Modeling and Reconfiguration of Solar

Photovoltaic Arrays under Non-Uniform Shadow

Conditions

A Thesis Presented

by

Dung Duc Nguyen

to

The Department of Electrical and Computer Engineering

in partial fulfillment of the requirements

for the degree of

Doctor of Philosophy

in

Electrical Engineering

Northeastern University

Boston, Massachusetts

July, 2008
 Modeling and Reconfiguration of Solar
Photovoltaic Arrays under Non-Uniform Shadow
Conditions

by

Dung Duc Nguyen

Abstract

Mass production and use of electricity generated from solar energy has become very

common recently because of the environmental threats arising from the production of

electricity from fossil fuels and nuclear power. The obvious benefits of solar energy are

clean energy production and infinite supply of daylight. The main disadvantage is the

high cost. In these photovoltaic systems, semiconductor materials convert the solar light

into electrical energy. Current versus voltage characteristics of the solar cells are non-

linear, thus leading to technical control challenges. In the first order approximation,

output power of a solar array is proportional to the irradiance of sunlight. However, in

many applications, such as solar power plants, building integrated photovoltaic or solar

tents, the solar photovoltaic arrays might be illuminated non-uniformly. The cause of

non-uniform illumination may be the shadow of clouds, the trees, booms, neighbor’s

houses, or the shadow of one solar array on the other, etc. This further leads to

nonlinearities in characteristics.

2
Because of the nature of the electrical characteristics of solar cells, the maximum power

losses are not proportional to the shadow, but magnify nonlinearly [1]. Further, shadows

of solar PV array can cause other undesired effects:

• The power actually generated from the solar PV array is much less than designed.

At some systems, the annual losses because of the shadow effects can be reached

10%. Thus, the probability for “loss of load” increases [2].

• The local hot spot in the shaded part of the solar PV array can damage the solar

cells. The shaded solar cells may be work on the negative voltage region and

become a resistive load and absorb power. Bypass diodes are sometimes

connected parallel to solar cells to protect them from damage. However, in most

cases, just one diode is connected in parallel to group of solar cells [3], and this

hidden the potential power output of the array.

This proposed research will focus on the development of an adaptable solar array that is

able to optimize power output, reconfigure itself when solar cells are damaged and create

controllable output voltages and currents.

This study will be a technological advancement over the existing technology of solar PV.

Presently solar arrays are fixed arrays that require external device to control their output.

In this research, the solar array will be able to self-reconfigure, leading to the following

advantages:

• Higher efficiency because no external devices are used.

3
 • Can reach maximum possible output power that is much higher than the

maximum power of fixed solar arrays by arranging the solar cells in optimized

connections.

• Elimination of the hot spot effects.

The proposed research has the following goals: First, to create a modeling and computing

algorithm, which is able to simulate and analyze the effects of non-uniform changing

shadows on the output power of solar PV arrays. Our model will be able to determine the

power losses in each solar cell and the collective hot spots of an array. Second, to propose

new methods, which are able to predict the performance of solar PV arrays under shadow

conditions for long term (days, months, years). Finally, to develop adaptive

reconfiguration algorithms to reconfigure connections within solar PV arrays in real time,

under shadow conditions, in order to optimize output power.

4
 ACKNOWLEDGMENTS
First, I would like to express my sincere gratitude to my advisor, Prof. Brad Lehman for

believing in me and giving me this opportunity. I am thankful to Professor Lehman for

his thorough guidance, constant encouragement and altruistic support for all these years.

His extensive vision and creative thinking have been the source of inspiration for me

throughout this work. I truly appreciate all the hard work that he did for me so that I

graduated on time. It was an exhilarating learning experience to be one of his students.

I am grateful to my committee members, Professor Bahram Shafai and Professor Sagar

Kamarthi for their valuable suggestions and help.

I want to specially thank Vietnam Education Foundation and Graduate school of

Northeastern University for their financial and technical support in my research work. I

would also like to thank Dr. Yaman Yener, Director of the Graduate School, and Mr.

Brian McGuire from Graduate School.

It has been a great pleasure to work in the Department of Electrical and Computer

Engineering in Northeastern University. I would like to thank the ECE administrative and

management staffs, namely Cynthia Bates, Faith Crisley, Fernando Dellaporta, Paulette

Deery and Sharon Heath for their countless help during my entire time at Northeastern

University.

5
I want to specially thank Dung Nguyen and Son Nguyen from Massachusetts Institute of

Technology for their technical support in my research work.

I wish to express my indebtedness to my friends and colleagues, Dr. Anindita

Bhattacharya, Dr. Ting Qian, Dr. Rosa Ciprian, Dr. Florent Boico for their valuable

support to my work. My colleagues in the power electronics research group made my

stay in this group very fruitful and memorable. I really appreciate all the help and support

as well as value all the connections that I made working in this group.

I would like to whole-heartedly thank my family for their infinite generosity. My wife, Dr.

Hoa Nguyen, has been unfailing in her support. She has listened to, reflected on, and

supported me in every possible way for the past five years. This thesis would not have

been possible without her.

I would like to thank my daughter, Nhi Nguyen, who teaches me some Taekwondo

techniques.

Finally, I am grateful for the constant support and encouragement from my parents Dr.

Uy Nguyen, Ms. Ninh Nguyen. Their encouragement was my inspiration in these years,

as I was balancing between research, study and family. I am also deeply indebted to them

for their generosity in forgiving me every time I missed any family duty for my studies.

6
Table of Contents
List of Figures ..................................................................................................................... 9
List of Tables .................................................................................................................... 11
List of Acronyms .............................................................................................................. 12
1. Introduction ............................................................................................................... 13
1.1 Background ........................................................................................................ 13
1.1.1 Photovoltaic becomes competitive ............................................................. 13
1.1.2 Various applications of solar PV systems ................................................... 15
1.1.3 I-V and P-V characteristic of the solar arrays ............................................. 21
1.2 Modeling and simulation solar PV system under non-uniform shadow
conditions ...................................................................................................................... 25
1.3 Methods to improve the performance of solar PV array under partial shadow
conditions ...................................................................................................................... 29
1.4 Maximum power point tracking ......................................................................... 32
1.5 Conclusion/Summary of thesis contributions .................................................... 34
2. Modeling and Simulation Solar PV Array under Shadow Conditions ..................... 39
2.1 Introduction ........................................................................................................ 39
2.2 Simulation of solar PV array under a passing cloud .......................................... 43
2.3 Comparison between the proposed numerical model and linear model............. 57
2.4 Simulation solar array with and without bypass diodes ..................................... 58
2.4.1 With bypass diode ....................................................................................... 59
2.4.2 Without bypass diode .................................................................................. 59
2.5 Solar PV arrays of different configurations ....................................................... 60
2.6 Conclusions ........................................................................................................ 63
3. Solar PV Array Shadow Evaluation using Neural Network ..................................... 64
and On- Site Measurement................................................................................................ 64
3.1 Introduction ........................................................................................................ 64
3.2 Performance evaluation for solar PV systems.................................................... 65
3.2.1 Numerical method ....................................................................................... 70
3.2.2 Photogram metric method ........................................................................... 70
3.2.3 Neural network method............................................................................... 71
3.2.4 The combined method................................................................................. 72
3.3 Proposed method ................................................................................................ 74
3.3.1 Shadow evaluation using neural network without shading factor
measurement ............................................................................................................. 74
3.3.2 Experimental set up..................................................................................... 79
3.4 Neural network structure .................................................................................... 80
3.4.1 Neural network structure............................................................................. 80
3.4.2 Input signals correlation analysis ................................................................ 82
3.4.3 The learning process ................................................................................... 83
3.4.4 The generalization check ............................................................................ 85
3.4.5 The performance analysis of the network response .................................... 85
3.4.6 Analysis and discussion .............................................................................. 86
3.5 Conclusions ........................................................................................................ 87
4. A Adaptive Reconfiguration of Solar PV Arrays ..................................................... 89
4.1 Introduction ........................................................................................................ 89

7
 4.2 Shadow variation on photovoltaic collectors in a solar field ............................. 92
4.3 Conventional methods of reconfiguration solar PV array conventional methods
94
4.4 Proposed a adaptive reconfiguration method ................................................... 102
4.4.1 “Bubble - Sort” method ............................................................................ 106
4.4.2 Model-based method ................................................................................. 109
4.5 Conclusions ...................................................................................................... 116
5. A Small Scale Solar Photovoltaic Test Bed............................................................ 118
5.1 Introduction ...................................................................................................... 118
5.2 Bubble-Sort method ......................................................................................... 119
5.3 Model-Based method ....................................................................................... 121
5.4 Comparison between “Bubble - Sort” and “Model-based” methods ............... 122
5.5 Partly shaded solar adaptive bank .................................................................... 122
5.6 The number of shaded fixed solar cells is larger than the number of solar
adaptive bank .............................................................................................................. 123
5.7 Conclusion ........................................................................................................ 125
6. Single - Stage Two - Input DC-DC Converter for MPPT Solar PV Arrays ........... 126
6.1 Introduction ...................................................................................................... 126
6.2 Solar PV array’s model and the conventional MPPT based on Buck-converter
topology ...................................................................................................................... 128
6.2.1 Conventional MPPT based on Buck converter ......................................... 128
6.3 Proposed MPPT based on reconfiguration converter ....................................... 131
6.4 The relationship between the input current and voltage with the output current
and voltage .................................................................................................................. 134
6.5 Simulation result .............................................................................................. 136
6.6 Conclusion ........................................................................................................ 139
7. Conclusions ............................................................................................................. 140
7.1 General description .......................................................................................... 140
7.2 Future research: Design of proposed reconfiguration converter ...................... 146
7.2.1 Input capacitor design for proposed converter. The comparison with
conventional Buck converter .................................................................................. 147
7.2.2 Output inductor design:............................................................................. 148
7.2.3 Output capacitor design ............................................................................ 149
APPENDIX A MATLAB source code for modeling solar PV arrays under shadow
conditions ........................................................................................................................ 151
APPENDIX B MATLAB code for modeling solar PV array using neural network method
......................................................................................................................................... 158
APPENDIX C MATLAB code for control solar PV array reconfiguration ................... 162
References ....................................................................................................................... 166

8
List of Figures

Fig. 1.1 Global PV market growth .................................................................................... 14

Fig. 1.2 Photovoltaic becomes competitive ...................................................................... 15
Fig. 1.3 The small scale stand-alone solar photovoltaic systems ..................................... 16
Fig. 1.4 The photovoltaic system with battery storage ..................................................... 17
Fig. 1.5 The building integrated solar PV systems ........................................................... 19
Fig. 1.6 Solar PV power plant ........................................................................................... 20
Fig. 1.7 Experimental I-V and P-V curve of solar PV arrays with different irradiation
levels. ........................................................................................................................ 23
Fig. 1.8 Shadow configurations ........................................................................................ 24
Fig. 1.9 Photovoltaic systems with parallel strings and central inverter .......................... 29
Fig. 1.10 Photovoltaic systems with a) String inverters b) Module inverters.................. 30
Fig. 1.11 P&O for Maximum power point tracking ......................................................... 32
Fig. 2.1 Solar PV array from sub-modules, controlled switches ...................................... 45
Fig. 2.2 An equivalent circuit of a solar cell .................................................................... 45
Fig. 2.3 A solar cell with bypass diode ............................................................................. 48
Fig. 2.4 Solar PV array’s common interconnections ....................................................... 50
Fig. 2.5 The non-uniform and changed irradiation is the effect of a passing cloud ......... 51
Fig. 2.6 Output power of solar PV array simulation is based on two different methods .. 58
Fig. 2.7 The sub-module of solar PV array with bypass diode ......................................... 58
Fig. 2.8 Output power of solar PV arrays with and without bypass diode ....................... 59
Fig. 2.9 Output power of solar PV array depends on interconnection .............................. 61
Fig. 2.10 The losses of maximum power depend on the shaded area. In SP one cell in
each column is fully shaded. In TCT, one cell in the same row is fully shaded ....... 62
Fig. 3.1 Example of annually global (total) irradiance on a horizontal surface and ambient
temperature. .............................................................................................................. 66
Fig. 3.2 Annual average values for system parameters at 187 Japanese field test sites for
4/1995-12/2001[1] .................................................................................................... 69
Fig. 3.3 The correlation between the optimal power current and short-circuit current .... 71
Fig. 3.4 The correlation between the optimal power voltage and open-circuit voltage.... 72
Fig. 3.5 The sun’s position at specific time at day............................................................ 77
Fig. 3.6 The input and output signals measurement circuit .............................................. 79
Fig. 3.7 Neural network’s structure .................................................................................. 81
Fig. 3.8 The squared error of the training process ............................................................ 84
Fig. 3.9 The network performance analysis ...................................................................... 86
Fig. 3.10 The measured and predicted output powers of solar PV arrays ........................ 86
Fig. 4.1 Shading by collectors in solar field ..................................................................... 93
Fig. 4.2 I-V curves with different irradiance level............................................................ 97
Fig. 4.3 Parallel arrangement ........................................................................................... 97
Fig. 4.4 Series arrangement .............................................................................................. 98
Fig. 4.5 Series-parallel arrangement ................................................................................ 99
Fig. 4.6 I-V curves for PV with series-parallel arrangement of full sun ........................ 100
Fig. 4.7 Practical circuit of the proposed reconfiguration: For m rows in the fixed solar
array, we proposed m solar cells in the adaptive bank ........................................... 104
Fig. 4.8 The switching matrix ......................................................................................... 105

9
Fig. 4.9 The flow chart of the control algorithm for “Bubble - Sort” method ............... 107
Fig. 4.10 The flowchart of the control algorithm for Model-based method ................... 111
Fig. 4.11 Solar array’s reconfiguration under non uniform illumination ...................... 115
Fig. 5.1 Solar PV array outdoor test platform ................................................................. 118
Fig. 5.2 Output voltage of solar PV array before and after reconfiguration in Two –
Bubble Sort Experiments ........................................................................................ 120
Fig. 5.3 Output voltage of solar PV array before and after reconfiguration in Two Model-
Based Experiments.................................................................................................. 120
Fig. 5.4 Output voltage of solar PV array before and after reconfiguration when solar
adaptive bank is partly shaded ................................................................................ 124
Fig. 5.5 Output voltage of solar PV array when the number of shaded fixed solar cells is
larger than the number of solar adaptive bank ........................................................ 124
Fig. 6.1 Traditional solar PV array MPPT based on Buck converter topology .............. 128
Fig. 6.2 The I-V and P-V curves before and after reconfiguration ................................ 130
Fig. 6.3 Proposed MPPT based on Reconfiguration converter ....................................... 132
Fig. 6.4 Two solar PV submodules are connected in series when S is turned ON ......... 132
Fig. 6.5 Two solar PV submodules are connected in parallel when S is turned OFF..... 133
Fig.6.6 The voltage stresses of the main switches .......................................................... 134
Fig. 6.7 The input voltage and output voltage relationship ............................................ 135
Fig. 6. 8 The voltage stress of transistor in a) buck converter and b) Reconfiguration
converter ................................................................................................................. 137
Fig. 6.9 The voltage stress of main diodes in a) Buck converter and b) Reconfiguration
converter ................................................................................................................. 137
Fig. 6. 10 The inductor current ripple in a) Buck converter and b) Reconfiguration
converter ................................................................................................................. 138
Fig. 6.11The input capacitor voltage ripple a) Buck converter with one C=100uF, Vin b)
Reconfiguration converter with two C=100uF, Vin/2 ............................................ 138

10
List of Tables

TABLE 2.1 The parameters of solar PV array for simulation .......................................... 57

TABLE 3.1 The correlation between the inputs signals ................................................... 83
TABLE 4.1 Comparison of the Reconfiguration Methods ............................................. 114
(m=rows, n=columns in fixed array) .............................................................................. 114
TABLE 6.1 Simulation parameters of conventional and proposed configuration.......... 136

11
List of Acronyms

Acronym Definition

IM output current of solar module

IC output current of solar cell
VM output voltage of solar module
VC output voltage of solar cell
ISCM short-circuit current of solar module
ISC short-circuit current of solar cell
RSM series resistance of solar module
RS series resistance of solar cell
RSHM shunt resistance of solar module
RSH shunt resistance of solar cell
a, a1 ideality factor of bypass diode
k Boltzman’s constant.
T cell operating temperature
q electron charge
s number of solar cells per column of a module
p number of solar cells per row of a module
I0 saturation current of the diode
IOBP saturation current of the bypass diode
Ai,j vertical element (i, j running position index)
Bi,j horizontal element (i, j running position index)
IAi,j current at element Ai,j
IBi,j current at element Bi,j
V total output voltage
I total output current
R load resistance
VAi,j voltage at element Ai,j
VBi,j voltage at element Bi,j
m number of vertical elements of a column
n number of horizontal elements of a row
RON resistance of turned ON switches
ROFF resistance of turned OFF switches
tp t1…t10 instant of time

12
1. Introduction

1.1 Background

1.1.1 Photovoltaic becomes competitive

The recent increase of energy demand from Asian countries as well as the traditional high

energy demand from industrial countries causes many undesirable effects:

• First, oil prices have doubled in the last five years, increasing from 44 USD/barrel

in 2003 to over 100USD/barrel in the beginning of 2008. In addition, there is a

forecast that fossil fuel will still be available in 2030, but there is a geographical

imbalance between the energy supply and demand that will eventually lead to the

exhaustion of fossil fuel [1].

• Second, greenhouse effects and environmental threats are arising from the

production of electricity from fossil fuels. The climate changes can cause floods

or dissertations over the world. With more lightly effects, the long dry periods

will lower water levels in hydro-electric facilities and may cause instability of the

main electricity supply source in many countries.

One solution to these problems is obvious, mass production and use of electricity

generated from solar energy, which is environmentally friendly. Benefits of solar energy

include clean energy production and inexhaustible supply of daylight. The solar

photovoltaic (PV) systems can supply energy without moving parts, operate noiselessly

and have minimum maintenance cost. A disadvantage is the high manufacturing cost

13
because the semiconductor materials used to convert the solar light into electrical energy

are scarce.

Fig. 1.1 Global PV market growth

That being said, the cost trend of photovoltaic is downwards, while fossil fuel price are

rising. The cost of PV modules and associated system components depend on developing

mass production techniques and facilities. A growing and stable market of PV system is

developing. Fig. 1.1 shows the previous data and prediction data of the global market of

PV from 1990 to 2013. The total Mega Watt Power of global solar PV market is doubled

for the period from 2005 to 2008 [3].

The study [3] indicates that the cost of electricity produced by PV could match the retail

price the domestic consumers pay for electricity by 2020, when the price of the consumer

electricity in Germany rise in average 2% per year. In sunnier countries like Spain, the

14
solar PV systems work more efficiently, the price of PV electricity in Spain could be

match the retail price in 2010, as shown in Fig. 1.2.

Fig. 1.2 Photovoltaic becomes competitive

1.1.2 Various applications of solar PV systems

A solar cell generates the DC current from sunlight. A solar array is created by

connecting individual solar cells together. The output current of a solar array depends on

solar insolation, the ambient temperature, the size and the configuration of the solar array.

In general, the larger area solar panels will produce more energy, and smaller solar panels

produce less.

The solar PV system can be used to supply DC power to load or can be connected

through DC-AC inverters to supply AC power to loads.

15
Here are some examples of Photovoltaic applications:

Stand-Alone Systems

There are widespread applications using small - scale solar arrays, such as data solar

disks, calculators, solar powered LED lights, to solar powered speedometers in the

highways, etc, see Fig. 1.3. These systems usually have small power range: from few

microwatts up to few hundred watts. The output voltages are in the range of few volts to

tents of volts. Typical solar panels weights only 3.66 mg per square millimeter allowing

mounting on a near portable electronics or consumer products without exceeding size of

weights limitations [2].

Fig. 1.3 The small scale stand-alone solar photovoltaic systems

16
Battery Storage

The output power of solar PV systems depends on solar insolation and varies with

daytime. In the night, solar PV systems do not generate energy [3]. To stabilize the output

power, sometimes, the solar PV systems are connected to some type of battery. It allows

the solar charger to be in the sun. The energy from the sun is collected by the solar panel

and converted into electricity and sent to the battery charger. The electricity is produced

and then used to charge a battery. This way equipment can be powered later by battery

indoors, on a rainy day, or at night.

a) Small solar photovoltaic cell with battery storage

b) Water pump system using large photovoltaic with battery storage

Fig. 1.4 The photovoltaic system with battery storage

17
Fig. 1.4 (a) shows that some solar battery chargers work with simple AA batteries, like

those that used to run a flashlight. Some work with special batteries, which can in turn be

used to charge other equipment like cell phones, solar calculators, speedometers in the

highways, LED, etc. Some charge the batteries within the electronics equipment, such as

the battery in a laptop computer.

In the rural, high mountainous and forest areas where the cost of building and distributing

conventional sources of energy is very high, solar photovoltaic systems are used to

charge large battery systems to produce power in the order of kilowatts. Fig. 1.4( b)

shows the water pump in a rural area is powered by a large photovoltaic system using

large battery storage. Thus, the type of solar battery charger that a person would use

depends in a large degree upon what the solar charger will be used for.

Building Integrated Photovoltaic (BIPV) system

The battery storage will increase the total cost of the solar PV system. One popular

solution is to connect the solar PV system directly to the utility grid. The equipment is

powered by the grid whenever there is insufficient solar insolation and the photovoltaic

complements the grid in “high-demand electricity” hours, reducing the need for

increasing the main power plant size to meet the load demand.

Sometimes conventional building materials such as the roof, skylights, or facades are

replaced by photovoltaic materials in parts of the building envelope. In particular, they

are increasingly being integrated in the construction of new buildings as a principal or

18
ancillary source of electrical power. Other existing buildings may also be retrofitted with

BIPV modules. By using integrated photovoltaic systems, the initial cost can be reduced

by saving the building materials and labor that would normally be used to construct the

part of the building that the BIPV systems modules replace. In addition, since BIPV

systems are an integral part of the design, they generally blend in better and are more

aesthetically appealing than other solar options. These advantages make BIPV one of the

fastest growing segments of the photovoltaic industry [3].

Fig. 1.5 The building integrated solar PV systems

Utility Power Production

Large-scale photovoltaic power plants can generate a wide range of powers up to few

megawatts that can supply enough electricity for some thousand houses, as shown in Fig.

1.6.

19
 Fig. 1.6 Solar PV power plant

Actually, solar arrays can be connected into standard modules, which consist of many PV

arrays working together. Then these modules can be connected in parallel to generate

required power. Because PV arrays are fairly easy to install and connect, there are many

advantages of solar PV to utilities in comparison with other kind of power plants. Firstly,

the solar PV plants can be built more quickly than conventional fossil or nuclear power

plants as demand of electricity increase over planning. Because of unequal load demand

of distributed electrical system, the PV power plants can be built where they are most

needed in the grid. PV power plants consume no fuel and produce no air or water

pollution while they silently generate electricity. So, they can be located in or very close

to the population while meeting local environmental regulations

20
And, unlike conventional power plants, modular PV plants can be expanded

incrementally as demand increases. This makes PV power an attractive option for utilities

that want or need to cut fuel costs [6].

1.1.3 I-V and P-V characteristic of the solar arrays

Fig. 1.7 shows the current versus voltage and power versus voltage for two different solar

PV arrays: Global P3-30W and UNI-solar 15W taken in under indoor laboratory

conditions. The artificial light source is the four 500W halogen lamps. From the above I-

V and P-V curves, we can notice that:

• The performance of a solar PV array strongly depends on the irradiation levels.

• While the performance specifications of solar cell modules are available, the data

provided for these arrays do not accurately reflect the amount of electric power

that may be generated from the solar panel at time of use, i.e. these experimental

results differ significantly than the characteristics provided by the manufacturer.

• I-V and P-V curves have strong non-linear characteristics.

• The output power and I-V curve of a solar panel vary significantly (in shape and

power) depending on “the manufacturer”.

Current versus voltage characteristics of the solar cells are non-linear, thus leading to

technical control challenges. In the first order approximation, output power of a solar

array is proportional to the irradiance of sunlight.

21
 The V-I Curves of Global P3-30

1.8
1.6 Series1
1.4
Series2
1.2
1 Series3
(A)

0.8 Series4
0.6
Series5
0.4
0.2 Series6
0
0 5 10 15 20 25 30
(V)

a) V-I curve for Global P3-30 solar array

The V-I Curves of Unisolar 15W

0.35
0.3
0.25
0.2 Series1
(A)

0.15 Series2
0.1
0.05
0
0 5 10 15 20 25
(V)

b) V-I curve for Unisolar15W solar array

22
 The P-V Curves of Global P3-30

14
12 Series1
10 Series2
8 Series3
(W)

6 Series4
4 Series5
2 Series6
0
0 5 10 15 20 25 30
P-V curve for Global
(V) P3-30 solar array

c) P-V curve for Global P3-30 solar array

The P-V Curves of Unisolar 15W

4
3.5
3
2.5
Series1
(W)

2
1.5 Series2
1
0.5
0
0 5 10 15 20 25
(V)

d) P-V curve for Unisolar-15W solar array

Fig. 1.7 Experimental I-V and P-V curve of solar PV arrays with different irradiation levels.

23
However, in many applications, such as solar power plants, building integrated

photovoltaic or solar tents, the solar photovoltaic arrays might be illuminated non-

uniformly. The cause of non-uniform illumination may be the shadow of clouds, the trees,

booms, neighbor’s houses, or the shadow of one solar array on the other, etc. This further

leads to nonlinearities in characteristics.

a) Shadow on portable solar array b) Shadow from one array cast on another array

Fig. 1.8 Shadow configurations

For example, Fig. 1.8(a) shows a portable, flexible solar array that is embedded into

fabric. These new generation solar arrays can be folded and carried by campers and

soldiers to remote locations. Often, they are left alone to charge batteries near trees,

fences, and have been reported to even be wrapped around telephone poles or trees [7].

Furthermore, for applications like BIPV and solar power plants in small spaces, as shown

in Fig. 1.8(b), shadows from one solar PV array are commonly cast on another array. In

conventional solar PV plants, the designer must solve the complex problem, the tradeoff

between “maximum energy output” and “minimum produced energy cost” by varying the

distance between the rows [8], [9].

24
For these new applications, it has been especially important to optimize performance of

the arrays in shadowed conditions. Because of the nature of the electrical characteristics

of solar cells, the maximum power losses are not proportional to the shadow, but magnify

nonlinearly [10].

The shadow of solar PV array can cause many undesired effects:

• The real power generated from the solar PV array is much less than designed, so

that the loss of load probability increases [11].

• The local hot spot in the shaded part of the solar PV array can damage the solar

cells. The shaded solar cells may be work on the negative voltage region and

become a resistive load and absorb power. Bypass diodes are sometimes

connected parallel to solar cells to protect them from damage. However, in most

cases, just one diode is connected in parallel to a group of solar cells [12] and this

hides the potential power output of the array.

1.2 Modeling and simulation solar PV system under non-

uniform shadow conditions

There are also other undesired effects of the shadows on solar PV arrays other than power

drop. Dynamic modeling of the effect of clouds on solar array is a difficult task. Cloud

conditions can dramatically change fast and the cost of such fluctuations with their effect

on other systems is important to understand.

25
One method proposed by [17] measures the changes in solar insolation over a one minute

time interval. With this method, solar insolation values may be measured in the

horizontal plane and subsequently used to calculate insolation levels for any desired

angle [18].

There have also been attempts to predict the solar PV array output power in clouds. An

estimation method used in [18] proposes that the power output of a PV system is

proportional to the insolation levels measured for the surface of a solar cell at any angular

position. However, the output power of a solar array not only depends on changes in

irradiance values, but also depends on other factors such as the interconnection of the

individual solar cells, solar cells parameters, and the use of bypass diodes [19], [20].

Thus this modeling approach is not complete, and one goal of this thesis is to make

improvement on it.

A shadowed solar cell acts like a load because it dissipates input current. In the presence

of shadows, where there is no exposure to sunlight, a solar cell will heat up and develop a

hot spot. To reduce the overall effect of shadows, bypass diodes are sometimes connected

across shadowed cells to pass the full amount of current while preventing damage to the

solar cell [12]. Thus, predicting the electrical characteristics of a solar PV array when

experiencing passing clouds is more complex.

26
Mathematical models [19], [20] have been used to accurately characterize the electrical

solar PV arrays. In [21] the effect of non-uniform illumination is considered on various

configurations of solar PV arrays without a bypass diode.

The first objective of this project is to propose a modeling and computing algorithm that

greatly expands on the ideas in [19], which simulates and analyzes the effect of non-

uniform changing shadows (i.e. a passing cloud) on the output power of solar PV arrays.

Our model will be able to determine the power losses in each solar cell and the collective

hot spots of an array.

In order to include the effects of shadowing on solar PV arrays, conventional methods

use an approach to determine the “shading factor,” which is defined as the ratio of the

non-shaded area to the total area of the solar arrays.

In real operating conditions, solar PV arrays are connected with Maximum Power Point

Trackers (MPPT) to track maximum output power. The MPP is often assumed linearly

proportional to solar irradiance. So shading factors also can be used to calculate MPP

power of shaded solar PV arrays. In reality, it is difficult to accurately estimate this

shading factor because shadows change their shape and move with time.

This leads to several approaches to evaluate the shadow effects on those solar PV arrays:

• In the numerical method, the solar irradiance on all solar cells is modeled based

on real time input data. The output power of the solar PV array at a given output

27
 voltage is then received by solving differential equations on small time-scales

(second, minutes) [19], [20]. This method requires unrealistic number of sensors

to measure the irradiance in each solar cell. Furthermore, accurate prediction of

the maximum output power of solar PV systems for long term (days, months,

years), is more challenging, due to difficulties in modeling the shaded area in the

solar PV systems.

• In the photogram metric method, the position of obstacles and their shadow are

estimated by using the triangulation with two photographs or more [21].

• Neural network method is used in [22], [23], [25] for predicting the dependence

of maximum output power of solar PV arrays on environmental factors. However,

these papers focus on estimating the performance of solar PV arrays under

uniform illumination.

• In the combined method, the neural network prediction model uses training data

created by the numerical method. The low accuracy is due to the low accuracy of

the modeling shadow objects [26].

From the above discussion, there is not a simple and accurate method to define the

shadow effects on shaded solar PV array for long term (seasonal or annual) in non-

uniform illumination.

This thesis proposes a novel method to define the function of relationship between the

maximum output power (MPP) of the shaded solar PV array and the environmental

factors, such as the solar irradiation levels, the sun’s position angles and the ambient

28
temperature. This function, which is similar to the shading factor, can fully characterize

the shadow effects on the solar PV array for long term and is called the shading function.

As such, the shading function can eliminate the inaccuracy caused by the complexity of

the shading factor’s calculation.

1.3 Methods to improve the performance of solar PV array

under partial shadow conditions

Fig. 1.9 Photovoltaic systems with parallel strings and central inverter

29
 a) String inverters

b) Module inverters

Fig. 1.10 Photovoltaic systems with a) String inverters b) Module inverters

30
Fig. 1.9 shows a photovoltaic system with parallel strings and central inverter.

Solar cells PV1, PV2,…, PVn are connected in parallel and/or series to create the solar

submodule and/or strings. Photovoltaic inverters convert the dc voltage from solar arrays

to ac voltage of a grid. The maximum power point (MPP) voltage of the solar array is

different from the grid voltage, so the DC-DC converter (MPPT) usually is integrated in

photovoltaic inverters [42].

For partly shaded solar arrays, or modules with different power outputs, the system with

central inverter may have high losses, the parallel connection of string inverters or

module inverters can reduce the losses [31], [32], [33]. Fig. 1.14 shows photovoltaic

systems with two different classifications of inverters.

An emerging research field, and the part of this research thesis, is to adaptively

reconfigure solar array connections in real time in order to track maximum output power.

Traditionally fixed solar PV arrays have hard wired interconnections between their solar

cells. These connections are not changed after installation. However, it is possible to

continuously rearrange solar cells in series and parallel connections [34] to facilitate the

photovoltaic system to work more as a constant power source even in different operating

conditions, i.e. insolation, temperature, loads,…etc. Recent research studies in [35], [36]

have started to develop methods to reconfigure solar cells to improve power output in

shaded conditions.

31
The research of [35][36] focuses mostly on how to build the arrays, and does not propose

real-time implement able control algorithm. Because of this, the methods proposed have

unrealistic number of switches and sensors that must use complex control algorithms to

determine when it turns the switch ON or OFF. Thus, there is a real need for more

research to develop reconfigurable solar array that can actually be built.

1.4 Maximum power point tracking

The general idea of MPPT is that a solar panel behaves similarly to a nonlinear current

source that has maximum power at a specific output voltage for each environmental

condition (temperature, illumination, etc.) [37]. By inserting a DC-DC converter between

the solar array and the load, it is possible to adjust the duty ratio of the converter to match

the impedance of the array to the load, producing the maximum output power of the array.

Fig. 1.11 P&O for Maximum power point tracking

Several methods are utilized to achieve peak tracking of the solar PV array’s power [29]:

32
 • In the Perturb & Observe method, the output power is measured and the duty ratio

in a converter is changed.The duty ratio decreases when the output power of the

array decreases. If the power increases, the direction is reversed. When the output

voltage of the solar array changes around its MPP point, the duty ratio then

fluctuates between positive increments and negative increments.

• There is another method that has a similar effect called Incremental Conductance.

Sometimes the MPP point is estimated based on a solar panel’s open circuit

voltage and/or short circuit current. However, this method is dependent on the PV

array’s technology.

• In the conventional methods, the main assumption is that the solar PV system is

uniformly illuminated. The remaining task of this research is to develop the

topology that is able to track more power from non-uniformly illuminated solar

PV systems. All these topologies can be derived from the basic canonical

switching cell.

The large difference between the input and output current or voltage in each period/cycle

will require the large energy storage elements (L and C). The size of the DC-DC

converters strongly influenced the size of the energy storage elements.

In the above topologies, the buck converter has the smallest voltage stress and energy

storage elements.

33
This thesis proposes a new topology based on Buck topology that is able to reduce the

voltage stress of the main switches to half of the input voltage.

1.5 Conclusion/Summary of thesis contributions

This proposed research will focus on the modeling, development, and design of an

adaptable solar array and system that is able to optimize power output, reconfigure itself

when solar cells are damaged and create controllable output voltages and currents

This study proposes a technological advancement over the existing technology of solar

PV arrays. Presently solar arrays are fixed arrays that require external device to control

their output. In this research, the solar array will be able to self-reconfigure, leading to the

following advantages:

• Higher efficiency because no external devices are used.

• Maximum possible output power that is much higher than the maximum power of

fixed solar arrays by arranging the solar cells in optimized connections.

• Elimination of hot spot effects.

The proposed research has the following goals: First, to create a modeling and computing

algorithm, which is able to simulate and analyze the effects of non-uniform changing

shadows on the output power of solar PV arrays. Our model will be able to determine the

power losses in each solar cell and the collective hot spots of an array. Second, to propose

new methods, this will predict the performance of solar PV arrays under shadow

conditions for long periods of time (days, months, years). Finally, the thesis develops

34
adaptive reconfiguration algorithms to reconfigure connections within solar PV arrays in

real time, under shadow conditions, in order to optimize output power. Alternatively, the

reconfiguration algorithms can be used to improve performance when cells are damaged.

The objectives are described in more detail in the followings

Objective 1

Create a modeling and computing algorithm capable of accurately predicting the effects

of non-uniform changing shadows, e.g., i.e. a passing cloud or a tree’s shadow on the

output power of solar PV arrays.

Chapter 2 develops a model that is able to determine the power losses in each solar cell

and the collective hot spots of an array. The proposed procedure of modeling and

simulation of solar PV array is the following:

First, when an accurate model is created for each solar cell, the proposed method will be

able to define the parameters of the model for solar cells. Then, the model for horizontal

elements in the network (the connection elements between solar cells) is created and the

parameters of the horizontal elements are defined. Second, a system of equations between

loops and nodes of the array is developed by using Kirchoff’s laws. The system of

equations is solved by the Newton-Raphson method. Last, the proposed model is used to

simulate the electrical behavior of the solar PV array under certain environmental factors:

solar irradiance, temperature, and loads. The proposed model is applied to check the

shadow effects on the solar PV arrays with different configurations. The accuracy of the

proposed model with other conventional models is compared.

35
Objective 2

Propose methods using neural networks that are able to predict the shadow effects on the

solar PV arrays with low computational efforts.

Chapter 3 develops a neural network method, combining the multilayer perceptrons feed-

forward network with a back-propagation algorithm, to accurately predict maximum

output power of solar photovoltaic arrays under shadow conditions. Using the solar

irradiation levels, the solar photovoltaic array’s temperature, and the sun’s position

angles as the input signals, and the maximum output power of the solar PV array as an

output signal, the training data for the neural network is collected by one-day on site

measurement. After training, the neural network model’s accuracy and generalization will

be verified by using test data. The goal is to have the model to predict the shadow effects

on solar PV arrays for long periods of time (days, months, years) with low computational

efforts. The proposed method is able to define the main factors that influence the output

power of the solar PV array under shadow conditions. The simple measurement

procedure for each factor is described.

The appropriate neural network structure for this data is chosen. The random procedure to

obtain the training data set and the test data set is described. The function of the

relationship between the inputs and outputs from the training data set is derived. Last, the

accuracy of the model is tested with the test data set. Thus, the function can be used to

predict the output power of a solar PV array with the new input data set.

36
Objective 3

Develop adaptive reconfiguration algorithms to reconfigure connections within solar PV

arrays in real time, under shadow conditions, in order to optimize output power.

Chapter 4 proposes to connect matrix switches between a “solar adaptive bank” and a

standard TCT fixed solar PV array. Various control algorithm approaches are derived so

the solar PV array reconfigures to produce optimal power. Control algorithms are

implemented in real-time. Simple “bubble- sort” as well as model based control

algorithm will be investigated. The method to measure the shadow patterns in the solar

PV array with the smallest number of voltage sensor is described. The method also

proposes the configuration that is most suitable to compensate the shaded cells by

reconfiguration. Control algorithms for reconfiguration solar array first time ( as our

knowledge) are proposed.

Objective 4

Validate all new models and reconfiguration algorithms in a small scale solar PV

prototype system.

Chapter 5 propose the procedure to build an experimental test bed with 10- 15 solar cells

and their necessary switch connections needed to test the reconfiguration algorithms in

Objective 3. The tests with different shadow patterns and with each reconfigurable

control algorithms derived. The advantages and disadvantages of proposed

reconfiguration methods are derived.

37
Objective 5

Develop an approach for solar array operation that merges features of external power

electronics devices to be within the solar array itself. Next generation topology is

proposed to continuously rearrange the two identical solar submodules in series and

parallel connections by using high frequency switches to facilitate the photovoltaic

system to work more as a constant power source in even in different operating

conditions (i.e. irradiance, temperature, loads,…etc). The output of the solar PV array

will be connected to the load through a high frequency filter without MPPT.

Chapter 6 develops a class of DC-DC converter where the voltage stresses of the main

switches will be reduced to a half of the input voltage, when the input sources are solar

arrays. The difference between input and output voltage or current for each cycle will be

reduced so that the size of storage elements (L and C) will be smaller in comparison with

conventional topologies. A simulation reconfiguration PV system with the battery load is

presented and is shown to verify the proposed topology. Chapter 6 will describe the

proposed scheme in detail.

38
2. Modeling and Simulation Solar PV Array under Shadow
Conditions

2.1 Introduction

As introduced in the Abstract, this thesis focuses on the research of two dominant fields

of solar PV system: (1) modeling and simulation solar PV array under shadow conditions

and (2) reconfiguration solar PV arrays to improve their performance under shadow

conditions. This Chapter mainly discusses the modeling and simulation of solar PV

systems under changing shadow conditions, a passing cloud.

Modeling and simulation solar PV arrays are important tasks in the design of solar PV

system. The designs of solar PV system need to meet the demands with the consideration

of applicable restrictions. The most important demand is the energy used per year, day

and hour of the user. The restrictions may include the land area, the solar irradiation

availability, the ambient temperature, the availability of the solar PV products in the

market… etc.

The design of a solar PV system includes understanding the environmental effects on

performance, and in particular the output power of the array under various shading

conditions.

The solar cell is the semiconductor device that directly converts the light energy to the

electrical energy. A solar cell has nearly the same behavior as a diode. Specifically, the

39
output power of a solar array strongly depends on the irradiance level of sunlight and

ambient temperature. The most conventional model of a solar cell is the one diode

model [37].

Because of low output current and output voltage of solar PV cell, in practice, solar cells

are connected in series and/or in parallel to create a solar PV array that is able to meet the

output voltage and/or output current requirement. In uniform illumination, the output

power of the array is equal the total output power of all solar cells. It can be simply

calculated by the multiplying the output power of one solar cell to the total number of

solar cells in the array.

But, in some applications, such as solar power plants, building integrated photovoltaic

systems (BIPV), and portable solar tents, it is common for the solar PV to become

illuminated non-uniformly. The cause of non-uniform illumination may be shadows from:

trees, booms, a neighbor’s houses, or even a shadow of one solar array to the other one.

The real power generated from the solar PV array may become significantly less than

designed. At times, this may lead to a complete “loss of load”. A shadowed solar cell

acts like a load because it dissipates input current. In the presence of shadows, where

there is no exposure to sunlight, a solar cell will heat up and develop a hot spot. The local

hot spot in the shaded parts of the solar PV array can damage the solar cells, and there

can be a long-term decrease in annual system performance due to shading. To reduce the

overall effect of shadows, bypass diodes are connected across shadowed cells to pass the

40
full amount of current while preventing damage to the solar cell. Thus, predicting the

electrical characteristics of a solar PV array under non-uniform shadow conditions is

more complex. It is vital to understand this prior to installation in order to effectively

determine whether the system is sufficiently cost efficient enough to install [38].

There are also other undesired effects of the shadows on solar PV arrays other than power

drop. Dynamic modeling of the effect of clouds on solar arrays is a difficult task. Cloud

conditions can change fast dramatically and the cost of such fluctuations with their effect

on other systems is important to understand.

One method proposed by [17], [39] measures the changes in solar insolation over a one

minute time interval. With this method, solar insolation values may be measured in the

horizontal plane and subsequently used to calculate insolation levels for any desired

angle [18].

There have also been attempts to predict the solar PV array output power in clouds. An

estimation method used in [18] proposes that the power output of a PV system is

proportional to the insolation levels measured for the surface of a solar cell at any angular

position. However, the output power of a solar array not only depends on changes in

irradiance values, but also depends on other factors such as the interconnection of the

individual solar cells, solar cells parameters, and the use of bypass diodes [19], [20]. Thus

this modeling approach is not complete, and one goal of this research is to make

improvement on it.

41
Mathematical models [19], [20] have been used to accurately characterize the electrical

solar PV arrays. In [20] the effect of non-uniform illumination is considered on various

configurations of solar PV arrays without a bypass diode.

This research propose a modeling and computing algorithm that greatly expands on the

ideas in [19], which simulates and analyzes the effect of non-uniform changing shadows

(i.e. a passing cloud) on the output power of solar PV arrays. The model will be able to

determine the power losses in each solar cell and the collective hot spots of an array.

Specifically, our model has the following advantages:

• Creates an accurate model for each solar cell: The users can define the parameters

of the model for solar cells for different type of solar cells.

• The model is applicable for any configurations of solar PV arrays: By modifying

models for horizontal elements in the network (the connection elements between

solar cells), define parameters of the horizontal elements: the wires and/or the

switches.

• After inputting known parameters a system of equations between loops and nodes

of the array, using the Kirchoff’s laws is created automatically: The Newton-

Raphson method is applied for solving the system of equations. So, the simulation

converges fast and accurately.

42
The proposed models can be used to simulate the electrical behavior of the solar PV array

for different changing environmental factors: solar irradiance, temperature and load.

The proposed model also is able to check the shadow effects on the solar PV arrays with

different configurations. This model shows the higher accuracy in comparison with other

conventional models.

2.2 Simulation of solar PV array under a passing cloud

From discussion in the Section 2.1., the modeling and simulation of solar PV systems are

simple task under uniform irradiance level.

In non-uniform illumination conditions, when the solar cells are illuminated differently,

the solar array is modeled as a network in Fig. 2.4 from vertical elements (solar cells) and

horizontal elements (connectors, diodes).

The proposed model has the following advantages:

• Able to simulate all common types of solar PV cells and modules in the market,

by inputting the solar cells parameters from manufactures or by curve fitting.

• Able to simulate solar PV system consisting from different types of solar cells or

solar submodules.

• Able to simulate solar PV systems under non-uniform illumination with different

shadow’s shapes or shadow’s moving direction.

43
 • Converge fast enough.

Detailed analysis for modeling and simulation solar PV array under shadow conditions is

provided in the below description. The following procedure is proposed for simulating a

solar PV array under a passing cloud in MATLAB.

Step 1

The solar PV array is modeled as a matrix with dimension (m x n): m represents the

number of solar cells connected in a series string, while n represents the number of

strings connected in parallel. Fig 2.4 shows the two common solar PV array

configurations that utilize the combinations of the connections. In this figure, each PV(i, j)

is an individual solar cell. By inputting different solar cell’s parameters, we can model

the solar arrays from the same kind of solar cells or the solar arrays from different kinds

of solar cells.

The solar cells are connected in series and parallel to create a solar array. Fig. 2.1 shows a

solar array: all the vertical components are solar cells or modules and all horizontal

components are controlled switches.

By turn ON and turn OFF the horizontal switches B (i,j) we have different common

configuration of solar PV array: the series-parallel configuration and total-cross-tied

configuration.

44
 Iout
+
PV(1,2) PV(1,n)
PV(1,1) PV(1,n-1)
B(1,1) B(1,n-1)

PV(2,2) PV(2,n)

PV(2,1) PV(2,n-1)
B(2,1) B(2,n-1)

B(m-2,1) B(m-2,n-1)

PV(m-1,n)
PV(m-1,2)
PV(m-1,1)
Vout
B(m-1,1) PV(m-1,n-1) B(m-1,n-1)

PV(m,n)

PV(m,1) PV(m,2) PV(m,n-1)

Fig. 2.1 Solar PV array from sub-modules, controlled switches

Modeling solar cells, the vertical elements of solar array

The classical one diode model of a solar cell is shown in Fig. 2.2

Rs VC

+
+

Isc Rsh Dsolar RLoad

Fig. 2.2 An equivalent circuit of a solar cell

The current source generates the photocurrent ISC, which is linearly proportional to the

solar irradiance level.

45
The current goes through the solar cell diode varies with the junction voltage and the

solar cell reverse-saturation current I0. The series resistance of solar cell RS – characterize

the voltage drop when the charge carries through the semiconductor junction to the

external load. Series resistance of solar cell is small. The shunt resistance of solar cell

RSH is large. The shunt resistance describes the leakage currents at the cell edges.

The relationship between the output current and output voltage of the solar cell is given

by:
V + IRS
VT V + IRS
I = I PH − I 0 (e − 1) − (2. 1)
RSH

Here, IPH- Photo generated current proportional to illumination.

The threshold voltage of solar cell is defined by:

AkT
VT = (2. 2)
q

Here

• A is ideality factor;

• K is Boltzman constant;

• q is electron charge;

• T is the solar cell temperature (K degree).

To simulate the characteristic (I-V curve) of a solar cell, the parameters RS, RSH, IO and A,

have to be known. By changing environmental factors: solar irradiation levels and solar

cell temperature, we can have the curves of different weather conditions.

46
Solar cell parameters approximation

For the simulation of a real solar cell, the parameters of solar cell or solar submodule are

received by I-V curve fitting approximation from manufacturer’s datasheet. This I-V

curve is received in standard test condition (STC) and the parameters of a solar cell will

be changed in the real working conditions.

For more accurate approximation, [27] proposes method to extract the solar cell’s

parameters based on experiment over wide range of operating conditions. The I-V curve

of a solar PV cell is received by 300 measurements of the output current and output

voltage over a wide range of solar irradiance level, ambient temperature and varied load.

After that, the parameters of solar cell and photo generated current in Eq. (2.1) are

transformed to the coefficients a, b, c, d, f, according the following transformations:

q qR
I PH + I S = a , I S = −b , = c , S = d , RSH = f
AkT AkT

This leads to the following simpler equation:

I = a + b × exp(cU + dI ) − (U + dI / c) / f (2.2)

The coefficients a, b, c, d and f are received by curve fitting the I-V curve received from

the described experiment. The physical knowing the parameters quantities IPH, IS, n, RS,

and RSH are found:

47
 q d
= A R =
I PH = a + b , I S = −b , ckT ,
S
c , RSH = f

Solar cell with bypass diode model

Rs VC
+ +

+
Isc Rsh Dsolar Dbypass RLoad

Fig. 2.3 A solar cell with bypass diode

Fig. 2.3 shows a solar cell with bypass diode. A shadowed solar cell acts like a load

because it dissipates input current. In the presence of shadows, where there is no

exposure to sunlight, a solar cell will heat up and develop a hot spot. To reduce the

overall effect of shadows, (by creating low reverse breakdown voltage for a solar cell),

bypass diodes are sometimes connected across shadowed cells to pass the full amount of

current while preventing damage to the solar cell.

The relationship between the output current and output voltage of the solar cell with

bypass diode is given by:

V + I RS ⎛ −V ⎞
VT V + IRS ⎜ ⎟
I = I PH − I 0 (e − 1) − − I bypass ⎜ e VT − 1⎟ (2. 3)
RSH ⎜ ⎟
⎝ ⎠

48
where, I0 is the reverse- saturation current of the solar cell, IBYPASS- the saturation

current of bypass diode.

Modeling horizontal elements, the wires or switches

In the above equation, RBi , j = RON if the relays are turned ON, R Bi , j = ROFF = 10 20 Ω if the

relays are turned OFF.

Different configurations of solar PV arrays and the maximum power losses of the shadow

Fig. 2.4 shows the two common solar PV array configurations that utilize the

combinations of the connections. In this figure, each PV(i, j) is an individual solar cell.

The solar cells are connected in series and parallel to create a solar array.

Simple series-parallel (SP) array: if all switches are OFF

When all solar cells, for example, PV1,1 , PV2,1, …, PVm-1,1, PVm,1, are connected in series

creating strings. Then all these strings are connected in parallel, as shown in Fig. 2.4 (a).

Total-cross-tied (TCT) array: if all switches are ON

When all solar cells are connected in parallel, for example PV1,1 , PV1,,2, …, PV1,n-1,

PV1,n creating modules. Then these modules are connected in series, as shown in Fig.

2.4 (b) [15].

49
 a) Series-parallel interconnection

b) Total-cross-tied interconnection

Fig. 2.4 Solar PV array’s common interconnections

50
 Fig. 2.5 The non-uniform and changed irradiation is the effect of a passing cloud

Step 2

Model a passing cloud

The shadow of a passing cloud (and/or a tree’s shadow, boom’s shadow) will reduce

current, resulting in non-uniform irradiance to solar PV array. The change of irradiance

will change the short-circuit currents and open-circuit voltages of each cell. In this case

we assume that the parameters and temperature of the solar cells are unchanged. The user

specifies the ratio of the irradiance in each solar cell at each instant of time (t1…t10)

according to irradiance standards (STC).

51
To simulate a moving cloud, we must first determine the distance, Di , j , between the solar

cell with index i, j at instant of time t p and the center of a cloud. It can be defined by the

following equation:

Di , j = (i − t P * v )2 + ( j − t P * v )2 (2. 4)

Realizable values for the ratio of the irradiance in each solar cell will range from 0 to 1,

and can be specified using the nominal distribution function in the following equation:

(
− Di, j − μ )2
2
I SCi e 2σ
, j
= 1 − 4 * (2. 5)
I SC 0 σ 2π

where, ISC0 is the sort circuit current of the non-shaded solar cell and ISCi,j are short

circuit current of the solar cell (i,j), μ- is the mean, σ – is the standard deviation.

In general, the two above equations model the cast of a moving shadow on a solar PV

array for instant of time (t1 …t10). The cloud is moving with the speed (v=1solar cell/one

second). The cloud is most darkness in the center and brighter in the bounds. For purpose

of illustration, we will simulate as an example the followings: at the start point (t1) the

center of the cloud falls in the lowest left solar cell. At the end point (t10) the center of the

cloud falls in the highest right side solar cell.

The irradiance is approximately proportional to the short- circuit current, so that the

effect of a passing cloud to a solar array may be modeled as the change of short circuit

current through all elements of m x n matrix. Fig. 2.5 illustrates the graphical output of

the MATLAB program. The resulting irradiation on each individual solar cell (10x10

cells in this example) represents a passing cloud.

52
Step 3

Inputting the irradiance data

Inputting the irradiance data ( I SCi , j , VOCi , j ) from Step 2, the numerical method of [5] (in

MATLAB) can be used to determine the total current I, sub-voltages U and sub-currents

for each cell. In this method, Kirchhoff´s laws are applied to provide a system of

equations relating all currents and voltages in the network. Each element has to be

described by the mathematical relationship between current and voltage. The Newton-

Raphson method is used to calculate the solution for the systems of equations.

Newton-Raphson numerical method [13].

In order to simulate the solar cell I-V curve for each specific solar irradiance level and

ambient temperature, the current I must be computed individually for each given voltage

V.

To determine the characteristic curve numerical methods are normally used because of

the non-explicit form of the equation of the solar cell curve. An equation for the solar cell

is given in equation (2.1). For a given voltage V, the current I is given by the root of

equation (2.1), where I is expected to be a single root. In order to determine the single

root, it is suggested to use the Newton- Raphson method. By using the first few terms of

the Taylor series of a function f (x) in the vicinity of a suspected root this method has

the following procedure to find a root.

53
The f (I ) about the point I = I 0 + ε is presented in the Taylor series by:

1 ''
f ( I 0 + ε ) = f ( I 0 ) + f ' ( I 0 )ε + f ( I 0 )ε 2 + ....
2

Using two terms only to first order,

f ( I 0 + ε ) = f ( I 0 ) + f ' ( I 0 )ε

Starting from an initial guess I 0 , this expression can be used to estimate the amount of

offset ε needed to land closer to the root, Setting f ( I 0 + ε ) = 0 and solving (2.1)

for ε = ε 0 gives

f (I 0 )
ε0 = − ,
f ' (I 0 )

which is the first-order adjustment to the root's position. By letting I1 = I 0 + ε 0 a new ε 1

can be calculated. The offset in the n is:

f (I n )
εn = −
f ' (I n )

The algorithm can be applied iteratively to obtain

f (I n )
I n +1 = I n −
f ' (I n )

For n=1, 2, 3,…, an initial point I 0 , the process can be repeated until it converges to

a fixed point , which is precisely a root.

54
This procedure can be unstable near a horizontal asymptote or a local extreme but it can

be prevented a good initial choice of the root's position,

System of equations for solar PV array

The controlled switches shown in Fig. 2.4 may be relays or transistors; their switching

action is described by the following equation:

V Bi , j = I Bi , j * R Bi , j (2. 6)

Equation (2.6) has the form:

f ( I Bi , j , V Bi , j ) = 0 (2. 7)

When a solar panel contains a large number of solar cells, computational time may be

reduced by dividing the solar panel into modules. Assume that each module is uniformly

illuminated. Each vertical element in Fig. 2.4 is a solar module, made up of (m x n) solar

cells and protected by bypass diodes as shown in Fig. 2.7. Assume that all solar cells are

illuminated by the same irradiance. In this case, the equation for the vertical components

will be the following:

⎡ ⎛ q
f ( I M , V M ) = I SCM − pI 0 ⎢exp⎜ (VM + I M RSM )⎞⎟ − 1⎤⎥
⎣ ⎝ sakT ⎠ ⎦

⎛ V + I M RSM ⎞ ⎡ ⎛ − qVM ⎞ ⎤
− ⎜⎜ M ⎟⎟ − I M + I OBP ⎢exp⎜⎜ ⎟⎟ − 1⎥ = 0 (2. 8)
⎝ RSHM ⎠ ⎣ ⎝ a1 kT ⎠ ⎦

55
Where, IM - output current of solar module, IC - output current of solar cell, VM -

output voltage of solar module, VC - output voltage of solar cell, ISCM - short-circuit

current of solar module, ISC - short-circuit current of solar cell, RSM - series resistance of

solar module, RS- - series resistance of solar cell, RSHM - shunt resistance of solar

module, RSH - shunt resistance of solar cell, q- electron charge, a- ideality factor.

I M = pI C,VM = sVC , I SCM = pI SC sR S sR

, R SM = , R SHM = SH
p p

The last component in (2.8) is the equation for the bypass diode. In this case, the bypass

diode is connected in parallel to the whole module of the solar array and the irradiance of

each instant of time (t1 …t10)

The relationship between the current and voltage of the solar panel has the following

form:

f ( I Ai, j , V Ai, j ) = 0 (2. 9)

dI (V )
From (2.8) and (2.9) we can define for each element of the solar PV array. Using
dV

MATLAB command in Appendix A, we can solve system of non-linear equations.

Step 4

56
Graphically display the influence of a passing cloud on the output power. Based on this

curve, we can predict the maximum, minimum and average electrical power generated

from a solar PV array for non-uniform and variable irradiance using MATLAB.

2.3 Comparison between the proposed numerical model and

linear model

The proposed model is applied to simulate the solar PV array with the input data as

shown in Table 2.1.

TABLE 2.1 The parameters of solar PV array for simulation

Parameters Symbol Value Units

Number of solar cells mxn 10x10

Series resistance RS 0.1 Ohm
Parallel resistance RSH 52 Ohm
Saturation current of solar cell diode IS 3.4e-07 A
Ideality factor A 6.6
0
Cell temperature T 25 C

Fig. 2.6 shows the output power of a solar PV array for shadow conditions simulated by

the linear approximation method, which assumes that the output power of solar PV array

proportional to average irradiance level, and the numerical method.

57
 180

160

140

120

Output power (W)

100

80

60

40

20
Numerical simulation
Linear approximation
0
t1 t2 t3 t4 t5 t6 t7 t8 t9 t10
Time

Fig. 2.6 Output power of solar PV array simulation is based on two different methods

2.4 Simulation solar array with and without bypass diodes

Using the above simulation methods, we can compare the simulated output powers of the

solar PV array with and without bypass diode.

Fig. 2.7 The sub-module of solar PV array with bypass diode

58
2.4.1 With bypass diode

Assume that we have a solar module that is composed of solar cells, where each one is

protected by a bypass diode. When current goes through a bypass diode, the solar cells

that are in parallel (i.e. in the same row) will be shorted. The remaining solar cells will

operate normally.

160

140

120
Output power (W)

100

80

60

40

20 With bypass diodes

Without bypass diodes
0
t1 t2 t3 t4 t5 t6 t7 t8 t9 t10
Time

Fig. 2.8 Output power of solar PV arrays with and without bypass diode

2.4.2 Without bypass diode

In this case, the output current of a solar module will be limited by the most shadowed

row. For conditions of shadow cover, using a bypass diode causes different voltage levels

59
to develop in a solar array. Alternatively, without the using of a bypass diode, different

current levels develop in a solar array. Assume that illumination conditions are the same

Fig. 2.1 shows that the average power of a solar array with bypass diode is higher than

the average power of a solar array without bypass diode, although the difference is very

small because bypass diodes will eliminate whole rows of solar cells in which they are

connected to parallel.

2.5 Solar PV arrays of different configurations

Fig. 2.4 shows the two common solar PV array configurations:

Simple series-parallel (SP) array, when all the controlled switches are OFF and modeled

as insulation resistance,

R Bi , j = R OFF = 10 20 Ω (2. 10)

Total-cross-tied (TCT) array, when all the controlled switches are ON and modeled as

conductance,

RBi , j = RON (2. 11)

In general, by combining different ON/OFF switching matrices, different configuration

and simulation results of the solar PV array may be obtained. In Fig. 18, the output power

of two configurations is shown. We can see that for this specific illumination condition

the total-cross-tied interconnection array gives more output power than the series-parallel

interconnection array.

60
The production cost of TCT interconnection solar PV array usually is higher than the

production cost of SP interconnection solar PV array. Base on simulation results, the

designer can trade off between the output energy of solar PV arrays and their production

costs.

160

140

120

100
Output power (W)

80

60

40

20 TCT configuration
SP configuration
0
t1 t2 t3 t4 t5 t6 t7 t8 t9 t10
Time

Fig. 2.9 Output power of solar PV array depends on interconnection

Fig. 2.10 shows a typical result of calculation of the power losses for SP and TCT for an

array with 100 solar cells.

In this example, new simulation model can be used to simulate worst case shading

effects on solar PV array. For SP connection, each solar cell in a different column

completely shaded can cause up to 10% extra loss when shaded. The percent of lost

61
power depends on nonlinear functions, as described in Chapter 1. For example, if 6 solar

cells from different columns are completely shaded, the estimated power loss in Fig.

2.11, the point A, will be around 48%. On the other hand, if the 6 cells are in 3 different

columns, then the power loss, the point B, is reduced to about 17%.

For TCT connection, each solar cell in the same row completely shaded can cause up to

10% extra power loss. If two cells in different rows are shaded, the power loss is the

same ad one row. Referring to Fig. 5 again, when the 6 solar cells in the same row

completely shaded, then the power loss of the array is around 48%, the point A. If the 6

shaded cells have only three shaded cells in the same rows, then the power losses is about

17%, the point B.

Number of shaded solar cells in the different columns

Fig. 2.10 The losses of maximum power depend on the shaded area. In SP one cell in each
column is fully shaded. In TCT, one cell in the same row is fully shaded

62
In the worst case of 10 fully shaded solar cells (10% of the total number of solar cells in

the solar PV array), the maximum output power can be reduced more than 90%. The

result is the same for both SP and TCT.

2.6 Conclusions

A modeling and computing algorithm has been derived that simulates and analyzes the

effect of non-uniform changing shadows (i.e. a passing cloud) on the output power of

solar PV arrays.

The advantages and disadvantages of the topologies have been studied. The software

implementation in MATLAB is shown to converge and is fast enough to be used within

real time controllers. The model is able to determine the power losses in each solar cell

and the hot spots of a shaded solar PV array as well as the PV output power. The model is

flexible enough to simulate solar PV arrays with various configurations, with or without

bypass diode.

63
3. Solar PV Array Shadow Evaluation using Neural Network
and On­ Site Measurement

3.1 Introduction

As introduced in the previous chapter, this thesis focuses on the research of two dominant

fields of solar PV systems: (1) modeling and simulation of solar PV array under shadow

conditions and (2) reconfiguration of solar PV arrays to improve their performance under

shadow conditions. This chapter mainly discusses solar PV array performance evaluation

under shadow conditions using neural network and on - site measurement.

This chapter proposes a novel method to define the function of relationship between the

maximum output power (MPP) of the shaded solar PV array and the environmental

factors, such as the solar irradiation levels, the sun’s position angles and the ambient

temperature. This function, which is similar to the shading factor, can fully characterize

the shadow effects on the solar PV array for long term and is called the shading function.

As such, the shading function can eliminate the inaccuracy caused by the complexity of

the shading factor’s calculation.

Specifically, our approach proposes the following procedure to define the shading

function for the solar PV array under shadow conditions:

• The shadow ratio is characterized by the solar altitude and solar azimuth angles

that are easily received from the information of the time of day for specific

64
 location. Thus, the input signals of neural network are the solar irradiation levels,

the sun’s position angles and the ambient temperature. The output signal of neural

network is the maximum output power of the solar PV array.

• The input signals and output signal of neural network are measured for the one-

day period, when the solar PV array is shaded partly by any nearby object. The

measured data set is divided into a training set, and a test set.

• After the learning process of neural network, the test data prove the accuracy of

the proposed model. The verified neural network model, the shading function, is

able to predict the shadow effects on the solar PV arrays under any shadow

conditions, i.e. with any solar irradiation levels, at any time of a day, when the

solar array is shaded, and with different ambient temperature, for a long term and

with low computational effort.

3.2 Performance evaluation for solar PV systems

A important step when designing large solar PV systems (BIPVs, solar power plants) is

to evaluate the energy yield per year, month and day, etc. The system performance and

the associated economic return can be calculated for the system based on this design

constraint.

In practice, the conversion efficiency of a photovoltaic cell under outside operating

conditions may be different from the one measured according to Standard Test

Conditions procedures. In the Standard Test Conditions the PV module power ratings are

65
for of 1000 W/m2 solar irradiance and 25°C cell temperature and 1.5 Air Mass. Fig. 3.1

shows an example of annually global (total) irradiance on a horizontal surface and

aambient temperature.

Fig. 3.1 Example of annually global (total) irradiance on a horizontal surface and ambient
temperature.

As described in the previous chapter, the output energy of solar PV system heavily

depends on the environmental factors. Specifically, typical irradiance levels and ambient

temperatures must be known to estimate the energy generated by any solar PV system.

Obviously, environmental conditions vary from an installation to an installation. In

general, guidelines may be used to simplify estimation including the places with the same

latitude should have similar solar irradiance levels and ambient temperatures.

In order to support the design of solar PV systems, the National Renewable Lab has been

collecting irradiance levels, ambient temperatures, and wind speed from all US states.

One of the Internet accessible tools for calculating solar PV system performance with

66
meteorological input data is PVWATTS. In this software the Typical Meteorological

Year (TMY2) database is collected from 239 stations in United States. First, the PV

systems designers have to select a location and according grid cells (40-km by 40-km)

from the station map and sets of desired PV system parameters. Then, the PVWATTS

will perform an hour-by-hour simulation that provides monthly and annually alternating

current (AC) energy output in kilowatts and energy value in dollars [38]. The main

limitation of PVWATTS is that this software cannot simulate solar PV systems more

precisely, for example, minute-by-minute simulation.

[1] shows that the fundamental principle of the solar PV system sizing and evaluation is

the energy balance between the generated energy and consumed load:

H A × A × η PS × K = E L × DP × R
(3. 1)

Here, HA= in-plane irradiation (kWh/m2); A= array area (m2); η PS = PV efficiency at

STC; K = performance ratio; EL= load energy consumption (kWh); DP= solar energy

dependence; R= design redundancy for future load increase, safety margin.

The name plate d. c. power rating of the solar PV system equals the sum of all the module

powers listed on the nameplates on the backsides of the individual PV modules.

The energy yield of grid-connected solar PV system is measured in kWh and evaluated

for a specified solar PV system by the following equation:

HA
E P = PAS × × K = PAS × YH × K = PAS × YP
GS (3. 2)

67
Here, EP= system generated electricity (kWh)

K = K H × K PH × K PT × K PA × K PM × K B × K C (3. 3)

The system performance ratio K is the most convenient value to evaluate the solar PV

system performance. This parameter is normalized by site irradiation and system size.

The system performance ratio K consists of various parameters as shown in Fig. 3.2, such

as the following:

• KH= irradiation modification factor that characterizes the shadows, soil age of

surfaces;

• KPH= incident angle dependent factor, due to module glass structure reflection;

• KPT=cell temperature factor, due to the negative temperature coefficient of Pmax;

• KPA=array circuit factor, consisting of series-connected model mismatch and

wiring resistive losses;

• KPM=load matching factor, caused by mismatch operation apart from Pmax point;

• KB= battery circuit factor, including batteries and their peripheral losses;

• KC= power conditioner circuit factor, including power conditioner and their

peripheral losses.

These parameters are only major parameters influencing the system performance ratio K

in actual PV systems, not all that can be considered theoretically. These parameters are

usually calculated and evaluated in KWh, but KW.

68
Fig. 3.2 Annual average values for system parameters at 187 Japanese field test sites for 4/1995-
12/2001[1]

To accurately define the above listed factors, modeling and performance prediction of

solar photovoltaic (PV) arrays is an important task in the design of solar PV systems.

From the above discussion, the simple and reliable method to evaluate the shadow effects

for long term (year, month and day) is necessary.

The shadows may be caused by the shadow from one submodule row on the other in the

solar PV plants or by the nearby object such as trees, booms, or cloud passing… In order

to include the effects of shadowing on solar PV arrays, conventional methods use an

approach to determine the “shading factor,” which is defined as the ratio of the non-

shaded area to the total area of the solar arrays.

69
So shading factors also can be used to calculate MPP power of shaded solar PV arrays. In

reality, it is difficult to accurately estimate this shading factor because the shadows

change their shape and move with time.

This leads to several approaches to evaluate the shadow effects on those solar PV arrays:

3.2.1 Numerical method

In the numerical method, the solar irradiance on all solar cells is modeled based on real

time input data. The output power of the solar PV array at a given output voltage is then

received by solving differential equations on small time-scales (second, minutes) [19],

[20]. This method requires unrealistic number of sensors to measure the irradiance in

each solar cell. Furthermore, accurate prediction of the maximum output power of solar

PV systems for long term (days, months, years) is more challenging, due to difficulties in

modeling the shaded area in the solar PV systems.

3.2.2 Photogram metric method

In the photogram metric method, the position of obstacles and their shadow are estimated

by using the triangulation with two photographs or more [21].

70
3.2.3 Neural network method

Neural network method is used in [22], [23], [25] for predicting the dependence of

maximum output power of solar PV arrays on environmental factors. However, these

papers focus on estimating the performance of solar PV arrays under uniform

illumination.

Required data for identifying MPP

The maximum power of solar PV system mainly depends on the insolation. In the reality,

the sort circuit current ISC, gives the most adequate information of the effective

insolation. A linear correlation exits between the ISC and the optimal output current IOP.

Fig. 3.3 The correlation between the optimal power current and short-circuit current

71
To estimate maximum output power from the solar PV system, the optimal voltage, VOP,

is also required. The correlation between two parameters is non-linear and shown in Fig.

3.4.

Fig. 3.4 The correlation between the optimal power voltage and open-circuit voltage

3.2.4 The combined method

In the combined method, the neural network prediction model uses training data created

by the numerical method. The low accuracy is due to the low accuracy of the modeling

shadow objects [26].

From the above discussion, for the operation planning of solar power systems, there is a

need of a simple and accurate method to define the shadow effects on shaded solar PV

array for long term (seasonal or annual) in non-uniform illumination.

72
This research propose a model based on neural network that greatly expands on the ideas

in [23], which use the insolation data and ambient temperature data to predict the output

power of solar PV arrays under non-uniform shadow conditions.

Specially, the proposed model based on neural network has the following advantages:

• The model can define the main environmental factors that result in reducing the

output power of the solar PV array under shadow conditions. By analysis the

correlation between the input data, the dimension of the input vector can be

reduced by eliminating the components of the vectors that are highly correlated

(redundant).

• The simple experimental method is proposed to collect data necessary for the

training neural network. The required experimental equipments is minimum and

inexpensive.

• The appropriate neural network structure for these data by combining the

multilayer perceptrons feed forward network with a back-propagation algorithm.

• By random selection technique, the training data set, the test data set can be

received from the only one experimental data set. There is no need for repeating

the experiment for many days. but the accuracy of the function of the relationship

between the inputs and outputs from the training data set is verified by the test

data set.

• The shading function proposed in this method is the most convenient form to

store the information about the shadow effects in specific location. The

conventional methods using the table to record the data of the shading factors for

73
 hour, day, month, year that requires large memory space. To receive any

conclusion about the shadow effects using tables would require more data

analysis than our proposed approach.

3.3 Proposed method

3.3.1 Shadow evaluation using neural network without shading factor

measurement

Detailed analysis for the proposed approach to obtain a neural network model for the

shadow effects is provided in the following description.

In the numerical method, the shadow effects on the solar PV arrays are estimated by the

shading factor, measured as the ratio of the shaded area to the total area of the solar PV

array. In the proposed method based on neural network, instead of using the shading

factor calculation, we directly find the shading function that characterizes the relationship

between the maximum output powers of the shaded solar PV arrays and the

environmental factors: the horizontal solar irradiance, the sun’s position, and the

temperature:

PMAX = f SHD (Time,E,Te mp) = f SHD (α S ,γ S ,I SC ,Temp) (3. 1)

Here, referring to (3.9) to define the following variables:

PMAX- maximum output power of solar PV array, E-solar irradiation level, ISC- short-

circuit current, α S , γ S - solar position’s angles, Temp- ambient temperature, Time- time

of day, fSHD- the shading function.

74
A neural network is an adaptive machine that resembles the human brain behavior. The

neural network has a massively parallel distributed structure and the brain’s ability to

learn and generalize. Neural network can be used to detect and derive the meaning from

complicated or imprecise data. To derive the relationship in (3.1), the inputs of neural

network are measured data from environmental factors: the horizontal solar irradiance,

the sun’s position, and the ambient temperature. The output of neural network is the

maximum output power of the shaded solar array. The neural network will process

information, the large measured data of inputs and output, and learn from the data. The

neural network will derive the function, which characterizes the relationship between the

output and input signals in the form of matrix of weights. For future different

environmental factors, the maximum output power of solar array under shadow

conditions will be received in output of neural network by inputting the corresponding

data of the time of day, solar irradiance and the ambient temperature.

The dependence of maximum output power of solar array on time, solar irradiation levels

and ambient temperature is discussed below.

The solar irradiance levels

The average value, called the solar constant, Eo, of the solar irradiance measured outside

the Earth’s atmosphere on a surface perpendicular to the solar irradiations, is:

75
 Eo = 1.367 ± 2W/m 2 (3. 2)

The value of solar irradiance E measured in the Earth’s surface is usually smaller than the

solar constant due to reflection by the atmosphere, absorption in the atmosphere, and

scatterings.

In the solar photovoltaic array, the photo generated current or a short-circuit current is

linearly proportional to the solar irradiation levels E and the coefficient C0:

I PH ≈ I SC = C 0 E (3. 3)

The position of the sun

The sun’s position is defined by the two solar angles:

Sun’s height (solar altitude or elevation) γs, solar or sun azimuth αs is shown in Fig. 3.5.

These angles depend on the specific location of the solar PV array, the date, time, and

time zone.

76
 Fig. 3.5 The sun’s position at specific time at day

The angle of solar altitude is calculated by the following equation:

The angle of solar altitude:

γ S = arcsin ( cos ω.cos ϕ. cos δ + sin ϕ.sin δ) (3. 4)

The angle of solar azimuth:

If Solar time ≤12.00 hrs:

sin γ S . sin ϕ − sin δ

α S = 180 0 − arccos
cos γ S . cos ϕ (3. 5)

If Solar time ≥12.00 hrs:

77
 sin γ . sin ϕ − sin δ
0 S
α = 180 + arccos (3. 6)
S cos γ . cos ϕ
S

Here δ is the solar declination, ω is the hour angle and ϕ is the latitude of the location.

In general, Eq. (3.4) to (3.6) can convert from the time to solar angles on the specific

location of the solar PV array. These angels can be used as the input data for the neural

network model.

Temperature dependence

With the rising ambient temperatures, the open circuit voltage decreases significantly and

the short circuit current increases lightly. As a result, the maximum output power of the

solar PV array will decrease.

The maximum output power

Maximum output power is measured by connecting the variable resistor R to the external

circuit of the solar PV array. The value of R is defined by the following equation:

V MPP
R =
LOAD I MPP (3. 7)

here, VMPP and IMPP are the maximum output power voltage and the maximum output

power current.

78
3.3.2 Experimental set up

In this experiment, the solar PV array is outside and partly shaded by the nearby boom

for some hours of day. Thus, the shadow is moving slowly with time.

Fig. 3.6 The input and output signals measurement circuit

The signals are measured as shown in Fig. 3.6. Each diode presents the solar cell in total-

cross- tied solar PV array. The darkened/colored diodes represent the shaded solar cells.

Each shaded cell may be fully shaded or only partly shaded. So, the solar cells may have

different shading factors. We assume that clouds have uniform effects on solar PV arrays

and are not included in this experiment.

79
The shaded solar array is connected to the variable resistor RLOAD to track maximum

output power from solar array. One un-shaded solar PV cell is connected to very small

RSENS to measure the irradiation level (short-circuit current Isc).

The input signals are the short circuit current ISC, and ambient temperature. The output

signal is maximum output power of the solar PV array

PMAX = I MPP × VMPP (3. 8)

The signals are measured with interval of 30 seconds at particular time, when solar panel

is shaded. Using PC, the time of each measurement is recorded. The positions of the sun,

the two angles of the sun, are calculated from the time of day based on the MATLAB

software.

All measured data are divided to the training set and test set for neural network.

3.4 Neural network structure

3.4.1 Neural network structure

Fig. 3.7 shows multiple layers of neurons with nonlinear transfer functions that allow the

network to learn nonlinear and linear relationships between input and the corresponding

output value. This is a two layer feed-forward configuration.

80
The above experimentally measured data set includes the input data set: the time, the

solar irradiance, the ambient temperature and the output data set: the maximum output

power of solar PV array. This data set is divided into training, and test subsets. The

design of the neural network is as follows:

First, the input layer consist of source nodes equal the number of input signals, the output

layer consist of nodes with equal number of output signals. A subset of measured data

(the training set) is used to train the network by the back propagation algorithm. This part

of network design is called learning.

Fig. 3.7 Neural network’s structure

81
Second, the performance of trained neural network is tested with the data from the test

subset. The input signals of test subset are presented to the network. Then, the output of

the network is compared with actual measured output.

In the specified neural network, the domain knowledge (here it is the relationship

between the maximum output power of shaded solar PV array and the solar irradiation

levels and the ambient temperature at a particular time) is captured by the value of free

parameters (here they are the synaptic weights wi and wj and biases bi and bj) of the

neural network. The weights are adjusted by back-propagation algorithm. An effective

procedure for performing this operation is illustrated in Fig. 3.7.

3.4.2 Input signals correlation analysis

In some situations, the dimension of the input vector can be reduced by eliminating the

components of the vectors that are highly correlated (redundant).

Table 3.1 shows the correlation between the input signals of the neural network. The

short circuit current and the solar altitude have the correlation coefficient 0.98, so we can

remove the solar altitude from the input signals vector.

82
Table 3.1 shows that the input signals: the solar azimuth angle, the short circuit current,

and the ambient temperature. The neural network has one output: maximum output power

of solar PV array.

TABLE 3.1 The correlation between the inputs

SOLAR SOLAR Temp

Correlation ISC
ALTITUDE AZIMUTH .

Solar altitude 1

Solar azimuth -0.8355 1

Isc -0.9822 0.8580 1

Temp. -0.4120 -0.0559 0.3666 1

3.4.3 The learning process

Once the network weights and biases have been initialized, the network is ready for

training. The network is trained for function approximation. The training process requires

a set of examples which include network inputs and target outputs. The experimental data

is divided into training, and test subsets. We will take 70% of the data for the training set,

30% the test set.

83
 Fig. 3.8 The squared error of the training process

During training, the weights and biases of the network are iteratively adjusted to

minimize the network output estimation error. The default performance function for feed

forward network is the mean square error MSE, the average squared error between the

network outputs and the target outputs. Fig. 3.8 shows the squared errors after different

epochs for training process. The error is calculated by equation:

N N
1 1
Er. =
N
∑ (ei )2 = ∑ (PMEASi − PESTi )2
i =1 N i =1
(3. 9)

Here, PMEAS, PEST– the measured and estimated maximum output powers of solar PV

array, N- the number of measurements in the training subset

The result here is acceptable, because the training set error and the test error have similar

characteristics.

84
3.4.4 The generalization check

We can use data from the test subset to check the generalization of the network. The

generalization characterizes the ability of the network to correctly map the data that was

not involved in the learning. We can improve the generalization by using a larger

network or regularization and early stopping method.

3.4.5 The performance analysis of the network response

To investigating the network response, we will put the test data set through the network

and will perform a linear regression between the network outputs and the corresponding

targets. Then, we perform a regression analysis between the network response and the

corresponding targets. We pass the network output (predicted maximum output power)

and the corresponding targets (measured maximum output power) to the linear regression

analysis function. It returns three parameters shown in Fig. 3.7. If we have a perfect fit,

the slope is close to 1, and y –intercept is near 0. The R=0.994 indicates that the output of

the neural network track the targets well.

85
 Fig. 3.9 The network performance analysis

3.4.6 Analysis and discussion

14

12

10
Power (W)

8
Measured
6 Predicted

4 Maximum
possible

0
8:24 9:36 10:48 12:00 13:12 14:24

Time of Day

Fig. 3.10 The measured and predicted output powers of solar PV arrays

86
Fig. 3.10 shows the two predicted curves, maximum output powers of shaded and non-

shaded solar PV array, and measured maximum output power of shaded solar PV array.

We can see that the predicted maximum output power is very close to the measured one

(dashed) for the non-uniform shading. The trained neural network consequently may be

used for predicting the maximum output power for any other shadow conditions at any

time of a day, with any levels of solar irradiance or ambient temperature.

3.5 Conclusions

A neural network based model of the shadow effect on the maximum output power of the

solar PV array is described. The maximum output power or maximum output power

losses of the shaded solar PV array can be calculated based on shading function without a

shading factor calculation.

We proposed to develop a neural network method, combining the multilayer perceptrons

feed forward network with a back-propagation algorithm, to accurately predict maximum

output power of solar photovoltaic arrays under shadow conditions. Using the solar

irradiation levels, the solar photovoltaic array’s temperature, and the sun’s position

angles as the input signals, and the maximum output power of the solar PV array as an

output signal. The training data for the model, the shading function, can be easily

collected on site by periodically measuring the maximum output power of solar PV

arrays and measurable environmental factors over typical period of a day. The accuracy

87
and generalization of the network are checked by comparing the simulation data to the

measured data.

The model/ shading function can be employed to accurately predict the maximum output

power of the solar PV array over long periods of time with low computational efforts.

The approach uses only readily available solar irradiation data at different times of a day,

wherever the solar PV array is installed.

88
4. A Adaptive Reconfiguration of Solar PV Arrays

4.1 Introduction

This thesis generally can be categorized into two parts.

Mass production and use of electricity generated from solar energy has become more

common recently, perhaps because of the environmental threats arising from the

production of electricity from fossil fuels and nuclear power. However, in many

applications, such as solar power plants, building integrated photovoltaic, or solar tents,

the solar photovoltaic arrays might be illuminated non-uniformly. The cause of non-

uniform illumination may be shadows from clouds, trees, booms, neighbor’s houses, or

even the shadow of one solar array on the other, etc.

Furthermore, for applications like BIPV and solar power plants in small spaces, as

shown in Fig. 4.1, shadows from one solar PV array are commonly cast on another array,

[5]. In conventional solar PV plants, the designer must solve the complex problem, the

tradeoff between “maximum energy output” and “minimum produced energy cost” by

varying the distance between the rows For these new applications, it has been especially

important to optimize performance of the arrays in shadowed conditions. Because of the

nature of the electrical characteristics of solar cells, the maximum power losses are not

proportional to the shadow, but magnify nonlinearly. The shadow of solar PV array can

cause many undesired effects:

• The real power generated from the solar PV array is much less than designed, so

that the loss of load probability increases [17].

89
 • The local hot spot in the shaded part of the solar PV array can damage the solar

cells

There are several approaches that have been proposed to reduce the effect of shadows on

a solar PV array output power:

• Bypass diodes are connected across shadowed cells to pass the full amount of

current while preventing damage to the solar cell. This method usually requires a

great number of bypass diodes that are integrated in the solar arrays. The

production of solar arrays with bypass diodes is more costly. Furthermore, the

power losses of solar PV arrays are not prevented completely because there are

the additional power losses when the current passes though the bypass diodes.

• In large systems, each of the solar sub-modules can be connected to its own

maximum power point (MPP) tracking DC-DC converter and can individually

operate near its own MPP. Thus, the efficiency of the whole system is

increased, [42], but the method requires a large number of DC-DC converters

(equal to the number of solar modules).

• An alternate research field, and the focus of this paper, is to adaptively

reconfigure solar array connections in real time in order to track maximum output

power. Traditionally fixed solar PV arrays have hard wired interconnections

between their solar cells. These connections are not changed after installation.

However, it is possible to continuously rearrange solar cells in series and parallel

connections to facilitate the photovoltaic system to work more as a constant

power source, even in different operating conditions (i.e. insolation, temperature,

90
 loads,…etc). Research studies in have started to develop methods to reconfigure

solar cells to improve power output in shaded conditions. This research focuses

mostly on how to build the arrays, and does not propose real-time implementable

control algorithms (the focus of this paper). Because of this, the methods

proposed have unrealistic number of sensor and switches that must use complex

control algorithms to determine when it turns the switch ON or OFF.

This chapter develops research for third approach above. Adaptive reconfiguration of

solar arrays is proposed that requires significantly fewer voltages or current sensors and

switches than [16-17]. Further, for the first time published, simple real time

implementable control algorithms that determine how to adaptively reconfigure solar cell

connections have been developed. The algorithms have been experimentally tested on a

small scale solar reconfiguration system.

Specifically, this chapter presents the following research contributions:

• A new method for reconfiguration of solar PV arrays in real time under shadow

conditions is presented. Solar cells from a (smaller) solar adaptive bank will be

connected to the (larger) fixed part of the solar array. The maximum power point

(MPP) of the whole array is tracked by a single common maximum power point

tracker (MPPT), instead of many MPPTs as shown in [9-13]. Since only a small

percentage of the solar arrays are reconfigurable, fewer switches and simplified

control algorithms are possible. In the uniform illumination conditions, all these

adaptive solar cells will be equally connected to all rows of the fixed part of solar

91
 PV array. In non-uniform illumination conditions, the number of the adaptive

solar cells connected to the shaded submodules depends on the shaded area of

submodules. The configuration is executed through the proposed switching matrix.

• Simple control decision algorithms are presented to determine when and how to

open and close switches between the fixed part and adaptive bank of the solar PV

array. The algorithms rely on model predictions that can be implemented in real-

time by a micro-controllers or Digital Signal Processors.

• An experimental adaptively reconfigurable solar PV has been built and tested to

verify proposed reconfigurations. It is shown that the approach proposed is able to

increase output power of solar PV array in real-time under shaded conditions by

30% for a typical experiment. The experimental results are presented in Chapter 5.

4.2 Shadow variation on photovoltaic collectors in a solar field

One of the new trends in solar PV energy is to install solar PV plants on the roof of big

city buildings. The investment cost of Building integrated photovoltaic (BIPV) power

plants is reduced because of the relatively low price of land and the minimum electrical

transmission losses. On the other hand, the limited roof area of BIPV power plants causes

additional unavoidable power losses from shadow cast by nearby trees, booms and other

nearby buildings.

92
If there is no space limitation; the distance between the rows can be large, the shadow

effects from one row to other is kept to a minimum. On the other hand, if the size of land

is limited or economy calculation is taken into consideration, the tradeoff between the

row’s distance and the shadow losses have to be considered [8], [9].

K=1 K=2 K=3

H
D

W
S N L

Fig. 4.1 Shading by collectors in solar field

Fig. 4.1 shows the shading by collectors in a solar field. The solar field has fixed length L

and width W. K is the number of solar collectors in the field. The use of many rows (K is

large) will increase both the solar PV area and the shadowing losses. A small number of

rows will reduce the solar PV area and shadowing effects but the energy output from the

solar field may be reduced. In the conventional solar PV plant design approach, the solar

PV system’s designer has to solve the complex problem, the tradeoff between “maximum

energy output” and “minimum produced energy cost” by varying D-parameter, the

distance between the rows, as shown in Fig. 4.1.

93
The relative shaded area aS:

a S = λ S × hS (4.1)

The relative length of the shaded area λ S :

LS d sin β + cos β sin γ

λS = = 1− × (4.2)
L λ cos β tan α + sin β cos γ S

Here, α ≥ 0 - the sun elevation angle, γ ≤ 90 - the solar azimuth angle, 0 ≤ λ s ≤ 90

The relative height of the shaded area:

HS d sin β + cos β
hS = = 1− (4.3)
H sin β cos γ
cos β +
tan α

0 ≤ hs ≤ 1

d is the normalized horizontal distance between the collectors:

D
d= (4.4)
H sin β

And λ is the normalized collector length

L
λ= (4.5)
H sin β

4.3 Conventional methods of reconfiguration solar PV array

conventional methods

There are several approaches that have been proposed to reduce the effect of the shadow

on the solar PV array output power:

94
 • Bypass diodes are sometimes connected across shadowed cells to pass the full

amount of current while preventing damage to the solar cell. This method usually

requires a great number of bypass diodes that are integrated into the solar arrays,

adding cost. The power losses in the solar PV array are only partially prevented

due to the added shorts within the whole module.

• Alternatively, in large systems, each of the solar sub-modules can be connected to

its own DC-DC converter and can individually operate near to its own maximum

power point. Thus, the efficiency of the whole system is increased [37], [40], but

the method requires a large number of DC-DC converters (equal to the number of

solar modules).

• It is possible to continuously rearrange solar cells in series and parallel

connections [34] to facilitate the photovoltaic system to work more constantly as

the power source in specific design operating conditions (i.e. insolation,

temperature, loads, etc). Recent research studies [35], [36] started to develop

methods to reconfigure solar cells in shaded solar PV array. However, it has a

drawback of requiring an unrealistic number of switches and sensors that must use

a complex control algorithm to determine when it turns the switch ON or

OFF [36].

In conventional methods, the solar PV array contains two parts: fixed part and adaptive

part. The fixed part is main part of solar array. Solar cells in fixed part have fixed

configuration, they are connected by wires.

95
The adaptive bank contains the reconfigurable solar cells. They can be connected

together or connected in parallel or in series with the fixed part.

If adaptive solar cells are connected in parallel, the output current of solar array increases.

If adaptive solar cells are connected in series with the fixed part, the output voltage

increases. The photovoltaic solar arrays are neither constant current nor constant voltage

source; therefore it needs special treatment when operating it as a power source. The

combination of cells in series and parallel arrangements results in specific maximum

power output under design operating conditions (i.e. insolation, temperature, loads, etc).

It is of course required to get the maximum power output from the photo voltaic system

as possible. So if we can let the photovoltaic system work as constant power source this

will be the best solution. Fig. 4.2 shows the well known I-V characteristics of a solar

array with superimposed a constant locus of maximum power points.

As shown in Fig. 4.2 it is clear that the locus of the maximum power points is nearly

perpendicular to the constant power locus. The best solution of course is to let the locus

of the maximum power points coincident with the maximum constant power locus.

96
 MPP

Fig. 4.2 I-V curves with different irradiance level

The possible arrangement of solar PV array:

Fig. 4.3 Parallel arrangement

97
Fig. 4.3 shows a parallel arrangement of solar PV array. The fixed part contains solar

cells (PV(1,1), PV(2,2), …, PV(m,n)). The adaptive part contains series-connected

strings of solar cells ( APV(1,1), APV(1,2),…,APV(m,1))…(APV(1,4), APV(2,4), …

APV(m,4)). By turning switches (B1, B2, B3, B4), these strings will be connected in

parallel with the fixed part. The output current IOUT will increase.

Fig. 4.4 Series arrangement

98
Fig.4.4 shows a series arrangement of solar PV array. The fixed part contains solar cells (

PV1(1,1), PV(2,2), …, PV(m,n)). The adaptive part contains parallel-connected

submodules of solar cells ( SPV(1,1), SPV(1,2),…,SPV(m,1))…(SPV(1,4), SPV(2,4), …

SPV(m,4)). By turning switches (S1, S2, S3, S4), these strings will be connected in series

with the fixed part. The output voltage VOUT will increase.

Fig. 4.5 Series-parallel arrangement

Fig. 4.5 shows a series-parallel arrangement of solar PV array. It is the combination of

series arrangement and parallel arrangement. By turning switches (S1, S2, S3, S4) or ( B1,

B2, B3, B4), the strings or submodules of adaptive solar cells will be connected in series

or parallel with the fixed part. Both the output voltage VOUT and output current will

increase. The maximum output power point will be the intersections point between the I-

V curves and the maximum output power locus. It is shown in Fig. 4.7.

99
 Fig. 4.6 I-V curves for PV with series-parallel arrangement of full sun

Fig. 4.5 shows a fully reconfigurable solar PV array. The solar cells are continuously

rearranged in series or parallel connections to get optimize connection. The drawback is

that there is unrealistic number of switches and sensors. Furthermore, there are many

possible configurations but a simple control algorithm to determine how and when to

adaptively turn switch ON or OFF has yet to be developed.

This research project develops a new approach for adaptive reconfiguration of solar

arrays. We propose methods that require significantly fewer switches and sensors than

those in [35], [36] and also have simpler implementation of the real time control

algorithm. Our methods will rely on model prediction from our new models from Chapter

2. This research develops adaptive reconfiguration algorithms to reconfigure connections

within solar PV arrays in real time, under shadow conditions, in order to optimize output

power.

100
We propose to connect a matrix switches between a “solar adaptive bank” and a standard

TCT fixed solar PV array. Various control algorithm approaches, the solar PV array is

reconfigured to produce optimal power. Control algorithms will be implemented in real-

time. Implementation of neural network as well as model based control algorithm will be

investigated.

The specific achievements of the proposed reconfiguration method are the followings:

• Proposing a simple method to monitor the shadow pattern in the solar PV array:

the shadow pattern is defined by using just two monitoring voltage sensors instead

of using one sensor for each solar cell.

• Defining the configuration that solar cells can be reconfigured to compensate the

shaded cell: after analyzing the common configurations of solar PV arrays, we

decided to choose the TCT configuration for the fixed part of the adaptive solar

PV array, because in this type of configuration, the shaded solar cells of a row of

fixed part can be compensated by connecting the illuminated solar cells in parallel

to this row.

• Proposing the two simple control algorithms for reconfiguration of the solar

array: the “bubble sort” control algorithm and the control algorithm based on the

model developed in Chapter 2 can be applied for large solar PV system.

101
4.4 Proposed a adaptive reconfiguration method

This section proposes a system architecture that permits adaptive reconfiguration of the

connections between solar PV arrays. A switching matrix that connects a small

reconfigurable bank of PV arrays with a larger non-reconfigurable bank of solar PV

arrays is proposed. Because only the adaptive bank is being reconfigured, the number of

switches and reconfiguration time seems computationally efficient. We propose two

different adaptive reconfiguration algorithms for the arrays, each of which will eventually

produce the same increase in power under shadow conditions. The first method is

simpler, and it relies on a serial “Bubble – Sort” approach, which switches the adaptive

PV arrays in one at a time. After each switching, the power of the total system is

analyzed and the next sort is implemented. In the second method, a model reference

approach is proposed that is able to predict power levels in each of the rows of the fixed

solar arrays and uses this prediction to simultaneously switch the connections of the

adaptive bank.

These reconfiguration control algorithms are applicable even when the some of the solar

cells of the adaptive photovoltaic bank are shaded: the control algorithms will

automatically connect the most illuminated solar cells in the adaptive bank to the most

shaded row of the fixed part. If a solar cell in the adaptive bank is shaded it will still be

connected to the fixed solar array, perhaps to a more illuminated row. Since there is a

parallel connection, the only implication is that there will be only a small amount of

current/ power added from the shaded adaptive solar cell. In either case, the two different

approaches eventually will produce the same increased power of the entire solar PV

102
system, and both are implementable on a Digital Signal Processor in real time. However,

the second approach is quicker, as we later experimentally demonstrate. Fig. 4.10 shows

the operation principle of the proposed method.

The fixed part (m rows and n columns)

The fixed part is the main part of the solar array and has most number of solar cells, as

shown in the Fig. 4. 7. The fixed part contains (m x n) solar cells (the solar cells (1,1),

(1,2), …(m, n-1), (m, n). All solar cells in the fixed part are connected by hard wire and

have fixed configuration, with TCT, the total-cross-tied interconnection. We can consider

that the fixed part of solar PV array has m PV “modules” connected in series.

The solar adaptive bank of solar cells

In this paper, the adaptive bank has m solar cells (the solar cells A1, A2, …, Am) not

connected together i.e. the number of adaptive solar cells is same as the number of rows

in the fixed part.

These adaptive solar cells can be connected in parallel to any PV module, from PV1 to

PVm, as seen in Fig. 4.11. We remark that fewer than m adaptive solar cells can also be

used. (The algorithms still work and there will still be power improvement after

reconfiguration, as shown in experimental Chapter 5). However, by selecting m adaptive

solar cells, there is the advantage that when there is uniform illumination and no need for

reconfiguration, then the adaptive bank can form one additional column to the fixed part.

In this case, the entire PV system will behave like an (m x (n+1)) array.

103
 Fig. 4.7 Practical circuit of the proposed reconfiguration: For m rows in the fixed solar array,
we proposed m solar cells in the adaptive bank

The switching matrix

The fixed part and solar adaptive bank are connected together through the switching

matrix. The switching matrix, as shown in Fig. 4.8, contains switches S(1, 1), S(1,

104
2), …S(m-1, m-1), S(m, m) connect each solar cell in the adaptive bank to any row of

solar cells in the fixed part of the PV array.

Fig. 4.8 The switching matrix

When the switch S (i, j) is ON, the solar cell Ai from the adaptive bank will be connected

to row j of a fixed part. Thus, only one set of switches in a column can be ON at a time.

The switches can be either relays or electrical switches. In the proposed method, each

switch carries just the current of one solar cell (or one submodule), which is normally 1A

~ 5A, even for large solar power plants. Thus, if mechanical relays are selected, the

arcing phenomena may not be burdensome since the current is so low. On the other hand,

small electrical switches (in SO8 packages) are made at suitable power and voltage

ratings for applications such as in Fig. 4.1. Furthermore, these small package sizes are

easily embedded/sewn into the fabric already. So, it seems that these portable, foldable

solar panels are prime candidates for this reconfiguration approach.

105
4.4.1 “Bubble - Sort” method

The flow chart of the “Bubble - Sort” method is shown in Fig. 4.9. In general, the solar

array is reconfigured by the following principle: if the voltage of the one row is smaller

than voltages of the other rows, it indicates that this row is the most shaded row. One

solar cell from the adaptive bank will be switched in parallel to this row. The process will

continue until all the solar cells of the solar adaptive bank are connected in parallel to the

rows of the fixed part.

Step 1

The solar adaptive bank and fixed part of the solar array are connected together in the

original configuration as in Fig. 4.9 (a). The voltage V1 and VOUT are monitored, where

V1 is the voltage produced by row 1 of the solar array. In the uniform illumination, the

output voltage: VOUT = m × V1 .

In the non-uniform illumination, two situations can occur:

The first case occurs when the first row is shaded and the other rows are non-shaded. In

this case, the voltage of first row is reduced and it is less than the threshold voltage:

V1 < δV , the adaptive reconfiguration starts. The second case occurs, when the first row is

non-shaded, and at least one of the other rows is shaded, the output voltage of the solar

array is not equal to m × V1 , but much less than that: VOUT − (m − 1) × V1 < δV In this case

also the adaptive reconfiguration starts.

106
We propose to connect a matrix switches between a “solar adaptive bank” and a standard

TCT fixed solar PV array. Various control algorithm approaches, the solar PV array is

reconfigured to produce optimal power. Control algorithms will be implemented in real-

time. Implementation of neural network as well as model based control algorithm will be

investigated.

Fig. 4.9 The flow chart of the control algorithm for “Bubble - Sort” method

107
Step 2

By the opening all switches S(1,1), S(2,2), …, S(m, m) , all the solar cells in the solar

adaptive bank are in the open circuits. Define, sort in the decreasing order: the open

circuit voltages of all solar cells of the adaptive bank:

V0 A1 > V0 A2 > .... > V0 Am

Step 3

Define the number of adaptive solar cells connected parallel to the shaded solar

submodules in the fixed part.

Sorting: First, measure the voltages of all submodules of the solar fixed part. Next, the

voltages of all submodules of the fixed part are sorted in increasing order:

V1 < V2 < ... < Vm , the voltages of the fixed part. Thus, the rows of the fixed part and each

adaptive solar cell have been renumbered according their sorting voltages.

Adding: Connect the solar cell with the maximum open circuit voltage of the solar

adaptive bank in parallel to the most shaded sub-module of the fixed part, which has the

smallest voltage. For example, if we switch solar cell A1 in parallel with the sub-module

in row 1, then after the first switching, the solar adaptive bank becomes:

V0 A 2 > V0 A3 > ... > V0 Am

The fixed part might become:

108
V2 < V1 < .... < Vm

Then the second switching occurs and the solar cell A2 is connected in parallel with the

sub-module in the row 2. We continue the reconfiguration process until all the solar cells

of the solar adaptive bank are connected parallel to the rows of the fixed part.

Step 4

When the shadow changes direction or shape, the voltage of the first row and the output

voltage are continuously being measured and compared, and give the command to repeat

the reconfiguration process, if the difference between them is above the fixed range, the

control circuit repeats the procedure in Steps 2-3.

4.4.2 Model-based method

The flow chart of the model - based method is shown in Fig. 4.10. The control algorithm

to determine how to connect and reconfigure the solar cells is based on a model-based

control method and contains the following steps:

Step 1

It is the same as Step 1 in the “Bubble - Sort” method

Step 2

109
Define the photo generated currents of all solar cells of the solar adaptive bank and of all

submodules of the fixed part.

We can use the model-based method to calculate the photo generated currents in Step 3.

The benefits of this approach over the previous method is it defines photo generated

currents of all solar cells of the solar adaptive bank and of all submodules of the fixed

part, so all switches can be controlled synchronously at the same time.

The photo generated currents of the solar adaptive bank: By the opening all switches

S(1,1), S(2,2), …, S(m,m), all the solar cells in the solar adaptive bank are in the open

circuits. By measuring the open circuit voltages of solar cells {VOCA1, VOCA2, .., VOCAm},

the photo generated currents IPHAj are estimated by the following equation:

VOCAj ⎛ ⎛ qVOCAj ⎞ ⎞
I Aj = + I S * ⎜ exp⎜⎜ ⎟ − 1⎟
⎟ ⎟ (4.11)
RSH ⎜ akT
⎝ ⎝ ⎠ ⎠

Here, RSH - shunt resistance of solar cell, submodule, IS - saturation current of the solar

cell,

The photo generated currents of the fixed part

All submodules of the fixed part are still working with load. Their photo generated

currents are calculated by the equation:

⎡ ⎛ q
IFj = IOUT + nIS ⎢exp⎜ (Vj + IOUTRSM )⎞⎟ −1⎤⎥ + ⎛⎜⎜Vj + IOUTRSM ⎞⎟⎟ (4.12)
⎣ ⎝ akT ⎠ ⎦ ⎝ RSHM ⎠

where, Vj={V1, V2, ..,Vm} - measured voltages of submodules, IOUT – the output current

of solar array, RSM - series resistance of solar cell, submodule, RSHM - shunt resistance of

110
solar cell, submodule, IS - saturation current of the solar cell diode, n – number of solar

cells in submodule.

Fig. 4.10 The flowchart of the control algorithm for Model-based method

111
These parameters of solar cell or solar submodule are received from manufacturers or

extracted based on experiment by V-I curve fitting. By updating temperature and voltage

parameters in the model (in real-time) it is possible to estimate photo generated currents

IPHAj and IPHFj of each solar cell or submodule by using[19] [20], and their dynamics

multi-physics models[51].

Step 3

Define the number of adaptive solar cells connected parallel to the shaded solar

submodules in the fixed part.

Sorting: First, the photo generated currents of all solar cells of the solar adaptive bank are

defined and sorted in the decreasing order: I A1 > I A 2 > .... > I Am , the current of the

adaptive bank

Next, the photo generated currents of all submodules of the fixed part are sorted in

increasing order: I F1 < I F 2 < ... < I Fm , the currents of the fixed part.

Thus, the rows of the fixed part and each adaptive solar cell have been renumbered

according their sorting predicted photo generated current.

Adding: Connect the solar cell with the maximum predicted photo generated current of

the solar adaptive bank in parallel to the most shaded sub-module of the fixed part. For

example, if we switch solar cell A1 in parallel with the sub-module in row 1, then after

the first switching, the solar adaptive bank becomes: I A2 > I A3 > ... > I Am

112
The fixed part might become: I F 2 < I F 1 + I A1 < .... < I Fm

Then the second switching occurs and the solar cell A2 is connected in parallel with the

sub-module in the row 2. We continue the reconfiguration process until all the solar cells

of the solar adaptive bank are connected parallel to the rows of the fixed part.

Step 4

When the shadow changes direction or shape, the voltage of the first row and the output

voltage are continuously being measured and compared, and give the command to repeat

the reconfiguration process. If the difference between them is above the fixed range, the

control circuit repeats the procedure in Steps 2-3.

Table 4.1 compares the previous reconfiguration structures [35][36]with the proposed

method. In the previous reconfiguration approaches, it is assumed that all solar cells are

“adaptive,” that is that they can all be reconfigured to each other. In our proposed

approach, we assume that there is a fixed array and that this array has m rows and n

columns of solar cells. Often the number of columns is made larger than rows, as the

currents in each columns are added together to increase power.

Notice that the proposed method always uses fewer sensor, and in the case when the

number of columns is significantly greater than the number of rows (n >m), as commonly

113
occurs, then there are significantly fewer sensors. Similarly, when n>m, the number of

switches also can be significantly decreased.

We remark that [16], [17] do not present methodologies to determine how to reconfigure

the solar arrays. Thus, we are unable to make comparisons in control algorithms.

TABLE 4.1 Comparison of the Reconfiguration Methods

(m=rows, n=columns in fixed array)

No Item Conventional Proposed

1 Number of solar cells Fixed: 0 Fixed: m(n)
(submodules) Adaptive: Adaptive: m
m(n+1)
2 Number of voltage + m(n+1) 2m+1
current sensors
3 Number of switches 2m(n+1) to 2m2
3m(n+1)
4 Control Algorithms N/A Bubble - Sort
Model-based

One example of the reconfiguration method is shown in Fig. 4.11. The solar PV array

from 10x10 solar cells in the fixed part, 10 solar cells in the adaptive part has the

interconnection in 4.11(a), in the uniform illumination, in Fig. 4.12(b), in the non uniform

illumination. The figures show that by using the proposed adaptive reconfiguration

method, when the number of shaded solar cells is less than the number of solar cells in

one row, the maximum power loss is at most equals the power of the one row. If no

adaptive bank and TCT connection, the shade would have up to 30% reduction in

114
maximum output power (3 rows). With the adaptive bank, there is only 10% reduction in

power.

Before configuration

b) After reconfiguration

Fig. 4.11 Solar array’s reconfiguration under non uniform illumination

115
4.5 Conclusions

This research develops a new method for reconfiguration of solar PV arrays under

shadow conditions to maximize the possible output power.

A summary of the research contribution for the achievement of Objective 3 is as follows:

• A new method for reconfiguration of solar PV arrays in real time under shadow

conditions has been developed: Solar cells from a (smaller) solar adaptive bank

will be connected to the (larger) fixed part of the solar array. The maximum

power point (MPP) of the whole array can be tracked by a single common

maximum power point tracker (MPPT), instead of many MPPTs as in [37], [38],

[39], [40]. Since only a small percentage of the solar arrays are reconfigurable,

fewer switches and simplified control algorithms are required. In the uniform

illumination conditions, all these adaptive solar cells will be equally connected to

all rows of the fixed part of solar PV array. In non-uniform illumination

conditions, the number of the adaptive solar cells connected to the shaded

submodules depends on the shaded area of submodules. The reconfiguration is

executed through the proposed switching matrix.

• Simple control decision algorithms to determine when and how the switches

between the fixed part and adaptive bank of the solar PV array are open and

closed. The algorithms will rely on predictive models that can be implemented in

real-time by a micro-controllers or DSP. To our knowledge, it is the first time that

such adaptive and implement able algorithms have been reported.

116
• The switching matrix is working in at relatively low switching frequency, so the

electromechanical relays can be used to increase the current going through panels.

The number of voltage sensors can be increased if it is desired to reduce the

calculation time. The simple sorting algorithm is applied to reduce the calculation

time.

• The proposed reconfiguration method helps avoid the local maximum power

points of solar PV arrays, thus the central maximum power point tracker can be

used to track the maximum output power of the whole solar PV array.

117
5. A Small Scale Solar Photovoltaic Test Bed

5.1 Introduction

Fig. 5.1 shows the solar PV array test platform. The experiment includes the following:

The fixed solar array is from 9 solar cells, 3 rows (m) and 3 columns (n) (TCT

connection). The solar adaptive bank combine from 3 solar cells (one string of 3 x1 cells)

connected to the fixed part (3x3). The (3x3) switching matrix, controlled by Agilent Data

Acquisition/Switch Unit 34970A, is connected between the fixed part and the solar

adaptive bank. In total, we use 18 switches in the switching matrix and 6 voltage sensors

(two of them are voltage monitors) for this test platform. The output load in all the

experiments is 10 Ω.

Fig. 5.1 Solar PV array outdoor test platform

118
The voltages of solar cells in the solar adaptive bank and solar submodules in the fixed

part are continuously measured and sent to PC running real-time MATLAB software.

The sorting algorithm previously described is implemented by the PC in real-time.

5.2 Bubble-Sort method

Fig. 5.2 shows two separate experiments: Experiment 1 is from 0< t < T3. In this time

interval, the figure shows the output voltage of the solar array when 4 solar cells are

partially shaded. From time equal 0 to T1, the output voltage is under uniform

illumination. From T1, solar PV array is shaded. The interval T1-T2 is time for

measurement and control. After T3, output voltage of solar array has been optimized and

the reconfiguration has been completed. Notice that from T3 to T4, clouds move across

the array and cause a change in the output voltage. However, no reconfiguration occurs

since the clouds affect both the adaptive and fixed part of the solar array equally. Hence,

this demonstrates a robustness of the proposed algorithms to falsely reconfigure when it

is not necessary.

The output power of solar PV array is calculated by the following equation:

2
VOUT (1)
POUT =
R

For this Experiment 1, there is a 62.3% increase in output power from before to after the

reconfiguration.

119
 T1 T2 T3 T4 T5 T6

Fig. 5.2 Output voltage of solar PV array before and after reconfiguration in Two – Bubble Sort
Experiments

T1 T2 T3 T4

Fig. 5.3 Output voltage of solar PV array before and after reconfiguration in Two Model-Based
Experiments

Experiment 2 of the Bubble-Sort method is performed from t > T4 in Fig. 5.2. At t < T4,

the solar system is in the configuration stage from the previous Experiment 1. All of a

sudden at t= T4, the shadow configuration is suddenly changed so that two cells in the

120
same row of the fixed part are fully shaded. The interval T4-T5 is time for measurement

and control. After T6, output voltage of solar array has been optimized and the Bubble-

Sort reconfiguration has been completed. Using the data in Fig. 5.2 with (8), it is

calculated that there is a 156% increase in output power from before to after the

reconfiguration. In both experiments, the Bubble-Sort reconfiguration takes about 18

seconds.

5.3 Model-Based method

Fig. 5.3 shows another two experiments. Experiment 3 is from 0< t < T2, and Experiment

4 is for t> T3. In Experiment 3, the output voltage of the solar array when 4 solar cells

are partially shaded is shown in the figure. From 0 to T1, the output voltage is under

uniform illumination. From T1, solar PV array is shaded. The interval T1-T2 is time for

measurement and control. After T2, output voltage increases of solar array after

reconfiguration. Notice that the voltage across the load is increased from T2, by 20%

indicating a 33% increases in power due to reconfiguration. The output voltage of solar

array initially drops during real time reconfiguration (between T1 and T2). This is because

the solar adaptive bank is disconnected from the load for short time to measure and sort.

During the transition, though, the fixed part is still supplying power for the load.

In Experiment 4, shown in Fig. 5.3 for t > T3, the shadow configuration suddenly changes

so that three solar cells in the fixed part are completely shaded. The three solar cells in

121
the adaptive bank remain illuminated. The reconfiguration starts at t = T3 from the

previous reconfiguration connections of Experiment 3, but this provides no technical

difficulties. From T3-T4, there is measurement and control. After T4, the switches are

controlled synchronously to reconfigure. For this experiment, the reconfiguration from

T3 to T4 led to dramatic increases of power of a 3900% improvement from before to after

reconfiguration.

5.4 Comparison between “Bubble - Sort” and “Model-based”

methods

The second reconfiguration method takes about 8 seconds. This method is faster because

all switches can be controlled synchronously at the same time, when the “Bubble - Sort”

method switches one at a time instead of all at once. The final reconfiguration state does

not differ between the two methods.

5.5 Partly shaded solar adaptive bank

At times, one can expect that the adaptive solar cells may also become shaded. As

previously discussed, all the algorithms work when this occurs. For the case when there

are still more illuminated adaptive solar cells than shaded fixed solar cells, there will still

be significant power improvements after reconfiguration. Fig. 5.4 shows the output

voltage of the solar cells when 2 fixed solar cells and one adaptive solar cell are fully

shaded. (Thus there are still 2 adaptive solar cells fully illuminated.) As in the previous

two experiments, from 0 to T1, the output voltage is under uniform illumination. After T1,

122
two fixed solar cells in the same row and one adaptive solar cell are fully shaded. The

interval T1-T2 is time for measurement and control. At T2 three illuminated adaptive solar

cells are connected in parallel with the shaded row at the same time. Using the data in Fig

5.4 with (8), it is calculated that there is a 65% increase in output power from before to

after the reconfiguration.

5.6 The number of shaded fixed solar cells is larger than the
number of solar adaptive bank

It is possible that many adaptive solar cells become shaded and that there will not be

sufficient number of them to compensate for all the shaded solar cells in the fixed part.

Alternatively, this could also occur if fewer adaptive solar cells are used, as we only are

recommending “m” number of adaptive cells. Still, there is benefit of the reconfiguration

algorithms to add power to the system, and both reconfiguration algorithms will still

work. Of course, less power can be added to the system.

Fig. 5.5 shows the output voltage of the solar cells when 4 fixed solar cells are fully

shaded, 3 of which are in the same row. There remain 3 fully illuminated adaptive solar

cells.

From 0 to T1, the output voltage is under uniform illumination. After T1, four fixed solar

cells in the same row are fully shaded. The interval T1-T2 is time for measurement and

control. At T2, three illuminated adaptive solar cells are connected in parallel with the

123
shaded row at the same time. For this experiment, the reconfiguration from T1 to T2 led to

dramatic increases of power of a 4125% improvement from before to after

reconfiguration.

T1 T2

Fig. 5.4 Output voltage of solar PV array before and after reconfiguration when solar adaptive
bank is partly shaded

T2
T1

Fig. 5.5 Output voltage of solar PV array when the number of shaded fixed solar cells is larger
than the number of solar adaptive bank

124
5.7 Conclusion

A new approach for adaptive reconfiguration of solar PV arrays under shadow conditions

is described. A matrix of switches is used to connect a “fixed” total- cross- tied array with

an adaptive array that can be reconfigured. Simple control algorithms are presented that

determine how to control the switches to optimize output power. An experimental

adaptively reconfigurable solar PV array has been built and tested to verify the proposed

configurations.

The switching matrix may still require a large number of switches, and this represents a

disadvantage of the approach. However, we show that when the number of rows in the

solar array is not too large, the number of switches and sensors is significantly reduced.

Further, we present four separate experiments which demonstrate the technical feasibility

of two different reconfiguration algorithms. The reconfiguration methods are shown to be

effective in improving solar array power, even when the number of shaded arrays is more

than the power provided by the adaptive solar modules to be switched into the system.

The Model-Based sorting algorithms have reduced calculation time.

125
6. Single ­ Stage Two ­ Input DC­DC Converter for MPPT
Solar PV Arrays

6.1 Introduction

In general, a solar array consists of many solar PV submodules, which are connected in

parallel and/or series to meet the output requirements. Insolation on each solar module

may be different, because of incident angle of solar light, the mismatch of solar

submodule parameters, and shadows of nearby objects or clouds.

In the parallel connection, a single shadow over a solar cell does not affect the power

delivered by the other solar cells in the submodule. Thus, it is the most robust

configuration for solar array under shadow conditions. However, the output voltage is

very low (~0.5V).

In the series connection, all the solar cells share the same current, IOUT. Thus, when one

solar cell becomes shaded, it directly affects the power delivered by the other cells. That

is, output current decreases throughout all the cells and there is a significant power drop.

As previously discussed, there are several approaches that have been proposed to reduce

the effect of shadows on a solar PV array output power:

In large systems, each of the solar sub-modules can be connected to its own maximum

power point (MPP) tracking DC-DC converter and can individually operate near its own

126
MPP. Thus, the efficiency of the whole system is increased [42], but the method requires

a large number of DC-DC converters (equal to the number of solar modules).

In this chapter, we propose the new class of DC-DC converters which is to adaptively

reconfigure solar array connections in real time in order to track maximum output power.

Specifically, this chapter presents the following research contributions:

Developing new topology to continuously rearrange the two non-uniform illuminated

solar submodules in series and parallel connections by using high frequency switches to

facilitate the photovoltaic system to work more as a constant power source even in

different operating conditions (i.e. insolation, temperature, loads,…etc). The output of

solar PV array will be connected to the load through high frequency filter without MPPT.

The new topology is single-stage two-input DC-DC converter. In this topology the

voltage stresses of the main switches will be approximately reduced to a half of the input

voltage.

The difference between input and output voltage or current for each cycle will be reduced

so that the size of output inductor will be smaller in comparison with conventional buck

topology.

127
6.2 Solar PV array’s model and the conventional MPPT based
on Buck-converter topology

6.2.1 Conventional MPPT based on Buck converter

Fig. 6.1 shows the traditional solar PV array MPPT based on Buck converter topology

Fig. 6.1 Traditional solar PV array MPPT based on Buck converter topology

These are the most common topologies used as MPPT’s

• Down or Buck converters

• Up or Boost converters

• Up-Down or Buck-Boost converters

All these topologies can be derived from the basic canonical switching cell.

Fig. 6.1 shows the conventional solar PV array MPPT based on Buck converter. In this

topology, two solar PV sub-arrays PV1 and PV2 are connected in series.

128
a
br
b
c

a) I-V curve before configuration

b ) I-V curve after reconfiguration

129
 c) P-V curves

Fig. 6.2 The I-V and P-V curves before and after reconfiguration

Fig. 6.2 shows the I-V curves and P-V curves for two PV modules connected in series

before and after reconfiguration by the proposed method. In the operating lines b and c,

the non-shaded solar module PV1 cannot generate the full power because the current of

this module is limited by the current of the shaded module PV2. So, for example, on

operating line b, the non-shaded PV1 operates at I-V point Sb1 and the shaded PV2

operates at (I-V) point Sb2. In series, the total array will operate at their voltage sum,

while at the same fixed current, i.e. operating point Sb.

In the operating line a, the non-shaded module PV1 generate full power, but the operating

point of the shaded module PV2 moves to the negative region. The PV2 works as a load

and causes losses in the system.

130
The output power is decreased:

Pout = Pout1 − Plosses 2 (6.1)

From above discussion, the main reason to cause the losses in the solar PV array is that

the currents of the series connected solar cells or solar submodules are limited by the

current of the shaded solar cell or the shaded solar submodule.

After reconfiguration, the operating point of the solar PV array moved from the point Sb

in Fig. 6.2(a) to the point Sbmax in Fig. 6.2(b). The maximum output power of the whole

solar PV array increases from the point PSb to PSbmax in Fig. 6.2(c).

6.3 Proposed MPPT based on reconfiguration converter

In the proposed scheme, the solar submodules will be reconfigured from series

connection to parallel connection in each cycle. The output of solar PV array will be

changed from 2Vpv in ON time to VPV in OFF-time. The voltage stress of high frequency

switches will be a half of input voltage (VPV). In the OFF-time, the solar PV array does

both things: charge capacitor and supply current to the load.

Interval 1 (0-DT):

The condition of this period is shown in Fig. 6. 5. In this interval, S is turned ON: V32>0

and V14>0, the D1 and D2 are turned OFF. The subarray PV1 and PV2 are connected in

series. The current goes through: PV2, switch S, PV1, output filter (Lo and Co) and load.

131
 Fig. 6.3 Proposed MPPT based on Reconfiguration converter

Fig. 6.4 Two solar PV submodules are connected in series when S is turned ON

132
 Fig. 6.5 Two solar PV submodules are connected in parallel when S is turned OFF

The detail will be described below:

Interval 2 (DT-T)

As shown in Fig. 6.6, the S is turned OFF during this interval: D1 and D2 turn ON, PV1

and PV2 are connected in parallel. The current IL0 goes through the two solar parallel

solar PV1 and PV2 submodules, two diodes D1 and D2, the output filter (Lo and Co) and

load.

The voltage stresses of the main switches are shwon in Fig. 6. 7. They equal a half of

input voltage

VIN
VSW = = VPV
2 (6. 1)

133
 Fig.6.6 The voltage stresses of the main switches

6.4 The relationship between the input current and voltage with
the output current and voltage

In the steady state, the average voltage of input capacitor equals zero, so we have the

following equation:

< VC1 >= 0

T DT T
1 1 I0
∫
C1 0
I C1dt =
C1 ∫ ( I 0 − I IN )dt + ∫ ( I IN −
0 DT
2
)dt = 0 (6. 2)

2 I IN
I0 = (6. 3)
1+ D

Here, IIN- the output current of each solar submodule

134
 Fig. 6.7 The input voltage and output voltage relationship

If we assume that the converter operates in steady state, the energy stored in each

component at the end of a commutation cycle T is equal to that at the beginning of the

cycle. That means that the current IL is the same at t=0 and at t=T (see figure 4).

Therefore:

ΔI LON + ΔI LOFF = 0 (6. 4)

We have:

VIN − V0 V / 2 − V0
× DT + IN × (1 − D)T = 0 (6. 5)
L L

From (12) we have:

⎛1+ D ⎞
V0 = ⎜ ⎟VIN (6. 6)
⎝ 2 ⎠

135
6.5 Simulation result

The parameters of solar PV array (submodule) and the battery are shown in the Table.6.1.

TABLE 6.1 Simulation parameters of conventional and proposed configuration

No Item Symbol Unit Con. Proposed

1 Open circuit voltage VOC V 22 22

2 Short-circuit current of first ISC1 A 3 3

submodule

3 Short-circuit current of ISC2 A 1 1

second submodule

4 Series resistance RS Ohm 0.01 0.01

5 Parallel resistant RSH Ohm 50 50

6 Battery voltage VBAT V 12 12

7 Battery resistance RBAT Ohm 0.5 0.5

8 Switching frequency FSW kHz 50 50

9 Filter inductor L uH 75 75

10 Input capacitor CIN uF 100 100x2

11 Output capacitor COUT uF 50 50

In the first simulation schematic, the two solar PV submodules are connected in parallel.

The configuration is fixed. The solar PV array charges the battery through the buck

converters.

136
In the second simulation schematic, the two solar PV submodules are reconfigurable,

connected by the proposed topology. The solar PV array charges the battery through the

buck converters.
30V 30V

20V 20V

10V 10V

0V 0V
300us 320us 340us 360us 380us 400us 300us 320us 340us 360us 380us 400us
V(M2:s,M2:d) V(M6:d,M6:s)
Time Time

a) b)

Fig. 6. 8 The voltage stress of transistor in a) buck converter and b) Reconfiguration converter

40V 40V

20V 20V

0V
0V

-20V
-20V 300us 320us 340us 360us 380us 400us
300us 320us 340us 360us 380us 400us V(D16:2,D16:1)
V(D15:2,D15:1) Time
Time

a) b)

Fig. 6.9 The voltage stress of main diodes in a) Buck converter and b) Reconfiguration converter

137
 3.0A 4.0A

2.0A 3.0A

1.0A 2.0A

0A 1.0A
300us 320us 340us 360us 380us 400us 300us 320us 340us 360us 380us 400us
I(L1) I(L1)
Time Time

a) b)

Fig. 6. 10 The inductor current ripple in a) Buck converter and b) Reconfiguration converter

21.500V
10.500V

21.375V
10.375V

21.250V 10.250V

21.125V 10.125V

10.000V
21.000V 700us 720us 740us 760us 780us 800us
300us 320us 340us 360us 380us 400us V(C7:2,C7:1)
V(C5:2,C5:1) Time
Time

a) b)

Fig. 6.11The input capacitor voltage ripple a) Buck converter with one C=100uF, Vin b)
Reconfiguration converter with two C=100uF, Vin/2

138
6.6 Conclusion

A new topology of the reconfiguration converter of solar PV arrays under shadow

conditions is described.

The voltage stress is reduced to a half of input voltage. The size of the energy storage

elements is reduced without using more switches.

One disadvantage, may be, the second solar PV array is switched between ground voltage

and output voltage of first solar PV array that can be the EMI source.

This thesis only analytically proposes then simulates the new topology. Simulations

verify the principles, but to fully understand the capabilities of the approach, experiments

should be performed ( see Chapter 7 for future research)

139
 7. Conclusions

7.1 General description

With the mass production and use of solar PV energy, the new goal for solar PV research

field is to have the breakthrough cost for solar energy. The optimized problem for

“maximum output power” and “minimum cost of energy” creates new challenges for

solar PV system research. With unlimited space, the solar PV array can be installed in

locations without worries from shading objects: trees, booms… and the collectors in solar

field installed with the large distance between the rows. With economy calculation, the

solar arrays are forced to work in the non-uniform illumination to increase the gross solar

photovoltaic area in the limited land, roof area.

In uniform illumination, the large solar PV array is a simple system of “scaled- up” solar

cells: the output voltage is equal the voltage of solar cell, multiplied by the number of

solar cell in series. The current equals the current of solar cell, multiplies the number of

parallel solar cells.

In the non-uniform illumination, the picture is changed. The solar array become the

system of mxn solar cells, each solar cell is presented by a source with non-linear

characteristics. Although the solar cells may have identical parameters, they still create

mismatch losses when they are connected in series/parallel. Because of nonlinear

characteristics in the solar cells, non-uniform illumination dramatically reduces the

output power of solar PV system. The maximum output power losses (%) are much larger

140
than the shading factors. The shadows move and change their shape with time. The

modeling the solar PV system under dynamic change of illumination is more difficult

task.

The shaded cell, in the worst case, becomes the resistor that consumes energy instead of

supplying. Because the breakdown voltage of solar cell is large, the shaded cell can create

the local hot spot caused damage solar cells. The need of accurate model to analyze and

simulate the solar system under shadow conditions is apparent.

Under a passing cloud, the output power of solar array is reduced greatly. In utility grid

connected PV systems, the fast reduction in supply power can cause instability of the

whole system. This requires the modeling and computing algorithm that simulates and

analyzes the effect of non-uniform changing shadows (i.e. a passing cloud) on the output

power of solar PV arrays. The software must also be fast enough to be used within real

time controllers.

This thesis, in Chapter 2, proposes the software implementation in MATLAB that meets

the above requirements. The model also model is flexible enough to simulate solar PV

arrays with different scale, various configurations, with or without bypass diode, from

different manufacturers.

141
When the parameters of solar system are changed with the time and environmental

conditions, such as temperature, wind speed, humidity…The model is easily updates new

parameters in real time, so that simulation of solar PV system is closer to the real system.

The model is able to simulate and detect the local hot spots of solar PV system that is

caused by mismatch of solar cells parameters or non uniform illuminations of solar

insolation. The local hot spots of solar PV systems are the most reason that causes

damage the solar cells or solar submodules or even whole systems. The solar cells under

shaded conditions act like loads. They dissipate heat and the semiconductors junctions

may be damaged. When the solar cells become fully illuminated, they cannot generate the

full power as designed and block the current of the strings subsequently reducing system

output power. The solar systems under non-uniform illumination may have much shorter

lifetime than normal systems. The proposed model in Chapter 2 can simulate the power

dissipation on each shaded solar cell which help define the solar cell temperature is in

range of normal working condition or not.

Chapter 3 proposes another approach to simulate and analyze the solar PV system

performance based on neural networks. The performance of a solar PV system under any

working conditions, uniform illumination or non-uniform illumination, can be predicted

using acquired data for short term. Then the collected data are used to generate the

functions to characterize the relationship between the performance of the solar PV system

and the influence of environmental factors. The main advantages of the proposed neural

142
network method are that it is possible to simulate and analyze the solar PV system

performance, without knowing the parameters of solar cells, or the system configuration.

Chapter 3, uses the solar irradiation levels, the solar photovoltaic array’s temperature, and

the sun’s position angles as the input signals, and the maximum output power of the solar

PV array as an output signal. The maximum output power or maximum output power

losses of the shaded solar PV array can be calculated based on shading function without a

shading factor calculation. A neural network structure is proposed, combining the

multilayer perceptrons feed-forward network with a back-propagation algorithm that can

accurately simulate maximum output power of solar photovoltaic arrays under shadow

conditions.

The detailed procedures to collect the training data for the model and to check the

accuracy and generalization of the network are described. By using only readily available

solar irradiation data at different times of a day, wherever the solar PV array is installed

the proposed approach can be employed to accurately predict the maximum output power

of the solar PV array over long periods of time with low computational efforts. The

disadvantage of proposed neural network method is the collected data needs to be cleaned

and checked their possible correlation before input to the neural network simulation tools.

Chapter 4 of this thesis proposes a practical reconfigurable solar array. The previous

approaches of reconfigurable solar PV array are unrealistic because of large number of

required sensors, switches and complicated control algorithm.

143
This thesis proposes a reconfiguration method of solar PV array that requires fewer

switches and simplified control algorithms by choosing the appropriate configuration of

main fixed part of solar array and reconfiguring only a small percentage of the solar

arrays. The reconfiguration is executed through the proposed switching matrix. In

uniform illumination, both fixed and adaptive parts are equally connected in parallel so

they can generate full power to the load. In non-uniform illumination, the number of the

adaptive solar cells connected to the shaded submodules depends on the shaded area of

submodules. They create the paths for currents in the non-shaded solar cells in the same

string and bypass the shaded cells. The output power will just be reduced by the amount

proportional to the shaded area. Two simple control decision algorithms are proposed to

determine when and how the switches between the fixed part and adaptive bank of the

solar PV array are open and closed. The algorithms will rely on predictive models that

can be implemented in real-time by a micro-controllers or DSP. To our knowledge, it is

the first time that such adaptive and implement able algorithms have been reported. For

large solar PV system, we propose using the switching matrix the electromechanical

relays to increase the current going through panels. The proposed reconfiguration

method helps avoid the local maximum power points as well as the local hotspots of solar

PV arrays, thus the central maximum power point tracker can be used to track the

maximum output power of the whole solar PV array.

In the proposed reconfiguration method, to define the shadow pattern of large solar PV

system, we propose the algorithm that use only minimum number of voltage sensors

144
(two). This algorithm can be used for monitoring shaded solar PV system in other

applications.

In Chapter 5, a small test experimented platform was created to validate the proposed

solar PV array reconfiguration methods and control algorithms. The experiment includes

the following: The fixed solar array is from 9 solar cells, 3 rows (m) and 3 columns (n)

(TCT connection). The solar adaptive bank combine from 3 solar cells (one string of 3 x1

cells) connected to the fixed part (3x3). The (3x3) switching matrix, controlled by

Agilent Data Acquisition/Switch Unit 34970A, is connected between the fixed part and

the solar adaptive bank. In total, we use 18 switches in the switching matrix and 6 voltage

sensors (two of them are voltage monitors) for this test platform. The experiments

demonstrate the technical feasibility of both control algorithms. The model-based

algorithm have reduced calculation time and show advantage over bubble sort method

when applying for large solar PV system because in model based method all solar cells in

the adaptive bank are reconfigured synchronously.

Four experiments with different possible shadow patterns are demonstrated and shown to

be effective in improving solar array power, even when the number of shaded arrays is

more than the power provided by the adaptive solar modules to be switched into the

system.

145
In the next section, the future work of this thesis is to build new topology of the

reconfiguration converter of solar PV arrays under shadow conditions. From first

investigation and analyzing of new topology, there may be many possible advantages:

• The voltage stress is reduced to a half of input voltage.

• The size of the energy storage elements is reduced without using more switches.

• The total output power may be higher than power output of shaded solar array

with only one central dc-dc converters.

• One disadvantage, may be, the second solar PV array is switched between ground

voltage and output voltage of first solar PV array that can be the EMI source.

7.2 Future research: Design of proposed reconfiguration

converter

To build a prototype to verify the advantage of the proposed single stage two-input dc/dc

converters for tracking MPP of solar PV array

The power source from a solar panel GP 3-30 is considered a "current-limited" voltage

source. The solar panel is divided in two parts, each part has the same number of solar

cells (6 solar cells) connected in series. Before the design we firstly set the input DC

voltage range. We define

• VPVMAX as the maximum input voltage, VPVMIN as the minimum input DC

voltage: VPVMAX =16V; VPVMIN =8 V; VOUT =12V

• Input power: PPVMAX =15W; PINMAX =30W

• Maximum output current of solar PV array: I OUT =2.3A

146
The maximum power from the solar panel to charge a12V Li-Ion battery can be achieved

by regulating the system bus voltage around the MPP through charge current reduction

when the total current demand from the system and battery charging exceeds the output

current capability from the solar panel. System power and battery charging power control

architecture are critical elements for designing a reliable solar panel powered system.

The switching frequency is 50 kHz, switching period is 20 μs . The duty ratio: D=0.5

Main switches:

Transistor S, the voltage stress is VPVMAX =8V and maximum current is 20V, 2A,

MTM862270LBFCT-ND

Rectifier diodes: D1-D2, maximum current is 1.2 A, MURA105T3, 2A, 50V.

7.2.1 Input capacitor design for proposed converter. The comparison

with conventional Buck converter

For the Buck converter:

The minimum size of input filter capacitors is defined by:

I IN (1 − D)T
C≥ (7.1)
Δv IN

For the proposed converter:

Current relation: With ripple frequency, the capacitor is assumed to be sorted, the average

voltage of capacitor equal zero. We have:

dVC1
I C1 = C1 (7.2)
dt

147
The minimum size of capacitor that ripple of the input voltage (Vpv1) less than

requirement:

I IN × D × (1 − D)T
C1 ≥ (7.3)
(1 + D) × Δv PVM

The comparison the size of capacitors of the both schemes:

If IIN=0.75 IOUT, D=0.75 for buck converter:

I IN (1 − D)T I IN × 0.25
C≥ = T (7.4)
Δv IN Δv IN

For proposed converter:

D=0.5

I IN × D × (1 − D)T I × 0.25
C1 ≥ = IN T (7.5)
(1 + D) × Δv PVM 1.5 × Δv PVM

C MIN × Δv IN
C1MIN = (7.6)
1.5 × Δv PVM

Input capacitor: C1 =C2=100 uF, 135D506X9025F2CAP, WET TANT 50UF 10% 60V

AXIAL

7.2.2 Output inductor design:

A rule of thumb, which we will use later for selecting a value of inductance in an LC

filter, is to start with an inductor value that results in a peak-to-peak inductor current that

148
is 10% of the full load current. For our example of a 12 V output, 2.4 A load, and 50 kHz

switching frequency, the inductor can be calculated from the formula V = L*(di/dt).

For the Buck converter:

The energy storage elements are sized by the following equation:

The minimum size of output inductor is defined by:

VOUT (1 − D)T
LBuck ≥ (7.7)
Δi L

For the proposed converter:

The minimum size of output inductor is defined by:

(VOUT − VIN / 2)(1 − D)T

LP ≥ (6. 7)
Δi L

Comparison:

If for both converter VOUT=0.75 VIN, with the same current ripple requirement:

2 LBUCKMIN
LPMIN = (7.8)
3

Output inductor: 0.5 mH, LHL13NB101K-ND, INDUCTOR 100UH 2.0A RADIAL

7.2.3 Output capacitor design

Both buck and proposed converter have the same equation for output filter capacitor

design:

149
 I OUT (1 − D)T
COUT ≥ (7.9)
ΔvOUT

Output capacitor: 135D506X9025F2CAP, 50UF 10% 60V AXIAL

150
APPENDIX A MATLAB source code for modeling solar PV
arrays under shadow conditions
MAIN
function [P] = main(Isc)
Isc=Isc
% %Short circuit current
[y,x] = size(Isc);
m = (x-1)*y;
n = x*(y-1);
z = m+n+1;
z1 = x*y; %indexes for VA
z2 = (x-1)*(y-1); %indexes for VB
z3 = 1; %indexes for I
V =20;
I= 14;
%VA =2.1*ones(y,x);
VA =2.1*ones(y,x);
VB = zeros(y-1,x-1);
Res = 0.01*ones(y-1,x-1);
u(1:z1,1) = reshape(VA',1,z1)
u(z1+1:z1+z2,1) = reshape(VB',1,z2)
u(z1+z2+z3,1) = I
%Algorithm
%start with u0
epsilon = 1e-4; Error = 1;
while (Error > epsilon)
dU = Jacobian(u, Res, Isc);
[ru] = r_f(u, V, Isc, Res);
u = u - 0.04*inv(dU)*ru
Error = norm(inv(dU)*ru);
end
Cout=u(z1+z2+z3,1);
%Power output
P=Cout*V
CLOUD SIMULATION
[X,Y] = meshgrid(0:1:9,0:1:9);
for i=1:10
R = sqrt((X-(i-1)).^2 + (Y-(i-1)).^2) + eps
Z = (0.5 + sin(R)./(1+R))
figure
Xlabel('m')
Ylabel('n')
Zlabel('Irradiation (Isc/Isc0)')
mesh(X,Y,Z,'EdgeColor','black')
surf(X,Y,Z)
colormap hsv
colorbar
az = -132;
el = 7;
view(az, el);

151
end
CURRENT CALCULTAION
function [C] = Curr(Isc, V)
q=1.6022e-19;
a=6.55;
k=1.3806e-23;
T=300;
I0=3.403e-07;
Rsh=53;
Rs=0
%Open circuit output voltage
mVT=(a*k*T)/q
Voc = mVT * log(Isc/I0+1)
b = q*Voc/(a*k*T);
sigma = Voc/(Rsh*Isc);
bbd = b;
Iodb = 15e-13;
Vt=0.026;
C = Isc - ( Isc*( (1-sigma)/(exp(b) -1))*( exp(b*V/Voc)-1) ) - (Isc*V/Voc*sigma)
function [dI] = dIdV(Isc, V)
q=1.6022e-19;
a=6.55;
k=1.3806e-23;
T=300;
I0=3.403e-07;
Rsh=53;
Rs=0
%Open circuit output voltage
mVT=(a*k*T)/q;
Voc = mVT * log(Isc/I0+1);
b = q*Voc/(a*k*T);
sigma = Voc/(Rsh*Isc);
bbd = b;
Iodb = 15e-13;
Vt=0.026;
dI = -(Isc * (b/Voc) * (1-sigma)/(exp(b)-1) * ( exp(b*V/Voc) ) ) - (sigma*Isc/Voc)
function [dI] = dIdV_B(R, V)
dI = 1/R;
JACOBIAN MATRIX
%Return the Jacobian matrix of r(u)
function [dU] = Jacobian(u, Res, Isc)

[y, x] = size(Isc);
m = (x-1)*y;
n = x*(y-1);
z = m+n+1;

dU = zeros(z+1,z+1);

z1 = x*y; %indexes for VA

z2 = (x-1)*(y-1); %indexes for VB

152
z3 = 1; %indexes for I

% for i = 1:(y)
% for j = 1:(x)
% VA(i,j) = u( (i-1)*x+j );
% IA(i,j) = Curr(Isc(i,j), VA(i,j));
% end
% end
for i = 1:(y)
for j = 1:(x-2)
VA(i,j) = u( (i-1)*x+j );
IA(i,j) = Curr(Isc(i,j), VA(i,j));
end
end

for i =1:y
for j=(x-1):x
VA(i,j) = u( (i-1)*x+j );
IA(i,j) = Curr1(Isc(i,j), VA(i,j));
end
end

for i = 1:(y-1)
for j = 1:(x-1)
VB(i,j) = u( (i-1)*(x-1)+j +z1 );
IB(i,j) = VB(i,j)/Res(i,j);
end
end

I = u(z1+z2+z3);

%ru(1) = sum(VA(:,1)) - V;
for i = 1:y
dU(1,1+x*(i-1)) = 1;
end

%for loops on the 1st row

i = 1;
for j = 1:(x-1)
index = (i-1)*(x-1) + (j-1) + 2;
%ru(index) = VA(i,j) - VB(i,j) - VA(i,j+1);
dU(index, (i-1)*(x)+j ) = 1;
dU(index, z1 + (i-1)*(x-1)+j ) = -1;
dU(index, (i-1)*(x)+j+1 ) = -1;
end

%for middle loops

for i = 2:(y-1)
for j = 1:(x-1)
index = (i-1)*(x-1) + (j-1) + 2;
%ru(index) = VA(i,j) - VB(i,j) - VA(i,j+1) + VB(i-1,j);

153
 dU(index, (i-1)*(x)+j ) = 1;
dU(index, z1 + (i-1)*(x-1)+j ) = -1;
dU(index, (i-1)*(x)+j+1 ) = -1;
dU(index, z1 + (i-2)*(x-1)+j ) = 1;
end
end
%for loops on the last row
i = y;
for j = 1:(x-1)
index = (i-1)*(x-1) + (j-1) + 2;
%ru(index) = VA(i,j) - VA(i,j+1) + VB(i-1,j);
dU(index, (i-1)*(x)+j ) = 1;
dU(index, (i-1)*(x)+j+1 ) = -1;
dU(index, z1 + (i-2)*(x-1)+j ) = 1;
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%
%ru(index+1) = sum(IA(1,:)) - I;
for j=1:x
dU(index+1, j) = dIdV(Isc(1,j), VA(1,j));
end
dU(index+1, z+1) = -1;

start = index+1;
%for nodes in the 1st column
for i = 1:(y-1)
j=1;
index = start + (i-1)*x + (j-1) + 1;
%ru(index) = IA(i,j) + IB(i,j) - IA(i+1,j);
dU(index, (i-1)*(x)+j ) = dIdV(Isc(i,j), VA(i,j));
dU(index, z1 + (i-1)*(x-1)+j ) = dIdV_B(Res(i,j), VB(i,j));
dU(index, (i)*(x)+j ) = -dIdV(Isc(i+1,j), VA(i+1,j));
end
%for middle nodes
% for i = 1:(y-1)
% for j = 2:(x-1)
% index = start + (i-1)*x + (j-1) + 1;
% %ru(index) = IA(i,j) + IB(i,j) - IA(i+1,j) - IB(i,j-1);
% dU(index, (i-1)*x+j ) = dIdV(Isc(i,j), VA(i,j));
% dU(index, z1 + (i-1)*(x-1)+j ) = dIdV_B(Res(i,j), VB(i,j));
% dU(index, (i)*x+j ) = -dIdV(Isc(i+1,j), VA(i+1,j));
% dU(index, z1 + (i-1)*(x-1)+j-1 ) = -dIdV_B(Res(i,j-1), VB(i,j-1));
% end
% end
%Unidentical
%--------------------------------------------
for i = 1:(y-1)
for j = 2:(x-2)
index = start + (i-1)*x + (j-1) + 1;
%ru(index) = IA(i,j) + IB(i,j) - IA(i+1,j) - IB(i,j-1);

154
 dU(index, (i-1)*x+j ) = dIdV(Isc(i,j), VA(i,j));
dU(index, z1 + (i-1)*(x-1)+j ) = dIdV_B(Res(i,j), VB(i,j));
dU(index, (i)*x+j ) = -dIdV(Isc(i+1,j), VA(i+1,j));
dU(index, z1 + (i-1)*(x-1)+j-1 ) = -dIdV_B(Res(i,j-1), VB(i,j-1));
end
end
for i = 1:(y-1)
for j = (x-2):(x-1)
index = start + (i-1)*x + (j-1) + 1;
%ru(index) = IA(i,j) + IB(i,j) - IA(i+1,j) - IB(i,j-1);
dU(index, (i-1)*x+j ) = dIdV1(Isc(i,j), VA(i,j));
dU(index, z1 + (i-1)*(x-1)+j ) = dIdV_B(Res(i,j), VB(i,j));
dU(index, (i)*x+j ) = -dIdV(Isc(i+1,j), VA(i+1,j));
dU(index, z1 + (i-1)*(x-1)+j-1 ) = -dIdV_B(Res(i,j-1), VB(i,j-1));
end
end
%----------------------------------------------------------------------------

%for nodes in the last column

% for i = 1:(y-1)
% j = x;
% index = start + (i-1)*x + (j-1) + 1;
% %ru(index) = IA(i,j) - IA(i+1,j) - IB(i,j-1);
% dU(index, (i-1)*x+j ) = dIdV(Isc(i,j), VA(i,j));
% dU(index, (i)*x+j ) = -dIdV(Isc(i+1,j), VA(i+1,j));
% dU(index, z1 + (i-1)*(x-1)+j-1 ) = -dIdV_B(Res(i,j-1), VB(i,j-1));
% end
%Unidentical
%---------------------------------------------------------------------------------
for i = 1:(y-1)
j = x;
index = start + (i-1)*x + (j-1) + 1;
%ru(index) = IA(i,j) - IA(i+1,j) - IB(i,j-1);
dU(index, (i-1)*x+j ) = dIdV1(Isc(i,j), VA(i,j));
dU(index, (i)*x+j ) = -dIdV1(Isc(i+1,j), VA(i+1,j));
dU(index, z1 + (i-1)*(x-1)+j-1 ) = -dIdV_B(Res(i,j-1), VB(i,j-1));
end
%Return the Jacobian matrix of r(u, V, Isc). Isc is a vector
%u = [VA, VB, I];

function [ru] = r_f(u, V, Isc, Res)

[y,x] = size(Isc);
z1 = x*y; %indexes for VA
z2 = (x-1)*(y-1); %indexes for VB
z3 = 1; %indexes for I

for i = 1:(y)
for j = 1:(x-2)
VA(i,j) = u( (i-1)*x+j );
IA(i,j) = Curr(Isc(i,j), VA(i,j));

155
 end
end

for i =1:y
for j=(x-1):x
VA(i,j) = u( (i-1)*x+j );
IA(i,j) = Curr1(Isc(i,j), VA(i,j));
end
end

for i = 1:(y-1)
for j = 1:(x-1)
VB(i,j) = u( (i-1)*(x-1)+j +z1 );
IB(i,j) = VB(i,j)/Res(i,j);
end
end

I = u(z1+z2+z3);

ru(1,1) = sum(VA(:,1)) - V;

%for loops on the 1st row

i = 1;
for j = 1:(x-1)
index = (i-1)*(x-1) + (j-1) + 2;
ru(index,1) = VA(i,j) - VB(i,j) - VA(i,j+1)
end
%for middle loops
for i = 2:(y-1)
for j = 1:(x-1)
index = (i-1)*(x-1) + (j-1) + 2;
ru(index,1) = VA(i,j) - VB(i,j) - VA(i,j+1) + VB(i-1,j)
end
end
%for loops on the last row
i = y;
for j = 1:(x-1)
index = (i-1)*(x-1) + (j-1) + 2;
ru(index,1) = VA(i,j) - VA(i,j+1) + VB(i-1,j)
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%
%index - 1 - (x-1)*y
ru(index+1,1) = sum(IA(1,:)) - I
start = index+1; %numbers of equations r(u) before node equations

%for nodes in the 1st column

for i = 1:(y-1)
j=1;

156
 index = start + (i-1)*x + (j-1) + 1;
ru(index,1) = IA(i,j) + IB(i,j) - IA(i+1,j)
end
%for middle nodes
for i = 1:(y-1)
for j = 2:(x-1)
index = start + (i-1)*x + (j-1) + 1;
ru(index,1) = IA(i,j) + IB(i,j) - IA(i+1,j) - IB(i,j-1)
end
end
%for nodes in the last column
for i = 1:(y-1)
j = x;
index = start + (i-1)*x + (j-1) + 1;
ru(index,1) = IA(i,j) - IA(i+1,j) - IB(i,j-1)
end

157
APPENDIX B MATLAB code for modeling solar PV array
using neural network method

Input data file

load Irradiance_Time
[Poutmax]= xlsread('NN_05_07','Poutmax')
Isc=Isc;
Ishadow=Ishadow;
Vout=Vout;
Time=[5/3/2007 11:52….]
Main file

location.longitude = -71.09177;
location.latitude = 42.34153;
location.altitude = 10;
[m,n]=size(Time)
for i=1:m
time.year = Year(i);
time.month = Month(i);
time.day = Day(i);
time.hour = Hour(i);
time.min = Minute(i);
time.sec = Second(i);
time.UTC = -4;
sun = sun_position(time, location);
z(i)=sun;
end
load Time_Isc_Poutmax
clear x1 x2 x3 y;
[m,n]=size(Time)
for i=1:m
x1(i)=z(i).zenith;
x2(i)=z(i).azimuth;
x3(i)=Isc(i);
x4(i)=Ktemp(i);
y(i)=Poutmax(i);
end
X = [x2; x3; y]
save allresults.csv, X'

p = [x2; x3; x4];

t= [y];

%Step 1: Normalized the inputs and target

[n,meanp,stdp,tn,meant,stdt] = prestd(p,t)
[pn,meanp,stdp] = prestd(p);
[ptrans,transMat] = prepca(pn,0.001);
%Step 2: Chek size of transformed data

158
[R,Q] = size(ptrans)
[ptrans,tn]= randomize(ptrans, tn)
%___________________________________________
% [R,Q] = size(p)
% [ptrans,tn]= randomize(p,t)

% Step3:The next step is to divide the data up into training

% and test subsets. We will take one fourth of the data for the
validation set,
% one fourth for the test set and one half for the training set. We
pick the sets as equally spaced points throughout the original data.
Trnumber=round(Q*0.7);
[ptr,ttr, ptest, ttest]=randpick(ptrans,tn,Trnumber,t)
%[ptr,ttr, ptest, ttest]=randpick(p,t,Trnumber,t)
% iitst =[ 2:4:Q];
% iitr = [1:4:Q 3:4:Q 4:4:Q];
% test.P = ptrans(:,iitst); test.T = tn(:,iitst);
% ptr = ptrans(:,iitr); ttr = tn(:,iitr);
% test2=ptrans(:,iitst);

%Trainng net
net=newff(minmax(ptr),[3 1],{'tansig','purelin'},'trainlm');
[net,tr]=train(net,ptr,ttr,[],[]);
net.performFcn = 'msereg';
net.performParam.ratio = 1;
net.trainParam.show = 5;
net.trainParam.epochs =300;
net.trainParam.goal = 2e-2;

% net.performFcn = 'msereg';
% % net.performParam.ratio = 0.5;
% % net.trainParam.show = 5;
% % net.trainParam.epochs = 200;
% % net.trainParam.goal = 1e-5;
% net.trainParam.show = 5;
% net.trainParam.lr = 0.05;
% net.trainParam.epochs = 235;
% net.trainParam.goal = 1e-10;
%Test
plot(tr.epoch,tr.perf)
legend('Training',-1);
ylabel('Squared Error'); xlabel('Epoch')

% plot(tr.epoch,tr.perf,tr.epoch,tr.vperf)
% legend('Training','Test',-1);
% ylabel('Squared Error'); xlabel('Epoch')
% figure
% %plot(tr.epoch,tr.tperf)
% plot(tr.epoch,tr.perf)
% legend('Training','Test',-1);
% ylabel('Squared Error'); xlabel('Epoch')
% plot(tr.epoch,tr.perf,tr.epoch,tr.tperf)
% legend('Training','Test',-1);
% ylabel('Squared Error'); xlabel('Epoch')

% Test data set

159
% an = sim(net,ptrans);
a = sim(net,ptest);
% a = poststd(an,meant,stdt);

figure
[m,b,r] = postreg(a,ttest);
ylabel('Predicted power(W)'); xlabel('Measured power(W)')

Sun position

function sun = sun_position(time, location)

% sun = sun_position(time, location)
%
% This function compute the sun position (zenith and azimuth angle at
the observer
% location) as a function of the observer local time and position.
%
% It is an implementation of the algorithm presented by Reda et Andreas
in:
% Reda, I., Andreas, A. (2003) Solar position algorithm for solar
% radiation application. National Renewable Energy Laboratory (NREL)
% Technical report NREL/TP-560-34302.
% This document is avalaible at www.osti.gov/bridge
%
% This algorithm is based on numerical approximation of the exact
equations.
% The authors of the original paper state that this algorithm should be
% precise at +/- 0.0003 degrees. I have compared it to NOAA solar table
% (http://www.srrb.noaa.gov/highlights/sunrise/azel.html) and to USNO
solar
% table (http://aa.usno.navy.mil/data/docs/AltAz.html) and found very
good
% correspondance (up to the precision of those tables), except for
large
% zenith angle, where the refraction by the atmosphere is significant
% (difference of about 1 degree). Note that in this code the correction
% for refraction in the atmosphere as been implemented for a
temperature
% of 10C (283 kelvins) and a pressure of 1010 mbar. See the subfunction
% «sun_topocentric_zenith_angle_calculation» for a possible
modification
% to explicitely model the effect of temperature and pressure as
describe
% in Reda & Andreas (2003).
%
% Input parameters:
% time: a structure that specify the time when the sun position is
% calculated.
% time.year: year. Valid for [-2000, 6000]
% time.month: month [1-12]
% time.day: calendar day [1-31]

160
% time.hour: local hour [0-23]
% time.min: minute [0-59]
% time.sec: second [0-59]
% time.UTC: offset hour from UTC. Local time = Greenwich time +
time.UTC
% This input can also be passed using the Matlab time format ('dd-
mmm-yyyy HH:MM:SS').
% In that case, the time has to be specified as UTC time (time.UTC =
0)
%
% location: a structure that specify the location of the observer
% location.latitude: latitude (in degrees, north of equator is
% positive)
% location.longitude: longitude (in degrees, positive for east of
% Greenwich)
% location.altitude: altitude above mean sea level (in meters)
%
% Output parameters
% sun: a structure with the calculated sun position
% sun.zenith = zenith angle in degrees (angle from the vertical)
% sun.azimuth = azimuth angle in degrees, eastward from the
north.
% Only the sun zenith and azimuth angles are returned as output, but a
lot
% of other parameters are calculated that could also extracted as
output of
% this function.

161
APPENDIX C MATLAB code for control solar PV array
reconfiguration
function []=main2()
%Step1: Origin_State
%[Org]=Origin_State2()
%Step2:Check shadow ratio
% Start control
cont=0
for k=1:5
[Vmin,Vmax]=StartDC2()
range=0.03*Vmax;
[Y1,Z1]=Scan()
if Vmin<range
% Step3: Define the photogenerated currents
% Step4:
%Disconnect fixed and added parts
[Open]=Open_State()
SW=zeros(3,3);
for i=1:3
%a) Measure
[Vfix]=StartDC1()
[SW]=Switches1(Vfix,i,SW)
% b) Adding: Switching command
[T]=Switches(SW)
[Y2,Z2]=Scan()
end
[Y5,Z5]=Scan()
else
end
end
% Step5:Check shadow ratio after reconfig
[Y6,Z6]=Scan

function [SW]=Switches1(Vfix,i,SW)
SW=SW
[A1,indexV]=sort(Vfix,'ascend');
m=i;
n=indexV(1);
SW(m,n)=1;
function [Iadd, Ifixed,Iout]=getVDC2(RA,RF,R)
s = serial('COM1');
set(s,'BaudRate',9600)
get(s,{'BaudRate','DataBits','Parity','StopBits'})
set(s,'Timeout',60); % to avoid long waiting when there is a problem...
fopen(s);
fprintf(s,'ABOR'); % abort running scan
query(s,'R?'); % delete data from the buffer
fprintf(s,'ROUT:OPEN (@ 311:338)');% Open all channels
fprintf(s,'ROUT:SCAN (@101:109)') ; % set the channels to be scanned
% Measure the Iadd

162
 %IA1
fprintf(s,'ROUT:CLOS (@ 341)')
fprintf(s,'MEAS:VOLT:DC? (@101)')
fprintf(s,'ROUT:OPEN (@ 341)')
%IA2
fprintf(s,'ROUT:CLOS (@ 342)')
fprintf(s,'MEAS:VOLT:DC? (@101)')
fprintf(s,'ROUT:OPEN (@ 342)')
%IA3
fprintf(s,'ROUT:CLOS (@ 343)')
fprintf(s,'MEAS:VOLT:DC? (@101)')
fprintf(s,'ROUT:OPEN (@ 343)')
% Measure the Ifix
%If1
fprintf(s,'ROUT:CLOS (@ 341)')
fprintf(s,'MEAS:VOLT:DC? (@101)')
fprintf(s,'ROUT:OPEN (@ 341)')
%If2
fprintf(s,'ROUT:CLOS (@ 342)')
fprintf(s,'MEAS:VOLT:DC? (@101)')
fprintf(s,'ROUT:OPEN (@ 342)')
%If3
fprintf(s,'ROUT:CLOS (@ 343)')
fprintf(s,'MEAS:VOLT:DC? (@101)')
fprintf(s,'ROUT:OPEN (@ 343)')
%--------------------------------------------------
x=str2num(fscanf(s))
%Sense resistors
Iadd=x(1,1:3)./RA
Ifixed=Iout .+ x(1,4:6)./RF
% Chanel 107 measure the current
% Chanel 108 measure the total output voltage
Iout=x(1,8)/R
fclose(s)
function [Y]=Monitor2()
s = serial('COM1');
set(s,'BaudRate',9600)
get(s,{'BaudRate','DataBits','Parity','StopBits'})
set(s,'Timeout',60); % to avoid long waiting when there is a problem...
fopen(s);
fprintf(s,'ABOR'); % abort running scan
query(s,'R?'); % delete data from the buffer
for i=1:20
fprintf(s,'MON:VOLT:DC? (@107)')
y(i)=str2num(fscanf(s));
end
Y=y
fclose(s)
% Step 1: Original state
% Each cell connects to each row
function [Open]=Open_State()

163
s = serial('COM1');
set(s,'BaudRate',9600)
get(s,{'BaudRate','DataBits','Parity','StopBits'})
set(s,'Timeout',60); % to avoid long waiting when there is a problem...
fopen(s);
fprintf(s,'ABOR'); % abort running scan
query(s,'R?'); % delete data from the buffer
fprintf(s,'ROUT:OPEN (@311:338)')
Open=1
fclose(s)

% Step 1: Original state

% Each cell connects to each row
function [Org]=Origin_State2()
s = serial('COM1');
set(s,'BaudRate',9600)
get(s,{'BaudRate','DataBits','Parity','StopBits'})
set(s,'Timeout',60); % to avoid long waiting when there is a problem...
fopen(s);
fprintf(s,'ABOR'); % abort running scan
query(s,'R?'); % delete data from the buffer
fprintf(s,'ROUT:OPEN (@311:338)')
fprintf(s,'ROUT:CLOS (@313;324;335)')%To origin state
Org=1
fclose(s)
function [Vout,I]=Scan()
s = serial('COM1');
set(s,'BaudRate',9600)
get(s,{'BaudRate','DataBits','Parity','StopBits'})
set(s,'Timeout',60); % to avoid long waiting when there is a problem...
fopen(s);
fprintf(s,'ABOR'); % abort running scan
query(s,'R?'); % delete data from the buffer
% fprintf(s,'TOT:CLE:IMM (@101:108)');
fprintf(s,'ROUT:SCAN (@101:108)') ;
fprintf(s,'ROUT:MON (@107)');
fprintf(s,'ROUT:MON:STATE ON');
%Enable monitoring
for i=1:10
fprintf(s,'ROUT:MON:DATA?');
Vout(i)=str2num(fscanf(s));
end
fprintf(s,'ROUT:MON (@108)');
fprintf(s,'ROUT:MON:STATE ON ')
%Enable monitoring
for i=1:10
fprintf(s,'ROUT:MON:DATA?');
I(i)=str2num(fscanf(s));
end
Vout=Vout;
I=I;

164
 fclose(s)
%Step2:Check shadow ratio
function [V]=StartDC1()
s = serial('COM1');
get(s,{'BaudRate','DataBits','Parity','StopBits'})
set(s,'Timeout',60); % to avoid long waiting when there is a problem...
fopen(s);
fprintf(s,'ABOR'); % abort running scan
query(s,'R?'); % delete data from the buffer
%fprintf(s,'ROUT:OPEN (@ 311:338)');% Open all channels
%fprintf(s,'ROUT:CLOS (@313,324,335)')%Go to original state
fprintf(s,'ROUT:SCAN (@104:106)') ; % set the channels to be scanned
fprintf(s,'MEAS:VOLT:DC? (@104:106)')% Measurement of rows voltage
%fprintf(s,'MEAS:TEMP TC,J,(@109)')% Measurement of temperature
x=str2num(fscanf(s));
[m,n]=size(x);
V=x
fclose(s)
% Adding: Matrix switches
function [T]=Switches(SW)
s = serial('COM1');
get(s,{'BaudRate','DataBits','Parity','StopBits'})
set(s,'Timeout',60); % to avoid long waiting when there is a problem...
fopen(s);
fprintf(s,'ABOR'); % abort running scan
query(s,'R?'); % delete data from the buffer
if SW(3,1) == 1
fprintf(s,'ROUT:CLOS (@333)')
end
if SW(3,2)==1
fprintf(s,'ROUT:CLOS (@323)')
end
if SW(3,3)==1
fprintf(s,'ROUT:CLOS (@313)')
end
if SW(2,1)==1
fprintf(s,'ROUT:CLOS (@334)')
end
if SW(2,2)==1
fprintf(s,'ROUT:CLOS (@324)')
end
if SW(2,3)==1
fprintf(s,'ROUT:CLOS (@314)')
end
if SW(1,1)==1
fprintf(s,'ROUT:CLOS (@335)')
end
if SW(1,2)==1
fprintf(s,'ROUT:CLOS (@325)')
end
if SW(1,3)==1; fprintf(s,'ROUT:CLOS (@315)'); end; fclose(s); T=1

165
References
[1] Kosuke Kurokawa “Energy from Desert” 2003, ISBN 1 902916417
[2] C.J.Winter, R.L. Sizmann, L.L. Van-Hull, “Solar Power Plants”, ISBN-3-540-18897-5,
ISBN 0-387-18897-5, Springer-Verlag, 1991.
[3] “Stand alone solar PV systems”, http://www.powerfromthesun.net
[4] Stuart R. Wenham, Martin A. Green, “Applied photovoltaic”, ISBN-10: 1-84407-401-3,
ISBN-13: 978-1-84407-401-3, Earth Scan, 2007.
[5] Falk Antony, Christian Durschner, “Photovoltaic for Professionals”, ISBN-13: 978-3-93459-
543-9, Earth Scan, 2007.
[6] http://www1.eere.energy.gov/solar/utility_scale.html
[7] F. Jeffrey, Power Film Inc., Private communication
[8] Dan Weinstock, Joseph Appelbaum “Shadow Variation on Photovoltaic Collectors in a Solar
Filed” The 23rd IEEE Convention of Electrical and Electronics Engineers in Israel, 2004.
nd
[9] Dan Weinstock, Joseph Appelbaum “Optimal Design of Solar Field” The 22 Convention of
Electrical and Electronics Engineers in Israel, 2002.
[10] S. H. Raushenbach, “Electrical output of shadowed solar arrays,” IEEE Transactions on
Electronic Devices, vol. 18, p. 483, Aug. 1971.
[11] V.Quaschning and R. Hanitsch, “Influence of shading on electrical parameters of solar
cells,” Photovoltaic Specialists Conference, 13-17 May 1996.
[12] M. S. Swaleh, M. A. Green,“ Effects of shunt resistance and bypass diode on the shadow
tolerance of solar cell modules,” Solar Cells, vol. 5, no. 2, pp.183–198, 1982.
[13] “Newton Raphson method”, http://mathworld.wolfram.com/NewtonsMethod.html
[14] Toshihisa Shimizu, Masaki Hirakata, Tomoya Kamezawa, and Hisao Watanabe,“Generation
Control Circuit for Photovoltaic Modules,” IEEE Transactions On Power Electronics,
Vol.16, No.3 May 2001, pp. 293-300.
[15] Narendra D. Kaushika and Nalin K. Gautam “Energy Yield Simulations of Interconnected
Solar PV Arrays,” IEEE Transactions on Energy Conversion, vol. 18, No. 1, March 2003, pp.
127-134.
[16] B. Marion,’ J. Adelstein,’ K. 8oyle.l H. Hayden.’ B. Hammond,’ T. Fletcher,’ B. Canada,’
D. Narang,‘ A. Kimber.3 L. Mit~hell.~ G. Rich,4 and T. fownsend4 “Performance
Parameters for Grid-Connected PV Systems”, IEEE 2005.
[17] Ward T. Jewell, Timothy D. Unruh “Limits on Cloud-Induced Fluctuation in Photovoltaic
Generation”, IEEE Transactions on Energy Conversion, Vol. 5, No. 1, March 1990.
[18] Tamura J., Nakamura H., Inoue Y., Otani K., Kurokawa K., ”A new method of calculating
in-plane irradiation by one-minute local solar irradiance” IEEE Conference on Photovoltaic
Energy Conversion, vol.3, 2003.
[19] V.Quaschning and R. Hanitsch, “Numerical simulation of current-voltage characteristics of
photovoltaic systems with shaded solar cells,” Solar Energy, vol. 56, no. 6, pp. 513–520,
1996.
[20] Dzung D Nguyen, Brad Lehman “Modeling and Simulation Solar PV Array under Shadow
Conditions”, IEEE Compel, July 17-19, 2006.
[21] Satsuki Ike , Kosuke Kurokawa ,” Photogrammetric Estimation of Shading Impacts on
Photovoltaic Systems”, 2005 IEEE.
[22] Francois Giraud, Ziyad M. Salameh, “Analysis of the Effects of a Passing Cloud on a Grid-
Interactive Photovoltaic System with Battery Storage using Neural Network,” IEEE
Transactions on Energy Conversion, Vol. 14, No. 4, December 1999.
[23] Takashi Hiyama, Ken Kitabayashi, “Neural Network Based Estimation of Maximum Power
Generation from PV Module Using Environmental Information,” IEEE Transaction on
Energy Conversion, Vol.12, No.3, September 1997.

166
[24] Takashi Hiyama, Shinichi Kouzuma, Tomofumi Imakubo, “Identification of Optimal
Operating Point of PV Modules using Neural Network for Real Time Maximum Power
Tracking Control,” IEEE Transaction on Energy Conversion, Vol. 10, No. 2, Jun 1995.
[25] L. Hontoria, 1. Aguilera, P. Zufiria, . Grupo de Redes Neuronales. ”A Tool for Obtaining the
LOLP Curves for Sizing Off-Grid Photovoltaic System Based on Neural Network”, IEEE
Conference on Photovoltaic Energy Conversion, May 11-18, 2003.
[26] A.Al-Amoudi and L.Zhang, “Application of radial basis function networks for solar-array
modeling and maximum power-point predict ion”, IEEE Transmission and Distribution, Vol.
147, No. 5, September 2000.
[27] Wilhelm Durisch and Jean-Claude Mayor “Application of A Generalized Current Voltage
Model for Module Solar Cells to Outdoor Measurement On a Siemen SM 110-Module,”
IEEE Conference on Photovoltaic Energy Conversion’ May 11-18, 2003.
[28] Shengyi Liu, Roger A. Dougal, “Dynamic Multiphysics Model for Solar Array,” IEEE
Transaction on Energy Conversion, vol. 17, No. 2, June 2002.
[29] Boico F., Brad Lehman, Shujaee K.,“ Solar Battery Chargers for NiMH Batteries,” IEEE
Transaction on Power Electronics, September 2007.
[30] N. D. Kaushika, R. Gera, and K. Kaushika, “Hot spot protection in solar PV modules and
arrays,” IEEE Nat. Conf. Solar Energy Syst., 1988, pp. 113–117.
[31] Rodriguez, C., Amaratunga, G.A.J., “Energy control for long lifetime photovoltaic ac
module inverter”, IEEE Power Electronics Specialists Conference, 2006.
[32] Hisao Watanabe, Toshihisa Shimizu, Gunji Kimura, “A Novel Utility Interactive
Photovoltaic Inverter with Generation Control Circuit,” IEEE Industry Electronics Society,
1998.
[33] Zhang, X., Ni, H., Yao, D., Cao, R.X., Shen, W.X.,” Design of Single Phase Grid-connected
Photovoltaic Power Plant based on String Inverters,” IEEE Industrial Electronics and
Applications Conference, 2006.
[34] El-Shibini, M.A., Rakha, H.H.” Maximum power point tracking technique,” Electro technical
Conference Proceedings, MELECON '89., Mediterranean, 11-13 April 1989, pp. 21-24.
[35] Chin Chang “Solar Cell Array Having Lattice or Matrix Structure and Method Of Arranging
Solar Cells and Panels,” Patent No.: US 6,635,817 B2, Oct. 21, 2003.
[36] Raed A. Sherif, Karim S. Boutros, Moorpark United States Patent “Solar Module Array with
Reconfigurable Tile,” Patent No. US 6,350,944 B1, Feb, 26, 2002.
[37] Volker Quaschning “Understanding renewable energy systems”, Earth Scan 2005, ISBN 1-
84407-128-6 (pbk.)
[38] Energy Trust Program http://www.energytrust.org/TA/solar/charts/How_to_use.html
[39] Yanagawa S., Kato T., Kai W., Tabata A., Yokomizu Y., Okamoto T., Suzuoki Y.,
”Evaluation of LFC capacity for output fluctuation of photovoltaic power generation systems
based on multi-point observation of insolation,” IEEE Power Engineering Society Summer
Meeting, vol.3, 2001.
[40] G. R. Walker , P. C. Sernia, “Cascade DC-DC Converter Connection of Photovoltaic
Modules,” IEEE Transactions on Power Electronics, Vol.19, No.4, July 2004.
[41] Shiratani, M., Iwashita, S., Bando, K., Inoue, T., Koga, K.,” Incorporation of Higher-Order
Silane Radicals Into A-Si:H Films of High Stability Against Light Exposure,” IEEE
Conference on Photovoltaic Energy Conversion, Vol. 2, May 2006 Page(s):1596 – 1599.
[42] Roman E., Ibanez P., Elorduizapatarietxe S., Alonso R., Goitia D., Martinez D. Alegria,
“Intelligent PV module for Grid-Connected PV system,” IEEE Industrial Electronics
Conference, 2-6 November 2004, pp. 3082-3087.
[43] Weidong Xiao, Nathan Ozog, William G. Dunford, “Topology Study of Photovoltaic
Interface for Maximum Power Point Tracking”, IEEE Transactions on Industrial Electronics,
Vol.54, No.3, June 2007.

167
[44] Kimiyoshi Kobayashi, Hirofumi Matsuo, Yutaka Sekine “Novel Solar-Cell Power Supply
System Using a Multiple-Input DC–DC Converter”, IEEE Transactions on Industrial
Electronics, Vol.53, No.1, February 2006.
[45] Roman E., Ibanez P., Elorduizapatarietxe S., Alonso R., Goitia D., Martinez D. Alegria,
“Intelligent PV module for Grid-Connected PV system,” IEEE Transactions on Industrial
Electronics, Vol.54, No.4, August 2006.
[46] El-Shibini, M.A.; Rakha, H.H.” Maximum power point tracking technique,” Electro
technical Conference Proceedings, MELECON '89., Mediterranean, 11-13 April 1989, pp.
21-24.
[47] Yinggrak Auttawaitkul, Boonnua Pungsiri, Kosin Chammongthai, Makoto Okuda “A
Method of Appropriate Electrical Array Reconfiguration Management for Photovoltaic
Powered Car”, The 1998 IEEE Asia-Pacific Conference on Circuit and Systems, 1998.
[48] Chin Chang “Solar Cell Array Having Lattice or Matrix Structure and Method Of Arranging
Solar Cells and Panels,” Patent No.: US 6,635,817 B2, Oct. 21, 2003.
[49] Raed A. Sherif, Karim S. Boutros, Moorpark United States Patent “Solar Module Array with
Reconfigurable Tile,” Patent No. US 6,350,944 B1, Feb, 26, 2002.
[50] Wilhelm Durisch and Jean-Claude Mayor “Application of A Generalized Current Voltage
Model foe Module Solar Cells to Outdoor Measurement On a Siemen SM 110-Module,” 3rd
World Conference on Photovoltaic Energy Conversion’ May 11-18, 2003.
[51] Shengyi Liu, Roger A. Dougal, “Dynamic Multiphysics Model for Solar Array,” IEEE
Transactions on Energy Conversion, vol. 17, No. 2, June 2002.
[52] Y. Zhu, B. Lehman, “Three Level Switching Cell for Low Voltage, High-Current DC-DC
converters”, IEEE Transaction on Power Electronics, Vol.22, No.5, September 2007.
[53] Masato H. Ohsato, Satoru Inarida, Gunji Kimura, Mitsuo Shioya” Advanced Circuit of Five-
Level PWM inverter with an Inductive Load”, IECON, 1991
[54] Tsai-Fu Wu, Chien-Hsuan Chang, Yong-Jing Wu “Single-Stage Converters for PV Lighting
Systems with MPPT and Energy Backup”, IEEE Transaction on Aerospace and Electronics
systems 1999.
[55] Martina Calais Vassilios G. Agelidis, ”Multi-level converters for Single Phase Grid
Connected Photovoltaic Systems- An overview”, IEEE, 1998
[56] Takashi Kawasaki, Satoshi Hosoda, Jeong-ho Kim, Kazuhiro Toyoda, and Mengu
Cho”Charge Neutralization via Arcing on a Large Solar Array in the GEO Plasma
Environment” IEEE Transaction on Plasma Science, October, 2006.
[57] Juan Dixon, Luis Morán,”High-Level Multistep Inverter Optimization Using a Minimum
Number of Power Transistors”, IEEE Transaction on Power Electronics, Vol.21, No.2 March
2006.
[58] Ziyad M. Salameh, William A. Lynch, “Performance Analysis of a PV- Powered Health
Clinic with Multi-Stage Dual Priority Regulator,” IEEE Transaction on Energy Conversion,
Vol.5, No.4, December 1990.
[59] Lijun Gao, Roger A. Dougal, Shengui Liu, Albena Iotova “Portable Solar System using a
Step-up Power Converter with a Fast-Speed MPPT and a Parallel-configured Solar Panel to
Address Rapidly Changing Illumination,” IEEE Applied Power Electronics Conference,
2007.
[60] N. Femia, G. Petrone, G.Spagnuolo, M. Vitelli, “Matching the Photovoltaic Field
Orientation to Load Requirement in Stand-Alone Distributed Power System,” IEEE Power
Electronics Specialist Conference, 2004.
[61] A. Mellit, M. Benghanem, A. Hadj Arab, A. Guessoum, K. Moulai “Neural Network
Adaptive Wavelets for Sizing of Stand-Alone Photovoltaic System,” IEEE International
Conference on Intelligent Systems, June 2004.

168
[62] Nobuyoshi Mutoh, Takaoshi Matuo, Kazuhito Okada, Masahiro Sakai, “Prediction Data-
Based Maximum Power Point Tracking Method for Photovoltaic Power Generation System,”
IEEE Power Electronics Specialists Conference, Vol. 3, 23-27 June 2002.
[63] Takashi Hiyama, Shinichi Kouzuma, Tomofumi Imakubo, “Identification of Optimal
Operating Point of PV Modules using Neural Network for Real Time Maximum Power
Tracking Control,” IEEE Transaction on Energy Conversion, Vol. 10, No. 2, Jun 1995.
[64] S. Amstrong, W. G. Hurley, “Self- Regulating Maximum Power Point Tracking for Solar
Energy Systems,”
IEEE Universities Power Engineering Conference, Vol. 2, 6-8 Sept. 2004.
[65] E. Maset , E. Sanchis-Kilders , J. B. Ejea , A. Ferreres , J. M. Blanes , A. Garrigós , J. A.
Carrasco , A. H. Weinberg, “New Power Conditioning System for Battery-free Satellite
Buses with Maximum Power Point Tracking,” IEEE Applied Power Electronics Conference
2007.
[66] G. B. Shrestha I,. Goel, “ A Study on Optimal Sizing Of Stand-Alone Photovoltaic
Stations”, IEEE Transactions on Energy Conversion, Vol. 13, No. 4, December 1998.
[67] Mohammad A. .S.Masoum, Hooman Dehbonei, Ewald F. Fuchs, “Theoretical and
Experimental Analyses of Photovoltaic Systems With Voltage-and Current-Based Maximum
Power-Point Tracking”, IEEE Transaction On Energy Conversion, Vol.17, No.4, December
2002.
[68] http://www.otm.uiuc.edu/techs/techdetail.asp?id=326
[69] http://www.eurekalert.org/pub_releases/2005-01/uot-nnp010705.php
[70] Kida, J., Tokuda, K., Ishihara, Y., Todaka, T.” Analysis of DC-DC converter for the maximum
power point control of photovoltaic,” Telecommunications Energy Conference, 5-8 Nov. 1991.
[71] Nishida Y., Aikawa, N.,Sumiyoshi, S.,Yamashita, H., Omori, H.,”A novel type of utility-
interactive inverter for photovoltaic system” IEEE Power Electronics and Motion Control
Conference, Vol. 3, 14-16 Aug. 2004.
[72] Duran-Gomez, J.L., Garcia-Cervantes, E., Lopez-Flores, D.R., Enjeti, P.N., Palma, L.,
“Analysis and evaluation of a series-combined connected boost and buck-boost dc-dc converter
for photovoltaic application,” IEEE Applied Power Electronics Conference and Exposition,
2006.
[73] Vorster F.J., van Dyk E.E., Leitch A.W.R., “Investigation on the I-V characteristics of a high
concentration, photovoltaic array”, IEEE Photovoltaic Specialists Conference, pp.1604 –
1607, 2002.
[74] V.A. Letin, M.B. Kagan, V.P. Nadorov, V.M. Rjevsky, A.G. Arakelov, B.E. Chertok,
A.N. Domoratsky, A.I. Patsiora, U.P. Semenov, ASZernov, “ Basic Results of Post-flight
Researches of Solar Array (SA) Returned to The Earth from The Orbital Space Station
“MIR” after 10.5 Years of Operation,” IEEE 2000.
[75] http://www.systems-thinking.org/modsim/modsim.htm
[76] Enrique Daroqui Raga, Josep Soler Costa, Salvador Segui, Francisco Gimeno, Salvador Orts
“Direct Coupling between Photovoltaic Module and A PWM Converter,” IEEE Power
Electronics Specialists Conference, 2004.
[77] G. Grandi, D. Casadei, C. Rosi “Direct Coupling of Power Active Filters with Photovoltaic
Generation Systems with Improved MPPT Capability”, IEEE Power Tech Conference, June
23-26, 2003.
[78] J. Henriques, P. Gil, A. Cardoso, A. Dourado ,“Scheduling of PID Controllers by Means of
A Neural Network with Application to a Solar Power Plant”, 2002 IEEE.

169