Sunteți pe pagina 1din 155

Graduate School ETD Form 9

(Revised 12/07)

PURDUE UNIVERSITY
GRADUATE SCHOOL
Thesis/Dissertation Acceptance

This is to certify that the thesis/dissertation prepared

By  

Entitled
         

     


For the degree of

Is approved by the final examining committee:


      
Chair
  

To the best of my knowledge and as understood by the student in the Research Integrity and
Copyright Disclaimer (Graduate School Form 20), this thesis/dissertation adheres to the provisions of
Purdue University’s “Policy on Integrity in Research” and the use of copyrighted material.

  


Approved by Major Professor(s): ____________________________________
____________________________________

Approved by:    


Head of the Graduate Program Date
Graduate School Form 20
(Revised 9/10)

PURDUE UNIVERSITY
GRADUATE SCHOOL

Research Integrity and Copyright Disclaimer

Title of Thesis/Dissertation:
         

For the degree of 


Choose  
your degree   

I certify that in the preparation of this thesis, I have observed the provisions of Purdue University
Executive Memorandum No. C-22, September 6, 1991, Policy on Integrity in Research.*

Further, I certify that this work is free of plagiarism and all materials appearing in this
thesis/dissertation have been properly quoted and attributed.

I certify that all copyrighted material incorporated into this thesis/dissertation is in compliance with the
United States’ copyright law and that I have received written permission from the copyright owners for
my use of their work, which is beyond the scope of the law. I agree to indemnify and save harmless
Purdue University from any and all claims that may be asserted or that may arise from any copyright
violation.

 
______________________________________
Printed Name and Signature of Candidate


______________________________________
Date (month/day/year)

*Located at http://www.purdue.edu/policies/pages/teach_res_outreach/c_22.html
ENERGY AND CFD MODELING OF ROOF INTEGRATED
PHOTO-VOLTAIC THERMAL SYSTEMS

A Thesis

Submitted to the Faculty

of

Purdue University

by

Soumya Anantharaman

In Partial Fulfillment of the

Requirements for the Degree

of

Master of Science in Civil Engineering

August 2012

Purdue University

West Lafayette, Indiana


UMI Number: 1529352

All rights reserved

INFORMATION TO ALL USERS


The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.

UMI 1529352
Published by ProQuest LLC (2012). Copyright in the Dissertation held by the Author.
Microform Edition © ProQuest LLC.
All rights reserved. This work is protected against
unauthorized copying under Title 17, United States Code

ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346
ii

सन्तोषः परमो लाभः सत्सङ्गः परमा ग�तः । �वचारः परमं �ानं शमो �ह परमं सख
ु म ्॥

Contentment is the highest gain, Good Company the highest course,

Enquiry the highest wisdom and Peace the highest enjoyment.


iii

ACKNOWLEDGEMENTS

I am grateful for the skills that I have acquired during these two years and I can
say with pride that the credit goes to my committee Chair and Advisor Dr.Panagiota
Karava. I would like to express my sincere gratitude to her for giving me the opportunity
to work in a relevant field, for imparting invaluable knowledge, for being patient and
guiding me throughout my stay at Purdue.

I have benefited greatly from the knowledge imparted by Dr.Horton and


Dr.Tzempelikos, it was invaluable for both – the research as well as for my career. I
would like to extend my thanks to both of them for I thoroughly enjoyed taking their
classes and discussing ideas with them. I would like to thank Dr. Luis Miguel Candanedo
who helped greatly by providing experimental data and valuable insights on the results
analysis. I also thank my fellow lab mates at Purdue: Jianjun Hu, Siwei Li and Ying
Chieh for the stimulating discussions and support.

Most importantly, I would like to thank my family - my parents: Jayanthi and


Anantharaman, brother: Arjun and grandparents: Kamala and Seetharaman for their
unconditional love and endless support. I would also like to extend my heartfelt thanks to
my boyfriend, Rangasayee for being completely understanding and encouraging me all
through; to all my friends, most importantly –Shirisha, Swetha, Rohini, Yun, Susa &
Chandru for making my stay at Purdue so much more enjoyable. Last but not the least - I
would like to thank God almighty, for none of this would have been possible without his
blessing.
iv

TABLE OF CONTENTS

Page

LIST OF TABLES ............................................................................................................ vii


LIST OF FIGURES ......................................................................................................... viii
NOMENCLATURE .......................................................................................................... xi
ABSTRACT ............................................................................................................ xvi
CHAPTER 1. INTRODUCTION ................................................................................. 1
1.1 Introduction ............................................................................................... 1
1.2 Background ............................................................................................... 1
1.3 Motivation ................................................................................................. 3
1.4 Problem Statement .................................................................................... 5
1.5 Objective and Scope .................................................................................. 6
1.6 Overview of the Thesis ............................................................................. 7
CHAPTER 2. LITERATURE REVIEW ...................................................................... 9
2.1 Introduction ............................................................................................... 9
2.2 Forced Convection Studies........................................................................ 9
2.3 Studies on PV/T Systems ........................................................................ 14
2.4 Numerical Studies of Heat Transfer in Channels .................................... 19
2.5 Mixed Convection Studies ...................................................................... 20
2.6 Effect of Turbulence................................................................................ 26
2.7 Limitations of Existing Models ............................................................... 27
CHAPTER 3. NUMERICAL STUDY OF CONVECTIVE HEAT TRANSFER IN
PV/T SYSTEMS ............................................................................................................. 29
3.1 Introduction ............................................................................................. 29
v

Page

3.2 Experimental Setup ................................................................................. 29


3.3 Numerical Approach ............................................................................... 30
3.3.1 Computational Domain, Grid & Near Wall Treatment .....................30
3.3.2 Solution Parameters...........................................................................31
3.3.3 Turbulence Model ..............................................................................33
3.3.4 Boundary Conditions .........................................................................36
3.3.5 Grid Independence .............................................................................39
3.4 Results and Analysis ............................................................................... 39
3.4.1 Verification of Temperature Profile in the Smooth Channel with
Experimental Data ...................................................................................... 39
3.4.2 Velocity Profile Verification for Asymmetric Heating .......................48
3.4.3 Average Nusselt Number for Forced Convection ..............................48
3.4.4 Effect of Grashof Number ..................................................................51
3.4.5 Average Nusselt Number for Mixed Convection ................................55
3.4.6 Comparison of New Correlations with Previous Results ..................56
3.4.7 Effect of Tilt Angle and Channel Length............................................60
3.5 Framing Effect on Nusselt Number ........................................................ 62
3.5.1 Average Nusselt Number for Forced Convection ..............................64
3.5.2 Average Nusselt Number for Mixed Convection ................................64
CHAPTER 4. TRANSIENT ENERGY MODEL FOR OPEN LOOP PV/T
SYSTEMS ............................................................................................................. 67
4.1 Introduction ............................................................................................. 67
4.2 Experimental Setup and Data Collection ................................................ 67
4.3 Modeling Approach................................................................................. 70
4.3.1 Necessity for Transient Solution ........................................................71
4.3.2 Solar Geometry – Computing the Incidence Angle............................71
4.3.3 Effect of View Factor and Inlet Air Temperature ..............................75
4.3.4 Effect of Interior Radiative Heat Transfer Coefficient ......................77
4.3.5 Effect Convective and radiative Heat Transfer Coefficient ...............78
vi

Page

4.3.6 Interior Convective Heat Transfer Correlation .................................80


4.3.7 Transient Model - Solution ................................................................82
4.4 Results – Model Performance ................................................................. 86
4.5 Parametric Analysis................................................................................. 92
4.5.1 Effect of Exterior Heat Transfer Coefficient ......................................92
4.5.2 Effect of Interior Heat Transfer Coefficient – Bottom .......................93
4.5.3 Effect of Capacitance & R-value of PV module .................................95
4.6 Case Study ............................................................................................... 99
4.6.1 End Energy Use of Eco-Terra house .................................................99
4.6.2 Control Logic ...................................................................................101
CHAPTER 5. CONCLUSIONS AND RECOMMENDATIONS ............................ 105
5.1 Introduction ........................................................................................... 105
5.2 Summary of the Research ..................................................................... 105
5.3 Contributions of the Research ............................................................... 106
5.4 Recommendations for Future Work ...................................................... 107
LIST OF REFERENCES ................................................................................................ 109
APPENDICES
Appendix A Solar Geometry and Perez Model ......................................................... 123
Appendix B Transient Model .................................................................................... 127
Appendix C End Use Energy Model Data................................................................. 133
vii

LIST OF TABLES

Table ..............................................................................................................................Page
Table 2.1 Correlations for forced convection ................................................................... 12
Table 2.2 Correlations for mixed convection ................................................................... 24
Table 3.1 Average outlet air temperature of various turbulence models .......................... 34
Table 3.2 Impact of change in inlet TI on outlet air temperature ..................................... 38
Table 3.3 Grid independence Study .................................................................................. 39
Table 4.1. Components listed from top of the PV module to the bottom ......................... 69
Table 4.2 Components listed from top of the PV module to the bottom .......................... 95
viii

LIST OF FIGURES

Figure .............................................................................................................................Page
Figure 1.1. Distribution of electricity consumption, US-EIA(2010) .................................. 2
Figure 1.2. Graph charting the increase in use of PV in residential sector over 20 years &
Pie chart representing end use distribution of electricity in residential buildings
(US-EIA, 2011) ....................................................................................................... 3
Figure 1.3. Schematic of a typical air-based open-loop PV/T system ................................ 5
Figure 1.4. I-V curve for a PV module and the effect of temperature on the maximum
power point at same irradiation level (DGS, 2005) ................................................ 5
Figure 2.1 BIPV/T system – simplified model by Bazilian and Prasad (2002)................ 16
Figure 3.1. Schematic of PV/T of Eco-Terra House (Adapted from Chen, 2009) ........... 30
Figure 3.2. Grid with resolution near the wall .................................................................. 30
Figure 3.3. Plot of y* values on the upper wall for different air speed ............................. 32
Figure 3.4 Boundary condition used in the CFD model ................................................... 37
Figure 3.5 Variation of air temperature due to TI ............................................................. 38
Figure 3.6 Placement of thermocouples in the channel .................................................... 40
Figure 3.7 Temperature rise for UHF boundary condition in CFD .................................. 41
Figure 3.8 Variation of air temperature at mid-depth along the length of the channel for
different flow rates (Re = 1284, 2479, 4094, 7097) - experimental and CFD
results .................................................................................................................... 43
Figure 3.9 Variation of air temperature at mid-depth along the length of the channel and
surface temperature for velocity = 1.4m/s (Re = 7097) ........................................ 44
Figure 3.10 Variation of air temperature at mid-depth, along the length of the channel for
velocity = 1.4m/s (Re = 7097) .............................................................................. 44
ix

Figure ............................................................................................................................Page
Figure 3.11Variation of temperature across the depth of the channel for velocity = 1.4m/s
(Re = 7097) ........................................................................................................... 46
Figure 3.12 Variation of temperature across the depth of the channel for different flow
rates (Re = 1284, 2479, 4094, 7097) - CFD and experimental results ................ 46
Figure 3.13 Variation of temperature across the depth of the channel at the outlet for
different flow rates (Re = 1284, 2479, 4094, 7097) - CFD and experimental
results .................................................................................................................... 47
Figure 3.14 Velocity profile variation across the depth at various flow rates .................. 48
Figure 3.15 Average Nusselt versus Reynolds profile for forced convection .................. 50
Figure 3.16 Petukhov’s upper limit curves (forced convection range) ............................. 53
Figure 3.17 Petukhov’s upper limit curves (mixed convection range) ............................. 53
Figure 3.18 Metais and Eckert (1964) map for vertical tubes .......................................... 54
Figure 3.19 CFD data for Metais and Eckert map for various heat fluxes ....................... 54
Figure 3.20 Nusselt versus Re/Gr0.5 for mixed convection............................................... 56
Figure 3.21 Comparison of forced convection correlations.............................................. 57
Figure 3.22 Comparison of mixed convection correlations .............................................. 59
Figure 3.23 Variation of local Nusselt number in a smooth channel................................ 61
Figure 3.24 Variation of local Nusselt number in a channel with wooden framing ......... 61
Figure 3.25 Photo of the wood framing in Eco-Terra house (Chen, 2009); Sketch of PV/T
& Close-up of channel with framing .................................................................... 63
Figure 3.26 Average Nusselt versus Reynolds profile in a channel with and without
wooden framing for forced convection ................................................................. 65
Figure 3.27 Average Nusselt versus Gr/Re0.5 profile in a channel with and without
wooden framing for mixed convection ................................................................. 66
Figure 4.1 (a) Experimental PV/T setup replicating the PV/T system of the Eco-Terra
house with and without amorphous PV modules attached (Candanedo, 2010); (b)
Photograph of the PV/T roof in Eco-TerraTM house (Chen, 2009) ....................... 68
Figure 4.2 Cross section details of PV/T system .............................................................. 68
x

Figure ............................................................................................................................Page
Figure 4.3 Setup of the experiment, Adapted from the experiment at Concordia
(Candanedo, 2010) ................................................................................................ 70
Figure 4.4 Comparison between measured ambient and inlet air temperature, February
17th, 2009 .............................................................................................................. 77
Figure 4.5 BIPV/T thermal network model (Adapted from Candanedo, 2010) ............... 82
Figure 4.6 Measured average speed in the channel and CHTC, February 17th, 2009 ...... 87
Figure 4.7 Measured wind speed and derived heat transfer coefficients, February 17th,
2009....................................................................................................................... 87
Figure 4.8 Comparison of outlet air temperatures for smooth channel, February 17th,
2009....................................................................................................................... 89
Figure 4.9 Comparison of outlet air temperatures for channel with framing, November
11th, 2009 .............................................................................................................. 90
Figure 4.10 Thermal and electric output for different correlations for February 17th,
2009....................................................................................................................... 91
Figure 4.11 Thermal and electric output for smooth and ribbed channels ....................... 91
Figure 4.12 Parametric analysis for determining the impact of exterior CHTC on Outlet
Air and PV Efficiency........................................................................................... 93
Figure 4.13 Parametric analysis for assessing the impact of bottom CHTC, February 17th
2009....................................................................................................................... 94
Figure 4.14 Parametric analysis for assessing the effect of thermal capacitance ............. 96
Figure 4.15 Parametric analysis for R-value – variation of outlet temperature, variation of
the PV node temperature and overall efficiency on February 17th, 2009 ............. 97
Figure 4.16 Thermal and Electric efficiency of the PV/T system on January 12th, 2008 for
Eco-Terra house, Quebec .................................................................................... 100
Figure 4.17 Control logic for operation of fan behind a PV/T system ........................... 102
Appendix Figure
Figure C1 Graph depicting variation of solar irradiation and outside air temperature on
Jan.12th ,2008………………………………………………………………… 133
Figure C2 Graph depicting wind speed and wind direction on Jan 12th 2008………. 134
xi

NOMENCLATURE

Symbols

A Area
Ac Area of the PV/T system exposed to the sun
Acv Area of the control volume
AM Air mass modification parameter
Cpv Thermal capacitance of PV per unit area
Cmix Thermal capacitance of mix layer per unit area
cpv Specific heat of air
Cμ Coefficient used in wall treatment for SST k-ω model
cp Specific heat capacity
d Depth of the control volume
Dh Hydraulic diameter
Dω Cross diffusion term
et Equation of time, minutes
Epv Electrical energy produced by PV
Fij View factor between surfaces
F1,F2 Blending functions for the SST k-ω turbulence model
f Friction factor
g Acceleration due to gravity
G Solar Irradiation per unit area
Gk Generation of turbulence kinetic energy due to the mean velocity gradients
Gω Generation of ω
xii

Gr Grashof number
Gz Graetz number
h Hour angle
ho Exterior convective heat transfer coefficient per unit area
hr Interior radiative heat transfer coefficient per unit area
hct Interior convective heat transfer coefficient near top surface
hcb Interior convective heat transfer coefficient near bottom surface
hci Interior convective heat transfer coefficient of the attic space
i Time step indicator
IT Total global irradiation
k Thermal conductivity
k Turbulent kinetic energy used in SST k-ω model
kp Turbulent kinetic energy at center point P of wall- adjacent wall
L Length of the channel
Lstd Standard meridian
Lloc Local meridian
m mass flow rates
Mref Standard air mass modifier at reference condition
M Air mass modifier at given position of the sun
Mw Molecular weight of gas
n Day number of the year
Nu Nusselt number
Pelect Electrical power per unit area
Pop Operating pressure
Pr Prandtl number
qtop Heat flux on top of the channel
qsky Convective and radiative loss to sky per unit area
qcv Heat recovered in control volume per unit area
Q Heat transfer
Qin Heat removed by convection and radiation inside the channel
xiii

Qout Heat removed by convection and radiation outside on PV surface


R Universal gas constant
Ra Rayleigh number
Re Reynolds number
Rinsu Thermal resistance of insulation and plywood layer
Rpv Thermal resistance of the PV module (Sum of Rtefzel, Rmix)
Sconst Sutherland constant
S Actual absorbed solar radiation
Sk User defined source term for turbulence kinetic energy
Sref Absorbed solar radiation at reference condition
Sω User defined source term for turbulence specific dissipation rate
T Temperature
Tma Average air temperature
To Ambient air temperature
Tdp Dew point temperature
Tsky Sky temperature
Tpv PV cell temperature
Tpv_top PV temperature on top of the module
Tpv_bot PV temperature on bottom of the module/ top of the channel
Tpv_mid PV temperature at the middle of the module
Tinsu Interior insulation temperature
Tbulk Bulk air temperature inside the channel
t Time from midnight
tapp Apparent solar time
uτ Friction velocity
u Velocity vector component along i & j coordinates
Vwind Outside wind speed
Vch Air speed in the channel
W Width of the control volume
x Distance from the inlet
xiv

Yk Dissipation of k due to turbulence


Yω Dissipation of ω due to turbulence
y+ Dimensionless wall distance (normal)
*
y Dimensionless wall distance (normal)
yp Distance from the nearest cell to the wall
Zg Reference height of the wind direction
Z Height of the mounted PV

Greek Symbols

α Solar absorptance
α1, α2 Constant for calculating turbulent viscosity in the SST k-ω model
α* Coefficient for low Reynolds number correction for the SST k-ω model
α*∞,β1 Constants used in calculating α*
αw Power law exponent
alpha Solar altitude angle
β Volumetric thermal expansion coefficient
βmp Performance reduction factor of the PV module
β Intermediate variable in equation of time
Δx Length of the control volume
δ Declination angle
ηstc Efficiency of the module under standard testing conditions
ηpv Efficiency of the module
ε Longwave emissivity of the surface
θi Incidence angle
ψ Surface azimuth angle
φ Solar Azimuth
θ Inclination angle of the duct
γ Surface azimuth angle
θz Zenith angle
xv

τ Solar transmittance
ν Kinematic viscosity
ρ Reflectance of the PV surface
ρf Density of the fluid
μ Fluid viscosity
μo Reference viscosity
μt Boundary layer friction velocity
ω Specific dissipation rate of turbulent kinetic energy
σ Stefan Boltzman constant
σij Stress tensor due to molecular viscosity
σω Turbulent Prandtl number of turbulent specific dissipation rate
σk Turbulent Prandtl number of turbulent kinetic energy
ϕ Argument in the blending functions
xvi

ABSTRACT

Anantharaman, Soumya. M.S.C.E., Purdue University, August 2012. Energy and CFD
Modeling of Roof Integrated Photo-Voltaic Thermal Systems. Major Professor:
Panagiota Karava.

This Thesis investigates innovative methods to integrate actively cooled


Photovoltaic/Thermal (PV/T) systems into the sloped roofs of houses with the additional
function of recovering thermal energy in addition to electricity generation. An open-loop
PV/T system is considered with a typical Photovoltaic (PV) array on top of an inclined
rectangular smooth or ribbed channel that corresponds to a real installation. 2-D steady
Reynolds-Averaged Navier Stokes (RANS) Computational Fluid Dynamics (CFD)
simulations are performed in FLUENT to develop dimensionless correlations for the
convective heat transfer coefficient from the surfaces of the channel to the air for forced
and mixed convection regimes, expressed as Nusselt number (Nu). A transient thermal
network model for a PV/T system is developed in MATLAB using the developed
correlations and it is validated with experimental data from a full-scale experimental set-
up. The transient energy model is used to determine the air temperature at the outlet of
the channel, the surface temperature of the PV panel and the energy output. Results
indicate prediction of the outlet air temperature with better accuracy since the present
work includes the effect of buoyancy in the air flow within the channel at a lower air
velocity. The model can be used to assess and optimize the performance of different roof-
integrated PV/T systems considering different PV technology and it can be incorporated
into existing building energy simulation platforms to support the widespread adoption of
these systems into building design.
1

CHAPTER 1. INTRODUCTION

1.1 Introduction

In this chapter, an introduction to the research is provided. Section 1.2 details the
background behind the energy consumption in the US followed by the need for
renewable energy. An introduction to solar energy is provided followed by a discussion
about PV systems and their advantages. Section 1.3 discusses the motivation behind the
research followed by Section 1.4 that describes the problem statement where a simplified
energy balance of a Building-Integrated Photovoltaic/Thermal (BIPV/T) system is
explained. The objectives and scope of the Thesis are defined in Section 1.5 and the
layout of the Thesis is listed in Section 1.6.

1.2 Background

In the United States, Buildings account for 42% of the energy (source) consumed
with 23% being used by residential and 19% being used by commercial buildings
(Fig.1.1, US-EIA, 2010). To augment the growing need for energy and to reduce the
impact on the environment, renewable energy sources are now widely used.
Among renewable energy sources, solar energy is available abundantly all across
the globe. The earth receives 174 PetaWatts (PW) of incoming solar radiation of which
47% reaches the earth’s surface. This solar energy can be harnessed by use of many
technologies, the most popular of them being the Photovoltaic (PV) panels for generation
of electricity that can be used for power small houses, appliances and vehicles and solar
thermal used for harnessing the thermal energy to heat water, space heating and
producing steam for large scale electricity generation. In the Annual Energy Review by
EIA (2010), it was reported that there was a rapid growth in the photovoltaic industry
from 1990-2010 (Fig.1.2).
2

Figure 1.1. Distribution of electricity consumption, US-EIA(2010)

The Solar Energy Industries Association (SEIA) report tallies all types of solar
energy, and in 2007 the United States installed 342 MW of solar photovoltaic (PV)
electric power, 139 thermal megawatts (MWth) of solar water heating,
762 (MWth) of pool heating, and 21 (MWth) of solar space heating and cooling (DOE-
EERI, 2008). In a report by CleanTech (2008), it is projected that nearly 2% of the
nation's electricity comes from concentrating solar power systems, while PV systems will
provide more than 8% of the nation's electricity which is nearly 50,000 MW produced by
PV systems and more than 6,600 MW produced by concentrating solar power. As noted
in the report, solar power has been expanding rapidly in the past decade, growing at an
average pace of 40% per year. The split up of the total energy consumed by the
residential buildings is given in Figure 1.2b, with space and water heating accounting for
an average of 20% of the energy consumed in residential buildings. (Details being
provided online on their website:
http://www.eia.gov/emeu/recs/recs2001/enduse2001/enduse2001.html).

Measures such as using passive solar guidelines, orienting the buildings to face
south, using high insulation materials in the building envelope have aided in reducing the
heating loads, however this does not eliminate heating loads entirely.
Photovoltaic/Thermal (PV/T) systems serve a dual purpose of producing electricity and
thermal energy and hence are the focus of the present research.
3

1.3 Motivation

PV panels mounted on roofs experience very high temperatures due to absorption


of heat that is not converted to electricity. A study by Cross (1994) states that a fall in
panel efficiency in the order of 0.5% per degree higher than the standard operating
temperature of 25oC significantly reduces the panel rating (Fig. 1.4). Roof integrated
PV/T systems are functional components installed as a part of the building envelope on
the roof. While the module itself is used for electricity production, a part of the solar
energy which heats the modules is removed by circulating a fluid (air/water) beneath the
panels.

Figure 1.2. Graph charting the increase in use of PV in residential sector over 20 years &
Pie chart representing end use distribution of electricity in residential buildings (US-EIA,
2011)
There are two types of PV/T systems: open loop and closed loop. In both open
and closed loop PV/T systems, a fluid (usually air) is circulated beneath the PV module,
and it is later used to provide thermal energy to one or more functions in the building or it
is directly exhausted to the ambient air providing only a PV cooling function in this case.
The major difference between these two systems is that, the fluid circulated beneath the
PV module is not recirculated back in the open loop configuration. Open loop systems
have an inherent advantage over closed loop systems as they allow potential for fresh air
preheating. Closed loop air systems cause over heating of the PV modules thereby
4

reducing the efficiency. Studies by King (2000) indicate that degradation (due to
overheating of the modules) is more likely to occur in closed loop systems, more severely
in hot climates where there is significant potential for electricity generation with PV.
Though the outlet air temperatures are lower, open loop systems are preferred as the
combined thermal and electrical efficiency of the PV/T systems is higher. Figure 1.3
illustrates the schematic of a PV/T system and the terms which are used in the energy
balance. The PV module converts a portion of the incoming solar energy to electrical
energy (~6-20%) depending on the type of cell used), dissipating the rest of energy as
heat. An energy balance of a PV/T system can be given as:

𝛼𝐺 = 𝐸𝑝𝑣 + 𝑄𝑖𝑛 + 𝑄𝑜𝑢𝑡


1.1
Where 𝛼𝐺 represents the solar radiation absorbed by the PV panel, 𝐸𝑝𝑣 is the electrical
energy produced, 𝑄𝑖𝑛 is the heat removed by convection to air and radiation to bottom of
the panel by the circulating fluid beneath the panel, 𝑄𝑜𝑢𝑡 is the heat removed from the top
by convection and radiation due to the wind flowing over the panel.

Modeling of open loop PV/T systems is subject to several constraints such as their
complex physical geometry (e.g. framing elements, such as wooden battens used for
structural support), varying heat fluxes, thermal resistance and capacitance of the
modules, heating asymmetry, and Re numbers that span over laminar, transitional or
turbulent flow regimes. Models of differently complexity, steady and transient,
emphasizing different phenomena have been developed over the years. This research
aims to study the internal convective heat transfer coefficient in the channel, behind an
open loop PV/T system.
5

Figure 1.3. Schematic of a typical air-based open-loop PV/T system

Figure 1.4. I-V curve for a PV module and the effect of temperature on the maximum
power point at same irradiation level (DGS, 2005)

1.4 Problem Statement

This Thesis investigates innovative methods to integrate actively cooled PV


systems into the sloped roofs of houses in order to enhance their thermal performance,
optimize their overall performance and reduce uncertainty and risk. The PV/T system
under consideration is open-loop – where the air passing beneath the channel is the
6

outdoor air and the channel formed is between the mounted PV module and the roof
beneath. The air circulated in this channel beneath the PV serves a dual purpose of
improving the efficiency of the photovoltaic module as well as being used for purposes
such as producing domestic hot water, space heating and drying clothes (Chen, 2009). In
order to assess the overall performance of the PV/T system, it is important to predict the
convective heat transfer process behind the PV/T system with accuracy. This Thesis aims
to address this problem by modeling the convective heat transfer behind the PV/T system
and predicting correlations using a commercial computational fluid dynamics (CFD)
package, and using these as an input to a transient energy model (developed in
MATLAB) to predict the value of the temperature of air at the outlet as well as the PV
cell temperature. The objectives and scope are systematically listed below.

1.5 Objective and Scope

The overall objective of this Thesis is to develop models for open loop roof-
integrated PV/T systems in order to predict the variation of electrical and thermal
efficiency and to assess their energy performance. The specific objectives are enumerated
below:

1. Predict the convective heat transfer coefficients for a forced and mixed
convection regime in a PV/T channel with and without wooden battens and
develop generalized correlations using a CFD model verified with experimental
data.
2. Verify the newly developed correlations by comparison with other existing
correlations.
3. Utilize the new correlations in a lumped parameter transient model and assess
their performance by comparison to experimental data.
4. Perform a parametric analysis using the lumped parameter transient model and
develop concepts for enhanced thermal management and energy yield.
7

The Thesis investigates PV/T systems integrated in the roof of houses inclined at
30-45° considering (a) realistic geometry and configurations, (b) amorphous silicon (a-Si)
PV modules, which are widely available in the market, (c) heat fluxes and Reynolds
number range for the flow beneath the PV modules that correspond to real installations.
This Thesis extends previous work (Candanedo, 2010; Chen, 2009) on roof-integrated
PV/T systems in two ways:
1. Develops energy models with new convective heat transfer coefficient
correlations for PV/T systems based on CFD simulations.
2. It accounts for the effect of buoyancy on the convective heat transfer in the PV/T
channel which is significant for lower air flow rates.

These models can be used in potential future extensions of this work to support (a)
the design optimization of PV/T systems, using new PV cells with high efficiency such as
Copper Indium Gallium Selenide (CIGS) embedded into roofing materials, and the
development of prefabricated BIPV roof modules suitable for new construction and
retrofits; (b) the building integration of BIPV/T with HVAC systems (such as hybrid
solar-assisted heat pumps) through the development of advanced predictive control
strategies. Eventually, the developed models will be incorporated into existing building
energy simulation platforms to support the widespread adoption of BIPV/T systems into
building design.

1.6 Overview of the Thesis

In Chapter 1, the background and motivation for this study – describing the
growing need for renewable energy and the necessity for use of ventilation under a PV/T
system are described followed by the presentation of the Thesis problem statement,
research objectives and scope of the study.
A literature review is presented in Chapter 2. Forced and mixed convection
studies for channel flow with different boundary conditions and orientation are reviewed
and the relevant correlations that are applicable to PV/T systems are tabulated
chronologically. Numerical and experimental studies on PV/T systems are also presented.
8

Chapter 3 focusses on the numerical study of convective heat transfer in PV/T


systems. A CFD model is developed for a smooth channel for which the problem
parameters, modeling of the geometry, the boundary conditions involved and approach
are presented. This is followed by result analysis of the newly developed convective heat
transfer correlations by comparing it to other correlations from literature. An analysis is
also presented for a similar channel with wooden framing elements followed by a similar
comparative study with sources from literature.
Chapter 4 evaluates the performance of a transient open loop PV/T model with
the newly developed correlations for a channel with and without wooden framing. The
results are compared to experimental results obtained from tests conducted at Concordia
University, Montreal (Candanedo, 2010). Parametric analysis is performed to observe the
effect of exterior convective heat transfer coefficient, the capacitance and resistance of
the PV module on the electrical and thermal efficiency of the system. Lastly, an analysis
of the end energy use and a simple control logic are proposed.
Chapter 5 presents the summary of the conclusions, main research contributions
and recommendations for future work.
9

CHAPTER 2. LITERATURE REVIEW

2.1 Introduction

This chapter discusses the relevance of this research by presenting previous work
in the literature mainly focusing on flow through ducts and other PV/T and BIPV systems.
Section 2.2 presents a review on forced convection heat transfer in ducts and the
correlations employed followed by Section 2.3 which discusses the previous
experimental and numerical studies related to PV/T systems and Section 2.4 with focus
on numerical studies of heat transfer in channels. In Section 2.5 mixed convection studies
are detailed and Section 2.6 discusses the effect of turbulence, dealing with studies in
literature for ribbed channels and ducts. Finally, Section 2.7 discusses the limitation of
existing models.

2.2 Forced Convection Studies

When flow and heat transfer through heated ducts are considered, two boundary
conditions may be employed – Uniform Heat Flux (UHF) and Uniform Wall
Temperature (UWT). The heat transfer in a PV/T system is comparable to the heat
transfer in a duct, being heated from the top with UHF, the flow being classified either as
forced convection for a high fan speed, mixed convection for low fan speed and free
convection when the fan is switched off. For the case of flow categorized as forced
convection (with fan speed greater than 1m/s for this particular setup- explained in
sections 3 & 4), given below is a brief review of some papers that are relevant to this
research.
Martinelli (1947) conducted a study on heat transfer in a cylindrical tube with
UWT and turbulent flow of fluid (air, molten metals, etc). Based on earlier work by Von
10

Karman (1939) on heat transfer in the different layers near the wall, a Nusselt correlation
was developed that considers the use of the bulk air temperature to determine the heat
transfer coefficient and the friction factor. One of the main postulates of the Martinelli
equation is that it can explain the heat transfer behavior for different Prandtl numbers
(though the present study considers only air based open loop PV/T systems). It is
recommended by Kakac and Yener (1995) for turbulent flow in pipes and for parallel
plates.
Kays et al (1966) developed a correlation for a circular duct with uniform wall
heat flux. An expression for temperature distribution was developed for the laminar sub-
layer, buffer layer and turbulent core. Later, Kays and Crawford (1980) developed a
single equation for the temperature distribution and also a Nusselt correlation for a
similar duct. This equation was valid for Prandtl number between 0.5-5.0 and yielded
results in agreement with their previous work.
Sparrow et al (1966) investigated an asymmetrically heated rectangular duct for
10,000≤Re≤100,000 and Pr=1 and presented a Nusselt correlation. A longer thermal
entrance length was observed for an asymmetrically heated duct and decrease in Nusselt
number by a factor of 2-3 when compared to a symmetrically heated duct. Another
experimental work was performed by Tan and Charters (1969), who developed a
correlation for thermally fully developed turbulent flow for a horizontal rectangular
channel with asymmetric heating. The Reynolds number range was tested from 9500 to
22,000 and the heat flux was in the range of 800-900 W/m2. The results from Tan and
Charters experiment confirmed the findings from Sparrow’s experiment. Uniform wall
temperature boundary condition in both these studies mentioned above.
Petukhov (1963) developed a correlation which agrees within ±5% of the most
reliable heat transfer experiments (Kakac, 1987). This correlation too was developed for
symmetric heating in a vertical pipe. It was modified by Gnielinski (1976), extending it to
a larger Re range. It agrees with the Petukhov equation within -2% and +7.8%.
The Gnielinski correlation (1976) is based on comparisons with nearly 800
experimental measurements, 720 of which fall within ±20% of the predicted correlation.
Given the uncertainties of various investigations stemming from fluid property variation
11

and unaccounted effects like natural convection, this level of agreement is considered
satisfactory. For fully developed flows, the Nusselt numbers are valid over a wide range
of Re and Pr as shown in Table 2.1.
Dittus & Boelter (1930; 1985) developed separate correlations for heating and
cooling to account for the variation of fluid temperature. The correlations for
symmetrically heated vertical tubes were developed by averaging the work of various
researchers, valid for a large Re and Pr. Incropera and DeWitt (2006) recommend the use
of this correlation for Re>10,000 and L/Dh>10 and McAdams (1954) recommends this
correlation for 10,000<Re<120,000 and L/Dh>60. The correlations for forced convection
are enlisted in Table 2.1.
12

Table 2.1 Correlations for forced convection

Investigator Correlation Remarks

Average Nu, forced convection,


symmetrical heating where,
tw is the temperature of the wall,
tb the bulk air temperature,

𝑓 tc temperature at the center of the


𝑅𝑒. 𝑃𝑟. �
Martinelli 2 pipe.
𝑁𝑢 = 𝑡 − 𝑡
(1947) �𝑤 𝑏
� . 5. (𝑃𝑟 + 𝑙𝑛(1 + 5. 𝑃𝑟) + for air
𝑡𝑤 − 𝑡𝑐
𝑅𝑒 �𝑓 Pr=0.71 and 1000 ≤ Re ≤ 10000,
0.5. 𝑁𝐷𝑅. 𝑙𝑛 � �))
60 2
0.7<NDR<0.98.
(tw-tb)/(tw-tc) 0.78 to 0.84 for
2000<Re< 10000
(2.1)

Petukhov, 𝑓 Symmetric heating, vertical pipe.


(𝑅𝑒 − 1000). 𝑃𝑟
𝑁𝑢 = 2
2 Valid for 0.5≤Pr≤106 and
Kirillov, Popov �𝑓 3
𝐶 + 12.7
2
( Pr −1) 4000≤Re≤5x 105

(1963)
Where, C= 1.07+ 900/Re –[0.63/(1+10Pr)]

(2.2)

Average Nu, forced convection


and asymmetrical heating. Valid
Sparrow 𝑁𝑢 = 0.018 𝑅𝑒 0.8
𝑃𝑟 0.4
for 104≤ 𝑅𝑒 ≤105 Pr=0.7
(1966)

(2.3)
13

Table 2.1 Continued.

𝑓 Modified Petukhov equation, valid


(𝑅𝑒 − 1000). 𝑃𝑟
𝑁𝑢 = 2
Gnielinski 2 through 2300≤Re≤5x 106
�𝑓 3
(1976) 1 + 12.7 ( Pr −1)
2
(2.4)

�𝑓
𝑅𝑒. 𝑃𝑟
Kays & 𝑁𝑢 = 2 Valid for 0.5 ≤ Pr ≤ 5 and for
�𝑓
Crawford 0.833[2.25 ln �0.75 𝑅𝑒
2
� + 13.2 Pr −5.8 104≤Re≤5x 106.
(1980)
(2.5)

Developed for 0.7 ≤ Pr ≤ 120 and


Dittus-Boelter 2500≤Re≤1.24x 105
(1985) 𝑁𝑢 = 0.024 𝑅𝑒 0.8 𝑃𝑟 0.4 Symmetric heating and valid for
circular pipes.
(2.6)

Average Nu, forced dominated,


Tan and �−
0.03795𝐿𝑐
� asymmetrical heating. For
𝑁𝑢 = 0.0158𝑅𝑒 0.8 + (0.00181𝑅𝑒 + 2.92)𝑒 𝐷ℎ
Charters horizontal duct, Re>9500.
(Eicker, 2003)
(2.7)
14

The Martinelli, Petukhov and Gnielinski correlations (For Re>4000) were


evaluated with the Prandtl, Karman, Nikuradse (PKN) friction factor co-efficient (Kakac,
1987). This is a classical correlation with agreements between the prediction and the
results within 2% of experimental results and valid over a Reynolds number range
between 4000-107. The PKN friction factor, f, is given by the following formula:
1
= 1.7372 ln�𝑅𝑒. �𝑓� − 0.3946
�𝑓
2.8

2.3 Studies on PV/T Systems

This section deals with the review of experimental and numerical investigations
conducted related to PV/T systems, listed chronologically.
Karuter et al (1999) performed an experimental study to estimate the thermal
performance of a PV façade system using three different configurations and the findings
are as listed below:
1. Cell temperature increases by 20.7oC for PV elements without cooling
2. Actively cooled PV allows a reduction of cell operation temperature by up to
18oC at an air speed of 2 m/s.
3. If water is used a coolant behind the PV façade, it lowers operating cell
temperature by 20oC.
Brinkworth et al (1997) studied in detail the thermal regulation of PV/T cladding
with an aim to assess the PV performance and thermal output. A CFD model was
developed for a naturally ventilated vertical cladding and the results were compared to
experimental results and it was found that the flow was fully turbulent behind the PV
(~15-20%). It was concluded that the use of the model can be effective in selection of
basic design values such as the geometry of the duct in relation to the parameters studied.
Brinkworth et al (2000) modeled a naturally ventilated PV using a loop analysis where
the pressure drop of one full loop is equated to zero. The pressure drop across the system
due to friction is equated to the buoyancy forces (computed using the Boussinesque
approximation) where the frictional losses are a function of the viscous forces and
15

pressure loss occurring at the inlet and outlet. An analytical expression was formulated
for flow under a PV/T system and compared to experimental data from a solar simulation
lab in the UK and good agreement was observed. Postulates of this study are that it can
be adapted in a simple spreadsheet format but it has a main drawback of computing mean
velocity or the mean temperature since it is a 1D model. Good agreement was observed
for Grashof number till 105 beyond which distortion in velocity profile becomes
substantial.
Bazilian et al (2001) conducted a survey regarding BIPV/T systems. An
evaluation method was presented, that examined the energy and economic performance
of BIPV /T systems concluding that these systems had higher efficiency than regular
BIPV systems that generated electricity only. In another study conducted by Bazilian
(2002) using thermographic camera to evaluate the temperature gradient across BIPV/T
systems, it was concluded that it was an effective method to determine air leakage and
emissivity of the system. Brinkworth et al (2005) performed a theoretical study to assess
the optimal depth of the channel assuming constant heat flux on one side. It was
concluded that a value of length to hydraulic diameter (L/Dh) of 20 for ducts (smooth)
and 40 for very wide ducts is optimal for minimal temperature rise on the PV side. In yet
another study by the same author, Bazilian and Prasad (2002), a thermal network of the
BIPV/T system was presented (Fig.2.1), where the equations were solved using the
Engineering Equation Solver (EES) software. The model was validated with
experimental data and it was utilized to evaluate the yearly performance using data from
TMY2.
16

Figure 2.1 BIPV/T system – simplified model by Bazilian and Prasad (2002)

Tripanagnostopoulos et al (2001) conducted studies on hybrid PV/T systems with


different configurations – with and without glazing, water and air as circulating fluids and
top of the PV module with and without glazing for crystalline and amorphous silicon
cells. It was found that the setup using crystalline solar cells with a glazing on top under
which water was circulated as a fluid for enhancing thermal performance and booster
reflectors for enhancing electrical performance performed much better than simple PV/T
systems. The use of booster diffuse reflectors was found to considerably enhance the
performance of simple PV/T systems. Not only are PV/T cost effective, they also have a
better environmental impact when compared to standard PV modules
(Tripanagnostopoulos, 2004).
A study by Tonui and Tripanagnostopoulos (2006) the effect of using a thin flat
metallic plate or fins to augment heat transfer was investigated. Thermal efficiency was 5
and 3% more than the regular setup where the fins and thin metallic plate were used
respectively. It was recommended to use fins for high latitude locations and thin metallic
plate for low latitude locations as heat gains due to increased thermal efficiency of the
fins can be exploited. A mathematical model similar to Brinkworth’s model (2000) for
fictitious loop analysis was developed by the same authors (Tonui and
Tripanagnostopoulos) in 2007 to predict the induced airflow behind a PV/T system. It
was found to predict outlet air temperature within ± 2oC. The fins and thin metallic plate
17

configurations were tested with fins giving a better performance than thin metallic plates.
Results also indicate an optimum channel depth around 5-10 cm. Tripanagnostopoulos
(2007) found that placing a water heat exchanger behind PV rear surface was found to
give the best results for combined water-heat extraction and when this setup was
combined with diffuse collectors and also, a cost effective installation for horizontal roof
installation as achieved.
Charron and Athienitis (2006-a,b) conducted numerical and experimental studies
on double facades with integrated PV panels for two configurations – the front and
middle of the cavity. They also observed that the second configuration resulted in
increased combined efficiency (thermal and electric) and lower electricity generation
when compared to the first configuration. A 2D model was developed (Charron and
Athienitis (2006, b) and it was found to have good agreement with the 1D model
developed by the same authors. The same authors also observed that the values of CHTC
from literature were lower than those observed in experiments. The study was extended
by Liao et al (2007) who performed a CFD study on a PV/T system fitted outside a
double façade. In the study, measured temperature boundary conditions were used as
inputs to the CFD model and pressure was adjusted at the inlet and outlet to give the
required velocity. The fluid flow generated from the model was partially validated with
data obtained by a particle image velocimetry (PIV) system and generally, there was a
good agreement between computed and experimental temperature profiles. This study
reported Nusselt correlations for positions near and away from the PV/T façade.
An extensive review on flat plate PV/T collectors is presented in an article by
Zondag (2008). A historic overview of the research projects is listed according to the type
of coolant used, addressing the different research concerns. In order for PV/T systems to
be a commercial success, it was mentioned that general issues such as certification,
financing and awareness were equally important to the technical issues such as improving
efficiency for PV/T systems.
In a study on the performance of BIPV solar collectors by Anderson (2009), the
design of these systems was theoretically analyzed through the use of a modified Hottel-
Whillier model and the results were validated with experimental data. Results showed
18

that the key design parameters such as fin efficiency, thermal conductivity of the PV and
the lamination method on the overall efficiency of the system. An important finding was
that by integrating the PV/T system into the building resulted in a lower cost system since
the use of a PV with an insulated rear reduced the roof insulation cost as the air cavity
acted as passive insulating barrier.
Chen et al (2009) examined the effect of three integrated technologies – PV/T
system, ventilated concrete slab and a geothermal system and studied its effect on the
performance of a solar house with respect to efficiency, feasibility and performance in a
cold climate. The results indicated that the quasi-two dimensional steady state model
developed for simulating the thermal performance of the PV/T system is useful for
preliminary design and control of airflow in the PV/T system.
Norton et al (2010) presented a paper on enhancing the performance of BIPV
systems. It was found that the cost of a BIPV system could be lowered by reducing the
cost of the PV module and related expenditure and related efficiency. However, it was
found that considerable enhancement of a BIPV system was possible without
improvement in PV cell performance.
Da Silva (2010) modeled a hybrid PV/T solar system using a modular strategy
provided by Simulink/ Matlab. Parametric studies and annual simulations of PV/T
systems were preformed assuming an overall efficiency of 24% (15% thermal and 9%
electrical) for a house in Lisbon with 6m2 collector area and using p-Si cells. Results
indicated that a PV/T system could conveniently replace conventional solar thermal
systems. A collector enclosing vacuum or a noble gas at low pressure was tested and
results indicated that there was an increase in thermal efficiency and the decrease in
electrical efficiency was negligible. It was concluded that these systems (vacuum PV/T)
had a great potential and required further investigation.
Candanedo (2010) studied open loop PV/T systems with air as the fluid.
Experiments were conducted on an inclined PV/T system and correlations were
developed for forced convection dominated regime. A transient and steady state model
were proposed for evaluation of thermal/electrical performance. Effect of friction factors
on pressure drop, effects such as thermal capacitance of the PV and electrical modeling
19

of the PV were some of the other major contributions of this research. General guidelines
for the design and construction of PV/T systems were proposed. The main limitation of
this work was that it did not consider the effect of buoyancy at low flow rate which might
have affected the system performance. This work was used as the basis for the Thesis
developed presently by building on the recommendations suggested.

2.4 Numerical Studies of Heat Transfer in Channels

In order to obtain a better understanding of the turbulence in flow and heat


transfer, studies are conducted numerically using computational fluid dynamics (CFD) as
a tool. A few research papers are presented here that utilize these tools.
Shome et al (1995) performed a numerical analysis of a thermally developing
flow in the mixed convection regime in a horizontal tube with UWT. The study was
performed for viscous liquids from water to oil concluding that viscosity had a more
pronounced effect on friction factor than on the Nusselt number. An extensive literature
review was presented for developing laminar mixed convection and also, this study was
particularly useful in understanding the effect of boundary conditions on flows for mixed
convection described in Chapter 3.
In a numerical study by Ozsunar et al (2001), mixed convection for a rectangular
channel subjected to a UHF boundary condition was tested for Reynolds number between
500 and 1000 and Grashof number between 7x105 and 4x106. Effect of channel
inclination, surface heat flux and Reynolds number were tested using a CFD code and it
was concluded that buoyancy driven secondary flow could enhance heat transfer as much
as 250 percent.
Cadafalch et al (2003) studied the natural convection in an inclined channel with
asymmetric heating including the effects of radiation. Variation of channel length, inter-
plate spacing, inclination of channel, temperature difference between the wall and the
ambient and emissivity of the plates were studied. A normalized u-velocity was
compared to experimental results from an experiment using Particle Image Velocimetry
(PIV) and a good agreement was observed.
20

Yilmaz et al (2007) investigated turbulent natural convection in a vertical parallel


plate channel both experimentally and numerically for a channel with asymmetric heating.
A low Reynolds number (LRN) k-ε model was used and a correlation was presented and
compared with other studies in literature. This study provided valuable insights to the
turbulence in the near wall region, the turbulent kinetic energy (TKE) profile, bulk
temperature distribution and velocity distortion for natural convection.

2.5 Mixed Convection Studies

A simple way to classify the flow type is to examine the Richardson number. The
Richardson number is given as the ratio of the Grashof to the square of the Reynolds
number. Incropera (2006) cites the following criteria:
𝐺𝑟
≪ 1 𝑓𝑜𝑟𝑐𝑒𝑑 𝑐𝑜𝑛𝑣𝑒𝑐𝑡𝑖𝑜𝑛
𝑅𝑒 2
2.9
𝐺𝑟
≫ 1 𝑓𝑟𝑒𝑒 𝑐𝑜𝑛𝑣𝑒𝑐𝑡𝑖𝑜𝑛
𝑅𝑒 2
2.10
𝐺𝑟
~1 𝑚𝑖𝑥𝑒𝑑 𝑐𝑜𝑛𝑣𝑒𝑐𝑡𝑖𝑜𝑛
𝑅𝑒 2
2.11
A Richardson number range of up to 100 was used by Saelens et al (2002) for
mixed convection in turbulent, fully developed flow in a vertical tube. Joye et al (1989)
studied mixed convection in a vertical tube of L/D= 49.6 for Reynolds number ranging
from 360 to 35000 using a UWT boundary condition for assisted and opposed flow
conditions. Results were presented for mixed convection with a Richardson number range
between 0.2 and 50. Their experimental results indicated that the heat transfer for
opposed flow was double that for assisted flows and that for assisted flows, the effect of
natural convection is profound for laminar and transition regions.
Pavlovic et al (1991) reported results for mixed convection experiments in an
open cubic cavity of dimension 0.6m inclined at angles varying from 45-123o (0-90o –
aided flow and 90-180o – opposed flow) . The experimental setup was placed in a wind
21

tunnel with temperature of the surfaces ranging from 30-120oC and external flow ranging
from 0.3-1.2 m/s. A correlation was presented of the form:
𝑅𝑒 𝑛
𝑁𝑢 = 𝑁𝑢𝑁 [1 + 𝑐 � 0.5� ]
𝐺𝑟
2.12
An expression of a similar form is developed in the current study as well, explained in
Chapter 3.
Eicker et al (2003) presents a study in TRNSYS for a building with integrated PV.
The use of natural convection parameters was recommended even if the flow is fan-
driven. In the same study, the use of a comprehensive Nusselt number which accounts for
laminar and turbulent flow was proposed. Each of these Nusselt numbers in turn is given
as a function of a Reynolds number which is written as a function of free and forced
convection parameters, shown below:

2 2
𝑁𝑢 = �𝑁𝑢𝑙𝑎𝑚 + 𝑁𝑢𝑡𝑢𝑟𝑏

2.13
3
And 𝑁𝑢𝑙𝑎𝑚 = 0.644�𝑅𝑒𝑚 . √Pr

2.14
0.8
0.037𝑅𝑒𝑚 𝑃𝑟
𝑁𝑢𝑡𝑢𝑟𝑏 = −0.1
1 + 2.44 𝑅𝑒𝑚 (Pr 0.67 − 1)

2.15

2 2
Where 𝑅𝑒𝑚 = �𝑅𝑒𝑓𝑜𝑟𝑐𝑒𝑑 + 𝑅𝑒𝑓𝑟𝑒𝑒

2.16

And

𝑣𝐻
𝑅𝑒𝑓𝑜𝑟𝑐𝑒𝑑 =
𝜈

2.17
22

𝐺𝑟
𝑅𝑒𝑓𝑟𝑒𝑒 = �
2.5

2.18

A similar correlation was mentioned by Incropera (2006) and Kays (2005) of the
form 𝑁𝑢𝑀 = 𝑁𝑢𝐹 ± 𝑁𝑢𝑁 , where the superscript “M” denotes mixed convection, “F”
denotes forced and “N” denoted natural convection. The expression of this form is used
in Chapter 3 for justification of the newly developed correlation.
Petukhov et al (1976) devised equations for horizontal and vertical circular pipes
that extend the upper limits for the Grashof number beyond which the Nusselt numbers
will change by 1%. These are explained in detail in Chapter 3 to indicate that for pipes
with a low flow rate, forced convection correlations such as those cited above may not be
adequate in representing the convective heat transfer.
Jackson et al (1961) conducted an experimental study for mixed convection for a
horizontal tube with UWT boundary condition. The Graetz number ranged from 60-1300
and a semi empirical correlation was developed based on log mean temperature
difference which fit the laminar flow experimental data.
Brown and Gauvin (1966) conducted an experimental investigation to explain the
effects of buoyancy forces at low flow rates in horizontal cylinder with UWT thereby
explaining the conditions at which they aid or oppose the flow. It was found that for a
fully developed turbulent flow, temperature fluctuations were similar to pure forced flows
and the results in general agreed with a theoretical analysis published by Diessler (1962).
Oosthuizen et al (1970) presented an empirical analysis for mixed convection in
horizontal cylinders with the external flow along the direction of buoyancy, i.e, vertical
flow for 100 ≤ Re ≤ 3000. A correlation for forced convection was developed in terms
of the forced convection.
In a similar study by Jackson (1971) an analytical correlation of superposed
mixed convection in a horizontal cylinder with aided flow is presented. They proposed a
similar correlation and the agreement was good when compared to the correlation
23

proposed by Oosthuizen (1970), for a similar Reynolds and Grashof number range.
Osborne and Incropera (1985) conducted experiments for mixed convection thermally
developing flow in a horizontal duct with UHF. A Correlation was developed for the case
of asymmetric heating through horizontal plates with both plates at different heat fluxes.
Heat fluxes were tested for a range of 0-6000 W/m2 and Re< 2800.
Zhang et al (1998) studied mixed convection for asymmetric heating conditions in
a vertical square channel for assisted and opposed flows. It was observed that heat
transfer was more for opposed flow condition and the Nusselt ratio (Numixed/ Nuforced)
increases as the Richardson ratio increased for the Reynolds range of 400-10,000. The
experimental results were compared to correlations by Cotton and Jackson (1990), Hall
(1969), Yan (1986) and Gau (1992) and good agreement was obtained with Cotton and
Jackson correlation developed for opposed flow.
Guimaraes et al (2008) studied mixed convection in an inclined rectangular
channel heated by discreet sources. The Reynolds, Grashof numbers and inclination of
the duct were respectively: 1≤ Re ≤ 1000, 103 ≤ Gr ≤ 105 and 0o ≤ θ≤ 90o. It was
concluded that the inclination had s strong influence on flow and heat transfer for low
Reynolds number and that the lowest temperature distribution in the modules were for
inclination angles between 45o and 90o.
Chong et al (2008) studied the effect of the inclination a rectangular duct for
laminar and transition flows (for assisted and opposed conditions). They studied
Reynolds numbers were in the range of 420 to 2630 and the Gr was in the range of
6.8×103 to 4.1×104 for seven orientations ranging from -30 to +30. Although a
correlation was not developed, it was concluded that a maximum heat transfer occurred at
30° tilt angle and that for Reynolds numbers above 1800, the influence of the tilt angle on
the Nusselt number diminished. The correlations for mixed convection are enlisted in
Table 2.2.
24

Table 2.2 Correlations for mixed convection

Investigator Correlation Remarks

2
𝑁𝑢𝑚 = 2.67 [ 𝐺𝑧𝑚 + 7.57 Mixed in a
Jackson (1961) × 10−5 (𝐺𝑟𝑚 . 𝑃𝑟)1.5 ]1/6 horizontal tube
with UWT;
60≤ Gzm ≤ 3000 “m” denotes log
mean temperature
(2.19) difference.

Horizontal pipe
Oosthuizen, P.H 𝑁𝑢 𝐺𝑟 𝐺𝑟 2 with external
= 1 + 0.18 � 2 � − 0.011 � 2 �
and Madan, S 𝑁𝑢𝑓𝑜𝑟𝑐𝑒𝑑 𝑅𝑒 𝑅𝑒 flow of air,
(1970) perpendicular to
the pipe, along
100≤ Re ≤ 3000 the direction of
25,000≤ Gr ≤ 300,000 buoyancy.

(2.20)

Combined free
𝑁𝑢 𝐺𝑟 0.25 and forced
= �1 + 2 �
Jackson, T.W 𝑁𝑢𝑓𝑜𝑟𝑐𝑒𝑑 𝑅𝑒 convective
(1971) correlation for
flow along
direction of
(2.21) buoyancy.
25

Table 2.2 Continued

Vertical pipes:

𝐺𝑟𝑞 = Grashof number


1 range to show
(1.3 × 104 𝑅𝑒 2.75 . 𝑃𝑟 [𝑅𝑒 8 . 2.4 (Pr 0.667 −1)])/
when the effect of
(log(𝑅𝑒) + 1.15log (5𝑃𝑟 + 1) +
Petukhov (1976) buoyancy will
0.5Pr −1 . 8)
affect Nu
(2.22) numbers by more
than 1%
Horizontal pipes:

𝐺𝑟𝑞 = (3.10−5 𝑅𝑒 2.75 . Pr 0.5 [1


2 1
+ 2.4 �𝑃𝑟 3 − 1� 𝑅𝑒 −8 ]

(2.23)

Correlation
𝑁𝑢𝑥,𝑡 = 1.49 𝐺𝑧 0.33 for Re<2800 developed for
Osborne and (2.24) asymmetric
Incropera (1985) heating through
𝑁𝑢𝑥,𝑏𝑡 𝑅𝑎0.75 horizontal plates
= (1 + 0.00365 )
𝑁𝑢𝑥,𝐹 𝐺𝑧 with both plates
at different heat
400< Re< 2600 fluxes.
8.32x107 <Ra < 4.16x1010
(2.25)
26

Table 2.2 Continued

Correlation
𝑁𝑢 developed for
𝑁𝑢𝐹,𝑊𝑢 opposed flows in

= �1 + 1.25 a vertical annular


Wu, Xu and passage with a
−2 0.46
Jackson 𝐺𝑟𝑞 𝑁𝑢 heated core and
× 105 � 3.425 0.8 � � � �
(2002) 𝑅𝑒 𝑃𝑟 𝑁𝑢𝐹𝑊𝑢 insulated outer
core with water
Where, employed as the
0.74 0.4
𝑁𝑢𝐹,𝑊𝑢 = 0.042𝑅𝑒 𝑃𝑟 fluid.
Valid for 6000<Re<20,000
(2.26)

2.6 Effect of Turbulence

Han and Zhang (1991) investigated the effect of nine different rib shapes on two
walls of a heated square duct. Average heat transfer and friction factor were reported and
it was concluded that angled ribs and “V” ribs had maximum heat transfer and at an angle
of 60o, it was a maximum. The same authors (1992) investigated the effect of a surface
heat flux boundary condition on the same channel. Increased heat transfer was reported
when wither one or two of the walls were heated as opposed to heating all the four walls
and that the average Nusselt number tends to decrease for increasing Reynolds numbers.
In 1992, in another study by the same authors, seven different broken ribs were tested and
it was established that a broken “V” rib at 60o provided heat transfer 4.5 times higher
than a smooth channel.
Hu and Shen (1996) performed experiments from which detailed distribution of
internal heat transfer coefficients were obtained for a staggered array of ribs inclined at
45o with a combination of grooves. An averaged enhancement of 3-4 for ribs and 2.5-3.2
27

for a combination of ribs and grooves was observed. Eiamsa (2009) studied turbulent
forced convection heat transfer and friction characteristics in a rectangular duct with four
different rib types and observed up to 80% heat transfer enhancement when compared to
a smooth duct.
A CFD study was performed by Chaube (2006) on heat augmentation in a
rectangular duct with ribs for Reynolds range between 3000 and 20,000 was investigated.
A SST k-ω model was used and a detailed analysis of heat transfer variation in the inter-
rib region was done. Nine different shapes of ribs were examined and it was concluded
that the highest heat transfer enhancement was found with a rectangular rib of 3 × 5 mm.
Experimental investigations by Promvonge (2008) to assess forced convection heat
transfer and friction loss behavior through ribbed cross sections of various shapes and
orientations. It was found that an in-line rib arrangement provided higher heat transfer
than a staggered configuration for the same mass flow rate and the triangular rib arranged
in a staggered array showed higher thermal performance.
Tang (2012) conducted experimental and numerical investigations to study the
turbulent flow and heat transfer characteristics in a rectangular channel with crossed ribs
and grooves. An SST k-ω model was used to turbulence model against which the
experimental data was verified. Numerical results indicated that a rib angle of 45o showed
best overall thermo-hydraulic performance in the order of 18%-36% higher than the case
for a rib angle of 0o.

2.7 Limitations of Existing Models

Most of the correlations that have been developed are for conditions that are
unsuitable for PV/T systems. Listed below are few of the limitations due to which the
correlations cannot be generalized:

1. Validity of correlations for high Reynolds numbers (above 10,000) or low Reynolds
numbers (below 3000) - representative of turbulent or laminar forced convection.
2. Effect of buoyancy is not considered for heated channels with low flow rates.
28

3. Use of long heated lengths to establish fully developed conditions that may not be
representative of an actual setup since an actual BIPV/T roof may be shorter.
4. Many of these studies have been carried out in horizontal and vertical rectangular
channels and circular pipes, not many studies conducted for inclined channels.
5. Channels are heated symmetrically, not the case for PV/T systems where there is
heating asymmetry due to non-uniform fluxes.
6. Effect of turbulence due to wooden framing inside the channel.
To overcome the shortcomings of the existing models, a new model has been
proposed which is explained in Chapters 3 and 4.

.
29

CHAPTER 3. NUMERICAL STUDY OF CONVECTIVE HEAT TRANSFER IN PV/T


SYSTEMS

3.1 Introduction

This chapter presents the Computational Fluid Dynamics (CFD) model developed
to evaluate the convective heat transfer coefficient beneath the PV module in an inclined
roof integrated PV/T system. The problem formulation is explained in section 3.2,
followed by details about numerical modeling, grid geometry and input parameters in
section 3.3. The results and analysis for the smooth channel are presented in section 3.4
and a comparative analysis with a channel that includes a framing is listed in section 3.5.

3.2 Experimental Setup

The PV/T system considered consists of an amorphous PV module attached to a


metal sheet which is mounted at a small distance from the roof of the house. The house
considered for the purpose of this research is the Eco TerraTM Equilibrium demonstration
house in Quebec, Canada. The PV/T channel has the following dimensions: 5.5 m length,
10.2 m width with the distance between the PV module and the insulation beneath being
3.8 cm (Aspect ratio = 268), inclined at an angle of 30o (Fig.3.1). The plate beneath the
PV is made of steel with long wave emissivity of 0.8 and the bottom surface of the
channel consists of a plexiglass sheet with insulation having a long wave emissivity of
0.2. Two configurations are tested – a smooth channel and another with wooden framing
elements. For data verification, a smaller smooth channel of length 2.84 m, 0.387 m
width and depth of 4 cm inclined at an angle of 45o is considered for which experimental
data are available (Candanedo,2010).
30

Figure 3.1. Schematic of PV/T of Eco-Terra House (Adapted from Chen, 2009)

3.3 Numerical Approach

3.3.1 Computational Domain, Grid & Near Wall Treatment

Infrared photos of the BIPV/T system installation in the Eco TerraTM house
suggest that there was very little variation of temperature along the width (Chen, 2009),
and hence a 2D model is considered appropriate. The computational domain created in
Gambit is representative of a lengthwise cross-section of the roof mounted PV/T system
with dimensions 2.84x 0.387x 0.04 m (Aspect ratio= 9.675). Since the channel has a
simple rectangular geometry, a quadrilateral meshing scheme is chosen (Fig 3.2).

Figure 3.2. Grid with resolution near the wall


31

The commercial mesh generator GAMBIT 2.3.2 was used to create the grid. A mesh was
created such that the grid resolution was fine near the wall and coarse further away. A
dimensionless wall distance 𝑦 + or 𝑦 ∗ was used to characterize the required grid resolution
near the wall:
𝜌𝑓 𝑢𝜏 𝑦𝑝
𝑦+ =
𝜇
3.1
𝜌𝑓 𝐶𝜇0.25 𝑘𝑝0.5 𝑦𝑝
𝑦∗ =
𝜇
3.2
Where 𝑦𝑝 is the distance from the point P to the wall, 𝜌𝑓 is the density of the fluid, 𝑘𝑝 is
the turbulent kinetic energy at the point P, 𝜇 is the fluid viscosity at the point Pand 𝑢𝜏 is
the friction velocity, given by the following equation:
�𝜏𝑤
𝑢𝜏 =
𝜌𝑓
3.3
Where 𝜏𝑤 is the wall-shear stress. The value of 𝑦 + and 𝑦 ∗ depends on the choice of
distance for first cell from the wall when creating the boundary layer, which in turn
depends on the highest velocity used during problem formulation. The y* values for
different velocities are plotted below for a channel with length = 2.84 m (Fig 3.3) to
indicate that the requirement for 𝑦 ∗ <1 (Kays, 2005) is a satisfied.

3.3.2 Solution Parameters

The commercial CFD package FLUENT 12.1 was used to solve the equations
with double precision. The software uses the finite volume technique and it offers two
types of solvers; pressure based and density based. A pressure based solver was used to
account for low-speed incompressible flows. In the pressure based solver, the mass
conservation (continuity) of the velocity field is achieved by solving a pressure equation.
32

1.2
1.1
1 v = 0.6 m/s
0.9 v = 1.2 m/s
0.8 v = 2.4 m/s
0.7
v = 3.6 m/s
y*

0.6
0.5 v = 4 m/s
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5
Distance from Inlet (m)

Figure 3.3. Plot of y* values on the upper wall for different air speed

The fluid density is treated using the incompressible ideal gas approximation
since this is more accurate than the Boussinesq model. The Boussinesq model is
recommended for flows where the pressure variations are small enough for the fluid to be
treated as fully incompressible, but in order to obtain an accurate relationship between the
density and temperature, the incompressible ideal gas approximation is utilized, and the
equation is given below:
𝑃𝑜𝑝
𝜌𝑓 =
𝑅
𝑀𝑤 𝑇
3.4
Where R is the universal gas constant, Mw is the molecular weight of the gas and Pop is
the operating pressure. The thermal conductivity of the fluid was assumed to be 0.0242
W/m-K. For viscosity, the Sutherland approximation with three coefficients was
employed since it has an advantage of having lesser error of 2% when compared to the
power law which has an error of ± 4%.
𝑇 𝑇𝑜 + 𝑆𝑐𝑜𝑛𝑠𝑡
𝜇 = 𝜇𝑜 � � 3/2
𝑇𝑜 𝑇 + 𝑆𝑐𝑜𝑛𝑠𝑡
3.5
33

Where 𝜇 is the viscosity in kg/m-s, T is the static temperature in K, 𝜇𝑜 is the reference


value in kg/m-s, 𝑇𝑜 is the reference temperature and S is the Sutherland constant.

The specific heat cp is input as a piecewise polynomial with eight constants and it has the
form:
𝑐𝑝 (𝑇) = 𝐴1 + 𝐴2 𝑇 + 𝐴3 𝑇 2 + …
3.6
The pressure equation is derived from the continuity and momentum equations in
a way that the velocity field, corrected by pressure satisfies continuity and is solved
repeatedly till it satisfies a convergence criterion. A second order upwind discretization
scheme was used for momentum, turbulence parameters and energy while the pressure
interpolation was body force weighted, since this model considers the effect of buoyancy.
For the pressure velocity coupling the SIMPLE algorithm (Patankar and Spalding, 1972)
was used. The convergence criterion for the residuals of energy and continuity were set to
1e-06 while all other criteria were set to 1e-03 for all simulations performed in this study.

3.3.3 Turbulence Model

Turbulence is a phenomenon that consists of fluctuations of various flow


properties. For steady simulations, the instantaneous governing equations (conservation
of mass, momentum and energy) are time averaged resulting in a modified set of
equations Reynolds Averaged Navier Stokes (RANS equations) that are computationally
less expensive to solve. These modified equations have more unknown variables (Re
stresses, turbulent fluxes) than the number of equations and hence additional equations
are supplemented through turbulence closure models, which in turn are used to solve for
these unknown quantities. FLUENT offers several turbulence models of which a variant
of the k-ω turbulence model – Shear Stress Transport k-ω (SST k-ω) model developed by
Menter in 1994 is employed in the present study after comparison with other turbulence
models (Table 3.1). The outlet air temperature of these models is compared to the
experimental data obtained for February 17th, 2009 (discussed in detail in Section 3.4) for
v= 0.3 m/s to 1.4 m/s (Re= 1284-7097). The values shown in Table 3.1 are representative
34

for only case where velocity = 1.4m/s. It can be observed that the models predict the
outlet air temperature (averaged over 4 points) within 2% of each other and within 10%
of the experimental outlet air temperature at the chosen time. The SST k-ω model was
primarily chosen since it combines the positive aspects of the standard k-ε (i.e. free-
stream independence) and standard k-ω model (i.e. near wall modeling) by using a
blending function and also because it is recommended (Fluent, 2006) for use in wall
bounded flows. When flow over wooden framing elements is modeled, the SST k-ω is
preferred due to the advantage with using the blending function that captures the flow in
regions where there is flow separation (Chaube, 2006). Details for the SST k-ω model
are presented below.

Table 3.1 Average outlet air temperature of various turbulence models

Std. Realizable RNG Std. SST


Experimental
k-ε k-ε k-ε k-ω k-ω

Air
10.22 9.252 9.399 9.332 9.41 9.38
temperature
% variation
from - 0.095 0.089 0.094 0.087 0.089
Experimental

3.3.3.1 Shear Stress Transport k-ω Model

The SST k-ω model incorporates a damped cross diffusion derivative term in the
equation for the specific dissipation, ω. The definition of the turbulent viscosity is
modified to account for the transport of the turbulent shear stress and the modeling
constants are different. Transport equations for k and ω for this model are given as
follows:

𝜕 𝜕 𝜕 𝜇𝑡 𝜕𝑘
�𝜌𝑓 𝑘� + �𝜌𝑓 𝑘𝑢𝑖 � = ��𝜇 + � � + 𝐺𝑘 − 𝑌𝑘 + 𝑆𝑘
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑗 𝜎𝑘 𝜕𝑥𝑗
3.7
35

𝜕 𝜕 𝜕 𝜇𝑡 𝜕ω
�𝜌𝑓 ω� + �𝜌𝑓 ω𝑢𝑖 � = ��𝜇 + � � + 𝐺ω − 𝑌ω + 𝐷ω + 𝑆ω
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑗 𝜎ω 𝜕𝑥𝑗
3.8
It should be noted that Eq.3.1 and Eq.3.2 are written for compressible flow, but
they can be used for incompressible flow as well by treating the density as a constant. In
Eq.3.1 and Eq.3.2, 𝐺𝑘 represents the generation of turbulence kinetic energy due to the
mean velocity gradients. 𝐺ω is the generation of ω, 𝑌𝑘 and 𝑌ω represent the dissipation of
k and ω due to turbulence respectively and 𝐷ω represents the cross diffusion term. 𝑆𝑘 and
𝑆ω are user-defined source terms. 𝜎𝑘 and 𝜎ω are the turbulent Prandtl numbers for k and
ω respectively, and defined as
1
𝜎𝑘 =
𝐹1 1 − 𝐹1
𝜎𝑘,1 + 𝜎𝑘,2
3.9
1
𝜎ω =
𝐹1 1 − 𝐹1
𝜎ω,1 + 𝜎ω,2
3.10
The turbulent viscosity for this model is computed as shown below
𝜌𝑓 𝑘 1
𝜇𝑡 = .
𝜔 max � 1 , 𝑆𝐹2 �
𝛼 ∗ 𝛼1 𝜔
3.11
The coefficient 𝛼 ∗ damps the turbulent viscosity causing a low-Re correlation and is
defined as
𝑅𝑒
𝛼0∗ + 𝑅𝑒 𝑡
𝑘

𝛼 ∗ = 𝛼∞ � �
𝑅𝑒
1+ 𝑅 𝑡
𝑘

3.12
Where
36

𝜌𝑓 𝑘 𝛽𝑖
𝑅𝑒𝑡 = 𝛼0∗ =
𝜇𝜔 3

F1 and F2 are the blending functions and are given by


𝐹1 = tanh(𝜙14 )
3.13
Where

√𝑘 500𝜇 4𝜌𝑓 𝑘
𝜙1 = min �𝑚𝑎𝑥 � , � , �
0.09𝜔𝑦 𝜌𝑓 𝑦 2 𝜔 𝜎𝜔,2 𝐷𝜔+ 𝑦 2

1 1 𝜕𝜔
𝐷𝜔+ = max � 2𝜌𝑓 , 10−10 �
𝜎𝜔,2 𝜔 𝜕𝑥𝑗
And
𝐹2 = tanh(𝜙22 )
3.14
√𝑘 500𝜇
𝜙2 = max � 2 , �
0.09𝜔𝑦 𝜌𝑓 𝑦 2 𝜔

The model constants take the following values for the SST k-ω model (Fluent, 2006).
𝜎𝑘,1= 1.176 𝜎𝜔,1= 2.0 𝜎𝑘,2= 1.0 𝜎𝜔,2= 1.168

𝛼1 = 0.31 𝛼∞ = 1.0 𝑅𝑘 = 6.0 𝛽𝑖 = 0.072

3.3.4 Boundary Conditions

The choice of boundary conditions is critical to the problem as this reflects how
close to reality the model will be and also determines the accuracy of the results. In this
study, a velocity inlet boundary condition was used to simulate a fan at the end of the
channel that draws air through it. At the outlet, a pressure outlet boundary condition was
employed. Backflow is not expected since there are no sharp turns and the flow is being
37

forced by a fan, a default condition of 0 Pa gauge pressure was used. The top and bottom
walls were modeled as no-slip wall with zero roughness height. For purposes of
verification of the temperature profile, the top wall is assigned a measured temperature
profile using a UDF (user defined function) and for purposes such as developing a
correlation, a uniform heat flux boundary condition was used. The heat flux boundary
condition is based on the approximation that the solar heat flux incident on the top of the
PV panel is conducted uniformly though the PV module and on to the top of the channel.
This is a reasonable assumption given that all the material properties are uniform along
the depth. The bottom wall is well insulated and it is treated as an adiabatic wall with
zero heat flux. The boundary conditions are represented in the diagram below (Fig.3.4).
A nominal value for turbulence intensity (TI) - 10% is set at the inlet to account for sharp
edges. Given below is a graph (Fig.3.5) depicting the variation of air temperature along
the length for different TI at the inlet and for v= 1.4 m/s (Re= 7097) for UHF boundary
condition. The difference in percentage between outlet air temperatures for different TI
from the assumed TI of 10% is given in Table 3.2. The impact of inlet TI was checked for
higher and lower velocities (v= 0.4 m/s, 0.8 m/s, 1.6 m/s, 2 m/s) and similar results were
obtained.

Figure 3.4 Boundary condition used in the CFD model


38

280
TI= 5%
279
TI= 10%
278
TI= 15%
Air Temperature (K) 277
TI= 20%
276
275
274
273
272
271
0 0.5 1 1.5 2 2.5 3
Distance from inlet (m)

Figure 3.5 Variation of air temperature due to TI

It can be observed that the change in the inlet turbulence intensity from 5-20%
yields a change of less than 1% in outlet temperature. The change in the inlet turbulence
intensity changes the temperature profile only in the initial portion of the channel. Also,
Nemati (2009) cites that for a Reynolds number between 1400-6000, a core turbulence
intensity of 5-6% is expected using a formula based on Reynolds number (TI = 0.16*
Re)-1/8. However for flow behind PV/T systems, the core TI, Brinkworth (2000) cites is
between 15-20%. Hence the assumption of an inlet turbulence intensity of 10% is
justified.

Table 3.2 Impact of change in inlet TI on outlet air temperature

Outlet Air Temperature Variation


TI (%) (K) (%)

5 279.106 -
10 278.934 0.0616
15 278.886 0.0172
20 278.887 0.0168
39

3.3.5 Grid Independence

A grid independence study was carried out to achieve a balance between being
computationally inexpensive and obtaining accurate results. The surface averaged CHTC
was used to check the grid independence. Initially a grid of 1500x20 was chosen and the
surface averaged heat transfer coefficient was plotted refining it by a factor of 1.5-2 every
successive iteration. The results of the grid independence study are shown in Table3.3,
based which a 3000x30 grid with 90,000 cells was chosen for the present study.

Table 3.3 Grid independence Study

Surface Averaged CHTC


Grid size Cells % difference
(W/m2-K)
1500x20 30000 3.07035 -
2000x15 30000 3.2078 4.477
2000x30 60000 3.3528 4.520
3000x30 90000 3.5328 5.369
3000x40 120000 3.5572 0.691

3.4 Results and Analysis

This section presents CFD simulation results for heat transfer and air flow through
a 2D channel of dimensions 2.84x 0.04 m. A realistic PV/T thermal system is of a larger
length and hence, for the purpose of developing more generalized correlations (in terms
of Nusselt number) valid for an actual roof, a 2D channel of dimension 5.5x 0.038 m is
used for which the details are provided in the following sections.

3.4.1 Verification of Temperature Profile in the Smooth Channel with Experimental


Data
In order to examine the temperature profile in the channel, it is essential to
understand the experimental setup and placement of thermocouples first. In the
40

experimental setup, 40 thermocouples were placed along the middle of the channel from
the inlet to the outlet. Nine of these were placed along the top, bottom and center of the
channel to measure the temperature of surfaces and air (Fig.3.6). Thermocouples are
placed at distances 0, 0.237, 0.711, 1.185, 1.422, 1.659, 2.133, 2.607, 2.84 m in the duct
and average temperature of air at the inlet, outlet and at the middle of the channel were
measured by placing 3 additional thermocouples at each of these locations. The variation
of temperature in the channel can be divided into two sections – variation of temperature
along the length and variation of temperature across the depth of the channel.

Figure 3.6 Placement of thermocouples in the channel

3.4.1.1 Variation of Temperature Profile along the Length of the Channel

Incropera and Dewitt (2006) list the variation of temperature profile along the
length of the channel, for two different boundary conditions – Uniform heat Flux (UHF)
and Uniform Wall Temperature (UWT). Though these are applicable for symmetrically
heated vertical pipes, these results may be extended to inclined channels as well. For a
case with UHF, the temperature rise for the channel surface is exponential in the entrance
region and linear in the fully developed region whereas the temperature rise for bulk air
in the channel is linear for both regions. Similarly, for the case with UWT, given that the
channel surface temperature is fixed, the rise of bulk air temperature is exponential
through the length of the channel. Authors like Aung (1987), Brinkworth (2000) and
Zhang (1998) note in their work that the temperature profile of air inside a heated duct
with UHF boundary condition, for both symmetric and asymmetric heating will remain
41

similar, i.e, the bulk air temperature will rise linearly. These findings are consistent with
the findings from CFD. For a UHF boundary condition of 300 W/m2, it can be observed
that the top surface temperature rises exponentially in the developing region and then
rises linearly along the length, whereas the bulk air temperature rises linearly in both
developing and developed regions (Fig.3.7). There is a distinct difference between mixed
convection in pipes or ducts with (UHF) and (UWT). Shome et al (1995) observed that
for the case of mixed convection in horizontal ducts gives rise to secondary flows
(induced due to buoyancy) which increases pressure drop and heat transfer thereby
reducing the thermal entrance length and induces an early transition to turbulent flow.

350
Bulk Air Temperature Top surface temperature
340
330
320
Temperature (K)

310
300
290
280
270
260
250
0 1 2 3 4 5 6
Position from inlet (m)

Figure 3.7 Temperature rise for UHF boundary condition in CFD

In a duct with UHF, there would always exist a difference in temperature between
wall and bulk air resulting in secondary flow throughout the length of the duct whereas
for a duct with UWT, the secondary flows develop to a maximum and eventually
diminish to zero as the wall to bulk air temperature diminishes if the duct is sufficiently
long. Applying a similar analogy to the present case, we observe the variation of
temperature along the length of the channel (at 2 cm distance from the top and bottom
surface) shown below in Fig.3.8 and Fig.3.9. Results are shown for different flow rates
(Re = 1284, 2479, 4094, 7097) based on measured data that were collected on February
17, 2009. It can be observed that the values from the 2D-CFD model agree well with the
42

measured temperature up to a length of 1.4m (mid channel length). After this point, the
measured temperature values show a sudden jump whereas the values from CFD continue
to rise linearly (Fig.3.8, Fig.3.9). This is different from the expected linear rise in air
temperature. This abrupt shift in temperature profile may be attributed to factors such as
placement of additional thermocouples at the center and at the outlet of the channel and
the necessary wiring may have created blockage and additional turbulence considering
that is a narrow cavity and placing thermocouples in exact positions is a difficult task.
Also in the experimental set up, there may be inlet (i.e. flow disturbances and/or
separation at the entrance which propagate downstream) and wind-induced effects that
were not considered in the idealized channel flow considered in the CFD model since
velocity data were not available to be used as inputs and for validation. Also potential
leakage of air through the PV/T duct or the damper and all other uncertainties associated
with experiments performed under uncontrolled weather conditions (wind speed and
direction, etc) may be another reason for this difference. As it has been pointed out in
previous studies (Blocken, 2009), transitional flows in channels are more difficult to
model while buoyancy-induced instabilities at low flow rates may be another source of
complexity (Greig, 2012). It is interesting to note that the CFD results are in better
agreement with the experimental data for higher flow rates, where potential turbulence
effects, flow disturbances and instabilities would be smaller. The measured data are
plotted below with the corresponding experimental error calculated (Candanedo, 2010)
from the uncertainties in measurement by Laminar Flow Element, collector area, flow
rate and material properties (more information about the experimental set up is provided
in Ch. 4). It can be observed from Fig.3.10 that if the measured air temperature profile is
linearly extended, it matches well with the simulated data.
A 3D model was created with dimensions 2.84x 0.387x 0.04 m (aspect ratio =
9.675) to verify the accuracy of the results from the 2D model. The boundary conditions
for the model would all remain the same except for the addition of side walls that are
treated as adiabatic. For the 3D model, the grid cells remain same as that of the 2D model
along the X and Y axes (3000x 30) and the number of cells along the Z axis was taken to
be 100. The centerline is chosen for measurement of temperature along the length of the
43

channel, and this is plotted along with the measured data and the results from the 2D
model as shown above (Figure.3.10). It can be observed that the results from a 2D-CFD
model agree slightly better with the measured values than the results from a 3D-CFD
model. This could be due to the possibility of losing heat gained by the air to the side
walls in the CFD model than what may actually occur in an outdoor experiment, since the
sides are exposed to the outside air and ground reflected sunlight in a reality which may
mean that the side walls are not completely adiabatic.

1.4m/s_CFD 1m/s_CFD
320
0.5m/s_CFD 0.29m/s_CFD
315
1.4m/s_measured 1m/s_measured
310
0.5m/s_measured 0.29m/s_measured
305
Temperature (K)

300

295

290

285

280

275

270
0.00000 0.50000 1.00000 1.50000 2.00000 2.50000 3.00000
Position from inlet (m)

Figure 3.8 Variation of air temperature at mid-depth along the length of the channel for
different flow rates (Re = 1284, 2479, 4094, 7097) - experimental and CFD results
44

Measured data 2D model data


300
Mid-length Measured top surface temperature
Air Temperature (K) 295

290

285

280

275

270
0 0.5 1 1.5 2 2.5 3
Length (m)

Figure 3.9 Variation of air temperature at mid-depth along the length of the channel and
surface temperature for velocity = 1.4m/s (Re = 7097)

Measured air temperature 3D model data


2D model data Measured top surface temperature
295

290
Temperature (K)

285

280

275

270
0 0.5 1 1.5 2 2.5 3
Position from inlet (m)

Figure 3.10 Variation of air temperature at mid-depth, along the length of the channel for
velocity = 1.4m/s (Re = 7097)
45

3.4.1.2 Variation of Temperature Profile across the Depth of the Channel

Asymmetric heating inside the channel causes a large temperature gradient along
its depth with the consequence that the average air temperature is closer to the insulation
temperature than to the top surface. This can be confirmed by CFD analysis as shown
below. Data are presented for v = 1.4m/s (Re = 7097). It is evident from Figure 3.11 that
the average air temperature is closer to the temperature at the bottom of the channel since
the heat flux incident the top of the channel is much larger than the heat transferred from
the bottom of the channel to the airstream by means of convection. The insulation creates
nearly adiabatic conditions and the heat loss through the bottom is practically negligible.
The values of the measured temperature at the mid-length and the outlet were checked to
compare the temperature variation (Fig.3.12, Fig.3.13). The values predicted close to the
top surface by CFD (at 3.2 cm from bottom) that are important for CHTC calculations
(Tbulk is evaluated at 2 cm), are in reasonably good agreement with the measured values.
Also, the average outlet air temperature, which is important for calculating the thermal
efficiency of the PV/T system, predicted by CFD is closer (within 2-3oC) to the average
outlet air temperature measured for the four points than individual values. Similar
temperature profiles with high temperature differences between the top and bottom air
layers across the depth of a vertical BIPV/T channel were observed in the simulated and
experimental results by Liao (2007) for flows with Re in the same range.

The difference between the experimental and CFD results at points closer to the
bottom of the channel may have been because the channel depth is very small and the
experiment may have been affected by mixing of air due to the turbulence initiated by the
thermocouple wiring. In a small channel such as this one, the thermocouples and wiring
may behave in a similar manner acting as corrugations or fins that enhance heat transfer.
46

0.04

0.035
Distance from bottom (m)
0.03

0.025
0.71 m from entrance
0.02
1.42 m from entrance
0.015 2.13 m from entrance
0.01 Outlet

0.005

0
270 275 280 285 290 295
Temperature (K)

Figure 3.11Variation of temperature across the depth of the channel for velocity = 1.4m/s
(Re = 7097)

0.04

0.03 CFD_v=1.4m/s
Distance from bottom (m)

CFD_v=1m/s
0.03
CFD_v=0.49m/s
0.02 CFD_v=0.29m/s

0.02 Actual_v=1.4m/s
Actual_v=1m/s
0.01
Actual_v=0.49m/s
0.01
Actual_v=0.29m/s
0.00
270 275 280 285 290 295 300
Temperature (K)

Figure 3.12 Variation of temperature across the depth of the channel for different flow
rates (Re = 1284, 2479, 4094, 7097) - CFD and experimental results
47

0.04
CFD_v=1.4m/s
0.03
Distance from bottom (m) CFD_v=1m/s
0.03 CFD_v=0.49m/s

0.02 CFD_v=0.29m/s

Actual_v=1.4m/s
0.02
Actual_v=1m/s
0.01
Actual_v=0.49m/s
0.01 Actual_v=0.29m/s

0.00
270 275 280 285 290 295 300 305 310
Temperature (K)

Figure 3.13 Variation of temperature across the depth of the channel at the outlet for
different flow rates (Re = 1284, 2479, 4094, 7097) - CFD and experimental results

In a work on the investigation of flow behind a corrugated collector (Greig, 2012),


it was found that there was significant increase in heat transfer due to the corrugated
geometry since it enhanced turbulence properties and mixing. Also the temperature
difference between the air layers closer to the heated surface and the air layers closer to
the unheated surface was smaller for the vertical channel with a corrugated surface
compared to a channel with smooth surfaces. A detailed analysis of the thermo-fluid
behavior in the channel is interesting but out of the scope of this study due to the limited
information available from the experimental data. Instead the present study focuses on
the prediction of CHTC and the average temperature at the outlet which is important for
the evaluation of the energy performance in the PV/T system. This is discussed in
Chapter 4, along with the energy model with newly developed correlations where it can
be observed that the correlations predict outlet air temperature with acceptable accuracy.
Further experimental work is planned for additional verification of the CFD modeling
efforts presented here, under well-controlled conditions in a lab equipped with a solar
simulator (at Concordia University, Montreal) as well as in full-scale (Architectural
Engineering Research Labs, Purdue).
48

3.4.2 Velocity Profile Verification for Asymmetric Heating

Asymmetric heating has a considerable impact on the velocity profile of a heated


channel for both boundary conditions – UHF and UWT as it distorts the velocity profile.
For a channel with UWT, the profile is highly distorted for developing flow and
diminishes for a fully developed flow and completely disappears for symmetric wall
temperatures resulting in a parabolic profile. Whereas for a UHF case, the velocity profile
distortion diminishes for a fully developed flow, but unlike the case with UWT, the
velocity becomes parabolic only when Gr/Re=0 (Aung, 1987). Shown below (Fig.3.14)
are the velocity profiles for various velocities of 0.29, 0.49, 1, 1.4 m/s for which the
corresponding velocities are Re = 1284, 2479, 4094, 7097. It can be observed that as the
velocity in the channel is decreased and buoyancy increases, there is a shift in the cross-
sectional velocity towards the heated edge.

1.8 v=1.4m/s
v=1m/s
1.6
v=0.49m/s
1.4 v=0.29m/s
Velocity (m/s)

1.2
1
0.8
0.6
0.4
0.2
0
0.000 0.100 0.200 0.300 0.400 0.500 0.600
Non-dimensional distance from bottom, y/Dh

Figure 3.14 Velocity profile variation across the depth at various flow rates

3.4.3 Average Nusselt Number for Forced Convection

Average instantaneous CHTC values are obtained by using the control volume
technique. The heat flux, temperature of the surface, air temperature for each cell along
the length of the channel are first used to compute local CHTC whose average is taken
49

for average Nusselt number estimation. Previous studies have considered the use of two
CHTC – one for the top half of the channel and other for the bottom half (Candanedo,
2010; Hatton & Quarmby, 1963; Sparrow, 1966; Osborne & Incropera, 1985) and for
asymmetric heating in concentric circular tubes (Kays, 2005). Most cases cited above
dealt with flow over concentric tubes or flow over solar collectors that warranted use of
dual convective heat transfer coefficients. However, in this analysis, the bottom heat loss
can be considered negligible due to the high thermal resistance of the insulation. Since
the value of temperature for the bulk air and bottom insulation were not very different, it
has been proved unessential for calculation of heat removal inside the channel, later on in
this Thesis by means of sensitivity analysis. Their effect on outlet air temperature was
proved to be far less important than that of the computed CHTC on the top half of the
channel. For the present study, radiative heat transfer from the top to the bottom surface
of the channel was not taken into account as this accounted for 4-5W/m2 of heat gain by
the insulation which was computed to be 0.9 % of the net heat flux.
The data from CFD along the length (3000 cells) were used to evaluate the
convective heat transfer coefficient, the average of which is taken to determine the
average Nusselt number. The convective heat transfer coefficient for each control volume
in the channel was computed as:
𝑞𝑡𝑜𝑝 "
ℎ𝑐 =
𝑇𝑝𝑣_𝑏𝑜𝑡 − 𝑇𝑏𝑢𝑙𝑘
3.15
There 𝑇𝑏𝑢𝑙𝑘 is taken as the temperature of air along the middle of the channel at mid-
depth (2cm from top) of the channel
The Average Nusselt number is given as:
ℎ𝑐 𝐷ℎ
𝑁𝑢 =
𝑘𝑎𝑖𝑟
3.16
𝑣.𝐷ℎ
The Reynolds number for the channel is given by the equation, 𝑅𝑒 = , where v
𝜈

represents the velocity in the channel, Dh is the hydraulic diameter and 𝜈 is the kinematic
viscosity. The Average Nusselt number for the channel was calculated from local
50

distributions and it is plotted as a function of the Reynolds number in Fig. 3.15. A general
correlation is developed using regression analysis.

For forced convection, for 5000≤ Re ≤ 24000


𝑁𝑢𝑓𝑜𝑟𝑐𝑒𝑑 = 0.084𝑅𝑒 0.658 𝑃𝑟 0.4
3.17
This is 40-50% lower than the values for Nusselt number predicted by experimental
results (Candanedo, 2010). This can be attributed to a number of approximations during
the course of computation of the CHTC from experimental results and approximations by
the CFD software listed below.
60

50

40
Nusselt Number

30 Nu_fit
CFD data
20

10

0
0 5000 10000 15000 20000 25000 30000
Reynolds Number

Figure 3.15 Average Nusselt versus Reynolds profile for forced convection

In the experimental setup, 9 thermocouples were placed along the centerline; for
the calculation of Nusselt number, the value of air temperature at the mid-point is used to
represent Tbulk, which is an approximation. Also, a control volume based approach was
used for computation of heat transfer coefficients where, the heat gained from one control
volume to the other was approximated by the heat gained by the air due to convection on
the top half and by long wave radiation at the bottom. This approach reduces the accuracy
of predicting the actual heat transfer coefficient in an asymmetrically heated channel
51

which is a function of the heat flux incident on top of the channel and convection that
occurs through the entire depth of the channel. Experiments performed with similar PV/T
systems (Brinkworth, 2000; Huang, 1996) with a uniform heat flux boundary condition
have been performed in a lab setting where the conditions were simulated to suit the
needs of the experiment, whereas this was an outdoor experiment where there were many
factors that could not be controlled such as the external irradiation, wind speed and
direction, humidity inside the channel etc.
The channel is modeled as a smooth surface with zero roughness height since
modeling of roughness in FLUENT is difficult owing to roughness height being more
than the first cell height in the viscous region. The channel in reality has a certain
roughness which causes a pressure drop and this controls the temperature variations
within the channel. This phenomenon is an important aspect which should be considered
while using CFD. The biggest advantage of using CFD is that it eliminates the
uncertainties that are present in experiments. The accuracy of the CFD model however
depends on the boundary conditions that define the problem.

3.4.4 Effect of Grashof Number

Grashof number is a dimensionless number which is given as the ratio of the


buoyancy to viscous force acting on a fluid. It is given by:

𝑔𝛽. 𝑠𝑖𝑛𝜃. �𝑇𝑝𝑣_𝑏𝑜𝑡 − 𝑇𝑏𝑢𝑙𝑘 �. 𝐷ℎ3


𝐺𝑟 =
𝜈2
3.18
Where, g is the acceleration due to gravity, β is the volumetric expansion coefficient, 𝜃 is
the inclination angle of the duct, 𝑇𝑝𝑣_𝑏𝑜𝑡 is the temperature of the top of the channel/
bottom of the PV module, 𝑇𝑏𝑢𝑙𝑘 is the bulk air temperature inside the duct and 𝐷ℎ is the
hydraulic diameter.
Petukhov (1976) presented equations (Eq.2.22 -2.23 in Ch. 2) that establish the
limit of Grashof number for forced convection in vertical and horizontal pipes when
52

buoyancy will affect the heat transfer. If these limits are surpassed, the actual Nu number
was predicted to be higher than at least 1%. The values of Petukhov’s equations are
plotted with the data from the CFD simulations at higher and Reynolds numbers in
Figures 3.16 and 3.17 respectively.
It can be observed that the Grashof numbers computed from the CFD simulations
for high Reynolds number exceed Petukhov’s limits by only a small percentage whereas
for a lower Reynolds number for the same heat flux incident on the top surface, the
Grashof numbers exceed by several orders of magnitude Petukhov’s upper limit curves.
This indicates that forced convection corrleations used for a low Reynolds numbers
(Re<8000) may not be valid .
Kakac (1987) presents a chart by Metais and Eckert (1964) in his book on single
phase convective heat transfer that helps us chart the regime in which the data fall under.
This chart is valid for both UWT and UHF boundary conditions for vertical pipes. The
values of Reynolds number are plotted against the product of Gr.Pr.Dh/L (Fig.3.18,
Fig.3.19) to determine the range under which the data may be classified under.
CFD data for various heat fluxes is mapped and it can be observed that most of
the data falls in the mixed convection regime under both laminar and turbulent flow
regimes. The Ra numbers calculated from the CFD simulations are in the range 1x106 to
2.5x106.
53

62500000

12500000
Grashof number

2500000
Gr from CFD
Petukhov vertical
500000
Perokhov Horizontal

100000
0 5000 10000 15000 20000 25000
Reynolds Number

Figure 3.16 Petukhov’s upper limit curves (forced convection range)

100000000

10000000
Gr from CFD

1000000 Petukhov Verical


Grashof number

Petukhov Horizontal

100000

10000

1000
0 1000 2000 3000 4000 5000 6000
Reynolds Number

Figure 3.17 Petukhov’s upper limit curves (mixed convection range)


54

Figure 3.18 Metais and Eckert (1964) map for vertical tubes

10000
450W/m2

400W/m2

1000 500W/m2

300W/m2
Re

100 200W/m2

10

1
1000 10000 100000 1000000
Gr.Pr.D/L

Figure 3.19 CFD data for Metais and Eckert map for various heat fluxes
55

3.4.5 Average Nusselt Number for Mixed Convection

The criteria for a flow to be classified as mixed convection have been mentioned
in Chapter 2. In this section, using the data from CFD for the temperature of the surface
and bulk air for each control volume, the value of this ratio can be computed and the
velocities during which mixed convection occurs can be identified. Pavlovic (1991)
quantified mixed convection in a cubical enclosure by formulating an equation (Eq.2.12).
The value for natural convection for the case of an inclined channel asymmetrically
heated can be calculated from an equation by Azevedo and Sparrow (Eq.3.20). Shown
below is the graph (Fig. 3.20) for the data obtained from CFD simulations. Fitting the
curve for the data, using regression, the values for the constants are obtained as c= 0.935
and d= 0.539.

Hence, for mixed convection, for 1100≤ Re ≤ 3200


𝑅𝑒 0.539
𝑁𝑢_𝑚𝑖𝑥𝑒𝑑 = 𝑁𝑢_𝑛𝑎𝑡[ 1 + 0.935. � 0.5 � ]
𝐺𝑟
3.19
𝐷 0.25
Where, 𝑁𝑢𝑛𝑎𝑡 = 0.645 �𝑅𝑎. 𝐿 �

3.20

It may be noted here that the Reynolds range is not continuous for the correlations
developed, i.e., for forced convection 5000 ≤Re ≤ 24000 whereas for mixed convection,
1100≤ Re ≤ 3200. This can be explained with the limits for which mixed convection is
determined. Mixed convection is determined as a function of the Richardson number
which has the Grashof number in it. The Grashof number is dependent on the difference
between the surface and bulk air temperature. For determining whether a flow lies in
mixed convection range or forced convection regime for a Re between 3200 and 5000 (~
v= 0.6-0.95 m/s) would depend on this difference. For a given velocity, if the incident
fluxes are high, the flow would be more influenced by buoyancy and it would be in the
mixed convection region whereas if the fluxes were lower, the flow would be in the
forced convection region. For the purpose of this study, the Reynolds number range
56

between 3200-5000 has been included in the forced convection region. The same analogy
holds true for the correlations developed for the case with wooden framing.

1.8
1.6 CFD data
1.4 Nu fit
1.2
(Nu/Nu_Nat) -1

1
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5
Re/Gr0.5

Figure 3.20 Nusselt versus Re/Gr0.5 for mixed convection

3.4.6 Comparison of New Correlations with Previous Results

Several correlations have been developed over a period of years for flow through
ducts of different geometries, different heating styles and for a wide variety of flows. A
PV/T system is unique in a sense that these standard correlations developed in heat
transfer may not be applicable due to many factors such as geometry, varying heat fluxes
and flow rates with Re and Gr within a range that may be specific to this application and
complex flow patterns due to of framing members, (Brinkworth, 2000; Brinkworth &
Sandberg, 2005; Eicker, 1999; Bazilian & Prasad, 2002). The developed correlations are
unique in that sense. Comparison of the newly developed correlations to existing
correlations is only for the purpose of verification. This sections deals with forced and
mixed convection correlations and evaluation of their performance.
57

3.4.6.1 Comparison of the New Correlation with Previous Results – Forced Convection

The various correlations tabulated in Chapter 2 are plotted with the correlation
developed (Table 2.1) in this study (Fig.3.21).

New Correlation Dittus Boelter (1985)


80 Kays and Crawford (1980) Gnielinski (1976)
Tan&Charters (1969) Sparrow (1966)
70 Martinelli (1947) Petukhov (1963)
Candanedo (2010)

60

50
Nusselt numbers

40

30

20

10

0
2500 7500 12500 17500 22500 27500
Reynolds numbers

Figure 3.21 Comparison of forced convection correlations

The Petukhov correlation (1963) modified by Gnielinski (1976) is in overall best


accord with the simulation data (Fig.3.21). Since the work by Petukhov, Gnielinski and
Kays & Crawford (1980) were developed for similar settings of circular duct with UHF,
their results are is agreement amongst themselves as well. The Martinelli correlation is
valid up to a Reynolds number of 10,000 and the match is excellent with the newly
developed correlation. The Dittus-Boelter correlation is a classic correlation to gauge
forced convection in heated ducts. It can be observed that the Dittus-Boelter correlation
over predicts the Nusselt number. Tan and Charters (1969; Eicker, 2003) derived a
58

correlation which Eicker (2003) recommend for use in solar collectors where the
difference between the surface and bulk air temperatures is very high. This equation
under predicts the Nusselt number for the present case since it was originally developed
using a very high heat flux (~900 W/m2) boundary condition for a very high temperature
difference (~50-60oC) between surface and bulk air. For the present case, neither the heat
flux, nor the temperature differences are that high, hence a higher Nusselt number is
expected. The heat flux that finally reaches the top of the channel is estimated to be
around 300-400 W/m2 and the temperature difference between the top of the channel and
the bulk air is around 10-15oC.

3.4.6.2 Comparison of the New Correlation with Previous Results – Mixed Convection

The Nusselt number correlation developed for the mixed convection based on
CFD simulations is compared against four other correlations (Table 2.2). The values of
Nusselt number from every correlation is plotted against the Richardson number
(Fig.3.22).
With the help of Metais chart (1964) and Petukhov’s (1976) equation, it has been
clearly established that for low Reynolds number flows, buoyancy will play a major role
and that use of forced convection equations will not be sufficient. Jackson (1971)
developed a correlation for combined convection using a horizontal cylinder with UHF.
This has been plotted along with the correlation developed, but it under predicts the heat
transfer. This can be explained with the inclination effect on buoyancy - buoyancy is
enhanced with the inclination of the duct as determined by Busedra and Soliman (2000)
and Chong (2008). A study on laminar mixed convection heat transfer was completed by
Osborne and Incropera (1985) and a correlation was developed for Reynolds number
range between 400 and 2600 & a Rayleigh range of 8.32x107 and 4.16x 1010. The present
study has air a fluid and also does not have such high buoyancy and hence this correlation
could not be applied.
59

Incropera and De Witt (2006) and Kays et al. (2005) have mentioned that a
common practice is to correlate mixed convection heat transfer with an expression of the
form shown below:
𝑛 𝑛 𝑛
𝑁𝑢𝑚𝑖𝑥𝑒𝑑 = 𝑁𝑢𝑓𝑜𝑟𝑐𝑒𝑑 + 𝑁𝑢𝑛𝑎𝑡𝑢𝑟𝑎𝑙
3.21
For the form of the equation mentioned above, the correlation for forced convection
developed during the course of the study can be used and for natural convection the
correlation from Azevedo and Sparrow (1985) can be used:
1.85
𝑛 𝐷 0.25
(𝑁𝑢𝑚𝑖𝑥𝑒𝑑 )1.85 = (0.084 𝑅𝑒 0.658
𝑃𝑟 0.4 )1.85
+ �0.645 �𝑅𝑎. � �
𝐿
3.22

20

18

16

14
Nusselt Number

12

10

8
New Correlation for Mixed Convection
6 New Forced Coorelation and Azevedo(1985)

4 Wu-Xu (2002)
Modified Brinkworth(2000) and Azevedo(1985)
2
Jackson (1971)
0
0 0.5 1 1.5 2 2.5

Gr/Re2

Figure 3.22 Comparison of mixed convection correlations


60

When these values are plotted with n=1.85, the values obtained for mixed
convection are within are within 14% of the values predicted by the correlation obtained
in the previous section. Wu et al. (2002) developed a correlation where the ratio of
effective Nusselt number to the Nusselt number for forced convection is given by a
function of the Grashof (based on heat flux), Re and Pr numbers.
Listed below is a comparison between the use of Brinkworth’s correlation (2000)
for natural convection and the newly developed correlation for forced convection, for use
in Eq.3.21. The difference between the natural convection correlation used previously
(Azevedo & Sparrow - Eq.3.20) and Eq.3.21 is primarily that Brinkworth’s correlation
for natural convection is a function of Re, Dh, L whereas Azevedo & Sparrow’s
correlation for natural convection is a function of Ra, Dh, L.

𝑛 𝐷
(𝑁𝑢𝑚𝑖𝑥𝑒𝑑 )1.85 = (0.084 𝑅𝑒 0.658 𝑃𝑟 0.4 )1.85 + (5.801 + 0.086 �𝑅𝑒. �) 1.85
𝐿
3.23
It can be observed that Eq.3.23 shows excellent agreement with the CFD data.
This may be because Brinkworth’s correlation for natural convection is developed for a
PV/T system, asymmetrically heated from one side. This, combined with the forced
convection correlation developed for an inclined channel is consistent with the expected
results for the mixed convection regime.

3.4.7 Effect of Tilt Angle and Channel Length

3.4.7.1 Effect of Channel Length

The correlations have been developed for a channel length of 5.5 m and are valid
for a longer channel length as well. It can be inferred from Figure 3.23 where the
variation of the local Nusselt number along the dimensionless length has been plotted. It
can be observed that, after the entrance region, the Nusselt number remains a constant
through the length of the channel. A discussion about the effect of entrance region and
the validity of the correlations for shorter lengths has been presented in Chapter 4, section
4.3.6. For channel with wooden framing elements, it can be observed (Fig.3.24) that at
61

every framing, there is a sudden jump in the local Nusselt number followed by a fall, for
each of the four framing elements. In such cases, the Nusselt correlation developed is
strictly valid only within the limits specified and for a similar geometry.

180
Re= 4987
160 Re=9750
140 Re=14524
Re=20519
120
Local Nusselt

100

80

60

40

20

0
0 10 20 30 40 50 60 70 80
x/D

Figure 3.23 Variation of local Nusselt number in a smooth channel

300
Re=13875
Re=11876
250
Re=8325
200 Re=5550
Local Nusselt

150

100

50

0
0 10 20 30 40 50 60 70 80
x/D

Figure 3.24 Variation of local Nusselt number in a channel with wooden framing
62

3.4.7.2 Effect of Tilt of the Channel

It has been indicated that the Nusselt number varies largely with the inclination
angle at when the flow rate is low, but as the Reynolds number is increased, the heat
transfer coefficients were found to be nearly independent of inclination angles (Chong,
2008). Experiments on mixed convection indicate an increase in Nusselt number at low
flow rates at inclination angles between 0-30o (Maughan, 1987). Studies by Busedra
(2000) conducted for mixed convection in inclined semi-circular ducts for buoyancy
assisted and opposed flow conditions indicate an increase in heat transfer up to a certain
angle. This angle was found to be between -20o and +20o for inclined semi-circular ducts.
Similar studies were conducted by Ozsunar (2001) in a horizontal and upward inclined
rectangular channel and it was determined that the Nusselt number increased from 0-30o.
The inclination of the PV/T system depends on the geographic location (latitude)
of the place at which it is installed and the weather conditions (i.e. snow accumulation).
For the purposes of this study, angles of 30o and 45o are used since this is the range of
most American sloped roof housing (NREL, 2012). In the experimental study conducted
by Candanedo (2010) it was indicated that the Nusselt numbers do not vary largely
between the two angles.
The Grashof number used during the course of this Thesis includes the effect of
inclination angle (Eq.3.18) hence at low flow rates, this effect is accounted for. Since the
correlations have been developed for inclination angles 30o and 45o, it is recommended to
use them within the given limits for accuracy.

3.5 Framing Effect on Nusselt Number

When PV modules are installed on the roof, they are mounted on wooden framing
elements as shown in the figure (Fig.3.25). These structural framing elements enhance
heat transfer as they increase the turbulence by interrupting the viscous sub-layer and
causing flow separation and reattachment (Chaube, 2006). A more realistic representation
of a roof mounted PV/T system would be modeling the members with the wooden battens
and then predicting the average Nusselt numbers. For this particular scenario, four
63

rectangular pieces of wood are inserted, equally spaced along the length of the roof,
starting at the leading edge spaced at 1.084 m. For this PV/T system, the ratio of P/e
(Pitch to rib height ratio) is 52.7 and the (rib height/ hydraulic diameter ration) e/Dh =
0.27.

Figure 3.25 Photo of the wood framing in Eco-Terra house (Chen, 2009); Sketch of PV/T
& Close-up of channel with framing

A 2D model is developed for developing the correlations since studies in literature


(Chaube, 2006) indicate that there is negligible secondary flow in a transverse ribbed duct
surface. The Nusselt number is computed using bulk air temperature every 0.25 m from
64

the leading edge and also above every ridge of the wooden battens. The remaining
procedure for computing the Nusselt number is the same as for the case without wooden
battens.

3.5.1 Average Nusselt Number for Forced Convection

Shown below (Fig.3.26) are the average Nu numbers plotted versus the Reynolds
number. For forced convection, for 6000≤ Re ≤ 18000

𝑁𝑢𝑓𝑜𝑟𝑐𝑒𝑑 = 0.08𝑅𝑒 0.732 Pr 0.4


3.24
The correlation above yields Nusselt numbers those are about 1.9-2 times higher
than the Dittus-Boelter correlation as well as the correlation for newly developed
correlation for forced convection regime for the case without the wooden framing. In a
study of heat transfer in a rectangular channel ribbed channel, Promvonge and Thianpong
(2008) and Chaube (2006) report Nusselt number values almost 2 to 4 times the value
obtained with the Dittus Boelter correlation for Reynolds number range between 4000
and 16000. Not many studies are available where the heat transfer characteristics are
dealt with in detail for channels with wooden framing as most studies focus on flow
features and other studies focus on adiabatic ducts.

3.5.2 Average Nusselt Number for Mixed Convection

Shown below (Fig.3.27) are the average Nu numbers plotted versus the Reynolds number.
For mixed convection, for 1100≤ Re ≤ 3200,

𝑅𝑒 0.303
𝑁𝑢𝑚𝑖𝑥𝑒𝑑 = 𝑁𝑢𝑛𝑎𝑡 [ 1 + 2.366. � 0.5 � ]
𝐺𝑟
3.25
Where, from Eq.3.20,
𝐷 0.25
𝑁𝑢𝑛𝑎𝑡 = 0.645 �𝑅𝑎. �
𝐿
65

The equation above yields Nusselt numbers that are about 1.7-1.8 times higher
than the correlation obtained for the channel in a mixed regime without wooden framing.
One thing which should be noted is that, in a study that involves the formulation of a heat
transfer correlation for a channel with wooden framing, the correlations developed are
strictly valid only for the geometry and flow mentioned above. The reason being, the heat
transfer coefficient depends on turbulence which increases or decreases, depending on the
type of obstruction and the number of obstruction elements (in this case, wooden framing)
present inside the channel.

CFD data

Framing Correlation
100
Non-Framing Correlation
90
80
70
Nusselt Number

60
50
40
30
20
10
0
5500 7500 9500 11500 13500 15500 17500
Reynolds Number

Figure 3.26 Average Nusselt versus Reynolds profile in a channel with and without
wooden framing for forced convection
66

CFD data for case w/Framing Correlation w/Framing Correlation w/o Framing

35

30

25
Nusselt Number

20

15

10

0
0.7 0.9 1.1 1.3 1.5 1.7 1.9
Gr/Re^0.5

Figure 3.27 Average Nusselt versus Gr/Re0.5 profile in a channel with and without
wooden framing for mixed convection
67

CHAPTER 4. TRANSIENT ENERGY MODEL FOR OPEN LOOP PV/T SYSTEMS

4.1 Introduction

This chapter presents the transient model developed to evaluate the thermal and
electrical output of the PV module in an inclined roof integrated system. Details about the
experimental setup and data collection are explained in section 4.2, followed by the
modeling approach explained in section 4.3. The results and analysis for the smooth and
channel and the channel with wooden framing are presented in section 4.4 and a
parametric analysis to verify the sensitivity of the model to input data is presented in
section 4.5. Finally in section 4.6, a case study is presented where the model is
implemented to assess the performance of the Eco-Terra house and a simple control logic
is proposed.
4.2 Experimental Setup and Data Collection

The model developed is based on the experimental setup at Concordia University


which is shown in Fig.4.1a (Candanedo, 2010). The PV/T thermal system tested is
similar (but shorter) with the system mounted on the roof of the Eco-Terra demonstration
house (Fig.4.1b, Chen, 2007). A brief description of the PV/T system is provided below.
The top of the channel consists of an amorphous silicon PV module with 6% efficiency
under standard test conditions (STC). The PV module consists of seven different layers
as listed from top to bottom – Tefzel, antireflective coating, amorphous silicon, backing
substrate, Tefzel, adhesive and a stainless steel sheet to which this is pasted. The PV
module used and its properties are listed in the table below with the components listed
from top of the module to the bottom). There is a gap of approximately 4 cm between the
PV module and the bottom of the channel, which are kept together by means of a wooden
frame. The bottom of the channel consists of 50 mm of polystyrene insulation R-10 (1.76
K-m2/W) and 9.5 mm thick plywood board (Fig.4.2).
68

Figure 4.1 (a) Experimental PV/T setup replicating the PV/T system of the Eco-Terra
house with and without amorphous PV modules attached (Candanedo, 2010); (b)
Photograph of the PV/T roof in Eco-TerraTM house (Chen, 2009)

The channel’s length (L) in the flow direction is 2.84 m, and its width is 0.387 m. The
channel is oriented with a due south azimuth angle and a tilt angle (β) of 45° (Fig.4.3).

Figure 4.2 Cross section details of PV/T system


69

Table 4.1. Components listed from top of the PV module to the bottom

Layer Parameter Value (SI)


Thickness 1 mm
Density 1750 kg/m3
TEFZEL
Specific Heat 1050 J/kg·K
Thermal Conductivity 0.24W/m·K
Antireflective coating Thickness 0.0005 mm
Thickness 0.0005 mm
Silicon Density 2330 kg/m3
Thermal Conductivity 148 W/m·K
Thickness 0.2 mm
Backing Substrate Density 7900 kg/m3
(Steel) Specific Heat 477 J/kg·K
Thermal Conductivity 14.9 W/m·K
TEFZEL Thickness 1 mm
Thickness 0.6 mm
Density 2150 kg/m3
Adhesive (Ethylene
propylene copolymer) Specific Heat 1100 J/Kg·K

Thermal Conductivity 0.2 W/m·K

Steel sheet Thickness 0.5 mm

The visible reflectance of the PV module was measured with a reflectometer as


4.3%. The long wave emissivity has been taken as 0.95 (Candanedo, 2010), the measured
emissivity of the steel plate is taken as 0.8 and that for an insulation is 0.2. The air flow
rate was a controlled parameter, measured with a laminar flow element (LFE) with an
accuracy of 0.4%. Other measured variables include top and bottom temperatures for the
top, middle and bottom of the channel interior by means of thermocouple, electrical
output of the PV module, solar irradiation, wind speed and relative humidity represented
given in Chapter 3.
70

Figure 4.3 Setup of the experiment, Adapted from the experiment at Concordia
(Candanedo, 2010)

4.3 Modeling Approach

In the present research, a transient model is developed based on previous work at


Concordia University (Candanedo, 2010; Chen, 2009) using a control volume based
approach. The present work extends the previously developed models in two important
ways:
1. The model uses new convective heat transfer coefficient correlations developed
through CFD simulations.
2. It accounts for the effect of buoyancy on the convective heat transfer in the PV/T
channel which is significant for lower flow rates.
The need for transient modeling followed by the parameters involved is explained
in the following sub-section.
71

4.3.1 Necessity for Transient Solution

To observe the transient effect on heat capacity, Klein et al (1974) conducted


studies on a flat plate collector by using a single node having capacitance at every layer.
It was concluded that a zero heat capacitance model would provide equally good results
as a one node model. It was observed by both Klein et al (1974) as well as Duffie and
Beckman (1991) that for predicting long-term behavior of south facing collectors such as
useful heat, the transient effects would not be essential, however if the collector was not
south facing, it would become necessary to consider transient effects, i.e, consider the
capacitance of the PV module.
Ito (2006) conducted a transient analysis to evaluate the performance of a flat
plate collector using a fully explicit finite difference model. It was concluded that
transition effects due to capacitance of the PV were important for predicting the outlet
temperature and short-term behavior. It was also established that since the transient
model accounted for the effect of rapid changes (wind speed and solar radiation
fluctuations). It could be used for the development of robust control algorithms.
In the recent work by Candanedo (2010) it was concluded that the transient model
gives better results for a control volume based finite difference system when compared to
a steady state model. Hence, a transient model which considers the effect of capacitance
of the PV module is considered in the present study.

4.3.2 Solar Geometry – Computing the Incidence Angle

A number of factors determine the efficiency of the PV module and the amount of
heat gained by the air behind the module. The most important factor is the solar radiation
incident on the surface of the PV module. This depends on the geographic location, sky
conditions, the orientation and tilt angle of the roof that the PV module is mounted on, etc.
In order to determine the incident sunlight, the solar incidence angle needs to be
calculated. The incidence angle is a function of the day number, time of the day and the
geographic location that in turn determines the position of the sun in the sky. The
equations for calculating the solar incidence angle are listed below.
72

The equation for the equation of time (𝑒𝑡 ) is defined as the difference between the
standard time and the solar time given by the equation:

𝑒𝑡 = 229.2 (0.000075 + 0.001868 cos 𝛽 − 0.04089 sin 𝛽)


4.1

Where, 𝛽 is defined by an intermediate variable that depends on the day number:

2𝜋
𝛽 = (𝑛 − 1).
365
4.2

Next, the apparent solar time (𝑡𝑎𝑝𝑝𝑎𝑟𝑒𝑛𝑡 ) is determined. It is determined by the


daily apparent motion of the observed sun, defined as a function of the geographical
location and equation of time, given by the following equation:

𝐿𝑙𝑜𝑛𝑔.𝑠𝑡𝑑 − 𝐿𝑙𝑜𝑛𝑔 𝑒𝑡
𝑡𝑎𝑝𝑝𝑎𝑟𝑒𝑛𝑡 = 𝐿𝑠𝑡𝑑 + 4 � �+
60 60
4.3

The hour angle ( ℎ ) is calculated next. This is a measure in an equatorial


coordinate system for a particular location on the earth is zero and noon, negative before
and positive after. In a 24 hour period, it changes from 0 to 360 degrees:

15𝜋
ℎ = �𝑡𝑎𝑝𝑝𝑎𝑟𝑒𝑛𝑡 − 12� ∗
180
4.4
Solar declination (𝛿) and altitude (𝑎𝑙𝑝ℎ𝑎) are also need to be defined in order to
determine the solar incidence angle. Solar declination is the angle between the earth-sun
line and the equatorial. This is defined in terms of an intermediate variable defined above,
𝛽 which is a function of day number. Solar altitude is the angular height of the sun
measured from the horizon, positive during day and negative at night.
73

These are represented via equation as shown below:

𝛿 = 0.3964 − 22.9133 cos 𝛽 + 4.0254 sin 𝛽 + 0.3872 cos 2𝛽 + 0.0520 sin 2𝛽


− 0.01545 cos 3𝛽 + 0.0848 sin 3𝛽

4.5

𝑎𝑙𝑝ℎ𝑎 = sin−1(cos 𝐿𝑙𝑜𝑐 . cos 𝛿. cos ℎ + sin 𝐿𝑙𝑜𝑐 . sin 𝛿)


4.6

All the above calculated parameters are required for the calculation of the solar
incidence angle (𝜃𝑖 ), defined as the angle between the sun’s rays and a line normal to this
surface. This value directly determines how much of sunlight is incident on the PV
module.

𝜃𝑖 = cos −1(cos 𝑎𝑙𝑝ℎ𝑎 . cos(𝜑 − 𝜓) . sin 𝛽 + sin 𝑎𝑙𝑝ℎ𝑎 . cos 𝜃)


4.7

4.3.2.1 Incidence Angle Adjustments

The effect of the variation in optical properties (transmittance, reflectance and


absorptance) of significant surfaces needs to be accounted for as a function of the angle
of incidence (Candanedo, 2010). Fanney (2003) and King (2004) have considered
incidence angle adjustments in investigations while dealing with the electrical
performance of PV systems. A correction curve is developed for this specific amorphous
PV module calculated according to the procedure described by King (1997) and available
data from Sandia National Laboratories database (2006). The absorbed solar radiation, S,
is a parameter that significantly affects the electricity produced by the PV module. The
absorbed solar radiation at reference conditions is given by the following equation:
𝑆𝑟𝑒𝑓 = 𝐺𝑟𝑒𝑓 (𝜏𝛼)𝑏
4.8
𝑀𝑟𝑒𝑓 is defined as the air mass modifier at the standard reference condition. The
air mass ratio of air that the beam radiation has to traverse at a specific location and time
to the amount of air it must traverse if the sun were directly overhead and the air mass
74

modifier ratio 𝑀/𝑀𝑟𝑒𝑓 is expressed as a summation of empirical parameters that are a


function of the air mass. King (1998) uses the following expression for the air mass:
1
𝐴𝑀 =
cos 𝜃𝑧 + 0.5057 (96.080 − 𝜃𝑧 )−1.634

4.9

Where, 𝜃𝑧 is the zenith angle in degrees.

King (2004) uses the following expression for the air mass modifier:
4
𝑀
= � 𝑎𝑖 (𝐴𝑀)𝑖
𝑀𝑟𝑒𝑓
0

4.10

Where, 𝑎𝑖 are the constants obtained from PV Sandia database given by:

𝑀
= 1.047 + 0.00082115 𝐴𝑀 − 0.0259 𝐴𝑀2 + 0.0031736 𝐴𝑀3
𝑀𝑟𝑒𝑓

− 0.00011026 𝐴𝑀4

4.11

Duffie and Beckman (2006) provide different methods to compute the absorbed
solar radiation, S. For the case where there is global (direct + diffuse) measurement of
solar radiation IT on the plane of the module, a reasonable assumption is to assume that
the absorbed solar radiation is the following:

𝑆 = (𝜏𝛼)𝑎𝑣 𝐼𝑇 = 0.96(𝜏𝛼)𝑏 𝐼𝑇
4.12

Using the results by King et al. (1998), Duffie and Beckman (2006) arrive at the equation:

(𝜏𝛼)𝑏
𝐾𝜏𝛼𝑏 (𝜃) =
(𝜏𝛼)𝑛
4.13
75

For this module, the value of (𝜏𝛼)𝑛 has been taken as shown below:
1 − 𝜌 = 1 − 0.043 = 0.957

4.14

Then the previous equation can be written in terms of (𝜏𝛼)𝑛 as shown below:
𝑆 = 0.96(𝜏𝛼)𝑏 𝐼𝑇 = 0.96𝐾𝜏𝛼𝑏 (𝜃)(𝜏𝛼)𝑛 𝐼𝑇
4.15
The ratio of 𝑆/ 𝑆𝑟𝑒𝑓 can be expressed as:

𝑆 0.96(𝜏𝛼)𝑏 𝐼𝑇
=
𝑆𝑟𝑒𝑓 𝐺𝑟𝑒𝑓

4.16

Finally for performance evaluation, the measured radiation is multiplied by the correction
factors as shown below:
𝑀
𝑆 = 𝑆𝑟𝑒𝑓 . 0.96. 𝐾𝜏𝛼𝑏 (𝜃).
𝑀𝑟𝑒𝑓

4.17

4.3.3 Effect of View Factor and Inlet Air Temperature

4.3.3.1 Effect of View Factor

The majority of the models assume that the view factor between two opposite
surfaces in a channel is equal to 1. This assumption may not yield accurate results since
the considered geometry is a duct. Thus the view factor Fij is calculated as a function of
geometric parameters suggested by Incropera & De Witt (2006) using the equation below:
76

1
2 (1 + 𝑋 ′2 )(1 + 𝑌 ′2 ) 2
𝐹𝑖𝑗 = {ln � �
𝜋. 𝑋 ′ 𝑌 ′ 1 + 𝑋 ′2 + 𝑌 ′2

1 𝑋′
+ 𝑋 ′ (1 + 𝑌 ′2 )2 tan−1 1
(1 + 𝑌 ′2 )2

1 𝑌′
+ 𝑌 ′ (1 + 𝑋 ′2 )2 tan−1 1 − 𝑋 ′ tan−1 𝑋 ′ − 𝑌 ′ tan−1 𝑌′
(1 + 𝑋 ′2 )2

4.18

Where, x = length of the channel, y= width of the channel, L= depth of the channel

𝑋 ′ = 𝑥/𝐿 , 𝑌 ′ = 𝑦/𝐿

4.3.3.2 Effect of Inlet Air Temperature

In PV/T systems, the inlet air temperature is higher than the exterior air
temperature by a few degrees since the inlet air has been warmed by the thermal energy
released by the building’s façade. Saelens (2004) developed an energy model for a
building with a multiple skin facade and results showed that reliable energy assessment
depends on the right choice of boundary conditions, especially that of inlet air
particularly for cases that air is reused after flowing through a cavity. For the experiment
considered in the present study, the difference between exterior and measured inlet air
temperature can be observed in Fig.4.4. It can be noted that a difference of 2-6oC exists
between the ambient and inlet air temperature, thus be determined with certainty that if
the right temperature is not used, it could lead to error in the final outlet air temperature
which is a major output for a PV/T systems.
77

Figure 4.4 Comparison between measured ambient and inlet air temperature, February
17th, 2009

4.3.4 Effect of Interior Radiative Heat Transfer Coefficient

The radiative heat transfer is calculated as a product of the radiative heat transfer
coefficient and the view factor between which the radiation heat transfer occurs.
𝑞𝑟𝑎𝑑 = ℎ𝑟 . 𝐴𝑐𝑣 . 𝐹𝑖𝑗
4.19
Where,
4 4
𝜎�𝑇𝑡𝑜𝑝 − 𝑇𝑏𝑜𝑡𝑡𝑜𝑚 �
ℎ𝑟 =
1 1
(𝜀 + 𝜀 − 1)
2 3

4.20

𝜀2 and 𝜀3 are the long wave emissivities of the top and bottom surfaces, measured by a
hemispherical emissometer (This procedure for evaluating radiative heat transfer
coefficient is essentially for long-parallel surfaces, it is used here for as an
approximation). As already mentioned earlier, these values are found to be 0.8 and 0.2
78

respectively. 𝐹𝑖𝑗 is the view factor between the two cavity surfaces explained in the
previous section.
An alternate method of finding the radiation heat transfer is to use the mean
temperature of the surrounding surfaces (Tm) to calculate a linearization factor 4𝜎𝑇𝑚3
instead of using the surface temperatures. The radiative heat transfer coefficient, hr is
1 1
then given as 4𝜎𝑇𝑚3 /(ε + ε − 1) , assuming a view factor of 1 between the plates.
2 3

However this overestimates the actual radiative heat transfer by 1.5%. For accurate
calculation, Eq.4.19 is used in the energy model.

4.3.5 Effect Convective and radiative Heat Transfer Coefficient

4.3.5.1 Exterior Convective Heat Transfer Coefficient

During the experiment, the wind speed was measured at a height of 10 m above
the roof of the building where the PV/T system was placed. The measured wind speed is
adjusted to the height of the PV/T system using the following power-law formula from
ASHRAE (2012):
𝛼𝑤
𝑍
𝑉𝑤𝑖𝑛𝑑 = 𝑉𝑔 � �
𝑍𝑔

4.21

Where,𝑉𝑤𝑖𝑛𝑑 is the wind speed at the PV/T height and 𝑉𝑔 is the reference wind
speed, 𝑍𝑔 is 10m (the height at which the reference wind speed is measured) and Z is the
height at which the PV/T system is mounted. Here, 𝛼𝑤 is the power law exponent taken
as 0.36 to represent the terrain at the city center.
This wind speed 𝑉𝑤𝑖𝑛𝑑 is used to determine the exterior convective heat transfer
coefficient. The Sharples and Charlesworth (1998) equation is the correlation of choice
(Eq. 4.22) since in a parametric analysis performed by Candanedo (2010), it was
determined to yield the most accurate results. A similar parametric analysis determining
the impact of the choice of exterior CHTC on the outlet air temperature and PV
efficiency is presented in section 4.5.1.
79

ℎ𝑜 = 11.9 + 2.2 𝑉𝑤𝑖𝑛𝑑


4.22

4.3.5.2 Effect of Exterior Radiative Heat Transfer Coefficient

The exterior radiative coefficient can be calculated using the temperature of the
sky, temperature on the top of the PV module, long wave emissivity of the top layer. The
sky temperature, employed to calculate radiative heat losses to the exterior is obtained
with the following equation provided by Duffie and Beckman (2006):
𝜋. 𝑡
𝑇𝑠𝑘𝑦 = 𝑇𝑎 �0.711 + 0.0056 𝑇𝑑𝑝 + 0.013 𝑐𝑜𝑠 � ��
12

4.23

Where, 𝑇𝑎 is the temperature of the air and 𝑇𝑑𝑝 is the dew-point temperature of air.
The specific heat capacity of air 𝑐𝑝 has been calculated for conditions of temperature and
relative humidity measured at the inlet of the channel using the equation given below:
𝑐𝑝 = 𝑐𝑝𝑎 + 𝑊. 𝑐𝑝𝑣
4.24

Where, W is the air moisture content calculated using the relative humidity (RH)
and dry bulb temperature of air. 𝑐𝑝𝑎 is the specific heat of dry air taken as 1.0 kJ/(kga-K)
and 𝑐𝑝𝑣 is the specific heat of water vapor taken as 1.86 kJ/(kgv-K). The long wave
emissivity was measured and found to be 0.95. Thus piecing together the equations from
above, the equation for exterior heat transfer coefficient is then given by:

4 4
ℎ𝑟−𝑒𝑥𝑡𝑒𝑟𝑛𝑎𝑙 = 𝐹𝑟𝑜𝑜𝑓−𝑠𝑘𝑦 𝜀1 𝜀𝑠𝑘𝑦 𝜎 �𝑇𝑃𝑉 𝑇𝑂𝑃
− 𝑇𝑆𝑘𝑦 �
4.25
Here 𝜀𝑠𝑘𝑦 &𝐹𝑟𝑜𝑜𝑓−𝑠𝑘𝑦 are considered equal to 1. A more accurate formulation would
require accurate calculation of these parameters.
80

4.3.6 Interior Convective Heat Transfer Correlation

The average Nusselt correlations developed in the previous chapter using CFD
simulations are utilized to determine the rate of heat transfer from the channel to the air.
For the case with and without wooden framing, the correlations for mixed convection are
valid for 1100≤ Re ≤ 3200 while for forced dominated convection, the correlations are
valid for 5000≤ Re ≤ 24000 for the smooth channel and 6000≤ Re ≤ 18000 for the
channel with the wooden framing. The validity of the Reynolds number ranges are
explained in Chapter 3 (Section 3.4.5). The correlation for the case with forced-
dominated convection is given by Eq.4.26 while for mixed convection Eq. 4.27 was
utilized:

𝑁𝑢𝑓𝑜𝑟𝑐𝑒𝑑 = 0.084𝑅𝑒 0.658 𝑃𝑟 0.4


4.26
𝑅𝑒 0.539
𝑁𝑢𝑚𝑖𝑥𝑒𝑑 = 𝑁𝑢𝑛𝑎𝑡 [ 1 + 0.935. � � ]
𝐺𝑟 0.5
4.27
For the case with the wooden battens and forced-dominated convection Eq. 4.28 was
used while for mixed convection with Eq.4.29 was utilized:

𝑁𝑢𝑓𝑜𝑟𝑐𝑒𝑑 = 0.08𝑅𝑒 0.732 Pr 0.4


4.28
𝑅𝑒 0.303
𝑁𝑢𝑚𝑖𝑥𝑒𝑑 = 𝑁𝑢𝑛𝑎𝑡 [ 1 + 2.366. � � ]
𝐺𝑟 0.5
4.29
In both cases of mixed convection mentioned above:

𝐷 0.25
𝑁𝑢𝑛𝑎𝑡 = 0.645 �𝑅𝑎. �
𝐿
4.30
81

To understand the distribution of convective heat transfer coefficients over the


length of the channel, it is important to first examine the type of flow (laminar or
turbulent) and to judge the effect of development length for laminar and turbulent region
of the flow. For forced convection through pipes and ducts, the flow is classified based
on the Reynolds number – for Re<2300 (Laminar), 2300<Re< 4000 (Transitional) and
Re>4000, turbulent. Similarly for free convection, the classification was made based on
the Grashof number – for Gr<108 (Laminar), 108<Gr<109 (Transitional) and Gr>109
(Turbulent). However for mixed convection no such specific range exists. However, from
the CFD work in Chapter 3 and based on Metais map (Fig.3.18 and Fig.3.19), it can be
inferred that the flow falls in the mixed convection transition regime which tends towards
turbulent. In order to determine the development length for laminar flow, Kays (2005)
recommends using the entry length solution for a circular tube, which can be extended to
ducts by substituting the diameter with the hydraulic diameter as shown below:

𝑥
~ 0.05𝑅𝑒. 𝑃𝑟
𝐷ℎ
4.31

For turbulent region, White (2003) states that the development length is shorter and
estimates it using the formula:
𝑥 1
~ 4.4𝑅𝑒 6
𝐷ℎ
4.32
Convective heat transfer coefficient is higher at the entrance region before fully
developed conditions are established in the channel. The entrance length covers between
36-52% of the shorter channel of length 2.84 m and between 18-26% of the longer
channel of length 5.5 m for a range of Reynolds number between 1100 and 7500.
Another important consideration is the thermal development length indicated by the use
of the Graetz number.
𝐷ℎ
𝐺𝑧 = 𝑅𝑒. 𝑃𝑟
𝐿
4.33
82

A Graetz number of 1000 is indicative of a thermally fully developed flow. For a


shorter channel, the Graetz number varies between 18-148 and for a longer channel, it
varies between 9-80. The focus of this research is to study the final outlet temperature
and the average temperature behind the PV module and hence a single uniform value will
be used for the value of interior convective heat transfer coefficient (hconv) for a given
flow rate.

4.3.7 Transient Model - Solution

The following are the assumptions for the transient model:


• Properties of all the materials remain constant and uniform
• Temperatures of the surfaces (PV, steel sheet, insulation board) are assumed to be
uniform inside the control volume
• No edge effects are considered
• No temperature variations are considered across the width of the channel
• No air leakage or mixing with the exterior air after entering the channel
• No humidification or dehumidification of the air stream

Given below is a diagram of a cross section showing the thermal network (Fig.4.5)

Figure 4.5 BIPV/T thermal network model (Adapted from Candanedo, 2010)
83

The PV module is split up into 2 sections – the Tefzel and the Mix since these
have different R-values and capacitances (Candanedo, 2010). Also, the capacitances are
included only in the top part of the channel since it is exposed to dynamic weather
conditions such as the wind and the irradiance, whereas the bottom of the channel is
insulated.
Since the transient model takes into account the capacitance of the PV module, a
fully explicit finite difference scheme is employed that requires the solution of the
previous time step. The time step for the model is taken as 1 second for the sake of
stability (~1.9 seconds) since this is an explicit scheme, where the time -step is calculated
using the formula:

𝐶𝑝𝑣 𝐶𝑚𝑖𝑥
𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 ≤ min � , �
1 1 1 1
+ +
𝑅𝑇𝑒𝑓𝑧𝑒𝑙 /2 𝑅𝑇𝑒𝑓𝑧𝑒𝑙 /2 𝑅𝑚𝑖𝑥 /2 𝑅𝑚𝑖𝑥 /2
4.34

In this study, the values of ℎ𝑐𝑡 and ℎ𝑐𝑏 are the same. They have been indicated
differently only for the purpose of testing whether the given problem required the use of a
single or dual coefficients. A programming tool MATLAB is used to numerically solve
the equations for successive time step and control volume.
The equation for the top of the PV panel represents an energy balance of the form
𝑞"𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑒𝑑 = 𝑞"𝑟𝑎𝑑𝑖𝑎𝑡𝑒𝑑 + 𝑞"𝑐𝑜𝑛𝑣𝑒𝑐𝑡𝑖𝑣𝑒 . The value 𝜀1 represents the long wave
emissivity of the PV panel and ℎ𝑜 represents the Sharples and Charlesworth CHTC.

𝑇𝑀𝐼𝐷1,𝑖+1 − 𝑇𝑃𝑉𝑇𝑂𝑃 ,𝑖 4 4
− 𝜀1 𝜎 �𝑇𝑃𝑉 ,𝑖+1 − 𝑇𝑆𝑘𝑦,𝑖+1 � − ℎ𝑜 �𝑇𝑃𝑉 𝑇𝑂𝑃 − 𝑇𝑜,𝑖+1 � = 0
𝑅𝑇𝑒𝑓𝑧𝑒𝑙 /2 𝑇𝑂𝑃

4.35

A two node model is adopted where the resistance of the each layer is split
equally between the capacitances. The solar radiation is transmitted through the
encapsulate material Tefzel and is absorbed at the Tpv node.
84

The energy balance for the layers between the PV is represented by the equations given
below:

𝛥𝑡 𝑇𝑃𝑉𝑇𝑂𝑃 ,𝑖 − 𝑇𝑀𝐼𝐷1,𝑖 𝑇𝑃𝑉,𝑖 − 𝑇𝑀𝐼𝐷1,𝑖


𝑇𝑀𝐼𝐷1,𝑖+1 = 𝑇𝑀𝐼𝐷1,𝑖 + � + �
𝐶𝑃𝑉 𝑅𝑇𝑒𝑓𝑧𝑒𝑙 /2 𝑅𝑇𝑒𝑓𝑧𝑒𝑙 /2

4.36

𝑇𝑀𝐼𝐷1,𝑖 𝑇
� + 𝑀𝐼𝐷2,𝑖 + 0.96𝐾𝜏𝛼𝑏 (𝜃)𝑖 (𝜏𝛼)𝑛 𝐼𝑇,𝑖 − 𝑃𝑒𝑙𝑒𝑐𝑡,𝑖 �
𝑅𝑇𝑒𝑓𝑧𝑒𝑙 /2 𝑇𝑀𝑖𝑥 /2
𝑇𝑃𝑉,𝑖+1 =
1 1
+
𝑅𝑇𝑒𝑓𝑧𝑒𝑙 /2 𝑅𝑀𝑖𝑥 /2

4.37

𝛥𝑡 𝑇𝑃𝑉,𝑖 − 𝑇𝑀𝐼𝐷2,𝑖 𝑇𝑃𝑙𝑎𝑡𝑒,𝑖 − 𝑇𝑀𝐼𝐷2,𝑖


𝑇𝑀𝐼𝐷2,𝑖+1 = 𝑇𝑀𝐼𝐷2,𝑖 + � + �
𝐶𝑃𝑉 𝑅𝑀𝑖𝑥 /2 𝑅𝑀𝑖𝑥 /2

4.38

The heat gained by the air inside the control volumes is given below:

𝑞𝑟𝑒𝑐,𝑖+1 = �𝑇𝑝𝑙𝑎𝑡𝑒,𝑖 − 𝑇𝑏 �ℎ𝑐𝑡 + (𝑇𝑖𝑛𝑠𝑢 − 𝑇𝑏 )ℎ𝑐𝑏

4.39

The energy balance for the insulation on the bottom of the channel and the air is
represented by the equations given below:

𝑇𝑎𝑡𝑡𝑖𝑐,𝑖
𝑇𝑏,𝑖 ℎ𝑐𝑏 + + 𝑞𝑟𝑎𝑑,𝑖
𝑅𝑖𝑛𝑠 + 𝑅𝑝𝑙𝑦𝑤𝑜𝑜𝑑 + 1/ℎ𝑐𝑖
𝑇𝑖𝑛𝑠𝑢,𝑖+1 =
1
ℎ𝑐𝑏 +
𝑅𝑖𝑛𝑠 + 𝑅𝑝𝑙𝑦𝑤𝑜𝑜𝑑 + 1/ℎ𝑐𝑖

4.40

The heat flux that is transmitted through the PV module via the plate is transferred
to the air via radiation and convection. Here, 𝜀2 and 𝜀3 represent the long wave emissivity
of the steel plate and insulation.
85

−1
1 1 4 4
𝑞𝑟𝑎𝑑,𝑖+1 = 𝜎 𝐹𝑖𝑗 � + − 1� (𝑇𝑃𝑙𝑎𝑡𝑒,𝑖 − 𝑇𝑖𝑛𝑠𝑢,𝑖 )
𝜀2 𝜀3

4.41

As mentioned before, the efficiency of the PV depends on the temperature


attained at that node less the performance reduction factor 𝛽𝑚𝑝 . The total electricity
produced by the PV is calculated as the modified efficiency times the incident solar
radiation, given by the equation (Candanedo, 2010) below:
𝜂𝑃𝑉,𝑖+1 = 𝜂𝑆𝑇𝐶 + 𝛽𝑚𝑝 (𝑇𝑃𝑉,𝑖 − 25𝑜 𝐶)

4.42

𝑃𝑒𝑙𝑒𝑐𝑡,𝑖+1 = 𝜂𝑃𝑉,𝑖 𝐼𝑇,𝑖

4.43

The channel is divided into a number of control volumes which facilitates the
calculation. The output of one control volume is treated as the input of the next control
volume. As the temperatures are assumed to be uniform in each control volume, the bulk
air temperature variation within each control volume is taken to vary exponentially and is
computed as an integral over the area. The time step 𝑖 + 1 is written to indicate the next
time step, written as a function of the current time step, i.

𝐴𝑐𝑣 𝑞𝑟𝑒𝑐,𝑖
𝑇𝑜𝑢𝑡𝑙𝑒𝑡,𝑖+1 = 𝑇𝑖𝑛𝑙𝑒𝑡,𝑖 +
𝑚𝑖 𝑐𝑝 𝑎𝑖𝑟,𝑖

4.44

1 𝛥𝑥 ℎ𝑐𝑡 𝑇𝑃𝑙𝑎𝑡𝑒,𝑖 + ℎ𝑐𝑏 𝑇𝑖𝑛𝑠𝑢,𝑖


𝑇𝑏,𝑖+1 = � ( + (𝑇𝑖𝑛𝑙𝑒𝑡,𝑖
𝛥𝑥 0 ℎ𝑐𝑡 + ℎ𝑐𝑏

𝑊𝑃𝑉 (ℎ𝑐𝑡 +ℎ𝑐𝑏 )𝑥


ℎ𝑐𝑡 𝑇𝑃𝑙𝑎𝑡𝑒,𝑖 + ℎ𝑐𝑏 𝑇𝑖𝑛𝑠𝑢,𝑖 − 𝑚𝑖 𝑐𝑝
− )𝑒 𝑎𝑖𝑟,𝑖 ) 𝑑𝑥
ℎ𝑐𝑡 + ℎ𝑐𝑏

4.45

The results from the model are discussed in detail in the section below.
86

4.4 Results – Model Performance

The model was verified using as input measured data for solar radiation, exterior
temperature, wind speed, channel flow rate corresponding to February 17th, 2009 between
9.30 am and 1 pm for the channel without battens and November 11th, 2009 between 9.45
am and 12 noon for the channel with battens. The variation in velocity inside the channel
along with the CHTC from the model for the newly developed correlation as well as a
previously developed correlation by Candanedo (2010) is represented (Fig.4.6). Wind
speed variation along with exterior convective heat correlation by Sharples and
Charlesworth are shown below (Fig.4.7). Finally the results for the average outlet
temperature are presented with the measured solar radiation values and flow rates within
the channel (Fig.4.8 & 4.9) for channel without and with wooden battens.
It can be observed from Fig.4.6 that the average CHTC observed by using
Candanedo correlation, till 10.40am, was 10.624 W/m2K whereas using the new
correlation, CHTC was 5.78 W/m2K. After 10.40 am, when the flow rate is reduced the
Candanedo correlation predicts CHTC as 5.074 W/m2K whereas the new correlation
predicts 15.136 W/m2K. The increase in heat transfer is attributed to natural convection
which has been found to enhance heat transfer by 3-4 times more than the estimated heat
transfer by forced convection alone (Gnielinski, 1983).
The output from the energy model for every 1 second is averaged over 1 minute
before comparing it to the thermocouple measurements taken at intervals of 1 minute. For
the analysis on hand, the time required by the system to reach steady state is calculated
using 0.63RC. After five time constants, steady state is attained. The combined thermal
capacitance of the module and the metal roofing sheet is 7838 J/m2K; their combined
thermal resistance is 0.011 m2K/W, hence the time constant was computed to be 86.2
seconds. The approximate time to reach steady state is 7.2 minutes. During this time, the
temperatures in the system change rapidly and hence a true measure of the how well the
equations perform is after the system reaches steady state.
87

Figure 4.6 Measured average speed in the channel and CHTC, February 17th, 2009

Sharples and Charlesworth Wind speed

16 7

14 6

Windspeed at 10m (m/s)


12
5
CHTC (W/mK)

10
4
8
3
6
2
4

2 1

0 0
9:28 9:57 10:26 10:55 11:24 11:52 12:21 12:50
Time (hh:mm)

Figure 4.7 Measured wind speed and derived heat transfer coefficients, February 17th,
2009
88

The algorithm was fixed such that for a Richardson number (Gr/Re2) greater than
0.25, it would pick a forced convection correlation (Eq.4.26, Eq.4.28) and for a value
greater it would pick a mixed convection correlation (Eq.4.27, Eq.4.29). A comparison
has been drawn between pure forced convection correlations used versus the correlation
selected according to the Richardson number (Fig.4.8 & Fig.4.9). During the experiment,
the flow rate is changed manually in a gradual manner, the average values for which are
represented in the colored bars in CFM. In general, there is reasonably good agreement
between the correlation and the experimental data. The forced convection correlation
developed predicts the average outlet air temperatures reasonably well but it slightly
under predicts these values as the flow rate is decreased. The mixed convection
correlation developed initially over predicts the outlet air temperature, but as steady
conditions are reached from 11:00 am onwards, the values predicted by the correlation
match very well. To check the validity of the newly developed correlation for the case of
the channel with battens (Eq.4.28-4.29) data from November 11th, 2009 is plotted against
the results from the energy model from 9.45 am till 12 noon.
A good agreement between the measured and numerical results is observed. Since
the flow rate is changed gradually, there is no abrupt shift in temperature observed like in
the case of the channel without the wooden battens. Thus, the validity of the newly
developed correlations is checked by presenting them against experimental data and a
good agreement between them is observed.
Next, the output from the energy model is used to quantify the performance of the
system using the correlations as inputs. The outlet air temperature is used to evaluate the
potential thermal energy (Eq.4.46) and also, the electric output from the PV is determined.

𝑄𝑡ℎ𝑒𝑟𝑚𝑎𝑙 = 𝑚. 𝑐𝑝 . (𝑇𝑖𝑛 − 𝑇𝑜𝑢𝑡 )


4.46
The thermal and electrical outputs are evaluated (Fig.4.10) using the energy
model with the newly developed correlations (for forced and mixed convection) for the
weather conditions on February 17th, 2009. It can be observed that the two correlations
predict similar results for electric output but the thermal energy predicted by the new
89

correlation with forced correlation is 25.6 kWh as opposed to 33.378 kWh predicted by
the new correlation with mixed convection, which is 30% higher. Figure 4.11 shows
results for thermal and electric outputs are evaluated using the newly developed
correlations for mixed convection for a smooth and a ribbed channel for February 17th,
2009.

Figure 4.8 Comparison of outlet air temperatures for smooth channel, February 17th, 2009

It can be observed that the two correlations predict similar results for electric
output whereas the results for the thermal output vary significantly. The thermal output
for the smooth channel is lower than the thermal output for a channel with wooden
framing during periods of higher flow rate. The increased heat transfer may be due to the
increased turbulence due to the presence of wooden framing elements present inside the
channel. As the flow rate reduces, it can be seen that the difference in thermal output for
90

the two cases becomes closer and at a low flow rate, the thermal output of the smooth
channel is slightly better, this may be due to the absorption of some heat by the wooden
framing at lower velocity.
The total thermal output for the smooth channel was 33.378 kWh whereas for the
channel with wooden framing it was 35.215 kWh. Overall, the channel with the wooden
battens performs better in terms of thermal output and hence is recommended for use in
roof integrated PV/T systems.

Figure 4.9 Comparison of outlet air temperatures for channel with framing, November
11th, 2009
91

Figure 4.10 Thermal and electric output for different correlations for February 17th, 2009

Figure 4.11 Thermal and electric output for smooth and ribbed channels
92

4.5 Parametric Analysis

For the all the parametric analysis performed the same weather data from
February 17th, 2009 was used for the smooth channel of length 2.84 m.

4.5.1 Effect of Exterior Heat Transfer Coefficient

The choice of exterior convective heat transfer correlation impacts the


temperature on top of the channel which in turn influences the heat gained by the air
inside the channel. For the exterior CHTC, the Sharples and Charlesworth (1998)
correlation is used which is listed below:

ℎ𝑜 = 11.9 + 2.2 𝑉𝑤𝑖𝑛𝑑


4.47
Where, the value of 𝑉𝑤𝑖𝑛𝑑 is evaluated at 1.5m above the ridge of the inclined PV/T
system. The exterior CHTC correlation is halved and doubled to observe the impact on
the thermal efficiency. Figure 4.12 depicts the impact of the exterior CHTC on outlet air
and thermal efficiency.
It can be observed that the choice of the exterior HTC has a direct impact on the
temperature of air at the outlet and hence the thermal efficiency, i.e., higher the value of
exterior CHTC, lower the thermal efficiency as more heat is convected away from the top
surface. The right choice of the heat transfer coefficient is important both inside and
outside the channel to evaluate accurately the performance of the PV/T system.
Candanedo (2010) stated that the Sharples and Charlesworth correlation was chosen over
McAdams (1954) and Test (1981) correlations as it predicted channel surface top
temperature closest to the measured value. A detailed investigation of convective heat
transfer from an inclined roof of a low-rise building (with application to BIPV/T systems)
was performed by Karava et al. (2011; 2012); however the correlations developed in this
study could not be used herein, as the experimental set up was placed on the roof of a
building and the approaching flow conditions would be different.
93

Figure 4.12 Parametric analysis for determining the impact of exterior CHTC on Outlet
Air and PV Efficiency

4.5.2 Effect of Interior Heat Transfer Coefficient – Bottom

This study considers only one convective heat transfer coefficient inside the
channel, with the value of the CHTC of the top and bottom being treated as the same in
the solution algorithm, i.e, ℎ𝑐𝑏 = ℎ𝑐𝑡 .The purpose of this analysis is to resolve whether or
not dual coefficients are required to gauge the heat transfer phenomena inside the channel.
The value of ℎ𝑐𝑏 is varied from zero to three times the value of ℎ𝑐𝑡 is used to observe the
impact as shown below (Fig.4.13).
94

Figure 4.13 Parametric analysis for assessing the impact of bottom CHTC, February 17th
2009

It can be observed that the variation in value of the outlet temperature is almost
negligible when the value of ℎ𝑐𝑏 is varied between half to three times the initial value
(also shown in Fig 4.13). However, when the value of ℎ𝑐𝑏 is set to zero, the outlet
temperature is slightly higher. This is a hypothetical condition (when ℎ𝑐𝑏 is equated to
zero) since convection always occurs in the channel. Also, because the (near adiabatic)
bottom does not contribute significantly to the amount of heat added to air, it can be
concluded with certainty that the value of the bottom heat transfer coefficient does not
impact the heat gained by the air nor the PV module temperature and hence for the
purpose of this study, it can be neglected.
95

4.5.3 Effect of Capacitance & R-value of PV module

The value of stability is checked before a parametric study can be conducted for
checking the impact of the capacitance and the resistance. The values for the capacitance
and resistance are tabulated below along with the critical time step required.

Table 4.2 Components listed from top of the PV module to the bottom

Rtefzel Rmix Cpv Cmix tcritical


R&C 4.17E-03 7.33E-03 1837.5 6000 1.915594
2R 8.34E-03 1.47E-02 1837.5 6000 3.831188
3R 1.25E-02 2.20E-02 1837.5 6000 5.746781
2C 4.17E-03 7.33E-03 3675 12000 3.831188
3C 4.17E-03 7.33E-03 5512.5 18000 5.746781

4.5.3.1 Effect of Capacitance

This study models and studies the impact of a time-dependent variation of


temperature in the PV/T system. In order to fully understand the impact of the
capacitance and resistance in the study, a parametric analysis is performed, varying the
values to see their impact on temperature variation at the channel outlet. It can be
observed that as the value of capacitance is increased to three times the original value, the
average outlet air temperature decreases (Fig.4.14). One more aspect that has to be noted
is that, as the flow rate reduces (and the PV/T system reaches steady state), the effect of a
higher capacitance does not have much of an impact – from 11:50 am to 12:50 pm, the
outlet air temperatures remain the almost the same. Thus it can be concluded that
increasing the capacitance delays the time to attain steady state, but once steady state is
attained, it does not have a pronounced effect on the outlet air temperature. During the
course of the day, depending on the specific requirements, the flow rate is expected to
vary within the channel if a control algorithm is implemented such that the flow rate is
higher during the day (higher radiation, higher PV temperatures and more cooling
required) and lower in the evenings and afternoons. In order to cater to varying flow rates
96

within the channel, a transient model is a valuable tool. Thus it can be concluded that the
temperature is sensitive to the capacitance only when the flow rate varies over small
periods of time.

Figure 4.14 Parametric analysis for assessing the effect of thermal capacitance

4.5.3.2 Effect of R-value

A recent study that included the effect of resistance of a PV module was


performed by Ji (2009) and Candanedo (2010). They concluded that even when the
resistance of the PV was small, the heat flux across the PV module represented a
difference of at least a few degrees between both the sides. It can thus be inferred that the
97

resistance of the PV module cannot be treated as negligible due to the presence of several
layers in the module. This can be demonstrated by varying the value of resistance of the
PV module (Fig.4.15).

Cell temperature (0.5R) Outlet air temperature (0.5R)


Cell temperature (R) Outlet air temperature (R)
Cell Temperature (2R) Outlet air temperature (2R)
Cell temperature (3R) Outlet air temperature (3R)
Overall efficiency (0.5R) Overall efficiency (R)
Overall efficiency (2R) Overall efficiency (3R)

39 33 16.5 9.7 CFM


50 CFM CFM CFM 50

40 40

Efficiency (%)
Temperature (oC)

30 30

20 20

10 10

0 0
9:28 9:57 10:26 10:55 11:24 11:52 12:21 12:50
Time (hh:mm)

Figure 4.15 Parametric analysis for R-value – variation of outlet temperature, variation of
the PV node temperature and overall efficiency on February 17th, 2009

It can be observed that by varying the resistance of the PV module to triple the
original thermal resistance values, the temperature of outlet air was found to increase.
The rise in temperature of the outlet air is contrary to our expectation that the temperature
98

of the outlet air will decrease with the increase in R-value. This can be explained as
follows.
The temperature profile from the top of the PV to the top of the channel is unique
as the temperatures rises initially from the top of the PV to the PV cells and then it falls
between the PV cells and the top of the channel. This is because that the solar radiation
that is incident on the PV cells is absorbed at this point and the Tefzel material is only the
encapsulate that allows solar radiation to pass through it. The PV cells then absorbs the
solar radiation which it transmits through conduction - upwards to the top of the PV and
downwards to the top of the channel. The material between the PV cells and the top of
the PV is denoted as Tefzel and material between the PV cells and the top is denoted by
Mix. The temperature between the top and the bottom of the PV module is dependent on
two factors – the R-value and capacitance of the Tefzel and Mix layers. The actual R-
value of the Mix is twice that of Tefzel and the actual capacitance of the mix of Mix is
four times that of Tefzel (Table.4.2). When the R-value of the PV module (both Tefzel
and Mix) is doubled, the heat conductance through the material decreases from the PV
cells to the top of the PV as well as to the top of the channel, thus heating the PV cells to
a higher temperature. Also, the heat transmitted through the material is delayed by the
respective capacitances due to which the top of the channel gets heated to a greater extent
than the top of the PV. Thus when the R-value of the PV module is doubled or tripled, the
value of temperature (which depends on the R-value and the capacitance) on the top and
bottom of the PV module also increase proportionately, the change in the top of the
channel being higher than the change in the top of the PV. This can be observed in the
graph (Fig.4.15). Thus it can be concluded that, for design purposes, the R-value of the
PV module needs to be a minimum, since the temperature of the PV cell is sensitive to
the change in thermal resistance of the module. Keeping in mind that the number one
design priority is to increase the electrical efficiency, it is suggested to use a material with
high conductivity beneath the PV substrate to enhance the efficiency of the PV module.
99

4.6 Case Study

4.6.1 End Energy Use of Eco-Terra house

In order to comprehend the effectiveness of the model developed, an end energy


use scenario of the Eco-Terra house is studied for a day. The Eco-Terra house is a near
net zero house constructed with a PV/T system on the roof (10.2 m x 5.5 m) which is
connected to a ventilated concrete slab (VCS/PV-T and passive solar design) that stores a
part of this heat to be used for space heating while the rest is used for providing DHW
through an air-water heat exchanger (Chen, 2009). Data from Chen’s work (2009) is
obtained for January 12th, 2008 and the estimated heating load for the Eco-Terra house
was estimated to be 78 kWh (without occupancy – no internal load/ electrical appliances;
the PV/T system and VCS were not in use on this day).

The simplifications for the model are as observed below:

1. The data from TMY3 (air temperature, wind speed, diffuse and normal radiation)
are extracted to use as input to the model developed (the data are for the nearest
city Sherbrooke). Given the location and day number, the angle of incidence and
Irradiation are calculated from the Perez model (Appendix A). The highest wind
speed on this day was observed between 11 am and noon (~1.26m/s). For
Sherbrooke, the prevailing wind direction is from north-west, i.e. from the back of
the south-facing PV/T system. The energy model was modified to simulate the
values for a longer channel of dimensions 5.5x 10.2x 0.038 m.
2. The fan speed is set to run at 0.8 m/s between sunrise and sunset for Jan 12th,
2008.
3. An air-water heat exchanger, with 80% efficiency is utilized for space heating
(Chen, 2009) and 90% of the electric energy produced by the PV is used by
resistance heaters (DOE website). The solar irradiation is represented along with
the thermal and electrical efficiency in Figure 4.16.
100

For a constant flow rate of 550 CFM, The electrical efficiency is between 5.7- 6.1%
and it dips during the noon time when the solar irradiation is the highest, whereas the
thermal efficiency shadows the irradiation line. The results from the simulations indicate
a total thermal energy production of 33.75 kWh and electrical energy production of 7.8
kWh and these account for 53.3% (combined efficiency ~ 30-35%) of the measured
heating requirements of the house on this day. In the work by Chen (2009), the thermal
efficiency of the system was found to be not less than 20% during peak hours and the
electric efficiency of the PV was around 6%. It can be inferred that the transient model
can be used for energy analysis, to determine the energy that can be obtained from a real
PV/T system.

Figure 4.16 Thermal and Electric efficiency of the PV/T system on January 12th, 2008 for
Eco-Terra house, Quebec
101

4.6.2 Control Logic

As already mentioned earlier, the main design priority for a PV/T system is to
enhance the efficiency of the PV followed by enhancing the thermal efficiency. In order
for the PV module to function efficiently, the fan controlling the flow rate must be
adjusted such that not too much of the energy generated by the module is lost on running
the fan. A simple control logic has been proposed (Fig.4.17) based on earlier work on
PV/T systems by Chen (2009), Candanedo (2010), Candanedo (2011) and the present
study.

It can be observed that the priority is given for the PV cell temperature – if the
temperature of the PV cell (evaluated by the transient model) is higher than 25oC, the fan
is set to run at full speed till it drops. However, if the temperature of the PV cell is lower
than 25oC, the fan speed is decided depending on the irradiation level as seen in the flow
chart below. Similarly, the outlet air can be circulated to heat water (DHW), utilized for
space heating or exhausted depending on the temperature attained after circulating
beneath the PV module. This logic can be used as a building block to create more
advanced models and predictive control strategies for future research.
102

START

Location, weather data, Roof

specifications, Tattic

θ, G

Yes
G<300 W/m2
No Action
& To<5°C

No

Yes

300 ≤G
Vinlet =0.74Vmax
≤ 500

No

Yes

501 ≤G
Vinlet =0.83Vmax
≤ 800

No

A B

Figure 4.17 Control logic for operation of fan behind a PV/T system
103

A B

Yes

801 ≤G
Vinlet =0.91Vmax
≤ 950

No

G> 950
Vinlet =Vmax

Transient Energy model

Toutltet & TPV

Yes
Toutlet >
Drying clothes
o
15 C

No

C D

Figure 4.17 continued


104

C D

Yes
Toutlet - Twater
HX
> 6o C

No

Exhaust

FINISH

Figure 4.17 continued


105

CHAPTER 5. CONCLUSIONS AND RECOMMENDATIONS

5.1 Introduction

This chapter summarizes the Thesis by highlighting the findings and listing the
limitations in the study that was undertaken. In Section 5.2, a brief summary of the
research is provided followed by the research contributions in Section 5.3. Finally,
recommendations for future work are listed in Sections 5.4.

5.2 Summary of the Research

This objective of this work has been to develop models for open loop roof-
integrated PV/T systems to assess their overall performance. The first half of the work
was numerical, where a CFD tool FLUENT was executed to gauge the heat transfer in a
narrow duct behind a PV/T system for tilt angles of 30o and 45o and verified using
experimental data at 45o. Heat transfer correlations have been developed for a channel –
smooth and with wooden framing elements. After determining the importance of
buoyancy in the duct and establishing the limits for forced and mixed convection, the
effect of buoyancy aided flow was studied and correlations were represented in terms of
the Nusselt number, expressed as a function of the Reynolds and Prandtl number for
forced convection and in terms of Reynolds and Grashof number for mixed convection.
When compared against correlations within the limits of applicability, a good agreement
was noticed for a smooth channel with forced and mixed convection. Effect of heating
asymmetry on velocity and temperature profiles were investigated and it was concluded
that heating asymmetry did not have as much an impact on temperature profile across the
depth as it did on the velocity profile. The applicability of the new correlations was tested
106

for channel length and tilt and it was determined that the correlation can be used for
channel lengths greater than 1.5 m and for tilt angles between 30o and 45o.

The second half of the work dealt with the verification of the newly developed
correlations which was tested by implementing them in a transient thermal network
model and comparing them to experimental results. The thermal network model used one
heat transfer coefficient since only the top half of the channel that impacts the outlet air
temperature which was one of the required outputs from the model. The other output was
the PV efficiency. The agreement for both a channel with and without framing was found
to be very good for a combination of mixed and forced convection correlations
implemented through the algorithm. Parametric analysis was conducted for testing the
impact of thermal resistance (R-value), capacitance and effect of using a convective heat
transfer for the bottom half of the duct (hcb). While it was found that the thermal
resistance impacted the PV cell efficiency adversely and the thermal efficiency favorably,
the effect of changing the capacitance and bottom heat transfer correlation was found to
have negligible impact of the electrical and thermal efficiencies. A case study on the Eco-
Terra house was performed and it was found using the model that for the heating
requirements, the PV/T system was able to supply almost 53 % of the energy. Based on
the Irradiation levels and the PV cell temperature, a simple control logic was proposed
based on present findings and the work by Chen (2009) and Candanedo (2011).

5.3 Contributions of the Research

The main contributions of the Thesis can be summarized as follows:

1. The convective heat transfer coefficients are quantified and Nusselt number
correlations are developed for 4 cases – for a smooth channel (forced and
mixed convection regime) and channel with wooden framing (forced and
mixed convection regime) using CFD simulations with the SST k-ω
turbulence closure model. These correlations were verified against other works
from literature and from experimental data for both channels.
107

2. A transient lumped parameter model was developed based on the work by


Candanedo (2010) and the performance of the PV/T system was evaluated
using the newly developed correlations.
3. Parametric studies were performed to evaluate the impact of input parameters
such as R-value, capacitance, exterior CHTC and interior CHTC near the
bottom of the insulated channel. This was useful in understanding the
parameters that are required to optimize thermal and electrical energy
efficiency.
4. A simple control logic was also developed which can be developed for further
enhancement of model performance and predictive control strategies.

5.4 Recommendations for Future Work

This section addresses the research limitations and provides recommendations for further
research.
1. The experimental data obtained for verification of the correlations were from
an outdoor experiment where there were parameters that could not be
controlled (variable irradiation levels, wind speed and direction). It is
recommended to perform the experiment under well-controlled conditions
(using a solar simulator in a lab with controlled temperature settings)
considering flow conditioning measures (e.g. flow straightens, rounding the
edges of the duct, etc) at the inlet to eliminate disturbances.
2. To understand better the distribution of thermal gradient inside the channel,
experimental work can be conducted placing very few thermocouples inside
the channel to eliminate turbulent mixing by thermocouple wiring.
3. From the study, it was evident that asymmetric heating distorts the velocity
profile. To aid in better understanding of the flow field (with regard to
turbulence), Particle Image Velocimetry (PIV) can be utilized.
4. The increasing demand for renewable energy and advancement in technology
requires investigation of the performance of new materials. Implementation of
the model to facilitate and optimize the next generation of PV/T system using
108

third generation PV cells made of materials such as Amorphous Silicon (a-Si),


Gallium Arsenide (GaAs) and Copper Indium Gallium Selenide (CIGS),
embedded in them. These PV cells have an increased efficiency at low cost.
Developing concepts for improved energy yield and transient combined
thermal-electrical models for these innovative BIPV/T systems would be an
interesting extension of the present work.
5. Further investigation with regard to the channel depth, slope and inclusion of
fins/ battens is valuable in terms of easing the building integration of PV/T
systems while reducing pressure drops and enhancing thermal management
and heat recovery. Consideration of wind effects at a specific height for
predicting inlet air temperatures would provide valuable information.
LIST OF REFERENCES
109

LIST OF REFERENCES

Anderson TN, Duke M, Morrison GL, Carson JK. (2009). Performance of building
integrated photovoltaic/thermal (BIPVT) solar collector. Solar Energy 83, 445-455.

Aung W, Worku G, (1986). Developing Flow and Flow Reversal in a vertical Channel
with Asymmetric Wall Temperatures. ASME J. Heat Transfer. 108, 299-304.

Aung W, Worku G. (1986). Theory of Fully Developed, Combined Convection Including


Flow Reversal. ASME J. Heat Transfer. 108, 485-488.

Aung W, Worku G. (1987). Mixed Convection in ducts with asymmetric wall heat fluxes.
ASME J. Heat Transfer. 109, 947-951.

Bazilian M, Groenhout NK & Prasad D. (2001). Simplified numerical modelling and


simulation of a photovoltaic heat recovery system. In 17th European photovoltaic solar
energy conference, pp. 2387-2390. Munich, Germany.

Bazilian MD, Kamalanathan H & Prasad DK. (2002). Thermographic analysis of a


building integrated photovoltaic system. Renewable Energy 26, 449-461.

Bazilian MD & Prasad D. (2002). Modelling of a photovoltaic heat recovery system and
its role in a design decision support tool for building professionals. Renewable Energy 27,
57-68.
110

Blocken B, Defraeye T, Derome D, Carmeliet J. (2009). High-resolution CFD


simulations for forced convective heat transfer coefficients at the façade of a low-rise
building. Building & Environment 44, 2396-2412.

Brinkworth BJ, Cross BM, Marshall RH & Hongxing Yang. (1997). Thermal Regulation
of Photovoltaic cladding. Solar Energy 61, 169-178

Brinkworth BJ. (2000). Estimation of flow and heat transfer for the design of PV cooling
ducts. Solar Energy 69, 413-420.

Brinkworth BJ, Marshall RH & Ibarahim Z. (2000). A validated model of naturally


ventilated PV cladding. Solar Energy 69, 67-81.

Brinkworth BJ. (2000). A procedure for the routine calculation of laminar fre and mixed
convection in inclined ducts. International Journal of Heat and Fluid Flow 21, 456-462.

Brinkworth BJ & Sandberg M. (2005). A validated procedure for determining the


buoyancy induced flow in ducts. Building Service Engineering Research and Technology
26, 35-48.

Brown CK & Gauvin WH. (1965). Combined free and forced convection: 1. Heat transfer
in aiding flow. The Canadian journal of Chemical Engineering 1, 306-312.

Busedra AA & Soliman HM. (2000). Experimental investigation of laminar mixed


convection in an inclined semicircular duct under buoyancy assisted and opposed
conditions. International Journal of Heat and Mass Transfer 43, 1103-1111.

Cadafalch J, Oliva A, Van Der Graaf G, Albets X. (2003). Natural Convection in a Large,
Inclined Channel with Asymmetric Heating and Surface Radiation. Transactions ASME
125, 812-820.
111

Candanedo J. (2011). A Study of Predictive Control Strategies for Optimally Designed


Solar Homes. Ph.D Thesis, Concordia University, Montréal.

Candanedo L. (2010). Modelling and Evaluation of the Performance of Building


Integrated Open Loop Airbased Photovoltaic/Thermal Systems. Ph.D Thesis, Concordia
University, Montréal.

Charron R. (2004). One- and two-dimensional modelling of ventilated façades with


integrated photovoltaics. In Building, Civil and Environmental Engineering, pp. 191.
Concordia University, Montreal.

Charron R & Athienitis AK. (2006a). Optimization of the performance of double façades
with integrated photovoltaic panels and motorized blinds. Solar Energy 80, 482-491.

Charron R & Athienitis AK. (2006b). A two dimensional model of a double façade with
integrated photovoltaic panels. Journal of Solar Energy Engineering, ASME 128, 160-
167.

Chaube A, Sahoo PK, Solanki SC. (2006). Analysis of heat transfer augmentation and
flow characteristics due to rib roughness over absorber plate of a solar air heater,
Renewable Energy 31, 317-331.

Chen Y. (2009). Modeling and design of a solar house with focus on a ventilated concrete
slab coupled with a building integrated photovoltaic/thermal system. In BCEE, pp. 163.
Concordia University, Montréal.

Chen Y, Athienitis AK, Berneche B & Y. P. (2007a). Design and simulation of a building
integrated photovoltaic thermal system and thermal storage for a solar home. In 2nd
Canadian Solar Buildings Conference. Calgary.
112

Chen Y, Athienitis AK, Galal KE & Poissant Y. (2007b). Design and Simulation for a
Solar House with Building Integrated Photovoltaic-Thermal System and Thermal Storage.
In ISES Solar World Congress, Beijing, China, pp. 327-332.

Chen, Y. (2009). Modeling and Design of a Solar House with Focus on a Ventilated
Concrete Slab Coupled with a Building-Integrated Photovoltaic/thermal System. MASC
Thesis, Concordia University, Montréal.

Chong D, Liu J & Yan J. (2008). Effects of duct inclination angle on thermal entrance
region of laminar and transition mixed convection. International Journal of Heat and
Mass Transfer 51, 3953-3962.

Cotton MA, Jackson JD. (1990). Vertical tube air flows in the turbulent mixed convection
regime calculated using a low Reynolds k-ε model, International Journal of Heat and
Mass Transfer 33, 275-286.

Cross BM. (1994). Development, testing and first installations of an integrated solar roof
system. IEEE First world conference on photovoltaic energy conversion 1, 1020-1023.

Da Silva RM, Fernandes JLM. (2010). Hybrid photovoltaic/ thermal (PV/T) solar
systems simulation with Simulink/Matlab, Solar Energy 84, 1985-1996.

De Soto W, Klein SA & Beckman WA. (2006). Improvement and validation of a model
for photovoltaic array performance. Solar Energy 80, 78-88.

DGS. (2005). Planning and installing photovoltaic systems. James & James.

Dittus FW & Boelter LMK. (1930). Heat transfer in automobile radiators of the tubular
type. University of California Publications 2, 443-461.
113

Dittus FW & Boelter LMK. (1985). Heat transfer in automobile radiators of the tubular
type. International Communications in Heat and Mass Transfer 12, 3-22.

Duffie JA & Beckman WA. (1991). Solar Engineering of Thermal Processes, 2nd ed.,
Wiley, New York, pp. 289–291.

Duffie JA & Beckman WA. (2006). Solar engineering of thermal processes. John Willey
& Sons, Inc., Hoboken.

Dutta S, Zhang X, Khan J, Bell D. (1999). Adverse and favorable mixed convection heat
transfer in a two-sided heated square channel, Experimental Thermal and Fluid Science
18, 314-322.

Eiamsa-ard S, Promvong P. (2009). “Thermal characteristics of turbulent rib-grooved


channel flows”, Int. Commun. Heat Mass Transfer, 36, 705-711.

Eicker U. (2003). Solar technologies for buildings. West Sussex.

Fanney AH, Dougherty BP & Davis MW. (2003). Short term characterization of Building
Integrated Photovoltaic Panels. Journal of Solar Energy Engineering, ASME 125, 13-20.

FLUENT Inc. (2006). FLUENT 6.3 User's Guide. Lebanon.

Gau G, Yih KA, Aung W. (1992). Reversed Flow structure and heat transfer
measurements for buoyancy-assisted convection in a heated vertical duct, ASME Journal
of Heat Transfer 114, 928-935.

Gnielinski V. (1976). New Equations for Heat and Mass Transfer in Turbulent Pipe and
Channel Flow, International Journal of Chemical Engineering, vol. 16, pp. 359-368.
114

Gnielinski V. (1983). Forced convection in ducts, vol. 2. Hemisphere Publishing


Corporation, New York.

Greig D. (2012). An Experiment Investigation on the Flow Behavior in a Transpired Air


Collector. M.Sc Thesis, University of Western Ontario, Ontario.

Guimaraes PM, Menon GJ. (2008). Combination free and forced convection in an
inclined channel with discrete heat sources, International Communications in Heat and
Mass transfer 35(10), 1267-1274.

Hall WH, Jackson JD. (1969). Laminarization of a turbulent pipe flow by buoyancy
forces, ASME Paper 69-HT-55.

Han JC, Zhang HM , Lee CP. (1991). Augmented heat transfer in square channels with
parallel, crossed and V-shaped angled ribs, ASME, Journal of Heat Transfer 113, 590–
596.

Han JC, Zhang HM , Lee CP. (1992). Influence of surface heat flux ratio on heat transfer
augmentation in square channels with parallel, crossed, and V-shaped angled ribs, ASME,
Journal of Turbomachinery 114, 872–880.

Hatton AP & Quarmby A. (1963). The effect of axially varying and unsymmetrical
boundary conditions on heat transfer with turbulent flow between parallel plates.
International Journal of Heat and Mass Transfer 6, 903-914.

Hu ZJ, Shen J. (1996). “Heat transfer enhancement in a converging passage with discrete
ribs”, International Journal of Heat and Mass Transfer 39, 1719-1727.
115

Huang CC, Lin TF. (1996). Numerical Simulation of Transitional Aiding Mixedd
Convective Air Flow in a Bottom Heated Inclined Rectangular Duct, International
Journal of Heat and Mass Transfer 39 (8), pp.1697-1710.

Incropera FP & De Witt DP. (2006). Fundamentals of heat and mass transfer. John
Wiley & Sons.

Ito S, Kashima M & Miura N. (2006). Flow control and unsteady-state analysis on
thermal performance of solar air collectors. Journal of Solar Energy Engineering, ASME,
128, 354-359.

Jackson TW, Spurlock JM, Purdy KR. (1961). Combined Free and Forced Convection in
a Constant Temperature Horizontal Tube. AIChE J. 7, pp.38-45.

Jackson TW, Yen HH. (1971). Combined forced and free convective equations to
represent combined heat-transfer coefficients for horizontal cylinders, Journal of Heat
Transfer 93, pp 247-248.

Ji J, He H, Chow T, Pei G, He W & Liu K. (2009). Distributed dynamic modeling and


experimental study of PV evaporator in a PV/T solar-assisted heat pump. International
Journal of heat and mass transfer 52, 1365-1373.

Kakac S, Shah RK & Aung W. (1987). Handbook of single-phase convective heat


transfer.

Kakac S & Yener Y. (1995). Convective Heat Transfer, vol. 1. CRC Press, Boca Raton,
Florida.
116

Karava P, Jubayer C, Savory E. (2011). Numerical modeling of forced convective heat


transfer from the inclined windward roof of an isolated low-rise building with application
to photovoltaic-thermal systems. Applied Thermal Engineering 31, 1950-1963.

Karava P, Jubayer C, Savory E, Li, S. (2012). Effect of incident flow conditions on


convective heat transfer from the inclined windward roof of a low-rise building with
application to photovoltaic-thermal systems. Journal of Wind Engineering & Industrial
Aerodynamics 104-106, 428-438.

Kays WM. (1966). Convective Heat and Mass Transfer, 1st edition, McGraw-Hill.

Kays WM, Crawford ME. (1980) Convective Heat and Mass Transfer, 2nd edition,
McGraw-Hill, New York, pp. 161-163.

Kays W, Crawford M & Weigand B. (2005). Convective heat and mass transfer.
McGraw Hill.

King DL, Boyson WE & Kratochvil JA. (2004). Photovoltaic Array Performance Model.
Sandia National Laboratories.

King DL & Kratochvil JA. (1997). Temperature Coefficients for PV Modules and Arrays:
Measurement Methods, Difficulties, and Results In 26th IEEE Photovoltaic Specialists
Conference, pp. 6. Anaheim, California.

King DL, Kratochvil JA & Boyson WE. (1997). Measuring solar spectral and angle-of-
Incidence effects on photovoltaic modules and solar irradiance sensors. In 26th
IEEEPhotovoltaic Specialists Conference, pp. 5. Anaheim, California.
117

King DL, Kratochvil JA, Boyson WE & Bower WI. (1998). Field experience with a new
performance characterization procedure for photovoltaic arrays. In 2nd World Conference
and Exhibition on Photovoltaic Solar Energy Conversion pp. 7. Vienna, Austria.

King DL, Kratochvil JA & Boyson WE. (2000a). Stabilization and Performance
Characteristics of Commercial Amorphous-Silicon PV Modules. In IEEE photovoltaic
specialists conference, pp. 1446-1449. USA.

King DL, Quintana JA, Kratochvil JA, Ellibee DE & Hansen BR. (2000b). Photovoltaic
module performance and durability following long-term exposure. Progress in
photovoltaics 8, 241-256.

Klein, S. A., Duffie, J. A., and Beckman, W. A. (1974). Transient Considerations of Flat-
Plate Solar Collectors, ASME J. Eng. Power 96A, pp. 109–113.

Krauter S, Araujo RG, Schroer S, Hanitsch R, Salhi MJ, Triebel C, Lemoine R. (1999).
Combined Photovoltaic And Solar Thermal Systems For Facade Integration And
Building Insulation. Solar Energy 67, 239-248.

Liao L, Athienitis AK, Candanedo L, Park K-W, Poissant Y & Collins M. (2007).
Numerical and Experimental Study of heat transfer in a BIVP-Thermal system. Journal
of Solar EnergyEngineering, ASME 129, 423-430.

Martinelli RC. (1947). Heat transfer to Molten Metals. Transactions ASME 69, 947-959.

Maughan JR & Incropera FP. (1987). Experiments on mixed convection heat transfer for
airflow in a horizontal and inclined channel. International Journal of Heat and Mass
Transfer 30, 1307-1318.

McAdams WH. (1954). Heat Transmission. McGraw-Hill.


118

Metais B & Eckert ERG. (1964). Forced, mixed and free convection regimes. Journal of
Heat Transfer, 295-296.

Nikuradse J. (1966) Gesetzmabigkeiten der turbulenten stromung in glatten rohren,


Forsch. Arb, NASA TT F-10, 359.

Norton B, Eames PC, Mallick TK, Huang MJ, McCormack SJ, Mondol JD, Yohanis YG.
(2010). Enhancing the performance of building integrated photovoltaics, Solar Energy,
DOI: 10.1016/ j.solener.2009.10.004 .

Nualboonrueng T, Tuenpusa P, Ueda Y, Akisawa A. (2012). The performance of PV-T


systems for residential application in Bangkok, Progress in Photovoltaics: Research and
Applications. DOI: 10.1002/ pip.2181.

Oosthuizen PH, Madan S. (1970). Combined convective heat transfer from horizontal
cylinders in air, Journal of Heat Transfer 92, 194-196.

Osborne DG, Incropera FP. (1985). Laminar, Mixed Convection Heat Transfer for Flow
between Horizontal Parallel Plates with Asymmetric heating, International Journal of
Heat and Mass Transfer 28, pp.207-217.

Ozsunar A, Baskaya S, Sivrioglu M. (2001). Numerical Analysis of Grashof Number,


Reynolds and Inclination Effects on Mixed Convection Heat Transfer in Rectangular.
International Communications in Heat and Mass Transfer 28, pp.985-994.

Patankar SV & Spalding DB. (1972). A calculation procedure for heat, mass and
momentum transfer in three-dimensional parabolic flows. International Journal of Heat
and Mass Transfer 15, 1787-1806.
119

Pavlovic M, Penot F. (1991). Experiments in the Mixed Convection Regime in an


Isothermal open Cubic Cavity, Experimental thermal and Fluid Science 4, 648-655.

Petukhov BS, Kirillov VV. (1958). The Problem of Heat Exchange in the Turbulent flow
of Liquids in Tubes. Teploenergetika 4, pp.63-68.

Petukhov BS, Popov VN. (1963). Theoretical Calculation of Heat Exchange in the
Turbulent Flow of Liquids in tubes of an Incompressible Fluid with Variable Physical
Properties. High Temp. 1, pp.69-83.

Petukhov BS. (1976). Turbulent flow and heat transfer in pipes under considerable effect
of thermogravitational forces. In Heat transfer and turbulent buoyant convection, ed.
Spalding DB & Afgan N, pp. 701-717. Hemisphere Publishing Corporation, Washington.

Prandtl L. (1952). Fuhrrer durch die Stomungslehre, Vieweg, Braunschweig. p.359.

Promvonge P & Thianpong C. (2008). Thermal performance assessment of turbulent


channel flows over different shaped ribs. International Communications in Heat and
Mass Transfer 35, 1327-1334.

Rolle K., (2000) Heat and Mass Transfer, 1st edition, Prentice Hall, Upper Saddle
River, NJ.

Saelens D., (2002) Energy Performance Assessment of Single Storey Multiple-skin


Facades, PhD Thesis, Katholieke Universiteit Leuven.

Saelens D, Roels S & Hens H. (2004). The inlet temperature as a boundary condition for
multiple-skin facade modelling. Energy and Buildings 36, 825-835.
120

Sharples S & Charlesworth PS. (1998). Full-scale measurements of wind-induced


convective heat transfer from a roof-mounted flat plate solar collector. Solar Energy 62,
69-77.

Shome B, Jensen MK. (1995). Mixed Convection Laminar Flow and Heat Transfer of
Liquids in Isothermal Horizontal Circular Ducts, International Journal of Heat and Mass
Transfer 11, pp. 1945- 1956.

Sandia National Laboratories. (2006). Database of Photovoltaic Module Performance


Parameters.

Sparrow EM, Lloyd JR & Hixon CW. (1966). Experiments on Turbulent heat transfer in
an asymmetrically heated rectangular duct. Journal of Heat Transfer Transactions of the
ASME 88,170-174.

Tan HM & Charters WWS. (1969). Effect of thermal entrance region on turbulent forced
convective heat transfer for an asymmetrically heated rectangular duct with uniform heat
flux. Solar Energy 12, 513-516.

Tan HM & Charters WWS. (1970). An experimental investigation of forced convective


heat transfer for fully developed turbulent flow in a rectangular duct with asymetric
heating. SolarEnergy 13, 121-125.

Tang X, Zhu D. (2012). Experimental and Numerical Study on Heat Transfer


Enhancement of a Rectangular Channel with Discontinuous Crossed Ribs and Grooves.
Fluid Dynamics and Transport Phenomena 20, 220-230.

Test FL, Lessmann RC & Johary A. (1981). Heat transfer during wind flow over
rectangular bodies in the natural environmental. Transactions of the ASME J Heat
transfer 103, 262-267.
121

Tonui JK & Tripanagnostopoulos Y. (2007). Improved PV/T solar collectors with heat
extraction by forced or natural air circulation. Renewable Energy 32, 623-637.

Tonui JK & Tripanagnostopoulos Y. (2008). Performance improvement of PV/T solar


collectors with natural air. Solar Energy 82, 1-12.

Tripanagnostopoulos Y. (2007). Aspects and improvements of hybrid


photovoltaic/thermal solar energy systems. Solar Energy 81, 1117–1131.

Tripanagnostopoulos Y, Bazilian M, Zoulia I & Battisti R. (2000a). Hybrid PV/T system


with improved air heat extraction modification. Proceedings of the PV in Europe- from
PV technology to energy solutions conference.

Tripanagnostopoulos Y, Nousia T & Souliotis M. (2000b). Low cost improvements to


building integrated air cooled hybrid PV thermal systems. Proc 16th European PV Solar
Energy Conf, 1874-1877.

Tripanagnostopoulos Y, Nousia T, Souliotis M & Yianoulis P. (2002). Hybrid


photovoltaic/thermal solar systems. Solar Energy 72, 217-234.

Von Karman T. (1934). Turbulence and Skin Friction, Journal of Aerospace Science 4,
pp. 1-20.

Von Karman T. (1939). The Analogy between Fluid Friction and Heat Transfer,
Transactions ASME 61, pp. 705-710.

Wu T-H, Xu Z & Jackson JD. (2002). Mixed Convection Heat Transfer to Water Flowing
Through a Vertical Passage of Annular Cross Section: Part 2. Chemical Engineering
Research and Design 80, 246-251.
122

Yan WM, Lin TF. (1986). Natural convection heat transfer in vertical open channel
Flows with discrete heating, Proceedings of The Third National Conference on
Mechanical Engineering, CSME, pp. 157-167

Yilmaz T, Fraser SM. (2007). Turbulent Natural Convection in a Vertical Parallel-Plate


Channel with Asymmetric Heating, Internal Journal of Heat and Mass Transfer 50,
2612-2623.

Zhang X & Dutta S. (1998). Heat transfer analysis of buoyancy-assisted mixed


convection with asymmetric heating conditions. International Journal of Heat and Mass
Transfer 41, 3255- 3264.

Zondag HA. (2008). Flat-plate PV-Thermal collectors and systems – A review,


Renewable and Sustainable Energy Reviews 12, 891-959.
APPENDICES
123

Appendix A Solar Geometry and Perez Model

The following lines of code represent the calculation for solar incidence angle that was
formulated for the purpose of this study.

% Solar calculations for every minute of the day - BIPV

clc
clear

tstart = tic;

tol=0.01; % Tolerance for calculations

%Defining the t indices


h = 1:24; % No. of hours
hh = (1:24)'; % Hours in the day
m=0:(24*60-1); % minutes in the year
mm = (1:60)'; % minutes in every minute
n=12; % day of the year
interper = 1:((24-1)/1439):24; % interpolating for every minute of the
year

%Location information for Lafayette


Lstd=75 ; %degrees, East positive
Lloc=71+(54/60); %degrees, East positive
Lat=45+(24/60); %degrees, North positive
lat=deg2rad(Lat);

clear Lat
%---------------------------------------------------------------------
%Solar Geometry

% Fractional year, radians


B = ((2*pi)/365)* (n - 1 + ((hh-12)./24)'); % For every hour of the
year

%Equation of Time
et = 229.18*(0.000075 + 0.001868.*cos(B) - 0.032077.*sin(B) - ...
0.014615*cos(2.*B) - 0.040849.*sin(2.*B)); % minutes

% Declination
delta = 0.006918 - 0.399912.*cos(B) + 0.070257.*sin(B) - ...
0.006758.*cos(2.*B) + 0.000907.*sin(2.*B) - ...
124

0.002697.*cos(3.*B) + 0.00148.*sin(3.*B); % radians


deltam = interp1(h,delta,interper); % interpolating for every minute of
the year

% Time offset
offset = et - 4* (Lloc - Lstd);

%Solar Time
hhrep = reshape(repmat(hh,[1 60])',24*60,1); % for every minute of the
year
hhrep=hhrep';
minrep = repmat(mm',[1 24]);
tsol=(hhrep.*60)+minrep+reshape(repmat(offset,[60,1]),24*60,1)'; % for
every minute of the year

%Hour Angle
ha = deg2rad(tsol./4 - 180); %radians (all the minutes in the year)

%Zenith Angle for all the minutes in the year


z =
(acos(cos(lat).*cos(deltam).*cos(ha)+sin(lat).*sin(deltam))); %radians
day = z<(pi/2);
night = z>=(pi/2);
% z(night)=pi/2.;

%Solar Azimuth for all the minutes in the year


phis= acos((cos(z).*sin(lat) - sin(deltam))./(sin(z).*cos(lat)))...
.*(ha./abs(ha)); %radians

%Incidence angle
%(as a function of Beta and Psi, which are input in radians)
ftheta = @(psi, beta) min(pi/2, day.*acos(cos(z).*...
cos(beta) + sin(z).*sin(beta).*cos(phis-psi)) + ...
night.*(pi/2.)) ; %radians

ftheta2 = @(psi, beta) min(pi/2,


day.*acos((sin(deltam).*sin(lat).*cos(beta)) - ...
(sin(deltam).*cos(lat).*sin(beta).*cos(psi)) + ...
(cos(deltam).*cos(lat).*cos(beta).*cos(ha)) + ...
(cos(deltam).*sin(lat).*sin(beta).*cos(psi).*cos(ha)) + ...
(cos(deltam).*sin(beta).*sin(psi).*sin(ha))) + ...
night.*(pi/2.) ); %radians

% clear B delta et ttrep tsol phis lat Lloc Lstd night day
%--------------------------------------------------------------------

%Angle of incidence on given window


beta=pi/6.;
psi=0;
theta =(ftheta(psi,beta)); %South Vertical Wall
% figure, plot(1:24*60,rad2deg(theta(1:24*60)),1:24*60,rad2deg...
% (theta3(1:24*60)),1:24*60,rad2deg(theta4(1:24*60)))
125

%--------------------------------------------------------------------
%Final Illuminance Transmitted

[Ib, Id, It] = perez_pv(beta, theta, z);

function [Ib, Id, It, Eb, Ed, Et] = perez_pv(beta, theta, z)

tol = 0.001;

if (length(z) ~= length(theta))
fprintf('Arrays need to be of equal lengths\n')
end

%Weather Data
weather=load('weather.mat','c','d','e','f','g');
Iex=sparse(rot90(weather.c));
Ibn=sparse(rot90(weather.d));
Idh=sparse(rot90(weather.e));
Ebn=sparse(rot90(weather.f));
Ebn(1,25:24*31)=Ebn(1,25:24*31)*100;
Edh=sparse(rot90(weather.g));
Edh(1,25:24*31)=Edh(1,25:24*31)*100;

ref_g=repmat(0.8,1,length(z)); %Winter reflectance


ref_g((24*61)+1:24*330)=0.33; %Summer/Spring/Autumn reflectance

clear weather

%--------------------------------------------------------------------
% PEREZ MODEL

a=max(0,cos(theta));
b=max(cos(deg2rad(85)),cos(z));

%Clearness (in bins)


zcube=(5.535e-6).*(z.^3);
clearness=(zcube+((Idh+Ibn)./Idh)./(ones(1,length(z))+zcube));
clearness(Idh<=0)=0;
clearbin=repmat(9,[1,length(z)]);
%9 represents the lack of solar radiation, i.e. night time
clearbin(clearness>=1 & clearness<1.065)=1;
clearbin(clearness>=1.065 & clearness<1.23)=2;
clearbin(clearness>=1.23 & clearness<1.5)=3;
clearbin(clearness>=1.5 & clearness<1.95)=4;
clearbin(clearness>=1.95 & clearness<2.8)=5;
clearbin(clearness>=2.8 & clearness<4.5)=6;
clearbin(clearness>=4.5 & clearness<6.2)=7;
clearbin(clearness>6.2)=8;
clearbin=int8(clearbin);

%Brightness (delta)
bright=Idh./(Iex.*cos(z));
126

bright(abs(Iex.*cos(z))<=tol)=0;
bright = sparse(bright);

%IRRADIATION

Ib = max(0,Ibn.*cos(theta));

f1=[0 0 0; -0.008 0.588 -0.062; 0.130 .683 -.151; .330 0.487 -.221;...
.568 .187 -.295; .873 -0.392 -.362; 1.132 -1.237 -0.412;...
1.060 -1.6 -0.359; 0.678 -0.327 -0.250];
f2=[0 0 0; -0.06 0.072 -0.022; -0.019 0.066 -0.029;...
0.055 -0.064 -0.026; 0.109 -0.152 -0.014; 0.226 -0.462 0.001;...
0.288 -0.823 0.056; 0.264 -1.127 0.131; 0.156 -1.377 0.251];

F = [max(0,rot90(f1(clearbin,1))+(bright.*rot90(f1(clearbin,2)))+...
(z.*rot90(f1(clearbin,3)))) ; ...
rot90(f2(clearbin,1))+(bright.*rot90(f2(clearbin,2)))+...
(z.*rot90(f2(clearbin,3)))];

Id = max(0,Idh.*(bsxfun(@minus,1,F(1,:)).*(1+cos(beta))./2. +
F(1,:).*(a./b)...
+F(2,:).*sin(beta)));

It = Ib+Id+(Ibn.*cos(z) +
Idh.*ref_g.*((ones(1,length(z))+cos(beta))/2.));
It=sparse(It);

%ILLUMINANCE

Eb = max(0,Ebn.*cos(theta));

f1=[0 0 0; 0.011 0.57 -0.081; 0.429 0.363 -0.307;...


0.809 -0.054 -0.442; 1.014 -0.252 -0.531;...
1.282 -0.42 -0.689; 1.426 -0.653 -0.779;...
1.485 -1.214 -0.784; 1.17 -0.3 -0.615];
f2=[0 0 0; -0.095 0.158 -0.018; 0.05 0.008 -0.065;...
0.181 -0.169 -0.092; 0.275 -0.35 -0.096;...
0.38 -0.559 -0.114; 0.425 -0.785 -0.097;...
0.411 -0.629 -0.082; 0.518 -1.892 -0.055];

F = [max(0,rot90(f1(clearbin,1))+(bright.*rot90(f1(clearbin,2)))+...
(z.*rot90(f1(clearbin,3))));rot90(f2(clearbin,1))+...
(bright.*rot90(f2(clearbin,2)))+(z.*rot90(f2(clearbin,3)))];

Ed=max(0,(Edh.*(bsxfun(@minus,1,F(1,:)).*(1+cos(beta))./2. +
F(1,:).*(a./b)...
+F(2,:).*sin(beta))));

Et=Eb+Ed+(Ebn.*cos(z) + Edh.*ref_g.*((ones(1,length(z))+cos(beta))/2.));
Et=sparse(Et);
end
127

Appendix B Transient Model

The following lines of code represent the calculation for the transient model that was
formulated for the purpose of this study.
tic; % keeps track of the time elapsed for calculation
clear all
clc
DATA = load('data_file_nv_home_Y.txt'); % loads the text file
containing the values
SR1 = load('SR_4_home_Y1.txt');
Stuff= load('Stuff_x_home_Y1.txt');

%---- Values of the constants-sigma,emittances-----


sigma = 5.67E-8;
e1 = 0.20; % insulation
e2 = 0.80; % steel plate
e3 = 0.95; % PV panel
efi = 0.06; % efficiency of the panel

%-----R-values of the different materials------


Rsteel = 8.5E-5;
Rtefzel = 4.17E-3;
Rmix = 7.22E-3;
Rinsul = 1.9;

matrix = size(DATA);
rows = matrix(1,1);

%-----CAPACITANCES-----------------------------
Cap_pv = 1837.5; % capacitance o the PV module
Cap_b = 30; % capacitance of the PV module bottom
CAP_MIX = 6000; % capacitance of the mix layers

%----DIMENSIONS OF THE CHANNEL----------------


WidthPV = 0.387;
DeltaPV = 2.84;
displate = 0.04;
Le = (4*WidthPV*DeltaPV)/(2*WidthPV+2*DeltaPV); %effective length
Ar = WidthPV*DeltaPV; %area of the channel
Ac = 0.881; %area of collector exposed to sun

% Preallocating for speed


Tpv = zeros(1,rows);
ho=zeros(1440,1);
128

%--------INITIAL VALUES-----------------------

Tpv(1) = DATA(1,2);
Tma(1) = DATA(1,3);
Tb(1) = DATA(1,4);
qrad(1) = DATA(1,5)/Ar;
Pelect(1) = 2/Ac;
qsky(1) = DATA(1,6)/Ar;
Toutlet(1) = DATA(1,7);
qcv(1) = 1.332/Ar;

%-----------------------------------------

for i= 1:1440
ho(i) = 11.9+(2.2*Stuff (i,2)); % Sharples Correlation
end

velocity= Stuff(:,4);
Tsky= Stuff(:,3);
To= Stuff(:,1);
Tattic= Stuff(:,1);
Tinlet= Stuff(:,7);
mflow= Stuff(:,5);
cp_air= Stuff(:,6);

IA_Cor= SR1(:,2);
SR_temp = SR1(:,1).*IA_Cor*0.96; % SR= 0.96*It - Solar radiation
SR_temp1= SR1(:,1).*IA_Cor*0.8;
dT = ((24-1)/(86399)); % time interval for which we want the data for
dT1=((24-1)/1439);
T_t1= 1:dT1:24;
T_t = DATA(:,1); % time column

%INTERPOLATED VALUES-----------------------
T_t_int = DATA(1,1):dT:(DATA(rows,1));
T_t_int1 = DATA(1,1):dT:(DATA(rows,1));
T_t_int = T_t_int';
T_t_int1= T_t_int1';
T_t1= T_t1';
ho_int = interp1(T_t1,ho,T_t_int1,'linear');
SR_int = interp1(T_t1,SR_temp,T_t_int1,'linear');
SR_int1 = interp1(T_t1,SR_temp1,T_t_int1,'linear');
Tsky_int = interp1(T_t1,Tsky,T_t_int1,'linear');
To_int = interp1(T_t1,To,T_t_int1,'linear');
Tinlet_int = interp1(T_t1,Tinlet,T_t_int1,'linear');
mflow_int = interp1(T_t1,mflow,T_t_int1,'linear');
cp_air_int = interp1(T_t1,cp_air,T_t_int1,'linear');
Tattic_int = interp1(T_t1,Tattic,T_t_int1,'linear');
velocity_int= interp1(T_t1, velocity,T_t_int,'linear');
hcb_int(1)= 4;
hct_int(1)= 4;
%-------------------------------------------------------------------
129

n = numel(To_int); % total number of values of Tpv_top there will be


(n-1) times the equation will be solved

%---INITIAL VALUES OF CONSTANTS TO DESCRIBE EXPONENTIAL RISE--------


C1(1) = (hct_int(1)*Tpv(1)+hcb_int(1)*Tb(1))/(hct_int(1)+hcb_int(1));
C2(1) =
(WidthPV*(hct_int(1)+hcb_int(1)))/(mflow_int(1)*1225*cp_air_int(1));

%----NUMBER OF SECTIONS-------------
nsec = 6;
k = 1:nsec; %counter for the 6 CV
CAP_PV = Cap_pv;
CAP_B = Cap_b;
SECLENGTH = DeltaPV/nsec;

% Preallocating for speed


TPV_TOP =zeros(nsec,1);
TMID1 =zeros(nsec,1);
TPV =zeros(nsec,1);
TMID2 =zeros(nsec,1);
TPV_BOT =zeros(nsec,1);
TB =zeros(nsec,1);
TMA_int =zeros(nsec,1);
QRAD =zeros(nsec,1);
C_ONE =zeros(nsec,1);
C_TWO =zeros(nsec,1);
PELECT = zeros(nsec,1);
QSKY =zeros(nsec,1);
QCV= zeros(nsec,1);
TOUTLET = zeros(nsec,n);
beta_film= zeros(nsec,1);
Gr= zeros(nsec,1);
Re = zeros(nsec,1);
Ri = zeros(nsec,1);
Nu_nat = zeros(nsec,1);
TPV_BOT_avg = zeros(nsec,1);
TMA_int_avg = zeros(nsec,1);
hct_int=zeros(nsec,1);
hcb_int=zeros(nsec,1);
hct= zeros(nsec,1);

%---INITIAL VALUES IN SECTIONS------


for k=1:nsec
TPV_TOP(k,1) = Tpv(1);
TMID1(k,1) = Tpv(1)-0.5;
TPV(k,1) = Tpv(1)-0.8;
TMID2(k,1) = Tpv(1)-1;
TPV_BOT(k,1) = Tpv(1)-1.2;
TB(k,1) = Tb(1);
TMA_int(k,1) = Tma(1);
QRAD(k,1) = qrad(1)/nsec;
QCV(k,1) = qcv(1)/nsec;
C_ONE(k,1) = C1(1);
C_TWO(k,1) = C2(1);
130

PELECT(k,1) = Pelect(1)/Ar;
QSKY(k,1) = qsky(1)/nsec;
end

TB(1,2)=5;
TMID2(1,2)= 4;
TPV_BOT(1,2)= 10;
TMA_int(1,2)= 5;
Gr(1)= 4E05;
Ri(1)= 0.05;
Nu_nat(1)= 2;
Re(1)=500;
hcb_int(1)= 4;
hct_int(1)= 4;
hct(1)=4;

%--------------------------- Main Loop --------------------------------

for k = 1:nsec
for i = 1:n-1

% Number of times the equation is solved


% Calculation for CHTC

beta_film(k,i+1)= 2./(TPV_BOT(k,i)+ TMA_int(k,i)+ 273.15*2); %


Evaluated at film tmeperature
Gr(k,i+1)= (9.81.* beta_film(k,i+1).* 0.866.* (TPV_BOT_avg(k,i)- ...
TMA_int_avg(k,i))*0.074.^3)./(1.67E-05^2); % Grashoff number
Re(i+1) = (velocity_int(i+1)*0.074)./ (1.67E-05); % Reynolds number
Ri(k,i+1)= Gr(k,i+1)./(Re(i+1).^2); % Richardson number
Nu_nat(k,i+1)= 0.645*(Gr(k,i+1).*0.71*0.074/2.887).^0.25; % Natural
Convection

if Ri(k,i+1)<0.25
hct(k,i+1) = 0.024/.074*0.084*(Re(i+1).^0.658)*0.71^0.4;
else
hct(k,i+1) = Nu_nat(k,i+1)*(1+
0.935*(Re(i+1)./(Gr(k,i+1).^0.5)).^0.539);
end

hct_int(k,i+1)= hct(k,i+1);
hcb_int(k,i+1)=hct_int(k,i+1);

% ENERGY BALANCE EQUATIONS


%-----ENERGY BALANCE IN THE PV MODULE------------------------

TPV_TOP(k,i+1) = fzero(@(x)EBTOP(x,TMID1(k,i),To_int(i),Tsky_int(i)...
,Rtefzel,ho_int(i)),TPV_TOP(k,i));

TMID1(k,i+1) = TMID1(k,i) + (1/CAP_PV)*((TPV_TOP(k,i)-TMID1(k,i))/...


(Rtefzel/2) + (TPV(k,i)-TMID1(k,i))/(Rtefzel/2));

TPV(k,i+1) = (((SR_int(i)- PELECT(k,i))*(Rtefzel/2)*...


131

(Rmix/2)) + TMID1(k,i)*(Rmix/2) + TMID2(k,i)*(Rtefzel/2))/(Rmix/2 +


Rtefzel/2);

TMID2(k,i+1) = TMID2(k,i) + (1/CAP_MIX)*((TPV(k,i)-


TMID2(k,i))/(Rmix/2)+...
(TPV_BOT(k,i)-TMID2(k,i))/(Rmix/2));

TPV_BOT(k,i+1) = fzero(@(x) EBBOT(x,TMID2(k,i),TMA_int(k,i),TB(k,i),...


Rmix,hct_int(k,i+1)),TPV_BOT(k,i));

%-----ABSORBER PLATE------
TB(k,i+1) = ((TMA_int(k,i)*hcb_int(k,i+1)) + (Tattic_int(i)/Rinsul) +
QRAD(k,i))...
./ (hcb_int(k,i+1)+(1/Rinsul));

%-------------OTHER EQUATIONS-------------
C_ONE(k,i+1) =
((hct_int(k,i+1)*TPV_BOT(k,i))+(hcb_int(k,i+1)*TB(k,i)))/(hct_int(k,i+1
)...
+hcb_int(k,i+1));
C_TWO(k,i+1) =
(WidthPV*(hct_int(k,i+1)+hcb_int(k,i+1)))/(mflow_int(i)*1225....
*cp_air_int(i));

if k <= 1
TMA_int(k,i+1) = (1/SECLENGTH)*quad(@(x)(C_ONE(k,i) + (Tinlet_int(i)-...
C_ONE(k,i))*exp(-C_TWO(k,i)*x)),0,SECLENGTH);
else
TMA_int(k,i+1) = (1/SECLENGTH)*quad(@(x)(C_ONE(k,i) + (TOUTLET(k-1,i)-
...
C_ONE(k,i))*exp(-C_TWO(k,i)*x)),0,SECLENGTH);
end

if k <= 1
TOUTLET(k,i+1) = Tinlet_int(i) +
QCV(k,i).*(Ar/nsec)/(mflow_int(i)*1225....
*cp_air_int(i));
else
TOUTLET(k,i+1) = TOUTLET(k-1,i) +
QCV(k,i).*(Ar/nsec)/(mflow_int(i)*1225....
*cp_air_int(i));
end

QCV(k,i+1) = (hct_int(k,i+1).*(TPV_BOT(k,i)-TMA_int(k,i))) +
(hcb_int(k,i+1).*...
(TB(k,i)-TMA_int(k,i)));
QRAD(k,i+1) = (0.8479)*sigma*((1/e1+1/e2-1)^-
1)*((TPV_BOT(k,i)+273.15)^4....
- (TB(k,i)+273.15)^4);
QSKY(k,i+1) = sigma*e3*((TPV_TOP(k,i)+273.15)^4 -
(Tsky_int(i)+273.15)^4 );
PELECT(k,i+1) = (efi - (0.00014*(TPV(k,i)-25)))*SR_int1(i);

TPV_BOT_avg = mean(TPV_BOT,1);
132

TMA_int_avg = mean(TMA_int,1);

end
end

%------------END OF BIG LOOP------------------

TPV_avg = mean(TPV,1);
TPV_TOP_avg = mean(TPV_TOP,1);
TPV_BOT_avg = mean(TPV_BOT,1);
TB_avg = mean(TB,1);
TOUTLET_avg= mean(TOUTLET,1);
TMID2_avg= mean(TMID2,1);
TMID1_avg= mean(TMID1,1);
TMA_int_avg= mean(TMA_int,1);

a2= zeros(210,1);
a4= zeros(210,1);

a1=TOUTLET(6,34200:46800);
f=60;
for i=1:210
a2(i)= sum(a1(((f*(i-1)+1)):f*i+1))/f;
end
a3=TPV_BOT_avg(34200:46800);
for i=1:210
a4(i)= sum(a3(((f*(i-1)+1)):f*i+1))/f;
end
close all
toc

plot(Tinlet_int(1:86400,1),'DisplayName','Tinlet_int(1:86400,1)','YData
Source','Tinlet_int(1:86400,1)');figure(gcf)
xlabel('Time in seconds for full day')
ylabel('Temperature in deg.C')
133

Appendix C End Use Energy Model Data

The first graph (Fig.C1) represents the Solar Irradiation (W/m2) and outside
temperature (oC) as a function of time on Jan 12th, 2008. The second graph (Fig. C2)
depicts the variation of the wind speed and wind direction on the same day. The direction
(true or geographic, not magnetic) from which the wind blows is expressed in 10s of
degrees, 9 = 90 degrees true or an east wind, and 36 = 360 degrees true or a wind blowing
from the geographic north pole. A value of zero (0) denotes a calm wind.

Irradiation Outside air temperature


700 5.0

4.0
600
3.0
500 2.0
Irradiation (W/m2)

Temperature (oC)
1.0
400
0.0
300
-1.0

200 -2.0

-3.0
100
-4.0

0 -5.0
0:00 4:48 9:36 14:24 19:12 0:00
Time (hh:mm)

Figure C 1 Graph depicting variation of solar irradiation and outside air temperature on
Jan.12th, 2008
134

Wind Speed outside Wind direction


30 35

30
25

Wind direction (10s o)


25
20
Wind speed (km/h)

20
15
15

10
10

5
5

0 0
0:00 4:48 9:36 14:24 19:12
Time (hh:mm)

Figure C 2 Graph depicting variation of wind speed and wind direction on Jan.12th, 2008

S-ar putea să vă placă și