Abstract: The innovation of information technology provides significant opportunities for the development of the water industry, but the application of artificial intelligence algorithms in flood forecasting in lake-plain river network areas faces certain difficulties due to the insufficient historical measured data. Taking the Yundong Enclosure in Wuxi City of Jiangsu Province as the research area, this paper elaborates on the implementation path of intelligent forecasting and dispatching for urban floods and waterlogging. It utilizes hydrological and hydrodynamic models to generate dispatching plan sets, and employs Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNN) to establish intelligent forecasting methods for water levels at key points in the urban river network and rapid calculation methods for urban waterlogging risk thresholds. Guided by the FEDE(forecasting, early warning, drilling and emergancy plan), a smart flood forecasting and dispatching system is designed to integrate professional and intelligent models to serve business scenario applications. The development of system functions provides practical insights for the integration, intelligence, and operationalization of flood forecasting and dispatching in coastal Taihu Lake cities.