| 161 | 0 | 175 |
| 下载次数 | 被引频次 | 阅读次数 |
国家水网建设正从工程连通和数字化感知转向精细化调控，数字孪生水利建设也在持续提升预报、预警、预演、预案等业务支撑能力。然而，从辅助决策转向自主决策和闭环控制，水网系统需应对气象水文不确定性、水动力过程非线性、工程系统强耦合、调度规程难以结构化表达，以及跨业务目标难以协同等问题。对此，本文提出融合水网大模型与机理智能体的自主运行水网架构：以水网大模型作为认知决策中枢，以机理智能体作为专业执行单元，面向防洪调度、水资源配置、输配水控制、生态调控和水环境校核等任务，完成水文水动力计算、优化调度、控制指令生成和约束校核。认知决策层、专业执行层和反馈校核层通过标准化接口与状态反馈形成闭环协同，贯通业务需求识别、方案生成、工程控制和运行反馈。本文进一步提出在环测试验证体系和自主运行能力分级判定框架，为国家水网由辅助决策向自主运行演进提供架构参考。
Abstract：The development of national water networks has entered a stage characterized by refined regulation, moving beyond engineering connectivity and digital perception. Simultaneously, digital twin water conservancy continues to enhance its support for forecasting, early warning, scenario simulation, and contingency planning. However, the transition from decision support to autonomous decision-making and closed-loop control requires water network systems to address several challenges: meteorological and hydrological uncertainty, hydrodynamic nonlinearity, strong coupling among engineering components, difficulty in formalizing operating rules, and coordination across multiple operational objectives. To address these challenges, this paper proposed an autonomous water network operation architecture that integrated a water-network large model with mechanism-based agents. In this architecture, the water-network large model served as the cognitive decisionmaking hub, while mechanism-based agents acted as professional execution units for tasks such as flood control operation, water resources allocation, water conveyance and distribution control, ecological regulation, and water environment verification. These agents undertook hydrodynamic computation, optimal operation, control command generation, and constraint checking. The cognitive decision-making, professional execution, and feedback verification layers formed a closed-loop coordination through standardized interfaces and state feedback, linking business demand recognition, scheme generation, engineering control, and operational feedback. The paper further proposed an X-in-the-loop testing and verification system and a capability grading framework for autonomous operation, providing an architectural reference for the evolution of national water networks from decision support toward autonomous operation.
[1]中共中央，国务院.国家水网建设规划纲要[A].2023.
[2]轩玮，李博远，汪习文.深入贯彻落实党的十九届六中全会精神推动新阶段水利规划计划工作高质量发展——访水利部规划计划司司长石春先[J].中国水利，2021(24):1-3.
[3]成建国.数字孪生水网建设思路初探[J].中国水利，2022(20):18-22+10.
[4]李钟宁，刘洁.数字孪生技术赋能水利工程建设管理路径[J].中国科技信息，2025(18):152-154.
[5]BHANDARI P,CREIGHTON D,GONG J,et al.Evolution of cyber-physical-human water systems:Challenges and gaps[J]. Technological Forecasting&Social Change,2023,191:122540.
[6]FU G,JIN Y,SUN S,et al. The role of deep learning in urban water management:A critical review[J].Water Research,2022,223:118973.
[7]WANG J,FU G,SAVIC D. Leveraging large language models for automating water distribution network optimization[J]. Water Research,2026,288:124536.
[8]林圣德，管华明. AI大模型在水利知识库构建中的应用研究[J].水利规划与设计，2025(12):102-105+131+137.
[9]WANG L,MA C,FENG X,et al. A survey on large language model based autonomous agents[J]. Frontiers of Computer Science,2024,18(6):186345.
[10]NEGM A,MA X,AGGIDIS G. Deep reinforcement learning challenges and opportunities for urban water systems[J]. Water Research,2024,253:121145.
[11]HUNG F,YANG Y C E. Assessing adaptive irrigation impacts on water scarcity in nonstationary environments:A multi-agent reinforcement learning approach[J]. Water Resources Research,2021,57(9):e2020WR029262.
[12]ZHANG Z,TIAN W,LIAO Z. Towards coordinated and robust real-time control:A decentralized a p p r o a c h f o r c o m b i n e d s e w e r o v e r f l o w a n d urban flooding reduction based on multi-agent reinforcement learning[J]. Water Research,2023,229:119498.
[13]ADEDEJI K B,HAMAM Y. Cyber-physical systems for water supply network management[J].Sustainability,2020,12(22):9555.
[14]DENG L,GUO S,YIN J,et al. Multi-objective optimization of water resources allocation in Han River basin(China)integrating efficiency,equity and sustainability[J]. Scientific Reports,2022,12:798.
[15]XU B,SUN Y,HUANG X,et al. Scenariobased multiobjective robust optimization and decision-making framework for optimal operation of a cascade hydropower system under multiple uncertainties[J]. Water Resources Research,2022,58(4):e2021WR030965.
[16]YU L,WU X,WU S,et al. Multi-objective optimal operation of cascade hydropower plants considering ecological flow under different ecological conditions[J]. Journal of Hydrology,2021,601:126599.
[17]KONG L,QUAN J,YANG Q,et al. Automatic control of the middle route project for Southto-North Water Transfer based on linear model predictive control algorithm[J]. Water,2019,11(9):1873.
[18]VAN OVERLOOP P J,CLEMMENS A J,STRAND R J,et al. Real-time implementation of model predictive control on Maricopa-Stanfield Irrigation and Drainage District’s WM Canal[J].Journal of Irrigation and Drainage Engineering,2010,136(11):747-756.
[19]CASTELLETTI A,FICCHI A,COMINOLA A,et al. Model predictive control of water resources systems:A review and research agenda[J]. Annual Reviews in Control,2023,55:442-465.
[20]LITRICO X,FROMION V. Modeling and control of hydrosystems[M]. London:Springer,2009.
[21]F E N G S ,S U N H ,YA N X ,e t a l. D e n s e reinforcement learning for safety validation of autonomous vehicles[J]. Nature,2023,615(7953):620-627.
[22]WU R,WANG R,HAO J,et al. Multiobjective multihydropower reservoir operation optimization w i t h t r a n s f o r m e r-b a s e d d e e p r e i n f o r c e m e n t learning[J]. Journal of Hydrology,2024,632:130904.
[23]TIAN W,XIN K,ZHANG Z,et al. Flooding mitigation through safe and trustworthy reinforcement learning[J]. Journal of Hydrology,2023,620:129435.
[24]PESANTEZ J E,ALGHAMDI F,SABU S,et al.Using a digital twin to explore water infrastructure impacts during the COVID-19 pandemic[J].Sustainable Cities and Society,2022,77:103520.
[25]MESAROVIC M D,MACKO D,TAKAHARA Y.Theory of hierarchical,multilevel systems[M]. New York:Academic Press,1970.
[26]雷晓辉，张峥，苏承国，等.自主运行智能水网的在环测试体系[J].南水北调与水利科技（中英文），2025,23(4):787-793.
[27]SOMERS R J,DOUTHWAITE J A,WAGG D J,et al. Digital-twin-based testing for cyberphysical systems:A systematic literature review[J].Information and Software Technology,2023,156:107145.
[28]何立新，史博阳，张峥，等.引调水渠道控制系统硬件在环测试平台设计与实现[J].南水北调与水利科技（中英文），2025,23(5):1036-1046.
基本信息:
中图分类号:TV213.4
引用信息:
[1]雷晓辉,陈凯歌.融合水网大模型与机理智能体：构建自主运行水网新范式[J].中国水利,2026,No.1035(09):1-10.
基金信息:
国家自然科学基金(U25A20357)
2026-05-19
2026-05-19
2026-05-19