Skip to main content
05RESEARCH / PUBLICATIONS

Turning discovery into evidence.

From open questions to verifiable methods: our research outputs and scholarly work.

Research

59 public outputs
02

Publications

2026PUBLICATIONS

  • IEEE Transactions on Smart Grid

    Condition-Driven Direct-to-LEO Satellite Multi-UAV Inspection for Offshore Wind Farms

    Y. Wen, Q. Zhou, Z. Xia, Q. Zhong, Y. Shen, and T. Sun

    Research summary · Draft — pending author review

    This condition-driven multi-UAV inspection approach uses direct-to-LEO satellite links for offshore wind farms. It connects turbine electrical anomalies with flight paths, jointly schedules sensor activation and inspection priorities, and formulates a Markov decision process solved with K-means-prepartitioned SAC. Simulations across sensing loads and communication conditions reduce inspection completion time; post-processing shows that earlier intervention can shorten abnormal operation and output restrictions.

    2026
  • IEEE Transactions on Smart Grid

    Optimal Integrated Satellite-Terrestrial Network Design for Microgrid Clusters Under Wildfires

    X. Zhang, Z. Lian, R. Xie, C. Wu, and Q. Zhou

    Research summary · Draft — pending author review

    This work designs satellite–terrestrial networks for remote microgrid clusters whose terrestrial communication infrastructure is damaged by wildfire. It models fire evolution and link failures to assess reliability, uses ALT-Fiedler alternating optimisation to restore topology with limited satellite terminals, and applies semidefinite programming to balance convergence and delay robustness. Simulations demonstrate restored connectivity and reliable coordinated operation with limited terminals.

    2026
  • IEEE Transactions on Smart Grid

    Distributed Resilient Control for Satellite-Aided Remote Microgrid Clusters

    Z. Lian, F. Liu, H. Yu, Y. Zhu, Y. Chen, and M. Shahidehpour

    Research summary · Draft — pending author review

    An observer-based distributed resilient control strategy addresses measurement noise and communication delay in satellite-assisted remote microgrid clusters. Distributed observers compensate relative-output estimation errors, with secondary controllers and parameter selection guided by H∞ performance. Theory and OPAL-RT real-time experiments verify voltage restoration and accurate current sharing under combined noise and delay, including strong noise and substantial satellite-link latency.

    2026
  • IEEE Transactions on Industrial Electronics

    Secure Microgrid Clusters Under Integrated Satellite-Terrestrial Networks: A Differential Privacy-Preserving Control Strategy

    Z. Lian, H. Yu, Y. Chen, M. Shahidehpour and Q. Zhou

    Research summary · Draft — pending author review

    This differential-privacy-preserving control strategy protects operational data exchanged by microgrid clusters over satellite–terrestrial networks. It adds random noise with non-decaying variance and uses distributed consensus control with time-varying gains to preserve convergence and stability. Theory, OPAL-RT tests and digital signal processor experiments show frequency regulation and active-power sharing while meeting specified privacy targets.

    2026
  • IEEE Transactions on Smart Grid

    Reinforcement Learning-Based Multi-UAV Deployment for Power Corridor Monitoring With Satellite Internet

    K. He, Q. Zhou, Z. Xia, S. Yuan, Q. Zhou, and C. Sun

    Research summary · Draft — pending author review

    This satellite-supported reinforcement-learning method deploys multiple UAVs for large power corridors with uneven risk and limited resources. A joint UAV–satellite monitoring framework weights heterogeneous monitoring needs by risk, while a contextual multi-armed bandit coordinates sensing resources and UAV placement to balance energy and data collection. Simulations outperform compared methods in energy efficiency, data acquisition and convergence speed.

    2026
  • IEEE Transactions on Smart Grid

    Cognitive Radio Empowered Smart Grid

    Z. Xia, Q. Zhou, M. Shahidehpour and Z. Shuai

    Research summary · Draft — pending author review

    This cognitive-radio-enabled smart-grid architecture jointly designs wireless communication resources and power-system control. Distributed generators communicate with a main controller through adaptive wireless links. Finite-blocklength models describe short-packet delay and error rates in a multi-objective optimisation with finite-time stability constraints. Simulated satellite-assisted offshore turbine control and remote renewable generation show coordinated latency, reliability and communication–control costs.

    2026
  • International Conference on Machine Learning (ICML)

    Is Task-Specific Training Necessary for Anomaly Detection?

    Xingwu Zhang, Guanxuan Li, Paul Henderson, Gerardo Aragon Camarasa, Zijun Long*

    * Corresponding author(s): Zijun Long

    Research summary · Draft — pending author review

    RAD is a retrieval-based anomaly detection framework without task-specific training, addressing reliance on training and the fidelity–stability tension of reconstruction scores. It stores normal-image feature memories and matches test regions through multi-level retrieval to reduce cross-class and cross-region mismatches. Experiments on MVTec-AD, VisA, Real-IAD and 3D-ADAM show strong detection in standard and few-shot settings, supporting anomaly detection through pretrained features and normal-example retrieval.

    About this work

    * Corresponding author.

    Source code

    2026
  • International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR)

    ADaFuSE: Adaptive Diffusion-generated Image and Text Fusion for Interactive Text-to-Image Retrieval

    Zhuocheng Zhang, Xingwu Zhang, Kangheng Liang, Guanxuan Li, Richard McCreadie, Zijun Long*

    * Corresponding author(s): Zijun Long

    Research summary · Draft — pending author review

    ADaFuSE is a lightweight adaptive fusion model that addresses noise introduced by static fusion of diffusion-generated images in interactive retrieval. A dual-branch design uses adaptive gating to balance image and text reliability, followed by a semantic-aware mixture of experts for fine-grained cross-modal relations. It integrates without modifying the backbone. Experiments on four standard benchmarks improve on DAR, with greater robustness to noisy and longer interactive queries.

    About this work

    * Corresponding author.

    2026
  • IEEE International Conference on Robotics and Automation (ICRA)

    RoboEye: Enhancing 2D Robotic Object Identification with Selective 3D Geometric Keypoint Matching

    Xingwu Zhang, Guanxuan Li, Zhuocheng Zhang, Zijun Long*

    * Corresponding author(s): Zijun Long

    Research summary · Draft — pending author review

    RoboEye addresses occlusion, viewpoint changes and visually similar objects in warehouse recognition. It first ranks candidates with 2D semantic features, then evaluates 3D feature quality and enables geometric reranking only when useful, using geometry-aware features and keypoint confidence to distinguish objects. Experiments improve first-rank recall over RoboLLM. Inference uses RGB images without additional 3D input, avoiding unnecessary geometric computation.

    About this work

    * Corresponding author.

    Source code

    2026
  • IEEE Transactions on Industrial Electronics (TIE)

    Regionally Routed Attention U-Net With Latent Fading Kalman Filtering for Real-Time EIT

    Tianyi Deng, Zhaosheng Teng, Shin Takahashi, Bojun Liu, Qiu Tang, Zijun Long*

    * Corresponding author(s): Zijun Long

    Research summary · Draft — pending author review

    RNU-Net balances structural reconstruction accuracy and low latency for real-time electrical impedance tomography. It combines one-dimensional ConvNeXt, regionally routed attention and U-Net, with latent fading Kalman filtering for dynamic sensor-domain denoising at low signal-to-noise ratios. Synthetic data, physical tank measurements and independent KIT4 tests evaluate reconstruction and cross-domain performance; mobile GPU and standard CPU timing demonstrates the feasibility of accurate real-time EIT monitoring.

    About this work

    * Corresponding author.

    2026
  • Automation of Electric Power Systems

    Short-Term Photovoltaic Power Forecasting Using Information Gain from Satellite–Terrestrial Micro-Observation Stations

    Jianru Qian, Quan Zhou, Yao Wen, et al.

    Research summary · Draft — pending author review

    This approach improves short-term photovoltaic forecasting with sparse weather observations, low data resolution and unstable communication. Mutual information guides micro-station placement; numerical weather forecasts, XGBoost and physical models predict hourly trends; and a temporal convolutional network with multi-head attention compensates minute-scale fluctuations. Seasonal cases show improved accuracy, responsiveness and stability over comparison models.

    2026
  • Automation of Electric Power Systems

    Cyber–Physical Coordinated Recovery of Integrated Satellite–Terrestrial Distribution Systems Under Wildfires

    Shiyuan Qiu, Quan Zhou, Xuan Zhang, et al.

    Research summary · Draft — pending author review

    This study connects disaster sensing, firefighting dispatch and load restoration through a coordinated cyber–physical recovery strategy. A LEO satellite-supported three-layer framework jointly optimises fire monitoring, air–ground firefighting resources and terminal deployment, solved with linearised two-stage optimisation. Simulations on an IEEE 33-bus system and a 10-node sensing network show shorter recovery times, more restored load and improved resource allocation.

    2026

2025PUBLICATIONS

  • IEEE Transactions on Industrial Informatics

    Distributed cyber resilient control strategy for remote DC microgrids under integrated satellite terrestrial networks

    Z. Lian, Y. Zhu, F. Guo, C. Peng, and Q. Zhou

    Research summary · Draft — pending author review

    This distributed cyber-resilient controller addresses simultaneous denial-of-service attacks and communication noise in remote DC microgrids using satellite–terrestrial networks. Intermediate observers estimate neighbouring voltage and power-sharing states, with compensation for noise and attacked links. Theory establishes voltage restoration and accurate power sharing under the stated conditions, with OPAL-RT and digital signal processor cases validating the strategy.

    2025
  • IEEE Transactions on Industry Applications

    Zero-Trust Security Strategy for Renewable Power Plant Clusters Under Integrated Satellite-Terrestrial Networks

    C. Peng, Q. Zhou, Y. Shen, K. He, Z. Lian, and Z. Shuai

    Research summary · Draft — pending author review

    This zero-trust strategy protects renewable plant clusters from false-data injection through open satellite–terrestrial networks. A lightweight blockchain without a consensus mechanism records identities and interactions to create traceable dynamic network boundaries. A convolutional network analyses identity and data features, and Dempster–Shafer evidence theory combines trust assessments. Simulations and hardware-in-the-loop tests validate attack detection and customised access control.

    2025
  • IEEE Transactions on Industrial Informatics

    Neural Differential Equations-Driven Predictive Control for Improving Microgrids Stability With Satellite Internet

    Y. Shen, Q. Zhou, J. Qian, and Z. Shuai

    Research summary · Draft — pending author review

    SNPC uses neural differential equations for predictive control of remote microgrids affected by satellite-link noise and large disturbances. It combines model-based interpretability with data-driven flexibility to predict centre-of-inertia trajectories and converter dynamics and optimise stabilising signals across control-mode changes. Simulations and hardware-in-the-loop tests on an IEEE 38-bus system and a real islanded microgrid, using measured satellite delays, demonstrate improved transient stability and convergence.

    2025
  • IEEE Energy Sustainability Magazine

    Pioneering Climate Change Adaptation: Sustainable Power Systems Enhanced by Integrated Satellite-Terrestrial Networks

    Y. Shen, Q. Zhou, and J. Zhao

    Research summary · Draft — pending author review

    This study examines satellite–terrestrial networks as a means of adapting sustainable power systems to climate-driven extreme weather. It organises physical, information and benefit-related challenges through a four-stage Aware, Assess, Adjust and Act framework. Communication, navigation, remote sensing and heterogeneous data support disaster awareness, stability assessment, emergency adjustment and recovery, providing technical pathways and application analysis for climate-resilient power systems.

    2025
  • International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR)

    Diffusion Augmented Retrieval: A Training-Free Approach to Interactive Text-to-Image Retrieval

    Zijun Long, Kangheng Liang, Gerardo Aragon Camarasa, Richard McCreadie, Paul Henderson

    Research summary · Draft — pending author review

    DAR is a diffusion-augmented framework for interactive text-to-image retrieval without task-specific fine-tuning. A large language model refines conversational requirements, while diffusion models generate multiple intermediate visual representations to capture semantic and visual intent. Experiments on four benchmarks show performance comparable to fine-tuned methods for simple queries and stronger retrieval for complex multi-turn queries, reducing ongoing training and update costs.

    2025
  • ACM Transactions on Recommender Systems (TORS)

    Enhancing Recommender Systems: Deep Modality Alignment with Large Multi-Modal Encoders

    Zixuan Yi*, Zijun Long*, Iadh Ounis, Craig Macdonald, Richard McCreadie

    * Equal contribution: Zixuan Yi; Zijun Long

    Research summary · Draft — pending author review

    This work studies deep alignment in large multimodal encoders for recommendation, where separately encoded images and text can remain poorly aligned. It compares CLIP, VLMo and BEiT-3 using pretrained features, recommendation-data fine-tuning and end-to-end training, and analyses each modality’s contribution. Across four datasets, large multimodal encoders improve user and item representations, with further gains from end-to-end training; dual-stream CLIP offers a favourable balance of effectiveness and efficiency.

    About this work

    * Equal first authors.

    2025

2024PUBLICATIONS

  • IEEE Transactions on Industry Applications

    Spatiotemporal Precise Routing Strategy for Multi-UAV-Based Power Line Inspection With Integrated Satellite-Terrestrial Network

    K. He, Q. Zhou, Z. Lian, Y. Shen, J. Gao, and Z. Shuai

    Research summary · Draft — pending author review

    This spatiotemporal routing strategy supports UAV power-line inspections when disasters disrupt terrestrial communication. It coordinates satellite internet and flying ad hoc networks, balancing communication paths and UAV energy use under fluctuating signals. Simulations show more stable inspection communication and longer fleet operating times than conventional routing algorithms, supporting faster fault location and grid recovery.

    2024
  • IEEE Transactions on Industry Applications

    LEO Satellite Internet Assisted Data-Driven Predictive Control for Remote Islanded Microgrid Stability Enhancement

    Y. Shen, Q. Zhou, S. Fang, C. Peng, K. He, and Y. Li

    Research summary · Draft — pending author review

    A LEO satellite-assisted data-driven predictive controller addresses stability risks in remote islanded microgrids during typhoons, heavy rain and other disturbances. Satellite links connect converters, a two-layer gated recurrent unit predicts dynamics with less sensitivity to model mismatch, and dynamic consensus coordinates control. Cases based on a real remote microgrid demonstrate reduced oscillations, improved coordination and maintained stability while accounting for satellite delay.

    2024
  • IEEE Transactions on Smart Grid

    Integrated satellite-terrestrial network framework for next generation smart grid

    Y. Shen, Q. Zhou, Y. Wen, Z. Shuai, and Z. J. Shen

    Research summary · Draft — pending author review

    This next-generation smart-grid framework integrates satellite and terrestrial networks to address wide coverage, complex environments and emergency resilience with high renewable penetration. It combines satellite communication, navigation and remote sensing, using information entropy to coordinate aerial and terrestrial resources for urban, remote and disaster-response scenarios. A cyber–physical platform based on a real experimental LEO constellation demonstrates feasibility.

    2024
  • European Conference on Computer Vision (ECCV)

    Understanding and Mitigating Human-Labelling Errors in Supervised Contrastive Learning

    Zijun Long, Lipeng Zhuang, George Killick, Richard McCreadie, Gerardo Aragon Camarasa, Paul Henderson

    Research summary · Draft — pending author review

    This study examines how real human labelling errors differ from synthetic label noise and create incorrect positive pairs in supervised contrastive learning. It introduces SCL-RHE, a contrastive objective designed to resist the relatively low error rates found in real annotations. Experiments on multiple vision benchmarks show improved robustness relative to the compared representation-learning and noise-handling methods.

    2024
  • International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR)

    CFIR: Fast and Effective Long-Text To Image Retrieval for Large Corpora

    Zijun Long, Xuri Ge, Richard McCreadie, Joemon Jose

    Research summary · Draft — pending author review

    CFIR is a coarse-to-fine index-shared retrieval framework for complex long-text queries over large image collections. Entity-based ranking first selects candidates, followed by summary-based reranking, with a decoupled BEiT-3 encoder supporting vector similarity in both stages. Experiments on AToMiC show higher recall than the compared large multimodal models with substantially lower training and retrieval time.

    2024
  • IEEE International Conference on Robotics and Automation (ICRA)

    RoboLLM: Robotic Vision Tasks Grounded on Multimodal Large Language Models

    Zijun Long, George Killick, Richard McCreadie, Gerardo Aragon Camarasa

    Research summary · Draft — pending author review

    RoboLLM provides a unified framework for robotic segmentation, recognition and defect detection, reducing the engineering effort of integrating separate models. It adapts the pretrained knowledge of a BEiT-3 multimodal backbone to multiple warehouse perception tasks. Experiments on the real-world ARMBench dataset outperform comparison baselines and demonstrate a route to simpler model integration and task transfer.

    2024
  • IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

    MultiWay-Adapter: Adapting Large-Scale Multimodal Models for Scalable Image-Text Retrieval

    Zijun Long, George Killick, Richard McCreadie, Gerardo Aragon Camarasa

    Research summary · Draft — pending author review

    MultiWay-Adapter addresses the computational cost of adapting large multimodal models and the shallow alignment of existing efficient fine-tuning methods. A modality alignment enhancer strengthens deep image–text relationships with a small set of trainable adapter parameters. Experiments show reduced training time with competitive retrieval performance and few additional parameters, supporting resource-efficient adaptation of large multimodal encoders.

    2024
  • IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

    LaCViT: A Label-aware Contrastive Training Framework for Vision Transformers

    Zijun Long, George Killick, Richard McCreadie, Gerardo Aragon Camarasa, Zaiqiao Meng

    Research summary · Draft — pending author review

    LaCViT addresses the limitations of cross-entropy-only fine-tuning for vision transformer generalisation and transfer. It uses target-task labels to organise contrastive learning, improve the pretrained feature space and separate classes before downstream classification. Experiments on eight standard image classification datasets show improved performance for the evaluated transformers, demonstrating the value of label relationships during fine-tuning.

    2024
  • European Conference on Artificial Intelligence (ECAI)

    CLCE: An Approach to Refining Cross-Entropy and Contrastive Learning for Optimized Learning Fusion

    Zijun Long, Lipeng Zhuang, George Killick, Richard McCreadie, Gerardo Aragon Camarasa, Zaiqiao Meng

    Research summary · Draft — pending author review

    CLCE addresses limited generalisation in cross-entropy training and the dependence of contrastive learning on large batches and insufficient attention to hard negatives. It combines label-aware contrastive learning, cross-entropy and hard-negative mining. Experiments on twelve benchmarks show improved Top-1 accuracy in few-shot and transfer settings with less reliance on very large training batches, offering a practical approach under limited hardware resources.

    2024
  • Automation of Electric Power Systems

    Innovative Applications of Satellite Internet for Modern Power Systems

    Quan Zhou, Ke Zhou, Qingren Jin, Zhikang Shuai

    Research summary · Draft — pending author review

    This work reviews satellite internet capabilities for power systems with extensive coverage, diverse applications and demanding disaster resilience. It proposes application paths for planning, situational awareness, inspection and emergency response, examining cooperation across high-, medium- and low-altitude orbits and communication, navigation and remote sensing. A zero-trust satellite–terrestrial security architecture provides a technical framework and practical reference for subsequent engineering and application research.

    About this work

    Automation of Electric Power Systems, 2024, volume 48, issue 6, pp. 42–61. The paper covers representative applications, key technologies and security models for satellite-enabled power systems and has received more than ten citations. Supported by the National Key R&D Program of China (2023YFC3321703).

    2024

2023PUBLICATIONS

  • IEEE Transactions on Smart Grid

    Hierarchical Time-Series Assessment and Control for Transient Stability Enhancement in Islanded Microgrids

    Y. Shen, Y. Peng, Z. Shuai, Q. Zhou, L. Zhu, and Z. J. Shen

    Research summary · Draft — pending author review

    HTSAC provides hierarchical time-series assessment and control to reduce transient instability in islanded microgrids with grid-forming converters. Gated recurrent units predict early post-fault trajectories, while optimised input windows, quantile loss and emergency ride-through control enable timely intervention. Simulations and experiments on a real islanded microgrid in southern China validate stability assessment and emergency control and reduce trial-and-error design effort.

    2023
  • International Conference on Information Systems for Crisis Response and Management (ISCRAM)

    CrisisViT: A Robust Vision Transformer for Crisis Image Classification

    Zijun Long, Richard McCreadie, Muhammad Imran

    Research summary · Draft — pending author review

    CrisisViT helps emergency agencies analyse large volumes of social-media disaster imagery when manual review time is limited. It adapts vision transformers to crisis recognition and uses Incidents1M to enrich learning across disaster type, relevance, humanitarian category and damage severity tasks. Experiments on standard crisis-image benchmarks show improved classification over prior methods, with additional crisis imagery further improving accuracy for emergency information filtering and situational assessment.

    2023
03

Projects

  • Projects

    Optimising Integrated Satellite–Terrestrial Power-System Emergency Response to Extreme Disasters

    This project addresses disaster awareness, communication assurance and emergency response for the Shanghai power grid. Satellite remote sensing, UAV imagery and ground observations will identify and locate damaged facilities, assess risks and vulnerabilities, and produce risk maps and graded assessments for repairs and resource scheduling. LEO satellites, UAV relays and terrestrial networks will support emergency communication when ground infrastructure is damaged. Resource allocations and routes will consider transport conditions, grid status and task priorities to support repairs and load restoration. The methods will be integrated into an intelligent disaster dispatch and response platform for assessment, decisions, task assignment and tracking, with validation in representative disaster scenarios to improve response efficiency and resilience.

    2026
  • Projects

    Satellite-Enabled UAV Disaster Assessment and Coordinated Emergency Resource Scheduling

    Wuxi power grid project; industry-funded collaboration.

    The project addresses public network outages, unstable links, inefficient data fusion and inadequate scheduling decisions during extreme disasters. It aims to establish an engineering framework for reliable communication, intelligent disaster awareness and multi-objective emergency decisions. First, integrated space–air–ground networking and lightweight UAV satellite terminals will support remote control and reliable data return, improving link robustness and survivability. Second, multimodal data fusion will enable rapid interpretation, key-target recognition and disaster situational assessment. Third, multi-objective models and solution strategies will coordinate emergency resources according to disaster conditions and operational constraints, improving rescue deployment efficiency.

    —
  • Projects

    Rapid Disaster Information Acquisition Through Integrated Satellite Communication, Navigation and Remote Sensing

    This project addresses natural-disaster monitoring and warning through rapid information acquisition with integrated satellite communication, navigation and remote sensing. It studies satellite–terrestrial information fusion and analysis, emergency satellite-to-terminal data transmission technology and equipment, and integrated disaster monitoring. The work targets information fusion and coordinated transmission to improve monitoring and warning efficiency and accuracy.

    —
  • Projects

    Satellite Thermal Infrared Imaging and Precise Identification of Small-Scale Wildfires Affecting Power Grids

    Subtask of a National Key R&D Program project.

    The work examines noise and spatiotemporal correlations in power-grid wildfire detection and develops enhancement methods for aliased multiband infrared signals. It studies thermal-radiation edge gradients to identify small fires, fuses multiple data sources for fire-spread modelling, and develops rapid graded risk alerts. The target is satellite wildfire monitoring at 60 m × 60 m resolution under clear-sky conditions.

    —
  • Projects

    Satellite-Internet-Based Disaster Risk Warning and Emergency Response Strategies for Modern Distribution Systems

    The project studies coordination between satellite internet and modern distribution systems for disaster warning and emergency response. An integrated space–air–ground architecture will combine communication, navigation and sensing, clarifying how satellite networks support information transmission, disaster perception and response. Multiple satellite–terrestrial data sources will identify key disaster features and quantify risk factors, informing ensemble warning models for complex data. Coordinated satellite–terrestrial situational awareness and hierarchical resource allocation will address post-disaster imbalances in communication and power resources and support distribution-system recovery. A cyber–physical co-simulation platform combining physical and virtual components will validate the models and strategies to strengthen resilience.

    —
  • Projects

    Agile Hybrid Ad Hoc Networking for Satellite Terminals and Border-Management Communication Equipment

    Task within a National Key R&D Program project.

    —
04

Patents

  • Patents

    基于角度的星地融合新型配电系统风险异常识别与量化表征方法及相关装置

    周全(5)、练之洁(1)、吴梦轲(学)(2)、刘继洋(学)(3)、余会洁(学)(4)

    About this work

    在星地融合新型配电系统下,现有的总体风险异常识别与量化表征是对当前系统风险状态进行识别与量化表征,以减小电气设备故障、网络攻击等综合风险对新型配电系统的影响,但仍存在一定的局限性:(1)随着信息层面的加入,新型配电系统运行的数据维度增大,采用距离判断异常点的识别效果逐渐减弱,难以精准识别异常数据点;(2)对异常状态的量化不具体,表征方式不明显,难以直观表示系统的风险异常状态;(3)对电力-非电力等耦合因子辨识不明确,未能综合考虑电力与信息等多层面的风险因子。

    2025
  • Patents

    面向新型电力系统的业务-模型-数据联合驱动星地融合网络资源智能协同调度方法及系统

    周全(1)、仇世元(学)(2)、张轩(学)(4)、帅智康(5)

    About this work

    目前面向新型电力系统业务的通信传输资源与计算资源调度仅考虑传统地面通信网络,无法实现电力业务全面覆盖,未实现星地融合网络电力应用以及网络资源的有效利用。

    2025
  • Patents

    基于卫星互联网的输电走廊山火蔓延预测与风险评估方法及相关装置

    周全(1)、周家伟(学)(2)、练之洁(3)、帅智康(5)

    About this work

    现有的针对山火蔓延趋势预测的方法大致可以分为基于物理模型、基于统计模型以及基于概率模型的预测方法。然而此类方法只能在固定条件或历史数据的基础上进行预测,缺乏对突发环境变化的动态响应,难以满足快速变化的火灾环境需求。针对输电走廊跨越区度大、周边环境复杂多变的场景,常规的利用历史火灾数据来预测火势蔓延走向不够准确,且未充分考虑地理、电力负荷分布等人为因素对于山火蔓延的影响,缺乏对高风险区域的重点监测以及基于实时数据的动态调整自适应能力,导致预测结果实时性、准确性不足。

    2025
  • Patents

    天空地一体、通导感融合的电力系统典型灾害关键诱发因素辨识方法

    周全(1)、刘继洋(学)(2)、帅智康(5)、李炳绪(4)

    About this work

    现有的星地数据融合下新型电力系统典型灾害关键诱发因素辨识方法,仍存在如下缺陷:(1)天空地一体、通导感融合下,对新型电力系统采集的数据通常涵盖视频、图像、文本等多模态数据,而目前系统多分析单一类型的数据,例如仅限于图像或数值数据,对于结合视频、图像、文本等多模态数据的综合分析并未给予足够的重视。(2)天空地一体、通导感融合的新型电力系统灾害监测数据含有离散和连续两种数据特征,而将连续特征直接离散化处理后,会导致较大主观性和随机性,同时造成大量信息的浪费。(3)对于天空地一体、通导感融合下新型电力系统中的高风险低概率事件,如极端天气条件的设备表现,往往因数据稀少而难以进行有效分析。然而,这些罕见高危的事件对电力系统安全运行影响不容忽视。

    2025
  • Patents

    Unified Representation of Power-System Safety Risk Using Integrated Space–Air–Ground Communication, Navigation and Sensing

    周全(1)、刘继洋(学)(2)、帅智康(5)、练之洁(4)

    About this work

    在天空地一体、通导感融合的新型电力系统源-网-荷-储全流程风险态势感知架构下,电力系统是由发输变配用各领域、源网荷储各环节、技术体制各层面紧密耦合形成的有机整体,在提取新型电力系统典型灾害关键诱发因素后,构建新型电力系统安全风险态势统一表征方法将存在如下缺陷:(1)在电力设备层,大量自动化、信息化设备融入电网,系统故障波及范围增大,设备耦合关系更为复杂,天空地一体、通导感融合的新型电力系统安全风险态势统一表征的难度增大。(2)在监测模块层,天空地网络模块对电力灾害特征虽存在监测角度等差异,但采集特征间的耦合性却不能忽视。(3)在系统环节层,由传统电力系统相对分散的发输变配用单向过程,转化为高度协同且耦合性较强的源网荷储环节,某一环节出现问题对整个系统的影响不能忽视。

    2025
  • Patents

    Communication Load Modelling and Situational Awareness for Satellite–Terrestrial Modern Power Systems

    周全(1)、刘继洋(学)(2)、李炳绪(4)、帅智康(5)

    About this work

    现有技术缺点:
    (1)通信负荷运行机理不明确、时空随机性未充分考虑。现有的电力通信负荷预测大多基于统计学和历史数据进行分析,缺乏对通信负荷变化规律的深入理解和分析,导致预测结果往往存在较大误差,影响电力业务的通信保障效果。
    (2)通信负荷需求响应、时空聚合问题未充分考虑。电力系统中的通信负荷不仅受到电力负荷和设备状态的影响,还与调度、故障处理、应急响应等因素密切相关。现有方法在进行通信负荷需求响应分析时,常常面临困难。尤其在不同区域、不同时间尺度下,通信负荷的变化呈现出强烈的非线性和异质性,导致时空聚合分析变得非常困难。
    (3)通信负荷时空变化趋势未充分考虑。由于通信负荷时空特征的复杂性,现有的预测方法在提取通信负荷的时空特征时面临很大困难。特别是电力系统中的负荷突增、故障处理和灾难应急等极端情况,通信负荷的时空变化趋势往往呈现出高度的不规则性和波动性。现有的预测模型多采用传统的时间序列方法或单一的数据源,难以全面捕捉通信负荷的时空变化趋势,无法准确预测其未来的变化模式和趋势。

    2025
  • Patents

    新型电力系统星地融合通信网络建模与资源编排方法及系统

    周全(1)、文铫(学)(2)、钱健如(学)(4)、帅智康(5)

    About this work

    现有技术缺点:
    1、现有技术在支持新型电力系统多元业务方面存在明显不足,尤其表现在资源的按需分配能力上。目前,多元业务类型(如低时延、高带宽、实时监控等)对计算资源、带宽资源和链路资源的需求具有高度动态性和差异性。然而,现有技术无法根据业务需求的优先级、特性和时空分布灵活调整资源分配策略,导致资源利用效率低下,难以满足新型电力系统在高可靠性、低延迟和实时性方面的严格要求,这一缺陷制约了新型电力系统的通信网络性能优化及业务服务质量的提升。
    2、卫星高动态性导致星地融合网络拓扑频繁变化,服务节点需要频繁切换,导致服务中断和性能下降,当前的星地融合网络中,计算资源、通信资源和带宽资源等分配效率低下,现有的服务功能链部署策略未能充分考虑卫星节点的计算、内存资源和链路带宽的动态变化,导致资源利用率低。

    2025
  • Patents

    面向输电线路走廊广域监测的平急一体星地融合通信网络建模与终端设计方法及相关装置

    周全(1)、蒋鹏(学)(2)、彭辰(学)(3)、帅智康(5)

    About this work

    现有技术缺点:由于输电线路跨度长,公网往往存在较大通信盲区,数据传输能力有限,数据安全面临较大风险等问题,而搭建电力专网往往需要较大投资,且需要很好抵御输电线路沿途各类恶劣自然条件,投资与维护成本巨大。现有通信技术难以实现输电线路通信全覆盖,在线路出现故障时不能将线路的关键监测数据及时回传至数据中心,让电网工作人员采取应对措施。虽然,已有学者提出多种通信方式融合的输电线路监测通信网络,通过无线mesh接力与电力光纤结合、无线中继传输与地面蜂窝网络结合、电力专网结合北斗卫星导航等通信方式传输输电线路监测数据。但是,目前提出的监测网络仍存在以下几个缺陷:(1)虽然考虑到多种通信网络的融合,但缺乏针对卫星互联网与现有地面网络协同配合的系统研究。(2)对节点通信能力的配置与通信资源的选址研究还不够深入。(3)针对面向输电线路走廊的平急一体星地融合通信网络的可靠最优数据传输机制还存在许多研究空白。

    2025
  • Patents

    基于低轨卫星互联网的无人机空中基站智能调度方法及相关装置

    周全(5)、帅智康(1)、何可人(学)(2)、华品元(学)(3)

    About this work

    大量输电走廊地处偏远地区,普遍存在大量通信盲区,传统的地面监控和通信网络建设及维护成本高,覆盖范围有限,难以满足输电线路走廊的实时监控需求。无人机辅助通信由于其灵活的移动部署能力,为远程输电线路走廊的全面监控提供了新方案。无人机可以在走廊上空巡逻,从沿线部署的监控传感器实时收集数据,监测输电线路的状态、设备状况、极端天气的影响以及周围环境的变化。然而,偏远山区地形起伏和通信盲区对建立通信链路和数据传输构成了重大挑战。传统的地面通信基站在这些地区难以建立,无法为无人机提供稳定的远程通信服务。

    2025
  • Patents

    基于卫星互联网的模型-数据联合驱动输电线路覆冰状态感知与运行风险评估方法及相关装置

    周全(1)、张函宇(学)(2)、周斌(4)、帅智康(5)

    About this work

    输电线路所处场景复杂多变,所涉跨度大,覆冰常发生在人迹罕至的山区。采用地面网络开展覆冰监测,存在数据传输能力有限、更新频率慢与覆盖范围小等问题,且地面装置的铺设也需考虑输电线路沿途的恶劣自然条件,投资与维护困难。现有的覆冰预测方法多为依赖历史经验的传统模型,同时对气象数据的准确度要求高。而目前的气象数据多通过地面网络获取与传输,存在数据传输时效性低与准确度差的问题,无法为技术人员评估覆冰灾害风险提供准确全面的数据支撑。传统风险评估方法中,解析法难以处理系统中随机因素的影响,蒙特卡洛法计算多维问题效率又较低,风险指标构建不成体系。

    2025
  • Patents

    星地融合的自组网混合路由协议与故障自愈方法、装置、设备及存储介质

    周全(1)、蒋鹏(学)(2)、帅智康(3)

    About this work

    现有技术方案往往针对单一优化目标,缺乏能耗与时延之间灵活平衡的有效机制。链路稳定性较少考虑未来变化情况,常规路由协议或算法通常只根据当前测量值来判断链路是否可用,缺少对未来链路波动的考虑,在实际环境变化较快的网络中易出现中途断链或持续丢包,降低通信稳定性。故障切换与自愈速度慢,容错能力有限,现有的一些自组网协议在链路或节点故障时,需要进行较大范围的路由发现或泛洪才能完成切换,切换时延高且容错能力不足。算法智能化程度不够,传统路由协议(如距离向量、链路状态等)在动态场景下的优化依赖预先设定好的度量标准,对网络的历史运行数据、节点能量状况、链路可靠性统计等缺乏深入利用。很多方案在大型动态拓扑里难以自适应,很难通过“学习”来持续改进路由决策,缺少快速故障自愈和路径切换的“经验积累”机制。

    2025
  • Patents

    Unified Representation of Power-System Safety Risk Using Integrated Space–Air–Ground Communication, Navigation and Sensing

    Quan Zhou; Jiyang Liu; Chenhao Sun; Zhijie Lian; Zhikang Shuai

    About this work

    Differences in disaster observation and complex coupling between risk factors make power-system risk difficult to represent accurately. This method builds a safety-risk knowledge graph to capture relationships between equipment, events and states, combines space–air–ground observations in spatial and temporal risk models, and dynamically updates the graph and models by comparing predictions with observed risk data. It improves risk representation and prediction, supporting identification of high-risk operating areas, cascading-failure warning, prevention and emergency planning.

    2025
  • Patents

    Communication Load Modelling and Situational Awareness for Satellite–Terrestrial Modern Power Systems

    Quan Zhou; Jiyang Liu; Zhaoyue Xia; Bingxu Li; Zhikang Shuai

    About this work

    Strong temporal variation and spatial differences make communication loads in satellite–terrestrial power systems difficult to model. This method classifies communication loads, aggregates them in time and space using demand-control strategies, and applies graph and temporal convolution to extract dependencies for load prediction and situational awareness. It reduces analysis complexity and improves forecasting accuracy and timeliness, supporting network scheduling and communication quality for power services.

    2025
  • Patents

    Identification of Key Disaster Triggers in Power Systems Using Integrated Space–Air–Ground Communication, Navigation and Sensing

    Quan Zhou; Jiyang Liu; Chenhao Sun; Bingxu Li; Zhikang Shuai

    About this work

    This method addresses underused multimodal data, information loss from discretising continuous features, and difficulty identifying rare high-risk factors. It builds a disaster-event database from images, video and text, evaluates continuous and discrete features in parallel, and combines probability distributions with fuzzy-set theory to identify key triggers. The method captures rare dangerous factors and supports more comprehensive disaster warning.

    2025
  • Patents

    基于低轨卫星互联网的变电站空中隐患监测方法及系统

    周全(1)、钱健如(学)(2)、帅智康(3)、文铫(学)(4)、周家伟(学)(5)

    About this work

    变电站作为电力系统的关键基础设施,其安全运行对于保障社会经济活动至关重要。然而,变电站面临着来自无人机恶意侵入和飘挂物等意外威胁的挑战,这些威胁可能对电力系统的稳定运行造成严重影响。

    2024
  • Patents

    天地一体化电力信息网络安全加密控制方法及系统

    周全(1)、朱艳阳(学)(2)、熊点杰(学)(4)、帅智康(3)、练之洁(5)

    About this work

    随着电力信息化系统的广泛应用,针对电力信息网络的攻击事件频繁发生,各类网络攻击手段和攻击技术层出不穷,传统的安全防御方法无法应对日益智能化、协同化的网络攻击威胁。

    2024
  • Patents

    基于卫星互联网的无人机电力巡检不间断通信方法及系统

    周全(1)、蒋鹏(学)(4)、何可人(学)(2)、帅智康(3)、练之洁(5)

    About this work

    卫星互联网网络下无人机广域智慧电力巡检过程中面临的由于云衰导致的无人机通信失联问题

    2024
  • Patents

    天地一体化电力信息网络的零信任安全防护实现方法及系统

    周全(1)、沈阳(学)(2)、朱艳阳(学)(4)、帅智康(3)、练之洁(5)

    About this work

    解决现有技术电力系统安全性较差的问题

    2024
  • Patents

    天空地一体化光伏发电出力预测方法及系统

    周全(1)、文铫(学)(2)、周家伟(学)(4)、帅智康(3)

    About this work

    光伏出力受自然因素和环境因素因素影响,具有明显的间歇性和波动性,导致太阳能发电量不稳定,无法时刻满足客户的用电需求,还可能引起系统电压的波动和闪变,严重影响电力调度决策以及需求响应管理。基于此,亟需开展高精度的光伏功率预测,以有效降低光伏并网时对电网的影响。

    2024
  • Patents

    一种面向新型配电系统灾后场景的通导融合无人机功率优化分配方法

    周全(1)、仇世元(学)(2)、帅智康(4)

    About this work

    现有传统配电通信系统在自然灾害下较易损毁,无法应对极端事件,造成新型配电系统灾后场景缺乏可靠通信保障。同时,新型配电系统场景复杂,高精度的定位服务能够帮助抢修人员快速定位损毁节点。因此,亟需研究一种面向新型配电系统场景的灾后通信与导航技术,使其在极端自然灾害导致电力通信终端或通讯设备损坏时能够迅速响应,协助新型配电系统进行灾后恢复重构。

    2024
  • Patents

    基于卫星互联网的电力山火防灾应急资源优化调度方法及系统

    周全(1)、张轩(学)(2)、仇世元(学)(3)、帅智康(4)

    About this work

    目前,虽然已有部分电网企业采用地面无线通信技术和有限的卫星通信手段进行应急管理,但在火灾规模扩大、地理位置复杂或地面设施损毁严重的情况下,传统通信手段往往难以满足全局掌控和精确调度的需求,这给应急资源的快速调度带来了前所未有的挑战。基于此,本专利提供了一种基于卫星互联网的电力山火防灾应急资源优化调度方法,该方法不仅能从多维度实现火灾的监测和信息的传输,还可以提高灾区的通信资源利用率并减少电力山火的整体灭火时间。

    2024
  • Patents

    天空地多模态数据融合的超短期光伏功率预测方法及系统

    周全(1)、周家伟(学)(2)、刘继洋(学)(3)、帅智康(4)

    About this work

    随着全球能源结构向可再生能源转型,光伏发电因其清洁、高效的特性受到广泛关注。然而天气、地理环境等因素的复杂影响,光伏功率输出具有较大的不确定性。因此,通过实时监测云层的动态变化,尤其是云层的覆盖率、移动速度、厚度等关键因素,利用云图数据来预测光伏电站在短时间内的功率输出,从而实现高时效和高空间分辨率的预测。

    2024
  • Patents

    一种基于低轨卫星互联网的电力廊道空中隐患监测方法及相关装置

    周全(1)、钱健如(学)(2)、帅智康(3)、文铫(学)(4)

    About this work

    现有的空中隐患监测方法主要包括地面摄像头监控、无人机巡检和雷达监测等方式。其中,地面摄像头监控适用于固定位置的连续监测,但其监测范围有限,难以覆盖偏远和复杂地形区域。无人机巡检具有灵活性,可用于特定区域的隐患排查,但由于电池续航限制,巡检频率较低,难以实现长时间连续监测。雷达监测能够在一定程度上检测空中障碍物,适用于天气状况较好的环境,但在恶劣气候条件下其性能显著下降。此外,这些方法对快速移动的空中隐患响应不够及时,难以提供有效的预警和应急处理。

    2024
05

Other

No public records yet.