ARTICLE IN PRESS

Control Engineering Practice 17 (2009) 11571173

Contents lists available at ScienceDirect

Control Engineering Practice

journal homepage: www.elsevier.com/locate/conengprac

An adaptive temperature control law for a solar furnace

B. Andrade da Costa , J.M. Lemos
INESC-ID/IST/TU Lisbon, R. Alves Redol 9, 1000-029 Lisboa, Portugal

a r t i c l e in fo

abstract

Article history:
Received 14 August 2008
Accepted 7 May 2009
Available online 5 June 2009

This paper describes the development of an adaptive control law based on the exact feedback
linearization and Lyapunov adaptation of the process dynamics applied to a solar furnace. The algorithm
resulting from these underlying design principles is approximated in order to relate it with an adaptive
PI controller with feedforward. The controller is tested on a 6 kW solar furnace model that represents a
plant installed at the Odeillo Processes Materials and Solar Energy Laboratory (Oriental Pyrenees in the
South of France). The adaptive features allow to tackle the problems posed by knowledge uncertainty
about furnace dynamics. It is concluded that the specications related to material testing are met.
& 2009 Elsevier Ltd. All rights reserved.

Keywords:
Solar furnace
Thermal stress tests
Nonlinear control
Exact linearization
Lyapunov method

1. Introduction
Nowadays there is a huge concern about the use of energy
sources that contribute to the climatic change problem. To
mitigate the problem, the exploitation of solar energy has been
considered in several ways, by converting it to thermal and
electric energy, using either thermal solar elds or photovoltaic
systems/panels. Solar energy is also used in other elds such
as the synthesis of high-temperature materials, and in material
testing (Fernandes et al., 2006), thermal stress (Fernandes,
Amaral, Rosa, & Shohoji, 2000), or to recreate the conditions of
the reentry of spacecrafts in Earths atmosphere. It is worth to
mention that the development of the Odeillo solar plant was
motivated by research on materials to be used in space rockets,
guided missiles and nuclear plants (Boyle, 1996), thereby avoiding
the problems of testing them with direct radioactive elements.
All those applications of solar energy systems pose interesting
problems from the automatic control point of view (Berenguel,
Camacho, Garcia-Martin, & Rubio, 1999; Lemos, 2006), such as the
presence of disturbances, non-linearities, and variable delays.
Solar furnaces concentrate solar energy in a limited area,
around the focus of a concentration mirror, or Fresnel lens, and
allow to attain high temperatures. Despite its interest, there is a
scarce number of references on the specic topic of solar furnace
control. A major exception is the work of Berenguel et al. (1999)
where modeling and control of a 20 kW furnace located at
Plataforma Solar de Almeria (southern Spain) is presented. The
work reported therein includes several types of PID with gain
 Corresponding author. Tel.: +351 213100259; fax: +351 218417499.

E-mail addresses: bac@comp.ist.utl.pt (B. Andrade da Costa), jlml@inesc-id.pt

(J.M. Lemos).
0967-0661/$ - see front matter & 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conengprac.2009.05.001

scheduling and a self-tuning controller. In Paradkar and Feliachi

(2002) a controller is proposed to compensate disturbances in a
solar furnace, and in Lacasa, Berenguel, and Yebra (2006) a
controller based in fuzzy logic is evaluated in a solar furnace for
copper sintering. In Garcia-Gabin, Zambrano, and Camacho (2009)
a sliding mode predictive controller is evaluated and applied to a
solar air conditioning plant, and in Kojima, Taniwaki, and Okiami
(2008) the problem of positioning a exible solar array is
addressed. In Costa et al. (2008a) experimental results are
described with a PI for temperature control in the same furnace
considered here.
In this paper the exact feedback linearization method together
with Lyapunov adaptation (Slotine & Li, 1991) is used to design an
adaptive controller for controlling the temperature of a sample in
a solar furnace (Costa et al., 2008b). The prototype is the 6 kW
solar furnace of Odeillo solar complex. In addition to the
algorithm yielded by the direct application of these techniques,
a modied version is also presented. This has the advantage of
having a structure that is comparable to an adaptive PI controller
with feedforward, thereby rendering the commissioning easier.
The contributions of this paper consist in the application of the
above-mentioned algorithms, in order to develop a new modied
structure, and its demonstration by simulation in a realistic
furnace model.
This paper is organized as follows. Section 2 describes the solar
furnace plant, in particular the thermal model subsystem. Section 3
describes the design of the adaptive controller using the exact
linearization method. Simulation results are shown and discussed.
Section 4 describes the modication of the adaptive controller and
results are presented. Section 5 presents the stability analysis of
the closed loop system with the modied adaptive controller.
Section 6 draws conclusions.

ARTICLE IN PRESS
1158

B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

2. Plant description

Direct Solar Radiation, (8h00m00s17h59m59s)

1200

2.1. Heliostat
The heliostat, Fig. 1, operates in closed-loop control and follows
the movement of the Sun with accuracy, in such a way that the
parabolic concentrator located at the top of the building, inside
the laboratory, Fig. 1, receives the Sun beam always along the
same direction. This allows the focus to be in the same place
during operation of the solar furnace. With the heliostat operating
in closed-loop control, it may be assumed that the tracking of the
Suns position is perfect and there is no need to consider the
effects of heliostats dynamics in the temperature of the sample.
The closed-loop mode is selected whenever the measured direct
Suns power is higher that 300 W=m2 . Below 300 W=m2 the
furnace does not operate.
The solar direct radiation is not constant but can exhibit some
changes due to air moisture, dust, clouds, and during the day and
season. Fig. 2 shows two data records taken in (2003/05/03) and
(2006/05/25). The quantity of energy reected by the heliostats
mirror (Fig. 1) depends on its area and also on the cleanliness of
the mirror, which will decrease with time due to dust deposition.
Depending on time and dust level, the reected power can vary
from 85% to 90% of the available Suns power.

Day 030503

1000

Irradiance [W/m2]

The plant to control is the 6 kW solar furnace of the Odeillo

Process Materials and Solar Energy Laboratory. It is made of four
subsystems: the heliostat system, the parabolic mirror and its ux
distribution, the shutter system and the temperature control
system.

800

Day 060526

600

400

200

0
0

0.5

1.5

2.5

time [s]

3.5

4.5
x 104

Fig. 2. Direct solar radiation evolution with time. Examples show disturbances
due to moisture in air, dust and clouds.

2.2. Parabolic concentrator

The parabolic concentrator is built with small hexagonal
mirrors that direct the solar energy to the location where the
sample to be tested is placed. The maximum available power at
the focus is 6 kW. The position of the sample can be adjusted by
manually commanding the position of the supporting arm, Fig. 3,
that can be moved in the NorthSouth, WestEast and updown
directions using the operating console.
Assuming that the shutter is not present, the ux on the focus
is approximated by a Gaussian function. The focus has a diameter
of 6 cm, and receives 95% of the power concentrated by the
parabolic mirror. The ux inside the area of the focus is not
uniform but, for practical purposes, it may be considered that a
circle with radius of 2.5 cm centered at the focus receives a

Fig. 1. Schematic of the Odeillo 6 kW solar furnace.

Fig. 3. Shutter of the 6 kW solar furnace (top). Test tube in the focus with a SiC
sample (center).

uniform ux. The size of the focus can be changed by moving the
supporting arm in the updown direction.
2.3. Shutter
The shutter (Fig. 3) is made of 10 moveable aluminum blades
with a thickness of 2.0 mm. It has a circular like shape with a
radius of 0.5 m. The blades are moved by a brushless motor (from
the Parvex manufacturer) using a gear mechanism. The subsystem
motor-blades is controlled with a Digivex controller, which
receives positional commands (reference signal) to position the
blades. The commands can be sent using the operating console or
by a personal computer through the data acquisition system
InstruNet.
The shutter of the solar furnace is able to quickly change the
incident power on the sample, an important factor for the purpose
of temperature control. The physical aspect of the shutter,
location, size and material were selected in order to yield a fast
time response. In the shutter subsystem there are two aspects to
be considered: the static function sfs : that describes the steadystate relation between the power available before the shutter and
the power available at the focus, and the dynamics of the shutter.
The static function sfs : depends on y, the angle of the shutter, and
y0 the minimal angle value below which there is no power at the

ARTICLE IN PRESS
B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

focus. As shown in Fig. 3, the blades are not in the horizontal

position but the sample has no Sun power. The value of y0 is 251.
Using a geometric analysis, the static function is described by
cosy0 us t90  y0 =100
sfs us t 1 
cosy0

(1)

Table 2
Specic heat of SiC as a function of temperature.
T (K)
C p T

with 0ous t  100. The shutter operates in closed-loop control

and is able to move the blades to its target position in 0.5 s, with a
5% overshoot and 0.1 s of rise time. This dynamics is much faster
than the temperature dynamics. For digital implementations of
the temperature controller, a sampling time of 0.5 s is used (Costa
et al., 2008a); this means that the digital temperature controller
can be seen as having a continuous-time behavior and the shutter
can be described by its static function sfs : only.

The temperature model of the sample is developed based on an

energy balance, such as the one made in Berenguel et al. (1999).
The following model is used to describe the temperature behavior
of the sample T s t (in Kelvin)
dT s t
a1 T 4s t  T 4e t  a2 T s t  T e t a3 Gs tsfs us t
dt

(2)

where T e t (in K) is the temperature of the environment that

contributes to losses by radiation and convection. The parameters
a1 , a2 and a3 are temperature dependent and are dened by the
following equations:

a1

T s sAs
C p T s m

a2

hconv T s ; T e As
;
C p T s m

a3

as As g f
C p T s m

(3)

The parameters in (3) are described in Table 1, where a sample of

SiC material is considered (C
- engel, 2003).
The SiC emissivity in the temperature range of 3001500 K is
described by

T s 0:87  2:222  105 T s  600

(4)

The SiC specic heat is described in Table 2.

The convection factor is
hconv T s ; T e 1:32T s  T e =Lc 0:25

(5)

The samples solar absorption factor as is modeled by

as 0:09 T s  867:5  2:2086  104

8
0:055 if as  0:055
>
>

as T s

<

as

>
>
: 0:9

if 0:055oas  0:9

(6)

if as 40:9

Table 1
Furnace and sample parameters values.
Parameter: description

Value

r kg m3 : Density of the material

3:10  103
Table 2

1

C p Jkg K1 : Material specic heat

m (kg): Mass of the sample

: Emissivity of the material

s Wm2 K4 : StefanBoltzmann const.
As m2 : Samples radiation incident area
Lc (m): Characteristic length

3:29  103
Eq. (4)
5:67  108
7:068  104
7:50  103
Eq. (5)

hconv Wm2 K1 : Convection factor

as : Samples solar absorption factor
g f : Furnace gain

Eq. (6)
1713.6

Gs W=m2 : Max. solar ux

1:0  103

The sample is assumed to be of SiC.

300
675

1400

400
880

Model Output

600
1050

800
1135

1000
1195

1200
1000
Temp. [K]

2.4. Temperature model of the sample

1159

800

Measured Temp.

600
400
200
0

500

1000
1500
Time [seg]

2000

2500

Fig. 4. Comparing the temperature model output with the measured temperature
in a SiC test.

and it was selected after the temperature model validation with

experimental results.
The power absorption of the sample is modeled by
as As g f Gs tsfs us t. The factor g f describes the concentration
power of the solar furnace and depends on the heliostat, parabolic
mirror and position of the supporting arm. The function sfs :
describes the nonlinear contribution of the shutter to the incident
power on the sample. The signal us t is in the range [0, 100] and
represents the command signal of the shutter.
The temperature model was validated1 using temperature
experimental data obtained from a SiC sample test with a PI
controller (Costa et al., 2008b). Fig. 3 shows the apparatus, where
the tip of the thermocouple is in contact in the lower side of the
SiC sample. The experimental command signal of the shutter and
the solar ux was supplied to the temperature model and the
model output was compared with the experimental temperature
measurements. The results are shown in Figs. 4 and 5; note that
there is a large mismatch at lower temperatures, this can be
explained in part by the type of thermocouple used, a B
thermocouple, that does not give reliable readings at lower
temperatures. For higher temperatures, between 800 and 1300 K,
the model provides a good description of the experimental results.
This model will be used in computer simulation tests despite it
does not provide a good description in the 300800 K range.
The parameters a1 , a2 and a3 , change with temperature T s
according to Table 3, where T e 300 K.
The output from the simulated model using a sequence of
steps in the shutter command is shown in Fig. 6. It can be
concluded that the temperature model has different local
dynamics which are dependent on the temperature level. In
similar cases the usual procedure to tackle the controller design
problem is to obtain linearized models (Berenguel et al., 1999) and

1
Note: using the experimental data, the temperature dependence of as was
modeled by a linear equation.

ARTICLE IN PRESS
1160

B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

100

us(t) [%]

80
60
40
20
0
0

500

1000
1500
Time [seg]

2000

2500

500

1000
1500
Time [seg]

2000

2500

Gs(t) [Wm1]

900
800
700
600
500
400

Fig. 5. Experiment signals: command of the shutter, us t and solar ux Gs t used in the temperature model evaluation.

where T R t represents the temperature reference and T s t

represents the temperature of the sample. Computing the time
derivative2 of the tracking error yields

Table 3
Process model parameters for SiC as a function of temperature.
T (K)

a1 T

a2 T

a3 T

300

1:584  1011

3:000  101

400

1:212  1011

3:466  103

2:303  102

600

1:010  1011

3:823  103

1:930  102

800

9:299  1012

4:019  103

2:438  102

12

3

2

1000

8:786  10

4:152  10

3:678  10

e_ s t T_ R t a1 T 4s t  T 4e t a2 T s t  T e t  a3 Gs tsfs us t
(8)
In order to use the exact linearization method, a virtual
manipulated variable vt is dened by the change of variable
vt T_ R t a^ 1 T 4s t  T 4e t a^ 2 T s t  T e t  a^ 3 Gs tsfs us t
(9)

use them to design local controllers, such as a bank of PID

controllers with a switching mechanism to select the adequate
controller. In the present case an adaptive control technique based
on a feedback exact linearization is used in the controller design.

Under perfect knowledge of the parameters (a^ i , 8i 2 f1; 2; 3g), the

transformed model becomes an integrator and may be stabilized
with the control law vt K L es t, where K L 40 is a tuning
parameter. This corresponds to the actual control law used to
manipulate the shutter:

3. The adaptive controller

us t s1
fs wt

The design of the adaptive controller comprises two main

steps:

wt

 rst, a feedback linearization control law is obtained, based on

the physical model Eq. (2);

 then, the parameters entering linearly on the model are

estimated using a Lyapunov method.
The motivation to explore this approach is to use the information
about the process, namely the structure of the temperature model,
the static characteristic of the shutter and the direct measurements of the Suns radiation to compensate the radiation
disturbances.
To develop the adaptive controller, Eq. (2) is used in conjunction with the tracking error es t dened by
D

es t T R t  T s t

(11)

D
where estimates a^ i of ai are used, with ai a^ i a~ i , 8i 2 f1; 2; 3g.
However, the presence of parameter errors a~ 1 , a~ 2 and a~ 3 may
generate an unstable closed loop. To avoid that problem, the
estimates a^ 1 , a^ 2 , a^ 3 ought to be adapted in such a way that es t
converges to zero. To tackle the problem, a candidate Lyapunov
function Ves t; a~ 1 t; a~ 2 t; a~ 3 t that jointly encompasses control
and estimated states is chosen
"
#
3  
X
1 2
1 2
a~ i t
(12)
es t
Ves t; a~ 1 t; a~ 2 t; a~ 3 t
gi
2
i1

with adjustable parameters gi 40.

2

(7)

T_ R t K L es t a^ 1 T 4s t  T 4e t a^ 2 T s t  T e t

a^ 3 Gs t
a^ 3 Gs t

(10)

_
by .

To simplify the notation the derivative operator d=dt will be represented

ARTICLE IN PRESS
B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

1161

Temperature
1400

Ts(t) [K]

1200
1000
800
600
400
200
0

200

400

600
time [s]

800

1000

1200

800

1000

1200

Shutter position
100

us (t) [%]

80
60
40
20
0
0

200

400

600
time [s]

Fig. 6. Steps responses of the temperature model with Gs t 600 W=m2 .

Computing the time derivative of and using Eq. (8) yields

dVes t; a~ i t
 K L e2s t
dt



1
a~ 1 T 4s t  T 4e tes t a_~ 1 t
g1


1
a~ 2 T s t  T e tes t a_~ 2 t
g2


1 _
a~ 3 Gs twtes t a~ 3 t

g3

(13)

Equating to zero the terms that are multiplying a~ 1 , a~ 2 and a~ 3 , and

considering that a~^ i a_^ i for i 2 f1; 2; 3g yields the following
adaptation laws:

a_^ 1 t g1 T 4s t  T 4e tes t
a_^ 2 t g2 T s t  T e tes t

(15)

a_^ 3 t g3 Gs twtes t

(16)

algorithm, seven parameters must be specied. The a^ 1 t 0 , a^ 2 t 0 ,

a^ 3 t0 were selected to be near the process parameter values for
T s 800 K. All the three computer simulations use a^ 1 t 0 9:2
1012 , a^ 2 t 0 4:0  103 , a^ 3 t 0 2:3  102 and the K L 10,
the difference is in the adaptation gains.
One problem with adaptive controllers is that they usually
have knob parameters that must be selected and they affect the
change rate of estimates, which is the case with g1 , g2 and g3 . To
guide the selection of these parameters, the integral form of the
parameter adaptation laws is used with the following assumptions: (i) a 50% error is assumed for the initial estimates; (ii) the
temperatures T s t and T e t are assumed to be near 800 and
300 K, respectively, corresponding to a possible nominal working
point. It is then possible to obtain

(14)

_ s t; ai  0. A standard argument based on

that results in Ve
LaSalles invariance principle allows to conclude that es t
approaches zero and a~ i ; 8i 2 f1; 2; 3g are bounded. The parameters
a1 , a2 and a3 are assumed to be constant in this derivation. In
practice, this means that the parameters should vary slowly when
compared with T s t.

g1

T 4s

ja^ 1 t  a^ 1 t 0 j
1:25  1023
 R




 t
4 R t
 T e  t0 es t dt
 t0 es t dt

(17)

g2

ja^ 2 t  a^ 2 t 0 j
4:0  106
 R

R
  t

 t
T s  T e  t0 es t dt  t0 es t dt

(18)

g3

ja^ 3 t  a^ 3 t 0 j
1:5  105
 R

R
  t

 t
Gs  t0 wtes t dt  t0 wtes t dt

(19)

3.1. Computer simulation results

where 0  w:  1:0 and Gs is assumed to be 800 Wm2 . These

results are now used to guide the selection of the g1 , g2 and g3
which are very small!

The performance of the adaptive controller is evaluated by

computer simulations. In the simulations, the term T_ R t of
Eq. (10) is set to zero, this means it will act as an error source
when the reference is not constant. Three simulation results are
shown with different adaptation gains. To set up the adaptive

3.1.1. Simulation I
In this simulation the adaptation gains are g1 1:0  1024 ,
g2 1:0  107 , g3 2:0  105 .
The results are shown in Figs. 7 and 8. At the beginning of the
simulation the controller was able to track the reference signal.

ARTICLE IN PRESS
1162

B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

RefT(t) and Ts(t) [K]

Temperature
1200
Ref (t)
T

1000
800
600
400
200
0

50

100

150

200

250
time [s]

300

350

400

450

500

50

100

150

200

250
time [s]

300

350

400

450

500

350

400

450

500

us (t) [%]

100

50

Simulation of Solar Power variability

Gs(t) [W/m2]

1000
800
600
400
0

50

100

150

200

250
time [s]

300

Fig. 7. Simulation I. Process output and reference signal, control signal and the simulated Gs t. Closed-loop control.

But at time near 50 s the control signal becomes saturated at its

maximum value. The tracking error es t being positive starts to
increase and according to Eq. (16), the estimate of a3 starts to
decrease and becomes negative. The wrong signal value of a^ 3 t
has the consequence that the control signals us t and wt are set
to zero3 due to the constraints imposed by the operating limits of
the shutter, and the adaptation of a^ 3 t is blocked. After this event
the control signal is blocked at zero and the controller is not able
to track the reference signal.

3.1.2. Simulation II
The adaptation gains are now selected to g1 1:0  1024 ,
g2 1:0  107 , g3 1:0  1010 , note that g3 was multiplied by a
factor of 5:0  106 . The aim is to avoid the problem of having a
negative value on a^ 3 t. The simulation results are shown in
Figs. 911. Fig. 10 is a zoom to show the results of the temperature
and shutter aperture. Note that the shutter controller is able to
move the blades to the target position in 0.5 s.
The adaptive controller was able to track the temperature
reference, but the estimates a^ 2 t, a^ 3 t are affected by the control
signal saturation, that can cause stability problems. Note that the
a^ 2 t, a^ 3 t recover to the values prior to the control signal
saturation. The parameter estimate a^ 3 t can be considered
constant. However, the problem reported in Section 3.1.1 was
avoided by selecting a very small adaptation gain value for g3,
which is by freezing a^ 3 t to its initial value.
3

Note that wt sfs us t.

3.1.3. Simulation III

The adaptation gains are now selected to g1 1:0  1022 ,
g2 1:0  107 , g3 1:0  1010 . The aim is to test higher
adaptation gain values for g1 and g2 . The simulation results are
shown in Figs. 12 and 13. The effect of the saturation signal causes
a huge tracking error es t, couple with higher adaptation gains
cause the divergence of a^ 1 t and a^ 2 t. In practical the adaptation
laws must include a mechanism to overcome the control signal
saturation.
The results show that the adaptive controller has a good
performance, but adaptation gains must be very small to achieve
the good results.
3.1.4. Comments
The simulation results presented in Sections 3.1.13.1.3 show
that the controller has good performance but that depends on the
correct selection of the adaptation gains g1 , g2 and g3 . There are
also the following points to be mentioned:

 The saturated control signal causes problems in the adaptation




of a^ 1 t, a^ 2 t, a^ 3 t, which can cause a misbehavior of the

controller, see Section 3.1.1 for the behavior of a^ 3 t.
The adaptation parameters g1 , g2 , g3 values are very small and
it was difcult to select them.
It is difcult to incorporate an anti-windup mechanism in the
control law dened by Eqs. (10) and (11).

To use the adaptive controller in practice the above points must

be solved.

ARTICLE IN PRESS
B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

1163

Controller gains Estimates of 1, 2, 3

x 108

1 (t)

0.5

0
0

50

100

150

200

250

300

350

400

450

500

300

350

400

450

500

300

350

400

450

500

time [s]
4
2 (t)

3
2
1
0
0

50

100

150

200

x 103

250
time [s]

3 (t)

10
5
0
5
0

50

100

150

200

250
time [s]

Fig. 8. Simulation I. Evolution of a^ 1 t, a^ 2 t, and a^ 3 t. Closed-loop control.

4. Modication of the adaptive controller for practical use

With these assumptions Eqs. (20)(22) are modied to

Due to the difculties described in Section 3.1.4 an analysis of

the adaptation equations of a^ 1 t, a^ 2 t and a^ 3 t is done and a
modication is proposed. The aim is to obtain an implementation
of the adaptive controller suitable to be used in practice.

From Eqs. (14)(16) the adaptation equations of a^ i t are

Z t
a^ 1 t a^ 1 t0 g1 T 4s t  T 4e tes t dt
(20)

a^ 2 t a^ 2 t0 g2
a^ 3 t a^ 3 t0  g3

a^ 2 t a^ 2 t 0 g2

t0
Z t

T 4R t  T 4e tes t dt

(23)

T R t  T e tes t dt

(24)

t0

a^ 3 t a^ 3 t 0

4.1. The modied adaptation law of the parameter estimates

a^ 1 t a^ 1 t 0 g1

t0
t

T s t  T e tes t dt

(21)

Gs twtes t dt

(22)

t0
t
t0

To prevent a^ 3 t to assume a negative or null value this variable

will be kept constant and equal to a^ 3 t 0 40. Note that a^ 3 is given
by Eq. (3) and a value can be computed for a^ 3 t 0 from the size and
the properties of the material to be tested.
Additionally the following assumptions are made:

 The environment temperature T e t can be considered constant

and is smaller than T R t and T s t.

(25)

R
_ t dt f tgt
Using now the relationship given by
f tg
R_
f tgt dt, Eqs. (23) and (24) are written as

a^ 1 t a^ 1 t 0 g1 T 4R t  T 4e t
 4g1 T_ R

t
t0

T 3R t

Z t

t
t0

Z t

es t dt
t0

(26)

es t dt
t0

es t1 dt1 dt

(27)

t0

Considering now the terms with the slope of the temperature

reference signal T_ R , in Eqs. (26) and (27) and assuming that they
can be discarded when compared with the other terms, then the
following parameter adaptation laws are obtained:

 The kes tk is assumed to be small such that T s t T R t, note

that es t T R t  T s t. This assumption is used to approximate the adaptation laws Eqs. (20) and (21) in the case that
the output of the process is near the reference signal and the
closed-loop system is stable.
The reference signal T R t is assumed to be a constant signal or
a ramp signal with a small slope T_ R . This will be details next.

es t1 dt1 dt

t0

a^ 2 t a^ 2 t 0 g2 T R t  T e t
 g2 T_ R

a^ 1 t a^ 1 t 0 g1 T 4R t  T 4e t
a^ 2 t a^ 2 t 0 g2 T R t  T e t
a^ 3 t a^ 3 t 0

t0
t

es t dt

(28)

es t dt

(29)

t0

(30)

ARTICLE IN PRESS
1164

B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

RefT(t) and Ts(t) [K]

Temperature
1200
1000
800
600
400
200
0

50

100

150

200

250

300

350

400

450

500

300

350

400

450

500

400

450

500

time [s]

us (t) [%]

100

50

0
0

50

100

150

200

250
time [s]

Simulation of Solar Power variability

Gs(t) [W/m2]

1000
800
600
400
0

50

100

150

200

250
time [s]

300

350

Fig. 9. Simulation II. Process output and reference signal, control signal and the simulated Gs t. Closed-loop control.

4.2. The modied control law

The modied control law is obtained by substituting Eqs.
(28)(30) in Eq. (11). Using algebraic manipulation the modied
control law assumes the following structure:
us t s1
fs wt


Z
wt U o t K P t es t K I t

(31)

es t dt

(32)

t0

U o t
D

K P t
D

K I t

a^ 1 t 0 T 4s t  T 4e t a^ 2 t0 T s t  T e t
a^ 3 t0 Gs t

(33)

KL

a^ 3 t 0 Gs t
g1 T 4R t  T 4e tT 4s t  T 4e t
KL

(34)

present case the estimates a^ 1 t 0 , a^ 2 t 0 and a^ 3 t 0 can be

computed from physical data using Eq. (3), and only g1 and g2
and K L are the controller knob parameters. In Section 5.2 a
redenition of g1 and g2 is presented and it is shown that only one
parameter, g, is needed, and Section 5.6 demonstrates that the
parameter g can be selected as g  K L =22 , with the equality
corresponding to have a double real pole for the closed-loop
assuming a perfect knowledge of the process parameters.
This simplies the parameter selection and improves the
numerical representation of variables and constants in a digital
computer with limited resources. Note that the parameters in
Eq. (33) can also be redened to improve the numeric representation, by using a constant based on the temperature reference
mean value.

g2 T R t  T e tT s t  T e t
KL
(35)

This new control law may be interpreted as being a PI controller

with a feed-forward term U o t, with adaptive gains, K P t, K I t.
The stability analysis is addressed in Section 5. Note that the
control law depends only on the initial estimates a^ 1 t 0 , a^ 2 t 0 ,
a^ 3 t0 , the adaptation gains g1 and g2 and K L , and there is only one
integrator, which can be modied to have an anti-windup
mechanism. Comparing this controller with a standard PI
controller, it seems that the proposed controller has too many
parameters to be tuned. However, it must be emphasized that Eqs.
(33)(35) show the explicit relationship between the process
parameters and the controller parameters, in practice to design a
PID controller, the process parameters must be known. In the

4.3. Simulation results with the modied controller

The modied adaptive controller was simulated with the
following parameter values, a^ 1 t 0 9:2  1012 , a^ 2 t 0 4:0
103 , a^ 3 t 0 2:4  102 , K L 10, and the adaptation gains
g1 1:0  1024 , g2 1:0  107 . These controller parameter
values are the same as in the simulation described in Section
3.1.2. The aim is to compare the results of the modied adaptive
controller with the unmodied adaptive controller in the same
test conditions.
The results are shown in Figs. 14 and 15. The modied adaptive
controller has a good performance, the results are similar as the
results described in Section 3.1.2 and the controller has a simpler
structure that can be easily interpreted.

ARTICLE IN PRESS
B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

1165

RefT(t) and Ts(t) [K]

Temperature
1200
1000
800
600
200

220

240

260
time [s]

280

300

320

220

240

260
time [s]

280

300

320

300

320

us (t) [%]

100

50

0
200

Simulation of Solar Power variability

Gs(t) [W/m2]

900
800
700
600
500
200

220

240

260
time [s]

280

Fig. 10. Simulation II. Zoom of Fig. 9 to show the temperature and shutter aperture. Closed-loop control.

Note that in both simulations the controllers do not have antiwindup reset mechanisms. With the new controller it is easy to
tackle the control saturation by including an anti-windup reset
mechanism in the integrator. This problem was not so easily
resolved with the rst adaptive controller.

with4
D
Les t
g1 T 4R t  T 4e tT 4s t  T 4e t

g2 T R tt  T e tT s t  T e t

(37)

Eq. (36) includes the process dynamics and the modied control
law. The aim is to nd the equilibrium points of system Eq. (36)
and to demonstrate their stability.

5. Stability analysis of the modied adaptive controller

This section presents the stability analysis of the closed-loop
system using the modied adaptive controller. The stability
analysis is based on Lyapunovs theory with the assumptions
described in Section 4.1. It is also assumed that the control signal
is not saturated, which means the stability analysis is local. To
address the saturation of the control signal an anti-windup
mechanism must be incorporated and the stability analysis will be
more complex but can be addressed using techniques described in
Glattfelder and Schaufelberger (2003).
Using Eq. (8) that describes the dynamics of the tracking error
as a function of the control signal, and the equations that dene
the modied control law, Eqs. (31)(35), then the tracking error
dynamics is described by
des t
a3
a3
K e t a1  a^ 1 t 0
T 4 t  T 4e t

dt
a^ 3 t0 L s
a^ 3 t 0 s


Z t
a3
a3
a2  a^ 2 t 0
Les t
es t dt
T s t  T e 
a^ 3 t0
a^ 3 t0
t0
(36)

5.1. Equilibrium points

D Rt
D
Dening two state variables, x1 t t0 es t dt and x2 t es t,
Eq. (36) can be written in a state space representation, which
yields an equilibrium point xe xe1 ; xe2 given by

8
<x
:

e1

DTe
GTe

T R t  T e t

(38)

xe2 0

with
D
DTe
a1

a^ 3 t0 a^ 1 t0 T 4R t  T 4e t
a^ 3 t0 a^ 2 t0
a2


a3
a1
a3
a2
T R t  T e t


(39)

4
In order to keep equations as simple as possible the variable T s t is not
substituted by T s t T R t  es t.

ARTICLE IN PRESS
1166

B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

Controller gains Estimates of 1, 2, 3

x 1011
5

1 (t)

0
5
10
0

50

100

150

200

250
time [s]

300

350

400

450

500

50

100

150

200

250
time [s]

300

350

400

450

500

50

100

150

200

250
time [s]

300

350

400

450

500

x 103
5

2 (t)

0
5
10
0

x 103

3 (t)

10

9.995

9.99
0

Fig. 11. Simulation II. Evolution of a^ 1 t, a^ 2 t, and a^ 3 t. Closed-loop control.

and5

parameters the following denition is proposed to g1 and g2 :

GTe g1 T 4R t  T 4e t2 g2 T R t  T e t2

(40)

It is worth to mention that the equilibrium point corresponds to

a null tracking error but the value at the integrator xe1 can be
positive, null or negative, depending on the process parameter
initial estimates. If, however, bounds are known such that
aimin  ai  aimax , for i 2 f1; 2; 3g then xe1 can be always positive if
a^ 1 t0 a1min , a^ 2 t0 a2min and a^ 3 t0 a3max .

g1 g12T 4R t  T 4e t2
g2 g12T R t  T e t2

(42)
(43)

with g40 being the new positive adjustable parameter which is

common to both g1 and g2 . With this denition of g1 and g2 ,
Eq. (40) is simplied to
D

GTe g

5.2. Redenition of g1 and g2

Looking at Eqs. (36) and (37) the parameters g1 and g2 are only
used as a weight on the integration of es t. If the ideal case is
^ 0 a1 ,
considered with the initial estimates selected as a1 t
^
^
a2 t0 a2 and a3 t0 a3 then Eq. (36) is simplied to
Z t
des t
es t dt
(41)
K L es t  Les t
dt
t0
The linearization of Eq. (41) at the equilibrium point xe 0; 0
generates a second order ordinary equation with two poles that
can be selected by choosing K L and by a combination of the two
parameters, g1 and g2 . To reduce the number of adjustable
5
Note
that
T 2e tT R t T e t.

in

Eq.

(39)

T 4R t  T 4e t=T R t  T e t T 2R t

(44)

The proposed control algorithm is now written as

us t s1
fs wt


Z
wt U o t K P t es t K I t

es t dt

(45)
(46)

t0

U o t

a^ 1 t0 T 4s t  T 4e t a^ 2 t 0 T s t  T e t
a^ 3 t0 Gs t
KL

K P t
D

K I t

a^ 3 t0 Gs t
g T 4s t  T 4e t

2K L

T 4R t  T 4e t

(47)
(48)

T s t  T e t
T R t  T e t

Note that if T s t T R t then K I t g=2K L .

(49)

ARTICLE IN PRESS
B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

1167

RefT(t) and Ts(t) [K]

Temperature
1200
1000
800
600
400
200
0

50

100

150

200

250
time [s]

300

350

400

450

500

50

100

150

200

250
time [s]

300

350

400

450

500

350

400

450

500

us (t) [%]

100

50

Simulation of Solar Power variability

Gs(t) [W/m2]

1000
900
800
700
600
0

50

100

150

200

250
time [s]

300

Fig. 12. Simulation III. Process output and reference signal, control signal and Gs t. Closed-loop control.

5.3. Stability region

where8

The stability region is investigated using the following

Lyapunovs function candidate6:

Dx2 t a1

St

1
1 a^ t
gx1 t  xe1 2 3 0 x2 t  xe2 2
2
2 a3

(50)

a^ 3 t0 a^ 1 t0 2
T R t T 2s t

a3
a1


a^ 3 t0 a^ 2 t0

T R t T s t a2
a3
a2

(53)

and9
Computing the time derivative of Eq. (50), and using Eq. (38), it is
possible to obtain



_  x2 t K L a1 a^ 3 t 0  a^ 1 t 0 T 2 t T 2 tT 2 t T 2 t
St
2
R
s
R
s
a3
a1


a^ 3 t0 a^ 2 t0
g
 x1 t

a2
a3
a2
2T R t  T e t
"
#)
2
2
T R t T s tT R t T s t
1
(51)

T 2R t T 2e tT R t T e t
that can be rewritten as7

_ x2 t K L Dx2 t 
St
2

g
T R t  T e t

Yx2 tx1 t

"
#
1 T 2R t T 2s tT R t T s t
1
Yx2 t
2 T 2R t T 2e tT R t T e t
D

(54)

Eq. (52) can now be used to nd the region in the x1 t; x2 t

_  0. This can be obtained if the expression in the
plane where St
right side of Eq. (52) delimited by fg is positive.10 Imposing this
condition the stability zone is dened by
8
>
< x toK L T R t  T e t Dx2 t T R t  T e t
1
g Yx2 t
g
Yx2 t
>
: 8x2 t 2 T R t  T max ; T R t  T e t

(55)

(52)

See footnote 7.
See footnote 7.
10
Note that 0  us t  umax which implies that T e t  T s t  T max . The
reference signal T R t must be selected according to T e toT R t  T max . From a
practical point of view T R t, T e t, T s t have positive values because these signals
are measured in Kelvin.
9

6
Note that in this analysis T R t and T e t are assumed to be constant possibly
evolving by steps but with T R t4T e t.
7
Note that T s t T R t  x2 t.

ARTICLE IN PRESS
1168

B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

Controller gains Estimates of 1, 2, 3

x 108
1
1 (t)

0.5
0
0.5
1
0

50

100

150

200

250
time [s]

300

350

400

450

500

50

100

150

200

250
time [s]

300

350

400

450

500

50

100

150

200

250
time [s]

300

350

400

450

500

2 (t)

0.5
0
0.5
1

x 103

3 (t)

10

9.995

9.99
0

Fig. 13. Simulation III. Evolution of a^ 1 t, a^ 2 t, and a^ 3 t. Closed-loop control.

This means that the line formed by the points xb1 t; xb2 t such
that
8
>
< x t K L T R t  T e t Dx2 t T R t  T e t
b1
g Yx2 t
g
Yx2 t
(56)
>
: 8x t 2 T R t  T max ; T R t  T e t
b2
dene the boundary line in the x1 t; x2 t plane that delimits the
stability region. The closed-loop system to be stable must have the
equilibrium point xe1 ; xe2 inside the region dened by Eq. (55).
Two cases are now analyzed, the rst one considers a perfect
knowledge of the process parameter values a1 , a2 and a3 and the
other case considers the estimates of the process parameters.

5.4. Stability of the equilibrium point xe , with perfect knowledge of

the process parameter values
With perfect knowledge of the values of a1 , a2 and a3 the
equilibrium point, Eq. (38), is given by xe 0; 0 and the
condition in (55) is simplied to
8
>
< x1 toK L T R t  T e t
g Yx2 t
(57)
>
: 8x2 t 2 T R t  T max ; T R t  T e t
The function Yx2 t (Eq. (54)) depends only on x2 t. It does not
depend on the controller parameters and on the process
estimates. The function K L =gT R t  T e t=Yx2 t is a continuous and positive function of x2 t for 8x2 t 2 T R t  T max ; T R t 
T e t and it does not have singular points. This is illustrated in
Table 4 where the function T R t  T e t=Yx2 t is evaluated in
three points. For simplicity at x2 t 0 it is assumed that

T R tbT e t, and for x2 t T R t  T max it is assumed that

T max 10T R t.
The controller gains K L 40 and g40 have the effect of
modifying the size of the stability region. This is also illustrated
in Fig. 16. The main conclusion is that the equilibrium point xe
0; 0 is inside the stability region and therefore it is stable.

5.5. Stability of the equilibrium point xe , when using process

parameter estimates
When the initial parameter estimates a^1 t 0 , a^2 t 0 , a^3 t 0 are
different from a1 , a2 , a3 then the equilibrium point xe xe1 ; 0,
with xe1 different from null, is not located at the origin of the
x1 t; x2 t plane. For this case the equations that dene the
stability region, (55), can be rewritten in a new form which
include the coordinate xe1 as shown in (58). To do that the right
side of (55), the part containing the process parameter estimates,
is multiplied by xe1 x1
e1 . Using Eqs. (38)(40) yields
8
1
>
< x toK L T R t  T e t x Dx2 t
1
e1
g Yx2 t
DTe Yx2 t
(58)
>
: 8x2 t 2 T R t  T max ; T R t  T e t
Note that if Dx2 t=DTe 1=Yx2 t41, 8x2 t 2 T R t 
T max ; T R t  T e t then the equilibrium point in included in the
stability region. If, however, Dx2 t=DTe 1=Yx2 t  1 then the
term K L =gT R t  T e t=Yx2 t (in particular K L =g) must be
adjusted to increase the stability region in order to contain the
equilibrium point xe .
The several terms and factors of Eq. (58) are analyzed to nd
their contribution to the size of the stability region.

ARTICLE IN PRESS
B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

1169

RefT(t) and Ts(t) [K]

Temperature
1200
1000
800
600
400
200
0

50

100

150

200

250
time [s]

300

350

400

450

500

50

100

150

200

250
time [s]

300

350

400

450

500

350

400

450

500

us (t) [%]

100

50

Simulation of Solar Power variability

Gs(t) [W/m2]

1000
800
600
400
0

50

100

150

200

250
time [s]

300

Fig. 14. Simulation of the modied adaptive controller. Process output and reference signal, control signal and Gs t. Closed-loop control.

Error Integral

Table 4
Evaluation of T R t  T e t=Yx2 t.

100

x2 t

T R t  T e t=Yx2 t

T R t  T e t
0

T R t  T e t
2
5T R t  T e t if T R bT e t

T R t  T max

1
556T R t

50

es(t)dt

 T e t if T max 10T R t

50

5.5.1. Assuming that a1 a^ 3 t 0 =a3  a^ 1 t 0 =a1 is dominant

If a1 a^ 3 t 0 =a3  a^ 1 t 0 =a1 is dominant, then Dx2 t=DTe is
simplied to

100

Dx2 t T 2R t T 2s tT R t T s t
2
DTe
T R t T 2e tT R t T e t

150
200
0

50

100

150

200

250

300

350

400

450

500

time [s]
Fig. 15. Simulation of the modied adaptive controller. Evolution of the
Closed-loop control.

es t dt.

In this analysis the relative magnitude of a1 a^ 3 t 0 =a3 

a^ 1 t0 =a1 and a2 a^ 3 t0 =a3  a^ 2 t 0 =a2 and their signs are
considered. Without loss of generality it is assumed that xe
xe1 ; xe2 is located at the right side of the x1 t; x2 t plane.
The term K L =gT R t  T e t=Yx2 t was analyzed in Section 5.4. It can be used to enlarge the stability region.

(59)

Evaluating now Dx2 t=DTe 1=Yx2 t it can be concluded, as

shown in Table 5, that it is a continuous function always positive
and have values greater of equal to 1. More at x2 t 0 the
function D0=DTe 1=Y041. The conclusion is that the
equilibrium point xe is included in the stability region for K L 40
and g40.
5.5.2. Assuming that a2 a^ 3 t 0 =a3  a^ 2 t 0 =a2 is dominant
If a2 a^ 3 t 0 =a3  a^ 2 t 0 =a2 is dominant, then Dx2 t=DTe is
simplied to

Dx2 t
1
DTe

(60)

ARTICLE IN PRESS
1170

B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

x2(t)=TR(t)-Te(t)

5.6. Selection of the controller parameters K L and g

x2(t)
P1

Boundary Lines

Stability Region
P2

x1(t)

"

dxt

KL(TR(t)-TR(t))
 (x2(t))

Equilibrium Point

The stability analysis presented in Sections 5.4 and 5.5 enables

the denition of conditions on K L =g to have a stable system, but
does not dene a strategy on how to select K L and g.
In order to tackle the selection of K L and g a linearization of
Eq. (36) at the equilibrium point is performed. The perturbed state
vector is dened as
x1 t  xe1

#
(64)

x2 t  0

The dynamics of the linearized system is represented by

P3

x2(t)=TR(t)-Tmax

Fig. 16. Stability region in the ideal case (perfect knowledge of the a1 , a2 and a3
values).

x2 t

T R t  T e t=Yx2 t

T R t  T e t
0

1
41 85 if T R bT e t
41 2 if T max bT R t

T R t  T e t=Yx2 t

T R t  T e t
0

1
o1 25

T R t  T max

o1

1
556

Computing the roots of the characteristic polynomial of Eq. (65),

and using Eqs. (66) and (67) yields

if T R bT e t
if T max 10T R t

K L T R t  T e t
1
xe1
g
Y0
Y0

(61)

r 1;2

(62)

5.5.3. Assuming that a2 a^ 3 t 0 =a3  a^ 2 t 0 =a2 and

a2 a^ 3 t0 =a3  a^ 2 t0 =a2 have different signs
Assuming that a2 a^ 3 t 0 =a3  a^ 2 t 0 =a2 and a2 a^ 3 t 0 =a3 
a^ 2 t0 =a2 have different signs, then the worst case is when
a^ 3 t 0 =a3  a^ 2 t 0 =a2 o0 and a^ 3 t 0 =a3  a^ 2 t 0 =a2 40. This
can be concluded by inspecting Dx2 t=DTe as a function of x2 t.
There is the possibility that the sign of a^ 3 t 0 =a3  a^ 2 t 0 =a2
forces Dx2 t=DTe o0 for x2 toT r t  T e t. To compensate this
situation K L =g must be selected according to the following
condition which is a generalization of Eq. (62):


D0
Y0 
KL
DTe
(63)
4xe1
T R t  T e t
g

v
u

 !2

@f 2 x1 ; x2  u
@f x1 ; x2 
t @f 2 x1 ; x2 

4 2



@x2
@x2
@x1
xe
xe
xe
2

(69)

A necessary condition on K L to have a stable linearized system is

that @f 2 x1 ; x2 =@x2 jxe o0, the other condition on g is needed and
can be selected by setting @f 2 x1 ; x2 =@x2 jxe 2  4@f 2 x1 ; x2 =
@x1 jxe 0 to have a double real pole.
In the special case with a^ 1 t 0 a1 , a^ 2 t 0 a2 , a^ 3 t 0 a3
those conditions imposed a relation between K L and g given by

which denes a minimum value on K L =g such that

Y0  1
T R t  T e t

(67)

(68)

In this case Dx2 t=DTe 1=Yx2 t 1=Yx2 t. The function

1=Yx2 t have values lower than 1, except at x2 t T R t  T e t.
This is illustrated in Table 6.
In order to include the equilibrium point xe in the stability
zone, the following condition must be fullled:

4xe1

(66)

T 2R t T 2e tT R t T e t
#


a^ 3 t0 a^ 2 t0
4T 34 t

a2
a3
a2
T 2R t T 2e tT R t T e t

x2 t

KL

(65)

xe



@f 1 x1 ; x2 
@f 1 x1 ; x2 

0;

 1
@x1
@x2
x
xe
e
@f 2 x1 ; x2 
a3
  a^ t g
@x1
3 0
x
e



@f 2 x1 ; x2 
a3
a^ 3 t0 a^ 1 t0


K L a1

a3
a1
@x2
a^ 3 t0
xe

Table 6
Evaluation of 1=Yx2 t.

xe1 o

3
@f 1 x1 ; x2
7
@x2
7
7 dxt
@f 2 x1 ; x2 5
@x2

Calculating the partial derivatives at the equilibrium point and

performing some algebraic manipulation yields

Table 5
Evaluation of Dx2 t=DTe 1=Yx2 t.

T R t  T max

@f 1 x1 ; x2
6
@x1
6
dx_ t 6
4 @f 2 x1 ; x2
@x1

KL
2

2

(70)

5.7. Simulation results

This section presents simulation results with the modied
adaptive control described by Eqs. (45)(49). The dynamics of the
shutter is simulated by a rst order lter with a time constant of
0.12 s. White noise Zt with limited bandwidth and power value
of 20 is added to the process output T s t. A second, rst order
lter with time constant 1.25 s is included and used to lter
T s t Zt. These two lters act as unmodeled dynamics, as a
remark, the unmodeled dynamics was not considered in controller design and in the stability analysis. The temperature
reference signal T R t is chosen to evolve by steps, this is used to
expose the control algorithm to the control signal saturation

ARTICLE IN PRESS
B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

1171

TR(t) and Ts(t) [K]

Temperature
1200
1000
800
600
400
200
0

50

100

150

200

250

300

350

400

450

500

300

350

400

450

500

350

400

450

500

time [s]

us (t) [%]

100

50

0
0

50

100

150

200

250
time [s]

Simulation of Solar Power variability

Gs(t) [W/m2]

1000
800
600
400
0

50

100

150

200

250
time [s]

300

Fig. 17. Simulation of the simplied adaptive controller. Process output and reference signal, control signal and Gs t. Closed-loop control, use of the anti-windup
mechanism.

Integral Error
500
400

es(t)dt

300
200
100
0
100
0

50

100 150 200 250 300 350 400 450 500

Fig. 18. Simulation of the simplied adaptive controller. Evolution of the

Closed-loop control, use of the anti-windup mechanism.

problem. The following anti-windup mechanism

(
x_ 1 t T R t  T s t if ut is not saturated
otherwise
x_ 1 t K L x1 t
is used to deal with the control signal saturation.

es t dt.

(71)

The controller parameters a^ 1 t 0 , a^ 2 t 0 , a^ 3 t 0 were selected to

be near the process parameter values at T s 400 K, the aim is to
simulate the lack of knowledge about the process parameter
values at higher temperatures. The following values are used in
the computer simulations, a^ 1 t 0 1:2  1011 , a^ 2 t 0 3:41:2
103 , a^ 3 t 0 2:3  102 , K L 0:2 and g 1:0  102 . The value
of K L was selected based on the level of noise present at the
process output Zt and the allowed level of shutter oscillation
induced by the feedback of Zt. The value of g was computed from
Eq. (70).
The simulation results are shown in Figs. 17 and 18.
The control algorithm is able to deal with the process
input saturation. The control signal us t is now more smooth
than in the previous computer simulations, due to the
selection of a lower gain for K L. This also contributes to reduce
the effects of the presence of unmodeled dynamics that can
cause stability problems, and also prevents the damage of the
shutter.
The results shown in Figs. 19 and 20 were obtained using
K L 1 and g 0:25, the gain was increased by a factor of 5. The
controller is able to track the temperature reference signal but the
control signal is too oscillatory and has a huge amplitude, this can
damage the shutter. This can be interpreted as a combination of
noise amplication and unmodeled dynamics excitation caused
by a huge value of K L .
The main conclusion is that the modied adaptive controller
shows a good performance and is easier to use.

ARTICLE IN PRESS
1172

B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

TR(t) and Ts(t) [K]

Temperature
1200
1000
800
600
400
200
0

50

100

150

200

250

300

350

400

450

500

300

350

400

450

500

350

400

450

500

time [s]

us (t) [%]

100

50

0
0

50

100

150

200

250
time [s]

Simulation of Solar Power variability

Gs(t) [W/m2]

1000
800
600
400
0

50

100

150

200

250

300

time [s]
Fig. 19. Simulation of the simplied adaptive controller. Process output and reference signal, control signal and Gs t. Closed-loop control, use of the anti-windup
mechanism.

rities present in the plant and disturbances associated to solar

radiation.
After development and evaluation of the adaptive controller
the following points can be drawn:

Integral Error
25

20

 The controller performed well in computer simulations.

 The controller has seven parameters to be tuned, which can be

15
es(t)dt

a problem.

10
To overcome the problems described above, an analysis of the
adaptive controller was done, and a modied version was
developed. The new version can be interpreted as a PI controller
with adaptive parameters and a feed-forward term. With this new
controller version it is easier to understand how to select the
controller parameter values. The evaluation of the simplied
control with computer simulation demonstrated that it has a good
performance.

5
0

50

100 150 200 250 300 350 400 450 500

Fig. 20. Simulation of the simplied adaptive controller. Evolution of the

Closed-loop control, use of the anti-windup mechanism.

es t dt.

6. Conclusions
This paper explores exact linearization and the Lyapunov
methods to design an adaptive controller for a solar furnace. The
aim is to develop a controller able to compensate the nonlinea-

Acknowledgments
The authors thank Dr. G. Olalde from PROMES-France for its
supporting in the SolControl project, which was nanced by EU
under the SolFace Contract number RITA-CT-2003-507091, and to
Prof. L. R. Guerra from the Materials and Engineering Department/
IST/TU LisbonPortugal. The experimental data used for the
simulations were obtained under the SolControl project. The

ARTICLE IN PRESS
B. Andrade da Costa, J.M. Lemos / Control Engineering Practice 17 (2009) 11571173

control algorithm was developed under the project SFERA, Grant

agreement no. 228296.
References
Berenguel, M., Camacho, E. F., Garcia-Martin, F. J., & Rubio, F. R. (1999).
Temperature control of a solar furnace. IEEE Control Systems, 19(1), 824.
Boyle, G. (1996). Renewable energypower for a sustainable future. Oxford: Oxford
University Press (in association with the Open University).
C
- engel, Y. A. (2003). Heat transfera practical approach (2nd ed.). New York:
McGraw-Hill.
Costa, B. A., Lemos, J. M., Guillot, E., Olalde, G., Rosa L. G., & Fernandes J. C. (2008a).
Temperature control of a solar furnace for material testing. In CONTROLO 2008,
8th Portuguese conference on automatic control.
Costa, B. A., Lemos, J. M., Guillot, E., Olalde, G., Rosa L. G., & Fernandes J. C. (2008b).
An adaptive temperature control law for a solar furnace. In MED08, 2008
Mediterranean conference on control and automation.
Fernandes, J. C., Amaral, P. M., Rosa, L. G., & Shohoji, N. (2000). Weibull statistical
analysis of exure breaking performance for alumina ceramic disks sintered by
solar radiation heating. Ceramics InternationalElsevier, 26, 203206.

1173

Fernandes, J. C., Oliveira, F. A. C., Granier, B., Badie, J. M., Rosa, L. G., & Shohoji, N.
(2006). Kinetic aspects of reaction between tantalum and carbon material
(active carbon or graphite) under solar radiation heating. Solar Energy, Science
DirectElsevier, 80, 15531560.
Garcia-Gabin, W., Zambrano, D., & Camacho, E. F. (2009). Sliding mode predictive
control of a solar air conditioning plant. Control Engineering Practice, 17(6),
652663.
Glattfelder, A. H., & Schaufelberger, W. (2003). Control systems with input and
output constraints. Berlin: Springer.
Kojima, Y., Taniwaki, S., & Okiami, Y. (June 2008). Dynamic simulation of stick-slip
motion of a exible solar array. Control Engineering Practice, 16(6), 724735.
Lacasa, D., Berenguel, M., Yebra, L., & Martnez D. (2006). Copper sintering in a solar
furnace through fuzzy control. In Proceedings of the 2006 IEEE international
conference on control applications, Germany.
Lemos, J. M. (2006). Adaptive control of distributed collector solar elds.
International Journal of Systems Science, 37(8), 523533.
Paradkar, A., Davari A., & Feliachi, A. (2002). Temperature control of a solar furnace
with disturbance accommodating controller. In Proceedings of the 34th
Southeastern symposium on system theory.
Slotine, J.-J. E., & Li, W. (1991). Applied nonlinear control. Englewood Cliffs, NJ:
Prentice-Hall.