中国水利

我国数智技术的水利领域应用从试点探索逐步进入体系化建设新阶段，深度嵌入日常管理、运行决策、工程控制的核心环节，在显著提升治理效能的同时，也使系统运行面临风险隐患。为系统识别数智水利应用风险，保障系统安全、可靠、可信、可控运行，本文基于数字孪生水利架构和运行原理，分析数智水利应用对数据、模型、算力、知识、网络及集控的六大强依赖特性，揭示强依赖特性所衍生的数据风险、模型风险、算力风险、知识风险、网络风险、集控风险等复合型挑战，并从制度构建、技术防护、人才强基、韧性提升四个维度提出立体化风险防控体系对策，涵盖国家战略与综合安全框架制定、法律法规与标准规范健全、协同监管与第三方审计机制建立、数据全生命周期治理、可信AI与模型综合治理、自主弹性算力与韧性网络建设、组织重塑与流程再造、数智人才强基、人机协同与知识管理、联邦式混合韧性架构设计、常态化攻防演练、极端情景应急预案制定等关键举措，为防范化解数智水利系统性风险提供技术支撑。

As the application of digital-intelligent technologies in China's water conservancy sector gradually transitions from pilot exploration to a new stage of systematic development, these technologies are becoming deeply embedded in the critical junctures of daily management, operational decision-making, and engineering control. While this integration substantially enhances governance effciency, it also exposes system operations to emerging risks and vulnerabilities. To systematically identify the risks inherent in digital-intelligent water conservancy applications and ensure their safe, reliable, trustworthy, and controllable operation, this study, based on the architecture and operational principles of digital twin water conservancy, identifies six strong dependency characteristics of these applications—namely, on data, models, computing power, networks, knowledge, and centralized control. The study further reveals the composite challenges arising from these dependencies, including data risks, model risks, computing risks, knowledge risks, network risks, and centralized-control risks. In response, a multi-dimensional risk prevention and control framework is proposed from four interconnected dimensions—institutional development, technical safeguards, capacity building, and resilience enhancement. Key measures encompass the formulation of national strategies and comprehensive security frameworks, the improvement of laws, regulations, and standards, the establishment of collaborative oversight and third-party audit mechanisms, full life-cycle data governance, trusted AI and comprehensive model governance, the development of autonomous and resilient computing power and network infrastructure, organizational restructuring and process re-engineering, the cultivation of digital-intelligent talent, humanmachine collaboration and knowledge management, the design of federal hybrid resilient architectures, routine red team vs. blue team cyber defense exercises, and the development of extreme scenario emergency plans. This research provides technical support for preventing and mitigating systemic risks in digital-intelligent water conservancy applications.