Innovative Robot Design for Cleaning Solar Panels

Abderrahim Chellal

1,4 a

, Jos

e Lima

1,2 b

, Ana Isabel Pereira

1,5 c

and Paulo Costa

2,3 d

Research Centre of Digitalization and Intelligent Robotics CeDRI, Instituto Polit

ecnico de Braganc¸a, Braganc¸a, Portugal

Robotics and Intelligent Systems Research Group, INESC TEC, Porto, Portugal

Faculty of Engineering of the University of Porto, DEEC, Porto, Portugal

Higher School in Applied Sciences of Tlemcen, Tlemcen, Algeria

ALGORITMI Center, University of Minho, Braga, Portugal

Keywords:

Robotics, Automation and Mechatronics, Solar Energy, Data Analytics and Simulation, Digital Image

Processing, Efﬁciency Increase.

Abstract:

The accumulation of dust particles on the solar panels decrease the total amount of solar energy received by

the Photovoltaic panel (PV) and, it has also been proven that the increase of temperature reduces the overall

efﬁciency. These effects have been studied by different researchers based on collected data comparison. As

both of these parameters have a negative effect on the efﬁciency of the solar panel, it is essential to keep them

clean and at low temperatures. The use of technologies, such as robots, is an effective way of carrying out

repetitive tasks at low cost and in a short time. This work consists of developing a robot capable of cleaning and

cooling the solar panels, based on images acquired through a camera positioned directly towards the panels,

thus maintaining periodic cleaning in order to increase its efﬁciency. The preliminary results demonstrate the

possibility of using this approach for the accomplishment of this task.

1 INTRODUCTION

Renewable energy plays an important role in the en-

ergy sector by replacing limited resources such as

oil and coal with abundant natural resources. One

of the most popular of these renewables is solar en-

ergy. Solar energy is radiant energy emitted by the

sun and harnessed by photovoltaic panels. It offers

various advantages over other available renewable en-

ergy sources, but suffers from several drawbacks.

The photovoltaic module is a set of cells connected to

each other in series and/or parallel, which produces a

current and a voltage when exposed to the sun. (Cam-

pana et al., 2013). Environmental factors such as tem-

perature, wind and rain greatly inﬂuence the reliabil-

ity and efﬁciency of solar panels and represent one

of the major problems of this technology. It has also

been shown that the accumulation of dust particles

signiﬁcantly reduces the overall efﬁciency, especially

in desert areas, but not only.

https://orcid.org/0000-0002-9190-6865

https://orcid.org/0000-0001-7902-1207

https://orcid.org/0000-0003-3803-2043

https://orcid.org/0000-0002-4846-271X

During long periods without rain, such as the summer,

dust accumulation can lead to daily losses of over

20%. (Zorrilla-Casanova et al., 2011), these losses

lead to a substantial decrease in the overall efﬁciency

of photovoltaic systems.

Figure 1 shows dirty and clean solar panels.

Figure 1: Dirty and clean Solar panels (JKB Energy, 2018).

The effect of dust accumulation on solar cells

varies with location and has been studied by a number

of researchers. In (Christian et al., 2015) It is revealed

that a 20% drop in efﬁciency was observed on photo-

voltaic panels installed in Freiburg, Germany, within

5 months of heavy soiling of the glass surface. In

264

Chellal, A., Lima, J., Pereira, A. and Costa, P.

Innovative Robot Design for Cleaning Solar Panels.

DOI: 10.5220/0010540102640270

In Proceedings of the 11th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH 2021), pages 264-270

ISBN: 978-989-758-528-9

the Algerian Sahara desert, it was found that the efﬁ-

ciency of solar panels decreased by 32% after only 6

months without any cleaning operation. (Mostefaoui

et al., 2019). In these cases, regular cleaning of the

modules is necessary. It is strongly recommended

that, when weather conditions are not extreme, the

photovoltaic panels are cleaned 3 to 4 times a year,

and this number should increase during dry periods.

Usually this is done regularly by human operators, but

it is time-consuming and, more importantly, gener-

ates additional operating costs ranging from C1000 to

C4500/MW depending on the region. (Stridh, 2012).

The objective of this paper is to propose a method-

ology, based on simulation, that describes an efﬁcient

and economical way to perform this repetitive, time-

consuming and costly task by using robots in an inno-

vative way, through the use of a camera, giving the

robot autonomy and greater work efﬁciency. This

document is divided into seven sections. The ﬁrst

section presents the overview of the work. The sec-

ond section promotes the work already done in this

area, followed in section 3, by a brief description of

the overall system architecture. The environment and

the simulation of the model are presented in section 4

where the robot simulation software is described. The

ﬁfth section introduces in general the algorithms ap-

plied in the different software, while the sixth section

highlights the results obtained, and ﬁnally ends with

a conclusion and future work.

2 RELATED WORK

Some reported automatic systems were proposed in

the literature to solve the dust accumulation issue

through robots. In (Ronnaronglit and Maneerat,

2019) a freely moving robot is suggested: it can go

in any direction and is able to climb a maximum of

20° slope inclination. In (Anderson et al., 2010), a

two part rail guided robot is designed, it consist of a ”

Drive System” that moves horizontally along the top

and bottom edges of the solar panel, and a ” Cleaning

Head ” that is driven vertically by cables.

In (Aravind et al., 2014) the proposed system is

made up of two sub-system, a robot cleaner and an

automated carrier cart, while the robot clean the PV

panel surface, the carries moves along side.

In (Sundaram et al., 2019) The focus was on the

concept of a robot that could be installed in any solar

farm, easy to use and particularly easy to transport. A

market survey and patent analysis on the use of Solar

cleaning robots is presented in (Grando et al., 2019),

which offers a general view of different approach and

technologies involved and commercialized.

In (Hashim et al., 2019) The proposed strategy al-

lows both monitoring of solar irradiation and cleaning

of the photovoltaic surface when necessary through a

camera mounted directly on the robot, under the con-

trol of a mobile application that also collects the data.

Another cleaning strategy using a camera was also

proposed in (Li et al., 2016), where a mapping frame-

work was developed using a Kinect V2 as an RGB-D

sensor.

3 SYSTEM ARCHITECTURE

The system architecture is based on two components,

the SimTwo software, where the robot is fully sim-

ulated, and a processing application developed on

Lazarus Software, a cross-platform IDE for applica-

tion development. These two platforms communicate

with each other using the User Datagram Protocol

(UDP). Camera image is compressed on jpeg format

and sent by port 9808 from SimTwo to the application

whereas on the opposite way, the SimTwo will receive

commands on port 9898 to control the motors.

The use of SimTwo is justiﬁed by the need to per-

form the simulation of the robot while Lazarus is used

to develop an application capable of processing the

image acquired by the camera in the SimTwo soft-

ware and controlling the robot. The application con-

sists of the analysis of the image, the determination

of the location and edges of the solar panels and the

exact coordinates of the dust on them, this will be fur-

ther discussed afterwards.

Figure 2 shows the main architecture of the pro-

posed system.

Figure 2: System Architecture.

The same approach is to be used in a real scenario,

replacing the SimTwo simulation environment block

with the real robot and camera. In this case, the cam-

era will send data directly to the application, and the

robot will communicate with the application via WiFi

Innovative Robot Design for Cleaning Solar Panels

265

or another network, depending on the infrastructure.

4 ENVIRONMENT AND MODEL

SIMULATION

SimTwo is a realistic mobile robot simulation soft-

ware with a multiple window interface that can sup-

port several types of robots with dynamic capability

(Gonc¸alves et al., 2013) and it is useful for devel-

opment and testing purposes. It is strongly recom-

mended to create the robot model with dimensions,

positions and part masses very close to the real parts,

since the dynamics will be calculated based on these

characteristics and it is desired that it behaves like a

real situation. Figure 3 represents an overview of the

different interface of the software.

Figure 3: SimTwo software - General view.

In this section, it will be discussed how the devel-

opment of the scene environment on the software was

carried out, ﬁrstly by modelling the solar panels pre-

sented in the CeDRI facilities, and next, by modeling

the main parts of the cleaning robots, which are the

frame and the mobile robot. The robot’s look and be-

havior are deﬁned in several XML format ﬁles and are

implemented in the ”scene editor” window. The main

scene ﬁle (scene.xml) deﬁnes the different sub-robot

parts incorporated in the main robot and their speciﬁc

construction ﬁle. The robot parts are deﬁned by sev-

eral solids (cuboid and cylinder) connected through

joints and articulations (slider and hinge). These

joints are parts of the shell elements since they are

solids without mass and do not modify the robot’s

physical properties, but are an essential part of the

collision simulation (Costa et al., 2011). The move-

ment occurs through a guide rails, called drive frame,

ﬁxed in the edges of the solar panel, allowing a ﬂuid

movement of the mobile robot and giving it the possi-

bility of mounting a steeply sloping panel. The drive

frame should be manufactured using aluminium al-

loy, offering moderate strength, good formability and

is suitable for welding (Aravind et al., 2014). Figure

4 presents the designed model on the software.

Figure 4: SimTwo Software - Robot and Solar panels.

Different models have been designed, but the two-

axis model with rails is the best solution. Allowing

the robot to glide over the panel with small wheels

offers less resistance, reducing the battery capacity,

which signiﬁcantly reduces the weight and price. A

total of four position sensors are used to provide in-

formation about the position of both the frame and

robot. Two of them, continuously checks if the frame

reaches the edges of the solar panel and two others

check for the mobile robot if it reaches the upper and

lower limits.

A model was made using the FUSION 360 soft-

ware, allowing a better understanding of its shape.

The proposed parts, designed on this CAD software,

are presented in Figure 5.

Figure 5: Robot Prototype design.

Fusion360 is a software developed by the com-

pany AUTODESK, it is a very complete software,

particularly easy to use and efﬁcient to make com-

puter assisted drawings.

The four drive wheels must be assembled in pairs,

SIMULTECH 2021 - 11th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

266

Figure 6: Automatic mode ﬂowchart.

each pair of wheels must be connected to the other by

a chain that allows only one motor to be used, which

ensures that there will always be at least one drive

wheel in contact with the edge when crossing the gaps

between the panels. (Anderson et al., 2010).

5 CONTROL ALGORITHMS

The software offers the possibility of implementing

the algorithm in a script in Pascal language, capable

of managing the movement of the robot while per-

forming the cleaning. It also allows to choose be-

tween three modes: manual mode, partial automatic

mode and full automatic mode, as well as the direc-

tion of movement. These different modes will be de-

tailed in the next sub-sections.

5.1 Manual & Partial Automatic Mode

The manual mode is the simplest one where the robot

follows the commands transmitted by pressing the

corresponding button (right, left, up, down), a more

in-depth discussion of it, is therefore unnecessary.

The automatic mode repeats continuously the

same loop, as it can be seen in Figure 6, the robot

check its position ﬁrst and decide which way to

move, assuming that the horizontal movement of the

frame to the left is the ﬁrst step of the operation. The

second step of the cycle would be the robot moving

vertically upwards, the 180 mm diameter brush is

activated and moves in a circular motion using DC

motors. The frame then moves to the left, towards the

next cleaning area, the robot then starts the descent

until it reaches the lowest point. These loops are

repeated until the end of the solar panels array is

reached. When the cycle is ﬁnished, the robot stops

and wait for instructions.

5.2 Fully Automated Mode

The fully automated mode involves the analysis of

the panels by a front mounted camera. The acquired

image, is regularly sent with a format of 319 horizon-

tal pixels × 220 vertical pixels and is processed pixel

by pixel in the developed application. The application

ﬁrst determines the number of panels installed, their

boundaries and their respective positions, and then

highlights them with a red line.

Next, the algorithm checks the RGB (red, green,

blue) colour variation of each pixel included in the

solar panel and when a signiﬁcant amount changes,

the algorithm records it as a dust particle, highlights

it with a green line, identiﬁes the position of this

particle (X, Y and panel number) and assigns it a

speciﬁc number (starting from 1). Then the algorithm

continues to scan the rest of the panels for other dust

particles or any other type of particles such as urban

pollution, leaves, pollen, bird droppings. If another

one is found, the algorithm records its corresponding

coordinates and assigns it the next number, like a

Region Labelling algorithm. Figure 7 shows the

Innovative Robot Design for Cleaning Solar Panels

267

Figure 7: Application interface - Camera Picture.

application interface.

Figure 8 shows the processing algorithm ﬂowchart

implemented on the software as described previously.

Figure 8: Picture processing ﬂowchart.

The particle coordinates are sent to the SimTwo

software and the fully automated algorithm imple-

mented on the simulated robot moves the drive frame

to the corresponding ”X” position and then moves

the mobile robot upwards. The cleaning action is

simulated using a paint tool implemented on the

software that changes the colour of the dirt particles

into the original background colour. As it moves,

the mobile robot cleans the part of the solar panel it

is passing through, changing its colour. The mobile

robot then moves downwards while the driving frame

moves to the resting position in the resting panel,

waiting for further instructions. Figure 9 represents

the fully automatic mode ﬂowchart implemented in

the software.

The

position corresponds

with the panel

number

Convert the

equivalent position to

the real position

Dirt

particule position

Received

Stand-by

Yes

Move to position "X"

"X"

reached ?

Yes

Move to the upper

solar panel position

Position

reached ?

Move to the rest

position

Position

reached ?

Yes

Figure 9: Fully automated mode ﬂowchart.

It is possible to control and debug the algorithm

using the ”spreadsheet” window since this window is

fully customisable and allows the direct display of the

value of variables (sensors, motor states) and the ad-

dition of buttons (Stop and Automatic, for example).

The purpose of the software part of the project is to

control the simulated hardware of the system, includ-

ing the mechanical components. In general, in a real

case, small microcontrollers such as the well-known

ATMEGA328P are preferred in such a system, as the

control algorithm can easily be implemented on them

(Akyazi et al., 2019).

SIMULTECH 2021 - 11th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

268

6 RESULTS

The SimTwo software has a chart plotting window

which is used in this section, as it offers the possibil-

ity to plot the robot’s properties at any given time (for

example: position, speed, consumption and orienta-

tion). During cleaning, using the automatic mode, the

frame on the panel moves half the width of the panels

to the left (decreasing the X position) and stops. Sen-

sors on both sides of the frame signal the reach of the

panel’s edge, at which the robot stops. The mobile

robot moves up or down depending on its last posi-

tion, it is made of plastic and will contain the micro-

controller, battery to run the motors, electronics and

will be equipped with a brush.

The proposed partial automatic mode has been

tested for a large number of simulations. Figure 10

presents the graphical result of one of the simulations

performed, where the reference position (0,0,0) is lo-

cated in the middle of the second panel, the robot

starts in the rest position, located on the right with

X = 1.5 m. The blue line represents the robots move-

ments in the ”X” axis, while the red dotted line repre-

sents its position in the ”Y” axis.

Figure 10: Mobile robot movement graphic - Automatic

Mode.

The fully automated mode keeps the robot in

standby while the panels are clean, as soon as a dirt

particle is detected by the application, the drive frame

starts to move towards the ”X” coordinates of the

dust particle, once reached the mobile robot starts its

upward movement, the application is able to detect

dirt particles of any shape and size, as can be seen in

Figure 7.

The Figure 11 are shown the evolution of the ”X”

and ”Y” coordinates of the mobile robot, the speed of

the drive frame as it moves towards the ”X” position

and the speed of the mobile robot as it moves towards

the upper part of the solar panel.

The robot movement is controlled with the use

of a PID (Proportional Integral Derivative) controller

taking as reference the position.

Figure 11: Mobile robot movement graphic - Fully Auto-

mated Mode.

The proposed approach based on simulation al-

lows to validate the methodology both in image pro-

cessing and robot control.

7 CONCLUSION AND FUTURE

WORK

As the earth is in the midst of an energy transition, it

is essential to be able to provide an efﬁciency that is,

or tends to be, qualitatively equal to that of conven-

tional electricity generation. With this proposed de-

sign, the main concerns about dust accumulation on

PV panels can be effectively addressed, avoiding un-

necessary losses to these panels, which can be as high

as 20%.

This paper proposed a mobile robot that was able

to perform the assigned cleaning of dirt whatever is

the panel’s slope. It is important to the robot to move

independently of the slope since it vary with location.

Each mode developed is suitable for a speciﬁc case:

the automatic mode can be used after storm or heavy

soiling of the panels, while the full automatic mode

is used for a day by day maintenance. The results

performed on the software algorithm and on the sim-

ulated robot allowed to validate the approach.

Further development will optimize the system to

Innovative Robot Design for Cleaning Solar Panels

269

be smaller, lighter, more user-friendly and focus on its

implementation on the real hardware with the devel-

opment alongside of a digital twin. The experiments

with a real robot will address an RGB camera as well

as a thermal camera in order to optimize the cleaning

process with a sensor fusion algorithm. A docking

station for the robot is also equated to charge the bat-

teries of the robot and increase its autonomy.

ACKNOWLEDGEMENTS

This work is ﬁnanced by National Funds through

the Portuguese funding agency, FCT - Fundac¸

para a Ci

encia e a Tecnologia within project

UIDB/50014/2020.

REFERENCES

Akyazi,

O., S¸ahin, E.,

Ozsoy, T., and Alg

ul, M. (2019).

A solar panel cleaning robot design and application.

Avrupa Bilim ve Teknoloji Dergisi, pages 343–348.

Anderson, M., Grandy, A., Hastie, J., Sweezey, A., Ranky,

R., Mavroidis, C., and MARKOPOULOS, Y. P.

(2010). Robotic device for cleaning photovoltaic

panel arrays. In Mobile Robotics: Solutions and Chal-

lenges, pages 367–377. World Scientiﬁc.

Aravind, G., Vasan, G., Kumar, T. G., Balaji, R. N., and

Ilango, G. S. (2014). A control strategy for an au-

tonomous robotic vacuum cleaner for solar panels. In

2014 Texas Instruments India Educators’ Conference

(TIIEC), pages 53–61. IEEE.

Campana, P. E., Li, H., and Yan, J. (2013). Dynamic mod-

elling of a pv pumping system with special considera-

tion on water demand. Applied energy, 112:635–645.

Christian, S., Stefan, B., and Michael, K. (2015). Impact

of soiling on iv-curves and efﬁciency of pv-modules.

Solar Energy, 112:259–262.

Costa, P., Gonc¸alves, J., Lima, J., and Malheiros, P. (2011).

Simtwo realistic simulator: A tool for the develop-

ment and validation of robot software. Theory and

Applications of Mathematics & Computer Science,

1(1):17–33.

Gonc¸alves, J., Lima, J., Costa, P. J., and Moreira, A. P.

(2013). Modeling and simulation of the emg30 geared

motor with encoder resorting to simtwo: the ofﬁcial

robot@ factory simulator. In Advances in Sustainable

and Competitive Manufacturing Systems, pages 307–

314. Springer.

Grando, M. T., Maletz, E. R., Martins, D., Simas, H., and

Simoni, R. (2019). Robots for cleaning photovoltaic

panels: State of the art and future prospects. Revista

Tecnolog

ıa Y Ciencia, (35):137–150.

Hashim, N., Mohammed, M., Selvarajan, R. A., Al-

Zubaidi, S., and Mohammed, S. (2019). Study on so-

lar panel cleaning robot. In 2019 IEEE International

Conference on Automatic Control and Intelligent Sys-

tems (I2CACIS), pages 56–61. IEEE.

JKB Energy, . (2018). Get your system cleaned & running

properlys. https://www.jkbenergy.com/blog/2018/4/

30/get-your-system-cleaned-running-properly.

Li, M., Zhang, M., Fu, Y., Guo, W., Zhong, X., Wang,

X., and Chen, F. (2016). Fast and robust mapping

with low-cost kinect v2 for photovoltaic panel clean-

ing robot. In 2016 International Conference on Ad-

vanced Robotics and Mechatronics (ICARM), pages

95–100. IEEE.

Mostefaoui, M., Ziane, A., Bouraiou, A., and Kheliﬁ1, S.

(2019). Effect of sand dust accumulation on pho-

tovoltaic performance in the saharan environment:

southern algeria (adrar). Environmental Science and

Pollution Research, 26(1):259–268.

Ronnaronglit, N. and Maneerat, N. (2019). A cleaning robot

for solar panels. In 2019 5th International Confer-

ence on Engineering, Applied Sciences and Technol-

ogy (ICEAST), pages 1–4. IEEE.

Stridh, B. (2012). Evaluation of economical beneﬁt of

cleaning of soiling and snow in pv plants at three eu-

ropean locations. In 2012 38th ieee photovoltaic spe-

cialists conference, pages 001448–001451. IEEE.

Sundaram, M., Prabhakaran, S., Jishnu, T., and Sharma, S.

(2019). Design and analysis of an autonomous clean-

ing robot for large scale solar pv farms. In Proc. Int.

Conf. Autom. Control, Mechtron. Ind. Eng., page 265.

Zorrilla-Casanova, J., Piliougine, M., Carretero, J.,

Bernaola, P., Carpena, P., Mora-L

opez, L., and de Car-

dona, M. S. (2011). Analysis of dust losses in

photovoltaic modules. In World Renewable Energy

Congress-Sweden; 8-13 May; 2011; Sweden, number

057, pages 2985–2992.

SIMULTECH 2021 - 11th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

270