Modelling of wind flow over complex terrain using OpenFoam

Abstract

Vattenfall is pursuing a massive expansion of wind power capacity in the following years. The most suitable emplacements for such purpose are already taken so sites which are not so favorable but still economically feasible are to be considered. In order to assess the viability of the complex terrain locations onsite measurements are taken as well as a CFD tools employment is convenient since saves time and cost relative to a complete field measurement program. The current commercial codes are considered somewhat over-priced and not sufficiently flexible when it comes to user-defined models, consequently, it is decided to develop the open source OpenFoam CFD tool. The present thesis project is inside a larger project with which Vattenfall expects to integrate by 2011 OpenFoam CFD tool and be able to make wind power production accurate predictions over complex terrain. OpenFoam CFD tool was validated using two different cases: a flat plane and an axisymmetric plane getting results which strongly resembled to Fluent commercial code results. New wall functions including roughness were added to OpenFoams current version (OF-1.5.) in order to be able to establish non-uniform variable roughness along the domain. In addition, a pre-processing tool was developed (roughnessToFoam) which permits an easy addition of the roughness data arising from map format file. The implemented wall functions were validated and turned out to work even better than the commercial code Fluent. Coriolis and gravitational forces were successfully implemented being validated with theoretical formulation. It was appreciated that coriolis force can have a considerable influence when we come to domains which can span up to 20 km at high latitude places, therefore, it is considered that is necessary to take it into account. Eventually, Openfoams pre-processing tool SnappyHexMesh was evaluated in order to analyze whether is useful or not to mesh complex terrains. Two cases were tried to be meshed and whereas the first one (bump case) was correctly meshed, when trying to mesh the second (Askervein hill) problems arose if many surface layers were wanted to be added. It was concluded that more research is needed in order to get the most of the meshing tool, since the big amount of variable needed to be set make it quite difficult to control correctly the process.