You are on page 1of 79

TECHNICAL PERFORMANCE

EVALUATION OF SOLAR
PHOTOVOLTAIC SYSTEMS
A case study of eight PV systems on the Swedish market installed at RISE
research facilities

AWAZ RASHID
GULALA SAFARI

School of Business, Society and Engineering Supervisor: Jon Person, RISE


Course: Degree Project in Industrial Engineering Patrik Ollas, RISE
and Management with Specialization in Energy Bengt Stridh, MDH
Engineering Examinor: Hailong Li
Course code: ERA402 Customer: RISE
Credits: 30 hp Date: 2021-06-13
Program: M.Sc. in Industrial Engineering and Email:
Management asd16002@student.mdh.se
gsi15001@student.mdh.se
ABSTRACT

The objective of this thesis was to evaluate the technical performance of eight small-scale PV
(photovoltaic) systems on the Swedish market from June to November 2020. Furthermore,
the aim of this thesis was also to filter measured data since faulty data usually emerge in field
measurements. Several filtering methods have been applied to remove faulty data such as
linear interpolated values, outliers, and abnormal power generation to assure the quality of
the data used for the evaluation. The measured parameters were power out from the inverter,
the plane of array irradiance, ambient temperature, and module temperature. Although the
module technology had a certain effect on the module temperature, in this study, the
installation method influenced the module temperature of the systems to a greater extent. It
was found that building applied photovoltaic (BAPV) systems had lower module
temperatures compared to building-integrated photovoltaics (BIPV) systems. However, a
clear impact of the installation method on the system’s performance could not be observed.
Systems 3 and 6 were BAPV respective BIPV systems and they were the best-performing
systems in terms of specific energy yield (kWh/kWp) and Performance ratio (PR). The
average PR over the period was 89% for system 3 and 91% for system 6. The specific energy
yield was highest during June, it was around 135 kWh/kWp for both systems. The results also
showed that systems with mono-Si technology performed better than systems with CIGS
technology in terms of PR, specific energy yield, and energy per area (kWh/m2). Among the
four studied thin film systems, system 2 which consisted of CdTe was the best performing
system in terms of PR and monthly specific energy yield. In comparison to the mono-Si
systems, system 2 had an average PR equal to system 7 which had the lowest average PR
among the mono-Si systems. Additionally, based on the results in this study, a slightly
undersized inverter can cause power curtailment during high irradiance intensities. However,
power curtailment due to a slightly undersized inverter did not contribute to significant
losses in energy.

Keywords: Building integrated photovoltaics, Building applied photovoltaics, Performance


evaluation, Performance ratio, Specific energy yield, Module technology, Module
temperature, Inverter size
PREFACE

This master thesis is a part of the Master of Science program in Industrial Engineering and
Management with Specialization in Energy Engineering at Mälardalen University in Västerås.
The degree project was conducted during spring semester 2021 and was carried out on behalf
of RISE. The thesis work uses a subset of data from PV test systems financed by
Energimyndighetens Testlab, and the results will support RISE in the evaluation of these test
systems.

We want to express our sincere gratitude to Jon Persson and Patrik Ollas, our supervisors at
RISE, for giving us the opportunity to take part in this ongoing project. We want to thank
them for taking the time to give us their precious guidance and knowledge throughout the
project.

Finally, we would also like to extend our thanks to our supervisor Bengt Stridh and examiner
Hailong Li, at Mälardalen University, for the feedback and support.

Västerås, June 2021

Awaz Rashid Gulala Safari


SUMMARY

To achieve the national climate goals, a restructuring of the energy system and an increasing
use of renewable energy is a prerequisite. Sweden has therefore set a target to reach 100%
renewable electricity production by 2040. The Swedish Energy Agency believes that 5-10% of
the electricity consumption could be supplied from solar energy. Since PV systems can make
a significant contribution to a sustainable energy system, it is essential to contribute to
innovation in the PV industry. Studies have shown that by monitoring PV systems and
measuring data under real-life operating conditions, the technical performance of PV systems
can be evaluated more accurately. However, several studies have also emphasized the
importance of good and reliable measurement data.

Although there are several PV systems on the market in Sweden, there is still limited
information about their technical performance under field conditions. The objective of this
thesis was to evaluate the technical performance of eight small- scale PV systems on the
Swedish market installed at RISE research facilities. Furthermore, the aim of this thesis was
also to filter measured data since faulty data usually emerge in field measurements, mostly
due to power outages, interruption f0r maintenance, as well as component and
communication failure. In this study, parameters such as power out from the inverter, the
plane of array irradiance, ambient temperature, and module temperature were measured.
Moreover, the performance evaluation was based on data from June to November. Various
methods were applied to identify and remove faulty data such as linear interpolated values,
outliers, and abnormal power generation to assure the quality of the data used for the
evaluation. Additionally, a number of performance parameters such as specific energy yield
(kWh/kWp), energy generation per area (kWh/m2) and performance ratio (PR) were used in
this study to determine the overall performance of each PV system. The temperature
behavior for each system was also considered relevant for the analysis. Furthermore, the
Pearson correlation coefficient for the measured parameters were determined in order to
assess the correlation between the various parameters.

The results in this study showed that there were large variations in the average module
temperature among the systems. System 8 which was a building integrated photovoltaic
(BIPV) system had the highest module temperature, mainly because it was directly installed
on the roof of a loft which was an essential precondition that affected the module
temperature. Since it was directly installed on the roof, an air gap was not present, the air
circulation was reduced. Moreover, the temperature in the loft increased significantly during
the summer months which caused the module temperature of system 8 to increase even
more. System 2 and 4 consisted of thin film technology and did also have relatively high
module temperatures owing it to the installation method. System 3 and 7 which were the only
two building applied photovoltaic (BAPV) systems in this study consisted of mono-Si and had
the lowest module temperature. Although the module technology had a certain effect on the
module temperature, in this study, the installation method influenced the module
temperature of the systems to a greater extent.
Based on the results in this study, a slightly undersized inverter can cause power curtailment
during high irradiance intensities, this was the case for systems 6 and 7. However, power
curtailment due to a slightly undersized inverter did not contribute to significant losses in the
energy for these systems.

Large variations in PR could also be observed in this study, it differed around 20 percentage
between the highest and lowest (system 4 excluded). The results in this study indicated that
the systems with mono-Si technology performed better than the systems with CIGS
technology in terms of PR, specific energy yield, and energy per area. Since system 4 had
operational issues throughout the studying period it was not comparable with the other
systems. System 8 had the second lowest PR, and it was most likely caused by the higher
module temperature. Among the four studied thin film systems, system 2 which consisted of
CdTe was the best performing system in terms of PR and monthly specific energy yield. In
comparison to the mono-Si systems, system 2 had an average PR equal to system 7 which had
the lowest average PR among the mono-Si systems.

A clear impact of the installation method on the performance of the PV systems could not be
seen in this study, systems 3 and 6 were BAPV respective BIPV and they were the best-
performing systems in terms of specific energy yield and PR. The average PR over the period
was 89% for system 3 and 91% for system 6. The specific energy yield was highest during
June, it was around 135 kWh/kWp for both of the systems. Even though systems 6 had a
higher PR and energy yield, systems 3, 5, and 7 could be preferred over system 6 since these
could generate more energy per square meter of roof space.

According to the theory, PR should increase in the winter as there are no losses caused by a
high module temperature. However, the same trend could not be seen for all of the systems.
For instance, PR for system 2 dropped at colder months. One possible reason could be that
the systems do not reach extremely high temperatures during the summer months in
Swedish climate. Moreover, it is worth noting that the results were based on data for only 6
months and that the winter months were not included. Since the PR is affected by the module
temperature, the weather-corrected PR were also calculated with two different equations.
One of the equations corrected the PR to a fixed temperature of 25 ◦C while the other
equation corrected it to a site-dependent average module temperature over all the studied
months for each specific system. The differences between these two methods were not large
since the obtained site-dependent module temperature was in the range of 23 – 29 ◦C
(system 8 excluded).

Evaluating the performance under field conditions is complicated since many factors
influence the monitoring of the PV systems and also the filtering process. The results
presented in this study are largely dependent on the choices made during the installation of
the test systems and also the choices made in the filtering process.
SAMMANFATTNING

För att uppnå de nationella klimatmålen har Sverige satt ett mål att ha 100% förnybar
elproduktion fram till 2040. Energimyndigheten anser att 5–10% av elförbrukningen kan
komma från solenergi i framtiden. Solcellssystem kan ge ett betydande bidrag till ett hållbart
energisystem och därmed är det viktigt att bidra till innovation inom solcellsindustrin.
Studier har visat att övervakning och mätningar under fältförhållanden kan möjliggöra en
mer noggrann utvärdering av tekniska prestandan av solcellsanläggningar. Flera studier har
även betonat vikten av tillförlitliga mätdata vid utvärderingen av prestandan för
solcellsanläggningar.

Trots att det finns flera solcellssystem på den svenska marknaden är mängden information
om deras tekniska prestanda under fältförhållanden fortfarande väldigt begränsad. Syftet
med denna studie var att utvärdera den tekniska prestandan för åtta småskaliga
solcellssystem på den svenska marknaden som finns installerade vid RISE:s
forskningsanläggningar. Målet med detta arbete var också att filtrera mätdata eftersom det
vanligtvis förekommer felaktiga värden i fältmätningar, främst på grund av avbrott för
underhåll samt komponent- och kommunikationsfel. I denna studie uppmättes effekten från
växelriktaren, solinstrålningen mot modulens yta, omgivningstemperaturen och
modultemperaturen. Prestationsutvärderingen av systemen baserades på uppmätt data från
juni till november. Ett antal metoder användes för att identifiera och ta bort felaktiga data
såsom linjär interpolerade värden, extremvärden och onormala effektvärden i relation till
solinstrålningen för att säkerställa kvaliteten på datasetet. Flera prestandaparametrar såsom
specifikt energiutbyte (kWh/kWp), energiutbyte per yta (kWh/m2) och
prestationsförhållande (PR) användes i denna studie för att bestämma den totala prestandan
för varje solcellssystem. Temperaturbeteendet för varje system ansågs också vara relevant för
analysen. Dessutom bestämdes Pearson-korrelationskoefficienten för de uppmätta
parametrarna för att visa korrelationen mellan de olika parametrarna.

Resultatet i den här studien visade på att det fanns stora variationer i den genomsnittliga
modultemperaturen bland systemen. System 8, som var ett BIPV system, hade den högsta
modultemperaturen eftersom modulerna var direkt installerade på taket på en vind, vilket
var en grundförutsättning som påverkade modultemperaturen. Det fanns ingen luftspalt
mellan taket och modulen för system 8 och detta minskade i sin tur luftcirkulationen.
Dessutom ökade temperaturen i vinden avsevärt under sommarmånaderna vilket bidrog till
en ytterligare ökning av modultemperaturen för system 8. Systemen 2 och 4 som bestod av
tunnfilm teknologi hade också relativt höga modultemperaturer, orsakad främst av
installationsmetoden. Systemen 3 och 7 som var de enda BAPV-systemen i denna studie
bestod av mono-Si och hade de lägsta modultemperaturerna. Även om modultekniken hade
en viss effekt på modultemperaturen, verkar installationsmetoden i denna studie påverka
systemets modultemperatur i större utsträckning.

Baserat på resultaten i denna studie kan en något underdimensionerad växelriktare orsaka


effektkapning när solinstrålningsintensiteten är hög, detta var fallet för system 6 och 7.
Effektkapning på grund av en liten underdimensionerad växelriktare bidrog dock inte till
stora energiförluster för dessa system.

Stora variationer i PR kunde också observeras i denna studie, det skilde sig cirka 20 procent
mellan högsta och lägsta (system 4 uteslutet). Resultaten i denna studie pekar på att
systemen med mono-Si teknologi presterar bättre än systemen med CIGS teknologi när det
gäller PR, specifikt energiutbyte och energi per yta. System 4 hade driftsproblem och var
därmed inte jämförbar med de andra systemen. System 8 hade det näst lägsta PR värdet och
det berodde med stor sannolikhet på den högre modultemperaturen. Bland de fyra studerade
tunnfilmsystemen var system 2, som bestod av CdTe, det bästa systemet när det gällde PR
och specifikt energiutbyte. Jämfört med mono-Si systemen, presterade system 2 lika bra som
system 7 som hade lägst PR bland mono-Si systemen. En tydlig inverkan av
installationsmetoden på PV-systemens prestanda kunde inte observeras i denna studie,
systemen 3 och 6 var BAPV respektive BIPV och var de bästa systemen när det gällde
specifikt energiutbyte och PR. Det genomsnittliga PR värdet under perioden var 89% för
system 3 och 91% för system 6. Det specifika energiutbytet var högst under juni, det var cirka
135 kWh/kWp för båda systemen. Trots att system 6 hade högre PR- och energiutbyte kan
systemen 3, 5 och 7 föredras framför system 6 eftersom dessa system kunde generera mer
energi per kvadratmeter.

Enligt teorin bör PR öka på vintern eftersom det inte finns några förluster orsakade av en hög
modultemperatur. Detta kunde dock inte ses för alla systemen, exempelvis sjönk PR för
system 2 under de kallare månaderna. En möjlig anledning kan vara att systemen inte når
extremt hög temperaturer under sommarmånaderna i svenskt klimat. Dessutom är det värt
att notera att resultaten baserades på data från endast 6 månader och att vintermånaderna
inte inkluderades. Eftersom PR påverkas av modulens temperatur beräknades också
väderkorrigerad PR med två olika ekvationer. En av ekvationerna korrigerade PR mot en
standardtemperatur på 25 ° C. Den andra ekvationen korrigerade PR mot en platsberoende
genomsnittlig modultemperatur under alla studerade månader för varje specifikt system.
Skillnaderna mellan dessa två metoder var inte stora eftersom den erhållna platsberoende
modultemperaturen låg mellan 23 - 29 ° C (system 8 uteslutet).

Att utvärdera prestanda under fältförhållanden är komplicerat eftersom många faktorer


påverkar övervakningen av solcellssystemen och även filtreringsprocessen. Resultaten som
presenterades i denna studie beror till stor del på de val som har gjorts under installationen
av testsystemen och under filtreringsprocessen.
CONTENT

1 INTRODUCTION .................................................................................................. 1
1.1 Background .............................................................................................................................. 1

1.2 Problem definition ................................................................................................................... 2

1.3 Purpose..................................................................................................................................... 2

1.4 Research questions ................................................................................................................. 2

1.5 Delimitation and scope of the study ...................................................................................... 3

1.6 Description of grid-connected solar PV systems ................................................................. 3

2 LITERATURE STUDY ......................................................................................... 6


2.1 Factors affecting PV system performance ............................................................................ 6
2.1.1 Plane of array irradiance ....................................................................................................... 6
2.1.2 Module temperature .............................................................................................................. 8
2.1.3 Influence of module technology on PV system performance ................................................ 9
2.1.4 Influence of installation method on PV system performance............................................... 10
2.1.5 Solar PV Inverter ................................................................................................................. 10

2.2 Performance metrics ............................................................................................................. 12


2.2.1 Performance ratio (PR)........................................................................................................ 14
2.2.2 Weather corrected performance ratio (PRcorr) ..................................................................... 15

3 METHOD ............................................................................................................ 17
3.1 Data acquisition ..................................................................................................................... 17

3.2 Data filtering ........................................................................................................................... 18


3.2.1 Identification and removal of faulty data .............................................................................. 18
3.2.2 Treatment of missing data ................................................................................................... 19

3.3 Performance parameters for PV systems ........................................................................... 19

4 CURRENT STUDY ............................................................................................. 21


4.1 Description of the systems ................................................................................................... 21

4.2 Measured parameters ............................................................................................................ 24

4.3 Identification of faulty data ................................................................................................... 25


4.3.1 Linear interpolated values and outliers ................................................................................ 25
4.3.2 Abnormal power generation ................................................................................................ 27

4.4 Removing faulty data............................................................................................................. 27


4.4.1 Removal of global and local outliers .................................................................................... 27
4.4.2 Removal of linear interpolated values ................................................................................. 29
4.4.3 Checking deviation from mean and removing invalid data .................................................. 29
4.4.4 Removal of abnormal power generation.............................................................................. 32

4.5 Replacing missing data ......................................................................................................... 33

4.6 Rate of missing data values.................................................................................................. 33

5 RESULTS ........................................................................................................... 35
5.1 Measured field data ............................................................................................................... 35

5.2 Performance ratio .................................................................................................................. 38

5.3 Monthly specific energy yield ............................................................................................... 40

5.4 Relationships between the different parameters ................................................................ 41


5.4.1 Performance ratio related to module temperature and plane of array irradiance ................ 41
5.4.2 Power as function of the plane of array irradiance .............................................................. 43
5.4.3 Module temperature as a function of the plane of array irradiance ..................................... 45
5.4.4 Pearson correlation ............................................................................................................. 46

6 DISCUSSION ..................................................................................................... 48
6.1 Impact of module technology and installation method on module temperature............. 48

6.2 Impact of inverter on the performance ................................................................................ 49

6.3 Variations in performance ratio and energy generation .................................................... 50

6.4 Method discussion ................................................................................................................ 52

7 CONCLUSIONS ................................................................................................. 54

8 SUGGESTIONS FOR FURTHER WORK .......................................................... 55

REFERENCES .......................................................................................................... 56
LIST OF FIGURES

Figure 1 A grid-connected PV system includes a number of components; the main components are

the inverter and the PV array. Other components are DC disconnect, AC disconnect, main

service, and utility meter. .......................................................................................................... 3

Figure 2 The I-V and P-V curve of a module where the short circuit current (ISC), open circuit voltage

(VOC), maximum power point (MPP), current at maximum power point (IMP), and voltage at

maximum power point (VMP) are marked................................................................................. 5

Figure 3 Plane of array irradiance consists of direct, diffuse and reflected solar radiation. ................. 6

Figure 4 Distribution of annual average global radiation over Sweden ................................................. 7

Figure 5 I-V and P-V curve for three irradiance levels at constant ambient temperature. The dashed

line represents the I-V curve while the lines represent P-V curve ........................................... 8

Figure 6 I-V and P-V curve at three ambient temperatures and constant irradiance. The dashed line

represents the I-V curves while the lines represent P-V curves ............................................... 9

Figure 7 Inverter efficiency as a function of DC power for three different input voltages. The

efficiency curve belongs to the Sunny Tripower 10.0 inverter ................................................ 11

Figure 8 European efficiency of an inverter as a function of voltage at MPP. The curve for European

efficiency belongs to the Sunny Tripower 10.0 inverter ......................................................... 12

Figure 9 Energy flow in a grid- connected PV system where solar irradiance is received by the PV

array which are later on converted into AC power. Performance parameters such as Yr, Yf

and PR can be used to determine the performance of a PV system at various levels. ........... 13

Figure 10 Comparison of annual noncorrected PR and weather corrected PR on simulated data ... 16

Figure 11 A schematic flow chart illustrating the steps which has to be accomplished for the

performance evaluation of the PV systems. ....................................................................... 17

Figure 12 PV system 1-6 on Roof 1. ..................................................................................................... 21

Figure 13 PV system 8 and 9 on Roof 2. ............................................................................................. 22

Figure 14 Position of sensors measuring the module temperature on system 5. .............................. 25

Figure 15 Linear interpolated values and local outliers for the module temperature of system 1

detected…………… ................................................................................................................ 26

Figure 16 Global outliers for module temperature of system 1 detected. .......................................... 26


Figure 17 Identification of faulty data for system 2 when power generation is plotted as a function

of plane of array irradiance. Data points which lie within the red marked areas are

examples of faulty data. ...................................................................................................... 27

Figure 18 Removal of a local module temperature outlier with the Hampel filter for system 1. ......28

Figure 19 Removing linear interpolated values for module temperature of System 1. ..................... 29

Figure 20 A comparison of the three temperature sensors’ deviation from mean for system 3. ......30

Figure 21 (a) Measured temperature from three sensors on system 3 (b) Sensor Tmod3,3 for system 3

starts to behave differently compared to the other sensors. .............................................. 31

Figure 22 Comparison of the ambient temperature sensors, where sensor 5 is behaving differently.

……………………………………………………………………………………………………………………………31

Figure 23 Removal of faulty data for system 2 with fixed and percentage limits, where power is

plotted as a function of the plane of array irradiance. ....................................................... 32

Figure 24 Average monthly module temperature for the whole period. Nighttime values (plane of

array irradiance values lower than 20 W/m2) are excluded from the module

temperatures. *System 8 is not completely comparable due to its placement. ................ 35

Figure 25 The behavior of the module temperatures for all the systems on a sunny day in June. ... 36

Figure 26 Average weekly Tamb for June-November. Nighttime values (plane of array irradiance

values lower than 20 W/m2) are excluded. ........................................................................ 37

Figure 27 Average weekly plane of array irradiance for June-November. The three sensors are

located at the systems within a radius of 100 m from each other. .................................... 37

Figure 28 Average weekly PR (not weather-corrected) for all the systems for June-November. ..... 38

Figure 29 Comparison of regular PR and weather-corrected PR (corrected t0 25 °C) for systems 3, 5

and 8 as a function of time. ................................................................................................ 39

Figure 30 Comparison of a weather-corrected PR with a site-dependent average module

temperature and correction to 25 °C for system 3, 5 and 8. ..............................................40

Figure 31 Monthly specific energy yield for June to November for each system normalized to

nameplate DC power...........................................................................................................40

Figure 32 Monthly energy generation for June to November for each system normalized to the

system area.......................................................................................................................... 41
Figure 33 (a) Performance ratio as a function of the plane of array irradiance for systems 2, 3, 5, 6

and 7. (b) Performance ratio as a function of the plane of array irradiance for systems 1, 4

and 8. Figures represent data for the whole period. .......................................................... 42

Figure 34 Scatter plot of the performance ratio as a function of module temperature for system 6

and 8 that are systems with the highest and the lowest performance ratio. Figures

represent data for the whole period. .................................................................................. 42

Figure 35 (a) Normalized power to the installed DC power as a function of the plane of array

irradiance (data points averaged to 15 minutes). (b) Linear regression for power as a

function of the plane of array irradiance (data points averaged to 15 minutes). Figures

represent data for the whole period. .................................................................................. 43

Figure 36 Normalized power to the system area as a function of the plane of array irradiance (data

points averaged to 15 minutes). The figure represent data for the whole period.............. 44

Figure 37 (a) Power as a function of the plane of array irradiance for system 2 and 3 (original data

with 15 seconds interval). (b) Power as a function of the plane of array irradiance for

system 7 and 6. Figures represents data for the whole period. Note the different y-axis

scales. .................................................................................................................................. 44

Figure 38 An illustration of how often the systems work at certain rates of its capacity. A value

equal to one indicates that the inverter is working at its full capacity. ............................. 45

Figure 39 (a) Module temperature as a function of plane of array irradiance for system 7 and

system 8. (b) Linear regression for the module temperature as a function of the plane of

array irradiance for all systems. Figures represent data for the whole period. ................. 46
LIST OF TABLES
Table 1 Efficiency of the most common module technologies. ........................................................... 10

Table 2 Specification of the eight PV systems. ....................................................................................... 23

Table 3 Number of sensors measuring the module temperature for each system. ............................ 24

Table 4 Number of sensors measuring the ambient temperature. ..................................................... 24

Table 5 Number of sensors measuring the plane of array irradiance. ................................................ 25

Table 6 Minimum and maximum values for each measured parameter. ...........................................28

Table 7 The value of R2 before and after removing abnormal power generation. .............................. 33

Table 8 The rate of missing data for each parameter after removal of linear interpolated values and

outliers (Night values included).............................................................................................. 34

Table 9 The rate of missing data after filtering abnormal power generation. This filtering was done in

two steps, removal with a fixed limited and a percentage limit from the regression line. ..... 34

Table 10 Site-dependent average module temperature that is irradiance-weighted for each system

during June-November (night values excluded). ................................................................... 38

Table 11 An overview on the average PR over the studied period, PR corrected to the site dependent

average module temperature and PR corrected to 25 °C. ......................................................40

Table 12 Pearson correlation for all the measured parameters for each system. ................................. 47
NOMENCLATURE

Symbol Description Unit


ISC Short circuit current A
MPP Maximum Power Point W
P0 Nameplate DC power W
Pac Power out from inverter W
PR Performance ratio %
Tamb Ambient temperature °C
Tmod Module temperature °C
VOC Open circuit voltage V
Yf Instantaneous system yield kWh/kW
Yr Reference yield Hours

ABBREVIATIONS

Abbreviation Description
AC Alternating current
BIPV Building-integrated photovoltaics
BAPV Building-applied photovoltaics
CdTe Cadmium telluride
CIGS Copper indium gallium selenide solar
DC Direct current
MAD Median absolute deviation
Mono-Si Monocrystalline silicon
MPPT Maximum power point tracker
PR Performance ratio
PV Photovoltaics
RISE Research Institutes of Sweden
STC Standard test conditions
1 INTRODUCTION

This study focuses on evaluating the technical performance of various PV systems in a


Swedish climate. The following section will clarify the importance of the performance
evaluation of PV systems in field operations. Furthermore, the problem definition, purpose
and research questions of this thesis are presented. To facilitate the understanding, a short
background about PV systems and an insight into some of the components and technologies
relevant for this study is given at the end of this section.

1.1 Background

The usage of fossil fuels over the past century has increased greenhouse gases in the
atmosphere, which has increased the average temperature on earth and accelerated climate
change (NASA, n.d.). To reduce the impact of climate change, the Paris Agreement was
adopted where the goal is to limit global warming below 2 °C and preferably to 1.5°C,
compared to pre-industrial levels (UN, n.d.). As a consequence of climate change and the
Paris Agreement, countries have set up national climate goals (EU, n.d.). To achieve the
national climate goals, a restructuring of the energy system and an increasing use of
renewable energy is a prerequisite (Regeringen, 2020). Sweden has therefore set a target to
reach 100% renewable electricity production by 2040 (Regeringen, n.d.). The Swedish
Energy Agency believes that 5-10% of the electricity consumption could be supplied from
solar energy (Energimyndigheten, 2018).

Since PV systems can make a significant contribution to a sustainable energy system, it is


essential to contribute to innovation in the PV industry and invest in this renewable
technology (ETIP PV, 2021). According to the Ministry of the Environment (2016), solar
applications in Sweden have evolved to become technically applicable and economically
feasible during recent years. The global development of solar technology and various national
subsidy schemes, stimulating the use of solar energy, has played a significant role in this
progress (Alight, 2019).The total installed capacity of grid-connected PV systems reached
1.09 GW at the end of 2020, which accounted for an increase of 0.392 GW compared to the
previous year (Energimyndigheten, 2021).The grid-connected market is mainly comprised of
distributed roof-mounted systems installed by homeowners, municipalities, and companies
(Lindahl, Dahlberg, & Westerberg, 2019).

The main concern of individuals who install photovoltaic systems is whether the system will
perform as expected (Ma et al., 2018). Manufacturers give information about the
performance of their panels in their datasheets. Woyte et al. (2014) clarify that these
documents rely on the data measured under standard testing conditions (STC) where the
irradiance is 1000 W/m2, the air mass is 1.5 and a cell temperature is 25°C. These conditions
are rarely encountered in field operation since the air mass depends on the geographical

1
location and the cell temperature varies continuously due to change in ambient temperature
and solar irradiance (Woyte et al., 2014). The information in the datasheets can be beneficial
to compare different panels but they do not give any information about the actual
performance of a PV system in real-life operating conditions (Ustun et al., 2018). Thus,
Woyte et al. (2014) point out that it is important to monitor PV systems and measure data
under field conditions to accurately evaluate the technical performance of PV systems. Livera
et al. (2020a) add that the quality of the measured data can also affect the performance
evaluation of PV systems. Thus, good and reliable measurement data is also a prerequisite to
building knowledge about PV systems and increasing the accuracy of the analysis (Livera et
al., 2020a).

1.2 Problem definition


There are various types of small-scale PV systems on the Swedish market. Each of these
systems has different characteristics and consists of a unique combination of components
and technologies which affect the performance in different ways. Although there are several
PV systems on the Swedish market, there is still limited information about their technical
performance under field operations. Since values estimated during the initial design differ
from values achieved under field conditions, it is essential to monitor and measure data from
these systems for more accurate analysis (Woyte et al., 2014). By evaluating the performance
under field conditions, one can assess the impact of various factors such as temperature and
irradiance on the performance of the PV systems. Furthermore, the performance of the PV
systems can also be related to their technical characteristics and components such as module
technology, installation method1, and inverter. Consequently, performance evaluations under
field conditions can serve as a guide for future decisions of system installers and customers.

1.3 Purpose
The purpose of this thesis is to evaluate the technical performance of eight PV systems on the
Swedish market installed at RISE’s facilities in Borås, Sweden. The technical performance of
each system will be analyzed and related to its technical characteristics. Furthermore, the aim
of this thesis is also to develop a method to filter the measured data acquired from the eight
PV systems.

1.4 Research questions

• How do module technology and installation methods affect the module temperature
and the performance of each PV system?
• How does the size of the inverter affect the performance of the PV systems?
• What kind of differences in the performance ratio and energy generation can be seen
between the PV systems?

1 In this study, installation method refers to whether the system is a BIPV or BAPV system.

2
1.5 Delimitation and scope of the study
This study concerns eight small-scale PV systems. The choice of PV systems was made to
have a representation of technologies available on the Swedish market. The systems were
installed in spring 2020 and the measured data used in this case study is from June to
November 2020. The measured data include plane of array irradiance (GPOA), ambient
temperature (Tamb), module temperature (Tmod), and the power generated after the inverter
(PAC). The wind speed will not be considered since this parameter has not been measured.
Since the parameters between the PV array and the inverter (DC quantities) are not
measured, they are not considered either. Thus, the analysis is on a system level and will not
focus on the performance of individual components. However, this study will relate the
inverter size to the performance of the PV systems. The evaluation will be on the technical
performance of each system. Key performance indexes for the evaluation are the
Performance Ratio (PR), specific energy yield, energy generation per area and temperature
behavior of the systems. However, this study will not pay attention to the optimization of the
design of the systems. Financial perspectives have also been excluded in the evaluation of the
PV system.

1.6 Description of grid-connected solar PV systems


A grid-connected PV system is tied to the local utility grid as illustrated in Figure 1. Since it is
tied to the local utility grid, the excess electricity can be sold by exporting it to the utility grid,
thus, storage is not required. This type of PV system has also access to backup power from
the utility grid if it cannot generate the required amount of electricity to meet the
instantaneous load demand. The main components in a grid-connected PV system are the
inverter and the PV array. (Solar Reviews, 2019)

Figure 1 A grid-connected PV system includes a number of components; the main components are
the inverter and the PV array. Other components are DC disconnect, AC disconnect, main
service, and utility meter.

The PV cell is the most fundamental component in a PV system, and it is comprised of two or
more thin layers of semiconducting material. When the semiconducting material is exposed
to sunlight it generates DC electricity via the photovoltaic effect. There are various
semiconducting materials with their specific composition and atomic structure (Kalogirou,
2014). Crystalline silicon (c-Si) cells are the oldest and still the most commonly used PV

3
technology. There are two main c-Si types: monocrystalline silicon (Mono-Si) and
polycrystalline silicon (Poly-Si), with a global market share of approximately 95%. Other
commonly used technologies are Cadmium telluride (CdTe), Copper indium gallium selenide
(CIGS), and amorphous silicon (a-Si). These are different types of thin film technologies and
they accounted for only 5% of the world market share in 2019 (Fraunhofer ISE, 2020).

One single PV cell cannot generate large amounts of electricity, therefore, several PV cells are
connected and encapsulated into a product called a module or panel. The number of modules
connected will determine the total electrical output of the PV system. The modules are
usually connected in series to a PV string or array (Corba et al., 2012). Roof-mounted PV
modules can either be classified as building-applied photovoltaics (BAPV) or building-
integrated photovoltaics (BIPV). In the case of BAPV systems, the modules are attached to
the roof by using additional mounting structures and moving rails. Whereas the modules of
BIPV systems are integrated into the roof (Kumar, Sudhakar, & Samykano, 2019). Moreover,
according to IEA PVPS Task 15, a PV system cannot be classified as BIPV if the modules do
not have a building-related functionality and if they do not replace conventional building
material (Berger et al., 2018). One positive aspect of BAPV systems is that they are easier to
maintain and also easier to install on existing buildings, while BIPV systems are aesthetically
more pleasant (Enquist, 2012).

The main task of the inverter in a grid-connected PV system is to convert DC power


generated from the PV array to AC power (Bengtsson et al., 2017). Since grid-connected PV
systems are connected to the local utility grid, the inverter has to synchronize the phase and
frequency of the current to match the utility grid (Solar Reviews, 2019). Moreover,
Vidyanandan (2017) explains that string inverters are the most commonly used inverter type
in a grid-connected PV system. This type of inverter is connected to multiple strings of panels
in parallel and normally includes one or more maximum power point trackers (MPPTs)
(Bengtsson et al., 2017). MPPTs are used to find the maximum power point (MPP) which
represents a unique point on the Current-Voltage (I-V) and the Power-Voltage (P-V) curves
for a PV module, as illustrated in Figure 2 (Vidyanandan, 2017). The short circuit current
(ISC), which is also marked in Figure 2, is the maximum current that is produced and at this
point, the voltage is zero. Open circuit voltage (VOC), which can also be observed in Figure 2,
is the maximum voltage that is delivered and it corresponds to the point where the current is
zero (Alternative Energy Tutorials, 2018).

4
Figure 2 The I-V and P-V curve of a module where the short circuit current (ISC), open circuit
voltage (VOC), maximum power point (MPP), current at maximum power point (IMP), and
voltage at maximum power point (VMP) are marked. (Sfirat, Gontean, & Bularka, 2018)

Bentgsson et al. (2017) mention that the advantage of string inverters with MPPTs is the
limited number of components and the lower cost. The drawback is that solar cells will
operate at the MPP that applies to the entire string or system. This sometimes means that
individual modules will work outside their MPP which contributes to losses in the system
(Bengtsson et al., 2017). One way to minimize losses according to Saavedra (2016) is to use
module optimizers (DC/DC optimizers) in combination with a string inverter. Module
optimizers can be connected so that they find the MPP for individual modules, for separate
loops in modules or, for small groups of modules. When each optimizer has found their MPP,
they communicate with each other to regulate the current through the modules to a common
level (Saavedra, 2016).

5
2 LITERATURE STUDY
While Section 1 provided a brief introduction to grid-connected PV systems and the related
components, as well as the importance of performance evaluation in field operation, this
section lays the theoretical foundation for the performance evaluation of the eight PV
systems. The impact of various factors on the performance of a PV system and their
relations to each other are included in Section 2.1. Furthermore, relevant performance
parameters for the evaluation of the PV systems are described in Section 2.2.

2.1 Factors affecting PV system performance


When studying related literature, one can note that several factors have a determinant impact
on the performance of a PV system. Some of these factors have positive effects on the
performance while others have negative effects. This section will only deal with those factors
which lie in the scope of this study such as plane of array irradiance, ambient temperature,
module temperature, module technology, installation method, inverter efficiency, and
inverter size.

2.1.1 Plane of array irradiance


The term ‘’global solar radiation’’ is used to quantify the amount of solar radiation received
at a horizontal surface on earth (Lindahl et al., 2019). However, in PV performance analysis
and modeling of PV arrays, the term plane of array irradiance is generally used. This term
quantifies the incident irradiance on a PV array and is composed of direct, diffuse, and
reflected solar radiation as presented in Figure 3 (PVPMC, n.d.). The direct radiation shines
directly on the PV panel, and according to Bengtsson et al. (2017) it dominates during sunny
days. Diffuse radiation is direct radiation that has been scattered in the atmosphere by
clouds, molecules, and particles. Thus, the diffuse radiation does not come from a specific
direction but the whole sky and is the highest during cloudy days. Solar radiation which
bounces on something else, such as the ground before it reaches the panel is referred to as
reflected radiation and it is especially high when there is snow or other reflective surfaces
(Bengtsson et al., 2017; Bengts nya villablogg, 2015).

Figure 3 Plane of array irradiance consists of direct, diffuse and reflected solar radiation.

6
Solar radiation depends on the air mass, weather conditions, location on earth, season of the
year, and time of the day (Lindahl et al, 2019; Energy Efficiency and Renewable Energy,
n.d.). The annual average global radiation over Sweden is presented in Figure 4. One can see
from Figure 4 that the intensity of global solar radiation is higher for southern Sweden
compared to the northern part of the country (SMHI, 2017). The amount of solar irradiation
received by a PV panel can be carefully measured by installing a pyranometer at the same tilt
angle as the PV panel (PVPMC, n.d.).

Figure 4 Distribution of annual average global radiation over Sweden. (SMHI, 2017)

According to Platon et al. (2015), the power output of a PV panel and the plane of array
irradiance have a strong linear relation. As the intensity of the solar irradiance increases, the
electrical power output from the panel also increases (Platon et al., 2015; Sauer et al., 2007).
This is also illustrated in Figure 5, where the increase in solar irradiance at constant ambient
temperature causes the current and power for the same voltage to increase (Depuru &
Mahankali, 2018). Several studies have also shown that, as ambient temperature increases
with the increase in solar radiation, the efficiency of solar radiation conversion into electricity
is reduced due to temperature increase in the PV cell (Kurpaska et al., 2018; Rani, Giridhar,
& Prasad, 2018). This will be explained more in detail in Section 2.1.2.

7
Figure 5 I-V and P-V curve for three irradiance levels at constant ambient temperature. The
dashed line represents the I-V curve while the lines represent P-V curve. (Depuru &
Mahankali, 2018)

2.1.2 Module temperature


The efficiency of a PV module defines how much of the received irradiance it can convert into
electrical energy. As a result of the advancement in the PV industry during recent years, the
conversion efficiency for PV modules has improved. Nowadays, efficiencies for PV modules
can even lie around 20% or slightly higher (Solar Reviews, 2021; NREL, n.d.). However, since
the PV modules have limited efficiency, the greater part of the irradiance is still converted
into heat (Duffie & Beckman, 2006; Siecker et al., 2017). The conversion to heat causes the
cell temperature to increase, which in turn contributes to an increase in current and a
reduction in voltage. Since the voltage decreases faster than the current increases, the power
output will therefore decrease. (Quaschning, 2005; Depuru & Mahankali, 2018)

The module temperature is to a great extent affected by other environmental factors such as
ambient temperature, irradiance, and wind (Woyte et al., 2014). More specifically, an
increase in the module temperature usually occurs due to high ambient temperatures and
high solar radiation (Kurpaska et al., 2018; Rani, Giridhar, & Prasad, 2018; Moharram et al.,
2013). A study conducted by Zdyb and Golkowski (2020) on four different module
technologies showed that the module temperature increased during high irradiance levels
because of an increase in ambient temperature and direct heating of the surface of the
module. Depuru and Mahankali (2018) state that high solar irradiation and low ambient
temperature is the most preferred environmental condition for a PV system. Figure 6
exemplifies how an increase in ambient temperature contributes to a decrease in the power
and voltage when the irradiance is constant.

8
Figure 6 I-V and P-V curve at three ambient temperatures and constant irradiance. The dashed
line represents the I-V curves while the lines represent P-V curves. (Depuru & Mahankali,
2018)

It is evident that the operating temperature of a PV module has a key role in the performance
of a PV system; and the temperature coefficient is therefore established for several
parameters of the module such as power, current, and voltage. The temperature coefficient
indicates the change in these parameters with respect to the temperature of the cell and these
are assumed to behave linearly. Manufacturers usually provide them as a reference for their
products. (King, Kratochvil, & Boyson, 1997)

According to King et al. (2004), the temperature of a PV module is normally measured by


placing a sensor on the back of the module. The disadvantage is that the temperature of the
module and that of the cell inside the module differs, which increases the errors in the
measurements (King, Boyson, & Kratochvil, 2004). Furthermore, the wind speed can also
contribute to a temperature difference higher than 5 °C among various modules in the same
system. Therefore, when monitoring the performance of a PV system, several sensors should
be placed in different positions on the PV modules so that the average temperature can be
determined accurately. (IEC, 2016)

2.1.3 Influence of module technology on PV system performance


The performance of PV systems is to a great extent affected by the efficiency of the module
technology (Kovacs, 2019). Modules with higher efficiency can generate more energy per
square meter of roof space and they can have higher energy generation over the lifespan of
the system (Solar Reviews, 2021). A higher module efficiency also lowers the costs of the rest
of the system (non-module components) on a per-watt basis (Wang & Barnett, 2019).
The efficiencies for the most common module technologies are presented in Table 1. As
observed, mono-Si and poly-Si have higher efficiencies than the thin film technologies.

9
Table 1 Efficiency of the most common module technologies. (Kovacs, 2019)

Module technology Efficiency [%]


mono-Si 18-22
poly-Si 16-20
CdTe 14-18
CIGS 14-16
a-Si 8-12

The PV modules are temperature-dependent regardless of technology (Jankovec & Topic,


2012). However, according to several studies, there are differences in how various module
technologies are affected by increasing module temperatures (Dubey et al., 2013; Adeeb et
al., 2019; Ye et al., 2014; Cañete et al., 2014). Thus, it is vital to choose an appropriate module
technology by taking into account the conditions of the specific location and its average
ambient temperature (Adeeb, Farhan, & Salaymeh, 2019). Since the thin film cells are
thinner than c-Si cells, the heat can be conducted more easily from the solar module,
compared to crystalline silicon solar cells (Toivola , 2010). A study conducted by Zdyb and
Golkowski (2020) on four different module technologies in Poland showed that the efficiency
of the poly-Si and CIGS modules decreased with rising temperature, while the efficiency of
the a-Si and CdTe modules were more stable. Kumar and Sudhakar (2019) did also note that
the performance of the CdTe module was less affected by an increase in ambient temperature
and irradiance compared to CIGS and c-Si modules.

2.1.4 Influence of installation method on PV system performance


According to Gonçalves et al. (2019), heat exchange to the exterior is significantly reduced for
BIPV systems, compared to the BAPV equivalent, since the modules are integrated into the
building. As a consequence, module temperatures can get as high as 80 °C depending on if air
gaps are present or not to provide natural ventilation (Woyte et al., 2014; Gonçalves et al.,
2019). Usually, a gap is created between the building element and the PV panel for BAPV
system to avoid extremely high temperatures. Ambient air is then circulated in this gap which
enables the removal of the produced heat. (Kalogirou, 2014).

A study by Kumar and Sudhakar (2019) showed that there were some variations in the
performance of BAPV and BIPV systems. It was found that the energy yield and the
performance ratio were somewhat higher for BAPV systems than for the BIPV systems, but a
significant difference in the performance of the systems could not be found.

2.1.5 Solar PV Inverter


During the conversion from DC to AC power, there are always losses. Thus, the more efficient
the inverter is, the more useful electrical energy it will generate (Bengtsson et al., 2017).
There are various ways to describe the efficiency of inverters. As illustrated in Figure 7, the
inverter efficiency varies with the output power. At very low output power, a substantial
decrease in the inverter efficiency can be noted.

10
Figure 7 Inverter efficiency as a function of DC power for three different input voltages. The
efficiency curve belongs to the Sunny Tripower 10.0 inverter. (SMA, n.d.)

The efficiency of an inverter can also be described with the European efficiency, which is the
weighted number considering how frequently the inverter operates at various power outputs.
The European efficiency is defined as (PVsyst, n.d.; Salam & Rahman, 2014):

𝜂!" = 0.03𝜂#% + 0.06𝜂%&% + 0.13𝜂'&% + 0.10𝜂(&% + 0.48𝜂#&% + 0.20𝜂%&&% eq.1

where 𝜂)% represents the inverter efficiency when operated at i% of its rated capacity in terms
of DC input. The factor which is multiplied with 𝜂)% is called the weighting factor and is
highly influenced by irradiance data. (Salam & Rahman, 2014)

The curve for the European efficiency for the same inverter as in Figure 7 is presented in
Figure 8. The performance of the inverter at the optimal power output can also be described
with the peak efficiency which is typically used as a criterion of the inverter’s quality.
Nevertheless, the European efficiency is more usable than the peak efficiency since it shows
how the inverter performs at different output levels during sunny days. (Mertens, 2014)
In Figure 8, one can see that the peak efficiency for this inverter is 98%.

11
Figure 8 European efficiency of an inverter as a function of voltage at MPP. The curve for
European efficiency belongs to the Sunny Tripower 10.0 inverter. (SMA, n.d.)

The size of the inverter in relation to the size of the PV array can be described with the DC-to-
AC ratio and defined as the ratio of the nameplate DC power to the inverter nominal AC
power (Aurora Solar, 2021). The selection of an appropriate inverter size in a PV system is
considered important since errors in the size of the inverter can constrain the performance of
the entire PV system (Usman, Tah, Abanda, & Nche, 2020). According to Jordan et al.
(2020), errors in the sizing of the inverter are a common reason for poor performance of an
installed PV system. Danoune et al. (2019) claim that the inverter size should be around 80–
90% of the PV array (DC-to-AC ratio of 1.25—1.11). This is because PV systems hardly
generate their maximum required power, mainly due to the effect of increasing module
temperature and low irradiance levels (Danoune et al., 2019; Partlin, 2015). Kratzenberg et
al. (2014) and Madkor et al. (2015) explain that the climatic conditions, the location where
the inverter is installed, technical characteristics of the inverter, and the photovoltaic
modules are some factors that should be considered when choosing inverter size. One of the
benefits of installing an undersized inverter is that a lower cost of delivered energy can be
realized (Partlin, 2015). Additionally, if the inverter is oversized, higher losses occur as the
inverter converts more energy at the lower power ranges, see Figure 7 for the inverter’s
efficiency characteristics as a function of power. Sangwongwanich et al. (2017) and
Kratzenberg et al. (2014) also emphasize that if the PV array is oversized in relation to the
inverter, the power from the array will be curtailed at the rated inverter power level during
periods with high irradiance levels and this will result in a loss of energy yield.

2.2 Performance metrics


A schematic illustration of the energy flow in a grid-connected PV system along with the most
important parameters is presented in Figure 9. The irradiance received by the plane of array
is defined as GPOA while the terms PDC and PAC describe the DC and AC power outputs,
respectively. As mentioned in Section 2.1.2, the module temperature (Tmod) and the ambient
temperature (Tamb) have a significant influence on the performance of a PV system, thus these
parameters are also included in Figure 9. It can also be observed that losses will occur both
before and after the inverter, which will affect the values of the performance parameters.

12
Figure 9 Energy flow in a grid- connected PV system where solar irradiance is received by the PV
array which are later on converted into AC power. Performance parameters such as Yr,
Yf and PR can be used to determine the performance of a PV system at various levels.

As Figure 9 illustrates, it is possible to do several yield comparisons for sub-systems. The


reference yield (Yr) defines the source of solar radiation for the PV system. The value of Yr
depends on the location of the system as well as the orientation of the PV array and the
weather variability (Shiva Kumar & Sudhakar, 2015). It is defined as the GPOA divided by the
reference irradiance (Gref) at STC (1000W/m2) (Nordmann et al., 2014):

G+,-./ eq. 2
Y* = (hours)
G+,*01

The instantaneous final yield (Yf ) is the energy generation normalized with respect to the
system size (Nordmann et al., 2014). Since Yf is normalized to system size, it is useful when
comparing the energy generation from PV systems of different sizes. The Yf represents how
many hours the PV system must be in operation at nameplate DC power to generate the same
amount of energy (Chioncel et al., 2010). Yf, is defined as the net energy output divided by the
nameplate DC power of the PV system measured under STC (Shiva Kumar & Sudhakar,
2015):

P/2 kWh eq.3


Y1 = 7 : or (hours)
P& kW

Specific energy yield (kWh/kWp) is defined as eq.3 but calculated for longer periods such as
monthly or annually. Specific energy yield is one of the most commonly used performance
metrics for PV systems of all sizes. Specific energy yield defines how much energy (kWh) is
produced for every kW of installed DC power. The specific energy yield of a PV module
technology should be determined for the long-term, preferably for one year to include the
seasonal effects. This metric is an appropriate way of measuring the performance of PV

13
systems because the generation is proportional to the systems potential gaining while the
installed power reflects the cost of the system. (Theristis et al., 2018)

2.2.1 Performance ratio (PR)


According to Woyte et al. (2014) the performance of PV systems is typically evaluated by
computation and analysis of the performance ratio, PR. PR is an internationally introduced
performance metric, which determines how effectively the PV system converts the captured
irradiance into AC energy, compared to what would be expected from the nameplate rating of
the PV system (Woyte et al., 2014). Dierauf et al. (2013) defines the PR as:

Y1 ∑P/2 eq.4
𝑃𝑅 = =
Y* G
∑[P& ∗ G-./,+ ]
*01,+

t = Point in time
PAC = Measured AC electrical generation (W)
P0 = Nameplate DC power (W)
𝐺345,6 = Measured POA irradiance (W/m2)
Gref, t= Irradiance at standard test conditions (1000 W/m2)

Woyte et al. (2014) mention that PR also quantifies the overall effect of losses, for instance,
losses due to inverter inefficiency, component failures, and cell mismatch. Marion et al.
(2005) add that PR values are typically calculated on a monthly or yearly basis. However,
values calculated for shorter time intervals, such as weekly or daily, may be useful for
identifying occurrences of component failures (Marion et al., 2005).

The PR value is affected by the ambient and module temperature, thus the PR changes
seasonally (Marion et al., 2005). Both Nordmann et al. (2014) and Marion et al. (2005)
clarifies that PR values are usually greater in the winter than in the summer because of a
significant decrease in the ambient and module temperature. However, seasonal soiling of
the PV modules could also cause differences in PR from summer to winter (Marion et al.,
2005). An investigation of the effect of temperature on the performance metrics of a PV
systems was conducted by Prakhya and Kotha (2018). The study showed that the PR reduced
significantly during March to June, as the average ambient temperature was above 29 °C
during this period (Prakhya & Kotha, 2018).

The PR value is highly independent of the site-specific solar radiation conditions and
therefore also independent of module orientations (Reich et al., 2012). Marion et al. (2005)
point out that a system with low PR in a location with high solar radiation might produce
more energy than a system with a high PR in a location with low solar radiation.
Consequently, the PR values do not represent the amount of energy produced and they can
therefore be used to compare various system designs or even equal system designs at
different locations. PR value also enables a comparison of the same system over several years,
and if a reduction of yearly PR values is noted, it might indicate a permanent loss in
performance (Marion et al., 2005).

14
The closer the PR value is to 100 %, the more efficient system is operating. However, in real-
life, a value of 100 % cannot be achieved because unavoidable losses always arise with the
operation of the PV systems (Woyte et al., 2014). Consequently, calculating the PR by
considering the rated maximum power at STC may not give exact results (Prakhya & Kotha,
2018).

2.2.2 Weather corrected performance ratio (PRcorr)


Usually, the PR for PV systems is evaluated by using eq.4. However, a drawback is that this
PR does not consider the cell’s temperature-dependency. Including this effect on PV module
output performance is essential since the power changes with the variation in the module
temperature. The PRs strong dependence on temperature can result in seasonal variations in
PR, these variations can be between ±2% to ±10% (Dierauf et al., 2013). As a result of the
temperature effect, Nordmann et al. (2014) suggest a correction of the PR definition, where
the influence of temperature is considered and where the seasonal variation of the PR is
reduced. By including the temperature-dependency the seasonal variability of the PR metric
is removed, without changing the PR value stated in the contract. Moreover, the difference
between the regular PR and the weather-corrected PR is that the latter contains a term to
translate power to the average operating module temperature. (Dierauf et al., 2013)

According to Nordmann et al. (2014), the PR is often corrected to standard temperature of


25 °C (STC) as shown in eq.5.

∑𝑃5: eq.5
𝑃𝑅7899 =
𝐺345,6 𝛾
∑(𝑃& ∗ 𝐺 )(1 − 100 (25 − 𝑇=8>,6 ))
9;<,6

Where, γ is the relative maximum-power temperature coefficient (in units of °C−1) that
corresponds to the type of installed module, Tmod,t is the cell temperature (in °C) in time step
t. The temperature coefficient of the power (𝛾) is the parameter that includes the effect of
temperature on the module output power.

While correction to 25 °C essentially solves the problem of seasonal variations, it may not be
justified as temperature varies constantly and in reality, modules do not generate power at 25
°C the whole year (Nordmann et al., 2014). Therefore, NREL suggests a more accurate
equation that includes site-dependent average module temperature, which the PR can be
corrected to, as shown in eq.6.

∑𝑃5: eq.6
PR ?@** =
𝐺 𝛾
∑(𝑃& ∗ 𝐺345,6 )(1 − 100 (𝑇=8>,ABC − 𝑇=8>,6 ))
9;<,6

Here, Tmod,avg is the site-dependent average annual cell temperature that is irradiance-
weighted and defined as:

15
∑HG-./,+ ∗ 𝑇=8>,6 I eq.7
TD@E_GHI =
∑G-./,+

Dierauf et al. (2013) presented Figure 10 where the weather corrected PR is compared to the
noncorrected PR during a year using eq.6.

Figure 10 Comparison of annual noncorrected PR and weather corrected PR on simulated data.


(Dierauf et al., 2013)

The reason it is called weather-corrected and not temperature-corrected is that this corrects
PR for most of the weather-related effects. The operating module temperature accounts for
the effects of both the ambient temperature, wind, and also heating from the sun. (Dierauf et
al., 2013)

16
3 METHOD
This section describes the methodology which is applied for the performance evaluation of
the eight PV systems. The different steps of the method are presented in Figure 11.
Performance data from the PV systems as well as ambient condition variables are recorded
continuously. The acquisition of the data will be described in the first section. The second
section concerns the procedures for data filtering and the last part clarifies how the
performance evaluation of the PV systems was completed.

1. Data acquistion 2. Data filtering 3. Performance ----


----parameters for -
----PV systems

•Module •Removal of •PR


temperature (Tmod) outliers, linear
interpolated •Specific energy
•Ambient yield
temperature (Tamb) values, and
abnormal power •Energy
•Plane of array generation per
irradiance (GPOA) generation
•Replacing area
•Power out from •Temperature
inverter (PAC) missing values
for GPOA, Tmod, behavior
and Tamb.

Figure 11 A schematic flow chart illustrating the steps which has to be accomplished for the
performance evaluation of the PV systems.

3.1 Data acquisition


The data acquired for the performance evaluation of the PV systems included measurements
of module temperature (Tmod), ambient temperature (Tamb), plane of array irradiance (GPOA),
and power generation after the inverter (PAC). Since all PV systems were not installed on the
same roof, there were two different data logging systems. One of the systems was a wireless
solution that collected data from six of the PV systems. Another logger collected data from
two other PV systems that were positioned on a different building. The temperatures were
measured with PT100 sensors and thermoelements. Although the temperature sensors had
high accuracy, the measurements will give higher errors since the sensors are highly affected
by their position on the module. The sensors were glued on the module; hence, it can also
affect the degree of error in the measurements. It is also worth mentioning that the sensors
measure the module temperature rather than the solar cell temperature. Thus, the values will
not be completely correct. As a result of these error sources, the temperature sensors were
assumed to have an uncertainty of ±0.5K.

The in-plane irradiance was measured with both pyranometers and silicon sensors with an
uncertainty of ±2%. The energy outputs were measured using energy meters with 1000
pulses per kWh and an uncertainty of ±1%. Data for irradiance and energy were acquired
with a temporal resolution of 30 seconds, while the data for temperature were acquired with
60 second resolution.

17
The data acquired from the two different loggers were later on synchronized into a common
time vector with a frequency of 15 seconds and missing data points were replaced with linear
interpolated values. The frequency of 15 seconds was chosen to avoid possible time shifts
with the two data loggers.

3.2 Data filtering


According to Livera et al. (2020b), faulty data usually emerge in field measurements. The
most common sources of faulty data are power outages, interruption f0r maintenance, and
component and communication failure. Faulty data can arise as missing data or outlying
values (Livera et al., 2020a). Kotu and Deshpande (2019) describe that an outlier is
anomalous data with respect to the majority of the dataset. Invalid data points can
significantly bias the results of the databased analysis. Therefore, these data points must be
identified and removed, and this can be done automatically or manually. (Livera et al.,
2020a)

3.2.1 Identification and removal of faulty data


In this study, faulty data such as outliers and linear interpolated values (originally missing
data points) were identified and removed in MATLAB. The methodology for identifying and
removing invalid data points are based on a number of steps from the IEC 61724 standard
(2016) and other PV data quality reports (Livera et al., 2020a; Li et al., 2020; Platon et al.,
2015). The methods used are summarized below:

• Comparisons among different sensors and visual inspection of plots:


Simplifies the identification of faulty data by visually reviewing the dataset.
• Removal of night-time data: Removes extraneous data values that could
introduce errors in analyses.
• Application of threshold ranges: By setting minimum and maximum values for
each parameter, outliers can be identified and removed.
• Plotting the power as a function of irradiance: Simplifies the identification
and removal of data not representative for a normal PV system operation since power
and irradiance have a strong linear relationship (Platon et al., 2015). A linear
regression can be performed on the dataset. The coefficient of determination (R2) is
determined before and after filtering to show the improvement of the irradiance-
power linear relationship. The formula for coefficient of determination is defined as
(Alm & Britton, 2014):
∑KJL%(xKJ − xLJ )' eq.8
𝑅' = 1 − K
∑JL%(xJ − xLJ )'

Where, 𝑥K is the predicted value of 𝑥, 𝑥̅ is the mean of 𝑥, and 𝑛 is the number of


observations.
• Application of Hampel identifier: There are several suitable statistical methods
available. One of the methods recommended by Livera et al. (2020a) is the Hampel
identifier which is also called Hampel filter. The Hampel identifier can be applied to

18
any data set and there is a built-in function for it in MATLAB. This statistical method
uses the median value and the median absolute deviation (MAD) to identify outliers.
If xi is the sample which is checked and the number of neighbors on either side of the
sample xi is k, then the moving window lengths are 2k+1, and the local median (mi) is
determined as (Pearson et al., 2015; MATLAB, 2021):

𝑚) = 𝑚𝑒𝑑𝑖𝑎𝑛(𝑥)MN , 𝑥)MNO% , 𝑥)MNO' , … 𝑥) , … , 𝑥)ONM' , 𝑥)ONM% , 𝑥)ON ) eq.9

Afterward, the standard deviation 𝜎) is determined as follows:

eq.10
𝜎) = 𝑘 ∗ 𝑚𝑒𝑑𝑖𝑎𝑛(|𝑥)MN , −𝑚) |, … , |𝑥)ON − 𝑚) |)

where k is the unbiased estimation of the Gaussian distribution and defined as:
%
𝑘= ≈ 1.4826
√';9<7 !" %/'
The median absolute deviation (MAD) is equal to the quantity 𝜎) /𝑘.
The sample xi is checked for a certain threshold 𝑛R to determine if it is an outlier or
not:

xJ , |xJ − mJ | < nS σJ eq.11


yJ = ]
outlier, | xJ − mJ | > nS σJ

If xi is declared as an outlier, it is replaced with mi.

3.2.2 Treatment of missing data


Recommendations for treatment of missing or invalid data can be found in the IEC 61724
standard (2016). Various methods for replacing missing data are recommended and it is also
mentioned that the analyzed interval can be treated as missing or invalid. In this study,
missing data points for module temperature, ambient temperature, and irradiance sensors
were replaced with data from other sensors (if available). If sensors were not available, it was
treated as missing data. Unlike the other parameters, any attempt to replace missing values
for the power were not made in this study, instead it was treated as missing data.

3.3 Performance parameters for PV systems


Several performance parameters such as specific energy yield and PR were used in this study
to determine the overall performance of each PV system. These parameters were thoroughly
explained in Section 2.2 and the equations used to determine these parameters were also
presented in the same section. Since the value of noncorrected PR explained in Section 2.2.1
varies with season, eq.5 was used to remove the influence of temperature. However, since the
modules do not always produce power at 25 °C, this equation was complemented with one
additional equation for weather corrected PR. In this equation, the PR is corrected to the
average module temperature instead of 25 °C. Aside from these parameters, the monthly
specific energy yield in (kWh/kWp), the energy generation per area (kWh/m2) were also

19
considered an important part of the performance evaluation. Moreover, the temperature
behavior for each system was also considered relevant for the analysis since this parameter
has a significant impact on the overall performance of each system. Thus, the filtered data for
irradiance, module temperature, and ambient temperature were of interest and were
presented as a function of time to find relationships and connections between these
parameters. The measured data for the performance evaluation of the PV systems do not
include nighttime data since these values are considered irrelevant.

Various combinations of parameters were plotted in MATLAB to see how they were related to
each other and also how the performance was affected by various conditions. The Pearson’s
correlation coefficient was also determined for various parameters, as a measure of the linear
association between two variables, ranging between +1 to -1. A value of “0” indicates that
there is no linear association between the two variables. A value greater than “0” indicates a
positive association. That is, as the value of one variable increases, so does the value of the
other variable. A value less than “0” indicates a negative association; that is, as the value of
one variable increases, the value of the other variable decreases. The Pearson’s correlation is
defined as (Berman, 2018):

∑V)L%(𝑥) − 𝑥̅ )(𝑦) − 𝑦L) eq.12


𝑟TU =
i∑V)L%(𝑥) − 𝑥̅ )' i∑V)L%(𝑦) − 𝑦L)'

where rxy is the Pearson’s correlation coefficient, n the sample size, 𝑥̅ the sample mean, and
analogously for 𝑦L.

20
4 CURRENT STUDY
In spring 2020 RISE installed eight small-scale PV systems at their research facility in
Borås to evaluate the performance of these systems on account of the Swedish Energy
Agency Test Lab. The choice of PV systems was made to represent the residential market in
Sweden. In the following section, a description of the PV systems and their respective
components and technologies are presented. Furthermore, this section includes an overview
of the measured parameters and the number of sensors used for measuring each
parameter. In addition to that, the identification and removal of faulty data are presented
and visualized for each parameter by applying the methodology explained in Section 3.

4.1 Description of the systems


The selected PV systems have a unique combination of components and technologies. The
systems differ in terms of the type of module technology, appearance, and installation
method. Module technologies like CIGS, CdTe, as well as Mono-Si, are included and there are
both BIPV and BAPV systems. One can also observe from Figure 12 that the module surface
of system 1 is designed as waves while some systems have a flatter module surface. All of the
systems were positioned to avoid shading and reflection from surrounding objects. Six of the
test systems are placed on Roof 1 as shown in Figure 12. They are installed at a higher altitude
compared to systems 7 and 8 which are installed on Roof 2, these two systems can be
observed in Figure 13.

Figure 12 PV system 1-6 on Roof 1.

21
Figure 13 PV system 8 and 9 on Roof 2.

System 8 which can be observed in Figure 13 is mounted directly on the roof of a loft and is
highly affected by the temperature inside the loft. Furthermore, there is an air gap of 5-10 cm
between the modules of system 7 and the roof, whereas this air gap does not exist for system
8 since it is a BIPV system.

The PV systems in this study either have an optimizer at module level or string level.
Additionally, all the PV systems have a peak power ranging between 3–5 kWp. The systems
are oriented due south (180◦ azimuth in SAM, System Advisor Model) and with a 45◦ tilt
angle to the horizontal. The latter since this represents a suitable angle at that altitude to
maximize the annual output. The specifications for all systems are presented in Table 2.

22
Table 2 Specification of the eight PV systems.

PV system

1 2 3 4 5 6 7 8

Nameplate
DC power 3080 2975 3900 3000 4080 3245 4800 3289
[Wp]:

System area
28.3 26.6 19.8 38.2 24.3 21.2 25.6 31.5
[m2]:

Specific
system area 9.19 8.94 5.08 12.7 5.96 6.53 5.09 8.33
[m2/kWp]:

Module
CIGS CdTe Mono-Si CIGS Mono-Si Mono-Si Mono-Si CIGS
technology:

Module
efficiency 2 10.88 11.18 19.67 7.85 16.79 15.31 18.75 10.44
[%]:
Temperature
coefficient of
-0.33 -0.21 -0.37 -0.35 -0.41 -0.42 -0.37 -0.33
power
[%°C]3 :
Inverter
(nom. AC- 3000 3000 4000 5000 4000 3000 4000 3000
power) [W]:

DC-to-AC
1.03 0.99 0.98 0.6 1.02 1.08 1.2 1.10
ratio4:

Inverter
Euro 96.2% 96.2% 97.5% 97.3% 97.3% 96.5% 97.3% 96.7%
efficiency 5:

Module
No No No Yes Yes No Yes No
optimizer:

BIPV/ BIPV/
Mounting: BIPV BAPV BIPV BIPV BAPV BIPV
BAPV 6 BAPV 6

2 Calculated from the formula: P=ηAG. (Stridh, et al., 2020)


3 Refers to the effect of the temperature on the power output and is taken from the datasheet of the
respective PV modules.
4 DC-to-AC ratio is defined as the nameplate DC power over the inverter nominal AC power. (Aurora

Solar, 2021)
5 The Inverter Euro efficiency for respective system is taken from the manufacturer's website.

6 Although the system aesthetically looks like BIPV it does not replace any building material, thus it

does not qualify for the definition of building-integrated solar cells according to IEA PVPS Task 15
(Berger, et al., 2018). In this study, this type of systems will be defined as BIPV/BAPV.

23
4.2 Measured parameters
The measured parameters for the performance evaluation of the eight PV systems are:

• Module temperature (°C)


• Ambient temperature (°C)
• Plane of array irradiance (W/m2)
• AC Power (W)

Table 3-Table 5 show the number of sensors measuring each parameter. As observed in Table
3, at least 3 sensors are measuring the module temperature for each system except for system
7 and 8 which has 1 and 2 sensors, respectively. For the performance evaluation, the mean
values for all sensors which are considered to give accurate values are determined and used
for each system.

Table 3 Number of sensors measuring the module temperature for each system.
Module temperatures
PV system sensors 1 sensors 2 sensors 3 Sensor 4
1 Tmod1, 1 Tmod1, 2 Tmod1, 3
2 Tmod2, 1 Tmod2, 2 Tmod2, 3 Tmod2, 4
3 Tmod3, 1 Tmod3, 2 Tmod3, 3
4 Tmod4, 1 Tmod4, 2 Tmod4, 3
5 Tmod5, 1 Tmod5, 2 Tmod5, 3
6 Tmod6, 1 Tmod6, 2 Tmod6, 3
7 Tmod7, 1
8 Tmod8, 1 Tmod8,2

The ambient temperature is measured in different positions, and as shown in Table 4, the
sensors on Roof 1 are relevant for the systems on that roof. When evaluating the performance
of system 1-6, the mean value of all sensors on Roof 1 is calculated and used. The two other
sensors on Roof 2 are relevant for system 7. Sensor 5 is placed in a loft and is only relevant
for system 8.

Table 4 Number of sensors measuring the ambient temperature.


Roof Ambient temperature sensors Applied to PV System
Tamb, 1
1 Tamb,2 1-6
Tamb,3
2 Tamb, 4 7

2 Tamb, 5 8

The irradiance is also measured in different positions with three sensors. The three different
radiation sensors are located at the systems within a radius of 100 m from each other.
It can be seen in Table 5 which sensor is applied to which system for the performance
evaluation of the PV systems.

24
Table 5 Number of sensors measuring the plane of array irradiance.
Plane of array
Roof Applied to PV System
irradiance sensors
1 GPOA1 1, 3, 4, & 5
1 GPOA2 2&6
2 GPOA3 7&8

4.3 Identification of faulty data


The first step is to identify faulty data by plotting the data measurements for the various
parameters as a function of time and visually inspecting the outcome. As mentioned in
Section 4.2, the parameters are usually measured with more than one sensor. For instance,
the module temperature of system 5 is measured with 3 different sensors on different
positions, which is shown in Figure 14. Since the measurements are affected by various
factors such as irradiance, ambient temperature, wind speed, and position of the sensors,
there will be natural differences in the measurements. This also applies to measurements of
the irradiance and ambient temperature. However, if there are significant differences in the
measurements or if the behavior of the measured temperature changes for a specific time
period it could indicate that it is faulty data. By comparing the sensors, it is possible to get an
overview of the type of erroneous data points that could arise and thereby making it easier to
develop of a code that identifies and removes the faulty data.

Figure 14 Position of sensors measuring the module temperature on system 5.

4.3.1 Linear interpolated values and outliers


After visually inspecting the data as a function of time, erroneous data points are detected. If
the data is linear for a long time period it may be linear interpolated values instead of actual
data, thus it should be removed from the dataset. Figure 15 shows how linear interpolated
values look like for the module temperature of system 1. From the figure it can be observed
that both sensors 1 and 3 are linearly interpolated for approximately one and a half days
between the evening on July 1st until the morning of July 3rd. From the same figure, it can
also be observed that sensor 3 gives a too high value on the 5th of July after midnight. This is

25
a local outlier and should also be removed from the dataset.

Figure 15 Linear interpolated values and local outliers for the module temperature of system 1
detected.

Global outliers which are more visible should also be removed. Figure 16 shows how global
outliers can look like for the module temperature of system 1.

Figure 16 Global outliers for module temperature of system 1 detected.

Figure 15 and Figure 16 are showed as examples. Some of these faulty data could also be
identified in the measurements for the ambient temperature, irradiance, and power.

26
4.3.2 Abnormal power generation
Visual inspection of the irradiance-power plot was deemed sufficient for detecting additional
outlying values and other faulty data. Under normal conditions, the solar irradiance and
power measurements follow a strong linear relation. Observations close to the irradiance–
power linear relationship line are considered as normal operation measurements, while
observations far from this line are considered as faulty operation data. Figure 17 shows the
power generation as a function of irradiance for system 2 and the data points which lie within
the red marked areas are examples of faulty data.

Figure 17 Identification of faulty data for system 2 when power generation is plotted as a function
of plane of array irradiance. Data points which lie within the red marked areas are
examples of faulty data.

4.4 Removing faulty data


The following section will describe how faulty data is removed and replaced with “NaN” in
MATLAB. Different filtering methods are used depending on the type of faulty data. As a final
step, the measured parameters are plotted as a function of time and inspected for the whole
time period. This is done to make sure that all data are filtered properly.

4.4.1 Removal of global and local outliers


To detect and remove global outliers, minimum and maximum values are determined for
each parameter. In Table 6, the minimum and maximum values for each parameter can be
observed. The values are based on assumptions relevant for conditions in Sweden and are
valid all year round. As observed from Table 6, the minimum plane of array irradiance is set
to 20 W/m2 to remove nighttime data as recommended in IEC 61724 standard (2016). The

27
maximum power for the PV system is defined as the installed power multiplied by a factor of
1.2, the factor 1.2 is based on an assumption.

Table 6 Minimum and maximum values for each measured parameter.

Parameters Minimum and Maximum values


Module temperature -25<Tmod<80
Ambient temperature -25<Tamb<40
Power 0 ≤ PAC< Nameplate DC power ×1.2
Plane of array irradiance 20 <GPOA<1200

These limits are useful for removing global outliers; however, they cannot eliminate local
outliers, so, additional methods are necessary. The built-in function for Hampel filtering,
introduced in Section 3.2.1, was used to remove and replace local outliers for the
temperatures with more accurate values. For this specific case study, k (number of neighbors)
was set to 40 while MAD was set to 3 in eq.11. An example of how the Hampel function
removes and replaces local outliers for sensor Tmod1,3 is illustrated in Figure 18.

Figure 18 Removal of a local module temperature outlier with the Hampel filter for system 1.

The Hampel filter is not used for the power and irradiance to avoid the removal of good data
points. The irradiance and power differ from the temperatures in the sense that they change
drastically with time, thus, the risk is that the Hampel filter will remove and replace values
that were initially good data points.

28
4.4.2 Removal of linear interpolated values
Linear interpolated values were removed for the different variables by developing a code.
If the derivative is constant for a specified period, the code will set the data points to “NaN”
for this interval. The time periods chosen for the parameters are also based on assumptions.
Since the irradiance and power change drastically, the time period chosen for these two
parameters is 10 minutes. The time period chosen for the temperatures is 50 minutes since
the temperature change is not as drastic as the changes in power and the irradiance. Figure
19 shows how a set of data points are removed for a module temperature of system 1 when it
is linear over 50 minutes.

Figure 19 Removing linear interpolated values for module temperature of system 1.

4.4.3 Checking deviation from mean and removing invalid data


An additional step is to filter the data for the module temperature, ambient temperature, and
irradiance manually. For the module temperature, this is done by determining the mean for
all sensors on the same system at timestep t as shown below:
W#$%&," (6)O W#$%&,( (6)O W#$%&,& (6)
𝜇 ( (𝑡) = (

Where µ 3 is the mean of all sensors at time, t.


Afterwards, the deviation from the mean value at times step t for each sensor is determined
and plotted.

TD@E(,E0H % (t) = µ( (𝑡) − TD@E(,% (t)


TD@E(,E0H ' (t) = µ( (𝑡) − TD@E(,' (t)
TD@E(,E0H ( (t) = µ( (𝑡) − TD@E(,( (t)

29
Whenever the deviation increases all of a sudden, the plotted dataset for each sensor as a
function of time is checked to find the cause of the high deviation. High deviations can be
caused by “spikes” that could not be removed with the previous methods. Whenever this type
of faulty data causes high deviation, it is removed manually by setting it to “NaN” for the
specific time the data points are inaccurate. Aside from that, high deviations can be caused by
sensors that have fallen off and are no longer measuring the module temperature. This was
the case for sensor Tmod3,3 of system 3. Figure 20 shows how the deviation of Tmod3,3 increases
at the beginning of August.

Figure 20 A comparison of the three temperature sensors’ deviation from mean for system 3.

By checking the data of sensor 3 as a function of time, it can also be seen that it gives accurate
values during the summer months while it behaves significantly differently after the summer,
as shown in Figure 21. Sensor 3 was later checked on-site, and it was confirmed that this
sensor had fallen off. Thus, the data from this sensor will not be used in the evaluation of
system 3.

30
Figure 21 (a) Measured temperature from three sensors on system 3 (b) Sensor Tmod3,3 for system 3
starts to behave differently compared to the other sensors.

The same concept is used for the irradiance and ambient sensors. However, the method for
the ambient temperature differs slightly. Since the fifth sensor, Tamb, 5 is placed in a loft, it
cannot be compared to the other sensors, therefore, it is not included when calculating the
mean value and deviations. Sensor 4 is placed behind Roof 2 and is less exposed to wind and
to the clear (cold) sky during the night. Therefore, sensor 4 gives higher temperature values
during night times compared to sensor 1-3. Figure 22 gives an illustration of the ambient
temperature sensors. It is obvious from the figure that sensor 5 has higher values and is a few
hours’ time-shifted to the right compared to the other sensors.

Figure 22 Comparison of the ambient temperature sensors, where sensor 5 is behaving differently.

31
4.4.4 Removal of abnormal power generation
The removal of additional faulty data for the power is performed by using another filtering
method. Linear regression is performed for the power and irradiance data by using the tool
basic fitting in MATLAB. Later, a code is developed in MATLAB where the purpose is to
remove faulty data that stand out and are far from the irradiance–power linear relation. This
is done by setting a fixed limit in which the data points can deviate from the linearity, the
limits for data points under and above the line can differ and are adjusted to the number of
invalid data points under and above the line. For example, if the power values above the
linear line in Figure 23 deviates with more than 600 W from the linearization, it is removed
from the dataset and if the data points under the linearization deviate more than 500 W it is
removed from the dataset. Moreover, since the installed capacity of the PV systems differs,
the limits chosen are also adjusted to the individual systems.

Despite the removal of faulty data with this method, faulty data at low irradiance levels
(GPOA<200W/m2) remained. Therefore, one more criterion is added to the code. The criteria
added is a percentage limit which the measured value could deviate from the value in the
linearized line. The percentage limits chosen are adjusted to the individual systems by
visually inspecting the plots, the upper limit is between 30% to 100% higher than the
regression line and the lower limit vary from 20% to 60% lower. It is different upper and
lower limits because the linear regression line is not always in the middle of the datapoints
that seem to be good points. Looking at Figure 23, it can be seen that the regression line is
shifted to the lower fixed limit line instead of being in the middle.

The result of the two filtering methods is shown in Figure 23. The purple data points are the
filtered data points while the yellow data points show measured data where only global
outliers and linear interpolated values have been removed.

Figure 23 Removal of faulty data for system 2 with fixed and percentage limits, where power is
plotted as a function of the plane of array irradiance.

32
Table 7 shows the coefficient of determination before and after filtering with this method. It
is evident from the table that the values of R2 have improved after this filtering method since
the values are closer to 1.

Table 7 The value of R2 before and after removing abnormal power generation.

PV system: 1 2 3 4 5 6 7 8
R2 Before: 0.9805 0.972 0.9826 0.9792 0.9808 0.9566 0.9839 0.9798
R2 After: 0.9883 0.985 0.99 0.9874 0.9887 0.9816 0.9923 0.9906
Improvement: 0.0078 0.013 0.0074 0.0082 0.0079 0.025 0.0084 0.0108

4.5 Replacing missing data


Missing data points for one temperature sensor are replaced by searching for “NaN” and
replacing it with the mean of the other workings sensors on the same system. For instance, if
sensor 1 for system 5 lacks data points for a specific period, it is replaced with the mean of the
two remaining sensors, i.e., sensors 2 and 3. If more than one sensor has missing values, the
replacement is done with the functioning sensor. Moreover, if all of the sensors lack data for a
specific time period it is treated as missing data in this study.

The sensors measuring the ambient temperature on Roof 1 are also replaced with each other
in the same way as the sensors for the module temperatures. Missing data for sensor 5
(Tamb,5), positioned in a loft on Roof 2, is not replaced with any sensor since this sensor is not
comparable with the others. However, sensor 4 which is another sensor on Roof 2 is replaced
with the mean of the 3 other sensors on Roof 1.

For the irradiance, the two sensors on Roof 1 were replaced by each other. The sensor was
positioned on Roof 2 was replaced with the mean of the sensors on Roof 1. Replacement of
missing data for the power was not possible since it was only one measuring sensor for each
system.

4.6 Rate of missing data values


After the filtering process in Sections 4.4.1-4.4.3, missing data rates for each parameter is
determined and presented. Table 8 presents the missing data rates after removing linear
interpolated values and outliers. It is possible to see that the values are quite low for many of
the temperature sensors. The module temperature of system 8 seems to have a much higher
missing data rate compared to the other systems. System 8 had two sensors and during
October both of the sensors were giving linear interpolated values for approximately 7 days.
Thus, it was not possible to fill in the missing data points.

33
Table 8 The rate of missing data for each parameter after removal of linear interpolated values
and outliers (Night values included).

PV system: 1 2 3 4 5 6 7 8
Tmod
0 0.81 0.81 0.81 0.81 0.35 0.81 16
Missing data [%]:
PAC
0.82 0.83 1.3 0.41 0.41 0.83 0.83 0.84
Missing data [%]:
Sensor: GPOA,1 GPOA,2 GPOA,3

GPOA
0.85 0.85 0.18
Missing data [%]:

Sensor: Tamb,1 Tamb,2 Tamb,3


Tamb
0.81 0 6
Missing data [%]:

After the filtering process in Section 4.3.2, missing data rates for the power are determined
again. It can be observed in Table 9 that the fixed limit has not removed many data points.
The percentage limit which was applied to the power-irradiance relationship removes many
invalid data points for lower irradiance levels (<200W/m2) and almost nothing for higher
irradiance levels. Often low irradiance conditions are associated with errors due to PV
components startup and it has to be removed in order to get reasonable values for the
performance evaluation.

Table 9 The rate of missing data after filtering abnormal power generation. This filtering was done
in two steps, removal with a fixed limited and a percentage limit from the regression line.

PV system: 1 2 3 4 5 6 7 8

PAC
Missing data after 2.2 1.08 3.02 1.9 2.5 1.13 1.04 1.9
fixed limit [%]
PAC
Missing data after
53 51 48 58 48 55 45 50
percentage limit
[%]

Since this study will look into the variations between eight PV systems, only complete
datasets are used to enable a fair comparison between the PV systems. It means that data
points that are not available for one of the PV systems at some timepoint or time interval will
not be included for the other systems either. This is done by using a function in MATLAB
called rmmissing. All the rows which include NaN will not be included in the performance
evaluation.

34
5 RESULTS

The filtered data along with the performance parameters presented in Section 2.2 will form
the basis of the performance evaluation of the eight PV systems. The results achieved from
filtering and from computing the performance parameters are presented in the following
section. Section 5.1 presents measured data of module temperature and plane of array
irradiance as a function of time. The noncorrected PR as well as the weather corrected PR
are presented in Section 5.2. Moreover, the specific energy yield and energy generation per
area for the systems is presented in Section 5.3. Section 5.4 shows the relations between
some parameters which were considered to be relevant for this study.

5.1 Measured field data


The monthly average module temperature for each system is shown in Figure 24. System 8 is
installed on a roof of a loft which has the highest average module temperature during the
studied months and the system with lowest temperature changes throughout the period. The
average module temperature of System 8 is approximately 7 ◦C higher than system 3 which
have the lowest temperature in June. July has lower values than June and August, which
follows the trend of the ambient temperature shown in Figure 26.

Figure 24 Average monthly module temperature for the whole period. Nighttime values (plane of
array irradiance values lower than 20 W/m2) are excluded from the module
temperatures. *System 8 is not completely comparable due to its placement.

Figure 25 shows how the module temperature for each system varies during a sunny day in
June. It is evident that system 8, a BIPV system consisting of thin film technology, has the
highest module temperature and reaches 70 °C as maximum. Moreover, systems 3 and 7 have
the lowest temperature and they are both BAPV systems consisting of mono-Si. One can also

35
note that systems 1, 5, and 6 have almost the same temperature during this specific day.
Systems 5 and 6 are both BIPV systems consisting of mono-Si, while system 1 is thin film.

Figure 25 The behavior of the module temperatures for all the systems on a sunny day in June.

Figure 26 shows the weekly average Tamb as a function of time. The highest average Tamb can
also be observed in August while the lowest average temperature is in October. The average
Tamb at the beginning of August increases and reaches values over 20 °C, however, in the
middle of the month it starts to decline slowly until it becomes lower than 20 °C.

36
Figure 26 Average weekly Tamb for June-November. Nighttime values (plane of array irradiance
values lower than 20 W/m2) are excluded.

In Figure 27, the variations of the weekly plane of array irradiance are shown. The total
irradiation levels are the highest for all of the sensors during August, reaching 40 kWh/m2. It
is also noticeable from Figure 27 that the irradiance level is slightly higher for GPOA,1 in
relation to the other sensors during the colder months (September-November).

Figure 27 Average weekly plane of array irradiance for June-November. The three sensors are
located at the systems within a radius of 100 m from each other.

37
5.2 Performance ratio
The average weekly PR’s (eq.4) during June to November are presented in Figure 28.
Systems 3 and 6 have the best average PR, equaling 91% and 89%, respectively as seen in
Table 11. Furthermore, these two systems perform almost equally during the summer
months, with only small variations in the PR. At the end of August, the PR of system 6 starts
to increase while the PR for system 3 are stable. As illustrated in Figure 28, some of the
systems also have higher PR during the summer months. The systems with the lowest PR are
systems 1, 2, 4, and 8, and all of them consist of thin film modules. System 4 that had
operation issues has the lowest PR and varies between 60—70%. System 8, which has the
highest module temperature, have a PR value between 70—80% and are also one of the
poorest performing systems.

Figure 28 Average weekly PR (not weather-corrected) for all the systems for June-November.

In Section 2.2.2, two equations were presented for the correction of the PR. In eq.6, the PR is
corrected with an average site-dependent module temperature which differs between the
systems while eq.5 corrects all the systems to an average temperature equal to 25 °C. The
site-dependent average module temperature for each system is calculated with eq.7 and the
results are presented in Table 10.

Table 10 Site-dependent average module temperature that is irradiance-weighted for each system
during June-November (night values excluded).

System 1 2 3 4 5 6 7 8
Average
26.5 30.2 25.4 29.9 27.4 27.1 25.6 32.1
Tmod [°C]

38
To reduce the seasonal variations of the PR presented in Figure 28 , the weather-corrected
PR is calculated using eq.5. In Figure 29, a comparison of the corrected PR and non-
corrected PR is done for systems 3, 5 and 8.

Figure 29 Comparison of regular PR and weather-corrected PR (corrected t0 25 °C) for systems 3, 5


and 8 as a function of time.

Furthermore, a comparison of the two equations for weather-corrected PR (eq.5 and eq.6) is
presented in Figure 30. As expected, the differences between the two weather-corrected PRs
depends on the module temperature presented in Table 10. For system 3, barely any
difference can be seen between the two corrected PR due to a site dependent Tmod equal to
25 °C.

39
Figure 30 Comparison of a weather-corrected PR with a site-dependent average module
temperature and correction to 25 °C for system 3, 5 and 8.

In Table 11, the average PR over the studied period for the three equations are compared. It
can be seen that the three equations give the same average PR for most of the systems. The
largest difference can be seen for system 8.

Table 11 An overview on the average PR over the studied period, PR corrected to the site
dependent average module temperature and PR corrected to 25 °C.

System 1 2 3 4 5 6 7 8
Average PR 0.74 0.82 0.89 0.65 0.85 0.91 0.82 0.72
Corrected to site
0.75 0.82 0.89 0.65 0.85 0.92 0.82 0.73
average Tmod
Corrected to 25 °C 0.74 0.82 0.89 0.66 0.85 0.92 0.81 0.74

5.3 Monthly specific energy yield


Figure 31 shows the monthly specific energy yield in (kWh/kWp) for each system. It can be
observed in Figure 31 that systems 3 and 6 have the highest specific energy yield throughout
the studied period. These two systems generate over 130 kWh/kWp during June and August.
System 3 performs slightly better during the months with the highest irradiation levels (June
and August), whereas system 6 performs better at periods with lower irradiance levels. The
three systems that have the lowest specific energy yield are systems 1, 4 and 8.

Figure 31 Monthly specific energy yield for June to November for each system normalized to
nameplate DC power.

The energy generation is also normalized to the system area. As Figure 32 shows, systems 3
and 7 have the highest values, 26 kWh/m2 and 24 kWh/m2 respectively in June. The four thin

40
film systems have less than 15 kWh/m2 which is considerably lower than the systems with
mono-Si technology.

Figure 32 Monthly energy generation for June to November for each system normalized to the
system area.

5.4 Relationships between the different parameters

5.4.1 Performance ratio related to module temperature and plane of array irradiance
Figure 33 indicates that system 4 has the lowest PR, around 80% as maximum. Systems 8
and 1 have also lower PRs compared to the other systems. System 6 have the highest, where
the majority of the measured values is near 1 and on some occasional points above 1. Another
system that seems to have high PR is system 3. The most obvious PR differences for the
systems are seen at low irradiance intensities (<400 W/m2). Systems 1, 4, 8 which consist of
CIGS technology have lower PR compared to the other systems at low irradiance intensities.

41
Figure 33 (a) Performance ratio as a function of the plane of array irradiance for
systems 2, 3, 5, 6 and 7. (b) Performance ratio as a function of the plane of
array irradiance for systems 1, 4 and 8. Figures represent data for the
whole period.

By plotting the PR as a function of the module temperature, a similar behavior can be


observed for almost all the systems. The PR increases at low temperatures while it starts to
decrease as the temperature reaches 20-30°C. As Figure 34 shows, the maximum PR for the
various systems is reached at different module temperatures and the maximum value for PR
is different for each system.

Figure 34 Scatter plot of the performance ratio as a function of module temperature for system
6 and 8 that are systems with the highest and the lowest performance ratio. Figures
represent data for the whole period.

42
5.4.2 Power as function of the plane of array irradiance
As Figure 35b indicates, normalized PAC and GPOA has a strong linear relation. However, it is
possible to see that some points start to deviate from this linear trend when the irradiance
increases, due to influence of a higher module temperature. The two systems with the highest
dependence on GPOA are systems 3 and 6. The two systems which are least dependent on the
GPOA are systems 4 and 8.

Figure 35 (a) Normalized power to the installed DC power as a function of the plane of array
irradiance (data points averaged to 15 minutes). (b) Linear regression for power as a
function of the plane of array irradiance (data points averaged to 15 minutes). Figures
represent data for the whole period.

Figure 36 shows the power generation per square meter. Even here, the mono-Si technologies
which includes systems 3, 5, 6 and 7 give better values than the thin film technologies.

43
Figure 36 Normalized power to the system area as a function of the plane of array irradiance (data
points averaged to 15 minutes). The figure represent data for the whole period.

Figure 37 shows the AC output power, PAC, as a function of the irradiance with a temporal
resolution of 15 seconds. It can be observed that PAC has been curtailed for systems 6 and 7,
since the power stops increasing with increasing GPOA.

Figure 37 (a) Power as a function of the plane of array irradiance for system 2 and 3 (original data
with 15 seconds interval). (b) Power as a function of the plane of array irradiance for
system 7 and 6. Figures represents data for the whole period. Note the different y-axis
scales.

44
Figure 38 shows how often the systems work at certain rates of its capacity. A value equal to
one indicates that the inverter is working at its full capacity. From Figure 38, it can be
observed that the systems often work at low capacities. Systems 4 seem to work at lower rates
than the other systems more often.

Figure 38 An illustration of how often the systems work at certain rates of its capacity. A value
equal to one indicates that the inverter is working at its full capacity.

5.4.3 Module temperature as a function of the plane of array irradiance


Figure 39a shows the behavior of the module temperature as a function of the plane of array
irradiance for systems 7 and 8, while Figure 39b shows trend for how the module
temperatures changes with irradiance for all systems. Both figures show that the module
temperature usually increases at higher irradiance levels. However, all systems are not
affected by high irradiance levels to the same extent. System 8 reaches the highest module
temperature, around 70 °C, while system 7 has the lowest module temperature. By looking at
Figure 24 and Figure 39it can also be stated that systems 3 and 7 have a slightly lower
temperature than systems 1, 5, and 6. The module temperature for system 7 never exceeds
60°C, and by looking at Figure 39b, one can state that the module temperature of systems 3
and 7 are, for the most part, below 50°C.

45
Figure 39 (a) Module temperature as a function of plane of array irradiance for system 7 and
system 8. (b) Linear regression for the module temperature as a function of the plane of
array irradiance for all systems. Figures represent data for the whole period.

5.4.4 Pearson correlation


In addition to the scatter plots above, the Pearson correlation was calculated to more easily
see how the different parameters are numerically correlated and the existing variations
among the systems. As Table 12 shows, there are some differences in the correlation between
the parameters for the systems. Compared to other systems, the power for system 7 is least
correlated to the module temperature. This is due to the fact that the module temperature is
the lowest according to Figure 39. From Figure 39, it can also be seen that system 8 has the
highest module temperature and thus also the highest correlation. For systems 1-6, the
correlation coefficient between GPOA and Tamb is approximately the same (between 0.38-0.39),
while it is 0.44 for systems 7.

46
Table 12 Pearson correlation for all the measured parameters for each system.

PAC GPOA Tmod


PV System
GPOA Tmod Tamb Tmod Tamb Tamb

1.00 0.81 0.37 0.82 0.38 0.79


1
1.00 0.83 0.41 0.83 0.39 0.74
2
1.00 0.81 0.38 0.82 0.38 0.80
3
0.99 0.84 0.36 0.86 0.38 0.74
4
1.00 0.83 0.38 0.83 0.38 0.78
5
0.99 0.81 0.37 0.83 0.39 0.78
6
1.00 0.78 0.43 0.79 0.44 0.86
7
1.00 0.86 0.24 0.88 0.27 0.61
8

47
6 DISCUSSION

This section aims to discuss the results presented in the previous section and the data
filtering process. When evaluating the performance ratio and specific energy yield, system
4 will be excluded from the comparisons since this system had operation issues throughout
the studied months.

6.1 Impact of module technology and installation method on module


temperature
The results indicate that the module temperature for the respective system increases
significantly during high irradiance levels because of an increase in ambient temperature and
direct heating of the modules. It is also possible to see a substantial decrease in the module
temperature for each system during the colder months. This has been evoked by lower
irradiance levels and changes in the ambient temperature during the different seasons.
Moreover, external factors such as precipitation and wind could have also affected the
module temperature.

According to Toivola (2010), the heat can be conducted more easily from the PV module for
thin film cells compared to c-Si cells. However, as presented in the previous section, systems
3 and 7, which consist of mono-Si, have the lowest module temperatures among the systems.
One can also note that those systems which have the highest module temperature are systems
2, 4 and 8, and they all consist of thin film technologies. However, the behavior of the module
temperatures for these systems is most likely influenced by how they are installed and not on
the module technology. Furthermore, the placing of the systems also has an impact on the
temperature of the modules, therefore, the installation place is an essential factor to consider
when installing PV systems. System 8, being a BIPV system, is directly installed on the roof of
a loft which is an essential precondition that affects the module temperature of this system.
The module temperature of this system is affected by the temperature inside of the loft.
During the summer months, the temperature increases in the loft which causes a substantial
increase in the module temperature. This relationship is also reflected in the Pearson
correlation coefficient. The Pearson correlation coefficient between the plane of array
irradiance and the ambient temperature for system 8 was 0.27, while it was around 0.38 or
higher for the other systems. Even though system 7 is placed on the same building as
system 8, it is not affected by the high temperature in the building to the same extent. There
is a 5-10 cm gap between the modules of system 7 and the tile, this increases the heat
exchange to the outside. System 7 is also partially installed on the roof overhang which
enables the additional exchange of heat to the outside.

System 2 is also a BIPV system which consist of CdTe, and even in the case of this system, the
installation method seems to contribute to a decrease in air circulation which causes the
module temperature of this system to increase. Nevertheless, systems 5 and 6, which are also
BIPV systems, have significantly lower module temperatures than system 2. Since system 5
and 6 consist of mono-Si technology, the conduction of heat should have been inferior to that
of system 2 according to Toivola (2010). Thus, the large difference in the module temperature
must be a result of various factors affecting the module temperature of system 2. For

48
instance, behind the modules of system 2, there is twinwall plastic sheets which can decrease
the air circulation and contribute to higher module temperatures.

There are very large differences in the module temperature of the various systems. At very
high irradiance levels, there can be more than 20 ◦C difference between the systems with the
highest and lowest temperature. As mentioned, the conditions of the systems were not the
same. Therefore, the large differences are clearly an effect of the various conditions for the
system. Systems 1-6 were placed on a higher altitude compared to systems 7-8 and therefore,
affected by wind to a greater extent. Moreover, it is also worth mentioning that system 8 is
the system which represents realistic conditions best. According to Lindahl et al. (2019),
homeowners do usually install roof mounted PV systems and many of the houses in Sweden
has a roof with a loft.

6.2 Impact of inverter on the performance


The size and efficiency of the inverter does also have an impact on the performance of the PV
systems. The Euro efficiency of the inverters in this study varies as much as 1.3 percentage,
thus, some differences in the energy generation can be due to the inverter efficiency.
However, a higher inverter efficiency does not necessarily mean a higher energy generation.
If an inverter is oversized in relation to the PV array, the inverter will work more often at
lower efficiencies, as shown in Figure 7. This might be the case with System 4, the DC-to-AC
ratio for this system is 0.6 where the size of the inverter is approximately the double size of
the PV array. Due to the small size of these systems, it has been difficult for the suppliers to
dimension properly based on available products. Figure 38a also shows that the inverter for
system 4 operates more often at lower power ranges compared to the other systems which
can normally be a consequence of an largely oversized inverter. However, the poor
performance of system 4 cannot solely be related to the large inverter size since the system
had operating issues.

From Figure 31, it can be seen that system 3 generates more energy compared to system 6
during June and August, which are two months with high solar irradiation. Even though the
inverter of system 6 was undersized in relation to the installed PV array according to the
specifics of the system, the nominal AC power of the inverter was adjusted by the supplier to
3.3 kW when the system was installed. Despite the adjustment, the inverter seems to
constrain the power generation at 3 kW during high irradiance levels, as shown in Figure 37.
This indicates that the adjustment is not giving the intended results and that some energy is
lost. However, choosing an undersized inverter can be a good choice. Despite the energy
losses, system 6 gave one of the highest specific energy yields which indicate that the energy
losses are low.

Looking at Figure 24, one can see that system 2 has the second-highest module temperature
throughout the whole period. However, despite the high module temperature, it performs
quite well in terms of PR and energy yield during the summer months, compared to system 7
which is a BAPV. From Figure 37 it can be observed that the power was curtailed for system 7
during high irradiance levels (summer months). System 7 had an undersized inverter, the
DC-to-AC ratio was 1.2. Thus, much of this indicates that there are losses in system 7 due to a
higher temperature coefficient and an undersized inverter that cuts off energy generation.

49
But since system 7 never reaches high module temperatures, the higher temperature
coefficient should not contribute to very large losses. System 7 did also have a module
optimizer, but since the systems are not shaded, this component can also cause some losses
in the system.

6.3 Variations in performance ratio and energy generation


The PR of the systems seem to differ unexpectedly much. At some points the difference
between the system with the highest PR and the system with the lowest PR is around 20%
(system 4 excluded). The CIGS technologies (systems 1 and 8) had the lowest PR during the
whole period and in addition to low PR values, systems 1, and 8 also had the lowest specific
energy yield. The lower PR and specific energy yield for system 8 can be explained by its high
module temperature. However, thin film technologies have a lower temperature coefficient
than c-Si technologies, so it is usually promoted as a technology that provides an advantage
at high module temperatures. Consequently, the temperature coefficient of the thin film
technologies might not give any advantage over the mono-Si technologies in this case.

The results indicate that PR is lower at low irradiance intensities for some of the systems,
those systems that consist of thin film appear to have lower PR than those that consist of
mono-Si. At low irradiance intensities, the temperature does not differ as much as during
high irradiance intensities, as shown in Figure 39. Thus, the large difference in PR should not
be caused by a pure temperature effect. When the irradiance intensity is about 200 W/m2,
the difference in the module temperature of the systems is approximately 5 ◦C between the
various systems. However, for higher irradiance intensities the temperature difference can be
about 20 ◦C at most. Thus, the difference between mono-Si and the thin film technologies
should have been even greater at high intensities if the temperature effect had been main
cause of the difference in PR at low irradiance intensities.

Among the four systems with thin film technologies, system 2, which is the only CdTe
technology, is the best performing system in terms of PR and specific energy yield. The
monthly average module temperature of this system is even 3-5 ◦C higher than that of system
1, but the PR of system 2 is significantly better. This might be due to the low-temperature
coefficient of system 2 (-0.21%, compared to -0.33% for system 1) which makes it more
efficient at high module temperatures. Moreover, system 1 has a wave-like module surface
which can also contribute to some losses since every part of the solar cell’s surface is not
perfectly oriented due to south.

Systems 3 and 6 are the best-performing system in terms of PR and specific energy yield.
Considering that system 5 has a module optimizer, a slightly lower temperature coefficient
and a somewhat higher inverter euro efficiency compared to system 6, it would be natural to
expect system 5 to perform better than system 6. However, the monthly specific energy yield
and PR throughout the entire period are higher for system 6. The lower PR for system 5 could
indicate that some losses are occurring in this system. The influence of optimizers has not
been evaluated in this study but even though system 5 has a module optimizer, it does not
necessarily mean that it will have a positive impact on the performance. Installing an
optimizer on module level can lead to further losses in the system if it is not compensated by
a corresponding improvement in mismatch losses, which can be a possible cause in this case

50
since these systems are not shaded. The analysis made above demonstrates that the values in
the specifications are not always reflected in the results achieved under real operating
conditions and once again, this indicates how beneficial it is to evaluate the performance
under real operating conditions.

One can also observe a very large difference between the thin film technologies and the
mono-Si technologies. The difference in the efficiency between the thin films and the mono-
Si is around 5-10% and even though the effect of temperature and the efficiency of the
inverter can affect the values in kWh/m2, it would be minimal compared to the efficiency of
the modules. Thus, this large difference can only be explained with the substantial differences
in the module efficiencies. These results do also indicate that the generation of 1 kWh for the
mono-Si requires a smaller system area compared to the thin films. Even though system 6
has a higher PR, systems 3, 5, and 7 might be preferred over system 6 since more energy can
be generated per square meter of roof space. One additional advantage is that the costs of the
rest of the system are also often lower on a per-watt basis when the modules have higher
efficiency.

Both Nordmann et al. (2014) and Marion et al. (2005) clarifies that the PR values are usually
lower in the summer than in the winter due to losses caused by high module temperatures.
However, a conclusion in line with this statement cannot be taken from this study since PR
for several systems seems to either stay on the same level or decrease in colder months. One
possible reason could be that the systems do not reach an extremely high temperature in the
summer months in the Swedish climate.

Compared to the non-corrected PR, the weather-corrected PR is lower during colder periods
while it is higher in the summer months, as Figure 29 indicates. This is the result of removing
the impact of increasing module temperature on the PR. This study used two equations to
determine the weather-corrected PR. The only difference between the two equations is that
eq.5 corrects the PR with a fixed module temperature of 25◦C while a site-dependent average
module temperature for each system is used in eq.6. This means that eq.6 gives a more fitted
correction for the individual systems. However, in Dierauf et al. (2013) eq.6. was introduced,
the calculations were done on cell temperature and simulated data which is not the case for
this study. On the other side, the differences between these two methods were not large since
the obtained site-dependent module temperature was in the range of 23 – 29◦C (system 8
excluded). This means that the differences between the noncorrected PR and weather-
corrected PR are not as large as Dierauf et al. (2013) presented in their report. One reason
could be differences in climatic conditions for Sweden and the climatic conditions for the
data used in their simulation.

51
6.4 Method discussion
Evaluating the performance under real- operating conditions is complicated since many
factors play a role in the monitoring of the PV systems and also the filtering process. The
results presented in this study are largely dependent on the choices made during the
installation of the test systems and also the choices made in the filtering process.

To determine which data points are good and which should be removed is not always easy,
especially at low irradiance levels. Several choices in the filtering process were made based on
assumptions and visual inspection of the plots. For instance, it is known that it is a linear
relationship between the power and irradiance, but not by how much the data points could
deviate from the linear line and still count as good data. The spread of the data points around
the linear regression line did also vary among the systems. The percentage limit used in this
study removed many data points at irradiance levels lower than 200W/m2, which generally
can be associated with errors due to the startup of the components (Dhimish & Alrashidi,
2020). When removing a large amount of data at low irradiance levels, there is a chance that
good data points are also removed. To avoid removal of good data, the plots has been
thoroughly reviewed and this is also the main reason to why the limits described in section
4.4.4 vary for the systems. To further enable comparison of the systems, a complete dataset
was used for the evaluation. This means that if a row contains a missing value the whole row
is removed from the dataset. The drawback of this method is that less data is available for the
evaluation, however, this step is a necessity for a fair comparison. For instance, with this
step, the weekly average PR values will be based on the same amount of data points for all the
systems. Another drawback is that the data for each system is not used to its fully potential
since some good data points are excluded, therefore, the most preferable would be to have a
filling strategy to fill all the missing values for the parameters. Especially for the power
generation since this parameter is causing the missing data rate to increase significantly as
shown in Table 9.

As already mentioned, the choices made at the time of installation can also affect the results.
As presented in Table 3, both System 7 and 8 did not have as many temperatures sensor as
the other systems. System 7 had one sensor, while system 8 had 2 sensors. Measuring the
module temperature with one sensor can give inaccurate values. For example, the speed of
the wind can contribute to a temperature difference higher than 5 °C among various modules
in the same system and the temperature differs depending on where on the module the
sensors are placed (IEC, 2016). In order to determine the average temperature more
accurately, a number of sensors should be placed in different positions on the PV modules. As
previously mentioned, measuring the temperature with the sensors that are placed on the
back of the module does not give the cell temperature which will affect the results. The
location of the sensors is not the same for each system which can also affect the measured
temperature values. Another thing that should be noted is that the surfaces of the back of the
module for each system differed and therefore, it could also have affected the results. System
2 consisted, for example, of glass modules, systems 1 and 8 were plastic modules whereas the
solar cell of system 6 was glued to sheet metal.

System 3 had initially three temperature sensors, but during the filtering process, it was
determined that one of these sensors had fallen off. Thus, it was not used in the performance
evaluation. It is much easier to determine if a sensor has fallen off by comparing it with other
sensors on the same system. Hence, it would have been harder to determine if the

52
temperature sensor on system 7 had fallen off. Having more than two module temperature
sensors could also have reduced the missing data rates for system 8. For system 8, 16% of the
module temperature data was missing. It would have been preferable if all missing data were
filled in. However, sometimes even all 3 sensors would give invalid data.

In this study PR was calculated over 6 months, however, calculating the PR over a longer
time period (i.e., one year) would have been preferable. It would have been easier to conclude
how the systems perform. For example, the PR of system 1 decreases significantly during the
colder months, it could be due to system degradation or something temporary. If the PR had
been calculated over a longer period, it would have easier to draw conclusions.

53
7 CONCLUSIONS
In this study, a technical performance evaluation of eight small-scale PV systems has been
conducted. The systems are evaluated in terms of performance ratio (noncorrected), specific
energy yield (kWh/kWp), and energy per area (kWh/m2). The importance of high-quality
data has also been highlighted. Several filtering methods have been applied to assure the
quality of the data used for this evaluation. Measured parameters are power, irradiance,
ambient temperature, and module temperatures.

From this study, it can be concluded that filtering the measured data is a prerequisite for PV
performance analysis since invalid data has appeared in numerous forms. It can also be
concluded that the monitoring of PV systems is complicated and that several factors should
be taken into consideration before installing PV systems. Due to not entirely comparable
conditions for the systems and several factors affecting the performance, a general conclusion
cannot be taken. Therefore, the conclusions drawn in this study apply to the specific systems
in this study. The module temperature for system 8 is not entirely comparable with other
systems due to its place of installation. Moreover, system 4 has not worked as intended
during the entire measurement period.

From this study, it could be observed that the thin film technologies had higher module
temperatures. However, the results of this study show that the installation method and the
external conditions for the systems had a greater impact on the module temperature than the
module technology itself. The two BAPV systems consisting of mono-Si had the lowest
average module temperature while the BIPV systems had significantly higher module
temperatures. However, a clear impact of the installation method on the performance of the
PV systems cannot be seen, system 3 and 6 which are BAPV respective BIPV are the best-
performing systems in terms of specific energy yield and PR. System 3 and 6 had an average
performance ratio of 89% respective 91% while system 8 that had the poorest performance
had a performance ratio equal to 72%. Systems 3 and 6 generated around 135 kWh/kWp as
the highest monthly specific energy yield in June while system 8 generated around 105
kWh/kWp in the same month.

Among the eight PV systems, systems with mono-Si technology performed better than the
systems with thin film technologies (CIGS) in terms of performance ratio, energy generation
(kWh/kWp), and energy per area (kWh/m2). Among the four studied thin film systems,
system 2 which consisted of CdTe cells were the best-performing systems in terms of a
performance ratio and monthly specific energy yield. System 2 had an average performance
ratio of 82% while other thin film systems had 74% and lower. In comparison to the mono-Si
systems, system 2 had an average PR equal to system 7 which had the lowest average PR
among the mono-Si systems.

Based on the results in this study, a slightly undersized inverter can cause power curtailment
during high irradiance intensities, this was the case for systems 6 and 7. However, power
curtailment due to a slightly undersized inverter did not contribute to significant losses in the
energy for these systems.

54
8 SUGGESTIONS FOR FURTHER WORK
Additional studies need to be done to address the limitations in this study. Some suggestions
are presented below:

• The analysis performed in this study was on system level and did not focus on the
performance of individual components. The measurements on the DC side would be
useful to analyze the impact of each component on the overall performance of the
systems.
• The module temperature can vary depending on where the sensors are placed. In this
study the placement of the sensors varied for the systems, therefore, thermal imaging
cameras could be used to get an overview of the temperature distribution of the solar
panels. This could increase the reliability of the measured data values.
• Since all of the systems were positioned to avoid shading and the number of
parameters was limited, the exact impact of the module optimizers on the
performance of the PV systems could not be determined. In future studies, it could be
interesting to investigate how effective the module optimizers are to minimize losses
due to shading compared to the string inverters with MPPTs.
• The data used for the performance evaluation of the PV systems in this study was
from June to November, but usually PR is calculated annually. Evaluation for a longer
time period would be preferred to further look into the behavior of each system
during various seasons.
• This study did not consider the economic aspects; therefore, it would be valuable to
do an economic evaluation and relate it to the technical performance of each system.

55
REFERENCES

Nordmann, T., Clavadetscher, L., Sark, W. G., & Green, M. (2014). Analysis of Long-Term
Performance of PV Systems. Different Data Resolution for Different Purposes. IEA.

Adeeb, J., Farhan, A., & Salaymeh, A. (2019). Temperature Effect on Performance of
Different Solar Cell Technologies. Journal of Ecological Engineering, vol.20, Issue 5,
pp.249-254. DOI: https://doi.org/10.12911/22998993/105543.

Alight. (2019, April 25). Solceller och PPA:s frammarsch i EU. Retrieved from Alight:
https://www.alight-energy.com/sv/nyheter/solceller-och-ppas-frammarsch-i-eu

Alm, S. E., & Britton , T. (2014). Stokastik. Sannolikhet och statistiskteori med
tillämpningar. Stockholm : Liber AB.

Alternative Energy Tutorials. (2018). Solar cell I-V characteristics. Alternative Energy
Tutorials. Retrieved from https://www.alternative-energy-
tutorials.com/photovoltaics/solar-cell-i-v-characteristic.html

Aurora Solar. (2021, April 14). Choosing the Right Size Solar Inverter. Retrieved from
Aurora Solar: https://www.aurorasolar.com/blog/choosing-the-right-size-inverter-
for-your-solar-design-a-primer-on-inverter-clipping/

Bengts nya villablogg. (2015, February 4). Skillnad mellan global, diffus och direkt
solinstrålning? Retrieved from Bengts nya villablogg:
https://bengtsvillablogg.info/2015/02/04/skillnad-mellan-global-diffus-och-direkt-
solinstralning/

Bengtsson , A., Holm , E., Larsson, D., & Karlsson , B. (2017). Skuggningshandbok. Design
av solcellssystem för minimerad inverkan av skuggning. Energiforsk.

Berger, K., Cueli, A. B., Boddaert, S., Buono, M. D., Delisle, V., Fedorova, A., . . . Chivelet, N.
M. (2018). International definitions of "BIPV". IEA PVPS task 15.

Berman , J. J. (2018). Principles and Practice of Big Data. Preparing, Sharing, and
Analyzing Complex Information. Elsevier. Secong edition. .

Cañete, C., Carretero, J., & Sidrach de Caradona , M. (2014). Energy performance of
different photovoltaic module technologies under outdoor conditions. Energy, Vol 65,
Feb., pp.295–302. DOI: https://doi.org/10.1016/j.energy.2013.12.013.

Chioncel, C. P., Kohake, D., Augustinov, L., Chioncel, P., & Tirian, O. (2010). Yield factors of
a photovolatic plant. Faculty of engineering-Hunedoara, Romania .

Corba , Z., Katic, V., Dumnic , B., & Milicevic, D. (2012). In-grid solar-to-electrical energy
conversion system modeling and testing. Thermal Science, Vol. 16, Suppl. 1, pp.
S159-S171. DOI: https://doi.org/10.2298/TSCI120224069C.

56
Danoune, M. B., Narimane , K., & Djafour, A. (2019). Experimental Evaluation of a
Photovoltaic System to Feed an AC Load. El Oued, Algeria: In Proceedings of the
International Conference on Communications and Electrical Engineering (ICCEE).
DOI: https://doi.rg/10.1109/CCEE.2018.8634493.

Depuru, S. R., & Mahankali, M. (2018). Performance Analysis of a Maximum Power Point
Tracking Technique using Silver Mean Method. Power Engineering and Electrical
Engineering. Vol 16, Issue 1. Department of Electrical and Electronics Engineering,
Institute of Aeronautical Engineering,India.

Dierauf, T., Growitz, A., Kurtz, S., Cruz, J. B., Riley, E., & Hansen , C. (2013). Weather-
Corrected Performance Ratio. (NREL) National Renewable Energy Laboratory.

Dubey, S., Sarvaiya, J. N., & Seshadri, B. (2013). Temperature Dependent Photovoltaic (PV)
Efficiency and Its Effect on PV Production in the World. A Review. Energy Procedia.
Vol 33, pp.311–321. DOI: https://doi.org/10.1016/j.egypro.2013.05.072.

Duffie, J., & Beckman , W. (2006). Solar Engineering of Thermal Porcesses. Third edition.
pp.759. John Wiley & Sons.

Energimyndigheten. (2018, 08 31). Solceller i Sveriges energisystem. Retrieved from


Energimyndigheten: http://www.energimyndigheten.se/fornybart/solelportalen/lar-
dig-mer-om-solceller/systemperspektiv-i-sverige/

Energimyndigheten. (2021, March 31). Nätanslutna solcellsanläggningar. Retrieved from


Energimyndigheten: http://www.energimyndigheten.se/statistik/den-officiella-
statistiken/statistikprodukter/natanslutna-solcellsanlaggningar/

Energy Efficiency and Renewable Energy. (n.d.). Retrieved from Solar Radiation Basics:
https://www.energy.gov/eere/solar/solar-radiation-basics

Enquist, S. (2012). Solel till akademiska sjukhuset. Möjlig genererad effekt och solcellers
estetiska konsekvenser. Gävle: Högskolan i Gävle. Akademin för miljö och teknik.

ETIP PV. (2021). Expert Input Paper – Eco-Design & Energy Labelling for Photovoltaic
Modules, Inverters and Systems in the EU. ETIP PV, SolarPower Europe, PVthin,
European Solar Manufacturing Council, IECRE.

EU. (n.d.). Paris Agreement | Climate Action . Retrieved from European Commission :
https://ec.europa.eu/clima/policies/international/negotiations/paris_en

Fraunhofer ISE. (2020). Photovoltaics Report. Freiburg. Retrieved from Fraunhofer ISE:
https://www.ise.fraunhofer.de/content/dam/ise/de/documents/publications/studies
/Photovoltaics-Report.pdf

Gonçalves, J., Hooff, T. V., & Saelens , D. (2019). Performance of BIPV modules under
different climatic conditions. WEENTECH Proceedings in Energy, Vol 5, pp.107-115.
DOI: https://doi.org/10.32438/WPE.6519.

57
Green, M. A., Emery, K., Hishikawa, Y., Warta, W., & Dunlop, E. D. (2015). Solar cell
efficency tables (Version 45). Progress in Photovoltaics, Vol 23, Issue 1, pp.1-9. DOI:
https://doi.org/10.1002/pip.2573.

IEC. (2016). Photovoltaic system performance - Part 1: Monitoring.

Jankovec, M., & Topic, M. (2012). Intercomparison of Temperature Sensors for Outdoor
Monitoring of Photovoltaic Modules. Journal of Solar Energy Engineering, Vol 135,
Issue 3. DOI: https://doi.org/10.1115/1.4023518.

Jordan , C. D., Marion , B., Deline, C., Barnes, T., & Bolinger , M. (2020). PV Field Reliability
Status–Analysis of 100,000 Solar Systems. Progress in Photovoltaics Research and
Applications. Vol 28, Issue 8, pp.737-859. DOI: https://doi.org/10.1002/pip.3262.

Kalogirou, S. (2014). Solar Energy Engineering : Processes and Systems. Processes and
Systems. Second edition. Elsevier Science & Technology.

Kaltschmitt, M., Streicher, W., & Wiese, A. (2007). Renewable Energy – Technology,
Economics and Environment. . Berlin/Heidelberg. Springer-Verlag Publishing, pp.
268–271.

King , D., Kratochvil, J., & Boyson , W. (1997). Temperature coefficients for PV modules and
arrays: measurement methods, difficulties, and results. Photovoltaic Specialists
Conference, 1997., Conference Record of the Twenty-Sixth IEEE. pp.1183–1186. DOI:
https://doi.org/10.1109/PVSC.1997.654300.

King, D., Boyson , W., & Kratochvil, J. (2004). Photovoltaic Array Perfromance Model.
Albuquerque, New Mexico.: Sandia National Laboratories. DOI:
https://doi.org/10.2172/919131.

Kotu, V., & Deshpande, B. (2019). Data science. Concepts and Practice. Chapter 2 - Data
Science Process. Second edition. Morgan Kaufaman. .

Kovacs, P. (2019). Marknadsöversikt för solcellsmoduler, växelriktare,


infästningsanordningar och kompletta system. Eskilstuna: Energimyndigheten.

Kratzenberg, M., Deschamps, E., Nascimento, L. R., Rüther, R., & Zürn, H. H. (2014).
Optimal Photovoltaic Inverter Sizing Considering Different Climate Conditions and
Energy Prices. Elsevier. Energy Procedia, Vol 57, pp.226-234. DOI:
https://doi.org/10.1016/j.egypro.2014.10.027.

Kumar, N. M., Sudhakar, K., & Samykano, M. (2019). Performance comparison of BAPV and
BIPV systems with c-Si, CIS T and CdTe photovoltaic technologies under tropical
weather conditions. Case Studies in Thermal Engineering. Vol 13. DOI:
https://doi.org/10.1016/j.csite.2018.100374.

Kurpaska , S., Knaga, J., Latała, H., Sikora, J., & Tomczyk, W. (2018). Efficiency of solar
radiation conversion in photovoltaic panels. Bio Web Conf., Vol 10, pp.30-149.

58
Faculty of Production and Power Engineering, University of Agriculture in Krakow,
Krakow, Poland. DOI : https://doi.org/10.1051/bioconf/20181002014.

Li, G., Duan, Z., Liang, L., Zhu, H., Hu, A., Cui, Q., . . . Hu, W. (2020). Outlier data mining
method considering the output distribution characteristics for photovoltaic arrays
and its application. Energy Reports, vol.6, pp.2345-2357. DOI:
https://doi.org/10.1016/j.egyr.2020.08.034.

Lindahl, J., Dahlberg, M. R., & Westerberg, A. O. (2019). National Survey Report of PV
Power Applications in Sweden 2019: Task 1 Strategic PV Analysis and Outreach.
IEA.

Livera, A., Theristis, M., Koumpli, E., Makrides, G., Stein, J. S., & Georghiou, G. E. (2020a).
Guidelines for ensuring data quality for photovoltaic system performance
assessment and monitoring. 37th European Photovoltaic Solar Energy Conference
and Exhibition.p.1352-1356. DOI: 10.4229/EUPVSEC20202020-5DO.2.4.

Livera, A., Theristis, M., Koumpli, E., Spyros, T., Makrides, G., Sutterlueti, J., . . . Georghiou,
G. E. (2020a). Data processing and quality verification for improved photovoltaic
performance and reliability analytics. Wiley. Progress in Photovoltaics. Vol 29, Issue
2. DOI: https://doi.org/10.1002/pip.3349.

Ma, J., Pan, X., Man, K., Li, X., Wen, H., & Ting, T. (2018). Detection and Assessment of
Partial Shading Scenarios on Photovoltaic Strings. Trans. Ind. Appl. Vol 54,
pp.6279-6289. DOI: https://doi.rog/10.1109/TIA.2018.2848643.

Madkor, A. N., Anis , W. R., & Hafez, I. (2015). The Effect Of Numbers Of Inverters In
Photovoltaic Grid Connected System On Efficiency, Reliability And Cost.
International Journal of Scientific & Technology Research. Vol 4, Issue 09.

Marion, B., Adelstein, J., Boyle, K., Hayden, H., Hammond, B., Fletcher, T., . . . Townsend, T.
(2005). Performance Parameters for Grid-Connected PV systems. Lake Buena Vista,
Florida.: Institute of Electrical and Electronics Engineers. DOI:
https://doi.org/10.1109/PVSC.2005.1488451.

MATLAB. (2021). Hampel. Outlier removal using hampel identifier. Retrieved from
MATLAB: https://www.mathworks.com/help/signal/ref/hampel.html

Mertens, K. (2014). Photovoltaics: Fundamentals, Technology and Practice. ISBN


9781118634165. Wiley .

Ministry of the Environment. (2016, 9 29). Summary of the Government’s budget initiatives
in the areas of environment, climate and energy. Retrieved from Government Offices
of Sweden: https://www.government.se/articles/2016/09/summary-of-the-
governments-budget-initiatives-in-the-areas-of-environment-climate-and-energy/

Moharram, K. A., Elhady, M., Kandil, H. A., & Sherif , H. (2013). Enhancing the performance
of photovoltaic panels by water cooling. Ain Shams Engineering Journal, Vol 4, Issue
4, pp. 869-877. DOI: https://doi.org/10.1016/j.asej.2013.03.005.

59
NASA. (n.d.). Causes| Facts-Climate change: Vital signs of the planet . Retrieved from
NASA. Global climate change. : https://climate.nasa.gov/causes/

NREL. (n.d.). Champion Photovoltaic Module Efficiency Chart. Retrieved from NREL:
https://www.nrel.gov/pv/module-efficiency.html

Partlin, S. (2015, December 15). 7 Reasons Why You Should Oversize Your PV Array.
Retrieved from SMA: https://www.sma-sunny.com/en/7-reasons-why-you-should-
oversize-your-pv-array-2/#comments

Pearson, R. K., Neuvo, Y., Astola, J., & Gabbouj, M. (2015). The Class of Generalized Hampel
filters. 23rd European Signal Processing Conference (EUSIPCO). DOI:
https://doi.org/10.1109/EUSIPCO.2015.7362835.

Platon, R., Martel , J., Woodruff, N., & Chau, T. Y. (2015). Online Fault Detection in PV
systems. IEEE Transactions on Sustainable Energy. Vol 6. pp.1200-1207. DOI:
https://doi.org/10.1109/TSTE.2015.2421447.

Prakhya, R. K., & Kotha , R. S. (2018). Temperature Effect on the Performance Metrics of
Gridtied SPV Plant. Electrical and Electronics Engineering Department, CVR College
of Engineering, Hyderabad, Telangana, India. International Journal of Engineering
Research and Technology. Vol 11, Issue 6, pp. 909-923.

PVPMC. (n.d.). Plane of Array (POA) Irradiance . Retrieved from PV Performance Modeling
Collaborative: https://pvpmc.sandia.gov/modeling-steps/1-weather-design-
inputs/plane-of-array-poa-irradiance/

PVsyst. (n.d.). European or CEC Efficiency. Retrieved from PVsyst:


https://www.pvsyst.com/help/inverter_euroeff.htm

Quaschning, V. (2005). Understanding Renewable Energy Systems. Pap/Cdr. Earthscan.


p.137.

Rani, P. S., Giridhar, M. S., & Prasad, S. (2018). Effect of Temperature and Irradiance on
Solar Module Performance. Department Of Eee, L.B.R.College of Engineering, India:
IOSR Journal of Electrical and Electronics Engineering (IOSR-JEEE). Vol 13, Issue 2,
pp 36-40. DOI: https://doi.org/10.21474/IJAR01/3720.

Regeringen. (2020). En samlad politik för klimatet. En sammanfattning av regeringens


klimatpolitiska handlingsplan. Retrieved from Regeringen:
https://www.regeringen.se/49c770/contentassets/b0f74b9a2a024cfcb1ea42966963a
bfb/en-samlad-politik-for-klimatet---en-sammanfattning-av-regeringens-
klimatpolitiska-handlingsplan.pdf

Regeringen. (n.d.). Mål för energipolitiken. Retrieved from Regeringen:


https://www.regeringen.se/regeringens-politik/energi/mal-och-visioner-for-energi/

Reich, N. H., Müller, B., Armbruster, A., Sark, W. v., Kiefer, K., & Reise, C. (2012).
Performance ratio revisited: is PR > 90% realistic? Progress in Photovoltaics

60
Research and Applications. Vol 20, Issue 6. pp.717-726. DOI:
https://doi.org/10.1002/pip.1219.

Saavedra, S. G. (2016). Analysis and simulation of shading effects on photovoltaic cells.


Univeristy of Gävle. Faculty of engineering and sustainable development. Department
og Building,Energy and Environemntal engineering. .

Salam, Z., & Rahman , A. (2014). Efficiency for photovoltaic inverter: A technological
review. 2014 IEEE Conference on Energy Conversion (CENCON). DOI:
https://doi.org/10.1109/CENCON.2014.6967497.

Sangwongwanich, A., Yang, Y., Sera, D., & Blaabjerg, F. (2017). Impacts of PV array sizing
on PV inverter lifetime and reliability. Conference: 2017 IEEE Energy Conversion
Congress and Exposition (ECCE). DOI: https://doi.org/10.1109/ECCE.2017.8096675.

Schweiger, M., Jahn, U., & Herrman, W. (2011). Factors Affecting the Performance of
Different Thin- Film PV Technologies and Their Impact on the Energy Yield.
Hamburg, Germany.: 26th European Photovoltaic Solar Energy Conference and
Exhibition, EU PVSEC 2012. pp.3640-3645. DOI:
https://doi.org/10.4229/26thEUPVSEC2011-4AV.2.56.

Sfirat, A., Gontean, A., & Bularka, S. (2018). A New Method for MPPT Algorithm
Implementation and Testing, Suitable for Photovoltaic Cells. Advances in Electrical
and Computer Engineering Vol. 18, Issue. 3, pp. 53-60. DOI:
https://doi.org/10.4316/AECE.2018.03008.

Shiva Kumar, B., & Sudhakar, K. (2015). Performance evaluation of 10 MW grid connected
solar photovoltaic power plant in India. Elsevier. Energy Reports. Vol 1, pp.184-192.
DOI: https://doi.org/10.1016/j.egyr.2015.10.001.

Siecker , J., Kusakana, K., & Numbi, B. (2017). A review of solar photovoltaic systems
cooling technologies. Renewable and Sustainable Energy Reviews., Vol. 79, pp.192-
203. DOI: https://doi.org/10.1016/j.rser.2017.05.053.

SMA. (n.d.). SUNNY TRIPOWER 8.0 / 10.0 with SMA SMART CONNECTED. Retrieved
from SMA: https://files.sma.de/downloads/STP8-10-3AV-40-DS-en-20.pdf

SMHI. (2017, March 20). Normal globalstrålning under ett år. Retrieved from SMHI:
https://www.smhi.se/data/meteorologi/stralning/normal-globalstralning-under-ett-
ar-1.2927

Solar Reviews. (2019, July 15). Grid-Tied, Off-Grid, and Hybrid Solar Systems. Retrieved
from Solar Reviews: https://www.solarreviews.com/blog/grid-tied-off-grid-and-
hybrid-solar-systems

Solar Reviews. (2021, May 11). Solar panel efficiency: most efficient solar panels in 2021.
Retrieved from Solar Reviews: https://www.solarreviews.com/blog/what-are-the-
most-efficient-solar-panels

61
Stridh, B. (2015, Februari 4). Skillnad mellan global, diffus och direkt solinstrålning?
Retrieved from Bengts nya villablogg:
https://bengtsvillablogg.info/page/32/?ak_action=reject_mobile&commentid=bb99
dbdf14

Stridh, B., Andersson, S., Campana , P., Carlund, T., Landelius, T., Lind, M., . . . Vadiee, A.
(2020). Improved calculation of PV power generation in Sverige. Mälardalens
Högskola, School of Business, Society and Engineering. Västerås:
Energimyndigheten.

Theristis, M., Venizelou, V., Makrides, G., & Georghiou , G. E. (2018). Energy Yield in
Photovoltaic Systems. McEvoy's Handbook of Photovoltaics: Fundamentals and
Applications. Third edition. Chapter: II-1-B. Elsevier. DOI:
https://doi.org/10.1016/B978-0-12-809921-6.00017-3 .

Toivola , M. (2010). Dye-sentized Solar Cells on Alternative Subtrates. Aalto University


School of Science and Technology.

Tossa, A. K., Soro, Y. M., Thiaw, L., Azoumah, Y., Lionel, S., Yamegueu, D., . . . Razongles, G.
(2016). Energy performance of different silicon photovoltaic technologies under hot
and harsh climate. Energy, Elsevier, vol. 103(C), pages 261-270. DOI:
https://doi.org/10.1016/j.energy.2016.02.133.

UN. (n.d.). The Paris Agreement . Retrieved from UNFCCC: https://unfccc.int/process-and-


meetings/the-paris-agreement/the-paris-agreement

Usman, Z., Tah, J., Abanda, H., & Nche, C. (2020). A critical appraisal of PV-systems´
performance. Buildings 2020, Vol 10, Issue 11, pp.192. Multidisciplinary Digital
Publishing Institute. DOI: https://doi.org/10.3390/buildings10110192.

Ustun, S. T., Nakamura, Y., Hashimoto, J., & Otani, K. (2018). Performance analysis of PV
panels based on different technologies after two years of outdoor exposure in
Fukushima, Japan. Renewable Energy, Vol 136, pp.159-178. DOI:
https://doi.org/10.1016/j.renene.2018.12.100.

Vanek , F., Albright , L., & Angenent , L. (2008). Energy Systems Engineering, Evaluation
and Implementation. . USA: Mc Graw Hill.

Wang, X., & Barnett, A. (2019). The Evolving Value of Photovoltaic Module Efficiency.
Applied Sciences. 2019, Vol 9, Issue6,pp. 1227.

Vidyanandan , K. V. (2017). An Overview of Factors Affecting the Performance of Solar PV


Systems. Power Management Institute, NTPC Ltd., NOIDA, India.

Viitanen, J. (2015). Energy efficient lighting systems in buildings with integrated


photovoltaics. ,Aaltouniversity.

62
Woyte , A., Richter, M., Moser, D., Reich, N., Green, M., Mau, S., & Beyer, H. G. (2014).
Analytical Monitoring of Grid-connected Photovoltaic Systems.Good Practices for
Monitoring and Performance Analysis. IEA.

Woyte, A., Richter, M., Moser, D., Reich, N., Green, M., Mau, S., & Beyer, H. G. (2014).
Analytical Monitoring of Grid-connected Photovoltaic Systems.Good Practices for
Monitoring and Performance Analysis. IEA.

Ye, J. Y., Reindl, T., Aberle, A. G., & Walsh, T. M. (2014). Performance Degradation of
Various PV Module Technologies in Tropical Singapore. IEEE Journal of
Photovoltaics, Vol 4, Issue 5. DOI: https://doi.org/10.1109/JPHOTOV.2014.2338051.

Zdyb, A., & Gulkowski, S. (2020). Performance Assessment of Four Different Photovoltaic
Technologies in Poland. Energies Vol 13, Issue 1, pp.196. Department of Renewable
Energy Engineering, Faculty of Environmental Engineering, Lublin University of
Technology. DOI: https://doi.org/10.3390/en13010196.

63
P.O. Box 883, SE-721 23 Västerås, Sweden Phone: +46 21 101 300
P.O. Box 325, SE-631 05 Eskilstuna, Sweden Phone: +46 16 153 600
E-mail: info@mdh.se Webb: www.mdh.se

You might also like