APPLICATIONS

Tracking and back-tracking

E. Lorenzo*, L. Narvarte and J. Muñoz
Instituto de Energía Solar, Grupo de Sistemas Fotovoltaicos, Universidad Politécnica de Madrid (IES-UPM), Ciudad Universitaria s/n,
28040 Madrid, Spain

ABSTRACT
This paper presents a review of back-tracking geometry not only for single axis but also for two-axis tracking and analyses
the corresponding energy gains. It compares the different back-tracking strategies with the ideal tracking in terms of energy
yield concluding, on the one hand, that back-tracking is more useñil for single horizontal axis than for the single vertical
one, and on the other hand, that back-tracking is more efficient when applied in the primary axis of a two-axis tracker.
Copyright © 2011 John Wiley & Sons, Ltd.

KEYWORDS
tracking; back-tracking

*Correspondence
E. Lorenzo, Instituto de Energía Solar, ETSI Telecomunicación, Ciudad Universitaria s/n, 28040 Madrid, Spain.
E-mail: lorenzo@ies-def.upm.es

1. INTRODUCTION energy gains. The results help us to understand why back-

tracking implementation in commercial products is today
Solar tracking is been used with PV flat modules for more restricted to single horizontal axis trackers, and suggest
than 20 years, and some studies [1] suggest that it will that back-tracking is still attractive for other trackers, in
maintain its interest while the PV module cost remains particular for the two axis ones.
more than 0.8€/Wp. More than 20 companies currently
offer products with different tracking possibilities: two
axis, single vertical axis, single horizontal axis and single 2. SINGLE HORIZONTAL AXIS
tilted axis [2-5]. When several trackers are placed in
a limited área, the problem of mutual shadowing arises. Figure la describes the geometry of a single horizontal,
A previous paper [1] has dealt with trackers on the N-S oriented axis, ideal tracking in the case of mutual
assumption of ideal tracking, le. when all the receiver shadowing. The ideal tracking angle, &>ro, is given by:
surfaces are always perpendicular to the Sun. Tracking x
geometry has been reviewed and the relationship between tan COID = - (1)
the yearly energy gain and land occupation has been z
analysed. where (x, y, z) are the Cartesian coordinates of the Sun
vector referred to a system with the x-axis pointing west,
Shade not only reduces the electric output power but
the y-axis pointing south and the z-axis pointing to the
also increases the risk of hot spots. Henee, the interest
zenith [1]. These coordinates are related to solar elevation,
of the so called back-tracking strategy, which consists of
ys, and solar azimuth, i/rs, angles by:
moving the surface angles away from then ideal valúes, just
enough to get the shadow borderline to pass outside x — cosj/ssini/fs
the border of the adjacent tracker. This way, first, shade is y — cosyscosi/rs (2)
ñilly avoided and, second, the loss due to the angle of z — sinys
incidence is minimised. Back-tracking geometry for single Shadow length, s, and shaded fraction of the PV
horizontal [6] and single vertical axis [7] was described generator área, FS, are given by:
some years ago. This paper presents a comprehensive
review of back-tracking geometry not only for single axis COS CÜJT, / ^ \

but also for two-axis tracking, and analyses the associated FS = max[0, ( 1 - ^ ) ]
 Zenith A better approximation is obtained by taking into
(a)
account the shaded blocks. A 'block' is defined here as
a group of cells protected by one bypass diode. A block
is shaded when at least one of its cells is shaded. An
aecurate empirical expression is [8]:
NSB
( 1 - F E S ) = (1-FS) 1 (10)
NTB + 1

where A^TB 1S the total number of blocks inside the

specific array and JVSB 1S the number of shaded blocks.
In that follows, the cases defined by Equations (8)-(10)
will be termed as 'optimistic', 'pessimistic' and 'realistic',
respectively.
(b) In order to analyse the back-tracking energetic impact,
Zenith
we have calculated the energy yield of grid connected
PV systems, with and without back-tracking, for the two
different representative electric layouts described in
Figure 2a,b. The PV module (Figure 2c) is made up of
48 solar cells and 3 bypass-diodes, arranged in such a way
that the number of blocks is 3 x 1 (horizontal x vertical).
Then, layout Figure 2a,b are characterised, respectively, by
90 x 1 and 90 x 2 blocks. Calculations have been made
with the IES-UPM own code, whose details can be found in
Reference [1]. This code considers the effect of changing
Figure 1. Geometry of a single horizontal, N-S oriented, axis absorptance of the PV module with incidence angle and
tracking. Projections on the vertical, E-W oriented, plañe: (a) also the effect of albedo. The selected location is
ideal tracking in the case of shadowing, (b) back-tracking. Amareleja, in South Portugal (Latitude: 38.2N; Longitude:
7.2W; solar radiation and ambient temperature data as
where LEW is the distance between axes. The back-tracking given by PVGIS [9]). Despite not being directly related
correction angle, a>c, is: with this paper, it is worth mentioning that there has been a
48 MW PV plant in this village, based on trackers, in
<"c = oijD-ctísx: (4)
routine operation since 2009 (http://www.acciona-ener-
gia.com). Figure 3 shows the evolution of the yearly energy
COS ü>c — ¿EW COS (ÜID (5)
yield versus distance between tracker axes, or 1/GCR.
It is worth mentioning that the ratio between the PV Table I gives numerical results for 1/GCR = 3.5, often
array área to total ground área, the so-called ground cover found in practical applications. The key result is that,
ratio (GCR), is given by: compared with ideal tracking, the back-tracking strategy
1 is always cióse to the optimistic case and better than the
GCR: (6) realistic ones. Therefore, back-tracking is highly attractive,
¿EW
which helps us to understand why it has been implemented
Now, let us consider a PV array affected by shading. At in single horizontal axis trackers for many years [10,11].
any instant, we can state: It is worth mentioning that PV modules at Figure 2 are
oriented in such a way that their short side is parallel to
FS=FNS(1-FES) (7)
the rotation axis. In fact, because the bypass disposition
where Ps and F N S represent the power delivered by the PV of these particular PV modules, the orientation with their
array with and without shading, respectively, and F E S long side parallel to the axis encompasses less shading
known here as the effective shading factor, whose valué sensitivity. However, back-tracking is still attractive also in
determines the power decrease and depends on the this case.
particular electric PV system layout (mainly solar cells
series and parallel association, and bypass diode disposi-
tion). Whatever the case, it encompasses two extremes.
The most optimistic is when the power reduction is just
3. SINGLE TILTED A N D
equal to the shaded fraction, FS, that is:
DEVIATED AXIS
F ES = FS (8) The geometry of back-tracking for a rotating single axis,
tilted an angle /SAXIS (the angle that the axis makes with the
And the most pessimistic is when the mere existence of
horizontal plañe) and aligned at an angle «AXIS relative to
any shade fully caneéis out the power. That is:
the N-S direction is an analogue to the previous case. It
FS > 0 =• FES = 1 (9) must be noted that the coordinates of the Sun (x7, y', ¿),
E. Lorenzo eí al. Tracking and Back-Tracking

(a)

wv1 2 3
V
30
c- (c)

(b)

c-

Figure 2. Electrical layouts of two different PV generators, (a) and (b), whose module has 48 solar cells and three bypass-diodes (c).

now referred to as a system with the y-axis coinciding with Figure 6 shows a case of shading from West. Ideal
the rotating tracking axis; the x-axis horizontal, roughly tracking angle, shadow length and shaded fraction are
pointing west and the z-axis pointing upwards, are given given by:
by:
tan coiD = - (12)
x' — x eos a AXIS —y sin a AXIS y
y = x eos ^AXJS + y eos / ^ s eos «AXIS ~z sin /3 AX1S s = b (eos /3AX1S + sin /3AX1S cot}/s) (13)
z' =x sin /3AX1S sin «AXIS +y sin /3AX1S eos «AXIS +Z eos /3AX1S
(14)
FS\v = FSl\yFS2\y
(11)
Where:
Figure 4 represents the energy yield for the same cases s—LEW sin \¡rs
as Figure 2, but now with the axes tilted /SAXIS — 20° and FSU and FS2«
deviated also at a AXIS = 20°. Table I gives numerical COS
1—LEW i/fs
results for 1/GCR = 3.5. Again, back-tracking is cióse to (15)
1
the optimistic case and better than the realistic cases. It is
worth mention that deviated axes, despite being not The back-tracking correction angle is again given by
optimal, offer sometimes practical interest for adapting to Equations (4) and (5)1. Obviously, shading from E is
non-flat terrains. symmetric to this case.
Figure 7 show a case of shading from SW. Shaded
fraction is now given by:
FSsw — FSlswFS2<; (16)
4. SINGLE VERTICAL AXIS
Where:
Let us consider a set of vertical trackers arranged as shown LNS eos T/^S+LEW sin jrs
FS1 SW = 1 s and
in Figure 5. A general tracker (like the striped one, in the tan i/f s -L E W /L N (17)
centre of the figure) can be shaded from five other trackers, FS2sw = (L EW +cos i/r s )/(L N S -sin i/r s )-L E W /L N S

respectively placed at its E, SE, S, W and WE. Of course, It is also possible to establish a back-tracking strategy
it is possible to also imagine shades from other than these
for avoiding shade from the SE and SW trackers. However,
five trackers. For example, from ones placed at ESE
and WSW. In higher latitudes shading can likely also 1
ln fact, Equation (5) is the solution for the particular case
oceur from NE and NW trackers. However, these cases /SAXIS = H/2, i-e. with the tracker vertical. But, it can be accepted
encompassed rather limited practical interest, so we have as a general solution without great error. The interested reader is
decided to avoid here the associated complexity. encouraged to consult the reference [7].
 Figure 8 shows energy yield versus the GCR for a single
vertical axis tracking, again at Amareleja. The geometry is
defined by the /SAxis = 40°, b=l/^2, L N S = 2.06; and
electric layout by the number of bypass diodes in the
horizontal and vertical direction of each tracker, 18 and 3,
—e— Optimistic
^^Pessimistic
respectively. Table I gives numerical results for 1/
-^—Layout-(a) Backtracking
Layout-(a) No backtracking
GCR = 4.5. Now, the key result is that, compared with
—i— Layaut-(b) No backtracking ideal tracking, back-tracking strategy is slightly better than
~ I,ayont-(ti) Backtracking
the realistic case and both are cióse to the optimistic case.
Obviously, this result is dependent on the particular
geometry and electric layout. In fact, as the number of
bypass diodes in the vertical direction increases, shading
Figure 3. Yearly energy yield versus the distance between axes
losses decrease and, therefore, back-tracking becomes less
for single horizontal axis tracking. Back-tracking is cióse to the
optimistic cases and better than the two realistic cases.
attractive. However the sensitivity of energy to layout is in
practice rather low. For example, changing the electrical
this would be scarcely practical, because it encompasses a layout from 18 x 3 to 18 x 6 led to reduce shading losses by
sudden 180° turn. For example, back-tracking for avoiding about 1%. Henee, one can say that back-tracking on vertical
SW shade forces trackers to move towards the SE until axis tracking tends to be roughly neutral in puré energy
they, shade and shaded, are precisely aligned with the Sun. terms. That helps us to understand why it is not implemented
Just at this moment, the trackers should turn 180°. in today's commercial vertical axis trackers. However, it
Otherwise, The Sun's rays will fall on the back surface of must be noted that shade is prone to inducing hot-spots,
the PV modules. Apart of that, because shading occurrence which can even result in PV module destruction after a
is normally associated to low Sun elevation angles, shading certain operating time [12] and that back-tracking fully
from the SE and SW used to be scarcely relevant. For avoids shade. Henee, even irrespective of the energy balance,
example, for the case considered in Table I (described we think back-tracking can still be attractive in practise.
below) energy losses due to shading from the E and W, and
SE and SW are 8 and 0.4%, respectively. For both reasons,
lack of practical sense and irrelevance have led us to 5. TWO AXES, PRIMARY
disregard the implementation of back-tracking to avoid VERTICAL
shading from SE and SW. Along the same lines, we have
also disregard designs involving any shading from S. It is Two-axis trackers are made up of an axis connected to the
easy to see that this leads to the condition: foundation, called the primary axis by us, and another axis
fixed to the primary, here called the secondary. The primary
¿NS
> eos ¡3A sin fiA tan (18) vertical axis, or pedestal, is today the most common two-
axis tracker. The figures and equations above for the single
Where <¡> is the site latitude and <5MAX — 23.5°the máximum vertical axis can also be applied here, once it is considered
Sun declination angle. that the /SAXIS is n o w continuously adjusted by rotating the

Table I. Yearly energy yield of grid-connected PV systems for several tracking types, with and without back-tracking, in a particular
location.

Tracking type Single axis Two axes

Horizontal N-S Tilted, 20° Vertica Primary vertica Primary horizonta

1/GCR 3.5 3.5 4.5 5.5

Without back-tracking
Optimistic 1893 2071 2132 2316 2225
Realistic (a) 1776 (6.2) 1919 (7.3) 2121 (0.5) 2302 (0.6) 2124 (4.5)
Realistic (b) 1695 (10.5) 1878 (9.3) 2103 (5.5)
Pessimistic 1695 (10.5) 1919 (7.3) 2086 (2.2) 2263 (2.3)
With back-tracking
On the primary axis 1873 (1.0) 2054 (0.8) 2129(0.14) 231 5 (0.04) 2217 (0.3)
On the secondary axis 2276 (1.7) 2124 (4.5)
On both axes 2287 (1.2) 2217 (0.3)

Valúes are given in kWh per kW of PV generator peak power. The numbers given in parentheses as percentages represent, the losses
with respect to the optimistic cases. It must be noted that, despite being set here to 100%, these optimistic cases still include some
shading losses due to mutual shading between trackers. For example, the energy yield with only one single axis tracker (1 /GCR —> oo)
is 2325, so the optimistic case as defined here includes 8.3%o shading losses.
 Optimistic
—'—Pessimistic
-^^Layout-(a) Backtracking
—*— Layout-(b) No Backtracking
-^—Layout-(b) Backtracking
-^^Layout-(a) No backtracking

Figure 4. Yearly energy yield versus the distance between axes

for single axis tracking tilted /¡AXIS = 20° and deviated also at
«AXIS = 20°. The back-tracking is, again, cióse to the optimistic
cases and better than the two realistic cases

secondary axis in such a way that:

(19)
2 2
tan /SJD = (* + y ) (20)

As regards back-tracking, two axes imply two different

possibilities. For example, shading from the West,
described in Figure 6a, can be avoided by rotating the
primary axis towards S (Figure 6b), Equations (4) and (5) - Figure 6. Geometry of single vertical axis tracking. Projections
or by rotating the secondary axis towards the horizontal, to on the horizontal plañe: (a) ideal tracking in a case of shading from
a tilt angle, /Smc- In this case, the back-tracking correction West, (b) corresponding back-tracking

angle is:

Pe = A D - A I (21)

MEiW
COS /SJDC = — e o s /S 1D (22)

Shading from the SW can also be avoided by rotating the

77777777 //////// secondary axis towards the horizontal. Now, the back-

SE SW W
-+- -f tan^s <
'-NS
w -y

Figure 5. Geometry of a field with single vertical axis trackers:

tilt angle, /¡AXIS; aspect relation b; spacing between adjacent
trackers in the North-South direction and East-West direction,
í_Ns and Z_EW. A general tracker (in the centre of the figure) can Figure 7. Geometry of single vertical axis tracking. Projections
receive shade from five trackers, placed at the E, SE, S, SW on the horizontal plañe: ideal tracking in a case of shading from
and W. South-West.
Figure 8. Energy yield versus the GCR for a single vertical axis
tracking at Amareleja.

tracking correction angle is:

LEW sin \¡ts + LNS eos \¡ts

eos /S1DC = eos /S1Dí (23)

Figure 10. Geometry of a two-axis tracking, being the primary

Figure 9 shows energy yield versus the GCR for a two-
horizontal, in a case of mutual shadowing.
axis tracking, primary vertical, again at Amareleja and with
b = 1/^/2, LNS = 2.06 and NTB = 18 x 3. The tilt angle has
fraction and back-tracking correction angle are given by:
been limited to 75°. Table I gives numerical results for 1/
GCR = 5.5. Three different back-tracking possibilities y (24)
tan /S1D =
í2+z2)l/2
have been considered: only on the primary, or vertical, axis
(this way, only shade from the E and W are avoided); only
on the secondary, or horizontal, axis (all shade, from the FS>, (25)
E, W, SE and SW are avoided) and on the two axes, each (¿/eos /Sn
time selecting the axis requiring the lowest back-tracking
¿NS
correction angle (all shade is again avoided). Now, the key eos /Sn eos /Sn (26)
result is that the back-tracking is significantly more
efficient on the primary axis than on the secondary one. It is worth commenting that, during the extreme hours of
the day, back-tracking on the primary axis - again,
Equations (4) and (5) - affeets the valué of the tilt angles
(ideal and back-tracking) of the secondary axis. Corre-
6. TWO AXES, PRIMARY sponding equations can be derived from geometric
HORIZONTAL considerations. In order to make the understanding of
the paper easier, these equations are not presented here, but
Figure 10 shows the geometry of a two-axis tracking, the they have been implemented in the corresponding software
primary being horizontal, in a case of mutual shadowing. This code.
kind of tracking has led to several commercial produets (see, Figure 11 shows energy yield versus the GCR for a two-
for e.g. [13]). The ideal primary axis rotating angle is given by axis tracking, primary horizontal, again at Amareleja
Equation (1). The ideal secondary axis tilt angle, /3n> shaded and with ¿ = 0.28 (e.g.: files of 5 x 1 modules, each
1 x yj2 m ), A^TB — 15 x 1. Practical reasons impose severe

****
211)0

3 2100
2000 -
—e— Optimistic
—B~ Pessimistic
i
% 1900 —a— Optimistic
— N o backtracking
5 —B— Pessimistic
——Vertical axis backtracking
—i—Horizontal axis backtracking —*- No backtracking
—*— Two-axis backtracking —**— Horizontal backtracking
1700 - -v— Óptima! backtracking
— Vertical backtracking
1600 -

Figure 9. Energy yield versus the GCR for a two-axis tracking,

primary vertical at Amareleja with b=1/j2, LNS = 2.06 and Figure 11. Energy yield versus the GCR for a two-axis tracking,
NTB = 1 8 x 3 . primary horizontal at Amareleja with b = 0.28 and NTB = 1 5 x 1 .
limits to LNS valúes. Otherwise, the tracking structure 2. Siemer J. An industry in transition. Global survey on
occupation becomes very low and, due to the correspond- tracking systems. Photon International 2009; 11: 138-
ing growth of required materials (steel and foundations), 189.
also very expensive. Here, we have set LNS = 0.58. Again 3. Kimber A, Mitchell L. Performance evaluation
because of practical reasons, the rotation and tilt angles standards for large grid-connected PV systems in
have also been limited to ±60 and 75°, respectively. the United States and Germany. 20th European
Table I gives numerical results for 1/GCR = 5.5. All Photovoltaic Solar Energy Conference, 2005;
these limits make this tracking type rather different than the 3229-3232.
primary vertical, and explain the lower energy collection 4. Baumgartner FP, Büchel A, Bartholet R. Solar Wings:
for this 1/GCR valué. A more detailed comparison is out of A new Lightweight PV Tracking System. 23rd Euro-
the scope of this work. However, concerning back- pean Photovoltaic Solar Energy Conference, 2008;
tracking, a similar conclusión can be made: back-tracking 2790-2794.
is significantly more efficient on the primary axis than on 5. García M, Vera JA, Marroyo L, Lorenzo E, Pérez E.
the secondary one. Moreover, back-tracking on only the Solar-tracking PV plants in Navarra: A 10 MW assess-
primary axis is likely to be a more optimal solution. ment. Progress in Photovoltaics: Research and Appli-
cations 2009; 17: 337-346.
7. CONCLUSIONS 6. Lorenzo E, Macagnan MH. Considerations in the
design of one axis tracking PV systems. Progress in
Back-tracking geometry for single-axis and two-axis Photovoltaics: Research and Applications 1994; 2:
tracking has been described and ideal tracking and 45-55.
back-tracking strategies have been compared in terms of 7. Lorenzo E, Pérez M, Ezpeleta A, Acedo J. Design of
energy yield for the representative cases. Numerical valúes tracking photovoltaic systems with a single vertical
have been presented for only a particular location and for axis. Progress in Photovoltaics: Research and Appli-
grid connected PV plants. However, general considerations cations 2002; 10: 533-543.
of the geometry governing Sun tracking lead us to propose 8. Martínez-Moreno F, Muñoz J, Lorenzo E. Experimen-
the general validity of key results: that back-tracking is tal model to estimate shading losses on PV arrays.
more attractive for the single horizontal axis than for the Solar Energy Materials and Solar cells 2010; 94:
single vertical one and, concerning two axes, that back- 2298-2303.
tracking is more efficient on the primary axis than on the 9. Photovoltaic Geographical Information System
secondary one. (PVGIS): http://sunbird.jrc.it/pvgis/
Finally, because back-tracking avoids shade, its 10. Pánico D, Garvison P, Wenger H, Shugar D. Back-
implementation in the current state-of-the-art of tracking tracking: a novel strategy for tracking PV systems.
technology is attractive not only for energy gain reasons 22nd IEEE Photovoltaic Specialist Conference, 1991;
but also for reducing the risk of hot-spots. 668-673.
11. Lorenzo E, Maquedano C. Operational results of the
lOOkWp tracking PV plant at Toledo-PV project. 13th
ACKNOWLEDGEMENTS European Photovoltaic Solar Energy Conference,
1995; 2433-2436.
Reviewer 1 provided extremely lucid comments to the first
12. Muñoz J, Lorenzo E, Martinez-Moreno F, Marroyo
versión of this paper.
L, Garcia M. An investigation into hot-spots in
two large grid-connected PV plants. Progress in
REFERENCES Photovoltaics: Research and Applications 2008;
1. Narvarte L, Lorenzo E. Tracking and Ground Cover 16: 693-701.
Ratio. Progress in Photovoltaics: Research and Appli- 13. SolarOptimus tracking system. (www.conergy.
cations 2008; 16: 703-714. com).