ARTICLE IN PRESS

APPLIED
ENERGY
Applied Energy xxx (2007) xxx–xxx
www.elsevier.com/locate/apenergy

Variable structure TITO fuzzy-logic controller

implementation for a solar air-conditioning system
J.N. Lygouras a,*, V.S. Kodogiannis b, Th. Pachidis a,
K.N. Tarchanidis c, C.S. Koukourlis d
a
Laboratory of Electronics, School of Electrical and Computer Engineering, Democritus University of Thrace, GR-67100 Xanthi, Greece
b
Centre for Systems Analysis, School of Computer Science, University of Westminster, London HA1 3TP, UK
c
Department of Petroleum Technology, Technological Education Institute of Kavala, GR-65404, Kavala, Greece
d
Laboratory of Telecommunications, School of Electrical and Computer Engineering, Democritus University of Thrace,
GR-67100 Xanthi, Greece

Accepted 9 July 2007

Abstract

The design and implementation of a Two-Input/Two-Output (TITO) variable structure fuzzy-logic controller for a
solar-powered air-conditioning system is described in this paper. Two DC motors are used to drive the generator pump
and the feed pump of the solar air-conditioner. The first affects the temperature in the generator of the solar air-condi-
tioner, while the second, the pressure in the power loop. The difficulty of Multi-Input/Multi-Output (MIMO) systems con-
trol is how to overcome the coupling effects among each degree of freedom. First, a traditional fuzzy-controller has been
designed, its output being one of the components of the control signal for each DC motor driver. Secondly, according to
the characteristics of the system’s dynamics coupling, an appropriate coupling fuzzy-controller (CFC) is incorporated into
a traditional fuzzy-controller (TFC) to compensate for the dynamic coupling among each degree of freedom. This control
strategy simplifies the implementation problem of fuzzy control, but can also improve the control performance. This mixed
fuzzy controller (MFC) can effectively improve the coupling effects of the systems, and this control strategy is easy to
design and implement. Experimental results from the implemented system are presented.
 2007 Elsevier Ltd. All rights reserved.

Keywords: Solar air-conditioner; Fuzzy-logic controller; MIMO system; Mixed fuzzy controller; Microcontroller

Introduction

Energy consumption in commercial and residential buildings is steadily increasing every year. The higher
living and working standards, the adverse outdoor conditions in urban environments and reduced prices of

*
Corresponding author. Tel.: +30 25410 79578; fax: +30 25410 79570.
E-mail addresses: ilygour@ee.duth.gr (J.N. Lygouras), kodogiv@wmin.ac.uk (V.S. Kodogiannis), ktarch@xan.forthnet.gr (K.N.
Tarchanidis), ckoukou@ee.duth.gr (C.S. Koukourlis).

0306-2619/$ - see front matter  2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.apenergy.2007.07.001

Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001
 ARTICLE IN PRESS

2 J.N. Lygouras et al. / Applied Energy xxx (2007) xxx–xxx

air-conditioning units, have caused a significant increase in demand for air-conditioning in buildings. Annual
energy use of room air-conditioners was 6 TJ in 1990, 40 TJ in 1996 and is estimated to reach 160 TJ in 2010.
As a result of the projected world energy shortage, the use of solar energy for environmental control is receiv-
ing much attention in the engineering-science literature. With suitable technology, solar cooling can help alle-
viate the problem. Thus, the use of solar power instead of electric power in air-conditioning systems is gaining
increasing interest during the last years. The trend towards reusable energy-sources is a significant reason for
this interest and especially for hybrid (solar-electric power) systems. The fact that peak cooling-demand in
summer is associated with high solar radiation offers an excellent opportunity to exploit solar thermal tech-
nologies that can match heat-driven cooling technologies. Of particular interest are urban areas, where adverse
outdoor conditions, as a result of higher outdoor pollution and the urban heat-island effect, encourage the use
of mechanical air-conditioning with a direct impact on peak electrical-energy use [1]. Commercial application
of solar energy for air conditioning purposes is relatively new. Lamp and Ziegler [2] gave an overview of the
European research on solar-assisted air-conditioning up to 1996. Tsoutsos et al. [3] presented a study of the
economic feasibility of solar cooling technologies.
Different heat-driven cooling technologies are available on the market, particularly for systems of above
40 kW, which can be used in combination with solar thermal-collectors. The main obstacles for large-scale
application, beside the high first-cost, are the lack of practical experience and acquaintance among architects,
builders and planners with the design, control and operation of these systems. For smaller-scale systems, there
is no market available technology. Therefore, the development of low power cooling and air-conditioning sys-
tems is of particular interest [4].
However, operating this type of system presents certain issues that must be addressed by the control strat-
egy. First, the primary-energy source (the Sun) cannot be manipulated, as in the case of any other conven-
tional thermal-process. Secondly, great disturbances exist in the process, mainly due to changes in
environmental conditions. There are also dead-times related to fluid transportation which are variable depend-
ing on the operating conditions. Finally, the cooling demand is variable, since it depends on the occupancy
rate and the kind of activity that is being carried out in the cooled space. Several control-strategies have been
tested at solar-power plants to address these problems [5,6]. Model Predictive Controllers have also received a
lot of attention in the last few decades, both within the research control community and in industry [7]. The
basic idea is to calculate a sequence of future control-signals, in such a way that it minimizes a multistage cost-
function defined over a control horizon.
In the recent years, the development of sophisticated control-systems, which often include standard proces-
sors (lP, lC, DSP) [8,9] and specific hardware components (ASSP, FPGA, ASIC) [10,11], has gained increas-
ing interest. The design of such systems represents a difficult and studious task. A new trend is to use
configurable logic-circuits (FPGA, CPLD, etc.) with processor units on the same board. This allows rapid
and efficient adaptation of the used board to new applications [12]. An interesting area of application of
the above type of controllers is the implementation of a DC motor controller, due to the wide range of appli-
cations of this type of motor [12–14]. With microcontrollers being available in different configurations, low
cost, small die sizes, large on-chip memory, power management features and high clock rates, it is reasonable
to apply fuzzy-logic in an inexpensive microcontroller to implement the fuzzy control loop.
In a related work presenting the basic aspects of a solar-powered air conditioning system [15], the imple-
mentation of a fuzzy-logic controller for a solar-powered air-conditioning system and its advantages have been
investigated. In the application described, one DC motor is used to drive the generator pump in the heating
cycle as is described below and a second DC motor is used to drive the feed pump in the cooling cycle of the
solar air-conditioner. A variable structure fuzzy-logic controller is proposed using a 7 · 7 fuzzy matrix to
assign the controller output with respect to the error value and the derivative of the error. The rotational speed
of the generator pump in the system is considered to affect the quantity of energy supplied to the power loop
and the rotational speed of the feed pump to affect the quantity of heat absorption from the environment. This
assumption is true if the two processes are considered as completely independent and thus the Two-Input/
Two-Output (TITO) system is considered uncoupled. In this system, the two inputs are the control voltages
applied to the two motors, while the two outputs are the rotational speeds of the two motors related directly
with the temperature of the heated liquid in the first loop and the pressure of the liquid in the second loop. In
practice, the situation is not as simple as this assumption, since there is a considerable dependence/coupling
Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001
 ARTICLE IN PRESS

J.N. Lygouras et al. / Applied Energy xxx (2007) xxx–xxx 3

between the working parameters in the two loops of the system. In this paper, this dependence is considered
and a more complex TITO model is studied.
Similarly, in [16], in order to control a MIMO system, a mixed fuzzy-controller has been developed to
improve the performance and to solve the problem of computational burden and dynamic uncertainty asso-
ciated with MIMO systems. In this work, a traditional fuzzy controller (TFC) is first designed from the view-
point of a single-input/single-output (SISO) system for controlling each degree of freedom of the MIMO
system. Then, an appropriate coupling fuzzy controller is also designed according to the characteristics of
the system’s dynamic coupling and incorporated into a TFC, to compensate for coupling effects between
the degrees of freedom. In [17], a self-learning fuzzy-logic system is proposed for control of unknown multi-
ple-input multiple-output plants. To verify the applicability of the proposed controller, a two-link robotic
manipulator with a complex dynamic model is studied, while in [18] this method is applied to a dynamic absor-
ber with a two-level mass-spring damper structure.
In [19], analytical structures of TITO Mamdani fuzzy PI/PD controllers are investigated with respect to a
conventional PI/PD control and variable-gain control: the characteristics of the fuzzy-controller are analysed
in relation to those of the corresponding TITO linear PI/PD control. The application of a self-organizing
fuzzy proportional-integral-derivative (SOF-PID) controller to a MIMO non-linear revolute-joint robot
arm is studied in [20], where the SOF controller is a learning supervisory controller, making small changes
to the values of the PID gains while the system is in operation.

Description of the solar air-conditioning system

In order to demonstrate this idea of a mixed fuzzy-controller, a practical application is proposed in the fol-
lowing sections. Numerical simulation with experimentation is employed to verify this concept. Several tech-
niques have been proposed in the literature for the direct conversion of the solar energy to heating, cooling or
both. The most common way for cooling is to exert mechanical pressure to the cooling liquid, usually using an
electric compressor and to pass it through a valve to a lower pressure chamber. The evaporation of the liquid
at this point is accompanied by heat absorption and thus the ‘‘cooling’’ effect is produced. In the case of solar
air-conditioning a different principle is adopted. The cooling cycle is based on the operation of a special ejec-
tor, as shown in Fig. 1. There are three loops in this configuration. In the first loop (i.e. the heating loop), a

Heating loop
Solar Panel

WG Heat Exchanger

QG
Generator
Pump

3 Generator
4
Feed Pump
Power loop
WF
QC
1 Ejector
2 Condenser

Throttle Valve Refrigeration loop

QE

6
5 Evaporator

Fig. 1. The solar air-conditioner.

Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001
 ARTICLE IN PRESS

4 J.N. Lygouras et al. / Applied Energy xxx (2007) xxx–xxx

working liquid – oil in our case – is heated in the high-efficiency solar panels to a temperature about 130 C.
The second loop (1-2-3-4-1) includes the ejector, the condenser, the feed pump and the generator. The third loop
(1-2-5-6-1) includes the ejector, the condenser, the throttle valve and the evaporator. The second loop is called
power loop, while the third is the refrigeration loop. The heating energy QG derived from the solar panels is
given to the power loop, in the generator, through the heat exchanger and by the aid of the generator pump.
At this point, the evaporation of the cooling liquid is achieved through heating, when it is passing from point 3
to 4. At point 4, the vapour has a low speed and high-pressure. This high-pressure vapour passes at high speed
through the ejector to the condenser. In the condenser, the vapour is converted again to liquid. This phase
change is accompanied by heat dissipation QC, which is rejected to the environment.
The cooling liquid coming out of the condenser passed through the pump, thus closing the power cycle. A
part of the cooling liquid, passing through the throttle valve to point 5, is fed at low pressure to the evapo-
rator. Here, the evaporation of the cooling liquid at point 6 is achieved due to the operation of the ejector.
When the high-pressure vapour coming from point 4 is injected into the ejector the resultant low-pressure
in the orifice at point 6, is forcing the conversion to vapour in the evaporator. Thus, vapour coming from
the evaporator is mixed with vapour coming from the generator and is converted to liquid in the condenser.
The evaporation, as is obvious, is accompanied by heat absorption QE from the environment, thus producing
the cooling effect in the evaporator. In the air-conditioning system the evaporator section is the indoor unit,
while the condenser section corresponds to the outdoor unit.
The purpose of the heating loop is the energy transfer in the form of heat from the incident solar-radiation
in the oil reservoir. The generator pump is circulating the oil in the loop which is heated by passing through
the solar panel. The achieved oil temperature is dependent on the intensity of the solar radiation as well as the
time the oil remains inside the panel. The circulating oil transfers the heat to the generator. The oil temper-
ature in the output of the solar panels must be about 125–135 C. The pump must adjust the oil circulation in
the solar panel loop in order to achieve the desired temperature. If the solar radiation is low, then the gener-
ator pump must reduce its speed so that the oil remains longer in the solar panel. On the other hand, if the
solar radiation is high, then the oil temperature increases rapidly. In this case the rotational speed of the pump
must be high so that the oil remains inside the panels for less time. For the implementation of the above logic,
the control system must observe the oil temperature in the solar panel outlet Toilpout. The purpose of the above
control system is to retain the oil temperature within the desired limits.
The oil is going out at lower temperature Toilpin from the generator and is driven again through the solar
panels. The transfer of energy in the generator unit increases the temperature of the cooling liquid which is
converted to vapour. This in turn is followed by a rise of the pressure PG of the vapour. Thus, the rotational
speed of the generator pump is directly related to the heat transfer from the solar panels to the refrigeration
loop and the rotational speed of the feed pump affects the pressure of the vapour in the power loop and con-
sequently the quantity of heat absorption from the environment. Table 1 indicates the upper and lower limits
for some critical parameters of the solar air-conditioning system.The two parameters, TG and PG, are affected
by the generator and feed pump speeds, respectively.
In practice, the situation is not exactly as simple as described in the previous paragraph, since there is a
considerable dependence between the working parameters in the first two loops of the system. In the first loop,
the parameter which affects the rotational speed of the DC motor is the temperature TG of the heated liquid
(measured in the generator). It must be retained as constant as is feasible. It is supposed that the solar panels
are capable of providing the appropriate energy QG to do so. In the second loop, the parameter which affects
the rotational speed of the DC motor is the pressure PG of the cooling liquid (measured in the generator’s
output). It should also be invariant. In a previous paper [15] and for simplicity reasons, the two loops were

Table 1
Values for some critical parameters
Pressure (PG) Temperature (TG)
Lower operational-limit 10.5 bar 102 C
Normal operational-point 14.5 bar 116 C
Upper operational-limit 16.5 bar 125 C

Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001
 ARTICLE IN PRESS

J.N. Lygouras et al. / Applied Energy xxx (2007) xxx–xxx 5

r1(k) + e1(k) u1(k)

- MIMO y1(k)
r2(k) e2(k) u2(k) MIMO
... + ... Fuzzy ... y2(k)
- ...
eJ(k) uQ(k) Plant
rJ(k) +
Controller yJ(k)
-

Fig. 2. Structure of a MIMO fuzzy-logic controller for a closed-loop MIMO system.

initially considered as completely independent and thus two identical SISO control systems have been
designed, based on this assumption. In practice, this is not feasible since the change in each input (rotational
speed of the generator and feed pump) affects the measured parameters in both loops. Thus, there is a con-
siderable interdependence between the working parameters of the system. In this investigation, this depen-
dence is considered and a more complex TITO model is evolved. The above system is a Two-Input
(temperature TG and pressure PG), Two-Output (generator pump and feed pump speeds) closed-loop control
system. The fuzzy inference logic applied the Max–Min product composition to operate the fuzzy-control
rules. Finally, this study employs the area method to defuzzify the output variables to attain the accurate
objectives for controlling this system.

Fuzzy-logic controller for MIMO systems

The general structure of a MIMO fuzzy-logic controller (FLC) for a closed-loop MIMO system is shown in
Fig. 2. The controlled system is supposed to have Q inputs and J outputs. The FLC calculates the Q inputs to
the plant. The closed-loop system has in total J inputs and J outputs: thus J errors are produced. The errors
ej(t) and their changes cej(t), j = 1, 2, . . ., J, are treated as linguistic FLC input variables. Although fuzzy-con-
trol theory has been successfully employed in many control-engineering fields, its control-strategies were
mostly used to control SISO systems or the fuzzy controller was designed based on the SISO system point-
of-view, in spite of the dynamic coupling of a MIMO control system. The difference between MIMO systems
control and SISO systems control is how to estimate and compensate the interaction among each degree of
freedom.
The matrix of fuzzy control rules for SISO systems has a dimension of F2, where F is the number of used
fuzzy sets. The direct1 fuzzy-controller design for MIMO systems encounters the following problems:

• The quantity of fuzzy rules increases exponentially as the system’s inputs increase.
• Cross coupling exists between the inputs and outputs of a physical system. In an ideal case, it can be
assumed that each input manipulates only one output. However, it is a special case to find in practice a
decoupled system which requires an equal number of inputs and outputs.

Generally, the maximum possible (i.e. upper bound) number of rules NR is given by:
N R ¼ r  F ðm;nÞ ð1Þ
where: r is the dimension of input vector u, m is the dimension of output vector y, F is the number of fuzzy sets,
and n is the number of input variables in the fuzzy rule (e.g. error, change of error, etc.) From Eq. (1), NR can
be seen to have a linear relation for increasing r while m is constant. But it has an exponential relation for
increasing m while r is constant. Therefore, if NR is large, this leads to a speed down of the inference mech-
anism of the fuzzy controller and much storage memory is required. This is a significant problem, especially in
the case where a microcontroller is used to implement the control functions, due to the limited internal mem-

1
Direct design means one controller for the overall system.

Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001
 ARTICLE IN PRESS

6 J.N. Lygouras et al. / Applied Energy xxx (2007) xxx–xxx

ory space included in those devices. The above equation affirms that the direct synthesis of a fuzzy controller
for a multivariable system is not practical. Thus it is preferable to develop an alternative design methodology
to overcome this problem.
A traditional fuzzy-logic controller, operating with system output error ei(t) and error change cei(t), was
chosen to control each degree of freedom of the MIMO system. Let the vector Y(t) represent the collection
of J plant outputs, and let the vector R(t) be the collection of J desired plant outputs. Also, let the collection
of Q plant inputs be denoted by U(t). The vector of errors is given by:
2 3 2 3
e1 ðkÞ r1 ðkÞ  y 1 ðkÞ
6 e ðkÞ 7 6 r ðkÞ  y ðkÞ 7
6 2 7 6 2 2 7
6 7 6 7
6: 7 6: 7
6
EðkÞ ¼ 6 7 ¼ RðkÞ  YðkÞ ¼ 66 7 ð2Þ
7 7
6: 7 6: 7
6 7 6 7
4: 5 4: 5
eJ ðkÞ rJ ðkÞ  y J ðkÞ
2 3
e1 ðkÞ  e1 ðk  1Þ
6 e ðkÞ  e ðk  1Þ 7
6 2 2 7
6 7
6: 7
6
C E ðkÞ ¼ 6 7 ð3Þ
7
6: 7
6 7
4: 5
eJ ðkÞ  eJ ðk  1Þ
Several types of fuzzy-control rules have been employed in the literature, as for example the method of Takagi
and Sugeno, by which the consequence of a control rule is a linear combination of the components of the crisp
vector H(t). A control rule can be given by:
Rule i : IF ðx1 is s1 ðh1 ðtÞÞ and x2 is s2 ðh2 ðtÞÞ and . . . and xM is sM ðhM ðtÞÞ
THEN
rk ðtÞ ¼ ak;0 þ ak;1 h1 ðtÞ þ ak;2 h2 ðtÞ þ    þ ak;M hM ðtÞ ð4Þ
The difficulty of employing traditional fuzzy-control theory for MIMO systems is how to overcome the cou-
pling effects among each degree of freedom. Because the dynamic model of MIMO systems possesses the usual
characteristics of non-linear and complicated dynamic coupling, it is difficult to develop: an accurate dynamic
model and to decouple it for controller design. Thus model-free intelligent control-strategies are employed for
designing the controllers of MIMO systems. Hence the traditional model based SISO system control scheme is
not easy to implement on complicated MIMO systems due to the large computational burden. Therefore, the
model-free intelligent control strategy is gradually gaining attention. In addition, the control rules and con-
troller computation are growing exponentially with respect to the number of considered variables. It is obvi-
ous that the difficulty of MIMO systems control is how to overcome the coupling effects among each degree of
freedom. Therefore, an appropriate coupling fuzzy controller must be incorporated into a TFC to control the
MIMO system to compensate for the dynamics coupling among each degree of freedom. This mixed fuzzy
controller (MFC) can effectively improve the coupling effects of the systems, and this control strategy is easy
to design and implement.
The concept of adding an appropriate coupling controller to compensate for the coupling effect is proposed
to enhance the control performance of the TITO control system of the solar air-conditioning system. The new
control approach combines a traditional fuzzy controller and an appropriate coupling fuzzy controller. Its
control block diagram is shown in Fig. 3. The design procedure of the mixed fuzzy-control strategy is to design
a TFC first, to control each degree-of-freedom of this MIMO system individually [15]. Then, an appropriate
coupling fuzzy controller is designed to compensate for the coupling effects of system dynamics among each
degree of freedom. Thus, a good control performance is obtained and also the total number of rules can be
kept small. In this diagram, the two inputs of the closed-loop system are the desired temperature TG and
Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001
 ARTICLE IN PRESS

J.N. Lygouras et al. / Applied Energy xxx (2007) xxx–xxx 7

Microcontroller

r1(k) + - e1 Fuzzy u1(k) + U1 y1(k)

Controller 1 +
T
e1 U2 →1 I
-1
Z Coupling Fuzzy T
e1 Controller 1→2 O

P
e2 Coupling Fuzzy
l
Z -1 Controller 2→1
e2 a
U1→ 2 n
r2(k) + Fuzzy
+ t y2(k)
Controller 2 u2(k)
- e2 U2
+

Fig. 3. Block diagram of the TITO mixed fuzzy-control strategy considered for the air-conditioner: the section implemented using a 12-bit
microcontroller is within the dashed lines.

the pressure PG of the coolant in the generator. The two outputs of the controller are the speeds of the gen-
erator and feed pump. In order to simplify the controller design, the two inputs, two outputs system, is sim-
plified by considering it as two subsystems having two inputs and one output.
Generally, the membership function of a fuzzy-logic controller is difficult to decide for an unknown control
system. In the solar air-conditioning system, triangular-type membership functions [15] have been used to con-
vert these input and output variables (ei(t), cei(t) and Dui(t)) into linguistic control-variables
(NL,NM, . . ., PM,PL), where the subscript i = 1, 2, 3, . . ., m is used for indicating the temperature or pressure
error, the error change and the control output in each degree of freedom. The fuzzy control rules employed to
control a MIMO solar air-conditioning system are listed in Table 2.

Table 2
Fuzzy rules employed to control a TFC for the solar air-conditioning system

ei

cei NL NM NS ZE PS PM PL

PL ZE PS PM PL

NS ZE PS PM PL PL
PM

PS NM NS ZE PS PM

ZE NL NM NS ZE PS PM PL

NS NM NS ZE PS PM

NL NL NM NS ZE PS
NM

NL NL NM NS ZE

Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001
 ARTICLE IN PRESS

8 J.N. Lygouras et al. / Applied Energy xxx (2007) xxx–xxx

In addition to the above design process, the actual control input voltage of the actuator for this TFC can be
written as:
ui ðkÞ ¼ ui ðk  1Þ þ Dui ðkÞ ð5Þ
where Dui(k) indicates the voltage increment of the ith degree on the k-step sampling interval; ui(k) and
ui(k  1) expresses the system control input-voltage of the ith degree on the k step and k  1 step sampling
intervals, respectively.
In a real MIMO system, the control output is influenced by more than one variable. According to the sys-
tem characteristics, these system variables are obviously interactive.
Y i ¼ P i ðu1 ; u2 ; . . . ; um ; Y 1 ; Y 2 ; . . . ; Y n Þ ð6Þ
or
Y l ¼ P l ðul ; Y 1 ; Y 2 ; . . . ; Y n Þ ð7Þ
where Pi(.) and Pl(.) are complicated coupling functions that are difficult to define and derive. According to the
analysis of the dynamic Eq. (6) or (7), it is well known that ui is the main effect and ul is the secondary effect for
the output Yi. Similarly, for the output, Yl, ul is the main effect and ui is the secondary effect. The main effect
on the system is controlled using a TFC. The secondary effect on the system is controlled by designing an
appropriate coupling fuzzy controller. It is clear that the control structure of the fuzzy control system is very
complicated when the input variable is multi-degree and the output variable is one degree, or more than one
degree. These parameters of a fuzzy control system are not easily decided, because the fuzzy control rules will
be generated as a geometric series, and much computing time will be required such that a general CPU is not
acceptable. Hence, the design procedure of this coupling fuzzy controller should be simplified to reduce the
computing burden during implementation. The input variable of the coupling fuzzy controller is chosen
mainly as two coupling affected factors, such that the structure of the fuzzy rules can be built conventionally.
However, both the input and output variables generally exceed two control degrees for MIMO systems, such
that the design of the coupling fuzzy controller becomes very complicated, and is difficult to implement.
According to the coupling situation among each degree of freedom for MIMO systems, which by way of
dynamics coupling characteristics analysis, a few of the mainly affected coupling factors are first considered
to design an appropriate coupling fuzzy controller to compensate for dynamic coupling effects and enhance
control performances of the systems. For example, output responses of nearer two degrees should be heavily
affected for a MIMO system, which is chosen as an input variable of a coupling fuzzy controller, and neglected
output responses of other degrees will be weakly affected. This design method for the coupling fuzzy controller
will be able to effectively decrease the difficulties of the coupling fuzzy controller design.
The fuzzy inference and defuzzification methods are similar to those of the TFC. The output of the coupling
fuzzy controller is chosen directly as the coupling control input-voltage. The main reason is that there is a dif-
ferent coupling effect for each sampling interval and it does not have an accumulating feature. The coupling
effect is incorporated into the TFC for each step to improve the system’s performance and robustness. There-
fore, the total control input-voltage of this MFC is represented as:
U i ðkÞ ¼ ui ðkÞ þ U ðkÞl!i ; i 6¼ l ð8Þ
where ui(k) expresses the systems control input-voltage of the ith degree of a TFC. U(k)l!i, represents the cou-
pling effect control of the lth degree relative to the ith degree of the coupling fuzzy controller and, since we are
dealing with a TITO system, i = {1, 2} and l = {1, 2}.

Knowledge-based design

Fuzzification

As has already been mentioned, the inputs of the controller are the desired temperature TG and the desired
pressure PG of the coolant in the generator. The limit values for temperature are between 0 C and 150 C,
while for pressure between 0 and 17 bar. Outputs of the controller are the speeds of the two pumps which
are controlled by the PWM technique. The duty-cycle variable has a value between zero and 1023 and it cor-
Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001
 ARTICLE IN PRESS

J.N. Lygouras et al. / Applied Energy xxx (2007) xxx–xxx 9

responds to 0% and 100% of the rational speed, respectively. From the inputs of the controller the following
fuzzy variables are used:

• The temperature error, which is determined as Terror = 116  TG,TG [Celsius degrees].
• The pressure error, which is determined as Perror = 145  PG,PG [bar].

To design a (linguistic) fuzzy model based on available expert knowledge, the following steps can be followed:

• Select the input and output variables, the structure of the rules, and the inference and defuzzification
methods.
• Decide on the number of linguistic terms for each variable and define the corresponding membership
functions.
• Formulate the available knowledge in terms of fuzzy ‘‘if-then’’ rules.
• Validate the model (typically by using data). If the model does not meet the expected performance, iterate
the above design steps.

It should be noted that the success of this method depends on the problem at hand, and the extent and quality
of the available knowledge. For some problems, the knowledge-based design may lead quickly to useful mod-
els, while for others it may be a very time-consuming and inefficient procedure (especially manual fine-tuning
of the model parameters). Therefore, it is useful to combine the knowledge-based design with a data-driven
tuning of the model parameters.
Since the additional coupling fuzzy controller is introduced to compensate for the unknown coupling effects
of MIMO systems, the mixed fuzzy-control strategy proposed here should be employed to control any com-
plicated dynamic coupling MIMO systems. Therefore, the application of this control strategy is not con-
strained to a specific MIMO control system. In order to evaluate the control performance of the mixed
fuzzy controller for MIMO control-systems, the solar air-conditioning system was constructed and tested
as shown in Fig. 1. As was mentioned above, the generator and the feed pump affect the temperature (TG)
and the pressure (PG) of the coolant liquid in the generator. Some general rules of how the two pumps affect
the two parameters (i.e. temperature and pressure) are needed. The two pumps affect the temperature and the
pressure as follows: the speed increment of the generator’s pump will cause increments in temperature TG and
pressure PG. The inverse is also valid. In parallel, an increment in the speed of the feed pumps results in a
reduction of the temperature and an increment of the pressure. On the other hand, reduction of the speed
of the feed pump will cause an increment of the temperature and a reduction of the pressure. The purpose
of the 2D fuzzy-logic system is to control the speed of the two pumps in a way that will allow one to maintain
the pressure around 14.5 bar and the temperature around to 116 C, which is the ideal operation point of the
air-conditioner.
The speed of each pump must be determined in such way that the system reaches the ideal operating-point
or be within the accepted limits around this point. This can be achieved if a set of linguistic rules are adopted
for the control of the two pumps. Some of these general rules are as follows:

• If the temperature TG is lower than the desired value, while the pressure is within normal operation limits,
then the feed pump speed must be reduced.
• If the temperature TG is higher than the desired value, while the pressure is within normal operation limits,
then the feed pump speed must be increased.
• If the temperature TG and the pressure PG are simultaneously increased towards the upper operation limits,
then the generator pump speed must be reduced.
• If the temperature TG and the pressure PG are simultaneously reduced towards their lower operation limits,
then the generator pump speed must be increased.
• If the pressure PG is higher than 16.5 bar then the generator pump must stop immediately and start again if
the pressure falls below 14 bar.
• If the pressure PG is lower than 10.5 bar, then the feed pump must stop immediately and start again if the
pressure reaches 13 bar.
Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001
 ARTICLE IN PRESS

10 J.N. Lygouras et al. / Applied Energy xxx (2007) xxx–xxx

Some general rules have been set for the coupling fuzzy controller: these rules are based on the fact that the
pressure PG of the coolant is affected ‘‘to a higher degree’’ by the speed of the generator’s pump and ‘‘less’’ by
that of the feed pump, while, the temperature TG is affected ‘‘more’’ by that of the speed of the feed pump and
‘‘less’’ by the generator pump. Therefore, during the designing of the fuzzy controller, the strategy that will be
followed is as follows. The variations of the temperature TG will cause greater variations in the speed of the
feed pump, while, the variations of the pressure PG will cause greater variations in the speed of the generator’s
pump. The controller acts properly in order to sustain the temperature and the pressure within the desired
limits of the air-conditioner, in order to achieve the ideal operation point. According to each degree of free-
dom for the dynamic coupling effect between the main and the secondary systems, the fuzzy-control rules of
the coupling fuzzy controller can be decided and adjusted based upon the systems output response. For exam-
ple, the coupling fuzzy control rules for controlling the TITO air-conditioning system are shown in Table 3.
The fuzzy inference and defuzzification methods are similar to those of the TFC.
The fuzzy sets that have been chosen for each variable for the coupling fuzzy controller are: NL (Negative
Large), ZE (Zero) and PL (Positive Large). The membership functions for each fuzzy set have been deter-
mined based on the rules that have been previously described, i.e. the variation of the temperature will affect
‘‘more’’ the speed of the feed pump, while the pressure will affect ‘‘more’’ the generator’s pump. Therefore, the
membership functions that have been set for the control of the generator’s pump are slightly different from
those that have been set for the control of the feed pump.

The defuzzification process

At the stage of the defuzzification, all the actions that have been activated are combined and converted into
a single non-fuzzy output signal, which is the control signal of the system. An important factor here is the
definition of the fuzzy set for the control signal (i.e. the output of the controller). The output levels of the
defuzzyfier are thirteen and their representation will be also triangular, their position depending on the
non-linearities existing in the system. To achieve this, we develop the control curve of the system representing
the I/O relation of the system and, based on that information, we define the output degree of the membership
function with the aim of minimizing the effect of the non-linearity.
The method that has been adopted in this research project for the estimation of the control signal is the
centroid method. It results in an output action that is associated with the centre of gravity of the active rule
outputs. For a vertically symmetric degree of the membership function, as in our case, the centroid of a fuzzy
set always corresponds to a single output value, regardless of the value of the input degree of membership. The
control signal is computed using the following relation taking into account all the fuzzy rules that fire, as well
as the weights of each rule:
Pn
j¼1 actionj xcentroidj
output signal ¼ R ¼ Pn K ð9Þ
j¼1 actioni

Table 3
Fuzzy-control rules employed for the coupling fuzzy controller for the solar air-conditioning system

ei

NL ZE PL
cei

PL ZE NL NL

ZE PL ZE NL

NL PL PL ZE

Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001
 ARTICLE IN PRESS

J.N. Lygouras et al. / Applied Energy xxx (2007) xxx–xxx 11

The variable K is a design parameter and is used to scale the output signal and to improve the performance of
the controller. The above control variables, in proportion to the values they take each time the loop is imple-
mented, will determine the one and only output of the system, which is the duty cycle of the PWM.

Experimental results

The hardware implementation of the system has been experimentally verified using the solar air-condition-
ing system. The control algorithm described has been used to drive the main two DC motors (the generator

START

Fuzzy Variables
Initialization

A/D Converter
Initialization

PWM Module
Initilization

A/D Conversion
TGact, PGact

T Error Calculation
e1(k)=TGref1(k)-TGact(k)

P Error Calculation
e2(k)=PGref2(k)-PGact(k)

Der. Error 1 Calculation

D1(error) = e1(k)-e1(k-1)

Der. Error 2 Calculation

D2(error) = e2(k)-e2(k-1)

Degree of Membership for TFC

Variables Calculation

Degree of Membership for

CFC Variables Calculation

Fuzzy Rules
Application

Defuzzification

Application of the new

duty cycles

Fig. 4. Flow chart implemented in the software TITO Fuzzy-controller.

Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001
 ARTICLE IN PRESS

12 J.N. Lygouras et al. / Applied Energy xxx (2007) xxx–xxx

15

10

y1(k)
5

0 100 200 300 400 500 600 700 800

Time (seconds)

10
y2(k)

10
0 100 200 300 400 500 600 700 800
Time (seconds)

6
u1(k)

0 100 200 300 400 500 600 700 800

Time (seconds)

10
u2(k)

-5
0 100 200 300 400 500 600 700 800
Time (seconds)

Fig. 5. Step responses of the system with SISO PI tuning (- - -) and MFC TITO fuzzy controller (___).

pump and the feed pump) of the solar air-conditioner. A tacho-generator is used in the feedback of each motor
for the actual rotational speed reading. There is a linear relation between the voltage of the tacho-generator
output waveform and the rotational speed of the motor shaft.A Pulse Width Modulation technique (PWM) is
used to drive the DC motor at the desired speed. The microcontroller computes the duration d of the PWM
waveform accordingly, and the appropriate signals are applied to the driving circuit (i.e. the DC chopper).
The control algorithms are implemented using a 16-bit microcontroller including an 8-channels 12-bit A/D
converter and, in order to avoid the use of an external D/A converter, the PWM technique is used (since the
microcontroller provides such a capability) to drive directly the power circuits. The command signal for the
first DC motor driving the Feed Pump is the pressure measured in the generator’s output, while the command
Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001
 ARTICLE IN PRESS

J.N. Lygouras et al. / Applied Energy xxx (2007) xxx–xxx 13

signal for the DC motor driving the generator pump, is the oil temperature measured in the solar panel output.
In order to achieve the desired correspondence between the input and the output variables in the two con-
trolled loops (pressure-rotational speed and temperature-rotational speed), a look-up table is used which is
stored in the external memory of the microcontroller. The look-up table operation is executed after each
A/D conversion. In Fig. 4, the flow chart implemented in the software fuzzy-logic controller is shown.
In Fig. 5, the step responses of the system with SISO PI tuning and MFC TITO fuzzy controller are shown.
It is obvious that the system under the MFC TITO controller behaves better compared with that of SISO PI
tuning especially at the transients.
Another attribute of this design is that in the control system, the fuzzy-logic is implemented through
software in a common, inexpensive, 16-bit microcontroller, which does not have special abilities for fuzzy
control.

Conclusions

An intelligent fuzzy-logic controller to adjust the rotational speed of two DC motors of a solar-powered air-
conditioner has been described. The aim was to investigate whether fuzzy control can be used to control a
TITO process. The original idea was to design a fuzzy controller for the process without any modelling phase.
In a second stage, the interactions between each degree of freedom are considered, based on information avail-
able from the real system data. This information is introduced in the system controller via the fuzzy-control
rules employed for the coupling fuzzy controller.
A mixed controller was developed for controlling TITO systems. A solar air-conditioner with two D.C.
motors was designed and constructed. A TFC was initially designed for each level only to control a SISO sys-
tem. Secondly, a coupling fuzzy controller was introduced into the traditional fuzzy controller to improve the
control performances of the TITO solar air-conditioning system. The system has been experimentally demon-
strated and its performance was verified. The experimental results show that both the TFC and the MFC are
stable for controlling this MIMO air-conditioning system. In addition, it can be observed, from the experi-
mental step-input response, that the control performance of this mixed fuzzy-control strategy is improved.
Therefore, this mixed intelligent-control strategy could be employed to control MIMO systems and improve
the control performance.
A disadvantage of the above control scheme is the intensive computation required and the size of memory
needed, especially during the stages of fuzzification and defuzzification. On the other hand, by using the micro-
controller, the controller can be implemented by using only a small amount of components and easily
improved to be an adaptive fuzzy controller. Comparison of conventional and fuzzy-logic control techniques
shows that fuzzy-logic can reduce the effects of non-linearities in a DC motor, improve the performance of the
controller and also, FLC can be designed and implemented in a much easier and quicker way. Finally, FLC
algorithms can be easily transferred to different motor sizes and allow a strong tolerance for electric-motor
parameter oscillations.

Acknowledgement

The work described in this article was fully supported by the European Union in the context of the project:
‘‘Solar Air-Conditioning System Using Very Low Cost Variable Plastic Ejector with Hybrid Potential for Dif-
ferent Markets (SACPEH).

References

[1] Papadopoulos AM, Oxizidis S, Kyriakis N. Perspectives of solar cooling in view of the developments in the air-conditioning sector.
Renew Sustain Energy Rev 2003;7:419–38.
[2] Lamp P, Ziegler F. European research on solar-assisted air-conditioning. Int J Refrig 1998;21:89–99.
[3] Tsoutsos T, Anagnostou J, Pritchard C, Karagiorgas M, Agoris D. Solar cooling technologies in Greece-An economic viability
analysis. Appl Thermal Eng 2003;23:1427–39.
[4] Grossman G. Solar-powered systems for cooling, dehumidification and air-conditioning. Solar Energy 2002;72:53–62.
[5] Camacho EF, Berenguel M, Rubio FR. Advanced Control of Solar Plants. London: Springer; 1997.

Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001
 ARTICLE IN PRESS

14 J.N. Lygouras et al. / Applied Energy xxx (2007) xxx–xxx

[6] Camacho EF, Rubio FR, Hughes FM. Self-tuning control of a solar power plant with a distributed collector field. IEEE Contr Syst
Mag 1992;12:72–8.
[7] Nunez-Reyes A, Normey-Rico JE, Bordons C, Camacho EF. A Smith-predictive based MPC in a solar air-conditioning plant. J
Process Contr 2005;15:1–10.
[8] He D, Nelms RM. Fuzzy logic average current-mode control for DC/DC converters using an inexpensive 8-bit microcontroller. IEEE
Ind Appl Conf 2004;4:2615–22.
[9] Govind N. Fuzzy Logic Control with the Intel 80C196 Embedded Microcontroller. Proceedings of ESC-West Conference. Miller
Freeman; 1995, ISBN 0-87930-388-3, pp. 411–424.
[10] Peyravi H, Khoei A, Hadidi K. Design of an Analog CMOS Fuzzy-Logic Controller Chip. Fuzzy Sets and Systems 2002;132:245–60.
[11] Saoud SB, Gerstlauer A, Gajski D. Codesign methodology of real-time embedded controllers for electromechanical systems. Am J
Appl Sci 2005;9:1331–6.
[12] Tipsuwan Y, Chow MY. Fuzzy Logic Microcontroller Implementation for DC Motor Speed Control. IEEE IECON’99. San Jose, CA
1999. pp. 1271–6.
[13] Mrozek B, Mrozek Z. Modelling and Fuzzy Control of DC Drive. 14th European Simulation Multiconf. ESM 2000, May 23–26,
Ghent, pp. 186–90.
[14] Dimitriadis CM, Lygouras JN, Tsalides PG. An Intelligent Fuzzy-Logic Controller for One-Dimensional Motion Control. IEEE
Computer Society Press; 1994, pp. 133–9.
[15] Lygouras JN, Botsaris PN, Vourvoulakis J, Kodogiannis V. Fuzzy-logic controller implementation for a solar air-conditioning
system. Applied Energy, in press.
[16] Ruey-Jing L, Bai-Fu L. Design of a mixed fuzzy-controller for multiple-input multiple-output systems. Mechatronics
2005;15:1225–52.
[17] Chunshien L, Priemer R. Fuzzy control of unknown multiple-input multiple-output plants. Fuzzy Sets and Systems 1999;104:245–67.
[18] Ruey-Jing L, Shiuh-Jer H. A mixed fuzzy controller for MIMO systems. Fuzzy Sets and Systems 2001;120:73–93.
[19] Ying H. TITO Mamdani Fuzzy PI/PD Controllers as Non-linear, Variable Gain PI/PD controllers. Int J Fuzzy Syst 2000;2:191–6.
[20] Kazemian HB. The SOF-PID Controller for the Control of a MIMO Robot Arm. IEEE Trans Fuzzy Syst 2002;10:523–32.

Please cite this article in press as: Lygouras J.N. et al., Variable structure TITO fuzzy-logic controller ..., Appl Energ
(2007), doi:10.1016/j.apenergy.2007.07.001