Abstract: Scientific flood control contingency plans serve as a crucial foundation for emergency management of flood disasters.However, urban flood control contingency plans continue to encounter challenges, including insufficient timeliness and difficulty in adapting to dynamic changes. The present paper puts forward a methodology for the systematic formulation and responsive generation of urban flood emergency response strategies, underpinned by knowledge graphs. The integration of deep learning with the Ratcliff/Obershelp fuzzy matching mechanism has enabled the development of a hybrid entity recognition model that facilitates rapid identification and extraction of entities and relationships in the flood control domain. A knowledge graph for urban flood emergency response plans is constructed using the Neo4j graph database, and a hybrid reasoning method combining case-based and rule-based reasoning is adopted to dynamically generate flood emergency response plans. The findings demonstrate that the entity recognition model attains an accuracy rate of 98.5%, and the constructed knowledge graph facilitates associative queries, semantic reasoning, and visual presentation of flood control emergency plan knowledge. This research contributes to the improvement of the efficiency of urban flood control emergency plan generation and provides a reference for urban flood mitigation decision support.