Computational fluid dynamics modelling of wind turbine wake losses in large offshore wind farms, incorporating atmospheric stability

Offshore wind power generation is projected to be the United Kingdom’s largest contributor to the European Union’s 2020 renewable energy target, with large numbers of wind turbines clustered into wind farms with capacities comparable to fossil fuelled power stations. The degree of power loss caused by the wake affected region behind each turbine is known to vary under different atmospheric stability conditions. Accurately predicting these losses for a variety of likely scenarios before new farms are built can significantly reduce the financial risk of private investment. The aim of this work was to investigate the structure of the offshore atmosphere and incorporate the findings relating to atmospheric stability into Computational Fluid Dynamics (CFD) simulations of large offshore wind farms to reduce financial investment risk in non-neutral stability conditions. This work incorporates three meteorologically established methods of calculating stability conditions into CFD simulations of large offshore wind farms using the Monin-Obukhov Similarity Theory (MOST). As MOST ideally requires meteorological parameters measured on-site using a mast for extended periods of time to obtain even a small collection of validation data, alternative methods of describing atmospheric conditions and corresponding wake behaviour are investigated which only require data obtainable by LiDAR. This has the potential to reduce the length of data collection campaigns, whilst also using more flexible instruments and thus increasing cost efficiency. The software front-end tool Windmodeller, which drives the ANSYS CFX software, is used to benchmark four separate two-equation turbulence models, each assuming neutral atmospheric stability conditions. Production data from four European offshore wind farms are used for validation purposes. Of these models, the Shear Stress Transport (SST) model consistently performed the worst, whilst modifying the RANS turbulence constant, 𝐶𝜇, only alters the location within a line of turbines where the standard 𝑘-𝜀 model was most accurate. The unsteady RANS model variation, which incorporates both the Coriolis effect and a stably stratified capping layer, was found III to have the smallest root-mean-squared error values for the largest wind farm and so was chosen to form the basis of the simulations incorporating atmospheric stability. The Obukhov Length required for MOST is incorporated into the CFD simulations using surface fluxes, water temperatures and atmospheric thermal gradients. There are only small variations in simulation accuracy between methods when simulating Neutral conditions, with the thermal gradient method performing best. Under stable conditions the sea surface temperature approach is most accurate, although it is also the least accurate under unstable conditions and was unable to generate the more extreme Unstable conditions. Although the flux method was less accurate than the gradient method in absolute terms, the variance of its errors at individual turbine locations was consistently smaller. The validation process for using MOST techniques was complicated by a lack of sufficient field data after the rigorous filtering required by the theory’s assumptions. The preliminary work using alternative methods of describing atmospheric conditions within CFD simulations did not suffer from a lack of validation data, but was unsuccessful at maintaining the required wind shear profiles across the whole domain. Recommendations are made to improve control over these parameters with models such as unsteady RANS, and to find a suitable successor to the actuator disc theory now wind shear values across a turbine are becoming significant.