Generación de un Modelo de Orden Reducido para el Cálculo Detallado del Viento en entornos con Orografía Compleja

Milla Val, Jaime
Montañés Bernal, Carlos (dir.)

Mateo Collazos, Pedro (ponente)

Universidad de Zaragoza, CIEN, 2020

Máster Universitario en Modelización e Investigación Matemática, Estadística y Computación

Resumen: This Master’s Thesis aims to develop a methodology to build a reduced order model (ROM) from numerical weather prediction simulations (NWP) to obtain equivalent results to those given by computational fluid dynamics (CFD) technics.
The ROM is built by fitting supervised learning regressor models to CFD products from NWP results.
The proposed ROM is used over a complex terrain located in the Pyrenees mountains of the Aragon (Spain) region.
The obtained results capture fairly well the general trend of the test data.
The $85\%$ percentile of absolute errors in the horizontal components of the wind is $\sim2$~[m/s] while for the vertical one is $\sim0.8$~[m/s].
These results seem to be strongly affected by altitude.
The numerical weather predictions simulations take around $2.5$~min while CFD require $\sim40$~min.
Once the ROM is trained, using it is nearly instantaneous so the computational advantage of using this methodology is huge.

Tipo de Trabajo Académico: Trabajo Fin de Master