Abstract: Aiming at extending the early warning period of flash flood and improving the accuracy of early warning, through the comprehensive application of parallel cloud computing, big data, artificial intelligence and mobile internet technologies; a flash flood forecasting and early warning cloud platform was constructed. The platform integrates the distributed hydrological model of the river basin and the measured and forecast rainfall information, and drives the simulation of flash floods in all river sections in real time, realizing flood forecasting of small basins and early warning of flash flood based on water level or flow threshold. Combined with the early warning of flash flood based on forecasting rainfall and monitoring rainfall or water level, the platform realizes multi-stage progressive flash flood prediction and early warning. By configuring access permissions for multi-level users, cloud-based deployment and multi-level users of the platform are realized. Taking the application of two river basins in Fujian province as example, the early warning effect of the platform is verified in terms of early warning period and accuracy.