CFD Simulation of a 3-Bladed Horizontal Axis Tidal Stream Turbine using RANS and LES J. McNaughton, I. Afgan, D.D. Apsley, S. Rolfo, T. Stallard and P.K. Stansby Modelling and Simulation Centre, School of MACE, The University of Manchester Oxford Tidal Energy Workshop Oxford, UK. 29-30 March 2012
Overview Intro Method Results Comp Con References 1 Introduction 2 Methodology 3 RANS Results 4 RANS and LES comparison 5 Conclusions & Further work
Introduction Intro Method Results Comp Con References ReDAPT (Reliable Data Acquisition Platform for Tidal): Project designed to accelerate the tidal energy industry. Aim to inspire market confidence in tidal stream turbines (TST). Deployment of a full scale 1MW turbine at EMEC in the Orkneys. This work: Using Code Saturne to accurately predict loading on tidal turbines. Assess the influence of turbulence and waves on TSTs. Current results compare Reynolds Averaged Navier Stokes (RANS) model and Large Eddy Simulation (LES) against those for a laboratory scale turbine.
Introduction Case overview CFD modelling of experimental study of Bahaj et al. (2005). 0.4 m radius, R, turbine pulled through a towing tank. Parametric study of force coefficients against tip-speed-ratio (TSR). Photo reproduced from Bahaj et al. (2005).
Methodology CFD Solver EDF s open-source CFD code, Code Saturne (Archambeau et al., 2004), is used to simulate turbulent flow past a TST. A sliding mesh method is developed in Code Saturne to allow for the TST rotation. Second order in space (central differencing). First-order in time with t 1.5 of rotation per time-step. The k ω SST RANS turbulence model is used to simulate the flow. Results are also compared against an LES that is also been performed as part of this project.
Methodology Sliding-mesh method Start in cell centre, I, with face-centre, F. I is projected through F to give a halo point, H. Search for, J, the nearest cell-centre to the halo-point. The face-centre value is given a Dirichlet condition with the value: φ F = 1 2 (φ I + φ H ), Interface I H F J Where: φ H = φ J + φ x (JH). J
Methodology Meshing strategy Geometry built in two parts, inner turbine and outer domain. Block-structured approach with hanging nodes to control cell-count. 2 million cells with 15 < y + < 200 at the walls. LES mesh is wall refined with 7.7 million cells.
RANS Results Flow-field Pressure iso-surfaces coloured by velocity are shown for the instantaneous flow-field for TSR = 6. Tip effects are clearly visible in the near wake whilst the mast creates main structures further down-stream.
RANS Results Flow-field Instantaneous velocity on centre-plane shows tip-effects and influence of mast on flow.
RANS Results Force coefficients Effect of blades passing the mast is clear from instantaneous force coefficients. Power Spectral Density (PSD) analysis of the C P shows peaks at 3 and 6 times the blade rotation frequency (f). 0.743 Force coefficients over one full rotation for TSR=6 0.742 0.741 0.74 0.739 0.408 0.407 C T C P 0.406 0.405 0.404 7.68 7.82 7.96 Time (s)
RANS Results Force coefficients k ω SST predicts force curve against TSR although under predicted by approximately 10%. LES shows gain in precision matching experiments within 3% except for the lower values of TSR. 1.2 1 Bahaj et al 2005 k-ω - Present work k-ω - McSherry et al 2011 LES 0.5 0.4 C T 0.8 C P 0.3 0.2 0.6 0.1 0.4 4 5 6 7 8 9 TSR 0 4 5 6 7 8 9 TSR
RANS and LES comparison Flow structures Iso-Q 0.5 (Ω ij Ω ij S ij S ij ) surfaces coloured by vorticity are shown for RANS and LES. LES maintains the tip vortices and captures structures in the wake better than RANS. k ω SST LES
RANS and LES comparison Pressure coefficients on the blades Mean pressure coefficients are shown on the blades at quarter length locations. Both RANS and LES predict similar behaviour although LES captures larger forces on pressure and suction surfaces. 2 r/r = 0.25 r/r = 0.50 r/r = 0.75 2 r/r = 0.25 r/r = 0.50 r/r = 0.75 1 1 -C P 0 -C P 0-1 -1-2 0 0.2 0.4 0.6 0.8 1 x / c k ω SST -2 0 0.2 0.4 0.6 0.8 1 x / c LES
Conclusions & Further work Conclusions: Sliding-mesh method successfully implemented in Code Saturne to simulate a rotating TST. Flow features are captured well by RANS and LES. RANS under predicts the force coefficients whilst LES is far more accurate. Difference in calculations is near wall modelling and higher order time-scheme used by Code Saturne for LES. Further work: Wall refined RANS simulation to compare with experiments. Assess influence of waves on full scale MCT comparing with data from EMEC.
Acknowledgements This research was performed as part of the Reliable Data Acquisition Platform for Tidal (ReDAPT) project commissioned and funded by the Energy Technologies Institute (ETI). The authors are highly grateful to EDF for additional funding and access to its High Performance Computing (HPC) facilities.
References Intro Method Results Comp Con References Archambeau, F., Mechitoua, N., and Sakiz, M. (2004). Code Saturne: a finite volume code for the computation of turbulent incompressible flows-industrial applications. International Journal on Finite Volumes, 1(1):1 62. Bahaj, A. S., Batten, W. M. J., Molland, A. F., and Chaplin, J. R. (2005). Experimental investigation into the effect of rotor blade sweep on the performance of the Marine Current Turbines. Proceedings of the Institution of Mechanical Engineers, Part A: Journal of Power and Energy, 215(5):611 622. Mcsherry, R., Grimwade, J., Jones, I., Mathias, S., Wells, A., Mateus, A., and House, E. F. (2011). 3D CFD modelling of tidal turbine performance with validation against laboratory experiments.
Sliding-mesh method Implementation: Code Saturne Subroutines Black - Unchanged. Red - Modified. Blue - New. findha - Search for closest cell-centre to halo-point. upcoef - Updates Dirichlet values on interface. caltri inivar usiniv tridim usclim findha navsto preduv resolp upcoef codits upcoef upcoef
Sliding-mesh method Implementation: Pressure-Velocity Loop Code Saturne s existing pressure-velocity loop is enabled by setting NTERUP > 1. Loop used to ensure continuity over the interface. Modification to include the RANS subroutines. tridim phyvar usclim navsto turb** P-V Loop phyvar
RANS and LES comparison Comparison of numerical set-up RANS (k ω SST) LES Time-scheme 1st order 2nd order Rotation per time-step 1.5 0.05 Space discretization Centred Centred Wall modelling Wall-functions Wall refined Mesh size 2.2 million 7.6 million