Flood forecasting method and system based on multi-scale memory and deep learning in data-scarce areas
By constructing a flood forecasting method based on multi-scale memory and deep learning, and combining graph convolutional networks and long short-term memory networks, the problem of not considering spatiotemporal characteristics in flood forecasting is solved, and high-precision flood forecasting is achieved for areas with scarce data.
Patent Information
- Application Number
- CN202510893633.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-30
- Publication Date
- 2026-07-24
- Estimated Expiration
- 2045-06-30
AI Technical Summary
Existing flood prediction models mainly handle missing values from a time perspective, failing to fully consider the complex spatiotemporal characteristics of flood runoff data, and their prediction accuracy needs further improvement.
A method based on multi-scale memory and deep learning is adopted to construct single-point spatiotemporal features by obtaining the temporal and spatial correlation of missing values. The generated values of flood runoff are generated using a multi-scale memory network, and flood forecasting is performed by combining graph convolutional networks and long short-term memory networks.
It improves the accuracy of missing value handling and prediction precision, better captures the spatiotemporal relationships between hydrological stations, enhances the performance and reliability of prediction models, and improves the accuracy of flood prediction in areas with scarce data.