A Deep Learning-Based Short-Term Targeted Early Warning Method for Rainstorm Disasters at New Energy Power Stations
By using deep learning-based methods, we can identify strong convective cloud systems at new energy power stations and correct rainfall patterns, thus solving the problems of accuracy, real-time updates, and differentiation in rainstorm disaster early warning for new energy power station clusters and achieving efficient targeted early warning.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- METEOROLOGICAL SERVICE CENT OF GUANGXI ZHUANG AUTONOMOUS REGION
- Filing Date
- 2026-04-21
- Publication Date
- 2026-06-30
AI Technical Summary
Existing technologies are insufficient to achieve accurate, real-time, and differentiated rainstorm disaster early warnings in new energy power plant clusters. In particular, when facing a large number of dispersed power plant clusters, they suffer from low efficiency, inconsistent standards, and slow response. Furthermore, they lack a dynamic correction mechanism for the rapid evolution of local severe convection, resulting in insufficient timeliness of early warnings.
Using a deep learning-based approach, early warning units are divided and a baseline analysis grid is constructed by acquiring spatial boundary data of new energy power stations and radar monitoring grid parameters. Combined with historical observations and expert rules, strong convective cloud systems are identified, and precipitation field correction is performed through a multi-branch fusion network to generate targeted early warning information.
It enables differentiated, real-time, and precise early warning for new energy power stations, and can directly deliver early warning results to specific early warning units, meeting the minute-level response requirements of new energy power stations for short-term heavy rainfall, and improving the hit rate and timeliness of early warnings.
Smart Images

Figure CN122313641A_ABST