River network flood real-time prediction method and system based on space-time diagram convolution

By constructing a river network flood prediction method based on spatiotemporal graph convolution, the problems of difficult modeling of river network topology and low computational efficiency in existing technologies are solved, achieving efficient and accurate flood forecasting with physical interpretability and adaptive capabilities, meeting the needs of real-time flood forecasting.

CN122286609APending Publication Date: 2026-06-26HEBEI UNIV OF ENG
0 Cites 4 Cited by

Patent Information

Application Number
CN202610063555.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-01-19
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Existing flood forecasting methods are difficult to effectively model the irregular spatial topology and hydraulic connections of river networks, and have low computational efficiency and complex parameter calibration, making it difficult to meet the requirements for real-time flood forecasting with high spatiotemporal resolution.

Method used

A river network flood prediction method based on spatiotemporal graph convolution is adopted. By constructing an initial river network topology map, hydrological stations are used as nodes and river connections are used as edges. A weighted graph is constructed by combining physical hydrological attribute information. A dynamic graph learning module is introduced to train the spatiotemporal graph convolutional network model. Features are extracted using spatial and temporal convolutional layers, and a physical constraint loss function is added.

Benefits of technology

It achieves efficient and accurate flood forecasting, possesses physical interpretability and adaptive capabilities, and can complete multi-site forecasts within seconds, meeting the needs of flood control emergency response and improving forecast accuracy and stability.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to the technical field of hydrological forecasting and artificial intelligence crossing, in particular to a river network flood real-time prediction method and system based on space-time diagram convolution. The method comprises the following steps: acquiring historical hydrological data of river network monitoring points, and constructing an initial river network topological graph according to a river channel connection relationship; based on the graph and historical data, a space-time graph convolutional network prediction model is constructed and trained, the model learns a spatial dependency relationship between sites through a space graph convolutional layer and learns a time sequence change rule of each site through a time convolutional layer, and physical constraint loss is introduced in training; and finally, performing real-time prediction by utilizing the trained model. The invention also discloses a corresponding prediction system. According to the method, river network physical topology and hydrological priori knowledge are deeply embedded into a deep learning framework, through dynamic graph learning and physical constraint, the model has both physical rationality and data adaptive ability, and rapid, accurate and reliable real-time simulation forecasting of the complex river network flood process is realized.
Need to check novelty before this filing date? Find Prior Art