Scale-resolving simulations and optimisation of wind farms in complex terrain
File(s)
Author(s)
Jané-Ippel, Christian
Type
Thesis
Abstract
Wind energy is a cornerstone of the global transition to net zero, yet wind farm efficiency is often constrained by wake interactions and complex terrain. Optimising turbine layouts to maximise power capture is therefore critical. Conventional methods, such as analytical wake models or steady RANS, are efficient in flat terrain but fail to resolve the unsteady turbulence and terrain–wake coupling dominating hilly landscapes. Large-eddy simulation (LES) provides the necessary fidelity, but computational costs have traditionally limited it to validation. This thesis addresses this gap by developing a framework coupling LES with Bayesian Optimisation (BO) for wind-farm layout optimisation in complex terrain.
Methodologically, the high-order solver WInc3D is extended using immersed boundary and wall-stress modelling to simulate realistic onshore farms. In parallel, BO is embedded within an HPC workflow, introducing physics-informed seeding strategies to efficiently initialise optimisation in high-dimensional design spaces.
Validation against wind-tunnel experiments and SCADA data confirms the solver reproduces boundary-layer inflows, terrain-induced flow, and turbine performance. A two-turbine optimisation around a Gaussian hill provides proof of concept, demonstrating that terrain features can be exploited through hub-height or wake steering. Finally, the framework is scaled to a 16-turbine farm in synthetic three-dimensional terrain. Here, physics-informed BO proves essential: while black-box BO struggles, seeding with flow-based layouts enables tractable optimisation, revealing consistent exploitation of ridges and avoidance of recirculation zones.
This thesis demonstrates that LES–BO is feasible for wind-farm design in complex terrain, with typical simulations taking one to four hours. It advances methodology by integrating high-fidelity simulation with optimisation, extending the understanding of mechanisms like ridge acceleration, recirculation, wake deflection, and valley sheltering. Ultimately, it establishes LES as a design-oriented framework for next-generation wind-farm optimisation.
Methodologically, the high-order solver WInc3D is extended using immersed boundary and wall-stress modelling to simulate realistic onshore farms. In parallel, BO is embedded within an HPC workflow, introducing physics-informed seeding strategies to efficiently initialise optimisation in high-dimensional design spaces.
Validation against wind-tunnel experiments and SCADA data confirms the solver reproduces boundary-layer inflows, terrain-induced flow, and turbine performance. A two-turbine optimisation around a Gaussian hill provides proof of concept, demonstrating that terrain features can be exploited through hub-height or wake steering. Finally, the framework is scaled to a 16-turbine farm in synthetic three-dimensional terrain. Here, physics-informed BO proves essential: while black-box BO struggles, seeding with flow-based layouts enables tractable optimisation, revealing consistent exploitation of ridges and avoidance of recirculation zones.
This thesis demonstrates that LES–BO is feasible for wind-farm design in complex terrain, with typical simulations taking one to four hours. It advances methodology by integrating high-fidelity simulation with optimisation, extending the understanding of mechanisms like ridge acceleration, recirculation, wake deflection, and valley sheltering. Ultimately, it establishes LES as a design-oriented framework for next-generation wind-farm optimisation.
Version
Open Access
Date Issued
2025-10-01
Date Awarded
2026-03-01
Copyright Statement
Attribution-NonCommercial 4.0 International Licence (CC BY-NC)
License URL
Advisor
Laizet, Sylvain
Palacios, Rafael
Sponsor
Natural Environment Research Council (Great Britain)
Imperial College London
Grant Number
NE/S007415/1
Publisher Department
Department of Aeronautics
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)