Solar Energy 148 (2017) 2535

Contents lists available at ScienceDirect

Solar Energy
journal homepage: www.elsevier.com/locate/solener

Energy gleaning for extracting additional energy and improving the

efficiency of 2-axis time-position tracking photovoltaic arrays under
variably cloudy skies
Stephanie White Quinn
Dept. of Electrical and Computer Engineering, Northeastern University, 409 Dana Research Center, 360 Huntington Ave., Boston, MA 02115, United States

a r t i c l e i n f o a b s t r a c t

Article history: Positioning a photovoltaic (PV) array in the optimal orientation increases the collection of solar radiation
Received 25 October 2016 and the production of electricity. Many methods for determining the optimal tilt angle of a fixed PV array
Received in revised form 16 January 2017 have been reported in the literature; however, few methods have been proposed for finding the optimal
Accepted 5 February 2017
tilt angle of a 2-axis time-position tracking PV array. This paper derives and validates a simple formula for
directly calculating the optimal tilt angle of a 2-axis time-position tracking PV array under varying sky
conditions. By modifying the conventional tilt angle as the sky conditions change, the tracking PV array
Keywords:
can glean the additional small amounts of irradiance that are overlooked and unused on cloudy days. The
Energy gleaning
Irradiance
validity of this formula was verified using 24 months of weather data from an installation in the north-
Photovoltaic systems eastern United States where clear skies occur about 39% of the time. Simulations indicated that modifying
Solar energy the conventional 2-axis tracking angles in response to changing cloud cover results in 2.3% increase in
2-axis tracking collected insolation and 2.4% increase in AC energy over a 24-month period. During hourly and sub-
hourly intervals with cloud cover, the increase in energy collection can reach up to 50%. The ability to
modify the conventional tracking angles in response to changing cloud cover allows the PV array to glean
the previously uncollected energy, thereby capturing more of the total available irradiance and increasing
electrical power production.
2017 Elsevier Ltd. All rights reserved.

1. Introduction with the irradiance collected by a fixed PV array that is positioned

1.1. Overview of problem and definition of energy gleaning
There are three types of photovoltaic (PV) installations fixed,
single-axis tracking, and 2-axis tracking. A review of several kinds
of tracking arrays is presented by (Mousazadeh et al., 2009). Com-
mercial 2-axis tracking arrays are divided into two categories:
trackers that use light sensors to find the brightest point in the
sky and trackers that use an astronomical algorithm based on the
suns apparent position.
Although more costly and complex than fixed PV arrays, the
advantage of 2-axis tracking arrays is the efficiency in which they
capture the solar radiation, especially on clear, sunny days. There is
an average of 34% more irradiance collected by a 2-axis tracking
array when compared with a fixed array in the same location
(Lubitz, 2011). For locations in the continental United States
(Lave and Kleissl, 2011) calculated a 2545% increase in irradiance
collected by a 2-axis time-position tracking PV array in comparison
E-mail address: quinn.st@husky.neu.edu
at optimal tilt and azimuth angles. They referred to (Kelly and
Gibson, 2009) in noting that a horizontal position may be more
optimal for a 2-axis tracking array when the amount of diffuse irra-
diance is great.
Although 2-axis trackers are typically installed in dry, sunny
locations with consistently clear skies, the number of installations
in cloudier locations has increased rapidly in recent years. The dis-
advantage of 2-axis tracking PV arrays that use an astronomical
algorithm is that they track the apparent position of the sun even
when it is obscured by clouds. On clear, sunny days, the total irra-
diation on the PV array is composed primarily of direct irradiation
from the sun. However, on cloudy days the direct irradiation is very
small and the total available irradiation is composed primarily of
diffuse irradiation that is scattered from the clouds and ground.
Several papers have reported that more irradiation could be col-
lected on cloudy days if the PV array did not track the sun
(Burduhos et al., 2015b; Gulin et al., 2013; Kelly and Gibson,
2009, 2011; Quesada et al., 2015; Quinn and Lehman, 2013).
Fig. 1 shows the irradiance measured on a cloudy day by two
sensors at a 2-axis time-position tracking installation located in

http://dx.doi.org/10.1016/j.solener.2017.02.005
0038-092X/ 2017 Elsevier Ltd. All rights reserved.
26 S.W. Quinn / Solar Energy 148 (2017) 2535
Fig. 1. 15-min measurements of Plane of Array (POA) irradiance and Global
Horizontal Irradiance (GHI) during an overcast day in Middlebury, Vermont
(October 29, 2014). More irradiance is measured by the GHI sensor that is fixed
horizontally (tilt angle of 0) than by the POA irradiance sensor that is mounted on
the 2-axis tracking PV array.
Vermont, USA. A sensor measuring Plane of Array (POA) irradiance
is mounted on one of the PV arrays tracking the sun. About 100 m
north of this POA sensor, a second sensor measuring Global Hori-
zontal Irradiance (GHI) is mounted at a tilt angle of zero degrees.
As seen in Fig. 1, the GHI measured by the horizontal sensor
exceeds the POA irradiance measured by the sensor that is tracking
the sun. This suggests that more irradiance could be captured by
adjusting the conventional 2-axis tracking tilt angle to a more hor-
izontal position. Each instance where the GHI exceeds the POA
irradiance represents an opportunity to collect more of the total
available irradiance and generate more electrical power.
The concept of collecting these additional small amounts of
solar energy during cloudy intervals is called energy gleaning.
Borrowed from the agricultural term for gathering the overlooked
crops that are left behind in the fields after the primary harvest,
energy gleaning collects the additional small amounts of power
that are overlooked and otherwise unused on cloudy days by
changing the orientation, in particular the tilt angle, of a 2-axis
tracking PV array. During hourly and sub-hourly intervals of vary-
ing cloud cover, the increase in collected irradiance can reach up to
50% (Kelly and Gibson, 2009). By capturing these small amounts of
additional power throughout the year, the annual energy produc-
tion of the PV installation could potentially increase by 0.53%,
depending on the installations location and climate (Quinn and
Lehman, 2013).
1.2. Definition of solar angles and conventional 2-axis tracking angles
Fig. 2 shows the solar geometry and angles used in modeling
irradiance. Two angles, the zenith angle hZ and the solar azimuth
angle cS, describe the apparent position of the sun. The zenith angle
is the angle between the zenith and the suns direct beam. The
solar altitude angle aS is the complement of the zenith angle. The
solar azimuth angle cS is the angle between the north-south axis
and the suns projection onto the horizontal plane.
The position of a tilted surface is defined by two angles, the tilt
angle b and the surface azimuth angle c. The tilt angle is the slope
of the surface with respect to the horizontal plane. The surface azi-
muth angle is the angle between the north-south axis and the
tilted surface. The angle of incidence h is the angle between a line
that is normal to the tilted surface and the incident light.
There are differences in the literature on the designation of the
starting point on the north-south axis. In this paper, the convention
in (Duffie and Beckman, 2006) is used where south is defined as 0,
north as 180 or
180, east
90, and west +90.
The conventional tracking angles for a 2-axis time-position
tracking array that is continuously tracking the sun under clear
skies (Braun and Mitchell, 1983; Duffie and Beckman, 2006) are:
Zenith
Normal to
z horizontal surface
S
North
West
S
East
South
Normal to tilted
Surface
Fig. 2. Solar geometry. aS is solar altitude angle, b is the tilt angle of the collector
with respect to the horizontal, c is surface azimuth angle, cS is solar azimuth angle, h
is angle of incidence, and hZ is zenith angle. All angles are in degrees. (Adapted from
(Duffie and Beckman, 2006)).
b hZ and c cS 1a; b
These angles assume that the only source of sunlight is the
direct beam from the sun. However, research published in the past
few years indicates that the conventional tilt angle of b = hZ is not
necessarily optimal under partially cloudy or overcast skies.
1.3. Review of methods for finding optimal tilt angles of 2-axis tracking
PV arrays
For fixed PV arrays, there have been many papers suggesting
methods for calculating the yearly or monthly optimal tilt angle
based on site parameters, weather conditions, cloudiness, or other
parameters (Calabro, 2013; Lave and Kleissl, 2011; Lubitz, 2011;
Quinn and Lehman, 2013; Yadav and Chandel, 2013). However,
for 2-axis time-position tracking PV arrays, only a few papers have
proposed methods for finding the optimal tilt angle under varying
sky conditions. These methods are briefly summarized in the fol-
lowing subsections.
1.3.1. Tracking advantage based on the ratio H/DTS
The concept of 2-axis tracking advantage (TA) was introduced
in (Kelly and Gibson, 2009). They defined the ratio H/DTS, where
H is the irradiance on a horizontal sensor and DTS is the irradiance
on a sensor pointed directly toward the sun. Six sensors mounted
on horizontal and tilted surfaces measured irradiance on four over-
cast days in Detroit, MI, where 50% of the days are cloudy. The H
sensors measured 2082% more irradiance throughout the day
than the DTS sensors, with a mean increase of 47%.
Additional experiments were conducted with four identical PV
arrays fixed at tilt angles of 0, latitude tilt, latitude
15 and lat-
itude + 15 with a LI-COR irradiance sensor mounted on each array
(Kelly and Gibson, 2011). The ratio H/DTS was calculated from irra-
diance data recorded for 30 min during solar noon, where H/
DTS = 1 represented the boundary between tracking and non-
tracking. There was a tracking advantage when H/DTS < 1, and
340 W/m2 was proposed as the threshold for determining whether
to track the sun or position the array horizontally. A 1% increase in
yearly irradiance for Detroit was estimated using this method. This
approach considers two components (direct irradiance and diffuse
irradiance) of the total available irradiance, but does not consider
the ground-reflected irradiance. It gives only two options for
 S.W. Quinn / Solar Energy 148 (2017) 2535
positioning the 2-axis tracking array: (i) place it in a horizontal
position when the sky is overcast, or (ii) track the suns path.
1.3.2. Tracking advantage based on the critical hourly global solar
radiation, IC
In (Quesada et al., 2015), the tracking advantage concept was
extended by using the Liu and Jordan isotropic irradiance model
(Duffie and Beckman, 2006) to include ground-reflected irradiance
in the total irradiance on a tilted surface, IT. The decision to track
the sun was based on the critical hourly global solar radiation, IC,
which was defined as the value of IH for which IT/IH = 1. Commer-
cial software calculated the electricity produced by a hypothetical
PV system connected to the grid. Under cloudy skies, the hourly
electrical energy produced by the horizontal PV panel was 18.8
46.9% greater than for the tracking panel on a winter day, and
2.324.7% greater on a summer day. The increase in electrical
energy was minimal (04.7%) on a cloudy winter day with high
ground reflectance due to snow.
In (Guillon et al., 2015) experiments were conducted using two
identical solar panels in Montreal, Canada one mounted horizon-
tally and the other mounted on a 2-axis tracker. During one cloudy
day, the measured short-circuit current, ISC, of the fixed horizontal
panel was 4.132.3% greater than the tracking PV panels ISC. On a
different cloudy day, the total electrical energy produced by the
fixed horizontal panel was 27.37% more than the tracking PV
panels energy.
The following algorithm was proposed in (Quesada et al., 2015):
Under clear skies or partially sunny/cloudy skies (IH > IC), the 2-axis
tracking array tracks the sun. When the sky is overcast and there is
no snow, the tracking panel is placed horizontally. When the sky is
overcast and snow is present, the panel is fixed in a south-facing
position at the locations optimal tilt angle. This approach assumes
that a tilt angle of 0 is the optimal solution when IT < IH. However,
simulations by (Gulin et al., 2013; Quinn and Lehman, 2013) indi-
cate that the optimal tilt angle depends upon sky conditions and is
not necessarily either hz or 0.
1.3.3. Computational methods using irradiance models and weather
data
In (Quinn and Lehman, 2013) the yearly POA irradiance was cal-
culated for a 2-axis tracking PV array in 21 U.S. locations in
MATLAB using Typical Meteorological Year 3 (TMY3) (Renewable
Resource Data Center, 2008) data. The hourly solar radiation on
the tilted surface was calculated as the tilt angle was varied from
0 to 90 in fixed steps of 1. The tilt angle corresponding to the lar-
gest POA irradiance was declared the optimal tilt angle for the
hour, and then the largest POA irradiance values for all hours of
the year were summed to obtain the annual POA irradiance.
Depending on the local climate, the yearly POA irradiance was
0.563.48% greater than the yearly POA irradiance calculated using
the conventional 2-axis tracking angles. For Detroit, the increase in
yearly POA irradiance was 1.88%, in comparison with 1% estimated
in (Kelly and Gibson, 2011). The increase in yearly POA irradiance
for Burlington, Vermont was 2.57%. Furthermore, the optimal tilt
angles for collecting maximum irradiance ranged between 0 and
hZ depending upon the sky conditions.
1.3.4. Differential evolution (DE) optimization
In (Gulin et al., 2013), differential evolution (DE) optimization
was used to generate 100 different daily trajectories for maximiz-
ing the electricity production of a 2-axis PV array under clear,
partly cloudy, and overcast skies. Weather data from Washington,
DC, and data on the energy consumed in positioning the array were
used to produce two types of trajectories a deterministic trajec-
tory and a stochastic trajectory based on the uncertainty in the
weather data and irradiance models. Optimal tilt and azimuth
27
angles were calculated at 15-min intervals. The daily trajectory
was set once before sunrise based on a day-ahead weather fore-
cast. Simulation results showed that on a clear day, the conven-
tional tracking angles (b = hZ and c = cS) were optimal for both
types of trajectories. On an overcast day, the optimal positions
were a tilt angle of 0 and an azimuth angle oriented due south
for the deterministic trajectory. However, for the stochastic trajec-
tory, the optimal tilt angles were found to be 2030 less than the
clear sky optimal angles (hZ). The azimuth angles were c = cS. On
the partly cloudy day, the irradiance profile was primarily diffuse
in the morning and direct in the afternoon. Throughout the day,
the optimal azimuth angle was cS for both types of trajectories.
During the afternoon (clear sky), the conventional tracking angles
(b = hZ) were optimal for both the deterministic and stochastic tra-
jectories. However, in the morning when the sky was cloudy, the
optimal tilt angle varied from about 15 to hZ for the deterministic
trajectory, and from 35 to hZ for the stochastic trajectory. The elec-
tricity produced by a 2-axis tracking array using the DE trajectory
was 0.37% greater than for an array using the conventional tracking
angles, and 1.27% greater than for an array that continuously
adjusted its tracking angles to maximize irradiance collection irre-
spective of the energy consumed by the tracker.
These results indicate that the optimal tilt angles depend on the
cloudiness of the sky and are not limited to the values of 0 and hZ.
This approach requires historical weather data to develop the daily
trajectories. Since the tracking arrays daily trajectory is set once
before sunrise, it may not be possible to alter the trajectory in
response to sudden changes in sky conditions.
1.3.5. Mathematical model of maximum global solar irradiance (MGSI)
In (Burduhos et al., 2012) the performance of a 2-axis tracking
PV module was compared with that of a south-facing PV module
fixed at the optimal tilt angle for three days (sunny, cloudy, and
partially cloudy) at a location in Romania. They reported that plac-
ing the tracking array in a horizontal position on cloudy days
resulted in increased irradiance collection and less power con-
sumption in moving the actuators.
A mathematical model for finding the direction of the maxi-
mum global solar irradiance (MGSI) was proposed in (Burduhos
et al., 2015b). The optimal elevation angle of the collector surface
was a function of the normal component of direct irradiance, the
normal component of diffuse irradiance, and the elevation angle
of the sun. The collectors azimuth angle was set equal to the solar
azimuth angle. Two tracking algorithms were simulated using
measured weather data for sunny, cloudy, and partially cloudy
days. The first algorithm changed angles every hour to track the
suns position. The second, adaptive algorithm, adjusted the track-
ing position in unequal steps toward the direction of the maximum
solar irradiance, Gmax. The adaptive algorithm showed an 8%
increase in captured irradiance on a cloudy day.
In (Burduhos et al., 2015a), they proposed using global horizon-
tal irradiance and diffuse irradiance sensors with the adaptive glo-
bal irradiance tracking algorithm to position the tracker in the
direction of the maximum solar irradiance. The decision to change
the tracking angle depended upon the measured clearness index
and the calculated threshold clearness indices for clear and over-
cast skies. The adaptive global irradiance tracking algorithm and
the sun-tracking algorithm were simulated using measured
weather data. On a cloudy day the adaptive algorithm collected
about 10% more solar irradiance and consumed less energy
because the tracking mechanism required fewer steps. On partially
cloudy and sunny days, both tracking algorithms collected about
the same amount of irradiance and required the same number of
steps.
The proposed adaptive algorithm indicates that the value of the
optimal tilt angle can vary between 0 and hZ depending upon the
28 S.W. Quinn / Solar Energy 148 (2017) 2535
direct and diffuse irradiances. However, it does not take into
account the ground-reflected irradiance, and it requires historical
weather data to establish the threshold values for clearness
indices.
1.3.6. Derived from mathematical model of solar radiation absorbed
on PV panels cell surface
In (Quinn and Lehman, 2013), a simple formula for estimating
the optimal tilt angle of a PV panel was derived by taking the
derivative of the equation for solar radiation absorbed on the cell
surface of a tilted photovoltaic panel (Duffie and Beckman, 2006),
and solving for b. The Parent S Formula is a function of latitude,
declination angle, hour angle, surface azimuth angle, ground reflec-
tance, transmittance-absorptance products of the diffuse and
reflected radiation streams, and clearness index. This formula
directly calculates the optimal tilt angle that maximizes irradiance
collection for a 2-axis tracking array for any surface orientation at
any time of the year. Simulations in MATLAB using TMY3 weather
data for 21 U.S. locations indicated that annual insolation increased
up to 2% for cloudier locations. The conventional tracking angles of
c = cS were used in the simulations.
1.4. Contribution and organization
The contribution of this research is a simple formula for calcu-
lating the optimal tilt angle of a 2-axis tracking PV array under
varying sky conditions. Placing the array in the optimal position
allows it to glean additional small amounts of solar energy
throughout the year. The proposed formula, called the S Formula,
has the following features:
 It takes into account all three components of irradiance (direct,
diffuse, and ground-reflected) because it is derived from the
equation for solar radiation absorbed on the cell surface of a
photovoltaic panel.
 It is a function of the conventional 2-axis tracking angle (hZ),
which is multiplied by an expression containing the clearness
index (g(kT) = Id/I), ground reflectance, and transmittance-
absorptance products of the direct, diffuse and ground-
reflected radiation streams.
 It helps in understanding the theory behind other researchers
simulations and experimental results; that is, the optimal tilt
angle is 0 during periods of overcast skies and hZ (zenith angle)
during periods of clear skies. It also shows why another angle
other than 0 or hZ may be optimal depending on the sky con-
ditions and ground reflectance.
 It calculates the optimal tilt angle during any interval of time.
 Its optimal tilt angles approximate those found using the com-
putational method when measured diffuse irradiance and Glo-
bal Horizontal Irradiance data are used.
The validity of the S Formula is investigated using two years of
measured weather data from a 2-axis tracking PV installation
located in Middlebury, Vermont, U.S. where clear skies occur about
39% of the year. The S Formulas optimal angles are compared with
the optimal tilt angles found by using the computational method
described in Section 1.3.3 and (Quinn and Lehman, 2013). The
yearly POA insolation and AC energy are calculated for three meth-
ods the S Formula, computational method, and conventional 2-
axis tracking.
This paper is organized as follows: Section 2.1 discusses the
installation parameters, instrumentation, and data quality and
screening. Section 2.2 explains the computational method for cal-
culating POA irradiance, AC power, and the optimal tilt angle. In
Section 2.3 the proposed S Formula is derived. Section 2.4 explains
the S Formula method for directly calculating the optimal tilt angle,
POA irradiance, and AC power. Section 3 presents the results of
these three methods (computational, S Formula, and conventional
2-axis tracking). Section 4 discusses and compares the results with
the simulations and experiments reported in the literature. Sec-
tion 5 presents conclusions and future work.
2. Methodology
2.1. Weather, irradiance and power data from a 2-axis tracking PV
installation
2.1.1. Site and location
This study used weather and power data from a PV installation
located at Middlebury College in Middlebury, Vermont, (latitude of
44.0N, longitude 73.2W). It is a 143 kW installation consisting of
34 AllEarth Renewables 2-axis trackers. Each tracking array holds
one string of twenty Evergreen ES-A 210 multicrystalline photo-
voltaic modules for a total DC capacity of 4200 W, and each array
is connected to an SMA America SB4000US 240V inverter. An
albedo (or ground reflectance) of 0.2 was assumed for the
simulations.
2.1.2. Data and instrumentation
Two 12-month periods of data were separately analyzed:
November 1, 2013 to October 31, 2014 and November 1, 2014 to
October 2015. About 15 days of data were missing from 2015.
The Draker Monitoring weather and power data for this location
are averaged and recorded at 15-min intervals.
Two sets of instrumentation are located at the PV installation.
One tracker has a Kipp & Zonen CMP11 pyranometer that measures
the Plane of Array (POA) irradiance and a Draker KL206 sensor
measuring the back-of-module temperature. This particular track-
ers DC current and DC voltage are also recorded.
Located about 109 m northwest of this particular tracker is a
second set of instrumentation. It consists of an Irradiance Inc.
RSR2 rotating shadowband radiometer with a LI-COR LI-200SA
irradiance sensor. According to the manufacturers specifications,
the LI-200SA has an error of 5% under daylight conditions and sta-
bility of less than 2% over a one-year period. The rotating shadow-
band radiometer has a shading ring and measures Global
Horizontal Irradiance (GHI), Diffuse Horizontal Irradiance (DHI),
and Direct Normal Irradiance (DNI). There is also a Campbell Scien-
tific CS215 ambient temperature and humidity sensor, and a Met
One 034B anemometer and wind vane.
A weather station at the Middlebury State Airport (Station ID
K6B0) records weather data and conditions about every 20 min.
It is located about 7.5 km southeast from the PV installation. The
reported weather conditions are clear, partly cloudy, mostly
cloudy, overcast, haze, fog, ice fog, thunder, light drizzle, moderate
drizzle, heavy drizzle, light rain, moderate rain, heavy rain, light
snow, moderate snow, heavy snow, light rain thunder shower,
heavy rain thunder shower, moderate thunder shower, and
unknown precipitation. These weather observations were used to
eliminate periods of precipitation from the original Draker data
set. About 38.07% of the daytime values were classified as clear
during November 1, 2013 to October 31, 2014, and about 39.82%
during the period of November 1, 2014 to October 31, 2015.
2.1.3. Data quality and screening
All data points occurring during nighttime intervals were elim-
inated first. About 25% of the daytime data points were removed
based on criteria suggested by (Li and Lam, 2000; Notton et al.,
2006) or for other reasons, as shown here:
 S.W. Quinn / Solar Energy 148 (2017) 2535 29

 Data occurred at sunrise and sunset, or when the zenith angle
was greater than 87.
 GHI was greater than the corresponding extraterrestrial irradi-
ance value.
 DNI was greater than the corresponding extraterrestrial irradi-
ance value.
 POA irradiance was less or equal to 0 (occasional negative val-
ues did occur).
 GHI was 25% more than the ideal Clear Sky GHI model as calcu-
lated in (Masters, 2004).
 DNI was 25% more than the ideal Clear Sky DNI model as calcu-
lated in (Masters, 2004).
 Data occurred during periods of precipitation as indicated by
the weather conditions recorded at Middlebury State Airport.
 AC power calculated in the PV_LIB Toolbox was less than 0 W.
2014 to October 31, 2015. The program was run for each 12-
month period.
Referring to Fig. 3, it was assumed that the surface continues to
track the sun from east to west (c = cs). During each time period,
the tilt angle was varied from 0 to 90 in one-degree increments
and the POA irradiance was calculated at each value of b using
the isotropic sky diffuse model (Duffie and Beckman, 2006; Liu
and Jordan, 1963). The optimal tilt angle boptimal was the angle at
which the maximum POA irradiance occurred during that particu-
lar time period. The value of maximum POA was then used to cal-
culate the AC power at that time.
POA irradiance and AC power at the conventional tracking
angles (b = hZ and c = cs) were also calculated in MATLAB and the
PV_LIB Toolbox with the same installation parameters and weather
data. This is not shown in Fig. 3.

Table 1 shows the weather conditions for the two 12-month 2.3. S Formula method for directly calculating the optimal tilt angle
data sets used in the simulations. Each data set represents a 15- under varying sky conditions
min interval. As can be seen in the table, clear skies occurred dur-
ing 43% of the intervals and overcast days about 31% of the time. In this section, the derivation of the proposed S formula is pre-
About 23% of the intervals were classified as either partly cloudy sented. The equations shown in (2)-(9) previously appeared in
or mostly cloudy. These intervals of varying cloud cover represent (Quinn and Lehman, 2013). Starting with (10), the derivation has
potential opportunities for energy gleaning. been revised to show the S Formula as a function of hz.
The solar radiation absorbed on the cell surface of a photo-
2.2. Computational method to find optimal tilt angles voltaic panel, IPV, is (Duffie and Beckman, 2006):



1 cos b 1
cos b
As mentioned in Section 1.3.3., one method for finding the IT;PV M sab Ib Rb sad Id sar Iqg
2 2
optimal tilt angle of a PV array is to use weather data and
2
irradiance models to compute the POA irradiance received by
the collecting surface as the tilt angle is varied from 0 to 90 where M is the air mass modifier, b is the tilt angle of the PV panel,
(Atmospheric Science Data Center, 2012; Lubitz, 2011). Fig. 3 Rb is the ratio of the direct radiation on a tilted plane to the direct
gives an overview of the process used in finding the optimal tilt radiation on a horizontal plane, qg is the ground reflectance, I is
angle. A program written in MATLAB and functions from the the irradiation on a horizontal surface, Ib is the direct irradiation
PV_LIB Toolbox were used in the simulations. The PV_LIB Toolbox on a horizontal surface, and Id is the diffuse irradiation on a horizon-
was developed for use in MATLab by Sandia National Laboratories tal surface. The symbols (sa)b, (sa)d, and (sa)r are the
(PV Performance Modeling Collaborative, 2012). The PVPMC transmittance-absorptance products of the direct, diffuse, and
website describes in detail the models that are the basis of each reflected radiation streams.
function. The value of Rb is calculated as:
The following parameters for the installation, previously given
in Section 2.1.1, were provided to the program: latitude, longitude, cos h
Rb 3
ground reflectance, PV module part number, inverter part number, cos hz
and the number of panels and strings. A soiling factor of 0.02 was
where the angle of incidence of the direct radiation on a surface, h,
assumed in calculating the AC power. The local timestamp was
is:
converted to solar time in the program. The 15-min weather data
(POA, GHI, DHI, DNI, and ambient temperature) that had been cos h b1 cos b b2 sin b b3 cos b b4 sin b b5 sin b 4
screened according to the criteria in Section 2.1.3 were input to
the program. There were 13,273 data points, each representing a and the zenith angle of the sun, hz, is calculated by:
15-min interval, for the period of November 1, 2013 to October cos hz cos d cos u cos x sin d sin u b3 b1 5
31, 2014, and 12,945 data points for the period of November 1,
with the coefficients b1, b2, b3, b4, and b5 defined as follows:

b1 sin d sin u 6a
Table 1
Description of the weather conditions for each data set used in simulations. Each data b2
sin d cos u cos c 6b
point represents a 15-min interval.

Weather Nov. 1, 2013 to Nov. 1, 2014 to b3 cos d cos u cos x 6c

observation Oct. 31, 2014 (13,273 Oct. 31, 2015 (12,945
data points) (%) data points) (%)
b4 cos d sin u cos c cos x 6d
Clear 42.53 44.35
Partly cloudy 12.61 12.34
Mostly cloudy 9.98 10.66 b5 cos d sin c sin x 6e
Overcast 32.54 30.65
Fog 0.99 1.24 where d is the declination angle, u is the latitude of the PV panel, c
Haze 0.32 0.43 is the azimuth angle of the PV panel, and x is the hour angle.
Thunder 0.20 0.32 Id = I  g(kT) is the hourly diffuse irradiation where g(kT) is
Thunder; fog 0.00 0.01
No reported observation 0.82 0.00
defined by correlations proposed by (de Miguel et al., 2001; Erbs
et al., 1982; Lam and Li, 1996; Mondol et al., 2008; Orgill and
30 S.W. Quinn / Solar Energy 148 (2017) 2535

Fig. 3. Finding boptimal by calculating POA irradiance as tilt angle is varied from 0 to 90. POAmax is the POA irradiance occurring at the optimal tilt angle, boptimal, and is used in
calculating the AC power in PV_LIB Toolbox.

Hollands, 1977; Reindl et al., 1990; Skartveit and Olseth, 1987) and The numerator in the first term of (10) is the angle of incidence of a
other researchers. vertical wall (b = 90) (Duffie and Beckman, 2006). This expression
kT is the hourly clearness index, and it is defined as the ratio of is equivalent to sin hZ cos (cS
c). Recall that there are two equa-
the hourly irradiation on a horizontal surface to the hourly tions for calculating the angle of incidence. One equation is shown
extraterrestrial radiation. as (4). The second is a function of the zenith angle, solar azimuth
angle, surface tilt angle, and surface azimuth angle (Duffie and
I
kT 7 Beckman, 2006):
Io
cos h cos hz cos b sin hz sin b coscs
c 11
Ib is the hourly direct (or beam) irradiation on a horizontal surface,
and it is calculated as: Setting (4) and (11) equal to each other and substituting b = 90
results in:
Ib I
Id 8

sin d cos u cos c cos d sin u cos c cos x cos d sin c sin x
Substituting (3)(8) into (2) results in the following expression:
  sin hz coscs
c 12
b1 b3 cos b b2 b4 b5 sin b
IT;PV M sab Io kT 1
gkT
cos hz Notice that the numerator of (10) is the same as the term on

 

1 cos b 1
cos b the left-hand side of (12). Substituting the term on the right-
sad Io kT gkT sar Io kT qg hand side of (12) into the numerator of (10), as well as substitut-
2 2
ing (5) into the denominator, yields a simpler and more meaning-
9
ful equation:
After taking the derivative of (9) with respect to b, setting it  " #
sinhz coscs
c 2sab 1
gkT
equal to zero, solving for tan b, substituting the constants shown tanboptimal
coshz 2sab 1
gkT sad gkT
sar qg
in (6a)(6e), and multiplying the numerator and denominator by " #
2(1
g(kT)), the following equation results: 2sab 1
gkT
  tanhz coscs
c

sindcos u cos c cosdsin u cos c cos x cosdsin c sin x 2sab 1
gkT sad gkT
sar qg
tanboptimal
sindsin u cosdcos u cos x 13
" #
2sab 1
gkT It is assumed that the 2-axis PV array will follow the sun using

2sab 1
gkT sad gkT
sar qg the conventional tracking angles of c = cS. Thus, the optimal tilt
10 angle can be found from:
 S.W. Quinn / Solar Energy 148 (2017) 2535 31

" #
2sab 1
gkT The following equations for Mean Bias Error, Mean Absolute
tan boptimal tan hz  Error, and Root Mean Square Error from (Zhang et al., 2013) were
2sab 1
gkT sad gkT
sar qg
used to analyze the errors between the optimal tilt angles found
14 through the computational method (Fig. 3) and the optimal tilt
As can be seen in (14), the optimal tilt angle is a function of the angles calculated using the S formula (Fig. 4),
zenith angle, the clearness index, the ground reflectance, and the r
1 XN
transmittance-absorptance products. Recalling from (1a) in Sec- MBE i1 i
y
xi 15
N
tion 1.2, the conventional 2-axis tilt angle is b = hZ, which can be
rewritten as tan b = tan hZ. Eq. (14) is the tangent of the conven- r
1 XN
tional 2-axis tilt angle multiplied by a function of g(kT), qg, and MAE jyi
xi j 16
(sa)b, (sa)d, and (sa)r. Since the clearness index, kT, is an indicator N i1

of sky conditions, this function modifies the conventional tilt angle r

as the sky conditions change.
Considering that Id = I  g(kT), and if measured data is available
for GHI and DHI, then the ratio DHI/GHI (that is, Id/I) may be sub-
stituted for g(kT) in (14).
2.4. Finding optimal tilt angles using the S Formula
Fig. 4 shows the process used to calculate POA irradiance and
AC power using the optimal tilt angles calculated by the S Formula
((14) in Section 2.3). It is assumed that the 2-axis tracking PV array
tracks the sun from east to west using the conventional 2-axis azi-
muth tracking angles (c = cS). The parameters and data sets
described in Sections 2.1.1 and 2.1.3, and used in Section 2.2 for
the computational method, were used in the calculations. A new
MATLAB program was written to directly calculate boptimal using
(14); however, the PV_LIB functions used by the computational
method were again used to calculate POA irradiance and AC power.
It was assumed that the transmittance-absorptance products were
ideal, so (sa)b, (sa)d, and (sa)r were set to 1. Since g(kT) = Id/I, the
measured values for Id and I were substituted in (14) to calculate
the optimal tilt angle.
1 XN
RMSE i1 i
y
xi 2 17
N
where yi = ith boptimal found using the computational method, xi = ith
boptimal calculated using the S formula, and N = number of data
points.
3. Results
This section presents the results of simulating three methods of
2-axis tracking (computational, S Formula, and conventional 2-axis
tracking) using measured weather data and the installation
parameters.
Fig. 5 displays the results of the computational method (Sec-
tion 2.2 and Fig. 3) for three days in 2014. The conventional 2-
axis tracking angles are shown for all daytime intervals, including
those intervals at sunrise and sunset when hZ may be greater than
87. On the overcast day (Fig. 5a), the optimal tilt angle is about 0.
On the clear day (Fig. 5b), the optimal tilt angle is near the conven-
tional 2-axis tracking angle of hZ, except around 06:30 when the
angle is near 0 during fog conditions. In Fig. 5c, a day of varying
sky conditions, the optimal tilt angle changes throughout the

Fig. 4. The S Formula calculates boptimal during each 15-min interval. boptimal is then used to calculate POA irradiance and AC power.
32 S.W. Quinn / Solar Energy 148 (2017) 2535
Fig. 5. Results of MATLab simulations showing the optimal tilt angles for three days
in 2014 (a) an overcast day (May 28th), (b) a clear day (August 19th), and (c) a day
with a mix of clouds and clear skies (August 27th). The red dotted trace represents
the optimal tilt angles calculated by the computational method where POA
irradiance is calculated as the tilt angle is varied from 0 to 90. Shown for
comparison are the conventional tilt angles of b = hZ (black solid line). Each step
represents a 15-min interval. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)
day, starting at about 0 before 08:00, varying between 25 and hZ
through 15:30, approaching 0, and finally ending at about hZ.
Fig. 6 shows the optimal tilt angles that were calculated using
the S Formula (Section 2.4 and Fig. 4) on the same three days.
The plots are similar to the ones shown in Fig. 5, with the calcu-
lated optimal tilt angle of about 0 on the overcast day (Fig. 6a),
near the conventional 2-axis tracking angle of hZ on the clear day
(Fig. 6b), and varying between 0 and hZ on the day with variable
weather (Fig. 6c).
The plots shown in Figs. 5 and 6 appear to be the same for both
the computational and S Formula methods. However, there were
sometimes small differences between the angles that are not visi-
ble in Figs. 5 and 6. To evaluate the errors between the two meth-
ods, the MBE, MAE, and RMSE were calculated (Table 2).
Fig. 7 shows the calculated POA irradiance values for a 2-axis
tracking array whose tilt is being adjusted using the S Formulas
optimal tilt angles. During most of the overcast day (Fig. 7a), more
POA irradiance is collected using the S Formulas optimal tilt
angles. On the clear day (Fig. 7b), the calculated POA irradiances
for both the S Formula and conventional methods are about the
same. On the day with variable weather (Fig. 7c), the S Formulas
tilt angles capture more of the POA irradiance during the periods
of overcast skies.
Fig. 6. Results of MATLab simulations showing the optimal tilt angles for three days
in 2014 (a) an overcast day (May 28th), (b) a clear day (August 19th), and (c) a day
with a mix of clouds and clear skies (August 27th). The blue dotted trace represents
the optimal tilt angles calculated by the proposed S Formula. Shown for comparison
are the conventional tilt angles of b = hZ (black solid line). Each step represents a 15-
min interval. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)
Table 2
MBE, MAE, and RMSE between the optimal tilt angles found using the S Formulas
method of directly calculating the optimal tilt angle, and the computational method
of varying the tilt angle to find the maximum POA irradiance.
Parameter Nov. 1, 2013 to Nov. 1, 2014 to
Oct. 31, 2014 Oct. 31, 2015
MBE, Mean Bias Error 0.0268 0.0311
MAE, Mean Absolute Error 0.2459 0.2464
RMSE, Root Mean Square Error 0.4579 0.4413
In reviewing the literature (Section 1.3), some researchers gave
a range of values for the percentage increase in POA irradiance
when the PV panel is placed horizontally during cloudy periods.
The histogram displayed in Fig. 8a shows the percentage increase
in POA irradiance when using the S formulas optimal tilt angles
versus the conventional tilt angles. Each column label indicates a
range of values. For example, the column labeled 15% means
10% < x%  15% where x% is the percentage increase in POA irra-
diance. In Year 1, 58.7% of overcast intervals showed an increase of
15% or more in POA irradiance using the S Formulas optimal tilt
angles. In Year 2, 55.5% of overcast intervals had a minimum
increase of 15%. Fig. 8b shows the net increase in POA irradiance
in W/m2 when using the S Formulas optimal tilt angles versus
the conventional tilt angles. The percentage of overcast intervals
 S.W. Quinn / Solar Energy 148 (2017) 2535
Fig. 7. Results of MATLab simulations showing the POA irradiance collected for
three days in 2014 (a) an overcast day (May 28th), (b) a clear day (August 19th),
and (c) a day with a mix of clouds and clear skies (August 27th). The blue dotted
trace represents the POA calculated by a hypothetical 2-axis tracking PV array using
the proposed S Formula to adjust the tilt angle. Shown for comparison is the POA
collected by the hypothetical array using conventional tilt angles of b = hZ (black
solid line). Each step represents a 15-min interval. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of
this article.)
with an increase in POA irradiance of 20 W/m2 or more was 46.2%
(Year 1) and 45.3% (Year 2).
Table 3 presents the yearly POA insolation and Table 4 lists the
yearly AC energy, as calculated in PV_LIB for three methods: con-
ventional 2-axis tracking, optimal angles found computationally,
and S Formulas optimal tilt angles. For the first year (Nov. 1,
2013 to Oct. 31, 2014), yearly POA insolation increased 2.35%
and yearly AC energy increased 2.42% when using the S Formulas
optimal angles. For the second year (Nov. 1, 2014 to Oct. 31, 2015),
yearly POA insolation increased 2.33% and AC energy increased
2.40% when using the S Formulas optimal tilt angles.
4. Discussion
As shown in (14) in Section 2.3, the tangent of the S Formulas
optimal tilt angle is equal to the tangent of the conventional 2-
axis tilt angle (hz), multiplied by a function of g(kT), qg, and (sa)b,
(sa)d, and (sa)r. Since the clearness index, kT, is an indicator of
sky conditions, this function modifies the conventional tilt angle
as the sky conditions change. For clear, sunny skies, the correlation
g(kT) approaches 0, thereby causing the modifying function (that is,
33
1000
# of 15- minute intervals
Year 1
800
Year 2
600
400
200
0
0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% 55% 60%
% increase in POA irradiance
(a)
800
# of 15- minute intervals
700 Year 1
600 Year 2
500
400
300
200
100
0
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Increase in POA irradiance (W/m2)
(b)
Fig. 8. Histograms showing the (a) percentage increase and (b) increase in W/m2 of
POA irradiance during overcast skies when optimal angles are calculated using the S
Formula versus using the conventional tilt angles. Number of overcast intervals is
shown for two 12-month periods Nov. 1, 2013 to Oct. 31, 2014 (Year 1) and Nov.
1, 2014 to Oct. 31, 2015 (Year 2).
the second term on the right-hand side of (14)) to approach 1. This
reduces to boptimal  hz, which is in agreement with the conven-
tional 2-axis tilt angle shown in (1a). In Fig. 6b, the S Formulas
optimal tilt angles are about the same as the conventional tilt angle
(boptimal  hz). The weather conditions for this day (August 19,
2014) were clear skies, except for a short period of fog occurring
at around 6:30 am.
For overcast skies the correlation g(kT) approaches 1, so the
numerator of the modifying function in (14) will approach 0. This
reduces to boptimal  0. Fig. 6a shows that the S Formulas optimal
tilt angles are about 0. On this day (May 28, 2014), there were
overcast skies starting at 6:30 am and continuing through the rest
of the day. The tilt angles calculated by the S Formula agree with
the recommendations of several researchers to place the PV mod-
ule in a horizontal position (bopt = 0) on overcast days (Burduhos
et al., 2012; Guillon et al., 2015; Kelly and Gibson, 2009, 2011;
Koussa et al., 2011; Quesada et al., 2015; Quinn and Lehman,
2013). In (Gulin et al., 2013) the PV modules optimal tilt angle
was 0 for the deterministic trajectory but 20 to 30 less than
the zenith angle for the stochastic trajectory.
On the overcast day (May 28, 2014), Fig. 7a shows that the S
Formulas POA irradiance exceeded the conventional trackings
POA irradiance during each 15-min interval. The increase in POA
irradiance ranged from 3.14% at midday to 52.49% before sunset.
On the same day (May 28, 2014) and 160 km north in Montreal,
Canada, the short circuit currents of two identical solar panels
were recorded under overcast skies (Guillon et al., 2015). One
panel was mounted at a tilt angle of 0 and the other mounted
on a 2-axis tracker. The short circuit currents were measured every
hour for 12 h. During overcast periods, the fixed horizontal panels
short-circuit current, ISC, was 4.132.3% greater than the tracking
PV panels ISC. These experimental results appear to support the S
Formulas simulation results for this particular day under overcast
skies.
34 S.W. Quinn / Solar Energy 148 (2017) 2535
Table 3
Yearly POA insolation calculated for three sets of tilt angles conventional 2-axis tracking, optimal angles found computationally, and S Formulas optimal tilt angles.
Method Nov. 1, 2013 to
Oct. 31, 2014
Insolation (kWh/m2)
Conventional 2-axis tracking 1616.82
Computational method 1654.83
S Formula method using measured DHI and GHI 1654.83
Table 4
Yearly AC energy calculated for three sets of tilt angles conventional 2-axis tracking, optimal angles found computationally, and S Formulas optimal tilt angles.
Method Nov. 1, 2013 to
Oct. 31, 2014
AC energy (kWh)
Conventional 2-axis tracking 6244.70
Computational method 6395.83
S Formula method using measured DHI and GHI 6395.87
In Fig. 6c, the S Formulas optimal tilt angles vary between 0
and hz. According to weather observations from the Middlebury
State Airport, the sky conditions changed throughout the day with
periods of mostly/partly cloudy, overcast, and clear skies. The trace
in Fig. 6c is similar to the traces in Fig. 6a and Fig. 6b; that is, the
optimal tilt angle is 0 during overcast skies and the optimal angle
is hz during periods of clear skies. However, as Fig. 6c shows, there
are also other angles neither 0 nor hzthat are optimal during
periods of varying cloud cover. This agrees with the simulations
of (Gulin et al., 2013) which showed that the optimal tilt angle var-
ied from about 15 to hZ for the deterministic trajectory, and from
35 to hZ for the stochastic trajectory.
The optimal tilt angles found using the computational method
(Section 2.2) were compared with the S Formulas angles. The
optimal tilt angles shown in Fig. 5 (computational method) and
Fig. 6 (S Formula) are similar. The low values of MBE, MAE, and
RMSE shown in Table 2 indicate that the S Formula is able to esti-
mate the computational methods optimal tilt angles. In this
study, measured DHI and GHI data from the installations sensors
were substituted into the S Formula; however, the errors may be
larger if using a correlation for g(kT). Identifying an appropriate
correlation for g(kT) to use in the S Formula is an area for future
study.
Fig. 8a shows the percentage increase in POA irradiance when
using the S Formulas optimal angles instead of the conventional
tracking angles. All 15-min intervals that occurred during overcast
skies throughout the year were distributed into bins of 5%. The lar-
gest number of overcast intervals had an increase in POA irradiance
of 5% or less. The remaining overcast intervals are somewhat
evenly distributed with increases in POA irradiance ranging from
10% to 50%. This range is similar to that reported in (Kelly and
Gibson, 2009) where the horizontal sensors measured 2082%
more irradiance throughout an overcast day than the sensors
pointed directly toward the sun.
Capturing these small amounts of increased POA irradiance
throughout the year potentially leads to an increase in the yearly
insolation and AC energy, as shown in Tables 3 and 4. The increase
in POA insolation was 2.35% for the first year and 2.33% for the sec-
ond year. These results are similar to the 2.57% increase in yearly
insolation that was calculated in (Quinn and Lehman, 2013) for
Burlington, Vermont, using the computational method, the Hay,
Davies, Klucher, and Reindl (HDKR) anisotropic sky model and
TMY3 weather data.
In this research, it was assumed that the ground reflectance was
0.2 throughout the year. An area for future study is how well the S
Formula estimates the optimal tilt angles at larger values of ground
Nov. 1, 2014 to
Oct. 31, 2015
Increase over conventional Insolation (kWh/m2) Increase over conventional
1570.50
2.35% 1607.02 2.33%
2.35% 1607.03 2.33%
Nov. 1, 2014 to
Oct. 31, 2015
Increase over conventional AC energy (kWh) Increase over conventional
6066.26
2.42% 6211.56 2.40%
2.42% 6211.60 2.40%
reflectance, such as when the ground is covered with snow. In
(Guillon et al., 2015), it was suggested when the sky is overcast
and snow is present, either falling or on the ground, that the 2-
axis tracking PV array should be fixed in a south-facing position
at the locations optimal tilt angle, rather than placed horizontally.
As mentioned earlier in this section, the S Formulas optimal tilt
angle is the tangent of the conventional 2-axis tilt angle multiplied
by a function of g(kT), qg, and (sa)b, (sa)d, and (sa)r. This multiply-
ing function modifies the conventional 2-axis tilt angle. A quick
calculation using the S formula may indicate why the horizontal
position is not optimal. Assuming that g(kT) = Id/I = 0.1 for an over-
cast sky, the multiplying term is 0.222 for qg = 0.2; however, the
multiplying term increases to 0.677 for qg = 0.8 for snow.
The goal of this research was to propose a simple, direct method
for calculating the optimal tilt angle under different sky conditions
at any given time, and to validate the method through simulations
using real weather data. For commercial 2-axis time-position
tracking arrays, the mechanisms for physically moving the arrays
and the intervals at which they change positions vary among man-
ufacturers. The S Formula allows different kinds of 2-axis tracking
arrays to be moved to the optimal position at the time increments
chosen by the designer or engineer. In a real-world implementa-
tion, data from two irradiance sensors, one measuring diffuse irra-
diance and the other measuring Global Horizontal Irradiance, may
be used in the S Formula to calculate the optimal tilt angle at the
time interval required by a particular 2-axis tracking array. After
showing the feasibility of the S Formula method through simula-
tions using 24 months of original weather data from a 2-axis track-
ing installation, we hope to arrange a research partnership and
conduct experiments in the near future.
5. Conclusions
This paper has proposed a simple formula for directly calculat-
ing the optimal position of a 2-axis time-position tracking PV array
under varying sky conditions. The S Formula is a function of hZ,
clearness index, ground reflectance, and transmittance-
absorptance products that mathematically explains the experi-
mental results reported in the literature. The optimal angles calcu-
lated by the S Formula reduce to the conventional tracking angles
b = hZ under clear skies and the tilt angle of b = 0 under overcast
skies. Under partially cloudy skies, the S Formula shows that there
exist intermediate values between 0 and hZ.
By adjusting the conventional 2-axis tilt angle to a more
optimal angle as the cloud cover changes, the PV array can glean
 S.W. Quinn / Solar Energy 148 (2017) 2535
additional small amounts of previously uncollected solar energy.
The collection of these small amounts of solar energy throughout
the year potentially increases the annual yield of the PV array.
Based on simulations using 24 months of weather data from an
installation located in Middlebury, Vermont, the results indicated
that adjusting the 2-axis tracking arrays tilt angle in response to
varying sky conditions allowed the array to capture more of the
available irradiance during 15-min intervals. For 45% of the inter-
vals with overcast skies, the minimum increase in POA irradiance
was 20 W/m2 using the S Formulas optimal angles versus the con-
ventional tilt angles. Using energy gleaning throughout the year,
the increases in insolation and AC power were 2.35% and 2.42%
for the first year, and 2.33% and 2.40% for the second year.
Future studies will focus on further validation of the S Formula
by running simulations using a third year of weather data from the
installation, and by conducting experiments with a 2-axis tracking
PV array that is adjusted using the proposed S Formula.
Funding
This research did not receive any specific grant from funding
agencies in the public, commercial, or not-for-profit sectors.
Conflict of interest
The author declared that there is no conflict of interest.
Acknowledgments
The author gratefully acknowledges Professor Rich Wolfson of
Middlebury College for providing access to the weather and power
data, and for his informative tour of the photovoltaic installation at
Middlebury College.
References
Atmospheric Science Data Center, 2012, Surface Meteorology and Solar Energy (SSE)
Release 6.0 Methodology, Version 3.1, March 1, 2012. <http://eosweb.
larc.nasa.gov/sse/documents/SSE6Methodology.pdf> (2016, Oct. 24).
Braun, J.E., Mitchell, J.C., 1983. Solar geometry for fixed and tracking surfaces. Sol.
Energy 31, 439444.
Burduhos, B., Diaconescu, D., Moraru, S., Badea, M., Grigorescu, C., 2012. Dual-axis
tracked vs. fixed PV: energetic response of one-year testing period in Romania.
In: 2012 13th International Conference on Optimization of Electrical and
Electronic Equipment, Brasov, Romania, pp. 979986.
Burduhos, B., Neagoe, M., Duta, A., 2015a. Adaptive stepwise orientation algorithm
for non-concentrated dual-axis solar tracking systems. In: Proceedings of the
14th IFToMM World Congress, Taipei, Taiwan, pp. 551558.
Burduhos, B.G., Visa, I., Neagoe, M., Badea, M., 2015b. Modeling and optimization of
the global solar irradiance collecting efficiency. Int. J. Green Energy 12, 743
755.
Calabro, E., 2013. An algorithm to determine the optimum tilt angle of a solar panel
from global horizontal solar radiation. J. Renew. Energy 2013, 112.
35
de Miguel, A., Bilbao, J., Aguiar, R., Kambezidis, H., Negro, E., 2001. Diffuse solar
irradiation model evaluation in the North Mediterranean belt area. Sol. Energy
70, 143153.
Duffie, J.A., Beckman, W.A., 2006. Solar Engineering of Thermal Processes. John
Wiley & Sons Inc, Hoboken, NJ.
Erbs, D.G., Klein, S.A., Duffie, J.A., 1982. Estimation of the diffuse radiation fraction
for hourly, daily and monthly-average global radiation. Sol. Energy 28, 293302.
Guillon, L., Quesada, G., Rousse, D.R., 2015. An experimental validation of the
concept critical solar radiation for solar tracking systems. Int. J. Appl. Sci.
Technol. 5, 111.
Gulin, M., Vaak, M., Peric, N., 2013. Dynamical optimal positioning of a
photovoltaic panel in all weather conditions. Appl. Energy 108, 429438.
Kelly, N.A., Gibson, T.L., 2009. Improved photovoltaic energy output for cloudy
conditions with a solar tracking system. Sol. Energy 83, 20922102.
Kelly, N.A., Gibson, T.L., 2011. Increasing the solar photovoltaic energy capture on
sunny and cloudy days. Sol. Energy 85, 111125.
Koussa, M., Cheknane, A., Hadji, S., Haddadi, M., Noureddine, S., 2011. Measured and
modelled improvement in solar energy yield from flat plate photovoltaic
systems utilizing different tracking systems and under a range of
environmental conditions. Appl. Energy 88, 17561771.
Lam, J.C., Li, D.H.W., 1996. Correlation between global solar radiation and its direct
and diffuse components. Build. Environ. 31, 527536.
Lave, M., Kleissl, J., 2011. Optimum fixed orientations and benefits of tracking for
capturing solar radiation in the continental United States. Renew. Energy 36,
11451152.
Li, D.H.W., Lam, J.C., 2000. Evaluation of slope irradiance and illuminance models
against measured Hong Kong data. Build. Environ. 35, 501509.
Liu, B.Y.H., Jordan, R.C., 1963. The long-term average performance of flat-plate solar
energy collectors. Sol. Energy 7, 5374.
Lubitz, W.D., 2011. Effect of manual tilt adjustments on incident irradiance on fixed
and tracking solar panels. Appl. Energy 88, 17101719.
Masters, G.M., 2004. Renewable and Efficient Electric Power Systems. John Wiley &
Sons, Hoboken, NJ.
Mondol, J.D., Yohanis, Y.G., Norton, B., 2008. Solar radiation modelling for the
simulation of photovoltaic systems. Renew. Energy 33, 11091120.
Mousazadeh, H., Keyhani, A., Javadi, A., Mobli, H., Abrinia, K., Sharifi, A., 2009. A
review of principle and sun-tracking methods for maximizing solar systems
output. Renew. Sustain. Energy Rev. 13, 18001818.
Notton, G., Poggi, P., Cristofari, C., 2006. Predicting hourly solar irradiations on
inclined surfaces based on the horizontal measurements: performances of the
association of well-known mathematical models. Energy Convers. Manage. 47,
18161829.
Orgill, J.F., Hollands, K.G.T., 1977. Correlation equation for hourly diffuse radiation
on a horizontal surface. Sol. Energy 19, 357359.
PV Performance Modeling Collaborative, 2012. PV_LIB Toolbox for MATLAB.
<https://pvpmc.sandia.gov/applications/pv_lib-toolbox/> (24 Oct 2016).
Quesada, G., Guillon, L., Rousse, D.R., Mehrtash, M., Dutil, Y., Paradis, P.L., 2015.
Tracking strategy for photovoltaic solar systems in high latitudes. Energy
Convers. Manage. 103, 147156.
Quinn, S.W., Lehman, B., 2013. A simple formula for estimating the optimum tilt
angles of photovoltaic panels. In: Control and Modeling for Power Electronics
(COMPEL), 2013 IEEE 14th Workshop on, Salt Lake City, Utah, pp. 18.
Reindl, D.T., Beckman, W.A., Duffie, J.A., 1990. Diffuse fraction correlations. Sol.
Energy 45, 17.
Renewable Resource Data Center, 2008, N.R.E.L., U.S. Department of Energy,
National Radiation Data Base 19912005 Update: Typical Meteorological Year
3. <http://rredc.nrel.gov/solar/old_data/nsrdb/1991-2005/tmy3/by_state_and_
city.html> (24 Oct 2016).
Skartveit, A., Olseth, J.A., 1987. A model for the diffuse fraction of hourly global
radiation. Sol. Energy 38, 271274.
Yadav, A.K., Chandel, S.S., 2013. Tilt angle optimization to maximize incident solar
radiation: a review. Renew. Sustain. Energy Rev. 23, 503513.
Zhang, J., Hodge, B., Florita, A., Lu, S., Hamann, H., Banunarayanan, V., 2013. Metrics
for evaluating the accuracy of solar power forecasting. In: 3rd International
Workshop on Integration of Solar Power into Power Systems, London, pp. 18.