Numerical Simulation of Low-Level Wind Shear Using CFD and LSTM Technology Based on the WRF Model
In an effort to elevate the precision of low-level wind shear forecasting， this paper amalgamates European Centre for Medium-Range Weather Forecasts （ECMWF） fifth-generation reanalysis data （ERA5） and National Centers for Environmental Prediction Final Operational Global Analysis （FNL） reanalysis da...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | zho |
| Published: |
Science Press, PR China
2025-04-01
|
| Series: | Gaoyuan qixiang |
| Subjects: | |
| Online Access: | http://www.gyqx.ac.cn/EN/10.7522/j.issn.1000-0534.2024.00119 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In an effort to elevate the precision of low-level wind shear forecasting， this paper amalgamates European Centre for Medium-Range Weather Forecasts （ECMWF） fifth-generation reanalysis data （ERA5） and National Centers for Environmental Prediction Final Operational Global Analysis （FNL） reanalysis data， high-resolution Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Digital Elevation Model （ASTER GDEM） terrain data， and real-time observational data from Lanzhou Zhongchuan Airport.It employs the Weather Research and Forecasting Model （WRF）， WRF integrated with Computational Fluid Dynamics （CFD）， and Long Short-Term Memory （LSTM） neural network methods to simulate and analyze two wind shear events at Lanzhou Zhongchuan Airport on April 15-16， 2021.The findings reveal that： （1） within grids smaller than 1 kilometer utilizing Large Eddy Simulation （LES）， the WRF model demonstrates superior performance in wind speed simulation for individual stations， yet it falls short when compared to the WRF model combined with Computational Fluid Dynamics （CFD） models in simulating near-surface horizontal wind field wind speeds； （2） concerning the simulation of two low-level wind shears encountered during aircraft landing， both Weather Research and Forecasting Model - Large Eddy Simulation （WRF-LES） and Weather Research and Forecasting Model - Computational Fluid Dynamics （WRF-CFD） models are capable of simulating the first wind shear， however， the second appears to be influenced by the potentially lower wind speed data input into the models， with neither model achieving the threshold for wind speed difference， necessitating further validation in future work； （3） under low wind speed conditions （6 meters per second）， the LSTM-based single-variable wind speed prediction model maintains an average absolute error of approximately 0.59 meters per second， effectively capturing the nonlinear relationship of wind speed changes under various terrain and circulation background conditions.Despite being constrained by WRF errors and incomplete observational elements， multi-variable wind speed prediction can achieve wind speed forecasting with higher computational efficiency and generalization capabilities while ensuring that the average absolute percentage error is less than 6.60%.This paper not only verifies the differences between WRF-CFD and WRF-LES coupling schemes in wind field and low-level wind shear forecasting but also explores the feasibility and accuracy of LSTM-based wind speed prediction， aspiring to offer new perspectives and methods for enhancing wind field simulation accuracy and reducing the time required for detailed wind field simulation. |
|---|---|
| ISSN: | 1000-0534 |