Abstract: The traditional water network business management model has problems such as decision-making lag, inconvenient comparison of multiple schemes, and insufficient quantitative support for risk assessment in scenarios such as supply-demand balance analysis, cross-regional water transfer, and sudden water and drought disaster disposal. To address this issue, with the goal of "knowledge driven, model-enabled", research and development of key technologies and systems for intelligent scheduling of water networks will be carried out. The research relies on knowledge service engines, large language models, and intelligent agent application systems to integrate multi-source hydrological monitoring data with water conservancy control policy requirements. It focuses on three technical breakthroughs, including water conservancy knowledge enhancement big model technology based on Retrieval Enhanced Generation(RAG), multi-source water network data fusion technology, and multi-agent modular generation technology. The system is built according to a three-tier architecture to create an integrated water network intelligent service system, covering functional modules such as intelligent analysis of water supply and demand balance,intelligent Q&A of water network business, intelligent generation of scheduling plans, and visualization of decision support. This system can assist in generating standardized annual and monthly water network scheduling plans, quickly matching historical operating conditions and disposal cases, and generating targeted decision recommendations, effectively shortening the decision-making process. Practical applications have shown that this system significantly improves the decision-making efficiency and emergency response speed of water network scheduling, achieving efficient reuse of water network business experience. The research results can provide reference and guidance for the construction of digital twin water networks and the intelligent operation and management of water networks.