You are on page 1of 14

Renewable Energy 63 (2014) 222e235

Contents lists available at ScienceDirect

Renewable Energy
journal homepage:

Near-wake characteristics of a model horizontal axis tidal stream
S.C. Tedds a, I. Owen b, R.J. Poole a, *
School of Engineering, University of Liverpool, Brownlow Street, Liverpool L69 3GU, UK
School of Engineering, University of Lincoln, Brayford Pool, Lincoln LN6 7TS, UK

a r t i c l e i n f o a b s t r a c t

Article history: The results of a detailed experimental investigation of the near-wake (up to seven turbine diameters
Received 1 February 2013 downstream) of a model horizontal axis tidal turbine (HATT) device in a large-scale recirculating water
Accepted 13 September 2013 channel facility are reported. An Acoustic Doppler Velocimeter is used to provide detailed three-
Available online 10 October 2013
dimensional mean and turbulent flow field information at five different depths across the full width
of the channel downstream of the turbine, giving the most complete three-dimensional velocities and
Reynolds normal and shear stress data set yet available. In addition the Reynolds-stress anisotropy tensor
is used to illustrate the degree of anisotropy of the Reynolds stress within the turbine’s wake. These
Tidal stream turbine
Acoustic Doppler velocimeter
results reveal the strongly anisotropic nature of the near-wake turbulence suggesting isotropic turbu-
measurements lence models should not be used to model near-wake dynamics. Finally the power-law decay rates of the
Turbulence statistics maximum normalised turbulent kinetic energy differ significantly from those found downstream of
Reynolds stress anisotropy grids, meshes or perforated disks, suggesting that previous modelling approaches, which neglected swirl
effects and modelled the turbine by absorption discs, may significantly over predict the turbulent kinetic
energy decay rate of HATT wakes.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction make them commercially viable, wake recovery length is crucial for
the appropriate spacing between turbines. Also, knowledge of the
Throughout the world there is a growing demand for energy wake is important so that potential effects on the seabed can be
produced from sustainable resources, with many governments investigated. Only a limited amount of research has hitherto been
setting targets for renewable production of electricity. The UK aims undertaken into the wakes behind tidal stream turbines. The
to provide 15% of total energy from renewable sources by 2020, this earliest of this research was conducted using an absorption disc to
is an increase from 3.3% produced in 2010 [1]. Harnessing the represent the turbine, both experimentally [2] and using Compu-
ocean’s energy is one way to meet these targets, and the energy can tational Fluid Dynamics (CFD) [3e5]. Experimental studies into the
be split into two categories: wave and tidal. One of the main ad- characterisation of the wake were conducted by Myers and Bahaj
vantages of tidal power is the predictability of the tides. There are [6], Stallard et al. [7,8], Rose et al. [9] and Maganga et al. [10].
two principal ways to harness tidal energy: tidal barrages or la- Further studies to compare experimental wake data to CFD were
goons which use the tidal range, and tidal stream turbines that use completed by Mycek et al. [11] and Rose et al. [12]. Table 1 includes
the tidal current. An advantage of tidal stream turbines is that they relevant details of studies undertaken to date.
minimise the impact on the marine environment as they allow Myers and Bahaj [2] conducted experiments in a 21 m tilting
water to pass straight through, and are usually fully submerged flume, which had a width of 1.35 m and depth of 0.4 m. The vertical
with no visual impact. velocity profile produced in the flume resembled a modified 1/7th
As well as characterising the power output for a tidal stream power law which was more uniform closer to the surface and
turbine it is also important to characterise the wake of a turbine. For approximated a profile measured at a full scale site and reported by
example, as turbines are likely to be placed in farms or arrays to the Carbon Trust [13]. Artificial bed roughness was added to the
flume to reduce the velocity and increase the shear stress in the
bottom third of the water column, again to create conditions that
* Corresponding author.
were as realistic as possible. The absorption discs (mesh discs) used
E-mail addresses: (S.C. Tedds), by Myers and Bahaj [2] were 100 mm in diameter and had porosity
(I. Owen), (R.J. Poole). (ratio of opened-to-closed area) varying between 0.48 and 0.35.

0960-1481/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved.

A computational study was conducted by Daly et al. [4] compared CFD simulations to the experi- mental absorption disc results conducted by Myers and Bahaj [2] discussed above.and three-dimensional CFD calculations. These simulations gave a first approach for the characteri- n kinematic viscosity of water (m2/s) sation of the wake and were concentrated on the near-wake and r density of water (kg/m3) effects on the free-surface close to the disc. by 5D downstream the TI was 10% which was still much formation on the far wake provided swirl effects are not still higher than the upstream level of 6%. 0.3 m and with air to a height of 0.1 m and thickness of 0.5D. which was 1. CT z 0.1D . and was located in the had dissipated. the thrust coefficient was matched.3% wake velocity deficit downstream increased. Myers centre of the water column. The turbulence model used was the kε model P power (W) which assumes isotropic turbulence. U mean upstream velocity (m/s) The two-dimensional domain was extended further to a three- v transverse velocity (m/s) dimensional domain. in turn this increase of and had an upstream velocity of 0. to a lowest point of roughly half its initial height. larger number of small holes had a greater thrust than discs with with a water depth of 0. Both the far-wake and the near- the mean velocity deficit is longer. The structure increased the turbulence intensity. where Myers and Bahaj [2] defined accuracy. For all discs tested. An Acoustic Doppler Velocimeter (ADV) device was was calculated using the ku shear stress transport (SST) model.8 m. v0 and w0 are the varying components of the flow. those with a which the experiments of Myers and Bahaj [2] were conducted. liminary studies indicated that the kε did not accurately model It was found that the porosity alone did not determine the thrust the flow conditions. 4 m wide and 2 m deep. It was found that as the disc was placed closer to the floor. especially near the free-surface. with the disc experimental results. the near-wake had some limitations in its (TKE) was also measured. giving a blockage ratio of 6. These results were compared there was some evidence that as the distance between the floor and to experimental data of the wake of a actuator fence in a channel. (1) 2 bulence levels in the wake were comparable. used.15 m above fewer large holes. and the ‘actuator fence’ to simulate an array of tidal turbines. The turbulent kinetic energy wake were compared. centre. TKE as but this. To enable direct comparison with the and Bahaj [2] investigated four different discs depths. [4] simulated the free-surface us- A load cell was attached to the disc so that thrust acting on it could ing a homogeneous coupled volume-of-fluid approach. The wake of the turbine the velocity deficit was larger than that with the smoother bed. Similar to the two- HATT (m) dimensional solution. the 18 m long. In the two-dimensional sim- R radius of turbine (m) ulations. 0.5 m. Harrison et al.75D above the floor of the measured inlet velocity distribution of the flume was used to set flume. The flow be measured. at 10D downstream the mean the water.33D.C.2D). 10 m long and had a CP power coefficient (P/0.9. in that the effects of swirl were not taken into account. was a reasonable method to estimate the effects of the far wake.66D and 0.1D. Sun et al. / Renewable Energy 63 (2014) 222e235 223 Sun et al. where a square disc was used which had an w spanwise velocity (m/s) area ratio to the flume cross-section of 17%. there was a free-surface drop behind the z spanwise distance behind the HATT (m) disc. in both two. These tests found that placed at various heights in a channel. Results from this so- T thrust (N) lution showed a substantial drop in the free-surface behind the u streamwise velocity (m/s) absorption disc. which sampled at a rate of 50 Hz with a sample volume of which is an extension of the kε model. port structure with a diameter of 0. which was bar above denotes this as a time average. Tedds et al. the disc increased. The computational domain used replicated a labora- A swept out area of turbine (m2) tory water channel.001 m. y transverse distance through the depth behind the then velocity drops directly behind the disc. however the drop was not as large for the three-dimensional l tip speed ratio (uR/U) model. placed in relation to the bed to minimise the effect of the bed. The centred at 0. The wake of the support important. the TKE levels in the wake increased. The trend and wake recovery of the far   1 wake was found to be similar in the two approaches and the tur- TKE ¼ u02 þ v02 þ w02 . to map the wake up to 20 disc diameters (D) downstream. in a recirculating water channel which was and as the distance between the disc and the floor decreased. u angular velocity (rad/s) Harrison et al. surface. When the water passed through the absorption disc and turbine radius (UR/n) approximately 38% of energy was dissipated. The model domain represented the flume in coefficient.1D. the authors argued. Using a disc with thrust coefficient.5rAU3) water depth of 1 m.5 m wide. Again the depth at which the disc was placed was altered Velocimeter (ADV). The disc was matched to that of Myers and Bahaj [2] with velocity profiles were virtually identical and any effects of the disc a diameter of 0. 0.8 m/s. The largest deficit was found to be at a height of 0.5rAU2) simulate the free surface and a no-slip wall boundary condition was D disc diameter used on the bed.15 cm3. To simulate which gave a reasonable approximation to the streamwise velocity a rocky seabed in their flume. Myers and Bahaj [2] placed artificial downstream of the fence. The support less suitable for placing tidal turbines than a smoother bed. the inlet velocity in the CFD model. As a support structure without the rotor was measured and was found consequence. [3] used an absorption zone centred at a height Re Reynolds number based on upstream flow velocity of 0. The results show that as HATT (m) the water flow reaches the disc it accelerates as it flows through. The changes in wake length with the height of the structure and rotor was measured at 5D downstream though the turbine provides a good indication of where a turbine should be width of the channel at five different heights (0. this resulted in a 10% x streamwise distance across the width behind the overall loss of kinetic energy in the flume. using an Acoustic Doppler bulence. [14] of an where u0 . [3] also used an absorption disc approximation to Nomenclature simulate a tidal turbine.2D. S. rocks with lengths between 6. A volume-of-fluid approach was used to CT thrust coefficient (T/0. this was used as pre- 0. Four discs with equal porosity were tested. 0. as Myers and Bahaj [6] measured the wake of a turbine and sup- they expected this type of bed to have the greatest effect on tur. again near to the free- experiments conducted by Myers and Bahaj [2] provide useful in. 0. Myers and Bahaj [2] suggested that a rocky seabed is to be significant.7 and 10 mm over a length of 4 m.

a Based on the radius and upstream velocity which is estimated by present authors based on limited data provided in the original papers. CFD 0.224 S. Exp 0. Harrison et al.9 RANS þ BE model gives a better agreement (2013) [5] to experimental data than the disc model. The effect of waves on the wake had a similar velocity trend to the uniform flow between 5D to 10D but the TI remained a lot higher downstream.4% TI at 7D downstream the (2011) [11] experimentally TI is 90% of the initial TI. CFD An actuator fence was used to simulate (2013) [14] arrays of turbines. firstly. due to the presence of the structure support and the rotation of the approach and RANS þ disk approach where turbulence is injected blades. The RANS results were compared to the showed a reasonable agreement to experimental data of [6] be- experiments conducted by Myers and Bahaj [6].3% z6% 3.4% 3.9 When disc was placed closer to the floor (2010) [2] mesh discs the velocity deficit persisted for longer. 13.000 a recovery of 70% of upstream velocity.2D) as the support structure was not present here. Exp 0. 8D and 10D downstream therefore the turbine diameter. The thrust velocity.373 Centreline measurements of velocity are (2011) [9] and 1. the 27. Batten et al.5% 10% and 25% Based on blade 0. Exp 0.3 m 18% 1.8 m 6. Results showed that by 22D downstream the velocity had not returned to the upstream value.567 consistent for the three devices. Rose et al.2  105 0. Overall it is clear that the support structure in this coefficient was overpredicted by the RANS þ BE approach particular set-up had a dominant effect on the overall wake. Tedds et al. With the TI and 0% in CFD levels of 2.8 m 6.9% and 14. with waves chord 30. proaches and experimental data were analysed through the depth gated.000. a RANS actuator disc approach þ blade element (BE) of the channel. 0. Maganga et al. Myers and Bahaj Exp 100 mm e 1.5% z5% 2. By 10D the velocity had recovered to 80% of the upstream velocity. Exp and CFD 0. CFD 100 mm e 1. At 5D.4% TI after 5D downstream the (2010) [10] deficit is almost negligible. CFD and Exp 0. The smallest deficit was found at the deepest point at the disc.5 compared with the results downstream of the centre of the turbine.5  105a With 14. The time Batten et al.5  104a z0.27 m 2. / Renewable Energy 63 (2014) 222e235 Table 1 Overview of previous tidal stream turbine wake studies.000 maximum deficit is always at the centreline. Two different computational models were investi. Mycek et al. [5] investigated the use of the actuator disc. Daly et al.7 m 5% 4. With two or more turbines the individual wakes of the devices can be seen up to 4D downstream but after the wakes merge to form a single deficit.9% at 10D downstream the flow is still non-uniform and TI levels are still much greater.14 m.5% z5% 2. compared to both the BEMT and experimental results. size of channel and velocity and TI time average velocities and TI of the two computational ap- was matched. The RANS þ BE can be used to predict the power co- measured (0. The CFD results were similar for the velocity measurements but underestimate the TI.9%. discussed above. Author/year CFD/ Turbine Blockage ratio TI Reynolds number CT Conclusions experimental diameter Sun et al. 5a Rose et al.6% TI at 10D downstream the TI is sill higher and the velocity deficit still visible.7 m 5% 2. with 4.6% and 14.5  105a With 14. At 5D downstream the TI from the RANS þ BE . (2008) [3] CFD 0. The CFD results under predicted the experimental results.C. the computational approach showed good agreement to the experimental results of a fence. efficient (CP) and thrust coefficient. and 0.3% 5% 3. averaged velocity and TI downstream of the turbine at the cen- Reynolds-average NaviereStokes (RANS) approach of predicting treline of both the RANS þ BE and the RANS þ disk approaches the wake of a tidal turbine.9 The trend of the wake recovery and (2010) [4] mesh discs turbulence levels is qualitatively similar in the CFD and experimental results.2  10 Rapid recovery between 2D and 8D but (2011) [12] by 25D still not fully recovered.5  104a z0. Stallard et al.5 m 17% First approach at estimating the near wake.2  10 0.9 Support structure had a large influence (2009) [6] up to 5D downstream. but (2013) [8] by 20D downstream the velocity was 80% of the upstream velocity.959.82 Rapid recovery up to 5D downstream with (2011) [7]. tween 5D and 10D downstream. 5a Myers and Bahaj Exp 0.000 and 246. it was found that by 10D from blade element momentum theory (BEMT). this model under predicted the The streamwise velocity was also measured from 3D to 10D CP at a tip speed ratio greater than 6.5 m chord 25. but gave a downstream the velocity was still less than 80% of the upstream reasonable approximation to the experimental results.4% 3.3% Based on the tip 0.25 m 3. 0.

Stallard et al.25 m. is a lake of approximately 400 m  133 m. A third. The negligible. 14. Rapid recovery of the wake was observed centreline always has the maximum velocity deficit.10 . 1.4 m  0. [9] because of the differing thrust co- efficients for each turbine. for the first 5D downstream. 4 m wide and 2 m deficit of just under 20%. The turbine used was four-bladed with a diameter of 1.5 m/s and 1.23 m below the surface.8] probed the effect of a series of turbines on the wake structure. [10] measured the wake of a three-bladed turbine. A 100 mm between points.6% and 14. [7. were taken every diameter in the streamwise direction behind the turbine.5. It was also noticed at the higher recovery in these conditions was similar over a range of 5D to 10D turbulence intensity level. It was found that with the 2. correctly 14. Measure- ments were taken between 0. The wake of a four-bladed.6%) the mean second experiment took place in a flume 35 m  0.8 Hz). therefore collecting a total of 825 samples per location. with a tip chord Reynolds number of 25. This experiment was undertaken in a flume which was 20 m wake recovery and that after 5D downstream the deficit is almost long. 5. placed at 0. however the turbulence intensity remains above 12% parison to the base case (4. sampling at 200 Hz for 60 s at each location (i.e.92 m (length  width  depth) at a mean velocity of 0.15 and 20 . 2. The effect of waves was briefly investigated.14 m diameter.5 m. All turbines were three bladed and had a diameter of 0. with a Reynolds number based on the blade chord of 30. much slower and by 20D downstream the wake still had a velocity with a diameter of 0.4%. [9] tested three devices and measured the wake using the turbine. It is difficult to compare these three results produced by Rose et al.57 m/s.25 s (frequency thrust were also measured at these yaw angles. Results from the study showed that with total of 1000 image pairs were taken for each test. but further downstream the TI levels of the model were in reasonable agree- ment with the experimental data. Velocimetry (LDV) system was used to measure the velocities behind Rose et al. with a sampling rate of 50 Hz for 100 s.7 m in a flume 18 m long. The Fig.285 m below the water surface. Different yaw angles of a single structure. was turbine. A two dimensional Laser Doppler at 10 D downstream.27 m (corresponding to an estimated 1/70th scale). Power and a waveheight of 50 mm and a peak period of 1.4] it is not entirely unexpected for the wakes to be different.000. The number of turbines tested at a time was 1. Further downstream the recovery was Maganga et al. and three rows. Tests were conducted with ve- turbine configurations the individual deficit was identifiable up to locities between 0. . An ADV system was again used which had a sampling rate of 64 Hz. The general Fig.52 m deep. at a velocity of 0. The site. As other work has shown that the wake is strongly affected by the thrust on the turbine and CT is not matched in these cases [2.6% TI).C. up to 10D downstream. and with different upstream 4D downstream but further downstream all the wakes merged into turbulence intensity levels of 4. Batten et al. The lateral spacing between the devices was 1.000 samples per point. Mont- gomery Lough. For a single turbine.000 samples per spatial point). These waves increased the “turbulence intensity” to around as the yaw angle increased from the optimum position (i.10 . 7 and 10. larger scale experiment was undertaken in a lake by towing a turbine.000. Tedds et al. over a 10 s the higher turbulence intensity level (14. 6. width of 5 m and lengthpof 12 m. the wake was measured directly downstream of the pattern produced by the three experiments is consistent as the centreline of the turbine. S. Measurements different techniques. with the turbine were investigated. the longitu- dinal spacing of devices considered was between 4 and 10D.000. and it was found that 0.42 m/s using a two-bladed turbine with a diameter of 0.7D and 1.75 m wide and 0. two.4%. At the lower turbulence intensity level (4. the thrust decreased. 2. A Nortek ADV was used to measure the velocities downstream.e. and tip chord Rey- nolds number of 27. 3.4%) there is a much faster period. both CP and CT decreased in com- downstream. 0 . / Renewable Energy 63 (2014) 222e235 225 model was greater than that of the experimental data.000.4% at the surface but the TI decayed through the depth. [5] concluded that the RANS þ BE model was a better method of modelling a turbine compared to the RANS þ disk model as the wake measurements showed a better agreement to that of the experiments and it removed the pragmatic approach of turbulence source terms added at the turbine location.5D and 3D in the longitudinal direc- tion and 1D laterally. The turbines in the arrays were configured into one. wake was measured using an ADV.7D with a spacing of acquired images at 100 Hz with 1000 ms between pairs of images.45 m. Schematic of the University of Liverpool water channel. measurements were also taken in measured using a Particle Imaging Velocimetry (PIV) system which the spanwise direction between 1. 3 and 5 deep. Schematic of the scale HATT. it was placed 1. 0. each point was sampled for 170 s therefore producing 10. These ex- periments were conducted in a flume with a water depth of 0. collecting 12. The wake aligned (0 )). with a Reynolds number based on the chord length of 246. sampling at 25 Hz for 33 s at each location. 2 or 3D. giving a blockage ratio of 5%. 0.45 m/s. the upstream flow ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi had a turbu- pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi lence intensity (TI ¼ ð1=3Þðu02 þ v02 þ w02 Þ ¼ ð2=3ÞTKE) of approximately 10% and a mean velocity of 0. combined with the sparse data sets.5 m/s.

4 and 6 ). The increase in blade pitch angles showed a smaller wake Fig. using a Lagrangian vortex particle method with a large eddy simulation turbulence model.C. [10] by Other comparisons of CFD and experimental work have been studying the interaction between two tidal turbines placed in se. Thus high upstream turbulence levels signifi- cantly reduce the wake. which were flow velocity of 0. As expected there was a drop in maximum power coeffi- levels. with a measure the three-dimensional velocities and TI. The results showed that there was a rapid and significant measured with a two component LDV system and were collected at recovery between 2D and 8D downstream. The turbine had interchangeable interaction of wakes of two turbines one placed downstream of the blades and 2 and 3 blade configurations were tested at different other. computational model. 3. z825 points per 2. the higher the maximum CP of the and prisms [15e18]. . 4. These results confirm a number of experimental studies cient. For the wake behind a single turbine with TI of 14.9% and 14. Experimental work with two turbines.92 m wide. The water long. however the mean velocity was underestimated by the Fig.42 m deep and 35 m three-bladed with a diameter of 0. The results showed a qualitatively similar profile.4% it was found that by 7D downstream the wake had recovered to 90% of its upstream turbulence intensity level and its uniformity had returned. using two component blade pitch angles (0 . 2 . downstream turbine.9% even after 10 diameters the profile was still non-uniform and the TI was still much greater than the upstream levels. presumably by enhancing the transfer of momentum between the free stream and the wake regions. The numerical study conducted by Mycek et al. Diagram to illustrate wake measurement locations downstream of the turbine. and numerically using CFD. An ADV was used to channel used was the same as that of Maganga et al. 8D and 10D downstream from the first bulence intensity levels in the wake remained higher than upstream turbine.2D behind the turbine. [11] furthered the work of Maganga et al. a circular cylinder turbine was placed from the first. Contours of the mean streamwise velocities behind a turbine. [11]. compared the power output of the velocity deficit is still clearly visible 10D downstream and the tur. Mycek et al. however the wake had each point for 100 s with a frequency of 717 Hz (therefore not fully returned to the mean upstream velocity by 25D down- 7001700 samples per point).226 S. also reported by Mycek et al. therefore giving a blockage ratio of 18%. [11] investigated an upstream flow with 0% TI. [12]. Whereas with a TI of 2. in which a single rotor with a ries. / Renewable Energy 63 (2014) 222e235 10D downstream of the turbine and 25 measurements were taken in the spanwise direction across the width of the flume every 100 mm.7 m (1/30th scale).e. The turbines investigated were ducted in a tank which was 0. The experiments were con- LDV. Tedds et al. The velocities and turbulence intensities were location). conducted by Rose et al. CP. [10]. The measurements were taken up to stream. second turbine at 4D.3 m was used. 0. by first measuring the wake of a single turbine and then the diameter of 0. 6D. for the downstream turbine (based on the velocity up- conducted on the influence of ambient turbulence intensity on the stream of the first turbine) and the further downstream the second near wake of different structures: a wind turbine.4%. this was done both experimentally. A com- parison of these results with experimental data was undertaken at 1.8 m/s and upstream turbulence intensity levels of sampled at 25 Hz for 33 s at each location (i.

The CFD compar. used a three-dimensional RANS stream of a turbine. A ku turbulence model was used near the walls and kε predicted the wake due to the assumption of an isotropic turbu- was used in the main flow. / Renewable Energy 63 (2014) 222e235 227 Fig. 6. In Testing was completed at the University of Liverpool using the the experimental studies the number of locations at which the flow high-speed recirculating water flume and a turbine with diameter has been characterised has been sparse or been concerned with the of 0. . Previous numerical studies have generally under- longer. The assumptions of isotropic turbulence lent flow field. In this study the Reynolds-stress anisotropy tensor is in this latter model meant that it was under-predicting the wake used to illustrate the degree of anisotropy of the Reynolds stress compared to that measured in the experimental study. turbulence statistics and Reynolds shear stresses data the experiments were conducted. Contours of the mean transverse velocities behind a turbine. but they made the channel set available. [12]. Contours of the mean spanwise velocities behind a turbine. giving the most complete three-dimensional simulation using the same width and height of the flume in which velocities. The research studies conducted to date have investigated the mean streamwise flow and turbulence intensity in single planes 2. This study investigates the near-wake (x/D  7) at two-bladed due to the corresponding higher CT. impact and the three-bladed turbine had a higher deficit than the far-wake effects. Fig. within the turbine’s wake. Tedds et al.5 m. Test facilities downstream of the turbine. with measurements confined to a centre plane through the turbine or planes normal to the flow. 5. S. five different depths across the full width of the channel down- ison conducted by Rose et al.C.

[20]. A regen resistor or dy- namic brake was used to apply an opposing torque to that Testing was undertaken in the University of Liverpool high- developed by the hydrodynamic forces on the turbine.1 A (max). The turbine has turbine was calculated from: been designed such that the number of blades can be varied from 2 up to a maximum of 6 blades and the blade pitch angle can be speedð%Þ 500 p u¼   2p ¼ speed  (2) varied. The torque throughout the and subsequently optimised using Blade Element Modelling theory. tests was set to 20% of the current limit. The streamwise Reynolds normal stress (i.e. 7.22]. standard deviation of the streamwise velocity) at a depth corresponding to the centre of the turbine behind a turbine. the angular velocity of the and nose cone that has been used in this study. in this case 10. three.228 S.5 m diameter. 2. The original Horizontal Axis Tidal Turbine (HATT) blade profile The torque applied was proportional to the maximum drive rate was selected by Ref. The rotational speed of the Fig. Complete details of the turbine can be found in Ref. proportionality constant of 0. Tedds et al. Therefore. which is related to the torque by a Advisory Committee for Aeronautics (NACA) for aerofoil sections. motor was measured as a percentage of the rated rotational velocity blade configuration HATT with the stanchion attachment point for the motor (500 rpm). angular velocity and po.2. [19] using methods suggested by the National current. / Renewable Energy 63 (2014) 222e235 2. 100 60 6 The turbine was connected to a Baldor brushless AC servomotor in order to measure/calculate the torque.82 Nm/A. Description of turbine was combined with a control system which in turn was pro- grammed via Workbench V5 [21.1. Experimental set-up wer generated via hydrodynamic loading. This motor speed re-circulating water flume.C. . 1 shows schematically the assembled 0. a schematic of which is shown Fig.

behind the turbine to 7D. measured at five different depths (transverse distances) at a l:/ELSEVIER/RENE/An Acoustic Doppler Velocimeter (Nortek number of streamwise distances behind the turbine. midway along the working bulence statistics were meaningful.85 m. Tedds et al. 2.15 m and 0.5 e collected at a sampling rate of 200 Hz and at least 10. The ADV used in this study had four working section which is 3. [24]. where the origin is defined at the point where the Fig. The ADV had a cylindrical sampling volume set to The conditions under which the experimental tests were made 198 mm2 (height of 7 mm and diameter of 6 mm).5 m and 1 m at a number of heights. 10 MHz receiving elements positioned around a 10 MHz trans- vide a depth range of between 0.C.5 m/s with a measured turbulence intensity of 2%. and taken in the by Ref. these mean velocities is estimated to be better than 1%. the design of which is described in detail 0. interference. giving a blockage ratio of approximately 16% [22.5D Vectrinoþ) was used to measure velocities and turbulence statis.4 m wide and can pro. The probe was submerged in the flow and focused on a the possibility of a uniform velocity profile ranging between location 50 mm away from the probe head to minimise flow 0.000 litres of water. Velocity To determine the thrust on the turbine a 50 kg strain gauge measurements were taken upstream of the turbine at distances of dynamometer was used. 8. S. It is estimated that this produced a of 50 mm between measurement locations. The statistical uncertainty in section.7 m long by 1. The near-wake was calibration which is accurate to about 1%. The turbine velocities were measured to ensure the mean velocities and tur- was located at a depth of 0.23]. .25 m. The flume uses a 75 kW motor-driven axial-flow impeller measuring velocity components in three dimensions using the to circulate 80. 0. Data was were 0. The turbulent kinetic energy at a depth corresponding to the centre of the turbine. This provides mitter.425 m. from 1. / Renewable Energy 63 (2014) 222e235 229 in Fig.8 m water depth with uniform velocity in the range 0. The force block was calibrated by applying a load from spanwise direction across the width of the flume with a maximum 1 kg to 25 kg in steps of 1 kg. The water flows into the acoustic Doppler principle.03 m/s and 6 m/s. ADV is a well known technique [25] for simultaneously every 50 mm. and in the spanwise direction at least tics.000 individual 1.

e. similarly it 3.230 S. Fig. The mean transverse (vertical) velocities in Fig. l ¼ uR/U ¼ 6. There is a small scale swirling motions. This asym- are illustrated in Fig. and then the surface effects can be seen between y/D ¼ þ0. asymmetry due to the whole turbine structure. . 4(d) and (e). Fig.1. 67 diameters downstream). CT ¼ T/ locities corresponding to the clockwise rotation of the blades. 5(a) where there are both negative velocities and positive ve- this results in CP ¼ P/0. this is clearest in ously the torque was set to 20% of the motor’s maximum torque. 6 shows the mean spanwise (side-to-side) velocities.5 and y/D ¼ 0. 5 show the effect bine the optimum blade pitch angle was set (6 ).15.22  105. For all the measurements downstream of the three bladed tur. tween negative and positive velocities. Mean flow are all less than 10% of the streamwise velocities. 3.e. the regions of high absolute velocities occur generally at the higher turbulent regions Fig.5D. this 0. 4(a) this can be seen at z/ at Fig. showing the effect of two rotations of the large- wise velocity to about 80% of the upstream velocity. metry in Fig. measurement window (i. Fig. 4(b) and (c) can be observed by noting that the ve- locities are larger at y/D ¼ 0. z/D ¼ 0). As stated previ. this is reversed. which 3. These higher turbulent regions alter be- bladed turbine measured in horizontal planes. / Renewable Energy 63 (2014) 222e235 blades meet in the centre of the hub.34.25. in Fig. including body and support.25 than at y/D ¼ þ0. in particular when we look ity deficit is at the tip of the blades. at the edge of the blades. Results and discussion can be seen at y/D ¼ 0. 4 shows the mean streamwise velocities behind the three.5 and in Fig. 4(d). These measurement locations asymmetry in the results due to free-surface effects. again. (e) at the centre (i. subtracting the drag of the body and support alone without blades Significant transverse velocities are still evident at the edge of the at identical upstream velocities) and Re ¼ UR/n ¼ 2. 9. 6(b) between 2 and 3D downstream there are negative D ¼ 0. Tedds et al.5 in Fig.5rAU3 ¼ 0. The largest veloc. of the swirl induced by the turbine blade rotation.C.25D depths. further downstream diameters downstream there is an overall recovery of the stream.5rAU2 ¼ 1 (this value was determined by measuring the force on same trend can be seen at 0. By seven values on one side and positive on the other.5 the velocities are greater than at y/D ¼ þ0. The Lumley triangle showing the turbulence anisotropy downstream of the turbine.

u0 . / Renewable Energy 63 (2014) 222e235 231 3. v and w are the mean velocities at a particular location. In Fig.C. 8.5D downstream there is a vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ! third smaller peak at z/D ¼ 0. S. 7 show the streamwise Reynolds normal stresses (u0 ) at the (v and w ) exhibit similar variations and are not shown for centre height and up to seven diameter distances downstream. The conciseness. 2 (5) N i¼1 i To investigate the anisotropy of the turbulent flow field behind the turbine the Reynolds-stress anisotropy tensor b can be used and is defined by Choi and Lumley [26]. The key point to take from the data for the Reynolds normal stresses is that there is a 0 u02 1 u02 þv02 þw02  13 uv u02 þv02 þw02 uw u02 þv02 þw02 maximum effect around the tip of the blades (z/D ¼ 0. 8 show spanwise distributions of the normalised turbulent and w0 are defined by the standard deviation in each direction. 8(a). Illustrations of the shapes formed in different regions of the flow field by the Reynolds-stress anisotropy tensor (after Simonsen and Krogstad (2005)).3. v0 Fig. Tedds et al. vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !ffi u where TKE is defined in equation (1). which is probably due to the presence u u 1 X N v0 ¼ t of the stanchion. Reynolds normal stresses of the three velocity components. The other two velocity components 0 0 Fig. ðvi  vi Þ2 . 2 (3) TKE levels peak at the edges of the blades at z/D ¼ 0. it can be seen that the u 1 X N 0 u ¼ t ðu  ui Þ . Decay of turbulence anisotropy downstream of the turbine u u 1 X N 0 w ¼ t ðw  wi Þ .j ¼ B uv  13 C: (6) @ u02 þv02 þw02 u02 þv02 þw02 u02 þv02 þw02 A has an impact on the flow field at this location. (4) N i¼1 vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ! 3.2. (a) Isotropic tur- bulence (b) Axisymmetric turbulence x > 0 (one large component) (c) Axisymmetric turbulence x < 0 (two large component) (d) One component turbulence (e) Two-component turbulence (axisymmetric) (f) Two-component turbulence. The Reynolds uw vw w02  1 normal stresses emanating from the blade tip can be expected to be u02 þv02 þw02 u02 þv02 þw02 u02 þv02 þw02 3 Fig. kinetic energy (TKE) downstream of the three bladed turbine. .5 and can still N i¼1 i be observed in all measurements up to seven diameters down- stream. and the generation of tip vortices.5) and this B C B v02 vw C effect can be seen up to 7D downstream showing the turbine still bi. 10. it can be seen that at 1. where N is the number of samples at a particular point and u. In Fig. Turbulent flow field higher due to the blade speed being maximum at this point.

C. Emanating from this isotropic origin there are two limiting sides of [26e29]. h and x. The side of the modelled the turbine wake using standard turbulence models ‘Lumley triangle’ when x > 0. . flow to the anisotropy invariant map. most studies have the triangle which are said to be axisymmetric. characterised by a pair of invariants.232 S. i. Tedds et al. As discussed in the introduction. where two fluctuating components have tended to zero resulting in Fig. Therefore knowledge of the of the Reynolds stress tensor is dominating over the other two degree of turbulence anisotropy induced by the turbine and the components and the shape of the tensor can be said to be ‘rod- rate of return to isotropy is important. / Renewable Energy 63 (2014) 222e235 The Reynolds-stress anisotropy can be more simply. this situation corre- sponds to isotropic turbulence: all three diagonal components of 0 0 0 6h2 ¼ b3ii ¼ bij bjk bki . and are restricted with respect to the realisability of the commonly. (7) when both invariants are zero. (8) the Reynolds stress tensor are equal and therefore u ¼ v ¼ w . 9. and is stresses.e. defined by. is where a single diagonal component which assume isotropic turbulence. As x increases further this side leads up to the limit point turbulent flow these invariants can be defined from the Reynolds. 11. At any point and time in any shaped’. At the origin 6x ¼ b2ii ¼ bij bji . The decay of maximum mean velocities downstream of the turbine. which is often referred to as 2 the ‘Lumley triangle’ [27] which is shown in Fig. x ¼ h ¼ 0.

As h increases along this side until the to observe the effect around the ‘outside’ of the turbine.C. The other limiting line of of the different possible ellipsoid shapes formed by the diagonal the ‘Lumley triangle’ when x < 0. and the shape of the tensor is undertaken to compare the effect of the turbine and the outer areas said to be ‘disc-shaped’. To aid in the visualisation of the eight turbulence anisotropy flow is mainly in the so-called ‘disc-shaped’ region. shown in Fig. where two of shapes (i. 10. 9(a). but further Fig. 9(a) shows the downstream data closest to the turbine. further downstream and Fig. disc-shaped etc) a graphic representation the diagonal components of the tensor are dominant. 12. dominant turbulent components (u and w ). 9(b) and (c) show that the tensor. . has been third smaller diagonal component. 9. is where two equal diagonal components of the Reynolds-stress tensor are shown in Fig. rod-shaped. The decay of the maximum Reynolds normal stresses downstream of the turbine. The line connecting the one.e.and two-component axisymmetric Fig. / Renewable Energy 63 (2014) 222e235 233 a one-component state of turbulence. and turbulence is all other possible states of the Reynolds stress tensor there are larger values for x and h. components of the Reynolds stress tensor are dominating over the The Reynolds stress anisotropy data plotted in Fig. showing the flow is highly where there are only two non-zero diagonal components of the anisotropic. Plotted in smallest diagonal component has tended to zero the limit point at this way it can be seen that the upstream (inlet) velocity has two 0 0 the side’s end corresponds to a two-component state of turbulence. S. Tedds et al. (a) The maximum Reynolds normal stresses of the streamwise velocity at various heights behind a turbine (b) The maximum Reynolds normal stresses of the transverse velocity at various heights behind a turbine (c) The maximum Reynolds normal stresses of the spanwise velocity at various heights behind a turbine.

In the streamwise and spanwise directions (Fig.5. u and w become smaller).4. 13. However at 7D the turbulence is still linearly decrease at the different heights. 11(bec)) can be seen to essentially 0 0 (i. The decay of the maximum Reynolds shear stresses and TKE downstream of the turbine. The maximum of the transverse isotropic as the two dominant axisymmetric components decrease and spanwise velocities (Fig. / Renewable Energy 63 (2014) 222e235 downstream the flow gradually becomes. 12(a) and (c)).2. 11(a) shows the maximum velocity deficit in the streamwise closest to the turbine are approximately double in magnitude to direction at the five different heights behind the turbine. more law with a decay exponent z1. . The centreline component of power law. as expected. such as The maximum of the Reynolds normal stresses of the three- the kε. where the values Fig. but there is no general strongly anisotropic and the effect of the turbine is still being felt. will always fail to capture the near-wake turbulent flow dimensional velocity field is shown in Fig. in the uv plane and vw plane. further down. kinetic energy is shown in Fig. Wake decay characteristics maximum Reynolds normal stresses decay at a similar rate with a power-law decay exponent of approximately 0.234 S. 13. the data that of the furthest measurement (7D downstream). Tedds et al.C. trend to the heights which have the maximum deficit. for different heights exhibit a similar trend. All the shear stresses decay with a stream the recovery is much slower. The maximum of the Reynolds shear stresses and turbulent ameters behind shows a rapid velocity recovery. and by four disc di- the streamwise velocity can be seen to be recovering with a power ameters downstream have decayed to a minimum level of 20% of Fig. the first three di.e. the 3. decrease in the maximum values further downstream. 12 and shows a field correctly. Thus turbulence models which assume turbulent isotropy.

Cabrera R. Strain 1983. Rep. Philos exponent most commonly found to be around 1. Energy 2012. Collings R. Bode SP. Interactions between tidal turbine as TKE (TI ¼ ð2=3ÞTKE) as shown in Fig. In: Ninth turbine has been determined experimentally using detailed three. [9] Rose S.D.3 [28]. The return to isotropy of homogeneous turbulence. et al. Southampton.18:123e75. 2011. UK 2011. tional fluid dynamics for the use of a turbine for extracting energy from the tide. [22] Mason-Jones A. Experimental study of turbulent flow in the lent kinetic energy (well fit by a power-law with an near wakes of single and tandem prisms. H. Such dif. Near wake properties of horizontal axis marine current ferences would indicate that simplified methods to model the wake turbines. Chick JP. [4]. Southampton. Southampton. Najafian G. Ph. Rossiter J.19(7):27e30. Experimental analysis of the flow field around horizontal [27] Lumley J. The authors would like to thank A. Turbulent stress invariant analysis: clarification of from tidal currents. Mason-Jones A. Southampton. University of Sydney. [10] Maganga F.6 dependent on exact CT value for example. Akinlade OG. Acoustic-Doppler velocimeter (ADV) for References laboratory use. Mason-Jones A. O’Doherty. D. / Renewable Energy 63 (2014) 222e235 235 the maximum level at two disc diameters downstream. Byrne CB.4(6.371:20120246.B. Comparison between CFD simulations and experiments for predicting the far wake of horizontal axis maximum turbulent kinetic energy decays with a power law with tidal turbines. Crespo A. The experimental pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi data of Harrison et al. Trans Royal Soc 2013. In: Ninth also like to thank H. Uppsala. [25] Lohrmann A. SI):498e509. Owen I. S. [17] Devarakonda R. Atcheson M. Phys Fluids 2005. Hahm T. Morris. McCombes T. United Kingdom Wave and Tidal Energy Study. UK 2011. et al. 820e9. Cardiff University. Fundamentals and advancements in hy- draulic measurements and experimentation 1994. Myers LE.37(2e3):218e27. Lee KH. Conclusions and outlook sation of flow effects on marine current turbine behaviour and on its wake properties. G. The design of a multi-purpose multi-component strain gauge dynamometer. 12033.4(6. stream of grids. O’Doh. Iss. Turbulent flows. Humphrey JAC. Mech 2001. such as the absorption disc Sweden 2009.371:20120293.44(1): erty. Cambridge University Press. O’Doherty T. Ocean Eng 1978. Finally the rate of decay of the maximum turbu. [3] Sun X. Rivoalen E.E. Performance assessment of a horizontal axis tidal turbine in a high velocity shear environment. Tedds et al.I. Sorensen JN.436:59e84. Germain G. Tur- of the Reynolds stress anisotropy tensor confirms the strong degree bulent wake of a finite circular cylinder of small aspect ratio. 2010. [15] Adaramola MS. Acknowledgements Experimental analysis of a model horizontal axis tidal turbine. Lumley JL. Whelan JI. Prickett and R. Mason-Jones. editor. systematic data set which may be used by numerical modellers to [14] Daly T. European wave and tidal energy conference. measurements. et al. Pinon G. [26] Choi KS. Prog kε) be avoided. Ayre RG. vol. [1] Department of Energy and Climate Change. decay rates of 1. [11] Mycek P. Jo CH. J Fluid 2011. King J. Myers LE. Laboratory-scale simulation of energy extraction [29] Simonsen AJ. Tech.371:20120159. Sitzki L. . Non-dimensional scaling of tidal stream turbines. Bahaj AS. wake of a tidal turbine. Grosvenor for design and production of the turbine. IET Renew Power Gener 2010. Investi- Thus at identical CT they predict a TKE decay rate almost twice as gating experimental techniques for measurement of the downstream near high as that measured here. In: 15th Australasian fluid mechanics conference. In: complex structure of both the mean velocity field and the turbulent Ninth European wave and tidal energy conference. Harrison ME. Sweden 2009. Southampton. [28] Pope SB. McCormick for help with ADV European wave and tidal energy conference. Philos Trans Royal Soc 2013. O’Doherty T. MacKinnon P. Modelling of the flow field surrounding tidal benchmark their results against for the simplified case of a single turbine arrays for varying positions in a channel. Feng T. [20] Mason-Jones A. Bahaj AS. 3D-simulation of the turbulent wake behind a wind turbine. we would [23] Tedds SC. Uppsala. Australia: The stream distances than could be obtained in the current experi. O’Doherty DM. suggesting that pre. SI):613e27. 13(d). Feasibility study using computa- predict the TKE decay rate of HATT wakes. 2010. 75. UK 2011. gests the use of turbulence models based on isotropy (such as the [16] Vermeer LJ. These measurements reveal the wakes: small scale experimental and initial computational modelling. Determination 2013. Experimental characteri- 4. exhibit power-law wakes: experimental study of a group of three-bladed rotors. HATT. may significantly over [19] Egarr DA.17(3): exponent 0. Krogstad PA. thesis. Owen I. In: Eighth European wave and tidal energy conference. flow field induced by the presence of the turbine. They provide a [13] Carbon Trust.94) is significantly different to that observed down. 2004. In: Eighth European wave and tidal energy conference. Renew Energy 2008. In: Ninth European the maximum turbulent kinetic energy downstream of a model wave and tidal energy conference. Bahaj AS. Byrne CB. mental facility are required to fully test this hypothesis. which neglected swirl effects and 2007. In: Science of making torque from windJ Phys Conf Ser. 2005. In: Ninth European wave and tidal energy conference. Byrne.17(8):88e103. Good A. Bergstrom DJ. Computational modelling of tubulent flows. Also. Int J Heat Fluid Flow 1996. Batten WMJ.W. Tidal turbine dimensional ADV measurements. O’Doherty DM. Bryden IG. Bahaj AS. P. p. Feng T. Wind turbine wake aerodynamics. Rivoalen E. an exponent of z0. Wang Y. Whelan J. J Fluid Struct of turbulence anisotropy induced by the blade rotation and sug. [6] Myers L. C. 219e27. Sp. Kraus NC.3e1. Experimental investigation of horizontal axis tidal stream turbines. UK 2011. Accuracy of the actuator disc-RANS grid-generated (or mesh generated) turbulence which has a decay approach for predicting the performance and wake of tidal turbines. existing terminology. Germain G. Prickett PW. Collings R. Johnstone C. Eyre and T. Adv Appl Mech axis tidal turbines by use of scale mesh disk rotor simulators. Ordonez S. Philos Trans Royal Soc turbine in a confining flume (blockage ratio ¼ 16%).C. Sydney. may significantly over predict the rate of decay of [7] Stallard T. Bahaj AS. Interations between tidal turbine wakes: experimental study of a group of 3-Bladed rotors. 2006. Sumner D. The [4] Harrison ME. UK 2011. [21] O’Doherty T. T. Aerosp Sci 2003. [2] Myers LE. et al. Pinon G. characteristics of tidal stream turbines. [18] Wussow S. Gaurier B. IET Renew Power Gener 2010. Morris T.M. Numerical and experi- The near-wake (x/D  7) of a model horizontal axis tidal stream mental study of the interaction between two marin current turbines. Data at further down. Variablity of UK marine resources. Final Report.9. approach [2e4]. Schenstead AJ. [24] Millward A. Morris CE. meshes or perforated disks. modelled the turbine by absorption discs. Poole RJ. when replotted [8] Stallard T. C. In: Pugh CA. Johnstone C.39(6e7):467e510. which is a slower rate than that of isotropic [5] Batten WMJ.33(6):1267e74. UK renewable energy roadmap. [12] Rose S.22(6e7):919e28. vious modelling approaches.