A distributed flood forecasting model based on coupling of physical mechanism and neural network
By constructing a correspondence between watershed grids and neural networks in the flood forecasting model and explicitly incorporating physical mechanisms, the problem of insufficient ability of existing models to reflect dynamic changes in hydrological processes and spatial generalization is solved, thus achieving high-precision and interpretable flood forecasting.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-18
- Publication Date
- 2026-06-26
AI Technical Summary
Existing flood forecasting models are insufficient in reflecting the dynamic changes of hydrological processes and in terms of spatial generalization ability. Physical mechanism models lack clear physical constraints, data-driven models are prone to overfitting, and the physical consistency and practical application effect of coupled models need to be improved.
By establishing a one-to-one correspondence between watershed grid units and neural network units, a flood forecasting neural network consistent with the watershed spatial structure is constructed. Physical mechanisms are explicitly integrated, and by combining a distributed hydrological model and a neural network, hydrological model parameters are dynamically calculated to achieve adaptive adjustment of runoff generation and confluence processes. The model parameters are then optimized through a backpropagation algorithm.
It improves the accuracy and spatial generalization ability of flood forecasting, enhances the interpretability of the model, and improves the simulation capability and engineering applicability under complex hydrological conditions.
Smart Images

Figure CN122287699A_ABST