Optimisation of Almería-type Greenhouse Ventilation Performance

Optimisation of Almería-type Greenhouse Ventilation Performance with Computational Fluid Dynamics F.D. Molina-Aiz, D.L. Valera and A.A. Peña Dept. of Rural Engineering University of Almería Carretera de Sacramento s/n 04120 Almería Spain

J.A. Gil Dept. of Rural Engineering E.T.S.I. Agrónomos y de Montes University of Córdoba Avda. Menéndez Pidal s/n 14004 Córdoba Spain

Keywords: Finite Elements Method, airflow, modelling, insect-proof screens, climatic control Abstract The aim of the present study was to evaluate effects of different parameters that influence the greenhouses ventilation, and recommend to the growers and greenhouses manufacturer companies the optimal designs. The parameters studied in the different simulations were: greenhouse width and span number, vent area, location and type of vent openings, presence of insect-proof screen with different porosity and incidence of plants. The effect of these parameters on the natural ventilation of an Almería-type greenhouse is analysed by means of Computational Fluid Dynamics (CFD), using the commercial program ANSYS/FLOTRAN v8.0 based on the Finite Elements Method. Reduction of 88% in the ventilation rate were observed when the spans number increased from 1 to 5, as consequence of the increase of the greenhouse width (9 to 45 m). The use of insect proof screens decreased around 50% the speed in the next area to the lateral windows and ventilation rate. The presence of a crop inside the greenhouse causes an effect of airflow lamination and reduces the airflow at the ground level, while air speed is greater in areas above the crop. The CFD predicted results showed that an Almería greenhouse with five roof leeward flap openings and two side vents resulted in higher air exchange rates than with windward flap opening. However, temperature distribution is not as good as with the windward ventilation. INTRODUCTION Ventilation is the main climatic control method of greenhouses in the province of Almería where there are more than 27000 hectares of greenhouses. Nowadays the Almería-type greenhouse is in course of standardisation, since in the few last years Almería manufacturers are exporting this greenhouse to other countries and climatic areas such as Central and South America, North Africa and China. The effectiveness of the ventilation may be related with the spatial distribution of cool air to maintain particular climatic conditions for the crop. The first simulations by means of Computational Fluids Dynamics (CFD) for the study of the ventilation in greenhouses were carried out by Okushima et al. (1989) who compared this numerical method with the experimental results obtained in a wind-tunnel (Sase et al., 1984). Although their results showed little correlation with the experimental data, probably due to the limited power of the available computer resources at that time, they provided important new information on the patterns of flow inside the greenhouses. This technique was not used for some time until the simulations of CFD were compared in a two-span greenhouse with the data obtained by means of sonic anemometry (Bot et al., 1996; Boulard et al., 1996). Mistriotis et al. (1997) studied Mediterranean type greenhouses using CFD in a two-dimensional grid and they concluded that CFD is a powerful tool to develop improvements in the design of greenhouses for efficient ventilation. Kacira et al. (1998) also evaluated the ventilation of a multi-span sawtooth greenhouse for several environmental conditions. Al-Helal (1998) utilized a CFD model to study natural ventilation in Proc. IC on Greensys Eds.: G. van Straten et al. Acta Hort. 691, ISHS 2005

433

arid region greenhouses, and compared the simulation results with those determined from energy and mass balance models with good agreement. Boulard et al. (1999) studied the natural ventilation (thermal and wind driven) in a reduced-scale greenhouse in two dimensions. Al-Arifi et al. (2001) studied the air movement in greenhouses equipped with a fan. Reichrath et al. (2000) using experimental data obtained good agreement of CFD prediction of the internal climate of a commercial 60 span Venlo-type glasshouse containing tomato crops. Haxaire (1999) was the first to study the effects of the crop on the airflow inside the greenhouses, determining the drag effects of the plants. He experimented in windtunnels, where he studied the relationship between the crop leaf area index and the pressure drop generated for different values of air velocity. This relationship has been used in later studies of CFD to investigate the influence of the canopy on the inside climate (Haxaire et al., 2000). He calculated the value of the drag coefficient of the canopy. Lee and Short (2000) also developed a model of CFD to determine the relationship between the air velocity and the pressure drop in the crop. Recently Bartzanas et al. (2002) studied the effect on the ventilation of a tunnel greenhouse caused by the presence of insect-proof screens. The effect of very fine antithrips and anti-aphids nets placed over the ventilation openings has been simulated in three dimensions in commercial greenhouses (Fatnassi et al., 2003). The airflow in Almería type greenhouses has also been analysed by means of two-dimensional (MolinaAiz et al., 2004) and three-dimensional (Campen and Bot, 2003) simulations. MATERIAL AND METHODS The CFD technique numerically solved the Navier-Stokes equations and the mass and energy conservation equations. In this work two-dimensional simulations have been carried out by means of a commercial CFD software packet (ANSYS/FLOTRAN v8.0) that applies the Finite Element method (FE) for the discretisation of the set of equations involved in ventilation processes. The conservation equations of the variable φ in three dimensions that describe the transport phenomena for flows in free convection are of the general form (Ferziger and Peric, 2002):

(

∂ ρCφ φ ∂t

) ∂(ρu Cφ φ ) +

i

∂xi

= Γφ

∂ 2φ + Sφ ∂xi2

(1)

where φ represents the common variable of interest as a concentration of the transported quantity in a non-dimensional form; ρ is the density of the air; Cφ advection coefficient; ui is the component of velocity vector in the direction i; Γφ is the diffusion coefficient and Sφ is the source term. These variables with their different terms are shown in the Table 1 for a steady state (first term in Eq. (1) is equal to zero) and incompressible flow. Numerical codes using a spatial discretisation are needed to solve the set of equations described by Eq. (1) and Table 1. The numerical grid used in simulations was an unstructured grid with higher density at the vent openings, where the flow was subject to strong gradients. The quality of the grid was checked carrying out two levels of modelling. A number elements close to the maximum level allowed by the software used in this work (64000 elements) and another around the half (32000 elements). In the case of the greenhouse with crop the necessary grid refinement and the problems of convergence didn't allow a reduction in the number of elements and we used a 58300 elements grid. In the case of the empty greenhouse a diminution in the number of elements didn't cause a significant variation in the accuracy of simulations (+0.2% of the measured temperature and –0.2% of the measured air velocity), neither it produce convergence problems. The relatively small accuracy increase between the different grids indicates that the grid dependency on the solution has become minimal. Finally, a grid with 28200 elements was used because it consumes the least computational time with reasonable accuracy. We selected a large 434

domain including the greenhouse (95 m long and 15 m high). This domain was chosen after modelling a domain of 200 m x 20 m and not obtaining a significant improvement in the results (0% for temperature, and 1.6% for air velocity). The boundary conditions for temperature at all surfaces of the greenhouse, used in simulations, are shown in Table 2. The boundary conditions prescribed a wall-type boundary condition (zero velocity) along the floor and the roof. The standard K-ε model (Launder and Spalding, 1974) assuming isotropic turbulence was adopted to describe turbulent transport. This model was used because their result corresponds better with experimental data for typical ventilation flows in greenhouses (Haxaire et al., 2000; Kacira et al., 1998). In order to include the effect of the nets placed over the vents in the study, the insect-proof screens were considered as porous media. The flow of air through a porous medium can be described by means of the Darcy-Forchheimer equation (Miguel, 1998): ⎛ ∂P ⎜ µ = −⎜ u+ρ ∂x ⎜Kp ⎝

⎛ Y ⎜ ⎜⎜ 1 / 2 ⎝Kp

⎞ ⎟u ⎟⎟ ⎠

⎞ ⎟ u⎟ ⎟ ⎠

(2)

where, µ is the dynamic viscosity of the fluid, Kp the permeability of the porous medium and Y the non-linear momentum loss coefficient or inertial factor. Equation (2) shows how fluid velocity is related to pressure drop, through the viscous resistance force, which appears due to the momentum transfer at the fluid interface (µ/Kp) and the pore inertia effects (ρY/Kp½). This term was introduced as the resistance distributed term in the porous medium (Rx) in the conservation of momentum equations (for vx), defined by Eq. (1) and values showed in Table 1. The values of the aerodynamic properties of the porous medium used as boundary conditions were calculated using wind tunnel tests (Valera et al., 2004). The crop was also simulated using the porous medium approach by the addition of a momentum source term, due to the drag effect of the crop, to the standard fluid flow equations. The drag form appearing in the source terms of momentum equations was modelled as (Yamada, 1982): ∂P = −C d Lρ u 2 ∂x

(3)

where, L is the crop leaf area index and Cd is the total canopy drag coefficient. A drag coefficient Cd=0.32 had been used. Haxaire (1999) used wind tunnel facilities to calculate this value of drag coefficient for a mature greenhouse tomato crop. Boulard and Wang (2002) have also successfully used this value of drag coefficient for a lettuce crop. RESULTS The differences between air temperatures measured into a commercial Almería type greenhouse with a roof rolling ventilators and two sidewall (Molina-Aiz et al., 2004) and those predicted by ANSYS/FLOTRAN were 0.1 to 2.1ºC. Average differences between the computational and the experimental data were 0.71ºC for temperature and 0.07 m s–1 for velocity. Figure 1 shows CFD predicted and measured air temperature profiles for this greenhouse. The differences between air velocities predicted by CFD models and those measured were from 0.00 to 0.36 m s–1 (0% to 65%). For the same climatic and boundary conditions, obtained by averaging the mean values measured in a commercial Almería type greenhouse (Molina-Aiz et al., 2004), we have simulated the climate for three cases: without insect-proof screens (Fig. 2 a-b), with a 10 x 16 threads cm–2 net (porosity 0.39) and with a 20 x 31 threads cm–2 (porosity 0.29) insect-proof screens. For all the insect-proof screens, the results showed a reduction of air speed induced by the nets, and air temperature rises in proportion to the decrease of air speed (Table 3). Similar results have been obtained for other greenhouse types, equipped with anti-insect or anti-thrips nets (Bartzanas et al., 2002; Fatnassi et al., 2003) 435

For low wind speed (v ≤ 2 m s–1), air entered through both side vents and left through the roof vent in the greenhouse with only one roof rolling ventilator (Fig. 3a). When wind speed was greater than 2 m s–1, air left the greenhouse through the roof vent and the leeward side opening (Fig. 3b). For the greenhouse with only one roof rolling ventilator, the average air exchange rates with plants (Fig. 4 a-b) was predicted to be less than without plants (Fig. 3 b) and the difference between the air exchange rates with and without plants increased as the wind speed increased. The CFD predicted results showed that an Almería greenhouse with leeward flap opening resulted in higher air exchange rates than with windward flap opening. For a west wind the two side vents and five leeward roof vents were predicted to act as inlet and outlet of airflows, respectively (Fig. 4 c-d). However, the air exchange rate decreased when we closed leeward vents and opened windward vents. In this case, the cold outside air entered the greenhouse through the windward side vent and the last three roof vents, placed on the leeward part of the greenhouse (Fig. 4 e-f). Inside hot air leaves the greenhouse through the leeward side ventilator and the two first windows (placed on the windward part of the greenhouse). These results are also in agreement with those obtained by Lee and Short (2000) for a multi-span greenhouse (four and one-half span) designed with a windward side vent and leeward roof vents. Reduction of 88% in the ventilation rate were observed when the spans number increased from 1 to 5, as consequence of the increase of the greenhouse width (9 to 45 m). For a greenhouse with 1 span, when wind speed was low (v ≤ 1 m s–1), air entered through both side vents and left through the roof vent. Two small and more or less symmetrical circulating cells were formed near the cover. When wind speed was greater than 1 m s–1, the wind created a jet-like flow that crossed the greenhouse from the windward to the leeward side opening. Air also left the greenhouse through the roof vent. A similar airflow pattern was found for greenhouses with 3 and 5 spans when wind speed was greater than 2 m s–1 (Fig. 3b). DISCUSSION Although the numerical model simulates quite well the ventilation performance of greenhouse, the three-dimensional calculation are preferable over the two-dimensional calculations when wind direction are not perpendicular, because it can play an important role in ventilation, and the accuracy of the CFD simulations can be improved using 3D models. To get a more realistic description of the flow field and temperature spatial heterogeneity within the greenhouse, heat and vapour exchanges between the crop and air, also should be applied to describe the influence of plants in microclimate in future CFD studies. The results are only valid for the specific case examined. Therefore, the conclusions, although they give a good qualitative idea about the influence of wind speed, can hardly be generalised. However, we can deduce from the results that Computational Fluid Dynamics is a very effective tool for the study of the ventilation in greenhouses and other agricultural buildings. ACKNOWLEDGEMENTS This research was partially supported by the Projects C03-159 and CR-UAL-0202. Literature Cited Al-Arifi, A., Short, T. and Ling, P. 2001. Validating the CFD model for air movements and heat transfer in ventilated greenhouses. 2001 ASAE Annual International. Meeting. July 29-Aug 1, Sacramento (USA). Paper No. 01-4056, pp. 1-16. Al-Helal, I. 1998. A computational fluid dynamics study of natural ventilation in arid region greenhouses. Ph.D. Thesis. Department of Food, Agricultural and Biological Engineering, The Ohio State University, Ohio (USA). Bartzanas, T., Boulard, T. and Kittas, C. 2002. Numerical simulation of the airflow and temperature distribution in a tunnel greenhouse equipped with insect-proof screen in the openings. Comput. Electron. Agric., 34: 207-221. 436

Bot, G.P.A., Boulard, T., Mistriotis, A., Papadakis, G., Picuno, P. and Scarascia-Mugozza, G. 1996. New techniques in greenhouse ventilation analysis. AgEng Madrid 1996. 96B-037, 1-9. Boulard, T., Meneses, J.F., Mermier, M. and Papadakis, G. 1996. The mechanisms involved in the natural ventilation of greenhouses. J. of Agric. and Forest Meteorology, 79 (1-2): 61-77. Boulard, T. and Wang, S. 2002. Experimental and numerical studies on the heterogeneity of crop transpiration in plastic tunnel. Comput. Electron. Agric., 34: 173-190. Boulard, T., Haxaire, R., Lamrani, M.A., Roy, J.C. and Jaffrin, A. 1999. Characterization and modelling of the air fluxes induced by natural ventilation in a greenhouse. J. Agric. Engng Res., 74: 135-144. Campen, J.B. and Bot, G.P.A. 2003. Determination of greenhouse-specific aspect of ventilation using three-dimensional computational fluid dynamics. Biosystems Engineering, 84 (1): 69-77. Fatnassi, H., Boulard, T. and Bouirden, L. 2003. Simulation of climatic conditions in fullscale greenhouse fitted with insect-proof screens. Agric and Forest Meteorol., 118: 97111. Ferziger, J.H. and Peric, M. 2002. Computational Methods for Fluid Dynamics. Springer, Berlín (Germany). Haxaire, R., 1999. Caracterisation et modelisations des ecoulements d’air dans une serre. Ph.D. Thesis, Université de Nice Sophie Antipolis. Faculté des Sciences, (France), 149 pp. (in French, with English abstract) Haxire, R., Boulard, T. and Mermier, M. 2000. Greenhouse natural ventilation by wind forces. Acta Hort. 534: 31-40. Kacira, M., Short, T.H. and Stowell, R.R. 1998. A CFD evaluation of naturally ventilated multi-span, sawtooth greenhouses. Transactions of the ASAE, 41 (3): 833-836. Launder, B.E. and Spalding, D.B. 1974. The numerical computational of turbulent flow. Comp. Method App. Mech. Eng., 3: 269-289. Lee, I.B. and Short, T.H. 2000. Two-dimensional numerical simulation of natural ventilation in a multi-span greenhouse. Transaction of the ASAE, 43 (3): 745-753. Miguel., A.F. 1998. Airflow through porous screens: from theory to practical considerations. Energy and Building, 28: 63-69. Mistriosis, A., Bot, G.P.A., Picuno, P. and Scarascia-Mugnozza, G. 1997. Analysis of the efficiency of greenhouse ventilation using computational fluid dynamics. Agric. and Forest Meteorology, 85: 217-228. Molina-Aiz, F.D., Valera, D.L. and Álvarez, A.J., 2004. Measurement and simulation of climate inside Almería-type greenhouses using Computational Fluid Dynamics. Agric. and Forest Meteorology, 125 (1-2): 33-51. Okushima, L., Sase, S. and Nara, M. 1989. A support system for natural ventilation design of greenhouses based on computational aerodynamics. Acta Hort. 284: 129–136. Reichrath, S., Ferioli F. and Davies, T.W. 2000. A simple computational fluid dynamics (CFD) model of a tomato glasshouse. Acta Hort. 534: 197-204. Sase, S., Takakura, T. and Nara, N. 1984. Wind tunnel testing on airflow and temperature distribution of a naturally ventilated greenhouse. Acta Hort., 148: 329-336. Valera, D.L., Molina-Aiz, F.D., Álvarez, A.J., Terrés-Nicoli, J.M., Madueño A. and López, J.A. 2004. Contribution to characterisation of insect-proof screens: experimental measurements in wind tunnel and CFD simulation. GREENSYS 2004, 8 pp. Yamada, T. 1982. A numerical model study of turbulent airflow in and above a forest canopy. J. Meteorological Society of Japan, 60: 439-454.

437

Tables

Table 1. Reduced form of the variables, φ, advection coefficient, Cφ , diffusion coefficient, Γφ , and source terms, Sφ , of the conservation equations defined by Eq. (1). Conservation equations Mass Momentum for x direction Momentum for y direction Energy Turbulent kinematic energy Turbulent dissipation rate

φ

Cφ 1 1 1 cp 1 1

1 vx vz T K

ε

Γφ

Sφ 0 –∂P/∂x+Rx ρ gz –∂P/∂z

0 µe µe

λ

Φ µt/µ Φ – ρ ε Cε1 µt/µ Φ ε/K – Cε2 ρ ε2/K

µt/σk µt/σε

Table 2. Temperatures used as boundary conditions for the simulation. Covering material temperatures (K) Span 1 309 Span 2 309 Span 3 310 Span 4 311 Span 5 310 Outside air temperature (K) 299 a

Greenhouse soil surface temperatures (K) Southwest side area (0-4 ma) 315 Southwest area (4-22 m) 318 Centre area (22-25 m) 318 Northeast area (25-41 m) 319 Northeast side area (41-45 m) 315 Outside surface soil temperature (K) 310

Distance from the windward side walls with crop

Table 3. Mean and standard deviation for air velocity and temperature difference between inside and outside. Insect proof-screen –2

20 x 31 threads cm 10 x 16 threads cm–2 without screen

438

(

∆T (º C )

σ ∆T (º C )

∆v m s −1

5.3 3.5 2.4

4.8 2.7 1.5

0.61 0.80 1.12

)

(

σ ∆v m s −1

0.49 0.66 0.85

)

Figurese (a)

Air temperature (ºC)

50 40 30 20 10 0.5

15

30

44.5

Distance from the windward side vent (m)

(b)

-1

Air velocity (m s )

1.5

1.0

0.5

0.0 0.5

15

30

44.5

Distance from the windward side vent (m)

Fig. 1. Measured (points) and CFD simulated (lines) longitudinal air temperature (a) and velocity (b) profiles at 1.5 m high for different wind speed. Greenhouse with crop:+ u=7.7 m s–1 ; □ u=1.8 m s–1 and empty greenhouse: O u=5.0 m s–1; X u=1.0 m s–1.

27.4ºC

34.4ºC

29.8ºC 29.8ºC

32.1ºC 36.7ºC

(a) wind

2 m/s

(b) Fig. 2. Simulated air temperatures (a) and velocity vectors (b) in an Almería type greenhouse without insect-proof screens in the openings for a wind of 5 m s–1. 439

(a)

wind

0.5 m/s

(b)

wind

2 m/s

Fig. 3. Simulated air velocity vectors in an Almería type greenhouse with insect-proof screens (porosity 0.39) in the openings for a wind of 2 m s–1 (a) and 5 m s–1 (b). (a) 29.6ºC

29.6ºC 26.0ºC

26.0ºC

26.0ºC 29.6ºC

26.0ºC 42.0ºC

(b) wind

1 m/s

(c) 28.6ºC 28.6ºC

31.3ºC

28.6ºC

34.5ºC 31.3ºC 37.4ºC 28.6ºC

40.4ºC

(d) wind

1 m/s

(e) 26.0ºC

31.0ºC

31.0ºC

34.6ºC

26.0ºC

27.4ºC 31.0ºC

34.6ºC

26.0ºC 27.4ºC

(f) wind

0.50 m/s

Fig. 4. Simulated air temperatures and velocity vectors in an Almería type greenhouse, for a wind of 5 m s–1, with: a roof rolling ventilators (1.8 m width) and two sidewall (1.5 m width) with a crop, LAI=2 (a-b); leeward flap ventilators of 0.4 m high (cd) and windward flap ventilators (e-f). In all cases ventilator was equipped with a insect-proof screen (porosity 0.39).

440