Wind environment at a roof-mounted wind turbine on

This article was downloaded by: [University of Guelph] On: 28 April 2014, At: 06:28 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Sustainable Energy Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/gsol20

Wind environment at a roof-mounted wind turbine on a peaked roof building a

a

Rohan Hakimi & William David Lubitz a

School of Engineering, University of Guelph, Guelph, Ontario, Canada, N1G 2W1 Published online: 23 Apr 2014.

To cite this article: Rohan Hakimi & William David Lubitz (2014): Wind environment at a roofmounted wind turbine on a peaked roof building, International Journal of Sustainable Energy, DOI: 10.1080/14786451.2014.910516 To link to this article: http://dx.doi.org/10.1080/14786451.2014.910516

PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/termsand-conditions

International Journal of Sustainable Energy, 2014 http://dx.doi.org/10.1080/14786451.2014.910516

Wind environment at a roof-mounted wind turbine on a peaked roof building Rohan Hakimi and William David Lubitz∗

Downloaded by [University of Guelph] at 06:28 28 April 2014

School of Engineering, University of Guelph, Guelph, Ontario, Canada, N1G 2W1 (Received 22 January 2014; final version received 26 March 2014) Knowledge of the wind resource above peaked roofs is necessary to determine whether installing small wind turbines on low-rise peaked roof buildings is feasible. The wind characteristics at a representative peaked roof barn in southern Ontario, Canada were investigated using a boundary layer wind tunnel and computational fluid dynamics. Field measurements at the barn were collected using sonic anemometers and compared with the simulation results. Wind speed amplification was confined to a region immediately above the roof and was relatively low for wind energy purposes. The presence of nearby trees or buildings adversely impacted wind speed amplification. Considering only wind-related factors, the placing of microwind turbines on roof peaks may be warranted. However, if sufficient space is available, it is recommended to place small turbines on a tower rather than on the peaked roof of a low-rise building. Keywords: small wind turbine; peaked roof; wind resource assessment; wind tunnel; wind speed; turbulence

Introduction A tower tall enough to rise above surrounding obstacles into higher speed winds is one of the largest costs of installing a small wind turbine. It has been suggested that total costs of small wind turbine installations could be reduced by placing the turbine on the top of an existing tall structure, such as a barn or storage building (Mann et al. 2006). Additionally, wind speeds over some parts of a building roof can be higher than wind speeds at a similar height in an open area. A wind turbine located in a favourable location on a building roof could potentially produce more power than a comparable tower-mounted turbine at the same height above ground. Roof-mounted small wind turbines will take longer to recover their embodied energy than larger, utility scale turbines; however, it is possible for roof top turbines to be net energy producers on a lifecycle basis (Mithraratne 2009). There are practical challenges associated with roof-mounted wind turbines. Wind resource assessments for roof-mounted wind turbines have often over-predicted wind speeds near the roof (Walker 2011). This was identified as a significant issue during the Warwick Wind Trials, a field study of 26 roof-mounted wind turbines in the UK (Encraft 2009), where virtually all studied turbines produced less power than predicted (James et al. 2010). Aesthetics, noise and vibration were also significant problems in some installations, almost all of which were on residential ∗ Corresponding

author. Email: [email protected]

© 2014 Taylor & Francis


2

R. Hakimi and W.D. Lubitz

buildings (Encraft 2009). There are currently few safety standards for rooftop turbines or their installation, which has been identified as a critical area for immediate work as rooftop turbines become more common (Smith et al. 2012). While these findings suggest placing wind turbines on residential buildings might not be advisable, concerns related to noise, safety and vibrations would be reduced for other non-residential structures (Grauthoff 1991). The amount of additional power generated by a roof-mounted wind turbine depends strongly on the wind regime above the building, which can be quite complex (Mertens 2005). Additional research on the rooftop wind field is needed, particularly on those aspects most relevant to wind turbine installations, such as mean wind speeds and turbulence intensities. Most prior studies in this area have been limited to buildings with flat roofs or examined a building primarily designed to integrate a large wind turbine (Campbell et al. 2001; Mertens 2003). More recent studies have examined buildings with peaked or curved roofs, which are much more common in rural and suburban areas than flat roofs. Heath, Walshe, and Watson (2007) used computational fluid dynamics (CFD) to study the wind resource that would be experienced by roof-mounted wind turbines on a single typical house in the open and surrounded by other similar houses, and noted that wind speed amplification effects seen when the house was in the open were almost entirely mitigated by placing the house within an array of similar houses. Watson, Infield, and Harding (2007) used a similar approach to derive a guideline for predicting maximum speeds above a particular house geometry, but noted that the uncertainty of even this specific model would be very high due to the many variables that cannot be captured in a simplified guideline. Ledo, Kosasih, and Cooper (2011) used CFD to study the wind energy potential above pitched, pyramidal and flat roof building forms within arrays of similar buildings, including several wind directions, and concluded that flat roofs had more favourable wind regimes when a range of directions were considered. Characterisation of rooftop wind regimes is still needed (Heath, Walshe, and Watson 2007). This study utilised a combination of field measurements, boundary layer wind tunnel (BLWT) experiments and CFD simulations to examine the feasibility of placing a small wind turbine on a non-residential peaked roof building in a rural setting. Methods The wind field above a case study building was investigated using a BLWT, and steady-state, three-dimensional CFD simulations. Results of the simulations were compared with field data collected using sonic anemometers. Site The study building is a storage barn in a rural area of southwestern Ontario, Canada (43.30◦ N, 80.55◦W), with a rectangular footprint 18.34 m by 24.43 m, and a height of 7.32 m. The roof is peaked with a slope of 1:3, and roofline oriented north-south. A second barn 7.62 m high, 30.48 m long and 15.24 m wide with a half-cylinder cross-section and long axis oriented east-west is located 14.2 m east of the study building. There is a small shed adjacent to the southeast corner of the study building that is 2.60 m tall at its peak and has a footprint of 4.80 m by 6.72 m. A dense row of pine trees aligned east-west, with heights to 12 m, is located south of the study building. The layout of the site is shown in Figure 1. Field experiments For the field experiments, a 4.1 cm diameter tubular vertical mast was placed at the roof peak midway along the roofline. A guyed mast was placed 44.2 m west of the west wall of the barn in a

International Journal of Sustainable Energy


Figure 1.

3

Site plan for the study barn vicinity.

flat, open alfalfa field. A Kaijo DA-600 sonic anemometer was placed on a boom projecting from the west side of the mast at 4.9 m to measure the approaching wind. A second DA-600 was placed on the roof mast at a height of 1.9 m above the roof peak, on a boom projecting 0.75 m west of the mast. DA-600 anemometers are extremely sensitive, but only accurately measure winds over a direction range of 60◦ . Both the anemometers were oriented horizontally and pointed west (into the prevailing wind direction), allowing winds from 240◦ to 300◦ to be accurately recorded. Prior to the collection of field data, all of the sonic anemometers were calibrated in the University of Guelph BLWT. They were also placed in an open field west of the study building at the same location as the guyed mast for one week at a height of 1.0 m. It was found that the anemometers agreed within 5% over a wind speed range of 2–8 m/s. Both anemometers were connected via a data acquisition board to a desktop PC located inside the barn. A total of 347.5 h of field data were collected over 19 days. Three components of wind speed plus temperature were simultaneously sampled at 20 Hz. These raw data were then post-processed to 10 min averages using MATLAB scripts. The DA-600 sonic anemometers are only accurate when the wind direction is within 30◦ of the long axis of the instrument. The 10 min average readings were filtered based on instantaneous wind directions to remove those with significant readings outside this range. A total of 100 ten-minute averages were used for analysis. Wind tunnel tests A 1:96 scale model of the study area was tested in the University of Guelph BLWT to determine the variability of wind speed and turbulence above the peaked roof. The area modelled included all of the buildings. The trees south of the study barn were modelled using small pieces of foam affixed to vertically oriented stiff wires. The model was centred on the wind tunnel turntable and rotated to simulate different wind directions. The University of Guelph BLWT is an open-return suction tunnel with a converging intake, flow straighteners and an 8 m flow development section followed by a 2 m test section (Figure 2). The test section is 1.2 m wide, 1.2 m tall, and was outfitted with a traversing system and a TSI IFA300 hotwire anemometry system. The hotwire was calibrated using a sensitive manometer and pitot tube. Output from the hotwire system was recorded at 1000 Hz using a National Instruments USB-6210 data acquisition system and LabView 8.0 software. Spires and roughness elements were utilised to produce a 30 cm deep boundary layer with a mean velocity shear exponent of α = 0.17 (Figure 3). Wind speed and turbulence intensity profiles were measured above the centre of the roof peak (at the location of the roof mast in the field study) for 16 equally spaced wind directions. Turbulence

4


Figure 2.

University of Guelph boundary layer wind tunnel. Dimensions in metres.

Velocity (m/s) Turbulece Intensity (%)

300

250

Height (mm)


350

200

150

100

50

0 0

2

4

6

8

10

12

14

16

18

Figure 3. Wind tunnel wind speed and turbulence intensity profiles.

intensity was calculated as the standard deviation of the wind speed over the sampling period divided by the mean wind speed. A series of profiles were also taken along an east-west transect bisecting the barn, while simulating a west wind.

CFD simulations CFD simulations of the case study barn were conducted using the commercial CFD software Fluent 6.2.16. (Fluent Inc., Lebanon, NH, USA). For this study, Fluent was used to solve a finite volume discretization of the steady-state, three-dimensional Reynolds-averaged Navier–Stokes (RANS) equations. The equations were solved iteratively, with velocity components calculated using a second order upwind discretization. The SIMPLE algorithm was used for velocity and pressure coupling. Standard k–ε, realisable k–ε, renormalisation group (RNG) k–ε, and the Reynolds stress turbulence closure models were used. Additional information on the implementation of these methods is given in Fluent (2005).



5

Figure 4. Fluent simulation domain geometry. The inner rectangular zone is finely meshed from the inlet on the right to the outlet on the left. The inner cylinder is more finely meshed and contains the model.

The final domain size of 605 m long, 420 m wide and 107 m high was selected to prevent flow distortion due to blockage, which can occur if blockage exceeds 3% (Mochida, Tominaga, and Yoshie 2006) and provide sufficient distance for wake reattachment (Franke et al. 2004). The domain was divided into several regions as outlined in Figure 4. In practice, the domain used for the simulations was made up of two distinct volumes, an inner cylindrical region centred on the central point of the barn and an outer region consisting of the remainder of the domain. The inner region was rotated inside the outer region to simulate an approaching wind from different directions. This approach allowed simulation of different wind directions while keeping the overall flow aligned with the domain. Hexahedral cells are known to produce lower truncation errors and exhibit better iterative convergence than tetrahedral cells near surfaces (Franke et al. 2004). However, attempts to mesh the two-region domain entirely or partially with hexahedral cells were either not successful or did not produce converged solutions because of problems at the interface between the inner and outer regions. Therefore, a non-uniform tetrahedral mesh was used throughout the domain. Sizing functions were used to vary the mesh density within each region of the domain. A nominal cell size of 0.5 m was used adjacent to the study barn, with 0.34 m cells adjacent to the eaves (Figure 5). This resolution was established based on established wind engineering guidelines (Franke et al. 2004; Mochida, Tominaga, and Yoshie 2006). A series of initial simulations confirmed that a further reduction in cell size had a minimal effect on the velocity profiles above the roof or the velocity distribution in the downstream wake. The cell size increased with height and lateral distance from the barn, to a maximum cell size of 15 m adjacent to the sides at the top of the domain. The expansion ratio between adjacent cells was kept below the maximum value of 1.3 as suggested by Franke et al. (2004) to minimise truncation errors. Solutions were iterated until residual values of velocity components, pressure and turbulence parameters reached less than 1 × 10−6 . No significant convergence issues occurred during simulations.


6


Figure 5. Mesh resolution near buildings. Study barn is lower left. Porous jump region used to simulate trees is on the right. North is to the left.

Boundary conditions The ground and barn were modelled as solid surfaces, with building surfaces assumed to be aerodynamically smooth. Based on the agricultural fields surrounding the site, the ground surface roughness height was estimated to be zo = 0.04 m (ESDU 1976). In Fluent, surface roughness is modelled as sand grain roughness Ks . For converting between the two surface roughness scales when using wall functions in Fluent, Blocken, Stathopoulos, and Carmeliet (2007) suggest 9.793z0 Ks = , (1) Cs where the roughness constant Cs = 0.5. A value of zo = 0.04 m gives Ks = 0.78 m. However, when using wall functions, the first computational node, which is at half the height of the cell, should be farther from the wall than Ks (Franke 2006). This implies that cells adjacent to the ground should have a height of at least 1.56 m; however, cells this large would have insufficient cell resolution to capture the flow features adjacent to the face of the barn. It was previously noted that to ensure sufficient resolution at the barn, a cell size adjacent to the wall of 0.5 m was needed. To maintain this cell size, a reduced roughness value of Ks = 0.24 m was used in the simulations to meet both wall function and resolution criteria, at the cost of using a lower than ideal surface roughness equivalent to zo = 0.01 m. Correct simulation of the atmospheric boundary layer requires longitudinal equilibrium throughout the volume, which can be difficult to achieve (Blocken, Stathopoulos, and Carmeliet 2007; Hoxey and Richards 1993). The inlet boundary conditions in initial simulations used profiles for mean velocity U, turbulent kinetic energy k and turbulence dissipation ε from Richards and Hoxey (1993) U∗ z + zo U= ln , (2) κ zo


U ∗2 k=√ , Cμ


ε=

U ∗3 , κ(z + z0 )

7

(3) (4)

where κ is von Kármán’s constant (taken as 0.4), Cμ is the k–ε model constant equal to 0.09, U ∗ = 0.364 m/s and zo = 0.04 m, corresponding to a power law velocity profile with shear exponent of α = 0.17. However, simulations with these conditions in an empty domain meshed equivalently to that used in the building simulations resulted in values of k at the downwind outlet reduced by over 50% relative to the inlet values. Other investigators have reported similar occurrences (Hargreaves and Wright 2007). This longitudinal non-equilibrium was addressed by using an iterative process to evolve an equilibrium boundary layer. Profiles of k and ε from the outlet of an empty-domain simulation were reapplied at the inlet in a subsequent simulation (while reapplying Equation (2) for the velocity profile). Profiles of U, k and ε reached equilibrium throughout the domain after three iterations of this process. The resulting inlet profiles, with length and time units of metres and seconds, respectively, were k = −1.007 × 10−11 z6 + 3.603x10−9 z5 − 5.006x10−7 z4 + 3.345x10−5 z3 − 1.030x10−3 z2 + 7.556x10−3 z + 2.748x10−1 , ε=

0.1014485626 , z

(5) (6)

and Equation (2). These profiles were imposed at the upwind face of the domain using a userdefined function in Fluent for the building simulations. Zero normal gradient boundary conditions were applied at the top and sides of the domain. The velocity in the direction normal to the face was set to zero, as were the gradients of all variables in the direction normal to the face. This ‘symmetry’ boundary condition (gradients normal to the surface are zero at the surface) has been used successfully and widely used in wind simulations (Franke 2006). At the downwind face, conditions of constant static pressure and zero gradients in the direction normal to the face were imposed.

Trees At the study site, a dense row of tall pine trees south of the barn forms a wall-like wind break approximately 12 m tall with a roughly constant optical porosity of 0.5 from top to bottom. Given the proximity of the trees to the south side of the barn, it was necessary to include them in the simulation as they would impact the wind field for the north and south wind directions. The most common methods of including trees in larger simulations when using CFD is to approximate them as porous ellipsoids or cones (Groß 1993). Modelling a dense row of individual trees would require a highly refined mesh in the vicinity of the trees, and the approximated shape would not give a physically correct simulation of a windbreak. Instead, it is common to model dense rows of trees as a single volume (e.g. Gromke et al. 2008). Mochida, Tominaga, and Yoshie (2008) review models that account for the effect of trees on the production and dissipation of turbulent kinetic energy, and ideally this effect would be incorporated into the simulation. However, a further constraint for this study was a desire to model the windbreak in a straightforward manner that could be easily applied in a commercial CFD code, without adding additional terms to the k and ε equations. Since the row of trees was dense enough to form a windbreak of roughly constant porosity and most CFD codes incorporate models for flow through porous media, it was

8



elected to model the row of trees as a single planar porous surface considering only the momentum deficit generated by the trees. The relative simplicity of this approach, and the resulting higher uncertainty in the predicted flow fields were felt to be acceptable due to the minor impact of the row of trees on most wind directions, and the lack of field data to validate more sophisticated modelling. Wind speed reduction by a porous surface depends on the size and porosity of the surface and the approach wind speed. The row of trees were estimated to have a drag coefficient Cd of 0.8, based on data from Grant and Nickling (1998), which gives the drag force on the row of trees Fd for a wind speed U as (7) Fd = 0.5Cd ρAU 2 , where air density ρ = 1.225 kg/m3 and the area of the porous surface is A = 404 m2 . An approach wind speed of U = 3 m/s results in an estimated drag force of Fd = 1780 N. The porous surface was implemented using the Fluent ‘porous jump’ model. The pressure drop P across the face of a uniform porous surface of finite thickness w is calculated using a modified version of Darcy’s law (Fluent 2005) μ 1 2 P = − (8) U + CPJ ρU w, φ 2 where air density ρ = 1.225 kg/m3 and viscosity μ = 1.8 10−5 N s/m2 . Permeability φ and constant CPJ are characteristics of the surface. At atmospheric scales, the permeability term is negligibly small, and P is a function of squared velocity that scales with the product of w and CPJ . Defining w = 1.0 m and performing iterative CFD simulations in an otherwise empty domain with the profiles of U, k and ε used for the barn simulations resulted in a value of CPJ = 0.773 m−1 to produce a drag force of Fd = 1780 N. While this approach was felt to be sufficient for this particular site and situation, a more sophisticated approach would be required at sites where trees would be expected to have a significant impact on the local wind resource.

Results All wind speeds were non-dimensionalised by a reference mean wind speed measured at a height of z/H = 2, where H = 7.32 m is the height of the study barn at the roof peak. Normalising velocities for the BLWT and CFD simulations were obtained by measuring the wind speed at the location of the barn in an empty domain or test section. Single wind direction: west wind The prevailing wind direction at the site is from the west, which corresponds to winds that are perpendicular to the roofline on the study barn. Figure 6 shows mean wind speed profiles predicted by the BLWT and CFD. The turbulence closure models used were standard k–ε, realisable k–ε, RNG k–ε and the Reynolds stress model. Standard default values were used for the constants in each model. Some flow features were consistently predicted by all the simulations. All simulations predict that wind speed amplification is greatest at the roof peak. A significant increase in wind speeds occurs in a limited region above the roof only a few metres deep (Figure 6). On the windward roof surface, mean velocities close to the surface (z/H < 05) are increased at a given height compared with the upwind profile (profile 1). The simulations converge to very similar wind speed profiles as


9

3

2.5


Height (z/H)

2

1.5

1

0.5

Wind tunnel Standard K-epsilon RNG K-epsilon Realizable K-epsilon Reynolds stress model

0 Non-dimensionalized velocity Figure 6. West wind non-dimensionalised mean velocity profiles along an east-west transect through the centre of the study barn. One horizontal grid spacing represents a non-dimensionalised velocity of 2. Data presented are from Fluent simulations using different turbulence closure models and the wind tunnel. Left side of the figure is upwind. Profile locations from left are (1) x/H = 2.52, (2) x/H = 1.89, (3) windward edge of roof at x/H = 1.26, (4) x/H = 0.63, (5) roof peak at x/H = 0, (6) x/H = −0.63 and (7) leeward edge of roof at x/H = −1.26.

height increases above approximately z/H = 1 (or lower for the upwind profiles). On the upwind side of the barn, there is some difference between the BLWT and CFD simulations adjacent to the ground, especially at the most upwind profile. This is due to reduced resolution of the CFD model in this location and the lower simulated surface roughness, relative to the wind tunnel measurements. The variation between the simulations is greatest in the region very close to the roof peak, and in the wake region above the leeward roof surface. Wind tunnel results show increasing velocities up the windward side of the roof, separation at the roof peak and a large wake region above the leeward roof. Of the CFD simulations, only the RNG k–ε closure predicted wind profiles above the leeward roof qualitatively similar to the wind tunnel measurements. The results from the CFD simulations using standard and realisable k–ε models did not predict separation at the peak, and predicted very shallow wake regions above the leeward roof. This results in an overestimation of wind speed at the peak. The Reynolds stress model maintains a similar profile after the peak though the velocity increases more gradually above the roof. It has been shown that the standard k–ε model does not reproduce the separation and the reverse flow downstream of a sharp edge such as a building roof peak (Mochida, Tominaga, and Yoshie 2006). The results shown here appear to support this finding. The RNG k–ε model was in closer agreement to the wind tunnel (and as will be shown below, the field measurements) than the realisable or standard k–ε models, and on this basis it will be the turbulence closure utilised in the remaining simulations and analysis.

10



Effect of wind direction on the rooftop wind regime Simulations investigating the wind speed at the roof peak as a function of wind direction were performed in the wind tunnel and using CFD with RNG k–ε closure. Figure 7 shows wind tunnel and CFD mean wind speed and turbulence intensity as a function of wind direction at four heights above the roof peak (h/H = 0.20, 0.26, 0.53 and 0.98, corresponding to actual heights of 1.46 1.92, 3.88 and 7.17 m above the roof peak, respectively). At the height closest to the roof, the wind speed shows the greatest variation with wind direction at the height closest to the roof (Figure 7(a) and 7(b)). Comparing the wind tunnel and CFD results, general qualitative features such as directions associated with peak and minimum values are in reasonable agreement; however, the magnitudes of mean velocity and turbulence intensity differ between the wind tunnel and CFD RNG k–ε simulations. This is likely due to differences in the simulated velocity profiles and turbulence levels. When the wind is parallel to the roof line, the wind speeds are lower at the peak than at the same height in an open area. While this effect is most pronounced for south winds, due to the wind speed deficit contributed by the wake of the trees upwind, wind speed deficit is almost as great for north winds. This suggests that, at least in the lowest few metres above the roof peak, additional turbulence is generated primarily by separation at the upwind gable end of the building itself, even when trees and other obstacles are upwind. Some of the difference between the CFD and wind tunnel simulations for the south wind case is due to differences in the modelling of the trees.

(a) NNW

1.2

h/H = 0.20 h/H = 0.26 h/H = 0.53 h/H = 0.98

N NNE

1 NW

(b) NNW NW

0.2

0

E

WSW

SW

0

W

SE SSW

WNW

W

SW

SE SSW

h/H = 0.20 h/H = 0.26 h/H = 0.53 h/H = 0.98

N NNE

SSE

(d) NNW

0.5

h/H = 0.20 h/H = 0.26 h/H = 0.53 h/H = 0.98

N NNE

0.4

NW

NE

NE

0.3 ENE

WNW

0.2

ENE

0.1 E

WSW

ESE SW

SE SSW

ESE

SSE

(c) NNW

E

WSW

ESE

NW

ENE

0.4

0.2

0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

NE

0.6 WNW

ENE

0.4

W

NNE

0.8

0.6 WNW

h/H = 0.20 h/H = 0.26 h/H = 0.53 h/H = 0.98

N

1

NE

0.8

1.2

SSE

W

0

E

WSW

ESE SW

SE SSW

SSE

Figure 7. Non-dimensionalised wind speed from (a) the wind tunnel and (b) Fluent (RNG closure), and turbulence intensity from (c) the wind tunnel and (d) Fluent (RNG closure) as a function of wind direction at four heights above the roof peak.



11

It is estimated that the ‘trees’ in the CFD simulation had greater porosity than the model trees used in the wind tunnel and this created less of a wind speed deficit downwind. As height above the roof increases, the predictions of the models converge as the effects of the trees decrease. The wind profile above the roof also converges towards the reference open country profiles as height increases. Overall, the turbulence intensity levels in the CFD simulations were roughly 40% less than corresponding wind tunnel measurements. Qualitatively, the features of the turbulence intensity predictions from the wind tunnel and CFD are similar (Figure 7(c) and 7(d)). The turbulence intensity above the roof peak is strongly dependent on wind direction. Winds blowing parallel to the roof line (north and south directions) are associated with elevated turbulence intensity. However, even a small deviation from parallel to the roof line substantially reduces turbulence intensity (and wind speed deficit). The increase in turbulence intensity for winds from the northnortheast, seen most prominently in the wind tunnel data (Figure 7(c)), is due to the roof peak being in line with the wake from the second barn for this wind direction. For winds parallel to the roof line at a height of h/H = 0.20 above the roof, both simulations predict turbulence intensities approximately four times greater than when winds are perpendicular to the roof line. This significant increase in turbulence appears to be caused almost entirely by the interaction of the wind with the building itself and should be taken into account if a wind turbine is being planned within the region close to the roof surface. This variation with direction decreases with height, and by h/H = 0.98 turbulence intensity as the function of wind direction is close to constant. The impact of the trees to the south and second barn to the east-northeast are reduced but still evident in the wind tunnel data at h/H = 0.98, while the impact of these nearby obstacles has almost disappeared in the CFD data at this height.

Comparison to field data Field data were collected for the case of a west wind. The velocity and turbulence intensity measurements from the rooftop anemometer (at 1.92 m or h/H = 0.26 above the roof peak) were compared to both the anemometer at the tower (4.9 m or z/H = 0.67 above ground) and to predictions from the simulations. Investigating the velocity ratio observed between the anemometers, it is apparent that the wind speeds at the roof anemometer were greater than those on the tower (Figure 8). It was judged that the mean velocity 1.92 m above the roof can be represented as 115% of the mean velocity at 4.9 m on the tower at wind speeds of interest for wind energy purposes (i.e. greater than 4 m/s). Turbulence intensities, although showing greater scatter than the velocity data, are about 15% lower at the anemometer above the barn roof (Figure 9). The highest turbulence intensities in the anemometer data are associated with the lowest mean wind speeds. Note that in Figure 9, the wind tunnel and CFD k–ε RNG results were virtually identical and are overlapping. The different simulation methods reproduced the field and wind tunnel data with varying degrees of accuracy that depended mostly on the method’s ability to reproduce the roof peak separation point and the wake region above the leeward roof surface. When modelling the barn with CFD, the k–ε RNG closure more accurately reproduced the field measurements than standard k–ε, for both velocity and turbulence. The standard k–ε model did not capture the separation occurring at the roofline, resulting in predictions of increased wind speed-up across the peak. A shear exponent α of 0.17 and a roughness height zo of 0.04 m were estimated for the alfalfa fields surrounding the barn. The wind tunnel spires and roughness elements were adjusted to reproduce this condition. A slightly lower value of zo = 0.01 m was used as a boundary condition for the CFD simulations. Unfortunately, the field measurements were inadequate for calculating

12


9 Wind speed at 4.9 m tower level (m/s)

Wind tunnel

7

Fluent standard K-epsilon Fluent RNG K-epsilon

6

1:1

5 4 3 2 1 0 0

1

2

3

4

5

6

7

8

9

Wind speed at 1.92 m above roof peak (m/s) Figure 8. Tower vs. barn mean wind speeds.

1 Turbulence intensity at 4.9 m tower level


Field data 8

Field data

0.9

Wind tunnel

0.8

FLUENT standard K-epsilon

0.7

FLUENT RNG K-epsilon 1:1

0.6 0.5 0.4 0.3 0.2 0.1 0 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Turbulence intensity at 1.92 m above roof peak Figure 9. Tower vs. barn turbulence intensities.

these parameters directly from the anemometer data, and it was impossible to verify if the simulated roughness values and profiles were representative of the field site.

Feasibility of roof-mounted wind turbines The goal of this study was to predict the difference in power output between a micro-wind turbine placed at the roof peak, and the same turbine placed on a conventional tower in an open area. For this analysis, the tower was considered to be in an open field sufficiently far removed from obstructions and that wind speed profiles are the same from all directions and follow the power law


13

160 140

Power (W)

120 100 80 60 40 20 0 –20


0

5

10

15

20

Wind speed (m/s) Figure 10.

Power curve of Southwest AIR-X Wind Turbine (From data in Van Dam et al. 2003).

relation. A Southwest Windpower AIR-X turbine having a rotor diameter of 1.14 m was chosen as a representative micro-wind turbine. The AIR-X power curve (Figure 10) utilised was measured at the National Renewable Energy Laboratory (Van Dam et al. 2003). A power law velocity profile with α = 0.17 was assumed at the tower. Wind speed was assumed to be Rayleigh-distributed with a mean wind speed at 10 m above ground (Uavg,10 ) of 7.0 m/s and ρ = 1.225 kg/m3 . For the west wind case (wind direction perpendicular to roofline), the ratio of mean wind speed 1.92 m above the barn roof (Uavg,W ) and mean wind speed 4.9 m above ground on the tower (Uavg,4.9 ) varied between field, wind tunnel and CFD results (Figure 8). Field data with wind speeds above 4 m/s suggest Uavg,W /Uavg,4.9 = 1.15, while wind tunnel results and CFD (k–ε RNG) are slightly lower and higher, respectively. The following analysis uses a consensus value of Uavg,W /Uavg,4.9 = 1.15. If Uavg,10 is known, the consensus value of Uavg,W /Uavg,4.9 and the power law give α 4.9 Uavg,W = 1.15Uavg,10 . (9) 10 Assuming Uavg,10 = 7.0 m/s gives the mean wind speed for a west wind direction 1.92 m above the roof as Uavg,W = 7.13 m/s. The power output of the wind turbine with a hub height 1.92 m above the barn roof in a west wind was calculated by numerically integrating the Rayleigh distribution over all hub-height wind speeds U, 2 ∞ π U π U PW = dU, (10) P(U) exp − 2 2 Uavg,W 4 Uavg,W 0 where P(U) is the power curve of the wind turbine (Figure 10). The wind power density per swept area for the west wind case (P/A)W was similarly calculated using 2 ∞ P π U U 3π dU. (11) = 0.5ρ U exp − 2 A W 2 Uavg,W 4 Uavg,W 0 If a steady west wind is the limiting case of wind always from a single direction, another limiting case is when the wind is equally likely to be blowing from any direction. It has already been shown that wind direction affects rooftop wind speeds (Figure 7(a) and 7(b)). The effect of wind direction on power output was calculated using a direction binning approach assuming that wind direction was evenly distributed between sixteen 22.5◦ wind direction bins. The wind speed

14


Table 1. AIR-X turbine power output and wind power density for a 7.0 m/s mean wind speed.

Location


Roof peak Tower Tower Roof peak Tower Tower

Wind direction

Height above ground (m)

Mean wind speed (m/s)

AIR-X power (W)

Wind power density (W/m2 )

West

8.24 8.24 11.2 8.24 8.24 6.35

7.13 6.77 7.13 6.83 6.77 6.48

53 50 53 48 50 48

424 363 424 430 363 318

All

distribution associated with a particular direction bin i was assumed to be Rayleigh-distributed based on an average wind speed for the bin of Uavg,i = Uavg,W (UWT,i /UWT,W ), where UWT,i and UWT,W are the wind tunnel measured mean wind speeds at 1.92 m above the barn roof, for direction i and the west direction, respectively. (Values of UWT are shown in Figure 7(a), h/H = 0.26.) The power output of the turbine (Pall ) is then 1 = 16 i=1 16

Pall

0

∞

2 π U π U dU, P(U) exp − 2 2 Uavg,i 4 Uavg,i

(12)

and the wind power density for evenly distributed wind directions is 2 16 P 1 ∞ 3π U π U = 0.5ρ dU. U exp − 2 A all 16 i=1 0 2 Uavg,W 4 Uavg,W

(13)

The AIR-X power output and (P/A) for both the tower and roof peak are given in Table 1. Mean wind speeds shown in Table 1 are for the specific corresponding locations and heights above ground. Wind speed-up occurs when the wind always blows from a direction perpendicular to the roofline with no upwind obstructions, corresponding to a west wind for the study barn. In this case, placing a wind turbine at 1.92 m above the roof peak (i.e. 8.24 m above ground) would produce the same power as placing it on a tower in an open area at a height of 11.2 m. This is due to amplification of the wind speed above the roof, and suggests that in this case placing the turbine in the rooftop location could be advantageous. Only a small fraction of locations, such as some tradewind sites or certain mountain passes, experience prevailing winds from a single direction. Alternatively, the wind direction could be assumed as uniformly distributed, meaning the wind comes equally from all wind directions. This effectively gives equal weight to all the wind effects at the roof top location associated with all wind directions. While this is no more physically realistic for most locations than a single wind direction, it represents a bounding case for comparison purposes. Table 1 shows that the overall turbine power output would be reduced if the wind direction was uniformly distributed, because this calculation includes wind directions associated with significant wind speed reductions at the rooftop location (e.g. when the wind is parallel to the ridgeline). The hub height of a tower-mounted turbine in an open area would only need to be 6.3 m to produce the same power as the turbine at the rooftop location with a uniform wind direction distribution, which is actually lower than the 7.32 m height of the barn. Additionally, for wind directions associated with reduced wind speeds above the roof (e.g. parallel to the roofline), the turbine will also be operating in greater turbulence intensity than it would experience on a tower, which could reduce turbine life and increase noise. The actual expected power output of an AIR-X turbine at the site was estimated using data from the nearby Elora Research Station weather station (43.642◦ N, 80.412◦W) for the calendar year


15

NORTH

15% 10% 5%


WEST

EAST

15 − 20 10 − 15 5 − 10 0−5

SOUTH Figure 11.

Measured 10 m wind rose at Elora Research Station 2006. Wind speeds in m/s.

Table 2. AIR-X turbine power output and wind power density based on year 2006 meteorological measurements at the Elora Research Station.

Location

Height above ground (m)

Mean wind speed (m/s)

Mean AIR-X power (W)

Wind power density (W/m2 )

Roof peak Tower Tower

8.24 8.24 14.2

4.12 3.87 4.25

20 16 20

113 85 112

2006. The measured average wind speed at 10 m above ground in the Elora data set was 4.00 m/s. The measured wind rose (Figure 11) shows that the prevailing wind direction is westerly. Hourly measurements of wind speed, direction, pressure and temperature were used with the AIR-X power curve (Figure 10) to predict density-corrected power output on an hourly basis, at both the rooftop location (8.24 m above ground) and on a tower in open country (Table 2). The calculation process was similar to the prior analysis, except that the measured hourly 10 m wind speeds and directions were used instead of assuming wind speed and direction distributions, and the power curve was linearly corrected based on the observed air density. The building orientation appears to be close to optimum for maximising power output of a rooftop wind turbine at this site. The distribution of wind directions (Figure 11) correspond to directions associated with wind speed-up at the rooftop turbine location (Figure 7(a)) and the rooftop turbine would benefit from a significant overall wind speed-up. The turbine would need to be placed on a tower 14.2 m high in an open area to achieve the same annual energy production (AEP) as the rooftop location (at 8.24 m above ground). The goal of any wind turbine installation is efficient electricity production. Table 2 shows that the predicted mean power output of the wind turbine based on actual wind measurements is 20 W, whether on the rooftop or a 14.2 m tower. This corresponds to an AEP of 175 kWh. The study site would be eligible for a provincial feed-in tariff rate of CAD $0.115/kWh (OPA 2014), which represents a revenue of CAD $20/year. A wind turbine with annual revenue this low is unlikely


16


ever to recoup initial equipment and installation costs. Such an installation would likely only be feasible in off-grid locations where supplying electricity is very expensive. Installing a larger turbine on the roof to increase power generation and therefore revenue would not be practical. Significant wind speed amplification occurs only in a region a few metres above the roof of a low-rise building (Figure 6), so larger turbines would benefit less from this effect. Installing a larger turbine could also be more expensive, since the added loads due to the turbine would likely require structural reinforcement of most low-rise buildings. The costs of mounting systems and structural reinforcements associated with rooftop wind turbine installations can in practice be greater than the cost of the rest of the wind turbine installation (Smith et al. 2012). Ambient wind speeds are typically low at the heights of low-rise buildings, and standard practice is to place even small wind turbines on towers of 15–30 m. At these heights (roughly z/H > 2) there is little significant difference in the wind resource between a stand-alone tower and a tower projecting upward from a peaked-roof building. Therefore, only micro-turbines with rotor diameters of less than a few metres could be expected to take significant advantage of wind speed amplification caused by a moderately sloped peaked roof such as the one investigated in this study. The vision of small wind turbines on low-rise building roofs is appealing, but is unlikely to be practical except in rare circumstances. Placing a wind turbine on a tower only a few metres higher than a building will usually result in similar power production, and the tower installation would minimise issues related to safety, structural loads and noise noted in the literature review. Since most low-rise peaked buildings are only a few storeys (i.e. less than 10 m), installing the turbine on a conventional tower of normal (e.g. 15–30 m) height would result in a greater power output, likely with lower installation costs and fewer complications related to safety and noise.

Conclusions The wind flow over a peaked roof barn was measured using sonic anemometers, and simulated in a BLWT and using the commercial CFD code Fluent. Wind tunnel and CFD simulations showed that the mean wind speed above the roof was increased in a region several meteres deep for most wind directions. Upwind obstacles, or wind directions parallel to the roofline, resulted in decreased wind speeds and increased turbulence intensity near the roof surface relative to reference open-field measurements. The different methods used to model nearby trees in the wind tunnel and CFD simulations influenced the predicted wind speeds above the barn roof. The proper modelling of tree wakes is likely to be the largest single source of uncertainty in this study. For similar situations where the wake of trees will have a significant impact on the building of interest, the use of tree models that consider impacts on turbulence generation and dissipation is recommended. When the field and simulation results were applied to predicting wind energy production potential above the roof, the wind direction distribution of the local winds was an important factor in predicting whether the wind turbine placed above a roof peak would require a lower overall height (including roof height) than a tower-mounted turbine to produce the same energy. Assuming a prevailing wind perpendicular to the roofline and unobstructed upwind of the building, a turbine mounted just above the roof would produce more energy than one at the same height on a conventional tower. However, if the wind blows regularly parallel to the roofline or from directions with upwind obstructions, the energy production of a roof-mounted wind turbine is significantly impaired compared with a one on a tower of a similar height in an open area. The expected power output of a rooftop AIR-X turbine at the study location was calculated to be a very low 175 kWh/year, based on one year of meteorological data from a nearby weather station. A rooftop turbine would not be economically feasible at the study location, even though


17

the building orientation and wind direction distribution maximise mean wind speed-up at the rooftop location. Performing simulations of roofs with varying slopes would provide a further insight into the effect of roof slope on the near-roof wind regime and wind energy production potential. This study would also benefit from the collection of field data over a longer time period, as the current data may not be an adequate representation of site conditions. Additional anemometers would allow the direct measurement of surface roughness and characterisation of wind profiles, and comprehensive comparisons between field and simulation data.


Acknowledgements The authors wish to thank Profs. Jon Warland and Bill Nickling of the University of Guelph for assistance meeting the equipment needs for this study. Several staff at RWDI Inc. contributed technical advice, including Michael Roth, Darryl Cann, Meiring Beyers and Hanqing Wu. Mr Hakimi was partially supported by a University of Guelph Undergraduate Research Assistantship and funding from the Natural Sciences and Engineering Research Council (NSERC) of Canada.

References Blocken, B., T. Stathopoulos, and J. Carmeliet. 2007. “CFD Simulations of the Atmospheric Boundary Layer: Wall Function Problems.” Atmospheric Environment 41 (2): 238–252. Campbell, N., S. Stankovic, M. Graham, P. Parkin, M. van Duijvendijk, T. de Gruiter, S. Behling, J. Hieber, M. Blanch. 2001. “Wind Energy for the Built Environment (Project Web).” Proceedings of the European wind energy conference, Copenhagen, Denmark, July 2–6. Encraft. 2009. “Warwick Wind Trials Final Report.” Encraft technical report. Accessed November 21, 2013. http:// www.warwickwindtrials.org.uk/ ESDU. 1976. The Response of Flexible Structures to Atmospheric Turbulence (ESDU Item No. 76001). 6th ed. London: Engineering Sciences Data Unit. Fluent 6.2: User’s Guide. Fluent Inc. 2005. http://www.fluent.com Franke, J. 2006. “Recommendations of the COST Action C14 on the use of CFD in Predicting Pedestrian Wind Environment.” The fourth international symposium on computational wind engineering (CWE 2006), 529–532. Yokohama, Yokohama, Japan, July 16–19. Franke, J., C. Hirsch, A. G. Jensen, H. W. Krus, M. Schatzmann, P. S. Westbury, S. D. Miles, J. A. Wisse, N. G. Wright. 2004. “Recommendations on the use of CFD in Wind Engineering.” Proceedings of the International Conference on Urban Wind Engineering and Building Aerodynamics. In: van Beeck JPAJ (Ed.), COST Action C14, Impact of Wind and Storm on City Life Built Environment. von Karman Institute, Sint-Genesius-Rode, Belgium, 5–7 May 2004. Grant, P. F., and W. G. Nickling. 1998. “Direct Field Measurement of Wind Drag on Vegetation for Application to Windbreak Design and Modelling.” Land Degradation and Development 9: 57–66. Grauthoff, M. 1991. “Utilization of Wind Energy in Urban Areas – Chance or Utopian Dream?” Energy and Buildings 16 (1–2): 517–523. Groß, G. 1993. “Numerical Simulation of Canopy Flows.” Berlin: Springer-Verlag. Gromke, C., R. Buccolieri, S. di Sabatino, B. Ruck. 2008. “Dispersion Study in a Street Canyon with Tree Planting by Means of Wind Tunnel and Numerical Investigations – Evaluation of CFD Data with Experimental Data.” Atmospheric Environment 42 (37): 8640–8650. Hargreaves, D. M., N. G. Wright. 2007. “On the Use of the k-ε Model in Commercial CFD Software to Model the Neutral Atmospheric Boundary Layer.” Journal of Wind Engineering & Industrial Aerodynamics 95: 355–369. Heath, M. A., J. D. Walshe, and S. J. Watson. 2007. “Estimating the Potential Yield of Small Building-mounted Wind Turbines.” Wind Energy 10: 271–287. Hoxey, R. P., P. J. Richards. 1993. “Flow Patterns and Pressure Fields around a Full-Scale Building.” Journal of Wind Engineering and Industrial Aerodynamics 50: 203–212. James, P. A. B., M. F. Sissons, J. Bradford, L. E. Myers, A. S. Bahaj, A. Anwar, and S. Green. 2010. “Implications of the UK Field Trial of Building Mounted Horizontal Axis Micro-Wind Turbines.” Energy Policy 38 (10): 6130–6144. Ledo, L., P. B. Kosasih, and P. Cooper. 2011. “Roof Mounting Site Analysis for Micro-Wind Turbines.” Renewable Energy 36 (5): 1379–1391. Mann, S., I. Harris, J. Harris. 2006. “The Development of Urban Renewable Energy at the Existential Technology Research Center (ETRC) in Toronto, Canada.” Renewable & Sustainable Energy Reviews 10 (6): 576–589. Mertens, S. 2003. “The Energy Yield of Roof Mounted Wind Turbines.” Wind Engineering 27 (6): 507–518. Mertens, S. 2005. “Wind Energy in the Built Environment.” Brentwood: Multi-Science Publishing. Mithraratne, N. 2009. “Roof-top Wind Turbines for Microgeneration in Urban Houses in New Zealand.” Energy and Buildings 41 (10)” 1013–1018.


18


Mochida, A., Y. Tabata, T. Iwata, H. Yoshino. 2008. “Examining Tree Canopy Models for CFD Prediction of Wind Environment at Pedestrian Level.” Journal of Wind Engineering & Industrial Aerodynamics 96 (10–11): 1667–1677. Mochida, A., Y. Tominaga, R. Yoshie. 2006. “AIJ Guideline for Practical Applications of CFD to Wind Environment Around Buildings.” The fourth international symposium on computational wind engineering (CWE 2006), 533–536, Yokohama. OPA. “FIT/microFIT Price Schedule (January 1, 2014).” Ontario Power Authority. Accessed February 23, 2014. http:/fit.powerauthority.on.ca Richards, P. J., R. P. Hoxey. 1993. “Appropriate Boundary Conditions for Computational Wind Engineering Models using the k-e Turbulence Model.” Journal of Wind Engineering and Industrial Aerodynamics 46, 47: 145–153. Smith, J., T. Forsyth, K. Sinclair, F. Oteri. 2012. “Built-Environment Wind Turbine Roadmap.” NREL/TP-5000-50499. Golden, CO: National Renewable Energy Laboratory. Van Dam, J., M. Meadors, H. Link, and P. Migliore. September 2003. “Power Performance Test Report for the Southwest Windpower AIR-X Wind Turbine.” TP-500-34756. Golden, CO: National Renewable Energy Laboratory. Walker, S. L. 2011. “Building Mounted Wind Turbines and their Suitability for the Urban Scale: A Review of Methods of Estimating Urban Wind Resource.” Energy and Buildings 43 (8): 1852–1862. Watson, S., D. Infield, and M. Harding. 2007. “Predicting the Performance of Small Wind Turbines in the Roof-Top Environment.” Proceedings of European wind energy conference, Milan, Italy, May 7–10.