We compute the loads on a model-scale tidal turbine with Blade Element Momentum (BEM) theory and Computational Fluid Dynamics (CFD) simulations, and we compare the results with towing tank tests. CFD simulations are wall-resolved, steady, Reynolds-averaged Navier-Stokes simulations with a k − ω SST turbulence model, where only a 120◦ wedge domain with a single blade is resolved in a non-inertial frame of reference. We undertake a detailed uncertainty analysis to identify the sources of error. BEM uncertainty is computed with a Monte-Carlo approach based on the differences in the predictions of CFD and Xfoil for the sectional lift and drag coefficients,while CFD uncertainty is based on the errors due to the finite number of iterations and spatial resolution.
The maximum error of CFD (8.0%) with respect to the experimental data is about half of that of BEM (15.5%) for the power (CP) and the thrust (CT) coefficients and both errors are within 4.1% for CFD and within 7.2% for BEM around the optimal tip-speed ratio (λ = 6.03). The BEM error is within the uncertainty associated with the imprecise knowledge of the sectional lift and drag coefficients. The sectional forces from CFD and BEM disagree at both the tip and the root, resulting in a substantial BEM underprediction of CP at high λ values (up to 15.5%), yet CT is well predicted (within 2.3%) at every λ. The CFD uncertainty is markedly smaller than the error, which is thus mostly due to a modelling error such as the turbulence model, the neglected effect of the support structure, the free surface, and the imprecise knowledge of the input conditions. Overall these results suggest that CFD provides both a maximum error and uncertainty that are substantially smaller than that of BEM, but both methods suffer from modelling errors that require further investigation.