A Typhoon Formation Prediction Method Based on Causal Perception Diffusion Network
By constructing a prediction model based on a causal perception diffusion network, the uncertainty and physical consistency issues in existing typhoon formation predictions are resolved, enabling accurate prediction of the time and location of typhoon formation and meeting real-time operational needs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG UNIV OF TECH
- Filing Date
- 2026-02-12
- Publication Date
- 2026-07-03
AI Technical Summary
Existing typhoon generation and prediction methods cannot accurately predict the time and location of formation, and lack physical consistency. Existing generation models rely on high-quality data and cannot meet real-time operational needs.
A predictive model based on a causal perception diffusion network is constructed. An adaptive knowledge distillation framework is designed by combining a teacher-student distillation architecture with multimodal physical variable constraints to achieve spatiotemporal modeling of the physical consistency of cloud evolution. Key physical driving factors are screened by using a causal discovery mechanism.
It achieves accurate prediction of the entire typhoon formation process, and can extract feature changes from the formation process of typhoon precursors, reduce uncertainty, and meet the needs of real-time prediction.
Smart Images

Figure CN122336360A_ABST