ERNATIONAL JOURNAL OF RENEWABLE ENERGY RESEARCH

F.M. Lorilla and R. B. Barroca, Vol. 12, No. 2, June, 2022

Flexible Dynamic Sun Tracking System (STS)

Employing Machine Vision Control Approach

Franch Maverick A. Lorilla*‡ , Renyl B. Barroca **

*Department of Energy Systems Management, College of Technology, University of Science and Technology of Southern
Philippines, Cagayan de Oro, Philippines, 9000
**Ateneo de Davao University, Davao City, Davao del Sur, Philippines 8000
([email protected], [email protected])

‡ Corresponding Author; Franch Maverick Lorilla, 8000, Tel: +63829096288227,

[email protected]

Received: 05.04.2022 Accepted: 26.04.2022

Abstract- In this study, a machine vision control approach for a sun tracking system (STS) is designed, implemented, and
performance is evaluated. The aim is to dynamically track the sun's centroid with high flexibility under low irradiation
conditions due to weather conditions such as cloud cover. The STS is designed to work independently in the absence of a
manual setup of the location's spatiotemporal data. The prototype used a 180 o FOV high-resolution camera as the primary
sensor for accurate image processing and adaptive control technique to regulate electrical signals to the two servo motors (pan
and tilt). The NVIDIA Jetson AI-Computing Board is used for the autonomous deployment of the tracker. It was shown in the
measurement that the sun's centroid tracking accuracy of the proposed tracker for Az (γ) and Al (α) is 0.23 o and 0.66o,
respectively, with the Solar Position Algorithm (SPA) while 0.59 o and 0.65o, respectively with the commercial solar tracker,
STR-22G. The results graphically and statistically show that the prototype using machine vision can measure accurately and
has the same tracking performance as compared with the two established measurements. The STS application based on
machine vision control approach can meet the requirements for a dynamic and flexible control system for designed Parabolic
Dish Solar Concentrators.

Keywords Renewable Energy, Computer Vision, Machine Vision, Image processing, Solar Tracking System, Parabolic
Concentrated solar technology

1. Introduction

One of the emerging technologies which have exhibited While this technology is not applied in industry, a
significant advances in energy harvesting is Concentrated residential application holds a promise. The key reason for
Solar Power (CSP) [1]. By concentrating the sunrays in a PDSC’s potential is the excellent thermal-to-electricity
reduced area, CSP systems can attain very high temperatures performance ratio, which is almost the same as the demands
[2]. For all practical purposes, CSP using Parabolic Dish placed on a domestic-use appliance, and their relatively low
Solar Collectors (PDSC) displays a maximum of noise and pollution (Ferreira et al., 2015). Over the past
approximately 25% efficiency in converting solar energy to decade, its implementation is restricted to prototypes of 10 -
electricity [3], [4]. PDSC system is CSP comprised of a 100kW capacity[6].
paraboloid-shaped reflector, a support structure, a receiver,
The technology's maturity is yet to be explored since
and a STS. The entire sun irradiation that impacts the
the traditional option for an efficient but costly energy source
parabolic dish is reflected towards its focus, where the
is a Photovoltaic system. Therefore, finding another reliable,
receiver is placed. This energy concentration allows the
safe, and cheaper energy source to reach off-grid/remote
PDSCs to achieve temperatures as high as 1500 oC [2], [5].
ERNATIONAL JOURNAL OF RENEWABLE ENERGY RESEARCH
F.M. Lorilla and R. B. Barroca, Vol. 12, No. 2, June, 2022

communities is a critical challenge to address. Interestingly, Center is estimated by simply dividing both axes by 2. The
there has been an increase in the market for economically center of the camera should match the center coordinate of
sound solutions in the various stages of solar dish the Sun's image by continually moving the pan and tilt
concentrators [7]. motors until it stops.
To raise solar concentrators' performance and make Based on the literature, the study's main objective is to
their implementation viable, many scientists and engineers develop a machine vision control approach for the proposed
have studied solar energy systems' efficiency [8][9]. In CSP STS using a high-resolution 180o FOV camera for accurate
technologies, STS is an essential part of improving the centroid estimation and the adoption of dynamic and flexible
overall energy production. Experiments also show that algorithms in image processing. Apart from other studies, the
tracking one or two axes can generate 20-50% more tracking performance of the proposed tracker is validated
electricity than those with a fixed structure [10]. STS is using a commercial STS, the STR-22G [19], and data from
accounted for about a 20% reduction of the operational costs the National Renewable Energy Laboratory (NREL) using
for CSP plants [11]–[13]. the established Solar Position Algorithm (SPA) which has
the highest accuracy and error less than 0.1187695°[20],
However, some problems such as energy usage,
[21]
reliability, weather, and maintenance also emerge and limit
the solar tracker's benefits [14]. Many authors have explored 2. Methodology
the recent use of low-cost and high-quality cameras for sun 2.1. Systen Architure
tracking applications. Due to the mature technology of image Figure 1 illustrates the complete system architecture. It is
sensors and more computing power become more available comprised of three main components:
in smaller packages, it is increasingly running to artificial a) Data acquisition unit - setup into three parts: built-in
intelligence (AI) and machine learning (ML) as a control GPS module, gyroscope with accelerometer, and
approach for an autonomous STS. Good results could be image sensor.
obtained through new control methods focused on AI and
b) Control unit - composed of a computing board and
machine vision (MV), according to Carballo et al. (2019).
servo motor drivers
This should address critical factors for a solar collector, such
as block shadow and cloud detections. In their study, CSP c) Mechanical unit - dual-axis (azimuth-altitude)
using heliostat and parabolic trough were implemented. mechanism consisting of servo motors.
Thus, an opportunity to prove its functionality to PDSC
The system works as follows. The tracker operation
would be one of this study's main interests.
starts by acquiring spatiotemporal data (date, time, latitude,
Authors like Sohag et al. (2016) used either LDR or and longitude of the location) of the experimental site
web camera mode to acquire the solar position. Control through a built-in GPS module. The accelerometer with
aproach is based on the grayscale conversion and gyroscope (MPU-6050) will properly orient inital azimuth
segmentation techniques to detect and estimate Sun's and altitude position with respect to true north. The solar
centroid. It was also proposed that it can work with the PLC position’s data acquisition stage is completed through a 180o
to control hundreds of Photovoltaic solar panels angle view image sensor (Sony IMX219) with an operating
simultaneously through the device. Lee et al. (2013) used the temperature of -20°C to +60°C and the ability to detect 3280
Cassegrain telescope to enlarge Sun's image and reduce the x 2464 resolution colors for better tracing efficiency.
radiation that might cause noises on the captured image. The
control strategy is based on converting the full-color image
into a defined level of Hue, saturation, and light (HSL) that
will result in good image binarization. Once the image is
binarized, the Sobel Method is implemented to recognize
Sun's boundary and 3-point formula is used to estimate the
Sun's centroid correctly. The tracking algorithm is tested and
established by setting up a laboratory for a sun image
simulator and the STS with an image-based sun position
sensor and tracking controller. It was evaluated that the
proposed system can come up with 0.040o accuracy with a
window setting of 5 pixels. Also, Garcia-Gil & Ramirez,
(2019) used a fisheye camera with a 180 angle view to
generate a panoramic image where a digital estimate of Fig. 1. System Architecture
azimuth and elevation angles is reflected. Using
The computing and control unit used is NVIDIA Jetson
accelerometer with a gyroscope, true north is set to correct
Nano, a single computing board based on a quad core ARM
the orientation of the angles. The captured RGB image is
converted to grayscale, and 256 grays for segmentation Cortex A57 running at 1.43 GHZ and 4 GB low-power
where black is set to 0 and white is set to 256. Centroid is DDR4 and a 128-core GPU capable of performing 472
estimated once the brightest portion of the image is located.

686
ERNATIONAL JOURNAL OF RENEWABLE ENERGY RESEARCH
F.M. Lorilla and R. B. Barroca, Vol. 12, No. 2, June, 2022

Billon computations/operations per seconds while only the NVIDIA Jetson. The entire image processing and
consuming 5-10W of power. Retailing from 99 USD, this algorithm to determine and record Sun’s position in real-time
board is a machine learning powerhouse and suitable for the were implemented in Open-Source Computer Vision
study’s interest. The GPS, accelerator, gyroscope, and the (OpenCV) using Python programming language via NVIDIA
motor drivers for servo motors (pan-tilt) were connected to Jetson. Table 1 covers components used in the study.

Table 1. Summary of components used and their detailed specification

Component Specifications Details Image
Sony IMX219 Camera Pixel 8 Megapixels
image sensor Chip Sony IMX219
Assembly Technique SMT (ROSH)
Resolution 3280 × 2464
Pixel Size 1.12µm x1.12µm
Field-of-View 200o
Price 1,380 PHP ≈ 27 USD
NVIDIA Jetson Nano GPU 128-core Maxwell
Development Kit-B01 CPU Quad-core ARM A57 @ 1.43
GHz
Memory 4 GB 64-bit LPDDR4 25.6 GB/s
Power 5-10 W
Price 5,818.75 PHP ≈ 99 USD
MG90S MicroServo Operating Voltage 4.8 to 67
Operating Speed @6.0V No 0.10sec / 60 degrees
load
Price 563 PHP ≈ 11.26 USD

Operating Voltage 3-5 V

Frequency 40-1000 Hz
No. of Channels 16
16-Channel 12-bit
PWM Servo Shield I2C Resolution 12 Bits
Interface Price 250 ≈ 5 USD
Operating Voltage 3-5 V
Gyro range + 250, 500, 1000, 2000 °/s
MPU-6050, 3-Axis
Acceleration range ± 2, ± 4, ± 8, ± 16 g
Accelerometer & Gyro
Price: 2.5 USD
2.2 Machine Vision image that can exclude other objects and unavoidable noises
for more accurate tracking. Figure 2 presents the steps in
estimating the centroid of the captured Sun’s image.
a) Full-Color Image Capture. The image sensor
dynamically tracks the Sun’s image in a 180o angle
view, as shown in Figure 2(a) with the short line of
code. The image is captured in a 24-bit RGB color
element combination.

Fig. 2. Image Processing Steps

Machine vision is implemented using image processing
algorithms to process, analyze, and identify sun images and
their centroid. The goal is to obtain a pure and clear Sun
Fig. 3. HSV Dynamic tracker
687
ERNATIONAL JOURNAL OF RENEWABLE ENERGY RESEARCH
F.M. Lorilla and R. B. Barroca, Vol. 12, No. 2, June, 2022

b) HSV Color Range Thresholding. Before treating a the midpoint of the bounding box dimensions is the
binary image, thresholds were applied on the lowest and estimated centroid.
highest HSV or Hue value, Saturation value, and Value
space to properly segment the image for better object
detection, as shown in Figure 2(b) with the line of code 2.3 Control Strategy and Programming
used. At the start, it becomes tedious to try to figure out
Once the centroid of the Sun (h,k) is calculated, the
what exact values to apply; thus, GUI for track bars is
algorithm will match its coordinates to the centroid of the
provided as shown in Figure 3 to allow flexible
camera frame (x,y) by sending appropriate control
configuration of the low and high settings of the filter.
During the experimentation, the best values were set. commands to the drive motors (negative feedback). There are
Hue is set to 130 (low) and 179 (high). Saturation is set two main control mechanisms to understand: the Sun moves
in the E/W and rises and sets in the northern and southern
to 0 (low) and 30 (high). Value is set to 255 to both low
hemispheres (azimuthal motion and altitude motion in
and high value.
relation to the observer on the earth's surface). This process
Since Sun’s image is brightest in the HSV, the object is repeated as the image sensor actively captures Sun’s
that is not the brightest in the image can be separated. image. The results of the machine vision control aproach
Saturation can be adjusted in real-time to exaggerate the algorithm have been shown in Figure 5.
difference between non-sun images such as cloud
Once initial spatiotemporal data and proper orientation
formations. The net effect is to generate more
of the tracker is identified and set, the image sensor starts
identifications without increasing false positives.
scanning its 180o field of view for the Sun’s location. Once
c) Image Binarization. To extract the Sun's bright spot, Sun’s image is detected, its central coordinates are identified,
HSV conversion to black and white (binarization) is and the offset is calculated to match the camera frame's
applied using a threshold which creates object masking central coordinates. The calculated offset has corresponding
in an image using Otsu's method shown in Figure 2(c) pixel values where the motors are dynamically adjusted to
with the partial line of code. In this method, pixel correct such offset. The y-axis offset has a PWM signal that
values were analyzed to find the best balance between moves the servo motor in pan motion, which tracks the
the two classes divided into two by minimizing the altitude, while the x-axis offset commands the servo motor in
histogram's variance denoted. tilt motion, which tracks the azimuth.
The final binarized image will be cleaned further by
removing the noises other than the Sun and filling in the
image's holes. Figure 4 (a-b) shows different conditions
of the Sun's image and its corresponding binary image.

Fig. 5. Control Strategy Algorithm for dynamic-flexible

motor servo operation
3. Results and Discussions
Experimental setup of the prototype is implemented at
Ateneo de Davao, Bangkal, Davao City at the CSP research
site (GPS: 7.0606297,125.55674). Figure 6(a) presents the
close-up view of the prototype with the parts label. The
image sensor’s tracker arm is comprised of two servo motors
Fig. 4 Different image capture of the sky during that enable solar tracking in terms of altitude (α deg) and the
experimental setup and calibration on April 2, 2021 with azimuth (γ deg). The bearing is aligned in parallel with the
their corresponding binary output. (a) 12 Noon – no clouds; vertical servo motor on the center end of the tracker's arm to
(b) 2:00 PM – fewer clouds and (c) 5:00 PM – more clouds. ensure flexibility when rotated on the horizontal axis. The
camera that serves as the image sensor is placed at the end of
d) Sun detection and centroid estimation. It is important to
the tracker arm covered with a transparent acrylic sheet (3-
distinguish the boundary of the sun image for locating
mm thickness) and protective cover to filter ultraviolet and
the centroid of the Sun. After identifying the largest
infrared wavelengths that can damage the image sensor. The
contour, a bounding rectangle is initiated around the
solar tracker enclosure has been manufactured using the
white spaces to quickly identify the central coordinates
available light-yellow colored plexiglass (acrylic plastic) as it
Sun (h,k), as shown in Figure 3(d). The intersection of
absorbs less heat and protects electronic components when
subject to direct sunlight. The prototype is placed beside the
688
ERNATIONAL JOURNAL OF RENEWABLE ENERGY RESEARCH
F.M. Lorilla and R. B. Barroca, Vol. 12, No. 2, June, 2022

standard measuring device, STR-22G, from ECKO secondary reference data and as the third source of azimuth
instruments, as seen in Figure 6(b). This device provides and altitude.
high tracking reliability, enhanced functionality with a fully
automated setup procedure through a built-in GPS receiver.
This device will serve as the primary reference data for
validating the performance of the prototype.
The while prototype measures and record solar position
data in real-time, STR-22G is communicating serial data
every 1s via RS232c to the NVIDIA Jetson using terminal
commands from the manufacturer’s manual. The data
gathered from both sources were saved in CSV format. The
data gathering and experimentation were conducted on April
4, 2021, from 6:00 AM to 5:30 PM. Prior to the day, weather
forecasting is verified for a sunny day, and experimentation
setup and calibration were arranged using Accuweather. The
Fig. 6. (a) Developed prototype, (b) Experimental Setup of
proponents also added reference data from NREL as
the prototype with the STR-22G Solar

Table 2. Recorded values of Azimuth (γ) and Altitude (α) in csv format.
SPA Algorithm STR-22G Proposed Tracker
Time (UT) NREL (ADDU) (Machine Vision)
γ α γ α γ α
5.00 84.08 -0.3 83.83 0.46 83.86 -0.27
5.50 84.53 3.19 92.55 3.95 93.29 3.93
6.00 84.82 5.58 84.57 15.20 87.01 8.02
6.50 86.16 17.85 84.96 18.61 84.62 17.51
7.00 86.41 20.32 85.18 20.47 87.37 22.51
7.50 87.64 32.69 86.52 32.84 87.96 34.14
8.00 87.88 35.16 86.80 35.27 87.22 35.59
8.50 89.13 47.56 88.38 47.69 89.20 48.38
9.00 89.41 50.04 88.61 50.24 88.83 50.27
9.50 91.02 62.44 90.91 62.53 91.53 63.06
10.00 91.43 64.91 91.59 65.14 93.35 66.67
10.50 95.02 77.3 97.84 77.54 97.70 77.17
11.00 96.56 79.77 95.89 79.95 97.01 80.89
11.50 240.74 87.4 240.26 87.23 239.09 86.23
12.00 255.24 85.08 254.76 85.00 256.48 86.79
12.50 266.76 72.79 265.93 72.79 267.97 74.83
13.00 267.44 70.31 266.45 70.23 267.39 71.25
13.50 269.64 57.92 268.24 57.84 269.03 58.71
14.00 269.96 55.44 269.36 55.42 270.90 56.97
14.50 271.34 43.04 269.74 43.01 268.31 41.61
15.00 271.59 40.56 270.00 40.50 270.59 41.15
15.50 272.81 28.17 270.42 36.05 269.47 27.22
16.00 273.05 25.7 270.67 33.49 270.46 25.50
16.50 274.31 13.35 272.42 15.92 272.64 13.57
17.00 274.57 10.89 272.71 13.51 273.71 11.89
17.50 275.71 1.3 275.19 1.88 276.06 2.17

Table 2 shows the result of the recorded values of Figures 7 and 8 present the graphical comparison of
Azimuth (γ) and Altitude (α) with a specific time interval. the three (3) sources measuring Azimuth (γ) and Altitude
The validation of the solar position derived from the (α). The graph shows that the proposed system has almost
prototype was determined through graphical comparison the identical tracking measurement as the SPA and STR-
and statistical validation from three different source under 22G.
the same condition, such as illumination and location.

689
ERNATIONAL JOURNAL OF RENEWABLE ENERGY RESEARCH
F.M. Lorilla and R. B. Barroca, Vol. 12, No. 2, June, 2022

A one-way analysis of variance was also conducted to

know if there is a significant difference on Azimuth and
Elevation angles when grouped according to the sources.
The Azimuth and Elevation angles were divided into three
groups in terms of SPA, STR-22G, and the proposed
prototype. There was no statistically significant difference
as the p-value is above 0.01 on the satisfaction of the three
groups: F = 0.001, p = 0.999 (Az-γ) and F = 0.015, p =
0.985 (Al- α).
Figures 9 and 10 show that the proposed prototype has a
sun’s centroid measurement that could be tracked in terms
of Azimuth and Altitude with an accuracy of 0.23o and
Fig 7. Graphical comparison of the 3 Sources in measuring 0.66o as compared with the SPA while 0.59o and 0.65 o
Azimuth (γ) respectively with the STR-22G. The results statistically
show that the prototype using machine vision can work
accurately and has the same performance as the two
established measurements. With almost 150 USD total
worth of prototype, this low-cost device helps present an
accurate solar position measurement through a dynamic-
flexible control of its servo motors.
4. Conclusions

In this paper, the design, implementation, and test of

the proposed prototype are based on a machine vision
control approach for a flexible-dynamic control strategy.
Fig 8. Graphical comparison of the 3 Sources in measuring The proposed prototype used an AI-based computing
altitude (α) board designed for remote applications and is also
composed of low-cost hardware and electronics to make
the product commercially feasible in terms of price. The
prototype applied all and tested of the recommendations
and useful algorithms/ techniques derived from similar
solar tracker studies using machine vision and image
processing. One of these is the use of high-resolution
cameras and the effective use of the 180o FOV to better
track accuracy in finding the centroid. Different techniques
in image processing and tracking customized options were
adopted to make better analysis for dynamic and flexible
tracking. As shown in Figure 4(b-c), the tracker can still
dynamically track and locate the Sun’s image’s centroid
under cloudy conditions.
Fig. 9. Variation of Az (γ) angles of SPA and STR-22G, On the other hand, a simple algorithm is also designed
PROTO and STR-22G, and PROTO and SPA for motor control in work. Based on the experiment, the
proposed prototype can work competitively in tracking
accuracy with the commercially available solar trackers. In
our future research, we intend to integrate these prototypes
using Artificial Intelligence in determining the estimated
value of DNI while capturing the video feed. The use of
wireless control of the motors to be used in the actual CSP
site. Also, it is also recommended that measured data and
calculated data could be done on summer solstice, solar
equinox, and winter solstice.
Acknowledgements

The researchers would like to thank the experts who

were involved in the validation of this study. Without
Fig. 10. Variation of Al (α) angles of SPA and STR-22G, them, the researchers might not meet their objectives in
PROTO and STR-22G, and PROTO and SPA doing this study. The authors would also like to
acknowledge Dr. Eleonor V. Palconit, who also
contributed to this study’s success and the MREC-ADDU

690
ERNATIONAL JOURNAL OF RENEWABLE ENERGY RESEARCH
F.M. Lorilla and R. B. Barroca, Vol. 12, No. 2, June, 2022

for allowing us to access the information gathered by the sun-tracking methods for maximizing solar systems
STR-22G sun-tracker. output,” Renew. Sustain. Energy Rev., vol. 13, no. 8,
pp. 1800–1818, 2009, doi: 10.1016/j.rser.2009.01.022.
References
[12] P. Roth, A. Georgiev, and H. Boudinov, “Design
[1] REN21, Renewables 2019 Global Status Report
and construction of a system for sun-tracking,” Renew.
Collaborative, vol. 105, no. July. 2019.
Energy, vol. 29, no. 3, pp. 393–402, 2004, doi:
[2] J. M. L. Butti, S. Rivier, L. Delgado, S. S. Rivera, and https://doi.org/10.1016/S0960-1481(03)00196-4.
J. E. N. McLeod, “Sun-tracking system design for
[13] G. C. Bakos, “Design and construction of a two-
parabolic dish solar concentrator,” Lect. Notes Eng.
axis Sun tracking system for parabolic trough collector
Comput. Sci., vol. 2236, pp. 447–452, 2018.
(PTC) efficiency improvement,” Renew. Energy, vol.
[3] M. A. Soderstrand, S. B. Lee, and P. Chung, “Mini- 31, no. 15, pp. 2411–2421, 2006, doi:
dish based hybrid Concentrated Solar Power (CSP) https://doi.org/10.1016/j.renene.2005.11.008.
system for home use,” Midwest Symp. Circuits Syst.,
[14] C. C. Wei, Y. C. Song, C. C. Chang, and C. B.
pp. 689–692, 2013, doi:
Lin, “Design of a solar tracking system using the
10.1109/MWSCAS.2013.6674742.
brightest region in the sky image sensor,” Sensors
[4] Q. Chen and Y. Wang, “Research Status and (Switzerland), vol. 16, no. 12, pp. 1–11, 2016, doi:
Development Trend of Concentrating Solar Power,” in 10.3390/s16121995.
2020 9th International Conference on Renewable
[15] J. A. Carballo, J. Bonilla, M. Berenguel, J.
Energy Research and Application (ICRERA), 2020, pp.
Fernández-Reche, and G. García, “Machine learning
390–393, doi: 10.1109/ICRERA49962.2020.9242893.
for solar trackers,” AIP Conf. Proc., vol. 2126, no.
[5] A. Giovannelli, “State of the art on small-scale July, 2019, doi: 10.1063/1.5117524.
concentrated solar power plants,” Energy Procedia,
[16] H. A. Sohag, M. Hasan, M. Khatun, and M.
vol. 82, pp. 607–614, 2015, doi:
Ahmad, “An accurate and efficient solar tracking
10.1016/j.egypro.2015.12.008.
system using image processing and LDR sensor,” 2nd
[6] G. Simbolotti, “Concentrating Solar Power Technology Int. Conf. Electr. Inf. Commun. Technol. EICT 2015,
Brief,” IEA-ETSAP and IRENA, 2013. no. Eict, pp. 522–527, 2016, doi:
10.1109/EICT.2015.7392008.
[7] F. V. Barbosa, J. L. Afonso, F. B. Rodrigues, and J. C.
F. Teixeira, “Development of a solar concentrator with [17] C. D. Lee, H. C. Huang, and H. Y. Yeh, “The
tracking system,” Mech. Sci., vol. 7, no. 2, pp. 233– development of sun-tracking system using image
245, 2016, doi: 10.5194/ms-7-233-2016. processing,” Sensors (Switzerland), vol. 13, no. 5, pp.
5448–5459, 2013, doi: 10.3390/s130505448.
[8] I. Stamatescu, I. Făgărăşan, G. Stamatescu, N. Arghira,
and S. S. Iliescu, “Design and Implementation of a [18] G. Garcia-Gil and J. M. Ramirez, “Fish-eye
Solar-tracking Algorithm,” Procedia Eng., vol. 69, pp. camera and image processing for commanding a solar
500–507, 2014, doi: tracker,” Heliyon, vol. 5, no. 3, p. e01398, 2019, doi:
https://doi.org/10.1016/j.proeng.2014.03.018. 10.1016/j.heliyon.2019.e01398.
[9] A. Varghese, A. M. Vasanthakumary, J. Freeman, and [19] Ecko Instruments, STR-22G Sun Trackers. Tokyo,
K. Achuthan, “Remote triggered solar energy 2012.
assessment using a pyrheliometer and a pyranometer,”
[20] I. Reda and A. Andreas, “SPA: Solar Position
2017 6th Int. Conf. Renew. Energy Res. Appl. ICRERA
Algorithm,” Astrophysics Source Code Library. p.
2017, vol. 2017-January, pp. 115–120, 2017, doi:
ascl:1504.002, Apr. 01, 2015, [Online]. Available:
10.1109/ICRERA.2017.8191251.
https://ui.adsabs.harvard.edu/abs/2015ascl.soft04002R.
[10] K. E. N’Tsoukpoe, H. Liu, N. Le Pierrès, and L.
[21] K. B. de Melo, M. Kitayama, L. R. Tavares, H. A.
Luo, “A review on long-term sorption solar energy
de Freitas, and M. G. Villalva, “Accuracy Analysis of
storage,” Renew. Sustain. Energy Rev., vol. 13, no. 9,
Sun Position Calculation Algorithms: Ineichen and
pp. 2385–2396, 2009, doi:
SPA,” in 2019 IEEE PES Innovative Smart Grid
https://doi.org/10.1016/j.rser.2009.05.008.
Technologies Conference - Latin America (ISGT Latin
[11] H. Mousazadeh, A. Keyhani, A. Javadi, H. Mobli, America), 2019, pp. 1–5, doi: 10.1109/ISGT-
K. Abrinia, and A. Sharifi, “A review of principle and LA.2019.8895282.

691

View publication stats