Experimental characterization of the near-wake of a cross flow water turbine with LDV measurement

Recent developments in tidal energy converter technology and turbine array projects require fast and robust simulation tools. An important challenge is now estimating the power production for a array of machines as well as optimizing the placement within a grid. The most common method to do so is the coupling of momentum sources with a flow simulation. This method, called the actuator disc model gives a good approximation of the turbine impact but shows discrepancies as far as the wake and speed recovery is concerned. This study establish useful data in the near-wake of a Darrieus (vertical axis) turbine. Its aim is to understand the processes in the wake of a cross-flow turbine in order to validate full geometry CFD simulations and build an adapted and simplified model. This problematic of the wake modeling is well known for axial flow turbines. Blade tip vortices as well as wake rotation provides a strong contribution to turbulence production that is not reproduced in basic actuator disc model. An effective modification of k-ε conservation equation has been proposed to account for non-linear scale interaction in the turbulence spectrum. It is based on the addition of source terms in conservation equations of k and ε to account for energy transfer from large scale turbulence to low scale turbulence proposed by Sanz to deal with canopy induced turbulence [7]. In the field of hydrokinetic convertors, the work of El Kasmi [5] for axial flow wind turbines and Roc et. al. [6] for marine axial turbine, both successfully adapted the modification to improve actuator disc model. The present study focuses on cross-flow water turbine (vertical axis) of Darrieus type. The counter rotating vortices visualized by Brochier are essentially two-dimensional and advocate for a different mixing process in the wake [7]. The understanding of the physical phenomenon in place is crucial for the construction of a reliable model for cross flow turbines. Several numerical models using RANS turbulence models coupled with sliding mesh methods demonstrated the ability to simulate the Darrieus turbine in 2D and 3D configurations. The comparison with experimental data showed a good precision for the flow dynamic in the rotor, as well as for the torque estimation [5]. The aim of the present experimental study is therefore to establish reference data on the flow in the wake of a Darrieus turbine in order to : • examine the precision of the complete RANS simulation for the wake description • understand the process responsible for the velocity recovery and adapt actuator disc model for a the cross flow turbine The hydrodynamic tunnel of the LEGI consists in a rectangular section of 0.25x0.7m and a length of 1m streamwise, inserted in a closed hydraulic loop of 30m. Small scale model of a cross-flow water turbine of Darrieus type is tested in it. Flow velocity varies between 1 to 2.3 m/s. It corresponds to a blade Reynolds number between 1.7x105 and 5x105. The flume lateral blockage ratio is 4, and a vertical blockage ratio of 1.03. The analysis presented here corresponds to a two-dimensional configuration. A twocomponent LDV system with back scattering is placed on a 3-axis traverse (see on figure 3). It allows measurements of stream-wise and transverse velocity in horizontal planes up to a distance of 3D behind the machine. The flow is seeded with 10 μm slivered hollow glass particles, for a sampling frequency up to 5 kHz. Time average velocity and turbulent kinetic energy are mapped at mid-blade horizontal plan behind the turbine. The investigation zone goes from 0.5 to 2.7 diameter in the stream-wise direction and is 2D wide. of 2D in the transverse direction. A constant step of ∆X = ∆Y = D/40 = C/8 is used. The velocity deficit produced by the full CFD simulation in the near wake shows a strong agreement with the 2D LDV measurement. Nevertheless the difference in turbulence levels and spreading of the sheared regions found here is likely to generate further differences. The transport of coherent structure within RANS model might be considered as a reason of the discrepancies. This analysis claims for an extension of the present study in the far wake to determine the validity of the RANS modeling approach, in the complex case of cross-flow turbines. As the simple modeling is concerned, the classical actuator disc model is not adapted directly to cross flow turbines. The study shows a direct effect of the momentum source distribution over the domain mesh. The complex distribution used by the authors seems to produce an over estimated level of turbulence. On the contrary, a simple uniform distribution predicts a too long length of recovery. The measurement data presented here will help to build a adjusted model with a good source terms distribution and/or turbulent sources of k and ε.