18th international conference on Sciences and Techniques of Automatic control & STA'2017-PID4615-CAD

computer engineering - STA'2017, Monastir, Tunisia, December 21-23, 2017

Control of relative humidity inside an agricultural

greenhouse
Amira LABIDI (1), Amine CHOUCHAINE(1), Abdelkader MAMI (1)
(1)Université de Tunis El Manar, Faculté de sciences de Tunis,
Laboratoire d’ Application de l’ efficacité Energétique et des Energies Renouvelables, (LAPER), Tunis, Tunisia.
amiralabidi1@gmail.com, amin.chouchain@laposte.net, abdelkader.mami@gmail.com

Abstract— The development of intelligent systems in

agriculture is required to improve plant production. The II. MODÉLISATION
first part of this paper is a general study of the greenhouse This section is devoted to presenting the greenhouse
and its climatic factors essential for the growth of the environment and its reduced physical model. The greenhouse
plant. In the second part, two controls have been adjusted is divided into four supposedly homogeneous medium: the
to assimilate the humidity control; the PID controller and cover, the indoor air, the crop and the soil. There are several
the fuzzy logic controller. energetic modes of exchange within the system with different
nature and scope.
Keywords—Greenhouse, humidity, PID, fuzzy logic,
simulation The different exchanges in the modelling of the
microclimate in a greenhouse are shown in Fig. 1:
I. INTRODUCTION
By the desire to increase the quality and quantity of
agricultural products, the development of greenhouse crops is
classified as an important technology giving a suitable
microclimate condition for plant growth. Recently, specialized
research has documented numerous modelling works to
understand the energetic exchange phenomena that govern the
indoor climate of greenhouses [1], [2], [3], [4].
We have developed a simplified model that describes the
system to simulate and control. The model linked to changes
in external climates such as air temperature, wind speed,
relative humidity... which have a considerable impact on
agriculture [5-6].
To ensure production that meets the needs of farmers while
adapting and mitigating climate change [7], it is necessary to
control the humidity and the temperature by acting on the
other characteristics of the environment (light under glass,
ambient temperature or humidity) to counter this phenomenon. Fig. 1. The different thermohydric exchanges in the modeling of the
microclimate in a greenhouse.
[8-9].
Various controllers have been developed to optimize the 1- Shortwave radiative exchange
climate of the greenhouses (the [on/off] control [10-11-12], 2- Longwave radiative exchange
Adaptive Predictive control [13-14-15], Neural Network [16,
17], Fuzzy Logic Controller [FLC] [18-19], and PID [20], 3- Convective and conductive exchanges
Nonlinear Adaptive Control [21] …). 4- Heat and mass exchange by transpiration and renewal
of air.
The paper is devoted to a physical model of the greenhouse
in section 2. Two techniques of controls have been used in the A set of dynamic equations describes a precise model of a
present work: the first in section 3 is based on the synthesis of greenhouse. Several models have been studied [22-23-24-25-
Proportional Integral and Derived. The second is based on the 26].
fuzzy logic in Section 4. After that, a comparative analysis of
dWa
the simulation results is established. Finally, a conclusion will U aVa Q Solar radiation  Q Evapotranspiration
end this paper in the last section. dt (1)
Q Ventilation
Q Dehumidifying

978-1-5386-1084-8/17/$31.00 ©2017 IEEE 109

Where Wa is the expression of absolute humidity of inside air The moisture loss rate of the dehumidification system is
1 given by:
( Kg .Kg 1 ), U a is the air density ( Kg .m 3 ), Va is the
volume of the indoor air in the greenhouse ( m ).
3 Q Dehumidifying U aVs Wdh  Wa  
The wave radiation absorbed by the greenhouse was
calculated by the following equation [27-8]. With Vs is the dehumidifier air flow  m 2 s 1  , Wdh is the
absolute humidity of the dehumidifier airflow having the
1
Q Solar radiation Rs (D p S p  AS plW p  D s S sW p ) (2) following expression (%):
O
Rd h
O 1 1 Wdh K Pae 
With is the conductivity of the cover ( Wm . K ), Rs 100
1
is the global solar radiation ( W .m ).
Rd h is the relative humidity of the dehumidifier airflow (%)
D p  D s : Coefficients of radiation absorption, by the cover, and
the soil (-). The experimental agricultural greenhouse studied in this
paper has a length of 1.5 m, width = 1 m and height = 1.15 m).
S p  S s  S pl  represents the surface of the cover, soil and the It is located in north of Tunisia with its axis parallel to the
plants ( m 2 ), A is a coefficient constant (-), and Wp is east-west direction as shown in fig. 2.

transmission coefficients of the cover (-).

The transpiration of the culture is given by:

ª§ ·º
U a S pl Ǭ J CnU 2
1
Q Evapotranspiration
O ¬© Ta 1  cdU 2  ¸¹ »¼

§ Rha ·
¨1  ¸ (3)
© 100 ¹
With J is the psychrometric constant ( Pa .K 1 ), S pl is the
surface of the plants ( m 2 ), U 2 wind speed ( m.s 1 ), Ta
1
temperature of inside air ( K ), Rha is the relative humidity
Fig. 2. External structure of the experimental greenhouse.
of the air (%).
The input/output parameters and the dimensions of the
The mass transfer obtained by air renewal is represented
experimental greenhouse have been used for this simulation.
by (4):
After that two control techniques have been applied to this
QVentilation U a Gv We  Wa   process to control its internal relative humidity.

With Gv is the airflow circulating inside the greenhouse

3
(m s ), Wa is the absolute humidity of indoor air
1
( Kg .Kg 1 ) having the following expression:
Rh
Wa KPa a (5)
100
and We is the absolute humidity of the outdoor air
1
( Kg .Kg 1 ), with the following expression:
Rhe
We KPae (6)
100 Fig. 3. Solar radiation

With Pa indoor air vapor pressure (Pa), Pae external air

saturation vapor pressure (Pa), Rha Indoor air relative
humidity (%), Rhe outdoor wind speed (%).

110
 The PID controller is used to drive the system output in
order to follow the setpoint, avoiding overshoots, and using
only the power required to go to the setpoint and maintain.
Concretely, in PID, we have 3 actions: P (Proportional), D
(Derived) and I (Integral).

A. Régulateur PI

Fig. 4. The inside and outside relative humidity.

The saturating atmosphere of the greenhouse at the beginning

of the day favors the proliferation and development of fungi Fig. 5. Structure of a classical IP
and various predators in the confined environment of the
greenhouse. The block diagram of the greenhouse model, which is set
with the PI controller shows the various I/O variables in the
The implementation of energy saving strategies during the
greenhouse model. The control variable is humidity ( Rdh ).The
night time (lowering of set temperatures, integration of heat
exchanger, installation of thermal screens, more tight seals) disturbance variables are solar radiation ( Rs ), indoor air
can lead producers to dehumidify their greenhouses, in order
temperature ( Te ), outdoor air humidity ( Rhe ) and wind speed
to slightly preserve the temperature from decreasing.
( U z ).These entrances to the greenhouse come from a real
The use of a dehumidifier to manage the humidity in
greenhouse culture, is a solution to consider. We studied the database of the greenhouse of Fig. 2, with Wa Rha .
dehumidifier type verodery-2009-Lcd. It is fully automatic
thanks to an incorporated hygrostat. The humidistat is located The conventional PI regulator directly connects the
on the top of the unit and can be set to the desired humidity control signal to the deviation signal E. Its time description is
level, depending on the need for dehumidification. as follows:
Rd h ( Kpe(t )  Ki ³ e(t )) (9)
TABLE I. VERODERY-LCD DEHUMIDIFIER SPECIFICATION The variable Rdh is the humidity of the air supplied by the
Model YL-2009LCD
Single phase power supply 220 ~ 240 V / 50 Hz dehumidifier. It consisted of 30% < Rdh < 100%.
Dehumidification capacity (liters / day) 11L / D (30 Ԩ 80% HR)
Power Consumption (W) 270 W The difference is defined as:
Compressor type Rotor Compressor
Water tank capacity
Refrigerant
1,6 L
R-134a without CFC
e(t ) Y (t )  Y (t ) 
The gas charge 120 g
Where Y (t ) is the setpoint and Y (t ) = Wa is the measured
Weight 11,2 kg
output.
Dimensions 35 x 21 x 54,5 cm
B. Adjustment of PI Parameters
III. CONTROLLER PID We have adjusted the PI parameters with the method of
The main purpose of this section is to define the basics of Ziegler and Nichols (ZN).
system regulation, where the control laws are synthesized To find the parameters of the PI controller, simply record the
using known techniques. The controller has been implemented step response of the process alone (without the controller),
to test their reliability to control the climate inside an then trace the tangent to the point of inflection of the curve. It
agricultural greenhouse, which present a very complex was then measured its slope p and the apparent delay
dynamic system. corresponding to the point of intersection of the tangent with
PID controllers are proving to be sufficient to solve a large the abscissa. We can therefore calculate the coefficients of the
number of control problems. They have been empirically regulator.
regulated, for example by the methods described by Ziegler Generally the Kp gains proposed by Ziegler-Nichols are too
and Nichols (1942). The fuzzy logic was established at the high. We must reduce Kp by a factor of 2 to obtain a
beginning of (1965) by Zadeh, it asserts itself as an satisfactory response.
operational technique. Used with others. Used alongside other Then we found:
advanced control techniques, it makes a discrete entry but Kp = 100;
appreciated in greenhouse climate control automation. Ki = 0.01.

111
 IV. FUZZY LOGIC CONTROLLER
The development of fuzzy controllers is usually based on The output control,
the expression in the base of the linguistic rules. It Where:
incorporates a set of rules to (logically) assign an output to an Com: control; H: High-Com; M: Average-Com; L: Low-Com.
input.
To control relative humidity, the greenhouse model typically The input variables of the fuzzy controller are the relative
has two types of input data: KXPLGLW\RIWKHLQGRRUDLUDQGLWVHUURUǻ+, where:
- Disturbance variables: outside weather, outdoor temperature, 'H H set po int  H inside , greenhouse (11)
external relative humidity, solar radiation and wind speed.
- Actuators: dehumidification. These current rules are chosen in order to have desirable
interior humidity. Therefore, if the humidity is high, we must
reduce it by using the dehumidification system. The
membership functions of this FLC controller are written
taking into account the variation of the relative internal
humidity to achieve the desired one by activating the actuator
at a suitable rate.

The membership functions of the input variables, which

are the humidity error and its variation, are plotted in Fig. 6.

Fig. 6. Structure of the dehumidification control of the greenhouse indoor air

with the fuzzy logic controller.

The structure of the dehumidification control of the greenhouse indoor air

with the fuzzy logic controller is given in Fig 5.

du
The operator ( ) makes it possible to obtain the
dt
variation of the error between two sampling instants. The
controller will therefore be sensitive to the humidity error and
its variation.

We have synthesized the Fuzzy logic controller of the

MANDANI type (MAX-MIN inference method) for the
humidity control.

TABLE II. HUMIDITY CONTROL BASE

Humidity
Variation Command
Error
H-Err N H-Com

H-Err Z H-Com

M-Err P L-Com Fig. 7. The membership of the error and the variation

M-Err N H-Com The function belonging to the control output variable, is

shown in Fig. 8.
M-Err Z M-Com

M-Err P L-Com

L-Err N M-Com

L-Err Z L-Com

L-Err P L-Com

Where:
Err: error; H: High-Err; M: Average-Err; L: Low-Err; N:
Negative; Z: Zero; P: Positive
Fig. 8. The membership functions of the command.

112
 V. COMPARISON
In this section, we present the simulation results
concerning the application of the fuzzy logic and PID control
on Fig. 9, 10, 11 and 12 respectively.
A series of setpoint changes was performed, and the process
behavior was simulated using the Matlab-Simulink
environment.

The responses show the internal relative humidity of the

greenhouse climate with a PI and fuzzy logic corrector for
several set points respectively from 60 % to 90%.

Fig. 11. The evolution of the relative humidity of the controlled air with
setpoint 80.

Fig. 9. The evolution of the relative humidity of the controlled air with
setpoint 60.

Fig. 12. The evolution of the relative humidity of the controlled air with
setpoint 90.

FIG. 9 shows that the FL controller is more efficient then

the PI to reach the setpoint with a small overshoot.
As been shown in figs. 10 and 11, the PI controller is quicker
than the FL controller to achieve the setpoint with a large
overshoot. However, according to FIG. 12 the response of the
PI presents a perfect result just when the setpoint equal to
90%. Otherwise, the LF controller follows the setpoint more
perfectly.
According to the simulations, the FLC performance is
better than the PI controller in terms of precision and rapidity
Fig. 10. The evolution of the relative humidity of the controlled air with
for the some set-points. In this simulations, the system
setpoint 70. response rapidity is better when using the PI controller there is
an overshooting problem as it can be seen in the figures 10
and 11.
The two controls PI and Fuzzy Logic have advantages
and disadvantages. The results allow us to confirm the
advantage of the fuzzy control compared to the classical
control “PI”. It can be concluded that the fuzzy regulation
presents an interesting solution for the humidity control.
On the other hand, the traditional PID controller loses its
performance, especially in terms of stability, overshoot and
dynamic accuracy, but, the fuzzy controller is more robust.
VI. CONCLUSION
In this work, the control of the indoor humidity of a greenhouse
is performed using two different types of correctors, PI and

113
Fuzzy Logic. The simulation results obtained were negotiated [10] H. AIT BOU, “Conception, développement et mise en place d’un
système de pilotage automatique du climat et de la fertirrigation goute à
and a comparative analysis was presented. The objective here
goute sous serre“, Thesis Faculté des sciences, Meknès, 2003.
is to show the advantages in robustness and accuracy of each
[11] L. Sbita, M. Souissi, M. Annabi, Validation du modèle de climat d’une
control law for controlling humidity which is one of the most serre agricole, SIPE 2, Béchar-Algérie, Novembre, (1994).
important parameters for climate management but also the [12] L. Sbita, “Modélisation thermo hydrique et étude en utilisant les réseaux
most difficult to control. de neurones de l’aération naturelle d’une serre géothermale du sud
The results showed that the indoor humidity has a faster Tunisien-Contrôle et commande par un système informatique
response time when using fuzzy controller than when using PI autonome“, Thesis Ecole Supérieure des Sciences et Techniques de
Tunis, pp.163, 1997.
controller for some set-points and a better accuracy for the
[13] J.B.Pomet, “La commande adaptative des systèmes non linéaires“,
others ones. In future work, the Fuzzy-Pi controller will be Thesis Ecole nationale supérieur des mines de Paris, 1989..
used to obtain an appropriate indoor relative humidity. [14] L. Ouesleti, G. Enea, M. Franceshi, “Commande adaptative
multivariable du climat d’une serre agricole, Revue d’automatique et de
productique appliquée“, vol3, n° 3-4, 1990.
[15] M. de Mathelin, J. Bara et J. Gangloff. “Commande numérique des
systèmes“, Thesis Ecole nationale supérieure de physique de Strasbourg.
References [16] F. Fourati, “Multiple neural control of a greenhouse,” Neurocomputing
139, pp. 138–144, 2014.
[1] O. Jolliet, L. Danloy , J.-B. Gay a, G.L. Munday, and A. Reist, [17] M. Taki, Y. Ajabshirchi, S. Faramarz Ranjbar, A. Rohani, and M.
“HORTICERN: an improved static model for predicting the energy Matloobi, “Heat transfer and MLP neural network models to predict
consumption of a greenhouse,” Agricultural and Forest Meteorology 55, inside environment variables and energy lost in a semi-solar
pp.265-294, 1991. greenhouse,” Energy and Buildings 110, pp. 314–329, 2016.
[2] M. Torrellas, A. Antón, J. Ignacio Montero “ An environmental impact [18] F. Lafont, and J.-F.Balmat, “Optimized fuzzy control of a greenhouse,”
calculator for greenhouse production systems” Journal of Environmental Fuzzy Sets and Systems 128, pp. 47–59, 2002.
Management, Vol. 118, pp.186-195, 2013. [19] M. Azaza1, K. Echaieb, F. Tadeo, E. Fabrizio, A. Iqbal, and A. Mami,
[3] R. Shamshiri, W. Ishak W. Ismail “ A Review of Greenhouse Climate “Fuzzy Decoupling Control of Greenhouse Climate Arabian Journal For
Control and Automation Systems in Tropical Regions” Journal of Science And Engineering, 2015.
Agricultural Science and Applications, Vol. 2, Issue 3, pp.175-182, [20] A. Chouchaine, E. Feki, A. Mami, “Stabilization Using a Discrete
2013. Fuzzy PDC Control with PID Controllers and Pole Placement:
[4] A. Vadiee, V. Martin “Energy management in horticultural applications Application to an Experimental Greenhouse,” Journal of Control Science
through the closed greenhouse concept, state of the art” Renewable and and Engineering, 2011.
Sustainable Energy Reviews, Vol. 16, pp.5087-5100, 2012. [21] S. Zeng, H. Hu, L. Xu, G. Li, “Nonlinear Adaptive PID Control for
[5] J. Fargues, C. Vidal, N. Smits, M. Rougier, T. Boulard, M. Mermier, P. Greenhouse Environment Based on RBF Network,” Sensors, 2012.
Nicot, P. Reich, P. Jeannequin, G. Ridray, J. Lagier, “Climatic factors on [22] D. Kolokotsa, G. Saridakis, K. Dalamagkidis, S. Dolianitis, I.
entomopathogenic hyphomycetes infection of Trialeurodes Kaliakatsos, “Development of an intelligent indoor environment and
vaporariorum (Homoptera: Aleyrodidae) in Mediterranean glasshouse energy management system for greenhouses“. Energy Convers.
tomato. Biol. Control 28, p.320–331, 2003. Manag. 51, pp. 155–168, 2010.
[6] A.M. Abdel-Ghany, T. Kozai, “Dynamic modeling of the environment [23] 0.Õ\DQ(%LQJ|O00HOLNRJOXDQG$$OERVWDQ³0RGHOOLQJDQG
in a naturally ventilated, fog-cooled greenhouse,” Renewable Energy 31, simulation of a hybrid solar heating system for greenhouse applications
pp.1521–1539, 2006. using Matlab/Simulink,” Energy Conversion and Management 72,
[7] E. Fitz-Rodrígueza, C. Kubotab, G.A. Giacomellia, M.E. Tignorc, S.B. pp.147–155, 2013.
Wilsond, and M. McMahone, “Dynamic modeling and simulation of [24] S. Bouadila, S. Kooli, S. Skouri, M. Lazaar, and A. Farhat,
greenhouse environments under several scenarios: A web-based “Improvement of the greenhouse climate using a solar air heater with
application,” Computers and Electronics in Agriculture 70, pp. 105–116, latent storage energy,” Energy 64, pp. 663-672, 2014.
2010.
[25] R. Ben Ali, E. Aridhi and A. Mami, “Dynamic model of an agricultural
[8] M. Souissi, “Modélisation et commande du climat d’une serre agricole,” greenhouse using Matlab-Simulink environment,” 16th international
Thesis, Faculty of Mathematical, Physical and Natural Sciences conference on Sciences and Techniques of Automatic control &
ELMANAR University, Tunisia, 2002. computer engineering - STA'2015, pp. 346–350, 2015.
[9] E. Feki, “Identification et commande d’une serre agricole, Application [26] M. Azaza, K. Echaieb, E. Fabrizio, A. Iqbal and A. Mami, “An
de l’approche multimodèle”, Thesis Ecole nationale d’ingénieurs de intelligent system for the climate control and energy savings in
Tunis, 2004. agricultural greenhouses,” Energy Efficiency, Springer
Science+Business Media Dordrecht, 2016.

114