You are on page 1of 6

1

Low cost two-axis solar tracker


with high precision positioning
A. Yazidil, Student Member, F. Betin , Member, G. Notton , G.A. Capolinol, Fellow
'Center of Robotics, Power Electrical Engineering and Control Systems of the University of Picardie,
15 av. Fran,ois Mitterrand, 02880 Cuffies, France (corresponding author: franck.beting u-picardie.fr).
2Laboratory of Sciences for the environment, university of Corse
Centre de recherche de Vignola, Route des Sanguinaires,
Ajaccio, Corse

Therefore, the instantaneous solar radiation collected by the


Abstract In this paper, the design of a low cost two-axis solar photovoltaic modules, assembled in a tracking system, is
tracker is introduced. The aim is to obtain a high precision higher than the critical irradiance level for a longer number of
positioning of the cell. The control-board is able to support hours than in fixed systems [8]. Tests have shown that up to
different control strategy such as classical techniques as PID 40% extra power can be produced per annum using a variable
algorithm or more advanced strategy such as fuzzy logic control
(FLC). Using the error signal, the tracking capacities of the elevation solar tracker [4, 5,6,7,9]. Vilela et al. [8] investigated
proposed approaches can be tested experimentally on the the effect of one-axis east-west tracking on solar radiation
experimental prototype built in our laboratory. received by a PV panel compared to fixed installation. Their
results showed that the irradiation collected by the tracker
plane is 19-24% higher than the one collected by the fixed
Index Terms- Position control, dc-motor, PID control, fuzzy logic system. In order to estimate the surplus of irradiance received
control on solar tracking, radiation data from 28 sites, especially in
moderate climate (between 540 and 1N latitude), were used
and Ratios of 1.34 (one axis) and 1.38 (double axis) were
found [10].
I. INTRODUCTION
But the use of a solar tracker can have another advantage than
pHOTOVOLTAIC (PV) technology for the conversion of only increase the daily solar energy, it increases the solar
sunlight into electricity is already a cost-effective method irradiance at given time allowing to reach a irradiance
for many applications worldwide. The solar electricity threshold under which the system can not run (for example a
finds applications in a number of systems for instance rural pumping system). Some photovoltaic pumping systems,
electrification, water pumping, satellite communications, grid- depending on their electric power and water head, require, to
connected applications and corrosion protection such as start operating, levels of collected irradiance higher than what
cathodic protection, etc. is known as critical irradiance level. That feature limits the
The efficiency and/or productivity of a PV system depends on daily operational time and the volume of water pumped. To
several factors such as climatic conditions (solar radiation, reduce its influence, the following technical means can be
ambient temperature and wind speed), optimal matching of the considered: (a) to use more efficient photovoltaic (PV) pumps;
system with the load, appropriate spatial placement of the (b) to increase the size of the PV array; and (c) to use solar
modules, i.e. placing the modules at an optimal inclination trackers [8]. The experiments carried out with a pumping
angle to the horizontal plane. system driven by a PV tracking collector show that the gains in
The majority of the PV systems used fixed solar modules; daily water volume are higher than the gains in collected
however, the tracking solar modules are also used for low- irradiation [8,11]. The water volume pumped by photovoltaic
power applications [1,2,3]. The daily average output of the PV pumping systems depends significantly on the critical
cells can be enhanced by a solar tracker, which forces sunlight irradiance (minimum irradiance level to start the operation of
to be incident normally (perpendicularly) to the PV cell at all the system). Large critical irradiance levels reduce the daily
times. operational time and consequently the water volume pumped
The benefit of using tracking devices is derived from the fact along the day [12]. As a result, the gain in terms of pumped
that the incidence angle of the direct component of solar water volume can be considerably larger than the gain in
radiation is more favorable, i.e., lower than or equal to the collected solar radiation. As a matter of fact, experiments with
angle that corresponds to fixed systems. The behavior of the a PV pumping system driven by a tracking device [8], showed
diffuse component for both systems is approximately the same. that, whereas, the increase in collected solar radiation was
around 22% (Recife-PE, Brazil), the gain in terms of water
volume was of the order of 39%.

1-4244-0232-8/06/$20. 00 C2006 IEEE 211

Authorized licensed use limited to: North China University of Technology. Downloaded on March 09,2010 at 00:36:04 EST from IEEE Xplore. Restrictions apply.
2

An ideal tracker would allow the PV cell to accurately point


towards the sun, compensating for both changes in the altitude II. MECHANICAL PART
angle of the sun (throughout the day) latitudinal offset of the
sun (during seasonal changes) and changes in azimuth angle. The complete mechanical system is shown in figure 1. The
The slow movement of the sun requires a damped system that photovoltaic module used (PWX 500 from Photowatt) is a
will also respond slowly and avoid an oscillatory movement. 4*9 polycristalline inch cells designed with double glass
Other desirable aspects would include the nocturnal technology (see figure 1). The module weight is about 9,2 kg
repositioning of the solar tracker to anticipate the alignment of and its size is 1042, 462, 39 mm. The typical power is 5OW
sunrise, opposite to that of the previous days sunset, reducing with a 12V configuration. You can see, on the figure 1, the
energy losses in the morning and also to include a vertical and the horizontal axis. The horizontal axis is
repositioning at an horizontal position when no direct radiation composed of an endless screw as shown on figure 2.
is available increasing the diffuse radiation on the PV panel.
Various sensing equipment and actuating devices, both
electrical (active) and mechanical (passive), are in use today
[12-13].
The study of solar tracking systems shows that generally they
have a vertical axial stand with rotary mechanism installed on
massive supports, and the tracking is done manually [14] or
automatically by fluid [15]. Poulek and Libra [16] describe an
automatic solar tracker that is driven by an electric motor: it
has a pyramidal stand and on one edge of the stand the solar
modules are installed.
The aim of this paper is to design a two-axis solar tracker that
is able to follow accurately the solar course with no oscillation.
For this, the PV cell will be driven by two dc-motors
controlled using a dedicated control board. In order to obtain a
high precision positioning of the cell, different control
strategies will be studied such as Proportional-Integral-
Derivative (PID) algorithm or more advanced strategy such as rig. 1. cu1omp1iee mcnanicnai systemI1
fuzzy logic control (FLC). This last algorithm is well-known
for its robustness between disturbances and its simplicity. To move the cell according to the vertical and the horizontal
Indeed, they do not require a lot of computational operations axis, two dc motors are used. They are classical 12 V DC
such as adaptive or predictive control algorithms. motors used for automotive applications (windscreen wiper).
The second part of this paper will be dedicated to the Figure 2 represent the DC motor used for the horizontal axis
mechanical design of the solar tracker. The choices for the with the endless screw.
mechanical parts, the motors and the sensors will be presented.
The third part of this paper will be dedicated to the control
board and the algorithm used to impose the position of the cell.
At the present time, the control system is not already on board
because we are testing different control algorithms. For this,
the Labview software is employed. This one allows to generate
the control algorithm using simple graphic tools and also to
obtain easily the control signals and the signal given by the
sensors. On this paper, the results obtained with a classical
Proportional-Integral-Derivative (PID) controller are shown.
Nevertheless, the basic concepts of fuzzy logic control, that
will be used later, will be introduced.

rig. 2: DU motor (norizontal ax-is

212

Authorized licensed use limited to: North China University of Technology. Downloaded on March 09,2010 at 00:36:04 EST from IEEE Xplore. Restrictions apply.
3

To obtain informations about the relative position of the cell, a sampling frequency, the parameters of the regulator can also
system composed of two photovoltaic generators (Solems 005- be modified on-line. Fig. 4 represents the chosen dialogue
048-016) is used. These components give a voltage and a interface for a Proportional-Integral-Derivative (PID)
current as a function of the solar radiation. regulator. On this interface, you can see:
The two cells are disposed on two different surfaces which - the evolution of the error signal between the two PV
present an angle of almost 90 between each other and an sensors for one axis,
angle of 450 with the surface of the photovoltaic cell (see - the evolution of PID output which is the control for
figure 3.a). The voltage differences between the two PV one dc motor,
sensors assess if the position of the cell is correct. Of course, - the chosen values P, I and D for the regulator that can
to control the vertical and the horizontal axis, two PV sensors be modified on-line,
systems as the one introduced just before are required. Fig. 3.b - the saturation values for the regulator that can be also
represents the PV sensors system of the vertical axis. changed on-line.

PV sensors

PV 4cel

PV cells

Fig. 4 : Chosen dialogue interface (PID control)


To follow the solar course during the day, the only two
informations are the error signals given by the two PV sensors
systems. Each error signal is used to control one axis: the
vertical or the horizontal one. Now, we will see first the
concept of Proportional-Integral-Derivative (PID) regulation
and then the theory about Fuzzy Logic Control (FLC)

B. PID Control
b) Sensors system for the vertical axis Proportional-Integral-Derivative (PID) control algorithm is
widely used in industry because of its simplicity.
Fig 3 : PV sensors system Indeed, the control u is obtained from the error e signal using
the following relation:

III. CONTROL PART u = Kp.e+Ki.


P' e.dt+Kdd- de (1)
d~ ~ ~ct
In this third part, we will present the control board used to Where Kp is the gain of the proportional action, Ki is the gain
regulate the two axis and the regulation algorithms which are of the integral part and Kd is the gain of the derivative part.
implemented on the control board. We will see that different With the proportional action, the controller output is
kinds of algorithms can be applied such as Proportional- proportional to the error or a change in measurement.
Integral-Derivative (PID) or fuzzy logic control (FLC) With integral action, the controller output is proportional to
the amount of time the error is present. Integral action
A. Control board eliminates offset.
At the present time, the control system is not already on board With derivative action, the controller output is proportional to
because we are testing different control algorithms. To define the rate of change of the measurement or error. The controller
easily the most suitable control algoritm, the Labview software output is calculated by the rate of change of the measurement
is employed in our laboratory. With this software, the signal with time. Derivative is often used to avoid overshoot and to
acquisitions, the control computation, the imposition of the stabilize loops.
control and the visualisation of the different signals are very PID control is a very efficient control method that is very often
easy since it can be made using simple graphic tools. The used in industry, but this technique presents two main

213

Authorized licensed use limited to: North China University of Technology. Downloaded on March 09,2010 at 00:36:04 EST from IEEE Xplore. Restrictions apply.
4

drawbacks. The first one is the determination of the 3


parameters Kp, Ki and Kd* Indeed, when you have no precise UA and B (x) = min [ uA(x), u(X)]
model of the system, it is very difficult to define the optimum UA and B (x) = UA(X)-UB(X) (3)
values for these three parameters. The second drawback is that UAor B (X) max [ uA(x), uB(X)]
this kind of regulator is not robust between plant parameters UAor B (X) [uA(x) + uB(x)] / 2
variations. In our case, we have no precise model of the whole
system and this last one can be submitted to variations of the Where uy(x) is the membership factor of the element x to the
mechanical environment. That's why we turn towards Fuzzy subset Y.
Logic Control (FLC) which does not require a precise model
for the design of the regulator and which is well-known for its Defuzzification: Finally, the fuzzy logic controller output
robustness capacities. expressed in linguistic codes must be converted back to
numerical values and scaled. This operation is called
"defuzzification". To determine the value of action to be taken
C. Fuzzy Logic from all the contributions, we can either choose the "maximum
In this third part, the fuzzy logic control algorithm [17] that value" or use the "center of gravity" methods. In the "center
will be used to control the position of the two axis of the of gravity" method, the action is given by the center of the
module is presented. summed area.
The fuzzy algorithm is based on the integration in an automatic
control system of the human operator expertise about the
control of a process. To adjust the output of the process, it is Presentation of the proposed FLC: The main problem with
generally necessary to use: fuzzy logic controller generation is to choose the parameters of
- the error signal equals to the set point minus the process this one. From our experience, we have chosen the following
output, structure (see fig. 5) to control the position of the cell. This
- the change of error equals to the error from the process approach has been still successfully applied to control step-
output minus the error from the last process output, motors [20], DC motors [21] or induction machines [22]. For
- and the error sum equals to the error from the process output one axis, to adjust the output of the drive unit, we use the error
plus the error from the last process output. signal between the two solar sensors and the change of error:
With those three contributions, a fuzzy controller can be easily e(k) = Ul(k) - U2(k)
built up. Once those values obtained, the computation of the Ae(k) = e(k) - e(k-1) (4)
regulator output is made in three steps [18]: fuzzification,
decision making, defuzzification. The structure of the controller for one axis is presented on the
figure 5.
Fuzzification: Firstly, input variables are normalized to an
"universe of discourse" using scaling factors. Using these FUZZY CORRECTOR
normalized quantities, the fuzzy logic controller inputs can be
described by membership factors for every linguistic codes.
This operation, which is called "fuzzification", requires the
definition of linguistic sets and their membership functions. In
general, seven linguistic sets are suffisant. The use of more
than seven linguistic sets does not increase the precision but
increases the time computation. The shape of the fuzzy set is
quite arbitrary and depends on the user's preference. For
simplicity, triangular and trapezoidal shapes are usually used.
Fig. 5: Structure of the controller
Decision making: The fuzzy controller output is deduced from
"fuzzified" inputs and fuzzy logic rules. The rules are The scale factors K1 and K2 are given by the open loop
combined to form a decision table for the fuzzy controller. response in the ex de space [18]. They are chosen in order that
This table consists of values showing the different situations the normalised inputs are well adapted to the universe of
experienced by the system and the corresponding control discourse for any working point. We have chosen seven
actions. The production of those rules can be heuristic (based linguistic sets (LP, MP, SP, ZE, SN, MN, LN) for the error,
on judgments of experienced plant operator) or general (built the error change and the output of the corrector.
by analogy with conventional regulators). The aim of those Symmetric triangular shapes are used for fuzzy sets as
linguistic rules is to obtain a non linear relationship between shown on figure 6:
system states and control action which gives a better control
action than linear alternative. Those rules are generally
expressed in if and then statments:
kf( . and . ) then ... or (2)
The operator and is usually realized with the function min or
product and the operator or with the function max or sum.

214

Authorized licensed use limited to: North China University of Technology. Downloaded on March 09,2010 at 00:36:04 EST from IEEE Xplore. Restrictions apply.
5

Membership IV. EXPERIMENTAL RESULTS


b --0.
This part is dedicated to the first experimental results that we
GN MN PN 2 PP 'MP GP
have obtained on our prototype. Fig. 8 represents the error
signal (blue) and the controller output (green) for the
horizontal axis when a PID regulation is imposed. As we can
see on this figure, when an error signal is detected on one axis,
I I the output of the PID regulator (green curve) makes the error
-

-3 -1.5 -0.5 0 0.5 1.5 signal (blue curve) converge until zero. We can also remark
Universe of discourse that there is no overshoot since the error signal reaches zero
Fig. 6: Membership functions ofthe controller without changing of sign. Furthermore, we can see that the
system is completely stable since there is no oscillation on the
The fuzzy control matrix used for inference is close to Mac steady state. The only oscillations during the transient state are
VicarWheelan one's [23]. This matrice is shown in table 1. due to the non linearities of the mechanical system (endless
screw).
ezie GN MN PN EZ PP MP GP Nevertheless, the response of the system is very slow since it
takes almost 17 seconds. This time may be decreased by using
GP EZ PP MP GP GP GP GP fuzzy logic control, but at the present time this has not been
MP PN EZ PP MP GP GP GP validated experimentally.
PP MN PN EZ PP MP GP GP
EZ GN MN PN EZ PP MP GP
PN GN GN MN PN EZ PP MP
MN GN GN GN MN PN EZ PP
GN GN GN GN GN MN PN EZ
Table 1: Fuzzy rules

The max-min inference mechanism is used to determine which


rules are active for any input data. The "center of gravity"
method has been chosen to determine the value of action.
Using this configuration, the characteristic of the controller
is the following:
Time (s)

Au,,
V. CONCLUSION
2
In this paper, a low cost two axis solar tracker has been
1 presented. This study has effectively begun in October 2005, it
0
corresponds to the starting point of the collaboration between
the university of Corse and the university of Picardie. Indeed,
-2 the two laboratories have decided to associate their knowledge
3 in electrical engineering and solar to electric conversion for
4
renewable energy.
The mechanical prototype is now finished and the first tests
have been realised with a classical Proportional-Integral-
Ae11 Derivative (PID) control algorithm. The experimental results
have shown that the system works and the next step consists
Fig. 7 : Characteristic of the fuzzy controller now in designing a more suitable control technique such as
Fuzzy Logic Control. Once the control step finished, we
should be able to define the part of the extra power that can be
produced using such kind of system. But before, we have to
define laws that reduce energy losses such as the repositioning
at the end of the day or the repositioning at a horizontal
position when no direct radiation is available.

215

Authorized licensed use limited to: North China University of Technology. Downloaded on March 09,2010 at 00:36:04 EST from IEEE Xplore. Restrictions apply.
6

VI. REFERENCES

[1] J.W. Twidell, A.D. Weir, Renewable Energy Resources, E [13]. Helwa, N.H., Bahgat, A.B.G., El Shafee, A.M.R., El
& F.N. Spon, London, 1986. Shenawy, E.T., 2000. Maximum collectable solar energy by
[2] G. Boyle (Ed.), Renewable Energy: Power for Sustainable different solar tracking systems. Energy Sources 22, 23-34.
Future, Oxford University Press in [13] Helwa, N.H., Bahgat, A.B.G., El Shafee, A.M.R., El
association with the Open University, New York, UK, 1996. Shenawy, E.T., 2000. Computation of the solar energy
[3] T. Markvart (Ed.), Solar Electricity, Wiley, New York, captured by
2000. different solar tracking systems. Energy Sources 22, 35-44.
[4] Kh.S. Karimov, J.A. Chattha, M.M. Ahmed, et al., J. Acad. [14] S. Dadamukhamedov, R.A. Muminov, M.N. Tursunov
Sci. Tajikistan XLV (9) (2002) 75-83. and M. Khamrit, Sol. Eng. (5) (1998) 86-88.
[5] M.J. Clifford, D. Eastwood, Design of a novel passive [15] T. Markvart (Ed.), Solar Electricity, Wiley, New York,
solar tracker, Solar Energy 77 (2004) 269-280 2000.
[6] Lorenzo, E., Perez, M., Ezpeleta, A., Acedo, J., 2002. [16]. V. Poulek, M. Libra, Sol. Energy Mater. Sol. Cells 51
Design of tracking photovoltaic systems with a single vertical (1998) 113-120.
axis. Progress in Photovoltaics 10, 533-543. [17] L.A. Zadeh, "Fuzzy Sets," Information and Control, vol.
[7] Patil, J.V., Nayak, J.K., Sundersingh, V.P., 1997. Design, 8, 1965, pp. 338-353.
fabrication and preliminary testing of a two-axes solar tracking [18] Y.F. Li, C.C. Lau, "Development of fuzzy algorithms for
system. Reric International Energy Journal 19, 15-23. servo systems," IEEE Control System Magazine, April 1989,
[8] Vilela OC, Fraidenraich N, Tiba C. Photovoltaic pumping pp. 65-72.
systems driven by tracking collectors: [19] M. Braee, D.A. Rutherford, "Selection of parameters for a
experiments and simulation. Solar Energy 2003;74:45-52. fuzzy logic controller," Fuzzy Sets and Systems, vol. 2, 1979,
[9] Gay CF, Yerkes JW, Wilson JH. Performance advantages pp. 185-199.
of two-axis tracking for large flat-plate photovoltaic energy [20] F. Betin, D. Pinchon, G. A. Capolino
system. In: Proceedings of 16th IEEE Photovoltaic Specialists "Fuzzy logic applied to speed control of a stepping motor
Conference, 1982. p. 1368-71. drive", IEEE Transactions on Industrial Electronics, vol. 47,
[10] Stefan Nann, Potentials for tracking photovoltaic systems n03, June 2000, pp. 610-622.
and V-troughs in moderate climates, Solar Energy, 45-6, pp. [21] A. Sivert, F. Betin, D. Pinchon
385-393. "Selection des parametres d'un regulateur flou a partir de la
[11] F raidenraich, N., Vilela, O.C., 2000. Performance of theorie des modes glissants: Application au moteur a courant
solar systems with non linear behavior calculated by the continu", Revue Internationale de Genie Electrique, vol.6, no
utilizability method. Application to PV solar pumps. Solar 6, 2003, pp. 609-636.
Energy69, 131-137. [22] F. Betin, A. Faqir, D. Pinchon, G. A. Capolino
[12]. J. Bione, O.C. Vilela, N. Fraidenraich. Comparison of "Fuzzy-logic control applied to position control of an
the performance of PV water pumping systems driven by induction machine", Electromotion, vol. vol.10, n04, July-
fixed, tracking and V-trough generators. Solar Energy 76 September 2003, pp.455-460
(2004) 703-711 [23] P.J. Mac Vivar Whelan, "Fuzzy sets for man-machine
interaction," Int. J. Man-Machine Studies, vol. 8, pp. 687-697,
1976.

216

Authorized licensed use limited to: North China University of Technology. Downloaded on March 09,2010 at 00:36:04 EST from IEEE Xplore. Restrictions apply.

You might also like