A method and system for downscaling nearshore extreme wind fields based on sparse physical supervision
By using a sparsely physically supervised hydrodynamic mapping operator network, the computational bottleneck and physical inconsistency problem in the reconstruction of complex coastal wind fields are solved, achieving efficient and accurate reconstruction of extreme wind fields and providing high-fidelity data support for offshore wind power and coastal engineering.
Patent Information
- Application Number
- CN202610731930.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-05-26
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies for long-term wind field surveys and risk reassessments in complex coastal zones suffer from problems such as smoothing masking true extreme values in coarse-grid reanalysis data, computational bottlenecks in high-precision numerical simulations, and physical inconsistencies in purely data-driven AI. These issues make it difficult to achieve efficient and physically consistent extreme wind field reconstruction with limited high-fidelity numerical simulation data.
A sparse physical supervision method for downscaling nearshore extreme wind fields is adopted. By constructing a hydrodynamic mapping operator network, a three-stage progressive training is used, including supervised base training, teacher-guided physical fine-tuning, and large-scale semi-supervised generalization. Combined with physical constraints such as kinetic spectrum matching, wind-pressure coupling, and friction inflow angle, high-resolution microscale nearshore wind and pressure fields are reconstructed.
It achieves efficient and physically consistent extreme wind field reconstruction, breaks through the computing power bottleneck of traditional numerical simulation, avoids false artifacts in pure data-driven models, significantly corrects the underestimation of extreme wind speeds in global reanalysis data, and provides high-fidelity data support for offshore wind power site selection and coastal engineering disaster prevention assessment.
Smart Images
Figure CN122413845A_ABST