Design and Analysis of Cooling Methods for Solar Panels

by Adam M. Palumbo

Submitted in Partial Fulllment of the Requirements for the Degree of Master of Science in Engineering in the Mechanical Engineering Program

YOUNGSTOWN STATE UNIVERSITY December, 2013

Design and Analysis of Cooling Methods for Solar Panels Adam M. Palumbo

I hereby release this thesis to the public. I understand that this thesis will be made available from the OhioLINK ETD Center and the Maag Library Circulation Desk for public access. I also authorize the University or other individuals to make copies of this thesis as needed for scholarly research. Signature:

Adam M. Palumbo, Student

Date

Approvals:

Dr. Ganesh Kudav, Thesis Advisor

Date

Dr. Hazel Marie, Committee Member

Date

Dr. Douglas M. Price, Committee Member

Date

Dr. Salvatore A. Sanders, Associate Dean, Graduate Studies Date

Abstract As the future progresses, many companies and industries are striving to achieve a greener approach to energy production by using solar energy. Solar panels that use PV cells (semiconductor devices used to convert light into electrical energy) are popular for converting solar power into electricity. One of the problems in using PV cells to extract energy from sunlight is the temperature eect on PV cells. As the solar panel is heated, the conversion eciency of light to electrical energy is diminished. Because solar panels can be expensive, it is important to be able to extract as much energy as possible. This thesis proposes cooling methods for the panel in order to achieve optimum eciency. To achieve this, various cooling methods have been proposed. A bare solar panel with no air velocity was used as a base model. This was tested and compared to bare solar panels cooled by heat sinks, in the form of extended surfaces such as plate ns, that can be mounted on the back surface of solar panels. These heat sinks were also tested for still air and dierent air velocities. Analytical calculations were also performed for the case of a bare panel with natural convection. Finally, computational models were made in ANSYS to obtain results that were compared with the experimental and analytical results. Other methods are discussed including using a pump to cool the panels using water or a coolant. Results showed that the heat sinks were only marginally eective; they resulted in a steady-state temperature of only a few degrees less than a solar panel without a heat sink. Due to these results, it is proposed that pump cooling would be far more benecial. With the correctly sized pump, the temperature can be made to closely match any desired value. The results are presented in the following thesis.

iii

Acknowledgements
Thesis Advisor Dr. Ganesh Kudav Thesis Committee Dr. Price, Dr. Marie Corporate Partners Q4 Manufacturing Solutions, Youngstown, OH Faculty Dr. Ganesh Kudav, Dr. Hazel Marie, Dr. Hyun W. Kim, Dr. Darrel Wallace, Dr. Daniel Suchora Additional Assistance Richard Laughlin, YSU Electrical Maintenance Dept. Maureen Adams, Department of Mechanical & Industrial Engineering Secretary Bob Schmitt and John Dodson, YSU Machine Shop Students Mike Kovach, Nick Matune, John Terzak

iv

Dedication
First and foremost, I would like to dedicate this thesis to my Grandma Mullett for working to instill knowledge in me starting at an extremely young age. You helped pique my curiosity and educate my mind when it was at its most fertile state for expansion. I have you to thank for all my academic achievements. I would like to thank Dr. Kudav and the rest of the Mechanical Engineering Program for giving me the great opportunity to work on a thesis project during my graduate studies at Youngstown State University. The amount of knowledge and experience I have gained during my thesis has been immeasurable. During the last two years, there have been countless people who have helped me with my project every step of the way. I would like to thank Dr. Kudav, Dr. Marie, Dr. Kim, Dr. Wallace, and Richard Laughlin from Electrical Maintenance for the equipment loaned and advice given on countless occasions. Maureen Adams was also a huge help in the various orders she placed for equipment I needed. I would also like to extend my gratitude towards my friends and classmates, Mike Kovach, Nick Matune, and John Terzak, as well as Bob Schmitt and John Dodson from the machine shop, for the many hours spent assisting me in assembling my experimental setup. I would also like to thank Dr. Suchora for the never-ending wisdom he has imparted on me during my college experience. Lastly, I would like to thank Youngstown State University and the Cushwa family for presenting me with the amazing opportunities I have received through the Cushwa

Fellowship position. Without all of you, none of this would have been possible. Thank you all!!

vi

Contents
Page 1 Introduction to Solar Energy 2 Photovoltaic Cells 2.1 2.2 2.3 2.4 2.5 General Information . . . . . . . . . . . . . . . . . . . . . . . . . . . Temperature Eect . . . . . . . . . . . . . . . . . . . . . . . . . . . . Increasing Eciency . . . . . . . . . . . . . . . . . . . . . . . . . . . Current-Voltage Characteristic of PV Cells . . . . . . . . . . . . . . . Scope of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 5 5 5 7 8 12 13 13 14 16 17 20 20 21 22 23 24 24 24

3 Analytical Calculations 3.1 3.2 3.3 3.4 3.5 Mathematical Model . . . . . . . . . . . . . . . . . . . . . . . . . . . Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . Analytical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Atmospheric Conditions . . . . . . . . . . . . . . . . . . . . . . . . . Experimental Calculations . . . . . . . . . . . . . . . . . . . . . . . . 3.5.1 3.5.2 3.5.3 3.5.4 External Forced Convection Over a Flat Plate . . . . . . . . . Natural Convection Over Flat Plate . . . . . . . . . . . . . . . Forced Convection Across Fins . . . . . . . . . . . . . . . . .

Natural Convection on Fins . . . . . . . . . . . . . . . . . . .

4 Computational Analysis 4.1 ANSYS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1 ANSYS Mechanical . . . . . . . . . . . . . . . . . . . . . . . .

vii

4.1.2 4.2 4.3 4.4 4.5

ANSYS Fluent . . . . . . . . . . . . . . . . . . . . . . . . . .

25 25 29 33 39 41 41 43 43 44 45 45 46 47 47 48 48 50 54 57 57 59 59 59 60 61

Bare Panel, Natural Convection, 2D . . . . . . . . . . . . . . . . . . . Bare Panel, Forced Convection, 3D . . . . . . . . . . . . . . . . . . . Fins, Natural Convection, 3D . . . . . . . . . . . . . . . . . . . . . . Fins, Forced Convection, 3D . . . . . . . . . . . . . . . . . . . . . . .

5 Experimental Analysis 5.1 5.2 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.1 5.2.2 5.2.3 5.2.4 5.2.5 5.2.6 5.2.7 5.2.8 5.2.9 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Solar Module . . . . . . . . . . . . . . . . . . . . . . . . . . . Pyranometer . . . . . . . . . . . . . . . . . . . . . . . . . . . Data Acquisition Unit . . . . . . . . . . . . . . . . . . . . . . Source Measurement Unit . . . . . . . . . . . . . . . . . . . . SMU Chassis . . . . . . . . . . . . . . . . . . . . . . . . . . . Laptops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Analog Output Module . . . . . . . . . . . . . . . . . . . . . . Light Array Control Board . . . . . . . . . . . . . . . . . . . .

5.2.10 Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.11 Custom Box . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.12 Ventilation Meter and Fan . . . . . . . . . . . . . . . . . . . . 5.2.13 Fins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 5.3.2 5.3.3 5.3.4 General Method . . . . . . . . . . . . . . . . . . . . . . . . . . Phase 1: Bare Panel with Natural Convection . . . . . . . . . Phase 2: Bare Panel with Forced Convection . . . . . . . . . . Phase 3: Panel with Fins . . . . . . . . . . . . . . . . . . . . .

viii

6 Results 6.1 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1.1 6.2 6.3 6.4 Fin Performance . . . . . . . . . . . . . . . . . . . . . . . . .

63 64 69 70 71 71 74 78 79 83

Analytical Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . Computational Results Vs. Experimental . . . . . . . . . . . . . . . . Explanation of Results . . . . . . . . . . . . . . . . . . . . . . . . . .

7 Proposed Cooling Method 8 Conclusion Bibliography Appendix A Screenshots

ix

List of Figures
Page Figure 1.1 Figure 1.2 Schematic Showing Solar Thermal Power [Solar Thermal Energy] Schematic Showing Photovoltaic Power Source [How Photo2 2

voltaic Solar Cells Work] . . . . . . . . . . . . . . . . . . . . . . . . . Figure 1.3 Schematic Showing Concentrating Solar Power (CSP) [A Power

Tower Power Plant] . . . . . . . . . . . . . . . . . . . . . . . . . . . . Figure 1.4 Figure 2.1 Agua Caliente Solar Project . . . . . . . . . . . . . . . . . . . Variation of Eciency as a Function of Module Temperature

3 4

[Nishioka et al., 2011] . . . . . . . . . . . . . . . . . . . . . . . . . . . Figure 2.2 Variation of Power as a Function of Cell Temperature [Due . . . . . . . . . . . . . . . . . . . . . . . . . . .

and Beckman, 2006] Figure 2.3 Figure 2.4 Figure 2.5 Figure 3.1 Figure 3.2 Figure 4.1 Figure 4.2 Figure 4.3

7 8 10 11 15 17 27 28

Equivalent Circuit of a Solar Cell [Ishibashi et al., 2012] . . . . I-V and P-V Curves [Desilvestro, 2008] . . . . . . . . . . . . . Temperature Eect on I-V Curves . . . . . . . . . . . . . . . . Boundary Conditions on Solar Panel . . . . . . . . . . . . . . Map Showing Solar Irradiation in the U.S. [Roberts, 2008] . . Bare Panel Model Created in ANSYS . . . . . . . . . . . . . . Meshed 2D Model . . . . . . . . . . . . . . . . . . . . . . . . . Temperature Distribution Throughout Bare Solar Panel with

W Natural Convection, I = 1000 m 2 . . . . . . . . . . . . . . . . . . . . .

28 30

Figure 4.4

Bare Panel Model Created in Solidworks . . . . . . . . . . . .

Figure 4.5 Figure 4.6 Figure 4.7 Figure 4.8 Figure 4.9

Closeup of Bare Panel Model Created in Solidworks . . . . . . Meshed Bare Panel Model . . . . . . . . . . . . . . . . . . . .

30 31

W m Temperature Distribution for Bare Panel, I = 1000 m 33 2 , V = 2.54 s

Model of Solar Panel with Heat Sink 1 . . . . . . . . . . . . . Model of Solar Panel with Heat Sink 2 . . . . . . . . . . . . .

34 35 36 36

Figure 4.10 Mesh Shown on Sink 1 Model . . . . . . . . . . . . . . . . . . Figure 4.11 Mesh Shown on Sink 2 Model . . . . . . . . . . . . . . . . . . Figure 4.12 Temperature Distribution Throughout Solar Panel with Sink 1 - I = 1000
W , m2

V=0

m s

. . . . . . . . . . . . . . . . . . . . . . . . . .

38

Figure 4.13 Temperature Distribution Throughout Solar Panel with Sink 2 - I = 1000
W , m2

V=0

m s

. . . . . . . . . . . . . . . . . . . . . . . . . .

38

Figure 4.14 Temperature Distribution Throughout Solar Panel with Sink 1 - I = 1000
W , m2

V = 2.54

m s

. . . . . . . . . . . . . . . . . . . . . . . .

40

Figure 4.15 Temperature Distribution Throughout Solar Panel with Sink 2 - I = 1000 Figure 5.1 Figure 5.2 Figure 5.3 Figure 5.4 Figure 5.5 Figure 5.6 Figure 5.7 Figure 5.8 Figure 5.9
W , m2

V = 2.54

m s

. . . . . . . . . . . . . . . . . . . . . . . .

40 42 42 43 44 45 46 46 47 48 49 49 50

Basic Experimental Setup . . . . . . . . . . . . . . . . . . . . Dimensions of Heat Sinks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Line Diagram of Major Components

Suntech STBKb 10 Watt Solar Module . . . . . . . . . . . . . Kipp & Zonen Pyranometer . . . . . . . . . . . . . . . . . . . NI 9219 DAQ . . . . . . . . . . . . . . . . . . . . . . . . . . . NI cDAQ 9172 Chassis . . . . . . . . . . . . . . . . . . . . . . NI PXIe-1073 5-Slot 3U PXI Express Chassis . . . . . . . . . NI 9264 Analog Output Module . . . . . . . . . . . . . . . . .

Figure 5.10 Paralax BASIC Stamp HoweWork Board . . . . . . . . . . . . Figure 5.11 Servo Connected to Dimmer Switch . . . . . . . . . . . . . . . Figure 5.12 Gold Solidstate Relay . . . . . . . . . . . . . . . . . . . . . . . xi

Figure 5.13 Front Panel of VI . . . . . . . . . . . . . . . . . . . . . . . . . Figure 5.14 Block Diagram of VI . . . . . . . . . . . . . . . . . . . . . . . Figure 5.15 Navigation Window . . . . . . . . . . . . . . . . . . . . . . . . Figure 5.16 Pyranometer VI . . . . . . . . . . . . . . . . . . . . . . . . . . Figure 5.17 Custom Box Stand . . . . . . . . . . . . . . . . . . . . . . . . Figure 5.18 Custom Box Top . . . . . . . . . . . . . . . . . . . . . . . . . Figure 5.19 Fully Assembled Box . . . . . . . . . . . . . . . . . . . . . . . Figure 5.20 Lights and Light Measurement . . . . . . . . . . . . . . . . . . Figure 5.21 Holmes Electric Fan . . . . . . . . . . . . . . . . . . . . . . .

51 52 53 54 55 55 56 56 57 57 58 58 60 61 62 64

Figure 5.22 VelociCalc Plus 8386 Ventilation Meter . . . . . . . . . . . . . Figure 5.23 Heat Sink 1, Smaller Heat Sink . . . . . . . . . . . . . . . . . Figure 5.24 Heat Sink 2, Larger Heat Sink . . . . . . . . . . . . . . . . . . Figure 5.25 Testing Performed with Still Air on Bare Panel . . . . . . . . Figure 5.26 Testing Performed with Wind on Bare Panel . . . . . . . . . . Figure 5.27 Testing Performed on Heat Sink Panel . . . . . . . . . . . . . Figure 6.1 Figure 6.2 Section of Master Spreadsheet . . . . . . . . . . . . . . . . . . Correlation of Power Output and Surface Temperature for Bare

Panel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Figure 6.3 Graph Comparing the Performance of the Heat Sink Panels vs.

66

W Bare Panel at Each Velocity for I = 1000 m . . . . . . . . . . . . . . 2

69

Figure 6.4

Clustered Column Graph Comparing the Results from Three 70 75 75 84 85

Methods for the Bare Panel with Natural Convection . . . . . . . . . Figure 7.1 Figure 7.2 Pump Cooling Schematic . . . . . . . . . . . . . . . . . . . . . Fluid Temperature . . . . . . . . . . . . . . . . . . . . . . . .

Figure A.1 Isometric of Sink 1 in Solidworks . . . . . . . . . . . . . . . . Figure A.2 Isometric of Sink 2 in Solidworks . . . . . . . . . . . . . . . . xii

Figure A.3 Light Control Program in BASIC Stamp . . . . . . . . . . . .

86

W m 87 Figure A.4 Bare Panel Starting at Room Temperature, I = 1000 m 2,V = 0 s W m Figure A.5 Bare Panel at Steady State, I = 1000 m 2 , V = 2.54 s . . . . . . W m Figure A.6 Bare Panel at Steady State, I = 1000 m 2 , V = 1.37 s . . . . . . W m Figure A.7 Bare Panel at Steady State, I = 1000 m 2 , V = 1.12 s . . . . . . W m . . . . . . . Figure A.8 Bare Panel at Steady State, I = 1000 m 2,V = 0 s

88 89 90 91

xiii

List of Tables
Page Table 3.1 Table 4.1 Table 4.2 Table 4.3 Table 4.4 Table 4.5 Table 6.1 Table 6.2 Table 6.3 Table 6.4 Table 6.5 Table 6.6 Table 6.7 Table 6.8 Table 6.9 Temperatures for Bare Panel with Natural Convection . . . . . Conductivity and Thickness of Solar Panel Layers . . . . . . . Bare Panel, Natural Convection Computational Data . . . . . . Bare Panel, Forced Convection Computational Data . . . . . . Fins, Natural Convection Computational Data . . . . . . . . . Fins, Forced Convection Computational Data . . . . . . . . . .
W . . . . . . . . . . . . . . Raw Experimental Data - I = 1000 m 2 W . . . . . . . . . . . . . . Raw Experimental Data - I = 1250 m 2 W . . . . . . . . . . . . . . Raw Experimental Data - I = 1500 m 2

16 26 29 33 37 39 65 65 65 67 67

Heat Transfer Coecients for Bare Panel, Forced Convection . Heat Transfer Coecients for Bare Panel, Natural Convection .

Heat Transfer Coecients for Heat Sink Panels, Forced Convection 68 Heat Transfer Coecients for Heat Sink Panels, Natural Convection 68 Comparison of Results for Bare Panel, Natural Convection . . . Comparison of Computational and Experimental Results For 71 70

Bare Panel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 6.10 Comparison of Computational and Experimental Results For Sink 1 71 Table 6.11 Comparison of Computational and Experimental Results For Sink 2 71 Table 6.12 Experimental Results Compared to Computational Results When Irradiance Decreased in ANSYS . . . . . . . . . . . . . . . . . . . . . 72

xiv

Nomenclature
h us N sol Te Ti Thermal diusivity Average convection heat transfer coecient Nusselt number for nned surface Coecient of volume expansion Solar altitude Exit temperature dierence, Ts Te Inlet temperature dierence, Ts Ti

Tlm Log mean temperature dierence m Q q o Declination angle of sun Mass ow rate Heat transfer rate Heat ux into panel - irradiance Emmisivity Panel eciency Kinematic viscosity xv

 A As B C cp Gd GR Gt

W Stefan-Boltzmann constant, = 5.67 x 108 m2 K 4

Apparant solar irradiation at air mass equal to zero Surface area Atmospheric extinction coecient Ratio of diuse irradiation on a horizontal surface to direct normal irradiation Specic heat capacity Direct radiation Reected radiation incident on surface Total irradiance on surface

GN D Normal direction irradiation h h1 h2 hrad hx keq L l Lc Lf in Hour angle Convection heat transfer coecient at top surface Convection heat transfer coecient at bottom surface Radiation heat transfer coecient Local convection heat transfer coecient Equivalent thermal conductivity for multiple layers Thickness of solar panel Latitiude Characteristic length Length of channel xvi

Lpanel Length of panel n Combination parameter for Nusselt number

N T U Number of transfer units N ub N ux p Pss Pr R RaL Ras Re b ReL Sf in T t Nusselt number for ow between plates Local Nusselt number Perimeter Power at steady-state Prandtl number Thermal resistance Rayleigh number Rayleigh number for nned surface Reynolds number for channel ow Reynolds number Spacing between ns Temperature Time

T (0) Temperature at top surface of panel T (L) Temperature at bottom surface of panel T 1 T 2 Ambient air temperature at top of panel Ambient air temperature at bottom of panel xvii

Te

Exit temperature

Tf ilm Film temperature Ti Tss Inlet temperature Steady-state temperature

Tsurr Temperature of sourroundings Ts V Temperature of radiating surface Velocity of air

xviii

Chapter 1 Introduction to Solar Energy

Renewable energy has become a large focus for many scientists and engineers in recent years, due to the growing concern of environmental pollutants given o by the burning of non-renewable fossil fuels. The use of solar energy is seen as one of the cleanest and most widely available alternatives to current sources of energy. There are many options available for collecting solar power. This is due to two sources of solar energy; thermal and photovoltaic [Shahsavar et al., 2011]. Solar energy is obtained by collecting the energy that the sun gives o as radiation. In general, thermal systems rely on the energy created by the radiation to heat up a uid. Photovoltaics, on the other hand, use the light given o by the sun to convert to electricity. There also exists a special method of collecting thermal energy known as Concentrating Solar Power (CSP), which involves using an optical device to concentrate the incoming radiation to a smaller area on a collector in order to minimize heat losses [Due and Beckman, 2006]. The higher temperatures created through CSP can be used to boil water and create steam. This in turn drives a turbine which converts the power into electricity. Examples of solar thermal power, photovoltaics, and CSP are shown in Figures 1.1, 1.2, and 1.3, respectively.

Figure 1.1: Schematic Showing Solar Thermal Power [Solar Thermal Energy]

Figure 1.2: Schematic Showing Photovoltaic Power Source [How Photovoltaic Solar Cells Work]

Figure 1.3: Schematic Showing Concentrating Solar Power (CSP) [A Power Tower Power Plant] Solar engineers have faced multiple problems for each method. For instance, there can be heat losses in thermal systems and only a percentage of potential light is actually converted to electricity in photovoltaics. Solar power can also be expensive in many ways. While the energy source itself is free, the systems used to collect and store this energy may not be economically feasible in some cases. For this reason, it is crucial for solar engineers to discover new and innovative ways to increase the eciency of solar units while simultaneously redesigning them in order to reduce the manufacturing and maintenance costs. There are two main methods of utilizing the suns energy: passive and active [Ogueke et al., 2009]. The passive method harnesses solar energy without the use of mechanical device. This can be as simple as placing large windows in a house to let sunlight in. The active method, however, uses mechanical systems such as solar panels and solar collectors. This thesis focuses primarily on active methods; more specically, through the use of photovoltaics. Many large scale solar energy systems exist in the United States. Currently, the photovoltaic system with the largest output is called the Agua Caliente Solar Project, created by NRG Energy. Array of solar panels is mounted on the ground and consist 3

of 5,200,000 modules manufactured by First Solar. Upon completion in 2014, the entire system is expected to generate 290MW of power. This is enough to provide power for approximately 100,000 homes [Agua Caliente Solar Project]. Figure 1.4 illustrates how large-scale the project is.

Figure 1.4: Agua Caliente Solar Project

Chapter 2 Photovoltaic Cells

2.1 General Information

Photovoltaic (PV) solar cells are typically made of crystalline silicon (Si). Circular PV cells are embedded into a plastic sheet in order to make a panel. It is known that the performance of PV cells is aected by irradiance, module temperature, and solar spectrum distribution [Fukishige, 2009]. This is due to the energy being converted from light energy in the photons to electrical energy. At the molecular level, the electrons become excited due to a high amount of thermal energy. This ends up dominating the electrical properties of the semi-conductor used in the solar panel. The temperature aects the voltage and current in the PV generator. While atmospheric temperature tends to have a small eect on the cells operating temperature, the solar irradiance and wind speed both have a large eect [Skoplaki et al., 2008].

2.2

Temperature Eect

PV cells are aected by temperature in a negative way due to the negative temperature coecient of crystalline silicon. This temperature coecient is estimated to be in the range of -0.4 to -0.5%/K [Lee et al., 2008]. This causes the eciency of the cells 5

(which are typically around 12% using a reference temperature of 25 ) to decrease. It has been shown that a decrease in the temperature of the solar panel can cause a 2% increase in the eciency of the PV cells, when the carrier uid (such as water) has a mass ow rate of 0.01 kg/s [Joshi and Tiwari, 2007]. An increase in eciency is vital to the development of solar panels, as it would save money while providing greater amounts of energy. PV cells have the potential to be extremely ecient. Even in the worst-case scenario, a PV cell has an eciency limit of 28.9% [Tayebjee et al., 2010]. Figure 2.1 shows the linear relationship of eciency versus the module temperature. As can be seen, the conversion eciency drops as the module temperature increases.

Figure 2.1: Variation of Eciency as a Function of Module Temperature [Nishioka et al., 2011] Since the radiation absorbed by the solar panel is converted to thermal energy as well as electrical energy, it is important to perform an energy balance on the system to see how ecient the solar panel will be at dierent operating temperatures. This allows the designer to project how much electrical power will be generated, so the design can be rened in order to generate the proper amount for the solar panels required load. The maximum power as a function of cell temperature can be graphed at dierent radiation levels, as shown in Figure 2.2. 6

Figure 2.2: Variation of Power as a Function of Cell Temperature [Due and Beckman, 2006]

2.3

Increasing Eciency

In recent years, there have been many attempts to curb the decrease in eciency. One such method is using a hybrid photovoltaic-thermal cell (PV/T) [Lee et al., 2008]. This method is shown to be eective for overall eciency; however, the eciency of the PV cells is still diminished. Jet impingement cooling devices have also been used to lower the temperature of solar panels. In this method, small pipes run through the plate that the PV cells are mounted on. Water runs through the pipes and the heat is transferred to the water. The water drains out constantly as new water is pumped in [Royne and Dey, 2007]. Another method adds textures to the outer layer of the solar panel. These textures allow the light to be trapped more eectively. Using a specic surface texture which creates a v-shaped geometry allows the eciency to increase as much as 52% for a 170 nm thick solar cell [Rim et al., 2007]. Similarly to adjusting the surface texture would be to create nned surfaces on the underside of the solar panel in order to increase the convection rate from which the heat leaves the back of the solar panel. The purpose of this thesis was to investigate how

dierent congurations of ns will aect the surface temperature and ultimately the eciency of the PV cells. Various computer models were created in which numerous n geometries were analyzed in order to nd an optimal design to keep the surface temperature of the solar panel as close as possible to the ambient temperature.

2.4

Current-Voltage Characteristic of PV Cells

To understand the performance of photovoltaic cells, it is important to nd the current-voltage characteristic, otherwise known as an I-V curve. The current-voltage characteristic is a graphical representation of the solar cells output current as a function of voltage. This characterization is achieved by approximating the solar cell (or module) with an equivalent circuit consisting of a current source in parallel with a single diode, as shown in Figure 2.3 [Ishibashi et al., 2008]. Using a diode to model the cells gives a nonlinear current-voltage characteristic, which does not obey Ohms Law. With zero irradiance, the solar cell will behave like a diode without a current source [Part II - Photovoltaic Cell I-V Characterization Theory and LabVIEW Analysis Code]. This will give an I-V curve that begins at a value of 0 amps. As the voltage increases, the current drops into the fourth quadrant of an xy graph, in the complex plane (real voltage value, imaginary current value). As light intensity is increased, the values needed to characterize the system are in the rst quadrant.

Figure 2.3: Equivalent Circuit of a Solar Cell [Ishibashi et al., 2012] 8

When the I-V curve is shown on a graph (as in Figure 2.4), it can be seen that the short circuit current (Isc ) can be read where the curve intersects the y-axis (voltage equals zero). Similarly, the open circuit voltage (Voc ) is shown where the I-V curve intersects the x-axis (where the current equals zero). Using Joules law, the voltage can be multiplied by the corresponding current at that point in order to nd the power generated by the solar cell. The power can be calculated along the entire curve in order to nd the maximum power. The voltage at this point is known as Vmpp , and the current known as Impp [Due and Beckman, 2006]. This is the most important point in the current-voltage characteristic, as it is used to nd the eciency of the solar cell. The curve can be seen to decrease with a small slope. Upon reaching maximum power, the power decrease is achieved faster as the negative slope increases to a larger number before reaching the open circuit voltage. The maximum power is often compared to the power found from the product of open circuit voltage and short circuit current (known as the theoretical power). The ratio between the two, respectively, is called the ll factor () [Desilvestro, 2008]. Ideally, the ll factor would be as close to 1 as possible in order to take advantage of the full potential of the solar cells. However, the ll factor in practice will always be less than unity. Once the ll factor is known for each level of irradiation, the maximum power can be calculated by multiplying the ll factor by the theoretical power [Desilvestro, 2008]. The power voltage curve (P-V curve) is often shown on the same graph as the I-V curve. This shows the power generated by the solar cell as a function of voltage, with the power on a secondary y-axis. The peak of this curve shows the maximum power generated. The P-V curve will cross through the origin at a power of zero. After the peak at maximum power, the curve will decrease before crossing the x-axis at the open circuit voltage. A graph showing the I-V curve, P-V curve, ll factor, open circuit voltage, short circuit current, and maximum power generated is shown in Figure 2.4 [Desilvestro, 2008].

Figure 2.4: I-V and P-V Curves [Desilvestro, 2008] The photovoltaic cells must be characterized with many dierent levels of irradiance to nd how the maximum power changes as a function of light intensity. It can be shown that the short circuit current of a solar cell will increase proportionately with an increase in irradiance. The open circuit voltage, however, will increase logarithmically. Since the short circuit current is nearly proportional to the incident radiation, it can be used to calculate the incident radiation [Due and Beckman, 2006]. This can be useful when the current voltage characteristic can be found but a pyranometer cannot be purchased to properly measure the irradiance. At each dierent irradiance level, it is crucial to keep the cells operating temperature constant. This is due to the temperature eects that will change the I-V curve. To show temperature eects, many dierent graphs should be made at each temperature, with each graph showing multiple traces of the current voltage characteristic at dierent levels of irradiance. As can be expected, the maximum power at a given irradiance will decrease as the cells operating temperature is increased.

10

The maximum power point has been known to decrease by as much as 19.5% when increasing the temperature from 20

to 50. The open circuit voltage can be ob-

served to decrease linearly with an increase in temperature [Desilvestro, 2008]. The temperature will typically have an increasing eect on the short circuit current at a rate of +0.05%/ , while an increase in temperature will decrease the open circuit voltage at a rate of -0.5%

. A graph showing this phenomenon is shown in Figure

2.5 [Part II - Photovoltaic Cell I-V Characterization Theory and LabVIEW Analysis Code].

Figure 2.5: Temperature Eect on I-V Curves [Part II - Photovoltaic Cell I-V Characterization Theory and LabVIEW Analysis Code] Values such as open circuit voltage, short circuit current, power generated, maximum power generated, voltage at maximum power, current at maximum power, and ll factor can be found using an I-V curve. Shunt and series resistances can also be approximated using an I-V curve. Losses exist in solar cells due to internal resistances that decrease eciency. To model these resistances, a shunt resistance in parallel and a series resistance are placed in the solar cells equivalent circuit, as can be seen in Figure 2.3. For maximum eciency, the shunt resistance should be as high as possible, with the series resistance kept at a minimum. These can be approximated on the I-V curve by taking the slope of the two main portions of the curve. The shunt resistance can be estimated as the inverse of the slope of the leftmost part 11

of the I-V curve (at the short circuit current), while the series resistance is found from the inverse of the slope of the right portion of the curve (at the open circuit voltage). This will give a better approximation of the shunt resistance. The series resistance, however, will be proportional but smaller than the resistance found at the open circuit voltage [Part II - Photovoltaic Cell I-V Characterization Theory and LabVIEW Analysis Code]. A better approximation of the series resistance is the dierent illumination level method, which involves using two dierent I-V curves at dierent irradiance levels. This method, however, will give errors if the temperature is not kept constant. Methods have been developed to compensate for this, such as the constant illumination level method. This method keeps the irradiance constant and factors in the temperature change [Ishibashi et al., 2008].

2.5

Scope of Thesis

The rest of this thesis investigates dierent methods for cooling the surface of solar panels in order to prevent the conversion eciency from diminishing. This was done in three ways. First, analytical equations were found and calculations were performed to nd the theoretical temperatures for the case of a bare panel with natural convection. Second, ANSYS was used to run various computational models. Finally, an experiment was setup in an attempt to match the results given computationally and by analytical calculations.

12

Chapter 3 Analytical Calculations

3.1 Mathematical Model

The equation for this model is derived from the energy equation. The energy equation for this case is a partial dierential equation derived from Fouriers law of heat conduction that describes the temperature distribution at any point in space or time. The equation must be solved using known boundary conditions. The basic equation is shown below for a three-dimensional plate [Mossayebi, 1987]. T t 2T 2T 2T + + x2 y 2 z 2

=0

(3.1)

It can be assumed that the temperature variation for this case is steady-state and one dimensional. In this case, the formula simplies to: 2T =0 x2

(3.2)

13

The general solution to this dierential equation can be solved by integrating twice. The result is shown in equation 3.3.

T ( x) = C1 x + C2

(3.3)

Equation 3.3 can be solved for C1 and C2 by applying the applicable boundary conditions discussed in the next section.

3.2

Boundary Conditions

The problem was rst investigated using analytical calculations. Basic heat transfer equations were used to model the solar panel as a plane wall with multiple layers, each with their own thermal properties. Literature research revealed that a typical solar panel is made of layers of glass, EVA, ARC, silicon, and Tedlar [Lee et al., 2008]. The thermal conductivity of these materials was referenced as well as the average thickness of each of the layers. Two boundary conditions were needed in order to solve the equations to nd the temperature at both surfaces of the solar panel. The rst set the solar heat ux on the top surface equal to heat conduction throughout the panel and heat that escapes the surface by convection. The second boundary condition set heat conduction equal to the heat convection at the back surface of the panel. The subscript 1 denotes a condition on the front of the panel, while 2 denotes a condition on the back surface. Figure 3.1 shows a schematic of the solar panel modeled as a 2D plane section with the appropriate boundary conditions. These two boundary conditions are shown in the following two equations, respectively.

14

Figure 3.1: Boundary Conditions on Solar Panel

keq

dT (0) =q o h1 [T (0) T1 ] dx dT (L) = h 2 [ T ( L) T 2 ] dx

(3.4) (3.5)

keq

The dierential equation was solved symbolically and yielded the following results: 2 L + h1 kT1 + h2 kT2 + qk h1 h2 LT1 + qh h1 k + h2 k + h1 h2 L

T (0) =

(3.6)

T ( L) =

h1 h2 L(T2 T1 ) qh 2 L + h1 h2 LT1 + qh 2 L + h1 kT1 + h2 kT2 + qk h1 k + h2 k + h1 h2 L (3.7)

These equations were backsolved to nd the convective heat transfer coecient. However, it was assumed that the h2 in the above equations was actually a com15

bined heat transfer coecient that included both convective and radiative eects (hcombined = hconv + hrad ). To compensate for this, the radiation value was calculated and subtracted out by rst assuming that the surface temperature under the panel (to which the radiation was transfered to) was equal to the ambient temperature. This required nding hrad through the following relation:
4 hrad (Ts Tsurr ) = (Ts4 Tsurr )

(3.8)

Once the radiative heat transfer coecient was found, it was subtracted from the combined radiation found in the previous equations. This gave a more accurate value to compare to the results found from the calculations performed with experimental data.

3.3

Analytical Results

Equations 3.6 and 3.7 were both solved using heat transfer coecients found from the experimental calculations. Because of the complexity of analytical calculations for nned surfaces, these equations only applied to the bare panel with natural convection. The assumption was made that the ambient temperature on both sides of the panel was 22 . Table 3.1 shows the calculated temperatures for a bare panel with natural convection. Table 3.1: Temperatures for Bare Panel with Natural Convection I
W m2

1000 1250 1500

T(0) ( ) 169.53044 204.18797 247.45344

T(L) ( ) 165.66324 199.29633 241.76036

16

3.4

Atmospheric Conditions

In order to demonstrate the conditions for a worst-case scenario, the atmospheric conditions were chosen in this section to model Phoenix, Arizona. The maximum air temperature in the summer months in Phoenix was found to be 41.06 [Climate Data Summary]. From a map showing the irradiation distribution throughout the United States, the average heat ux (available to photovoltaics) acting upon the top
kW h W surface of the solar panel was estimated to be 6.8 m 2 day or 283.33 m2 [Roberts, 2008]. A

map showing the solar irradiation distribution throughout the United States is shown Figure 3.2.

Figure 3.2: Map Showing Solar Irradiation in the U.S. [Roberts, 2008]

17

However, the daily average factors in non-peak hours, including hours after sunset when irradiance is virtually nonexistent. The peak irradiance acting on the solar panel would likely be much higher. The standard irradiance used for solar experiments is
W considered to be 1000 m 2 , which is assumed to be in ideal conditions [AL-Sabounchi,

1998]. This irradiance assumption was veried for Phoenix, Arizona by using the ASHRAE Clear Sky Model. The total irradiance, assuming a horizontal panel, is calculated using Equation 3.9:

Gt = GN D + Gd + GR

(3.9)

where Gt equals total irradiance, GN D equals normal direction irradiation, Gd equals direct radiation (adjusted for clearness), and GR equals reected radiation incident on the surface. Since this example uses a horizonal solar panel, GR will be equal to zero and drops out of the equation. This is due to the view factor from the ground to the solar panel being zero, meaning it will not reect any radiation onto the top of the panel. In order to calculate normal direction irradiation, many other values must be found rst. These values include apparant solar irradiation at air mass equal to zero (A), atmospheric extinction coecient (B), solar altitude (sol ), clearness number (CN ), sun declination ( ), hour angle (h), and the latitude of the location (l). Direct radiation is then calculated using normal direction irradiation (GN D ) and the ratio of diuse irradiation on a horizonal surface to direct normal irradiation (C). These values found from tables (chosen for the peak day of July 21st) and the calculations are shown in detail [McQuiston et al., 2005].

18

A = 1093

W m2

B = 0.186 C = 0.138 l = 33.43 h = 0, solarnoon

Equation 3.10 is used to nd a value used in the equation for declination. The value n is the day of the year. To match with the data found from tables, July 21st was chosen (day 202). N = (n 1)(260/365) (3.10)

= 0.3963723 22.9132745 cos N + 4.0254304 sin N 0.3872050 cos 2N (3.11) + 0.05196728 sin 2N 0.1545267 cos 3N + 0.08479777 sin 3N sol = sin1 [cos 1 cos cos h + sin l sin ] (3.12)

GN D = exp

A
B sin sol

(3.13)

Gd = CGN D

(3.14)

When the numerical values for GN D and Gd were substituted into equation 3.9, the
W total irradiation was found to be 925.0598 m 2 . This value is close to the standard value W of 1000 m 2.

19

3.5

Experimental Calculations

Once the speed of the wind and the steady state temperatures were found in each experimental case, the convection heat transfer coecient was calculated. The equations used for nding the heat transfer coecient were dierent for each case. These included external forced convection on a at plate, natural convection on a at plate, external forced convection on a nned surface, and natural convection on a nned surface. These calculations are shown in more detail in the following sections.

3.5.1

External Forced Convection Over a Flat Plate

This calculation was performed for the case with wind blowing across the back of the bare panel. First, the lm temperature was found by calculating the average temperature between the back surface and the ambient air. Once it was assumed that the ambient pressure was 1 atm, the properties of air were found in tables in Heat and Mass Transfer: Fundamentals & Applications, by Yunus A. C engel and Afshin J. Ghajar. These values included the thermal conductivity, Prandtl number, and kinematic viscosity. Then the Reynolds number was calculated using the wind velocity, panel length, and kinematic viscosity: V Lpanel

ReL =

(3.15)

This number was then referenced to the critical Reynolds number of 5X 105 . For the experiments performed, the Reynolds number was found to be less than the critical number; therefore, the ow was assumed to be laminar. The Nusselt number for a laminar case was then calculated as follows:
.5 1/3 N ux = 0.453Re0 x Pr

(3.16)

20

Since this is the local Nusselt number, the local convection heat transfer coecient was found as follows: hx = k Lpanel Nu (3.17)

However, the average heat transfer coecient needed to be determined. This required the equation to be integrated over the length of the entire backing. This calculation is shown in Equation 3.18. 1 Lpanel
0 Lpanel

= h

hx d x =

0.906kP r1/3 0.5 ReL Lpanel

(3.18)

3.5.2

Natural Convection Over Flat Plate

These next calculations were used to nd the convective heat transfer coecient for the bare back surface of the solar panel with still air, as well as the top of the solar panel in all cases. Once again, the process started with nding the lm temperature using the average between the ambient temperature and the measured temperature of the panel. The thermal conductivity, Prandtl number, kinematic viscosity, and coecient of volume expansion ( =
1 ) Tf ilm

were also used. The next step was to nd

the Rayleigh number using Equation 3.19, which is a measure of the relationship of buoyancy and viscosity of the uid [C engel and Ghajar, 2011]. g (Ts T )L3 c Pr 2

RaL =

(3.19)

Equation 3.19 makes use of the characteristic length, which is a ratio of the surface area of the plate and perimeter: Lc = As p (3.20)

21

From the Rayleigh number, the Nusselt number used to nd the convective heat transfer coecient was found for the case of 104 < RaL < 107 .
1 /4

N u = 0.54RaL

(3.21)

Finally, the heat transfer coecient was found from the relation used previously: k Nu Lc

h=

(3.22)

3.5.3

Forced Convection Across Fins

The following method for nding the Nusselt number for forced convection across ns was developed by Teertstra et al. The calculation for forced convection across ns starts with nding the Reynolds number. Unlike the forced convection over a at plate calculation, which uses the plate length in the Reynolds number calculation, Equation 3.23 uses the distance between the ns, Sf in . V Sf in

ReL =

(3.23)

The Reynolds number is then non-dimensionalized using the ratio between the channel width, Sf in , and the n length, Lf in . This gives the Reynolds number for channel ow. Re b = ReL Sf in Lf in (3.24)

22

Finally, the Nusselt number for ow between plates was found using Equation 3.25. Included in the equation is combination parameter, n. This number must be optimized (usually to the closest integer) in order for the equation to produce accurate results [Teertstra et al., 2000]. Re bP r 2
n n 1/n

N ub =

0.664

Re bP r

1 3

3.65 1+ Re b

(3.25)

3.5.4

Natural Convection on Fins

Natural convection across ns was calculated in a similar manner to the calculation for natural convection from the bottom of a at plate. After the relevent properties are found at the lm temperature, the Rayleigh number was once again found using Equation 3.19. The Nusselt number was then found as follows:
1 /4

N u = 0.27RaL

(3.26)

This was used to nd the Nusselt number for an inclined plate at 0 45 . us = 0.645[Ras (Sf in /Lf in )]1/4 N

(3.27)

The heat transfer coecient was then found from the common relation in Equation 3.17.

23

Chapter 4 Computational Analysis

4.1 ANSYS

To verify the analytical results found, ANSYS was used to model the heat transfer throughout the solar panel. ANSYS is a suite of computer software applications used to perform computer simulations for various cases (e.g., uid ow, harmonic response, rigid dynamics, etc.) The various programs contained within ANSYS were accessed from ANSYS Workbench, which works as a connecting point between them. The computational analysis was performed for each individual case, using dierent settings and one of two sub-programs within ANSYS. The sub-programs used (and accessed using Workbench) were ANSYS Mechanical and ANSYS Fluent.

4.1.1

ANSYS Mechanical

For steady-state heat conduction through a plane wall, a two-dimensional model was made for the sake of simplicity. This analysis excluded the need for uid ow analysis because the still air and heat ux boundary condition left only the temperature distribution through the solar panel to be investigated. Due to these simplied criteria, ANSYS Mechanical was chosen for the case of a bare panel with only natural

24

convection. ANSYS Mechanical is primarily a nite element analysis program used to perform static or dynamic analyses on linear and nonlinear structural systems. However, it can also perform a thermal analysis, as well as physics computations with systems such as piezoelectrics [ANSYS Mechanical]. For the bare panel with natural convection, only the thermal analysis was performed.

4.1.2

ANSYS Fluent

In addition to the bare panel with natural convection, three other cases were analyzed: (1) bare panel with forced convection and ns with both (2) natural and (3) forced convection. A uid ow analysis was required for each of these cases. For both cases of forced convection, the uid ow was modeled for the air owing underneath the solar panel. For natural convection with ns, uid ow was not necessarily needed as the air velocity was zero. However, a three-dimensional model was created for each case due to the more complicated nature of uid ow and ns. The natural convection case was analyzed by modifying the forced convection over ns case by simply changing the air speed to zero. These cases are best analyzed using ANSYS Fluent, a nite volume analysis program. Ansys Fluent is software included in the ANSYS suite that is used to perform computational uid dynamics (CFD) analyses. This includes wind over an airfoil, ow through a pipe, etc. The motion of the uid is calculated as well as the heat transfer (which gives the temperature distribution) and turbulence [ANSYS Fluent].

4.2

Bare Panel, Natural Convection, 2D

First, ANSYS Workbench was opened. The correct analysis system was then chosen to be Steady-State Thermal (ANSYS) which creates a project le appropriate

25

for the model that was needed. Then the setup for the model began by adding in the materials and specifying their properties (specically, thermal conductivity). Table 4.1 shows the thickness and thermal conductivity for each layer [Lee et al., 2008]. Table 4.1: Conductivity and Thickness of Solar Panel Layers Material Glass EVA1 ARC Si EVA2 Tedlar
W Conductivity ( m ) K 0.98 0.23 1.38 148 0.23 0.36

Thickness (m) 0.003 0.0005 0.08 106 0.000325 0.0005 0.0001

The geometry was then created using ANSYS DesignModeler, the computer-aided design (CAD) program used by ANSYS. An arbitrary length of 10 millimeters was chosen for the length of the panel, as the heat transfer is assumed to only be across the thickness. The dierent layers however were made to have the thicknesses that are typical of common solar panels, as shown in Table 4.1. Due to a minimum length value imposed by DesignModeler, the layer of ARC was omitted from the geometry due to its thickness of 0.08 103 mm. Therefore, the assumption was made that the eects of this layer were negligible, as its thermal resistance is extremely miniscule in comparison to the solar panels total resistance. A portion of the geometry created is shown in Figure 4.1. The next step in the computational analysis was to break up the geometry into elements in a process called meshing. This was completed using ANSYS Mechanical, a nite element analysis (FEA) application that allows the user to perform structural, thermal, thermoelectric, acoustic, and piezoelectric analyses [ANSYS Mechanical]. A rectangular mesh was created, with each layer being divided up into even smaller components. ANSYS Mechanical was also used to add the boundary conditions. A heat ux was added to the outer layer consisting of glass. Natural convection was then applied to the outside layer consisting of Tedlar, and the ends of the layers were 26

Figure 4.1: Bare Panel Model Created in ANSYS insulated in order to ignore heat transfer from the sides. Natural convection was also added to the glass layer. Ambient temperatures were assumed to be around room temperature at 22 , in order to match the conditions in the experimental setup. The heat transfer coecients for natural convection were taken from calculations from the experimental data. The meshed model is shown in Figure 4.2. The model was then solved, and results were given in the form of a model showing the temperature distribution throughout the solar panel. The top layer of the solar panel, which is made of glass and exposed to solar radiation, was found to have a steady-state temperature of 168.35

when a heat ux of 1000

W m2

was applied. The

bottom surface, which consists of Tedlar and experiences natural convection, had a temperature of 166.8 . Figure 4.3 shows the temperature distribution throughout the solar panel. The results obtained in the computational analysis were consistent with results found using analytical calculations. The found temperatures of the top surface (glass)
W had a percentage dierence of 0.699% for I = 1000 m 2 . The percentage dierence for

27

Figure 4.2: Meshed 2D Model

Figure 4.3: Temperature Distribution Throughout Bare Solar Panel with Natural W Convection, I = 1000 m 2 28

the bottom surface was 0.684% for the same irradiance. Any small dierences could exist because the computational model didnt contain the layer of ARC. Also, the assumption was made that the temperature distribution was uniform throughout the length. In reality, the temperature would slightly vary as the heat transfer approaches the edges. For further assurance, the model was resolved with an even ner mesh. The nal temperature distribution remained the same. It was then safely assumed that the results had converged. Table 4.2 shows the temperatures at the glass (T(0)) and back of Tedlar (T(L)) for each irradiance. Table 4.2: Bare Panel, Natural Convection Computational Data I
W m2

1000 1250 1500

T(0) ( ) 168.35 202.68 245.75

T(L) ( ) 166.8 200.71 243.46

4.3

Bare Panel, Forced Convection, 3D

The rst steps for each subsequent case were setup similarly to the analysis for the 2D bare panel with natural convection. The geometry, however, was created using Solidworks. This was used to draw a three-dimensional model adhering to the exact specications of the solar panel dimensions. An extra section was drawn which connects to the backing of the solar panel (i.e., Tedlar). This created a channel through which the air ow could be modeled. The geometry was saved as a parasolid le (extension .x t) for easy exportation to meshing in ANSYS Mechanical. Figure 4.4 shows the model created in Solidworks, while Figure 4.5 shows a magnied model where the layers (right-to-left: air, glass, EVA layer 1, ARC, Silicon, EVA layer 2, Tedlar) can be seen.

29

Figure 4.4: Bare Panel Model Created in Solidworks

Figure 4.5: Closeup of Bare Panel Model Created in Solidworks

30

After the geometry was imported, the next step was to create an ecient mesh. An automatic mesh was created at rst. After the next steps were taken and the full analysis was performed, the mesh was updated in an attempt to create more accurate results. This included customizing the mesh by setting the divisions on each edge to a pre-chosen number. Further, midside nodes (which gives each rectangular element 8 nodes for solution points instead of 4) were placed on each model for a second solution. This process was repeated until the results were considered sound, and there was little variation in the nal temperature distributions. The meshed bare panel can be seen in Figure 4.6.

Figure 4.6: Meshed Bare Panel Model

31

The next step was to open Fluent to calculate a solution. Before loading Fluent, an option was selected called Double Precision. Although slower to converge, this option can maintain more accuracy in cases with airow or complicated geometry. Upon opening Fluent, the default setup options were kept aside from turning on the Energy Equation. This allows for heat transfer to be calculated throughout the solar panel. The next step was to add the materials and their thermal conductivities. Each geometrical body was then selected and correlated with its proper material. Boundary conditions were then set for the glass top (heat ux, which models the irradiance) and the inlet/outlet of the air channel. The temperatures for the inlet and outlet were set to the ambient temperature of 299 K. The outlet was set as a pressure outlet, which allows outow caused by a velocity of a uid coming through the inlet. Finally, the velocity was set at the inlet. Mesh interfaces were then created at each interface between the dierent materials. This tied two walls into one, which allowed heat ow to travel across the materials without hitting an innitely small gap. The solution was then initialized and run. Residuals were modeled, such as energy (convergence criteria = 1e-0.6), continuity (convergence criteria = 0.001), and velocity (convergence criteria = 0.001). Once the residuals converged, the calculation was complete and temperature distributions could be viewed after opening the Results section in Workbench. Individual node temperatures at the center of the Tedlar backing (where the thermocouple was adhered during the experimental setup) were checked using the Probe tool. This process was repeated for each irradiance at each of the velocities used in the experimental setup. Table 4.3 shows the temperatures at the glass (T(0)) and back of Tedlar (T(L)) for each irradiance. Also, Figure 4.7
W shows the temperature distribution for the bare panel, with an irradiance of 1000 m 2

and a velocity of 2.54 m . s

32

Table 4.3: Bare Panel, Forced Convection Computational Data

I
W m2

1000 1000 1000 1250 1250 1250 1500 1500 1500

V (m ) s 1.12 1.37 2.54 1.12 1.37 2.54 1.12 1.37 2.54

T(L) ( ) 139.689 126.758 102.735 167.478 152.176 122.512 196.076 179.541 140.83

W m Figure 4.7: Temperature Distribution for Bare Panel, I = 1000 m 2 , V = 2.54 s

4.4

Fins, Natural Convection, 3D

The process for the nned solar panel with natural convection model started similar to the previous model. The solar panel and ns were drawn in Solidworks and imported into ANSYS Design Modeler using a parasolid le. The mesh was then created in the 33

manner discussed previously, which special attention paid to the elements on the ns. The mesh was rened until there was little variation in the temperature distrubition between the dierent meshes. The models and meshes for each heat sink are shown in Figures 4.8, 4.9, 4.10, and 4.11. Isometric views of Sink 1 and Sink 2 are shown in the Appendix.

Figure 4.8: Model of Solar Panel with Heat Sink 1

34

Figure 4.9: Model of Solar Panel with Heat Sink 2

35

Figure 4.10: Mesh Shown on Sink 1 Model

Figure 4.11: Mesh Shown on Sink 2 Model

36

The setup method was dierent due to the model having natural convection. As Fluent was opened, Double Precision was selected to ensure accuracy. First, gravity was activated and the components were set to mimic the panel being tilted at a 10 degree angle. Then operating conditions were set to keep the air temperature at 299 K
kg and the operating air density at 1.225 m 3 . The dierent materials were then added along

with their thermal conductivities. For air, the model diered from forced convection by setting the density option to boussinesq. This allows Fluent to model bouyancy driven ows with little to no density changes. Also, it was required that a thermal expansion coecient for air be specied as 0.003407. Under Solution Methods, the pressure setting was changed to Second-Order. The boundary conditions were then set to model natural convection, keeping the air temperatures at 299 K and the velocity inlet condition to zero. The model was then initialized and run until converged. This was repeated for each heat sink with each of the three heat ux conditions. Table 4.4 shows the temperatures at the base of the ns for each irradiance, and Figures 4.12
W and 4.13 show the temperature distributions for an irradiance of 1000 m 2.

Table 4.4: Fins, Natural Convection Computational Data Heat Sink 1 1 1 2 2 2 I

W m2

1000 1250 1500 1000 1250 1500

T( ) 130.72 144.92 177.167 112.244 125.283 137.343

37

Figure 4.12: Temperature Distribution Throughout Solar Panel with Sink 1 - I = 1000 W ,V=0 m m2 s

Figure 4.13: Temperature Distribution Throughout Solar Panel with Sink 2 - I = 1000 W ,V=0 m m2 s

38

4.5

Fins, Forced Convection, 3D

The setup for the nned solar panels with forced convection was identical to the setup for forced convection with the bare panel. The three-dimensional models for the nned solar panels shown previously, as well as the meshes previously shown, were used. The model was run for each air velocity at each irradiance for both nned panels. The temperatures at the base of the ns are shown in Table 4.5 for each irradiance. Figures 4.14 and 4.15 show the temperature distributions for sinks 1 and 2, respectively. Table 4.5: Fins, Forced Convection Computational Data Heat Sink 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 I
W m2

1000 1000 1000 1250 1250 1250 1500 1500 1500 1000 1000 1000 1250 1250 1250 1500 1500 1500

V (m ) s 1.12 1.37 2.54 1.12 1.37 2.54 1.12 1.37 2.54 1.12 1.37 2.54 1.12 1.37 2.54 1.12 1.37 2.54

T(L) ( ) 49.519 45.977 36.793 55.64 51.216 39.518 61.779 56.342 42.225 46.138 42.997 37.937 51.22 47.923 40.784 56.141 52.409 43.771

39

Figure 4.14: Temperature Distribution Throughout Solar Panel with Sink 1 - I = 1000 W , V = 2.54 m m2 s

Figure 4.15: Temperature Distribution Throughout Solar Panel with Sink 2 - I = 1000 W , V = 2.54 m m2 s

40

Chapter 5 Experimental Analysis

5.1 Objective

The main objective of the experimental analysis was to verify the results obtained by analytical calculations and computational methods. Variables in a real-world application can be compensated for in the experimental analysis, while the analytical and computational methods used simplied models and assumptions that gave idealized results. Care was taken to ensure that the experiment was setup in a uniform manner for each phase of testing. This was required as an assurance that the experiment was repeatable and reliable results could be obtained. New equipment was purchased in order to avoid technical errors and to make certain that the measurement tools were properly calibrated. The basic setup is shown in Figure 5.1. The fan can be seen blowing across the bottom surface of the solar panel, with the panel tilted at 10 . Also shown are the dimensions for the two heat sinks in Figure 5.2.

41

Figure 5.1: Basic Experimental Setup

(a) Heat Sink 1. Depth = 0.276m. Length = 0.337m.

(b) Heat Sink 2. Depth = 0.337m. Length = 0.276m.

Figure 5.2: Dimensions of Heat Sinks

42

5.2
5.2.1

Apparatus
General

Various equipment was used in the experimental analysis, which included the following: Suntech STBKb 10Watt solar module, NI PXI-4130 Power SMU, NI PXIe-1073 5-Slot 3U PXI Express Chassis with AC plus Integrated MXI-Express Controller, a PC laptop with National Instruments Labview 2011, thermocouple wire, a pyranometer, VelociCalc Plus Multi-Parameter Ventilation Meter 8386, a Holmes Products Corp. electric fan, NI 9219 24-Bit Universal Analog Input Data Acquisition (DAQ) Unit, NI cDAQ 9172 Legacy NI CompactDAQ Chassis, and a custom made box for containing the light given o by 90 Watt halogen ood lights manufactured by General Electric. Figure 5.3 shows a diagram of the major components used in the experimental setup. Use each section below to reference the component number in the line diagram.

Number 1 2 3 4 5 6 7

Component Suntech STBKb 10 Watt Solar Module Pyranometer DAQ Chassis SMU Chassis Power SMU Dell Vostro Dell Latitude

Number 8 9 10 11 12 13 14

Component Analog Output Module DAQ Dimmer Servo BASIC Stamp HomeWork Board Relay Thermocouple Wire

Figure 5.3: Line Diagram of Major Components

43

5.2.2

Solar Module

A smaller-scale solar panel was used as a representative area of a larger panel used in typical PV cell applications. A Suntech STBKb 10Watt solar module was used, which consists of a 12 by 6 array of PV cells. The panel has dimensions of 14.5in x 12.2in (0.3683m x 0.30988m), with a thickness of 0.70in (0.01778m). The module has a Peak Power Voltage (Vmpp) of 17.4 Volts, and a Peak Power Current (Impp) of 0.57 Amps, which gives a Peak Power (Pmpp) of 10 Watts. The maximum power is achieved during peak hours, assuming a total eciency of 8.8%. Two panels of the same model were used in order to perform the experiment for two dierent n congurations. The solar module is shown in Figure 5.4.

Figure 5.4: Suntech STBKb 10 Watt Solar Module

44

5.2.3

Pyranometer

A pyranometer manufactured by Kipp & Zonen (model SP Lite2) was used to measure the irradiance given o by the halogen lights. As can be seen in gure 5.5, the pyranometer was attached to a cord which is attached to the DAQ. The irradiance was indicated on a VI made in Labview. The lights were adjusted until the pyranometer
W measured an irradiance at a specic target value (1000, 1250, 1500 m 2 ).

Figure 5.5: Kipp & Zonen Pyranometer

5.2.4

Data Acquisition Unit

In order to obtain data from the experiment, the NI 9219 24-Bit Universal Analog Input Data Acquisition Unit was used. The DAQ acquired temperature data from the thermocouple as well as irradiance from the pyranometer. A chassis separate from the SMU chassis was used as a dock for the DAQ. The NI cDAQ 9172 Legacy NI CompactDAQ Chassis was used, as it is capable of measuring a variety of analog and digital signals read from numerous types of data acquisition units. The DAQ along with the chassis are shown in the following two gures, Figure 5.6 and Figure 5.7, respectively.

45

Figure 5.6: NI 9219 DAQ

Figure 5.7: NI cDAQ 9172 Chassis

5.2.5

Source Measurement Unit

In order to perform current/voltage sweeps across the solar panel to determine its eciency, a PXI Source Measure Unit (SMU) was used. The SMU can measure dierent voltages at a specic current and vice versa, which is useful in nding the systems characteristic IV curve. This IV curve is important in nding the maximum power generated by the solar cell at dierent light intensities. The specic SMU used in this experiment was the NI PXI-4130 Power SMU made by National Instruments, which is shown plugged into the SMU Chassis in Figure 5.8.

46

5.2.6

SMU Chassis

The SMU was plugged into a chassis (NI PXIe-1073 5-Slot 3U PXI Express Chassis with AC plus Integrated MXI-Express Controller), which docks the SMU and allows it to be connected alongside other PXI modules in the chassis ve peripheral slots. The chassis also allows connection to a PCI Express card (NI-8360), which allows a PC to be used as a host controller. The chassis with the SMU inserted into slot 1 is shown in Figure 5.8.

Figure 5.8: NI PXIe-1073 5-Slot 3U PXI Express Chassis

5.2.7

Laptops

Two laptops were used in the experiment. The rst laptop, a Dell Vostro, was equipped with Labview 2011. It also contained a slot used for the Express Card which connected to the Power SMU. This laptop was used to control the on/o relay for the lights, as well as measure the temperature, maximum power output, eciency, etc. The second laptop was a Dell Latitude E5500. A cable attached to the BASIC Stamp Homework Board ran to the laptop. Software for the Homework Board was used to control the rotation of the servo, which varied the dimmer switch to control the irradiance of the light array.

47

5.2.8

Analog Output Module

The relay required an analog input of 5 Volts to trigger on the lights. The NI 9264 16-Channel Analog Output Module was chosen for this purpose. Each channel consisted of a positive and negative terminal. The wires connected to the relay were connected into the terminals in channel 0. The NI 9264 is shown in Figure 5.9.

Figure 5.9: NI 9264 Analog Output Module

5.2.9

Light Array Control Board

In order to control various aspects of the lighting, a piece of plywood was cut to be used as a control board. First, a dimmer was installed to change the intensity. It can be very cumbersome to adjust a dimmer by hand and obtain an equal irradiance from a previous experiment, so it was desired to use computer software to control the dimmer. BASIC Stamp was used on a seperate laptop to write a small program which controls a small circuit, called a BASIC Stamp HomeWork Board (manufactured by Paralax). The circuit was specically wired in a way to control a small servo, also manufactured by Paralax. Plastic gears (gear ratio 1:1) were secured to the tops of the dimmer and servo, which were screwed into the board in a way that would interlock the gears. By changing a number within the BASIC Stamp program and selecting 48

RUN, the servo (and subsequently the dimmer) was rotated in small increments to achieve a more precise irradiance level in the lights. The BASIC Stamp HomeWork Board (with Duracell Procell 9V alkaline battery) and servo/dimmer switch are shown respectively in Figure 5.10 and Figure 5.11. The BASIC Stamp program is shown in the Appendix.

Figure 5.10: Paralax BASIC Stamp HoweWork Board

Figure 5.11: Servo Connected to Dimmer Switch It was desired to create a separate switch for the lights that can be controlled using computer software. A Gold solidstate relay (SAP4825D) was chosen for two purposes: rst, it sent a signal to the lights so a button could be pressed in the VI to turn the lights on/o. Secondly, the VI is designed so the user can input a limit temperature. If the thermocouple reads higher than this, the lights would automatically turn o until the temperature fell back below the limit temperature. This protects the solar panel from becoming damaged if the temperature exceeded the factory specied maximum operating temperature of 85

. The relay is shown in gure 5.12.

49

Figure 5.12: Gold Solidstate Relay

5.2.10

Software

The Dell Vostro laptop was installed with Labview 2011 to record the experimental readings from the DAQ and SMU. A Virtual Instrument, VI, is a specic program created within Labview that allows experiments to be recorded through signal manipulation and measurement. The specic VI used in the experiment was obtained from National Instruments website. This VI allowed the temperature to be recorded. It also displayed pertinent values such as voltage, current, and eciency. These values were used to show the IV-curve within the VI. Modications were made, such as converting the temperature to Fahrenheit, adding in a section to control the relay, and allowing the irradiance to be measured and specied prior to taking voltage measurements. A screenshot of the front panel (panel showing graphs and control knobs used for the experiment) as well as a screenshot of the block diagram (which shows the code and wiring that runs in the background of the VI) are shown in Figure 5.13 and Figure 5.14, respectively. Only a small portion of the block diagram is shown. More screenshots of the VI (after obtaining results for the bare panel at dierent velocities
W with an irradiance of 1000 m 2 ) are shown in the Appendix.

50

Figure 5.13: Front Panel of VI

51

52 Figure 5.14: Block Diagram of VI

Figure 5.15: Navigation Window

53

A second, smaller VI was also created in order to measure the irradiance before the experiment was set to start. Once the proper irradiance was achieved, the value was put into the main VI so the calculations for panel eciency could be calculated correctly. Figure 5.16 shows both the front panel and block diagram of the pyranometer VI.

Figure 5.16: Pyranometer VI

5.2.11

Custom Box

A custom-made box was designed in order to mount the solar panel over an array of 90 Watt halogen ood lights. The array consisted of four lights that are wired in parallel to a dimmer switch in order to adjust the light intensity to achieve a proper irradiance. The box itself was constructed by welding together a frame made of aluminum tubing. Flat aluminum panels were screwed to the outside to contain the light. The inner sides of the box were tted with insulation millboard to minimize loss of heat. The box sat on a seperate frame without aluminum sides so air could ow freely across the backside of the panel. Latches on the box and padding around the edges were used to create a tight t between the top and bottom sections, to ensure that no light would escape. With the bottom frame detached, the box could be set upside-down in order to provide easier access to the backside of the panel, where various cooling methods were used, such as water lm cooling. Figures 5.17, 5.18, and 54

5.19 show the bottom portion of the box (used as a stand), the top portion of the box (containing the array of halogen lights wired to the dimmer switch and the solar panel) and the fully assembled box, respectively.

Figure 5.17: Custom Box Stand

Figure 5.18: Custom Box Top

55

Figure 5.19: Fully Assembled Box The array of lights was contained within a piece of plywood, which sat on brackets inside the top portion of the custom box. A seperate piece of plywood was cut with a round hole in the center. This was placed between the two sections of the box (temporarily taking the place of the solar panel). This enabled a pyranometer to be placed in the hole in order to measure the irradiance given o by the lights before the solar panel was placed. Figure 5.20 depicts these two pieces of plywood.

(a) Halogen Flood Lights

(b) Plywood with Measuring Hole

Figure 5.20: Lights and Light Measurement

56

5.2.12

Ventilation Meter and Fan

A variable speed electric fan (manufactured by Holmes Products Corp.) was used to simulate airow at dierent velocities across the backing of the solar panel. The custom box was set at a 10 degree angle and the fan was laid underneath. The velocity was measured using a handheld VelociCalc Plus Multi-Parameter Ventilation Meter 8386. Three dierent velocities were measured using the device in units of feet per minute. The fan and ventilation meter are shown in Figures 5.21 and 5.22, respectively.

Figure 5.21: Holmes Electric Fan

Figure 5.22: VelociCalc Plus 8386 Ventilation Meter

5.2.13

Fins

Two dierent n congurations were chosen for the heat sinks. They were manufactured out of aluminum by Q4 Manufacturing Solutions located in Youngstown, Ohio. Each heat sink consisted of two pieces, which are adhered to the back of their 57

respective solar panels. This was done using a silicone manufactured by Silicone Solutions. The specic product used was SS-35 Thermally Conductive Silicone RTV. This was chosen for its ability to allow heat ow through the adhesive in order to minimize resistance. It was also chosen for its strength, which allowed the heat sinks to be mounted underneath the solar panel without detaching. The adhesive was made to be highly weather resistant, which is a large factor in solar applications. Figures 5.23 and 5.24 depict the rst heat sink with smaller ns and the second heat sink with larger ns and spacing, respectively.

Figure 5.23: Heat Sink 1, Smaller Heat Sink

Figure 5.24: Heat Sink 2, Larger Heat Sink

58

5.3
5.3.1

Method
General Method

The experimental setup started with attaching a thermocouple to the backside of the solar panel using electrical tape. This thermocouple was wired to the DAQ, as well as the wires leading to the relay. A pyranometer was set underneath the reserved piece of plywood, which was placed between the two sections of the box, so the irradiance could be measured at dierent light intensities. The lead wires from the solar panel were then inserted in the terminals of the SMU. The respective chassis for the SMU and the DAQ were connected to the laptop. The cable running from the BASIC Stamp HomeWork Board was inserted into a terminal in the second laptop. The solar power measurement VI in National Instruments Labview and the BASIC Stamp program were opened on their respective laptops. The VI was then started and the lights were turned on. At this point, it was necessary to allow the lights to sit for a period of time in order for them to heat up and approach steady-state. The radiation indicator attached to the pyranometer was then turned on. Adjustments were made
W W using BASIC Stamp in order to achieve irradiances of approximately 1000 m 2 , 1250 m2 W , and 1500 m 2 , depending on the stage of the experiment. Once the radiation indicator

showed a steady irradiance, the experiment was ready to begin.

5.3.2

Phase 1: Bare Panel with Natural Convection

The rst phase of experimentation consisted of measuring the eciency of the bare panel, using no cooling methods (natural convection, which relies solely on the bouyancy of air to naturally create an upward draft). Measurements were taken at the initial temperature, which was at or slightly above room temperature. This involved recording voltage, current, eciency, and temperature. The panel was then allowed to sit while the solar cells heated up to a steady temperature. Once steady-state 59

was achieved, a second set of data (also including voltage, current, eciency, and temperature) was recorded and everything was exported to an Excel spreadsheet. Screenshots were saved for each measurement which shows the IV curves each time a set of data was taken. This experiment was performed three times for the three dierent chosen irradiances, which gave a total of three sets of data. Figure 5.25 shows the entire experimental setup for Phase 1.

Figure 5.25: Testing Performed with Still Air on Bare Panel

5.3.3

Phase 2: Bare Panel with Forced Convection

It was desired to investigate the temperature at dierent air velocities. This consisted of starting the experiment identically to Phase 1, with the addition of a fan placed underneath the panel set to its highest setting ( 2.54 m ). Once the panel s

60

reached a steady temperature, the data was collected and the setting on the fan was changed to medium ( 1.37 m ). This process was repeated again with the low fan s setting ( 1.12 m ). The experiment was performed three seperate times for the three s irradiances. This gave an additional nine sets of data, for a total of twelve sets. The data was tabulated to be compared to the steady-state temperatures for the natural convection experiments. Figure 5.26 shows the experimental setup for this phase.

Figure 5.26: Testing Performed with Wind on Bare Panel

5.3.4

Phase 3: Panel with Fins

The third phase was then started with the goal of attaching two n congurations to the back of their respective solar panels. The heat sinks were attached to the solar panel using a highly conductive thermal adhesive. Thermocouples were attached to

61

the ns in roughly the center of the solar panel, where the temperature was previously taken without the heat sinks. Since two dierent heat sinks were manufactured, phases 1 and 2 were repeated for both sets. This gave an additional twleve sets of data for each heat sink, which increased the total number of data sets to 36. The data was once again exported to Excel in order to compare the temperature and eciency values with the data obtained from the bare panel. This also allowed the data to be compared between the dierent n congurations in order to nd which design was most eective. Figure 5.27 depicts one of the heat sinks mounted in the box and being measured.

Figure 5.27: Testing Performed on Heat Sink Panel

62

Chapter 6 Results
There are three dierent sections to the entire data set. As can be seen in a previous chapter, analytic equations were used to calculate the steady state temperature of the solar panel under idealized conditions. Preliminary analytical results were rst found to discover the aects of pin ns and plate ns under dierent congurations. While the actual analytical results used a heat transfer coecient calculated using experimental data, the preliminary results were found using a spreadsheet that was set up using arbitrary values for the heat transfer coecient in order to see the temperature at each value. It was found that plate ns would give a lower temperature at any given heat transfer coecient for a conguration similar to pin ns (i.e. number of pin ns widthwise equal to number of total plate ns). After this discovery, coupled with the fact that pin ns would be signicantly more dicult and expensive to manufacture, led to the decision to only use plate ns in the experimental setup. A small section of this spreadsheet is shown in Figure 6.1, depicting where the geometry could be changed to see how the surface temperature would be aected for dierent congurations.

63

Figure 6.1: Section of Master Spreadsheet

6.1

Experimental Results

The experimental data was useful in many ways. On a fundamental level (and most importantly), the data was used to represent real-world results for comparison with an idealized situation predicted by the analytical calculations. The data was then used to compare the many dierent situations in the experimental setup. For each irradiance tested, and for each individual solar panel, the results could be compared between the dierent cooling methods. Secondly, the three dierent solar panels could be compared in identical states. This provided evidence as to which heat sink provided the largest benet. It also showed how the application of either heat sink would provide a higher amount of cooling for a bare panel under the same conditions. Lastly, the data between each irradiance could be compared. This was done in an attempt to show that the cooling methods should provide the same benet, independent of the amount of the suns radiation. While many dierent values were recorded for each case, tables were made showing only the steady-state temperature, power output at steady-state, and the eciency of the solar panel. The tables for the three dierent
W W W irradiances of 1000 m 2 , 1250 m2 , and 1500 m2 are shown in Tables 6.1, 6.2, and 6.3,

respectively.

64

W Table 6.1: Eect of Temperature on Power Output/Eciency for Irradiance = 1000 m 2

Bare Panel

Sink 1

Sink 2

Tss ( ) Pss (W) Tss ( ) Pss (W) Tss ( ) Pss (W)

Wind 2.54 m/s 32.0547 3.31614 4.32844 28.7954 3.3087 4.31872 30.3525 3.59043 4.68647

Wind 1.37 m/s 33.2057 3.17732 4.14724 29.8608 3.33365 4.3513 31.242 3.64195 4.75371

Wind 1.12 m/s 34.128 3.19338 4.1682 31.2726 3.60498 4.70545 32.0984 3.30321 4.31156

Wind = 0 m/s 60.1916 2.81972 3.68048 64.4059 3.06134 3.99586 65.5033 2.91927 3.81042

W Table 6.2: Eect of Temperature on Power Output/Eciency for Irradiance = 1250 m 2

Bare Panel

Sink 1

Sink 2

Tss ( ) Pss (W) Tss ( ) Pss (W) Tss ( ) Pss (W)

Wind 2.54 m/s 33.3608 4.40317 4.59784 27.6544 4.29869 4.48874 30.2154 3.57338 3.73137

Wind 1.37 m/s 34.4098 4.37605 4.56952 28.9918 4.25768 4.44592 31.4584 3.53475 3.69103

Wind 1.12 m/s 35.3618 4.18621 4.37129 30.1841 4.35763 4.55029 32.5191 3.86096 4.03166

Wind = 0 m/s 65.9052 3.92068 4.09402 67.1114 3.9708 4.14636 68.967 3.16149 3.30127

W Table 6.3: Eect of Temperature on Power Output/Eciency for Irradiance = 1500 m 2

Bare Panel

Sink 1

Sink 2

Tss ( ) Pss (W) Tss ( ) Pss (W) Tss ( ) Pss (W)

Wind 2.54 m/s 36.2508 5.16061 4.49064 29.6765 5.85221 5.09246 34.1372 4.98171 4.33497

Wind 1.37 m/s 37.3939 5.0818 4.42206 31.358 5.97494 5.19925 34.2485 5.13099 4.46487

Wind 1.12 m/s 37.2835 5.11238 4.44868 32.8838 5.92447 5.15533 36.0272 5.59465 4.86833

Wind = 0 m/s 70.8141 4.37274 3.80505 74.8403 5.00712 4.35707 76.4637 4.40843 3.83612

As discussed previously, there exists a linear correlation between the power output of a solar panel and the surface temperature. Namely, the power output drops over time proportionally to the rise in surface temperature. The experimental results obtained in this thesis showed this to be true. Figure 6.2 shows the temperature data obtained for the bare panel. The rising temperatures in the graph are due to decreases the wind speed. The data is shown for all three irradiances. Equations for the linear relationships for each irradiance are shown next to each trend line. These equations can be used to predict the power output for temperatures outside of the range tested. 65

Figure 6.2: Correlation of Power Output and Surface Temperature for Bare Panel

66

The experimental results were used to nd the heat transfer coecients so they can be used to compare the results to other methods. Tables 6.4, 6.5, 6.6, and 6.7 show the lm temperature, Nusselt number, and heat transfer coecients for each of these cases. To see how the calculations were performed, refer to Subsections 3.5.1, 3.5.2, 3.5.3, and 3.5.4, respectively. Table 6.4: Heat Transfer Coecients for Bare Panel, Forced Convection I
W m2

V (m/s) 2.54 1.32 1.12 2.54 1.32 1.12 2.54 1.32 1.12

Ts ( ) 32.0547 33.2057 34.128 33.3608 34.4098 35.3618 36.2508 37.3939 37.2835

Tf ( ) 27.02735 27.60285 28.064 27.6804 28.2049 28.6809 29.1254 29.69695 29.64175

N ux 86.39 62.17 57.19 86.22 62.06 57.08 85.84 61.78 56.91

W m 2 K

1000 1000 1000 1250 1250 1250 1500 1500 1500

15.87 11.44 10.54 15.87 11.44 10.53 15.86 11.43 10.53

Table 6.5: Heat Transfer Coecients for Bare Panel, Natural Convection I
W m2

Ts ( ) 60.1916 65.9052 70.8141

Tf ( ) 41.0958 43.9526 46.40705

N ux 9.86 10.11 9.49

W m 2 K

1000 1250 1500

3.50 3.59 3.37

67

Table 6.6: Heat Transfer Coecients for Heat Sink Panels, Forced Convection Heat Sink 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 I
W m2

V (m/s) 2.54 1.32 1.12 2.54 1.32 1.12 2.54 1.32 1.12 2.54 1.32 1.12 2.54 1.32 1.12 2.54 1.32 1.12

Ts ( ) 28.7954 29.8608 31.2726 27.6544 28.9918 30.1841 29.6765 31.358 32.8838 30.3525 31.242 32.0984 30.2154 31.4584 32.5191 34.1372 34.2485 36.0272

Tf ( ) 25.3977 25.9304 26.6363 24.8272 25.4959 26.09205 25.83825 26.679 27.4419 26.17625 26.621 27.0492 26.1077 26.7292 27.25955 28.0686 28.12425 29.0136

N ux 0.81 0.43 0.37 0.82 0.44 0.37 0.81 0.43 0.37 3.24 2.31 2.08 3.24 2.30 2.08 3.22 2.29 2.07

W m 2 K

1000 1000 1000 1250 1250 1250 1500 1500 1500 1000 1000 1000 1250 1250 1250 1500 1500 1500

10.39 5.55 4.71 10.41 5.56 4.72 10.38 5.54 4.70 13.83 9.85 8.91 13.83 9.85 8.90 13.83 9.84 8.89

Table 6.7: Heat Transfer Coecients for Heat Sink Panels, Natural Convection Heat Sink 1 1 1 2 2 2 I
W m2

Ts ( ) 64.4059 67.1114 74.8403 65.5033 68.967 76.4637

Tf ( ) 43.20295 44.5557 48.42015 43.75165 45.4835 49.23185

N ux 14.64 10.16 7.24 10.09 10.23 10.47

W m 2 K

1000 1250 1500 1000 1250 1500

5.20 3.60 2.57 3.58 3.63 3.72

68

6.1.1

Fin Performance

The results showed that the heat sink panels performed better than the bare panel (with forced convection), as can be seen with the lower steady state temperatures. For natural convection, the heat sink panels actually performed worse, with a higher steady state temperature than the bare panel. Heat sink 1 can also be shown to perform slightly better than heat sink 2. However, the benets of heat sink 1 over heat sink 2 for all tests and the benets of the heat sinks versus the bare panel for forced convection were small. This leads to the conclusion that the added benet is insignicant considering the high cost of manufacturing the heat sinks. Figure 6.3 shows a graph which compares the performance for the two heat sink panels and the
W bare panel for an irradiance of I = 1000 m 2.

Figure 6.3: Graph Comparing the Performance of the Heat Sink Panels vs. Bare W Panel at Each Velocity for I = 1000 m 2

69

6.2

Analytical Comparison

Due to the complexity of the calculations, analytical results were found for only the case of natural convection with the bare panel. The temperatures for each irradiance in this case are shown in Table 6.8, in comparison with the computational and experimental results for the same case. Also shown are the percent dierences between analytical/computational and analytical/experimental. As can be seen, the percentage dierence is extremely small between the analytical and computational results. The experimental results, however, deviated signicantly from the other two methods. Refer to Figure 6.4 for a graph comparing the three methods. Table 6.8: Comparison of Results for Bare Panel, Natural Convection
W I (m 2) 1000 1250 1500

Tanal ( ) 165.66324 199.29633 241.76036

Tcomp ( ) 166.8 200.71 243.46

% Di Anal. Vs. Comp. 0.68 0.71 0.70

Texp ( ) 60.1916 65.9052 70.8141

% Di Anal. Vs. Exp. 93.40 100.60 109.38

Figure 6.4: Clustered Column Graph Comparing the Results from Three Methods for the Bare Panel with Natural Convection

70

6.3

Computational Results Vs. Experimental

The experimental and computational calculations were performed for each case. This includes the temperatures for each speed at the three irradiances, for the bare panel and both heat sinks. Tables 6.9, 6.10, and 6.11 show the temperatures compared for computational and experimental for the bare panel, sink 1, and sink 2, respectively. Table 6.9: Comparison of Computational and Experimental Results For Bare Panel
Bare 0 1.12 1.37 2.54 m/s m/s m/s m/s
W 1000 m 2 Comp. Exp. 166.800 60.1916 139.689 34.128 126.758 33.2057 102.735 32.0547 W 1250 m 2 Comp. Exp. 200.710 65.9052 167.478 35.3618 152.176 34.4098 122.512 33.3608 W 1500 m 2 Comp. Exp. 243.460 70.8141 196.076 37.2835 179.541 37.3939 140.830 36.2508

Table 6.10: Comparison of Computational and Experimental Results For Sink 1

Sink 1 0 1.12 1.37 2.54 m/s m/s m/s m/s
W 1000 m 2 Comp. Exp. 130.720 64.4059 49.519 31.2726 45.977 29.8608 36.793 28.7954 W 1250 m 2 Comp. Exp. 144.920 67.1114 55.640 30.1841 51.216 28.9918 39.518 27.6544 W 1500 m 2 Comp. Exp. 177.167 74.8403 61.779 32.8838 56.342 31.358 42.225 29.6765

Table 6.11: Comparison of Computational and Experimental Results For Sink 2

Sink 2 0 1.12 1.37 2.54 m/s m/s m/s m/s
W 1000 m 2 Comp. Exp. 112.244 65.5033 46.138 32.0984 42.997 31.242 37.937 30.3525 W 1250 m 2 Comp. Exp. 125.283 68.967 51.220 32.5191 47.923 31.4584 40.784 30.2154 W 1500 m 2 Comp. Exp. 137.343 76.4637 56.141 36.0272 52.409 34.2485 43.771 34.1372

6.4

Explanation of Results

As can be seen in the comparison tables, the computational and analytical results matched for the cases that the analytical results were found. However, the experimental results were in disagreement with the computational and analytical results. There 71

are many reasons why the experimental results are signicantly lower. Since the light intensity was adjusted with a dimmer in order to measure the temperature at the proper irradiance values (1000, 1250, and 1500
W ), m2

it is likely that the irradiance

measured using the pyranometer was higher than the radiation actually generated by the lights. The lights were in very close proximity to the solar panels/pyranometer. This causes the irradiance to vary widely as the pyranometer is moved into dierent positions around the panel. The maximum value could be seen to be directly underneath one of the light bulbs. As the pyranometer is moved, this irradiance value was shown to drop. A few readings were taken around the panel using the pyranometer after the initinal experiments, and an average was taken to get a better idea of the irradiance instead of the maximum value. However, it would have been more accurate to integrate the light intensity across the entire area of the panel. It can be shown that lowering the irradiance values by a specic factor in the analytical and computational calculations causes the temperatures to match up with the experimental data. Table 6.12 shows the results in ANSYS when the irradiance values were decreased by a . factor of approximately 4.5. The data is for sink 1, with an air velocity of 2.54 m s The factor by which the irradiance was decreased was adjusted until the temperatures approached the temperatures found in the experiment, until a factor of 4.5 was found to be the closest t. Table 6.12: Experimental Results Compared to Computational Results When Irradiance Decreased in ANSYS
W I(m 2) 1000 - 222 1250 - 277.5 1500 - 333

Texp ( ) 28.7954 27.6544 29.6765

Tcomp,adjusted ( ) 28.234 28.808 29.393

In the future, steps can be taken to avoid measuring an exaggerated irradiance. A simple x would be to use a light setup that is further away from the panels. This would cause the light intensity (and irradiance) to be more uniform across the solar 72

panel so it could be measured at a single point without worrying about variance as the pyranometer is moved around. Also, it would be helpful to measure the irradiance at as many spots as possible in order to calculate a more accurate average. Calculations would also be made that would integrate the irradiance across the panel. The panel also might not have had optimal thermal properties. The ARC layer (Anti-Reective Coating) may have not been applied properly to the panel. This would cause a large amount of the radition to be reected instead of absorbed, which would signinifantly lower the maximum temperatures of the panel when it reaches steady-state.

73

Chapter 7 Proposed Cooling Method

The data showed no signicant benet in using a heat sink versus a bare panel, as the increase in eciency was minimal. As an alternative, another cooling method was conceived. This would use a pump to send coolant into channels attached to the back of an array of solar panels (an array should be used, as the larger the array, the bigger the ratio between power saved to power used by a pump). The coolant would pump into a reservoir, where it would be dispersed to each channel that runs down the length of the array backing. At the bottom, the coolant would collect into another reservoir and be pumped into a cooling device to remove heat so it can repeat the cycle. A schematic showing the setup is shown in Figure 7.1. Ideally, the solar panel would operate at a constant temperature close to the ambient temperature. The tubes running along the back side can be assumed to have the same wall (surface) temperature as the solar panel backing. The temperature of the uid would approach the panel temperature asymptotically as it travels through the channels across the panel backing. Figure 7.2 is a diagram showing this relationship. Note the rate of exponential decay of the temperature dierence between the uid and surface in the direction of ow.

74

Figure 7.1: Pump Cooling Schematic

Figure 7.2: Fluid Temperature

If it was desired to design the channels so that the length enables the cooling uid to heat to a temperature within a degree or two of the constant cell temperature (which assures that the coolant is removing as much heat as possible), the surface temperature can be calculated. This would give an idea of the amount of heat that should be removed from the coolant once the heated cooling uid passes through a

75

cooling device. This calculation involves nding the log mean temperature dierence between the inlet and outlet temperatures of the coolant as they compare to the surface temperature. First, values for Ti and Te should be chosen. The value for Ti should be chosen to be relatively low, and the value for Te should be close to the ambient temperature (or the target temperature for the surface). The number of transfer units (NTU, see equation below) should be adjusted until the surface temperature is calculated to be within a degree or two of the exit temperature. The NTU of the uid is a dimensionless value that represents how eective the heat transfer system is [C engel and Ghajar, 2011]. As the NTU approaches a value of 2, the temperature of the surface should be very close to the exit temperature [C engel and Ghajar, 2011]. The NTU depends on the surface area, mass ow rate of the uid, constant average convection heat transfer coecient, and the specic heat capacity (N T U = calculation is performed as follows: Te eN T U Ti eN T U 1
hAs ). mc p

This

Ts =

(7.1)

In the next step, the mass ow rate should be calculated using the chosen value for NTU. This will allow the proper size pump to be chosen. Note that in this step, the mass ow rate may require the use of a pump that needs more power than desired. If this is the case, the calculations can be started over and iterations performed in order to nd the balance that suits the design requirements. This equation is derived from rearranging the NTU equation and solving for mass ow rate. hAs cp

m =

NT U

(7.2)

The last step involves calculating the heat transfer rate. This allows the size of the cooling device to be chosen, as it is known how much heat it must draw from the heated coolant. This calculation makes use of the log mean temperature dierence (LMTD). 76

The equations for LMTD and the heat transfer rate are shown below, respectively. Te Ti Te ln T i

Tlm =

(7.3)

= hAs Tlm Q

(7.4)

77

Chapter 8 Conclusion
In conclusion, the experiments proved that the heat sinks do indeed have a lowering eect on the solar panel temperature. Sink 1 was shown to perform slightly better than Sink 2. However, each sink only performed slightly better than the bare panel at wind speeds greater than zero. With stagnant air, the heat sinks actually performed worse as they began acting like an insulator. Even with constant wind, it would not be economically feasible to purchase a machined aluminum heat sink if the benet is as small as the experiments showed. Therefore, it is best to go with the previously mentioned cooling method using water and a pump. The price of a pump would be signicantly cheaper than purchasing an aluminum heat sink. The energy needed to power the pump would be minimal. Some of the energy saved from keeping the solar panel cooler could actually be used to power the pump. This would negate the need for an external power supply, thus simplifying the entire system. Perhaps the largest benet would arise from the systems ability to be adjusted in order to keep the solar panel surface as cool as desired. However, more research and experimental work must be done in the pump cooling method in order to solidify the analytical ndings.

78

Bibliography
[1] Agua Caliente Solar Project. First Solar. N.p., n.d. Web. 22 Apr. 2013. <http://www.rstsolar.com/en/Projects/Agua-Caliente-Solar-Project>. [2] AL-Sabounchi, A.M. Eect of Ambient Temperature on the Demanded Energy of Solar Cells at Dierent Inclinations. Renewable Energy 14 (1998): 149-55. Academic Search Complete. Web. 2 Aug. 2011. [3] ANSYS Fluent. ANSYS. ANSYS, Inc. Web. 4 Sept. 2013. [4] ANSYS Mechanical. ANSYS. ANSYS, Inc. Web. 11 July 2011. [5] Bergman, Theodore L., Adrienne S. Lavine, Frank P. Incropera, and David P. Dewitt. Fundamentals of Heat and Mass Transfer. 7th ed. Hoboken, NJ: John Wiley & Sons, 2011. Print. [6] C engel, Yunus A., and Afshin J. Ghajar. Heat and Mass Transfer: Fundamentals & Applications. 4th ed. New York: McGraw-Hill, 2011. Print. [7] Climate Data Summary. Western Regional Climate Center. search Institute. Web. 11 July 2011. Desert Re-

<http://www.wrcc.dri.edu/cgi-

bin/clilcd.pl?az23183>. [8] Desilvestro, Hans. What Physical Factors Aect Current-Voltage Characteristics of Dye Solar Cells? Dysol. Feb. 2008. Web. 19 Mar. 2012.

79

[9] Due, John A., and William A. Beckman. Solar Engineering of Thermal Processes. 3rd ed. Hoboken, NJ: Wiley, 2006. Print. [10] Ferguson, L. G., and L. M. Fraas. Theoretical Study of GaSb PV Cell Eciency as a Function of Temperature. Solar Energy Materials and Solar Cells 39 (1995): 11-18. ScienceDirect. Web. 2 Aug. 2011. [11] Fukushige, Shunichi, Kyoko Ichida, Takashi Minemoto, and Hideyuki Takakura. Analysis of the Temperature History of Amorphous Silicon Photovoltaic Module Outdoors. Solar Energy Materials & Solar Cells 93.6-7 (2009): 926-31. OhioLINK Electronic Journal Center. Web. 14 June 2011. [12] How Photovoltaic Solar Cells Work. Digital image. Total. N.p., n.d. Web. 22 Apr. 2013. <http://www.total.com/en/our-energies/alternative-energy/solarenergy/background/photovoltaic-922683.html>. [13] Ishibashi, Ken-ichi, Yasuo Kimura, and Michio Niwano. An Extensively Valid and Stable Method for Derivation of All Parameters of a Solar Cell from a Single Current-voltage Characteristic. Journal of Applied Physics 103 (2008): 0945071-94507-5. Academic Search Complete. Web. 19 Mar. 2012. [14] Joshi, Anand S., and Arvind Tiwari. Energy and Exergy Eciencies of a Hybrid Photovoltaic-thermal (PV/T) Air Collector. Renewable Energy 32 (2007): 2223241. Academic Search Complete. Web. 20 June 2011. [15] Lee, B., J. Z. Liu, B. Sun, C. Y. Shen, and G. C. Dai. Thermally Conductive and Electrically Insulating EVA Composite Encapsulants for Solar Photovoltaic (PV) Cell. EXPRESS Polymer Letters 2.5 (2008): 357-63. Academic Search Complete. Web. 15 June 2011.

80

[16] McQuiston, Faye C., Jerald D. Parker, and Jerey D. Spitler. Heating, Ventilating, and Air Conditioning: Analysis and Design. 6th ed. Hoboken, NJ: John Wiley & Sons, 2005. Print. [17] Mossayebi, Faramarz. Finite Element Analysis of Heat Conduction in the Absorber Plate of a Solar Collector. Thesis. Youngstown State University, 1987. Print. [18] Nishioka, L., T. Hatayama, Y. Uraoka, T. Fuyuki, R. Hagihara, and M. Watanabe. Field-test Analysis of PV System Output Characteristics Focusing on Module Temperature. Solar Energy Materials & Solar Cells 75.3/4 (2003): 665-71. Academic Search Complete. Web. 9 Aug. 2011. [19] Ogueke, N. V., E. E. Anyanwu, and O. V. Ekechukwu. A Review of Solar Water Heating Systems. Journal of Renewable and Sustainable Energy 1.4 (2009): 043106. Academic Search Complete. Web. 20 Apr. 2013. [20] Part II - Photovoltaic Cell I-V Characterization Theory and LabVIEW Analysis Code. National Instruments: Test, Measurement, and Embedded Systems. National Instruments Corporation, 4 Dec. 2009. Web. 3 Jan. 2012. [21] A Power Tower Power Plant. newable Energy. U.S. Digital image. Energy Eciency and ReWeb. 22 Apr. 2013.

Department of Energy, n.d.

<http://www.eere.energy.gov/basics/renewable energy/power tower.html>. [22] Rim, Seung-Bum, Shanbin Zhao, Shawn R. Scully, Michael D. McGehee, and Peter Peumans. An Eective Light Trapping Conguration for Thin-lm Solar Cells. Applied Physics Letters 91.243501 (2007). Web. 13 June 2011. [23] Roberts, Billy. Photovoltaic Solar Resource of the United States. Digital image. NREL. National Renewable Energy Laboratory, 20 Oct. 2008. Web. 11 July 2011. 81

[24] Royne, Anja, and Christopher J. Dey. Design of a Jet Impingement Cooling Device for Densely Packed PV Cells under High Concentration. Solar Energy 81 (2007): 1014-024. Academic Search Complete. Web. 10 July 2011. [25] Shahsavar, A., M. Salmanzadeh, M. Ameri, and P. Talebizadeh. Energy Saving in Buildings by Using the Exhaust and Ventilation Air for Cooling of Photovoltaic Panels. Energy and Buildings 43.9 (2011): 2219-226. Academic Search Complete. Web. 20 Apr. 2013. [26] Skoplaki, E., A. G. Boudouvis, and J. A. Palyvos. A Simple Correlation for the Operating Temperature of Photovoltaic Modules of Arbitrary Mounting. Solar Energy Materials & Solar Cells 92.11 (2008): 1393-402. OhioLINK Electronic Journal Center. Web. 14 June 2011. [27] Solar Thermal Energy. Digital image. Genesis Energy Solutions. N.p., n.d. Web. 22 Apr. 2013. <http://genesisenergysolutions.com/energy-solutions/solarthermal>. [28] Tayebjee, Murad J. Y., Louise C. Hirst, N. J. Ekins-Daukes, and Timothy W. Schmidt. The Eciency Limit of Solar Cells with Molecular Absorbers: A Master Equation Approach. Journal of Applied Physics 10.12 (2010): 124506. Academic Search Complete. Web. 13 June 2011. [29] Teertstra, P., M. M. Yovanovich, and J. R. Culham. Analytical Forced Convection Modeling of Plate Fin Heat Sinks. Journal of Electronics Manufacturing 10.4 (2000): 253-61. Web. 22 April 2013.

82

Appendix A Screenshots

83

Figure A.1: Isometric of Sink 1 in Solidworks

84

Figure A.2: Isometric of Sink 2 in Solidworks

85

Figure A.3: Light Control Program in BASIC Stamp

86

W m Figure A.4: Bare Panel Starting at Room Temperature, I = 1000 m 2,V = 0 s

87

W m Figure A.5: Bare Panel at Steady State, I = 1000 m 2 , V = 2.54 s

88

W m Figure A.6: Bare Panel at Steady State, I = 1000 m 2 , V = 1.37 s

89

W m Figure A.7: Bare Panel at Steady State, I = 1000 m 2 , V = 1.12 s

90

W m Figure A.8: Bare Panel at Steady State, I = 1000 m 2,V = 0 s

91