WANG Aoyu. Securing Data in AI Supply Chain[J]. Technology of IoT&AI, 2026, 58(3): 1-5.
As indicated in this report, the advancement of Artificial Intelligence (AI) relies on a complex supply chain ecosystem, in which data serves as a core component. When policymakers concentrate excessively on any single data element, it tends to result in unbalanced governance. To address risks comprehensively, the report proposes three approaches to security analysis: identifying and managing risks based on data states—including storage, transmission, and processing—threat actors, and supply chain governance. While many AI-related data risks can be mitigated using existing security standards, unique threats specific to AI, such as training data poisoning and neural backdoors, demand new defensive measures. Meanwhile, the report encourages the adoption of a “know your supplier” principle to strengthen the overall security of the AI supply chain.
MO Qinglin. Exploration of Artificial Intelligence-Related Technologies for Internet of Things Applications[J]. Technology of IoT&AI, 2026, 58(3): 6-10.
With the development of Internet of Things technology and the maturity of artificial intelligence algorithms, the deep integration of the two has ushered in a new era of intelligent applications. Internet of Things provides artificial intelligence with massive real-time data, while artificial intelligence endows Internet of Things systems with intelligent perception, autonomous decision-making, and optimized control capabilities. This paper systematically analyzes artificial intelligence-related technologies for Internet of Things applications, focuses on exploring the application principles and implementation methods of core technologies such as machine learning, deep learning, edge intelligence, and Natural Language Processing ( NLP) in Internet of Things, and reviews typical application scenarios such as smart cities, smart manufacturing, smart agriculture, and smart homes. The research indicates that the integration of artificial intelligence and Internet of Things will continue to deepen, and technologies such as edge intelligence, federated learning, and lightweight models will become key development directions, providing crucial technical support for building an intelligent society where everything is intelligently connected.
WANG Yao. IoT Intrusion Detection Method Based on Res-BiGRU Incorporating Attention Mechanism[J]. Technology of IoT&AI, 2026, 58(3): 11-15.
To address the challenges of class imbalance, insufficient temporal dependency modeling, and limited feature representation in Internet of Things (IoT) intrusion detection, this paper proposes a hybrid detection framework integrating adaptive Synthetic Minority Over-sampling Technique-One-Sided Selection (SMOTE-OSS) sampling, Residual-Bidirectional Gated Recurrent Unit (Res-BiGRU), and a Convolutional Block Attention Module (CBAM). The adaptive sampling strategy dynamically balances minority and majority classes, while the Res-BiGRU captures bidirectional temporal dependencies in network traffic. Furthermore, the channel-spatial attention mechanism enhances discriminative feature learning by emphasizing critical features and time steps. Experimental results on the UNSW-NB15 dataset demonstrate that the proposed method outperforms baseline models, particularly in detecting minority attack classes, confirming its effectiveness in complex IoT environments.
GUO Yu. High-Precision Measurement Technology for Optical Fiber Current Transformers in Smart Substations[J]. Technology of IoT&AI, 2026, 58(3): 16-20.
Smart substations have high requirements for the accuracy, bandwidth, and anti-disturbance performance of current measurement. Centering on the optical rotation measurement characteristics of Fiber Optic Current Transformer (FOCT), a measurement link composed of optical path stability control layout, polarization disturbance compensation, digital demodulation structure and error model fusion calibration is constructed. Comparative tests were carried out in combination with actual engineering scenarios, and quantitative analyses were conducted on five types of indicators. The results showed that the measurement sequences formed a concentrated distribution under dynamic conditions, providing a reference path for the engineering implementation of optical current sensing devices.
LIU Shaowei. Optimization Methods for Intelligent Sensor Networks in Chronic Disease Management[J]. Technology of IoT&AI, 2026, 58(3): 21-24.
Addressing the issues of high energy consumption, unstable data transmission, and poor network reliability in intelligent sensor networks for chronic disease management, a comprehensive optimization method is proposed. By constructing an overall framework for optimization strategies, system optimization is carried out from three dimensions: network topology, data transmission processing, and energy resource management. A multi-module collaborative optimization mechanism is established, and simulation experiments are conducted to verify the optimization effect. The results show that after using the optimization method, network energy consumption and network latency are reduced, and the success rate of data transmission is improved, proving that the optimization method can effectively enhance the overall performance of the chronic disease monitoring system.
ZHOU Fangliang, CHEN Ming, LYU Penghui. Distributed Collaborative Optimization of Cooling and Heating Supply Networks Based on Internet of Things[J]. Technology of IoT&AI, 2026, 58(3): 25-28.
Addressing the issues of complex energy coupling, multi-objective conflicts, and dynamic load fluctuations in the operation of combined cooling and heating supply networks, a distributed collaborative optimization method based on Internet of Things is proposed. This method collects and processes real-time data from the network through sensors, constructs a multi-objective optimization model using an improved consensus algorithm, and dynamically adapts the load by incorporating a rolling horizon strategy. Experiments show that compared to traditional centralized optimization, the designed method significantly reduces daily average energy consumption costs, narrows temperature fluctuations at the user side, shortens single calculation time, compresses communication data volume, and effectively improves system economy, robustness, and computational timeliness.
CHEN Xilin, LIANG He. Research on an Intelligent Detection Framework for Campus Network Security Vulnerabilities Based on Big Data[J]. Technology of IoT&AI, 2026, 58(3): 29-32.
In view of the characteristics of large-scale campus networks, heterogeneous terminals and complex types of vulnerabilities, a big data-driven intelligent detection framework for campus network security vulnerabilities is proposed. This framework achieves vulnerability feature modeling through multi-source security data collection and preprocessing, combined with big data analysis, and completes intelligent vulnerability detection and discrimination under a unified architecture. Experimental results show that this framework can effectively improve the accuracy and stability of campus network vulnerability detection, reduce false alarm risks, and has good engineering application value.
WANG Jianping. Research on Design and Implementation of a Cybersecurity Operations Platform Based on Internet of Things Technology[J]. Technology of IoT&AI, 2026, 58(3): 33-36.
In response to the increasing complexity of cybersecurity threats within Internet of Things environments, this study investigates the architecture and key technologies of a security operations platform. It outlines the design approach for multi-source security data fusion, event correlation analysis, and dynamic policy scheduling, while detailing the platform’s layered architecture, module functions, and implementation process. Performance testing demonstrates the platform’s significant advantages in data processing rate, response latency, and resource utilization efficiency, effectively supporting security operations in Internet of Things environments.
LI Ju, MENG Yan, LU Xia, YIN Yuedong, LIU Hongxiao. Research on Unattended Truck Scale Based on Internet of Things Devices#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(3): 37-40.
Traditional truck scales rely on manual operation, which is inefficient, costly and prone to errors. In response, it is proposed to utilize Internet of Things technology to build an unattended system for truck scales, integrating sensors, edge computing, 5G, and artificial intelligence algorithms. This will enable vehicles to be automatically identified and intelligently weighed mass, ultimately achieving early warning of abnormal situations. Experimental results show that the proposed system reduces the weighing mass time for a single truck from 3～5 min in the traditional mode to 45～60 s, improving the weighing accuracy and reducing labor costs. It provides an efficient and high-quality solution for industrial logistics scenarios.
NIU Jian, ZHAO Fuqing, LI Fen, WANG Kai. Application of Internet of Things-Information Technology Integration System in Smart Museums[J]. Technology of IoT&AI, 2026, 58(3): 41-47.
With the rapid development of the Internet of Things, artificial intelligence, and big data technology, traditional museums are undergoing a smart transformation. The Internet technology and intelligent Internet of Things is the core driving force behind this transformation. This paper explores the application framework and practical path of the Internet of Things-information technology integration system in the construction of smart museums. Firstly, it constructs a physical information integration technology architecture consisting of the intelligent perception layer, network transmission layer, data fusion layer, and smart application layer. Secondly, it deeply analyzes the specific application modes and effectiveness of this technology in three core business scenarios: intelligent protection of cultural relics, intelligent management of venues, and intelligent services for visitors, using the case of Yujiashan Archaeological Museum. Finally, it summarizes the current challenges faced, such as data fusion, standard unification, and cost-effectiveness, and looks forward to the future trends of technology integration and ecosystem construction, providing theoretical reference and practical guidance for the deepening construction of smart museums.
SHANG Jiatong. Application of a Fast FCM Algorithm Based on Spectral Clustering to Scanning Electron Microscope Image Segmentation[J]. Technology of IoT&AI, 2026, 58(3): 48-53.
Materials are the foundation of human survival and development. Scanning electron microscopy can display the micro-morphology of materials, facilitating the study of their organizational structure and properties. Currently, the information extraction of Scanning electron microscopy images mainly relies on manual annotation, which is both time-consuming and has poor accuracy. To improve the efficiency and accuracy of Scanning electron microscopy image segmentation, a fast Fuzzy C-Means (FCM) algorithm based on spectral clustering is proposed. This algorithm adaptively determines the number of clusters through spectral clustering and utilizes histogram information instead of traditional pixel counts, effectively reducing data redundancy and computational complexity. Experimental results show that the proposed algorithm exhibits high accuracy and efficiency in high-resolution image segmentation, offering significant advantages over traditional FCM algorithms. This method provides an innovative solution for high-resolution image segmentation and holds high practical value.
YU Xiaoqiu, SHEN Sen, ZHANG Haifeng, MA Xu, GE Qiyu. Optimization Method for Distribution Network Protection Setting Based on Intelligent Relays[J]. Technology of IoT&AI, 2026, 58(3): 54-58.
To address the shortcomings of traditional setting methods, an optimal setting method for distribution network protection based on intelligent relays is proposed, utilizing an improved genetic algorithm to achieve adaptive optimization of setting parameters. The intelligent relay collects operational data in real time via the IEC61850 protocol, while the master station performs online optimization calculations and remotely issues the parameters. Experimental results demonstrate the significant advantages of the optimization method in terms of protection action speed, selective coordination, and computational efficiency, particularly showcasing strong adaptability in scenarios involving distributed power sources, thereby laying the foundation for the deep application of distribution automation systems.
GUO Defeng, XIN Youqiang, DONG Fangjie, CHEN Jinming. Research on Co-Occurrence Patterns of Soundscape Communities Based on Unmanned Aerial Vehicle and Graph Neural Networks[J]. Technology of IoT&AI, 2026, 58(3): 59-64.
Addressing the monitoring challenges of complex soundscapes and multi-species co-occurrence patterns, existing fixed-point collection methods face issues of insufficient spatial coverage and difficulties in sound source separation. To this end, this paper proposes an analytical method that integrates an unmanned aerial vehicle-based mobile collection system with a Cross Time-Frequency Relationship Graph Neural Network (CTFR-GNN). The unmanned aerial vehicle is equipped with a high-sensitivity microphone array and positioning system, enabling continuous soundscape collection and dynamic path optimization in the air. At the model layer, sound spectral signals are modeled as time-frequency node graphs. Through Graph Neural Network (GNN) and self-supervised contrastive learning, co-occurrence features are extracted to construct an acoustic co-occurrence graph. Experimental results show that this method achieves a multi-label species recognition F1 score of 0.87, which is 15%～20% higher than that of Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN). It can effectively reveal the dynamic structure of community hierarchies, providing a new path for ecological acoustic monitoring.
HE Haixuan. Research on Unmanned Aerial Vehicle Path Planning Based on Improved Ant Colony Algorithm[J]. Technology of IoT&AI, 2026, 58(3): 65-68.
This study investigates an enhanced ant colony algorithm for unmanned aerial vehicle path planning, addressing the limitations of conventional methods including excessive blind optimization, low convergence efficiency, and susceptibility to local optima. The proposed strategy integrates target-oriented mechanisms, path potential evaluation, iterative feedback updates, and environmental adaptive adjustments. By constructing a multi-scale dynamic raster environment model and optimizing state transition rules with pheromone regulation mechanisms, the algorithm demonstrates improved adaptability to complex flight environments. Simulation results demonstrate that the improved algorithm outperforms traditional ant colony algorithms in both path length optimization rate and optimization efficiency, providing more practical optimal paths for unmanned aerial vehicle under dynamic constraints.
CHEN Xuechao. Noise Reduction and Recognition Method of Optical Fiber Acoustic of Oil and Gas Pipeline Based on Multi-classifier Fusion[J]. Technology of IoT&AI, 2026, 58(3): 69-72.
Addressing the issues of susceptibility to multi-source noise interference and low accuracy in event recognition in distributed optical fiber acoustic signals for oil and gas pipelines, this paper proposes a noise reduction and recognition method based on multi-classifier fusion. Firstly, the signal acquisition principle and the spatiotemporal frequency characteristics of typical events such as leaks, excavations, and vehicle passages are analyzed to clarify the noise interference mechanism. Secondly, adaptive filtering preprocessing is adopted to optimize the reference signal using spatial correlation, suppressing environmental noise and common mode interference. Lastly, 3 complementary classifiers, namely support vector machine, random forest, and one-dimensional convolutional neural network, are constructed. Based on the validation set, the weight coefficients are optimized, and a weighted probability fusion strategy is designed. Experiments show that this method improves the Signal-to-Noise Ratio (SRN) by an average of 1.7～2.1 dB after noise reduction. The overall accuracy of event recognition is 95.7%, and the recall rate for leaks is 97.8%. This method outperforms single classifiers and provides reliable technical support for oil and gas pipeline safety monitoring.
LIU Yong, LIU Xuezong. Research and Application of Data Processing System Based on Notification and Concurrent Processing[J]. Technology of IoT&AI, 2026, 58(3): 73-78.
In the field of urban rail transit signal and monitoring system, the real-time data processing mechanism is the key factor affecting the data processing performance of the real-time system. This paper discusses the insufficiencies of periodic scanning in the process of real-time data processing, and proposes a notification concurrent processing method. This paper focuses on the design scheme based on notification concurrent processing mode and the problem analysis of concurrent process, and discusses the corresponding solution. The notification based concurrent processing method mentioned in this paper has been used in the real-time data processing system dealing with large data capacity to ensure the real-time performance of the system.
WANG Hongyu. Multi-task Cropland Extraction Method Based on Boundary Semantic Perception#br#[J]. Technology of IoT&AI, 2026, 58(3): 79-84.
To address the issue of insufficient geometric accuracy caused by blurred boundaries in cropland extraction, a multi-task extraction method based on boundary semantic perception is proposed. A dual-branch architecture is constructed based on a Transformer backbone for boundary extraction and cropland prediction, introducing boundary information as a semantic constraint to enhance the geometric accuracy of cropland recognition. The model is composed of two core modules: a Boundary Enhancement Module (BEM), which is constructed to refine cropland contours using boundary features to mitigate boundary ambiguity, and a Multi-scale Feature Enhancement Module (MFEM), which is designed to integrate local and global information through multi-scale convolution to improve adaptability to cropland with varied shapes and sizes. Experiments on a Shandong dataset demonstrate that the proposed method significantly outperformed Convolutional Neural Network (CNN)-based models (U-Net and PSPNet) and Transformer-based models (Swin-Transformer) in terms of Intersection over Union (IoU) and F1-score, while achieving the lowest geometric error, validating its superior performance in cropland extraction tasks.
WANG Yajun, ZHOU Ping. Research on Multimodal Artificial Intelligence Perception Method in Edge Computing Environment#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(3): 85-88.
The edge computing architecture provides a low-latency and high-efficiency implementation path for multimodal artificial intelligence perception. By deploying lightweight neural networks at edge nodes and applying compression techniques such as model pruning, quantization, and knowledge distillation, the model size is reduced to 15 MB, and the inference latency is controlled within 95 ms. Combined with the attention fusion mechanism and distributed collaborative inference strategy, the multimodal perception accuracy is increased to 91.3%, effectively reducing network bandwidth occupation and system response latency in scenarios such as intelligent manufacturing and autonomous driving, demonstrating significant practical value.
ZHAO Lin, SONG Weiyang. Global Optimal Path Intelligent Planning Method for Fully Autonomous Mobile Robots[J]. Technology of IoT&AI, 2026, 58(3): 89-92.
To solve the problems of low efficiency, poor adaptability, and multi-objective trade-offs in path planning for fully autonomous mobile robots in complex dynamic environments, a global optimal path intelligent planning method for fully autonomous mobile robots is proposed. Using graph attention network and spatial reconstruction unit to extract global optimal path features of the environment, adopting improved Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm to achieve accurate division of static/dynamic obstacles and traffic areas, constructing Long Short-Term Memory (LSTM) motion prediction model and multi-sensor fusion obstacle trajectory prediction model. Conduct global path search under the improved ant colony algorithm framework, and introduce composite weights of path length, time, safety, and energy consumption. Using cubic B-spline curves for smooth optimization of the path and introducing curvature constraints to suppress the Runge phenomenon. The experiment shows that this method reduces the average path deviation by about 33% during low load periods, reduces the number of replanning times by about 66%, and increases the success rate of dynamic obstacle avoidance to 98%; During high load periods, the average path deviation is reduced by 40%, the number of re planning times is reduced by about 52%, and the success rate reaches 95%.
DU Pengqiang. Cyclic Cooperative Topology-Enhanced Small-Sample Skeleton Action Recognition Network[J]. Technology of IoT&AI, 2026, 58(3): 93-98.
In the data-driven era, achieving high-precision action recognition with limited labeled data remains a key challenge in the field of computer vision. To address the issues of poor generalization and structural overfitting in existing graph convolutional networks under small-sample scenarios due to their over-reliance on a single physical topology, this paper proposes a tri-view cooperative graph convolutional network. By breaking the limitations of traditional single topology structures, it constructs a complementary ternary topological framework through a topology augmentation mechanism. This framework comprises: a local structure view to maintain anatomical physical constraints, a global pose view to reinforce spatial configuration stability, and a latent correlation view to mine semantic coordination among limbs, thereby compensating for the scarcity of structural information under feature-scarce conditions. Furthermore, it proposes a novel cyclic cooperative attention mechanism, enabling dynamic feature calibration through circular interaction among heterogeneous views. This mechanism utilizes each topological feature stream as a query to retrieve complementary information from other views, enforcing bi-directional error correction during the feature evolution process. This allows the model to extract highly discriminative, intrinsic action features even with extremely limited data. Experimental results indicate that the proposed topological cooperative strategy exhibits outstanding data efficiency and significantly outperforms current mainstream models.
HE Jinlu, QIN Shan, YANG Zhiyu, LING Xianliang. Research on Key Technologies for Intelligent Contract Review Based on AI Large Models[J]. Technology of IoT&AI, 2026, 58(3): 99-102.
Aiming at the problems of low efficiency, insufficient risk identification, and difficulty in adapting to multiple types in contract review, a key technology for intelligent contract review based on Artificial Intelligence (AI) big model is proposed. It elaborates on the core principles of knowledge enhancement, semantic recognition, and reasoning chain modeling, introduces the model architecture, experimental data, and performance evaluation methods, and validates them through multiple types of contract examples. The results indicate that the system can effectively achieve collaborative optimization of clause semantic understanding and risk identification, and has broad application value.
CHEN Hongfeng. Automatic Baggage Detection Method for Airport Based on Multi-scale Feature Enhanced Transformer#br#[J]. Technology of IoT&AI, 2026, 58(3): 103-106.
Addressing the issues of significant target scale variations, severe item occlusions, and limited receptive field of traditional convolutional neural networks in airport security inspection for contraband detection, a Transformer detection model that integrates a multi-scale feature pyramid with adaptive spatial attention is proposed. This model constructs a hierarchical feature extraction architecture, combining the local features of convolutional networks with the global modeling capabilities of Transformers. Experiments conducted on the SIXray dataset demonstrate that the method achieves a mean Average Precision (mAP) of 91.7%, with particularly outstanding performance in tool detection.
FANG Jing. Exploring Path of Integrating AIGC Technology into Practical Teaching of Short Video Operations[J]. Technology of IoT&AI, 2026, 58(3): 107-110.
The Artificial Intelligence Generated Content (AIGC) technology is continuously developing and being widely adopted. The short video industry has undergone significant changes. The previous course teaching model is no longer able to meet the industry’s demand for versatile talents. This article reviews the development history of AIGC technology and its commonly used software tools, analyzes the current predicaments of short video operation teaching, explores the teaching advantages of AIGC in short video operation, and constructs a four-stage teaching application path of “technical cognition - software practice - scenario application - ability improvement”. It provides theoretical support and practical solutions for the deep integration of AIGC technology and short video operation,and helps cultivate high-quality talents with technical literacy, creative ability, and industry adaptability.
YANG Yue, WANG Ceren, LIU Wenjing. Lightweight Tomato Leaf Diseases Detection via Attention Mechanism[J]. Technology of IoT&AI, 2026, 58(3): 111-116.
This paper proposes a lightweight object detection model integrated with an attention mechanism to enhance the detection accuracy and efficiency of tomato leaf diseases and pests in complex field environments. By embedding a C2f-CBAM module into the backbone network, key feature responses across channel and spatial dimensions are strengthened. A PCBlock structure is introduced to achieve model lightweighting, while a bidirectional feature fusion path is constructed in the neck network to improve multi-scale contextual information integration. Additionally, the regression strategy of the detection head is optimized to increase inference speed. Experimental results show that the proposed method achieves an mean Average Precision (mAP) of 91.3% on a tomato disease and pest dataset, representing a 1.6 percentage point improvement over the baseline model. With only 5.87×106 parameters and 7.8×109 Number of Floating Point Operations (FLOPs), the approach significantly outperforms mainstream algorithms and demonstrates strong potential for field deployment.
LI Xindong, DI Mengyao. Small Target Detection of Corn Leaf Diseases and Pests Based on Improved YOLOv10n[J]. Technology of IoT&AI, 2026, 58(3): 117-122.
Corn leaf diseases and pests are characterized by small lesion spots and high texture-background coupling, posing significant challenges to identification. To address the issues of insufficient attention to small targets and limited localization stability, this paper proposes YOLOv10n-ACE, a detection model based on YOLOv10n. In terms of network architecture design, a lightweight channel attention mechanism is introduced to strengthen the representation of lesion-related features, and a bidirectional feature fusion structure is incorporated to enhance multi-scale semantic information interaction, thereby improving the model’s perception of small-scale lesions. For object localization, a geometric constraint optimization strategy is applied to improve localization robustness under ambiguous lesion boundaries. Experimental results and ablation studies demonstrate that the YOLOv10n-ACE model achieves improvements of 1.1 percentage points in mAP@0.5 and 1.8 percentage points in mAP@0.5:0.95 compared to the baseline model, validating the effectiveness of the proposed method and offering reliable technical support for intelligent disease and pest recognition.
LIU Xinglong. Trusted Data Sharing Method for Internet of Things in Cross-Domain Sensing Environments#br#[J]. Technology of IoT&AI, 2026, 58(3): 123-126.
Aiming at the problems of trust deficiency and permission management in cross-domain Internet of Things data sharing, a trusted data sharing method is proposed. A hierarchical trust measurement algorithm, an adaptive permission control framework, and a lightweight traceability mechanism are designed. Simulation tests show that the method achieves a response delay of 85 ms, a throughput of 265 transactions per second, and a security detection rate of 94.3%, effectively balancing efficiency and security, providing technical support for cross-domain data sharing.
GUO Zhengguo. Research on Multi-source Information Fusion Technology for Network Security Intrusion Detection Systems[J]. Technology of IoT&AI, 2026, 58(3): 127-131.
Traditional intrusion detection systems relying on a single data source face challenges such as insufficient detection accuracy and high false alarm rates. Multi-source information fusion technology leverages heterogeneous information sources such as network traffic, system logs, and application-layer data to construct a collaborative detection framework based on improved D-S evidence theory. The system employs a dynamic weight allocation model to adaptively adjust the credibility of data sources. Comparative experiments demonstrate that the multi-source fusion approach achieves a detection accuracy rate of 94.8%, an improvement of 9.6 percentage points compared to single-source detection, and reduces the false alarm rate to 3.8%. This provides an efficient and reliable technical solution for network security protection.
LIU Junjie, ZHAO Min, TAN Yixuan, HE Chenzidu. OTN Network Performance Degradation Prediction Technology Based on LSTM Network[J]. Technology of IoT&AI, 2026, 58(3): 132-135.
Addressing the practical issue of performance parameter degradation during Optical Transport Network (OTN) operation, this study investigates the application of Long Short-Term Memory (LSTM) networks in time-series prediction. It details multi-dimensional metric feature extraction, model architecture design, and optimization methods, while presenting prediction and early warning effectiveness under various operational conditions. Results demonstrate that the model exhibits high prediction accuracy and stability in complex scenarios, effectively supporting OTN performance degradation trend analysis.
JIANG Kai. Optimized Design of Remote Centralized Monitoring Solution Based on Unified Integration at Central Control Side#br#[J]. Technology of IoT&AI, 2026, 58(3): 136-140.
To enhance the efficiency of remote monitoring for new energy stations, a remote centralized monitoring and inventory solution based on unified integration at the centralized control side is proposed. This solution utilizes a keyboard, video, and mouse (KVM) along with a data fusion mechanism to study the optimization of the monitoring architecture under concurrent conditions across multiple stations. The results indicate that after unified integration and optimization at the centralized control side, the core performance indicators of the system have been significantly improved: response time has been shortened from 4.5 min to 2.3 min; mean time between failures has increased from 150 h to 170 h; failure rate has decreased from 10.5% to 7.3%; data transmission rate has increased from 100 Mb/s to 130 Mb/s; network bandwidth utilization has increased from 65% to 90%. Simultaneously, under the centralized monitoring and inventory mode, personnel allocation has been reduced from 12 to 7, annual labor costs have decreased from 1.8 million yuan to 1.05 million yuan, the number of monitoring terminals has been reduced from 35 to 18, total hardware power consumption has decreased from 9.4 kW to
6.2 kW, annual electricity consumption has decreased from 8.24×104 kW·h to 5.44×104 kW·h, network resource utilization has decreased from 65% to 38%, dedicated bandwidth demand has decreased from 500 Mb/s to 300 Mb/s, and annual comprehensive costs have decreased from 1.8 million yuan to 950 000 yuan. This demonstrates that the centralized monitoring and inventory solution exhibits significant advantages in system response performance, operational stability, and operation and maintenance cost control.
ZHANG Naisheng, LONG Yan, LU Dean, XIONG Bi. Research on the Safety Monitoring System of Hydropower Dam Based on Internet of Things[J]. Technology of IoT&AI, 2026, 58(3): 141-144.
In the safety monitoring of hydropower station dams, addressing the issues of poor real-time performance and data fusion difficulties caused by scattered sensor deployment and heterogeneous multi-source data, this paper designs an Internet of Things monitoring system based on an edge-cloud collaborative architecture. This system achieves unified access and reliable transmission at the sensing layer; performs data synchronization and feature extraction at the edge side; and conducts dynamic fusion and intelligent diagnosis at the cloud side to enhance system response and hazard identification capabilities. Experiments show that this system is real-time and has high accuracy in multi-source data alignment, providing a scalable and deployable engineering solution for dam safety monitoring.
YANG Bowen. Smart Construction Site Safety Control Platform Driven by Digital Twins#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(3): 145-148.
To build an efficient construction safety management system, a smart construction site safety control platform based on digital twins is designed. The platform adopts a three-layer architecture of “perception—twin—application”, integrating multi-source data collection, twin modeling, and intelligent warning algorithms to achieve dynamic monitoring and risk prediction of the environment, equipment, and personnel. The actual test results showed that the data transmission delay was 0.86 s, the risk identification rate increased to 94.7%, the alarm response time decreased from 92 s to 36 s, and the comprehensive risk index decreased from 0.72 to 0.28. The system significantly improved the accuracy and real-time performance of security control.
LI Fang. Research on Monitoring and Analysis Technology for Seismic Performance of Building Structures[J]. Technology of IoT&AI, 2026, 58(3): 149-152.
In this study, a real-time monitoring system is constructed, which combines multi-source sensor network and digital twin model to dynamically evaluate and intelligently warn the seismic performance of building structures. The system collects structural response data through multi-source sensor networks, establishes a virtual model with digital twin technology, combines multi-modal data fusion and feature extraction methods, and then uses deep learning algorithm to identify structural parameter changes and potential damage. From the perspective of practical engineering application, this technology has greatly improved the ability of early warning of seismic safety of building structures, and provided strong technical support for safety management of building structures and urban earthquake resistance and disaster reduction.
ZHAO Jin’e, ZHONG Ruiyan, HOU Huapeng, GUO Baocang, LI Chao, KONG Xiangbin. Research on Data Fusion and Visualization Technology of Multi-source Heterogeneous Data in New Energy Centralized Control Center#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(3): 153-157.
During operation, a new energy centralized control center needs to process multi-source heterogeneous data from wind power, photovoltaic, energy storage, and meteorological systems. These data differ significantly in structure, time scale, and correlation patterns, which leads to limitations in existing studies on holistic operation state presentation and situational awareness. Focusing on the operational monitoring requirements of new energy centralized control centers, this paper investigates multi-source heterogeneous data fusion and visualization technologies. By improving unified data modeling and correlation fusion methods and integrating the fusion results with visualization design, an intuitive and comprehensive presentation of system operating states is achieved.
LIU Feng. Application of Internet of Things Technology in the Renovation of Fire Prevention and Control Systems in Old Residential Areas[J]. Technology of IoT&AI, 2026, 58(3): 158-161.
Due to structural aging and facility degradation, old residential areas have formed high-frequency fire risks, with hidden dangers showing a trend of multi-source dispersion and dynamic evolution. Traditional prevention and control methods have obvious adaptability deficiencies in continuous monitoring and rapid response. This article analyzes the characteristics of fire hazards in old residential areas, points out the structural bottlenecks of traditional models in detection coverage, alarm links, and response rhythm, and proposes an Internet of Things technology transformation path consisting of smoke and temperature detection point reconstruction, electrical fault monitoring access, weak signal link strengthening, and platform linkage disposal. Based on low deployment costs, rapid deployment, and high-density perception capabilities, differentiated advantages are formed, and an intelligent prevention and control system with continuous coverage, stable transmission, and accurate judgment is constructed to provide innovative support for the scalable technical route of fire safety in old residential areas.
YE Zhenxiao. Construction and Application Empowerment of Vehicle-Pile-Grid Integrated Platform[J]. Technology of IoT&AI, 2026, 58(3): 162-167.
To address the challenges of data fragmentation, low collaboration efficiency, and insufficient value mining among “vehicles, charging piles, and the grid” in the new energy vehicle industry, this paper proposes an integrated big data platform. The platform achieves multi-source data fusion through unified identification and data governance, and is applied in 4 dimensions: government supervision, enterprise operation, public service, and industrial ecosystem. Practice shows that the platform can effectively enhance regulatory efficiency and corporate benefits, and offer a feasible pathway for the intelligent upgrading of the industry.
YANG Liushu. Design of AIoT Agricultural Product Supply Chain Collaborative Management System Adapted to Agricultural Scenarios[J]. Technology of IoT&AI, 2026, 58(3): 168-172.
In response to the significant information isolation effect in the production and sales of vegetables in Tonghai county, Yunnan province, the lag in processing unstructured data, and the high loss rate caused by the breakage of cold chain logistics, a supply chain collaborative management system architecture based on Artificial Intelligence of Things (AIoT) technology for agriculture is proposed. This system builds a three-level collaborative architecture of “end, edge, cloud”, suppresses the sensor noise in the strong interference environment of the plateau through data cleaning on the edge side, establishes a shelf life prediction model adapted to the high respiration rate characteristics of Tonghai leafy vegetables by using the improved Arrhenius equation, and realizes multi-subject trusted traceability through the alliance chain. During the three-month trial operation on the typical route from Tonghai to Guangzhou, the average spoilage rate of leafy vegetables decreased from 12.4% to 4.6%, the average timeorder response time was shortened by about 63.1%, and the average net profit of bicycles increased by about 39.3%. The attribution analysis shows that the reduction in loss is mainly attributed to the collaborative effect of edge warning and dynamic scheduling. This system provides an engineering solution that takes into account both technical adaptation and social embedding for the digital transformation of plateau characteristic agriculture.
XU Lijun. Intelligent Technology-Driven Reform Practice of Financial Big Data Course for Competency Development[J]. Technology of IoT&AI, 2026, 58(3): 173-176.
To address weak data processing competency in traditional financial education, this paper reforms the curriculum under the “data-driven, competency-oriented” concept, builds a 4 tier competency framework, introduces a Hadoop+Spark training platform, integrates real-world scenarios including isolation forest anomaly detection and XGBoost forecasting, and adopts project-based teaching. After reform, student skill scores improved by 34.1% and task efficiency by 60%, with notable results.
