formula is: f(X)=σ(W T X+b) Among them, X is the input feature vector, W is the weight vector, b is the bias term, σ(W T X+b) output range is between (0,1), if σ(W T If X+b)≥0.5, it is classified as urgent data, otherwise it is classified as normal data.

7. The cloud-edge collaborative method for AI intelligent IoT device data processing according to claim 1, characterized in that: The step S5 comprises the following steps: S51: Federated Learning and Data Aggregation: Aggregating data from multiple edge nodes, training a global model through federated learning, and ensuring data privacy. S52: Time series prediction and digital twin modeling, using time series models to predict equipment life and energy consumption, and building digital twins to simulate operating status; S53: Multimodal analysis and anomaly detection, integrating multi-source data, detecting anomalies and locating root causes through deep learning; S54: Model optimization and report generation, dynamically update cloud models, combine visualization tools to generate interactive reports, and support decision optimization.

8. The cloud-edge collaborative method for AI intelligent IoT device data processing according to claim 1, characterized in that: The step S6 comprises the following steps: S61: Parameter Optimization and Distribution: After analyzing global data, the cloud AI generates optimized parameters and distributes them to edge devices to adjust their operating strategies. S62: Lightweight incremental model updates, pushing incremental model updates via OTA technology, transmitting only the changed parts, reducing bandwidth usage and speeding up updates; S63: Dynamic performance monitoring and adjustment: real-time monitoring of edge device execution results, dynamic adjustment of optimization strategies based on feedback data, and closed-loop control. S64: Fault recovery and rollback mechanism. If updates and parameter adjustments cause exceptions, rollback to the stable version will be automatically triggered.

9. The cloud-edge collaborative method for AI intelligent IoT device data processing according to claim 1, characterized in that: The step S7 comprises the following steps: S71: Resource demand assessment, which uses AI algorithms to evaluate the computing resource requirements of edge and cloud devices in real time and generates scheduling requirements based on task priorities; S72: Dynamic task allocation, based on the resource manager, dynamically allocates tasks to the edge and the cloud; S73: Load balancing and elastic scaling, dynamically adjusting resource allocation based on device load status; S74: Energy efficiency optimization, combining equipment energy consumption data to optimize resource scheduling strategies.

10. A cloud-edge collaborative system for AI intelligent IoT device data processing, applied to a cloud-edge collaborative method for AI intelligent IoT device data processing according to any one of claims 1 to 9, characterized in that: include: Intelligent data acquisition module, using AI to dynamically adjust sensor sampling frequency and integrate multi-source data; Edge-side AI pre-processing module deploys lightweight AI models to detect anomalies and compress data in real time; Edge-cloud collaborative reasoning module, processing low-latency tasks at the edge and complex analysis in the cloud, dynamically allocating computing resources; Intelligent data transmission module, which optimizes transmission protocols and paths through AI; Cloud-based AI big data analysis module aggregates multi-node data to train global models, predict trends, and generate digital twins; Real-time feedback and self-optimization module sends optimization parameters to edge devices and dynamically updates the model; Adaptive resource scheduling module, which intelligently allocates edge and cloud resources based on AI and adjusts computing tasks on demand; The full-link visualization module locates the cause of the fault and generates a visualization report.

Images