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.

CN122336360APending Publication Date: 2026-07-03ZHEJIANG UNIV OF TECH
View PDF 0 Cites 0 Cited by

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

Technical Problem

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.

Method used

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.

Benefits of technology

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.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122336360A_ABST
    Figure CN122336360A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of meteorological disaster forecasting and discloses a typhoon formation prediction method based on causal perception diffusion network. By introducing multimodal physical variables to constrain the diffusion generation process and combining causal discovery mechanism to screen key physical driving factors, it realizes physical consistency spatiotemporal modeling of the complex process of cloud evolution. At the same time, an adaptive knowledge distillation framework is designed to overcome the problem of missing ERA5 data in real-time prediction, thereby realizing accurate prediction of the entire typhoon formation process.
Need to check novelty before this filing date? Find Prior Art