Modelling of Greenhouse Temperature Control using Adaptive … Paper/Volume 52/Issue 03/IJES...
Transcript of Modelling of Greenhouse Temperature Control using Adaptive … Paper/Volume 52/Issue 03/IJES...
International Journal of Engineering, Applied and Management Sciences Paradigms
Volume 52, Issue 03, Quarter 03 (July-August-September 2018)
An Indexed and Referred Journal
ISSN (Online): 2320-6608
www.ijeam.com
IJEAM
www.ijeam.com
29
Modelling of Greenhouse Temperature Control
using Adaptive-Neuro Fuzzy Inference System
Nidhi Rathi1 and Mr. Devender Kumar2
1M.Tech. Student, ECE, Gurgaon Institute of Tech. & Mgmt., Gurgaon, Haryana (India)
2Assistant Professor, ECE, Gurgaon Institute of Tech. & Mgmt., Gurgaon, Haryana (India)
Publishing Date: September 05, 2018
Abstract The greenhouse system is an important and effective
measure, which provides the plants with appropriate
environmental conditions for growing. To get proper
growth of crops, inside temperature and inside
humidity should be maintained at desired level with
respect to climate conditions. This paper presents a
design of a control system for a greenhouse using
geothermal energy as a power source for the plants
in order to reach predefined results for high yield, high
quality and low costs. Four controller techniques;
PI control, fuzzy logic control, artificial neural
network control and adaptive neuro-fuzzy control
are used to adjust the greenhouse indoor
temperature at the required value.
MATLAB/SIMULINK is used to simulate the
different types of controller techniques. Finally a
comparative study between different control strategies
is carried out.
Keywords: Artificial intelligent, Control,
Artificial neural network, Fuzzy logic control,
Adaptive neuro-fuzzy.
1. Introduction
Greenhouse crop production is a form of
protected agriculture where depending on the
technological level implemented, the degree of
environmental (aerial and root zone) control will
be sufficient to induce optimal plant growth. .A
greenhouse production systems usually refers to
fixed structures with a translucent glazing
material and with a varying degree of aerial
environment controllability (Jensen and Malter,
1995)[16].
The objective of greenhouse environment can be
further improved by adding heating, ventilation
and CO2 supply systems, in order to provide the
pre-eminent environmental conditions.
Conventional control techniques are difficult to
implement in greenhouse systems due to their
multi-variable and non-linear nature. Where
interrelations between internal and external
variables are complex (nonlinear physical
phenomena that govern these systems dynamics
are complicated). This provides justification for
the use of intelligent control techniques as a
good alternative. In this way, fuzzy logic as part
of AI techniques is an attractive and well-
established approach to solving control problems
[6]. In the last decade, a number of researchers
addressed the control design of the climatic
conditions in GHs. GH modelling and
identification have been studied by Ferreira et al.
(2002)[11], Cunha (2003)[8], and Bennis et al.
(2003)[4]. GH climate was developed using
intelligent controllers by Arvantis et al.
(2000)[2], Lafont and Balmat (2002)[19],
Sigrimis et al. (2002), Pasgianos et al. (2003),
Bennis et al. (2005) [5], Koutb et al. (2004)[18],
Bennis et al. (2003)[4], and Fourati and
Chtourou(2004)[12]. Numerous strategies and
control techniques have been proposed
(Márquez-Vera et al.[21], 2016; Revathi and
Sivakumaran, 2016; Coelho et al., 2005)[7],
model predictive control (Coelho et al., 2005;
Pin˜on´ et al.)[7], linear quadratic adaptative
control (Arvantis et al., 2000)[2], neural
networks control (Ferreira et al., 2002), fuzzy
control (Lafont and Balmat, 2004)[20]. In this
paper, modelling and control problem of
greenhouse indoor temperature are studied.
International Journal of Engineering, Applied and Management Sciences Paradigms
Volume 52, Issue 03, Quarter 03 (July-August-September 2018)
An Indexed and Referred Journal
ISSN (Online): 2320-6608
www.ijeam.com
IJEAM
www.ijeam.com
30
Various control techniques (PI control, fuzzy
logic control (FLC), artificial neural network
control (ANNC) and ANFIS) are presented.
Moreover MATLAB/SIMULINK is used to
validate the proposed types of controller
techniques. Finally a comparison study between
the controllers performance is carried out.
2. Artificial Intelligent
Artificial Intelligence (AI) is the field of
Computer Science that attempts to provide
humanlike abilities in computers. One of the
primary means by which computers are endowed
with humanlike abilities, is a Neural Network
(NN). The human brain consists of a network of
over a hundred billion interconnected neurons
that can process small amounts of information
and then activate other neurons to continue the
process. Neural networks are often not suitable
for problems in which it is necessary to know
precisely how the solution was derived. A NN
can be very useful for solving problems for
which they can be trained (Barrera et al.,
2000)[3]. ANFIS, FLC and ANNs techniques,
are types of AI techniques.Fuzzy logic is an
effective in feedback control system and easier to
implement. The computational structure of fuzzy
logic is composed of Fuzzification, Inference
engine and Defuzzication modules. The control
system computational structure of fuzzy logic is
composed of Fuzzification, Inference engine
implemented here is a single input-single output
system with input as ‘error in temperature’and
‘Ventilation rate’ as output variable (Satyajit
Ramesh Potdar et al., 2017)[26]. FLC is
commonly used in control systems. To overcome
system uncertainty and parameter variations, the
fuzzy logic controller is used to model the
control objective based on human knowledge,
understanding the system responses. A fuzzy
controller is a system which works on numerical
data and converts it into a symbolic form through
a data base (fuzzification). A logic of decision-
making (rule base) is implemented, thus it is
possible to provide a symbolic answer which
must be converted into a numerical data
(defuzzification) (Lafont and Balmat, 2002)[19].
ANN’s can be used to solve complex problems
where noise immunity is important (Gupta et al.,
2003)[14]. Two methods of ANN training which
are supervised training and unsupervised
training. Supervised training needs training set
where the input and the desired output of the
network are provided for several training cases,
while unsupervised training requires only the
input of the network, and ANN is supposed to
classify the data properly (Dadios, 2012)[9].
ANFIS is a kind of ANN that is based on
Takagi–Sugeno fuzzy inference system.
Figure 1: ANFIS Controller Structure
Since it combines both neural networks and
fuzzy logic principles, it has potential to capture
the benefits of both in a single framework. Its
inference system corresponds to a set of fuzzy
IF–THEN rules that have learning capability to
estimate nonlinear functions. Hence, ANFIS is
considered to be a universal estimator (Coelho et
al., 2005)[7]. ANFIS control is a hybrid method
consists of two parts which are gradient method
applied to calculate input membership function
parameters, and least square method is applied to
calculate the parameters of output function.
The structure of ANFIS is shown in Fig. 3 (Mote
and Lokhande, 2012) [10][23].
The individual layers of this ANFIS structure are
described below (Mote and Lokhande,
2012)[10][23]:
Layer 1: Every node i in this layer is adaptive
with a node function
1
Oi = μAi(x) (24)
International Journal of Engineering, Applied and Management Sciences Paradigms
Volume 52, Issue 03, Quarter 03 (July-August-September 2018)
An Indexed and Referred Journal
ISSN (Online): 2320-6608
www.ijeam.com
IJEAM
www.ijeam.com
31
(25)
where x is the input to node i, Ai is the
linguistic variable associated with this node
function μAi is the membership function of Ai,
and {ai, bi ,
ci} is
the
premise parameter set.
Layer 2: Each node in this layer is fixed node
which calculates the firing strength Wi of the
rule. i
The output of each node is the product of all
the incoming signals to it and is given by
(Mote and Lokhande, 2012[10][23]):
Oi2 =
wi = μAi (x) × μBi(y), I = 1, 2
Layer 3: Every node in this layer is a fixed
node. Each ith node calculates the ratio of the ith
rule’s firing strength to the sum of firing
strengths of all the rules. The output from the ith
node is the normalized firing strength given by
(Mote and Lokhande, 2012)[10][23]:
(27)
Layer 4: Every node in this layer is an adaptive
node with a node function given by (Mote and
Lokhande, 2012[10][23]):
Oi4=
wifiw (pi x+ qi y+ ri ), i= 1, 2
where wi is the output of Layer 3 and (pi, qi,
ri) is the consequent parameter set.
Layer 5: This layer comprises of only one fixed
node that calculates the overall output as the
summation of all incoming signals, i.e. (Mote
and Lokhande, 2012)[10][23].
(29)
Figure 2: Proposed PI controller using
MATLAB/SIMULINK
Figure 3: MATLAB/SIMULINK model of
fuzzy logic controller
3. Control system design
The effectiveness of the control depends on the
accuracy of the mathematical model which
describes the greenhouse dynamics variables
(Iga, 2008)[16]. AI can be used to control the
greenhouse climate in order to improve the
cultivation growth and to minimize the
production costs.
3.1. System design with PID controller
PID control is the method of feedback control
that uses the Proportional, the Integral and the
Derivative as the main tools (Hensel et al. 2012,
Beshi et al. 2011 and 2012). The purpose of
control is to make the process variable y(t) as a
suitable value named set-point yr(t). To achieve
this purpose, the manipulated variable u(t) is
changed at the command of the controller. In the
present application, the process variable y(t) is
the temperature and the manipulated variable u(t)
is the command of the controller. The
“disturbance” is any factor, other than the
manipulated variable, that influences the process
variable. In some applications, however, a major
disturbance enters the process in a different way,
or plural disturbances need to be considered. The
error e(t) is defined as: e(t) = yr(t) – y(t). Where,
yr(t) is the desired trajectory and y(t) is the
measured variable. The PID is a controller that
takes the present, the past, and the future of the
error into consideration. After digital
implementation was introduced, a certain change
of the control system structure was proposed and
was adopted in many applications. However, that
change doesn’t influence the essential part of the
analysis and design of the PID controllers. The
transfer function C(s) of the PID controller is:
International Journal of Engineering, Applied and Management Sciences Paradigms
Volume 52, Issue 03, Quarter 03 (July-August-September 2018)
An Indexed and Referred Journal
ISSN (Online): 2320-6608
www.ijeam.com
IJEAM
www.ijeam.com
32
The proposed PI controller is built in
MATLAB/SIMULINK as indicated in Fig. 4.
The optimal control parameters are obtained
using the trial and error method. The
proportional gain (Kp) and the integral gain (KI)
are equal to 500 and 0.0001 respectively.
Table 1: Rule base of fuzzy logic controller
e Ce
NL NM NS ZE PS PM PL
NL NL NL NL NL NM NS ZE
NM NL NL NL NM NS ZE PS
NS NL NL NM NS ZE PS PM
ZE NL NM NS ZE PS PM PL
PS NM NS ZE PS PM PL PL
PM NS ZE PS PM PL PL PL
PL ZE PS PM PL PL PL PL
Figure 4: Membership functions of pts (e, and
ce) and output
3.2. System design with fuzzy logic
controller
The proposed MATLAB/SIMULINK model
of FLC for greenhouse is presented in Fig. 5.
Input variables of the FLC are the error (e) and
change of error (ce) which are normalized by an
input scaling factor. The design of the fuzzy
controller is related to the choice of knowledge
base, decision making logic, and defuzzification
mechanism. The rule base of the proposed FLC
is shown in Table 1. The inputs and output
variables are partitioned into seven fuzzy subsets.
They are presented by seven membership
functions (triangular shape) which are NL
(Negative Large), NS (Negative Small), ZE
(Zero), PS (Positive Small), and PL (Positive
Large) as indicated in Fig. 6 [10]. The
composition operation is the method by which
the controlled output is generated. The Max–Min
method is used. The output membership function
of each rule is given by the minimum.
Defuzzificaion for this system is the centre of
gravity method which is simple and fast.
3.3. System design with neural network
controller
The construction of the networks affects the
learning method and the efficiency of control
unit meanwhile the approximating error depends
on the number of neurons in the hidden layer and
the network inputs. ANN consists of three layers.
Six neurons (error, air temperature, wind speed,
solar radiation, air humidity and floor
temperature) are in the input layer, and fifteen
nodes are in the hidden layer. The output layer
contains one neuron for control flow rate of bare
tube heating system. The LOGSIGMOID
transfer function is used for hidden layer and
PURELIN for output layer. The learning
algorithm is carried by the back-propagation
technique.
International Journal of Engineering, Applied and Management Sciences Paradigms
Volume 52, Issue 03, Quarter 03 (July-August-September 2018)
An Indexed and Referred Journal
ISSN (Online): 2320-6608
www.ijeam.com
IJEAM
www.ijeam.com
33
Figure 5: Mean Square Error of the Neural
Network
Figure 6: MATLAB/SIMULINK of the
Neural Network Control
Figure 7: The Proposed Structure of ANFIS
Controller
Table 2: Configuration of ANFIS controller
Configuration
Type Sugeno
Number of inputs 2
Number of outputs 1
Number of rules 9
And method Prod
Or method Probor
Implication Prod
Aggregation Sum
Defuzzification
Wtaver
Figure 8: ANFIS IF-THEN rule
Fig. 5 shows that the best validation
performance for the neural network which can be
obtained at epoch 1000. Hence the network is
working well. MATLAB/SIMULINK of the
neural network is presented in Fig. 8.[10]
International Journal of Engineering, Applied and Management Sciences Paradigms
Volume 52, Issue 03, Quarter 03 (July-August-September 2018)
An Indexed and Referred Journal
ISSN (Online): 2320-6608
www.ijeam.com
IJEAM
www.ijeam.com
34
3.4 System design with ANFIS controller
Selection of the appropriate membership
functions and the rule base affects ANFIS
controller accuracy. In the ANFIS controller,
neural network algorithm is used to select a
proper rule base which is accomplished using the
back propagation algorithm. The proposed
structure of ANFIS controller is shown in Fig. 9.
The configuration of ANFIS controller is
indicated in Table 2[10]. The used fuzzy
inference technique is Sugeno method. A hybrid
training algorithm is used where a combination
of the gradient descent algorithm and a least
squares algorithm is used for an effective search
for the optimal parameters. The main benefit of
ANFIS is that it converges much faster, since it
reduces the search space dimensions of the back
propagation method used in neural networks. The
suggested ANFIS If-Then rules are
Figure 9: Surface viewer of ANFIS controller
Figure 10: MATLAB/SIMULINK model of
the heating system using fuzzy controller
Figure 11: MATLAB/SIMULINK Model of
the Heating System using ANN controller
Figure 12: MATLAB/SIMULINK Model of
the Heating System using ANFIS Controller
Table 3: Meteorological data of Ras Sedr
Time
(h)
Air
temp.
(◦C)
Radiation (w/m2) Wind
speed
(m/s)
Humidity
(%)
1 11 0 5.08 49
2 11 0 4.8 50
3 11 0 4.8 50
4 11 0 4.9 52
5 9 0 5.3 54
6 10 2 5.6 56
7 10 39.48 5.8 58
8 10 195.6 6.2 48
9 13 470.05 6.3 40
10 15 535.93 6.8 31
International Journal of Engineering, Applied and Management Sciences Paradigms
Volume 52, Issue 03, Quarter 03 (July-August-September 2018)
An Indexed and Referred Journal
ISSN (Online): 2320-6608
www.ijeam.com
IJEAM
www.ijeam.com
35
11 17 637.22 7.2 33
12 17 680.52 7.5 34
13 17.5 675.56 7.6 36
14 18 606.13 7.7 34
15 18 482.15 7.8 32
16 17.5 317.7 7.6 33
17 17 151.78 7.5 36
18 16 17.26 7.3 38
19 15 0 7 42
20 14 0 6.6 42
21 13 0 6.011 42
22 12 0 5.6 43
23 12 0 5.4 43
presented in Fig. 10. Eight If-Then rules and
three Gaussian bell type of membership function
are used. The obtained surface viewer of ANFIS
controller is shown in Fig. 11.
4. Simulation Results
In order to predict the greenhouse
performance, MATLAB/SIMULINK is used to
model and simulate the heating system. Three
proposed controllers are simulated as shown in
Figs. 10–12 [10]. The simulated model is carried
out to test the system performance under various
climatic conditions. Mathematical modelling of
greenhouse subsystem using
Figure 13: Mathematical modelling of
greenhouse system using
MATLAB/SIMULINK
Table 4: MAE & RMSE of Controller
Techniques
Controller type MAE RMSE
PI 0.5851 1.595
ANN 3.3927 9.8941
FLC 0.0716 0.5998
ANFIS 0.0074 0.1977
MATLAB/SIMULINK is presented in Fig. 13.
Meteorological data (air temperature, wind
speed, and relative humidity) are obtained from
New and Renewable Energy Authority (2005) as
shown in Table 3.
The heat loss due to convection; conduction to
the greenhouse soil; condensation; ventilation;
infiltration and long wave radiation is given in
Fig. 14. It is clear that, the infiltration loss
represents the minimum heat loss component
International Journal of Engineering, Applied and Management Sciences Paradigms
Volume 52, Issue 03, Quarter 03 (July-August-September 2018)
An Indexed and Referred Journal
ISSN (Online): 2320-6608
www.ijeam.com
IJEAM
www.ijeam.com
36
Figure 14: Greenhouse energy loss variation
over the day
Figure 15: Greenhouse indoor temperature
variation
Followed by the conduction to greenhouse soil
heat, while the condensation loss represents the
largest percentage of total greenhouse heat loss.
Four controller techniques, PI, neural
network, fuzzy, and ANFIS are used to adjust the
greenhouse desired temperature. The response of
the greenhouse indoor temperature is given by
Fig. 15 for various control techniques. It is clear
that, the response under proposed ANFIS
controller is very smooth and nearly the set point
than that given by PI, neural network, and fuzzy
controller. Table 4 indicates the RMSE and MAE
of the proposed controllers. The RMSE and
MAE of ANFIS controller are very smaller than
given by the other controllers. The simulation
results demonstrated that, the system response of
the ANFIS controller is very acceptable and
faster than others.
5. Conclusion
Due to the physical dynamics involved in a
greenhouse, the synthesis of a climate controller
becomes a sophisticated task using traditional
control techniques. This paper represented an
approach to apply a conventional and AI
techniques in a greenhouse climate control. PI
control, fuzzy logic control (FLC), artificial
neural network (ANN) and adaptive neuro-fuzzy
(ANFIS) control represent useful tools for
solving the nonlinearity problem of greenhouse
modelling. Training data for the present study for
artificial neural network and ANFIS control was
randomly collected from several simulations in
MATLAB/SIMULINK. The simulation results
proved that ANFIS controller can be applied
successfully to control the greenhouse indoor
temperature because of its effectiveness and fast
response time.
References
[1] Alhanafy, T., Zaghlool, F., El Din Moustafa,
A., 2010. Neuro fuzzy modeling scheme for
the prediction of air pollution. J. Am. Sci. 6
(12), 605–6012.
[2] Arvantis, K.G., Paraskevopoulos, P.N.,
Vernados, A.A., 2000. Multirate adaptative
temperature control of greenhouses.
Comput. Electron. Agric. 26 (3), 303–320.
[3] Barrera, J., Terada, R., Jr R. H., Hirata, N.
S., 2000. T.Automatic programming of
morphological machines by pac learning.
Fundamenta Informaticae., 41(1-2): pp. 229-
258.
[4] Bennis, N., Duplaix, J., Enea, G., Halua, M.,
Youlal, H., 2003. Greenhouse climate
modeling and robust control. Comput.
Electron. Agric. 61 (2), 96–107.
[5] Bennis, N., Duplaix, J., Enea, G., Haloua,
M., Youlal, H., 2005. An advanced control
of greenhouse climate. In: Proceedings of
the 33 International Symposium “Actual
Tasks on Agricultural Engineering”, 21–
25th February, Croatia, pp. 265–277.
[6] Charaf eddine LACHOURI, Aissa
BELMEGUENAI, Khaled MANSOURI .,
2016, FPGA Implementation of Adaptive
Neuro-Fuzzy Inference Systems Controller
for Greenhouse Climate
[7] Coelho, J., de Moura Oliveira, P., Cunha, J.,
2005. Greenhouse air temperature predictive
control using the particle swarm
International Journal of Engineering, Applied and Management Sciences Paradigms
Volume 52, Issue 03, Quarter 03 (July-August-September 2018)
An Indexed and Referred Journal
ISSN (Online): 2320-6608
www.ijeam.com
IJEAM
www.ijeam.com
37
optimisation algorithm. Comput. Electron.
Agric. 49, 330–344.
[8] Cunha, Boaventura J., 2003. Greenhouse
climate models: an overview. In:
Proceedings of the EFITA Conference, July
5–9, Debrecen, Hungary, pp. 823–829.
[9] Dadios, E., 2012. Fuzzy Logic-Emerging
Technologies and Applications. InTech
Publisher.
[10] Doaa M. Atia, Hanaa T. El-Madany., 2016.
Analysis and design of greenhouse
temperature control using adaptive neuro-
fuzzy inference system.
[11] Ferreira, P.M., Faria, E.A., Ruano, A.E.,
2002. Neural network models in greenhouse
air temperature prediction. Neurocomputing
43 (1), 51–75.
[12] Fourati, F., Chtourou, M., 2004. A
greenhouse control with feed-forward and
recurrent neural networks. Simul. Model.
Pract. Theory 15 (8), 1016–1028.
[13] Ghoumari, M., Tantau, H., Megas, D.,
Serrano, J., 2002. Real time nonlinear
constrained model predictive control of a
greenhouse. In: Proceedings of 15th IFAC
World Congress, Barcelona, Spain.
[14] Gupta, M., Jin, L., Homma, N., 2003. Static
and Dynamic Neural Networks: From
Fundamentals to Advanced Theory. Wiley-
IEEE Press.
[15] Iga, J., 2008. Modeling of the climate for a
greenhouse in the north-east of México. In:
Proceedings of the 17th World Congress,
Korea, pp. 9558–9563.
[16] Jensen M H and A.J. Malter 1995.
Protected agriculture: A global review.
World Bank Technical Paper No. 253
[17] Koutb, M., El-Rabaie, N., Awad, H.,
Hameed, I.A., 2004. Environmental control
for plants using intelligent control systems.
In: Proceedings of Artificial Intelligence in
Agriculture (AIA’04), IFAC, Cairo, pp.
101–106.
[18] Lafont, F., Balmat, J.F., 2002. Optimized
fuzzy control of a greenhouse. Fuzzy Sets
Syst. 128, 47–59.
[19] Lafont, F., Balmat, J.F., 2004. Fuzzy logic
to the identification and the command of the
multidimensional systems. Int. J. Comput.
Cognit. 2, 21–47.
[20] Márquez-Vera, Marco A., Ramos-
Fernández, Julio C., Cerecero-Natale, Luis
F., Lafont, Frédéric, Balmat, Jean-Franc¸ois,
Esparza-Villanueva, Jorge I., 2016.
Temperature control in a MISO greenhouse
by inverting its fuzzy model. Comput.
Electron. Agric. 124, 168–174.
[21] Mesmoudi, K., Soudani, A., Bournet, P.,
2010. Determination of the inside air
temperature of a greenhouse with tomato
crop under hot and arid climates. Appl. Sci.
Environ. Manag. 5, 117–129.
[22] Mote, T., Lokhande, S., 2012. Temperature
control system using ANFIS. Int. J. Soft
Comput. Eng. 2, 156–161.
[23] New and Renewable Energy Authority,
Ministry of Electricity and Energy, 2005.
Wind Atlas for Egypt Measurements and
Modeling 1991-2005, Cairo, Egypt.
[24] Y. El Afou1, 2
, L. Belkoura2, M. Outanoute
1,
M. Guerbaoui1, A. Rahali
1, A. Ed-Dahhak
3,
A. Lachhab3, C. Join
4, B. Bouchikhi
1
Feedback Techniques Using PID and
PIIntelligent for Greenhouse Temperature
Control
[25] Satyajit Ramesh Potdar et. Al., 2017.
Greenhouse Air-Temperature Modelling and
Fuzzy Logic Control, IJEER.