Energy Conversion and Management 100 (2015) 429–439

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

Performance evaluation and validation of 5 MWp grid connected solar

photovoltaic plant in South India
Sivasankari Sundaram 1, Jakka Sarat Chandra Babu ⇑
Department of Chemical Engineering, National Institute of Technology, Tiruchirappalli, NH67, Tanjore High Road, Thuvakudi, Tiruchirappalli, Tamil Nadu, India

a r t i c l e i n f o a b s t r a c t

Article history: The main objective of this paper is to present the validated annual performance analysis with the mon-
Received 8 January 2015 itored results from a 5 MWp grid connected photovoltaic plant located in India at Sivagangai district in
Accepted 25 April 2015 Tamilnadu. The total annual energy generated was 8495296.4 kW h which averages around
Available online 26 May 2015
707941.4 kW h/month. In addition to the above, real time performance of the plant is validated through
system software called RETscreen plus which employs regression analysis for validation. The measured
Keywords: annual average energy generated by the 5 MWp system is 24116.61 kW h/day which is appropriately
Grid connected photovoltaic system
close to the predicted annual average which was found to be 24055.25 kW h/day by RETscreen. The pre-
Performance ratio
System efficiency
dicted responses are further justified by the value of statistical indicators such as mean bias error, root
Dependency plot mean square error and mean percentage error. The annual average daily array yield, corrected reference
RETscreen yield, final yield, module efficiency, inverter efficiency and system efficiency were found to be 5.46 h/day,
Exergy 5.128 h/day 4.810 h/day, 6.08%, 88.20% and 5.08% respectively. The overall absolute average daily capture
loss and system loss of the particular system under study is 0.384 h/day and 0.65 h/day respectively. A
comparison is also made between the performance indices of solar photovoltaic system situated at other
locations from the literature’s published. Furthermore the effect of input factors over the output of the
system is emphasized by regression coefficients obtained through regression analysis. In-depth analysis
dealing with energy and exergy of the system are also included to strengthen the study.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction the increasing demand and scarcity in conventional sources has

The trend for India’s energy consumption out of conventional
sources is found to increase with increased industrialization and
civilization aspects of the society (see Fig. 1).
The total consumption of energy from conventional sources in
India increased from 46,958 petajoules during 2011–2012 to
50,741 petajoules during 2012–2013, showing an increase of
8.06%. The per capita energy consumption increased from
3497.59 kW h in 2005–2006 to 6748.61 kW h in 2012–2013 with
a cumulative annual growth rate of 8.56% and an annual increase
of 8.76%. The estimated electricity consumption for various sectors
such as domestic, commercial, agriculture, industry, traction and
railways increased from 411,887e6 kW h during 2005–2006 to
852,900e6 kW h during 2012–2013, showing a cumulative annual
growth rate of 9.53%. The percentage increase in electricity con-
sumption for an annual period of 2012–2013 is 8.62% [1]. Thus
⇑ Corresponding author.
E-mail addresses: sivasankari66@gmail.com (S. Sundaram), sarat@nitt.edu (J.S.C.
Babu).
1
Tel.: +91 9444970220.
triggered the scientist to pave way for the development of research
in the field of renewable energy sources especially solar energy [2].
India is a tropical country located along the equatorial belt of the
earth with latitude lying between 7° and 37° which makes it to
receive enormous radiant power. There are about 300 clear sunny
days in most parts of the country per year with an average global
insolation of 4–7 kW h/m2/day [3]. Due to the vast solar potential,
the nation had eight initiatives launched under National action
plan on climate change addressing the remedies for balancing
the energy generation and demand. One of such initiative was
Jawaharlal Nehru National Solar mission (JNNSM) set from
January 2010 to deploy 20,000 MW of grid connected Solar power
by 2020 [4]. India has installed solar photovoltaic (PV) projects of
capacity amounting to 2208 MW out of which Tamilnadu con-
tributes 31 MW.
The International Energy Agency (IEA), under photovoltaic
power systems programme (PVPS) have framed a series of 13 tasks
[5] for the outreach of operation, performance and monitoring
solar photovoltaic plants under the platform of research and devel-
opment. As India, not being a member of an International Energy
agency, the studies and discussions on solar photovoltaic power

http://dx.doi.org/10.1016/j.enconman.2015.04.069
0196-8904/Ó 2015 Elsevier Ltd. All rights reserved.
430 S. Sundaram, J.S.C. Babu / Energy Conversion and Management 100 (2015) 429–439
Fig. 1. Energy consumption from conventional sources in India.
plants as per IEC 61724 standard is not available [6] and hence it
becomes imperative to document the performance of PV systems
for knowing its efficient operation. Also successful integration of
PV system includes knowledge on their operational performance
under varying climatic condition [7]. Monitoring input parameters
for performance study yields several advantages such as optimal
sizing and survival of the plant. Monitoring global horizontal irra-
diance (GHI) and ambient temperature helps us to predict the
energy output making the idea of installation over the locality
stronger. Monitoring module temperature helps us to quantify
the thermal losses of the PV array system which thereby affects
the energy and the exergic efficiency deliberately affecting the gen-
erated output power. Moreover the study on performance of the
plant helps in identifying the operational uncertainty (weather risk
involving variation in weather and system risk which involves
aspects of inverter conversion and control) which helps in improv-
ing the yield. Also annual average analysis extends itself in devel-
oping a theoretical model for predicting the daily average power
generated from the system which forms the hot topic of current
research. Additionally the storage of electricity is not required in
a grid tied plat as the electrical energy generated is fed to the grid
with ease of installation, operation and maintenance with less pay-
back period. Hence grid connected photovoltaic (PV) plant is
focussed in this analysis. Long term annual average analysis results
in the incorporation of seasonal variation such as rainfall, frost,
snow and intermittent problems occurring with balance of system
which is absent in short term performance analysis occurring for
less than a year.
The studies regarding the performance analysis of grid interac-
tive photovoltaic plants in literatures deal with the evaluation of
performance indicators. Some concrete literatures involving
annual analysis are as follows. Drif [8] studied the performance
characteristics of a 200 kWp grid connected PV system during
2000–2003. In 2011 Ayompe [9], conducted a performance study
on 1.7 kWp roof top grid connected plant during the year 2008–
2009. He has deduced the performance indicators and compared
the same with the reported results. Furthermore, Padmavathi and
Arul Daniel [6] analyzed the performance of 3 MWp grid connected
PV plant in Karnataka. Vikrant and Chandel [10], carried out a per-
formance analysis of 190 kWp plant and validated its performance
using PVsyst. In 2014 Trillo-Montero [11] developed a software
employing visual basic express for evaluating the system losses
and the performance of two different plants of installed capacity
217.6 kWp and 17.8 kWp.
This study hence aims to analyze the performance of 5 MWp
grid connected photovoltaic system in Sivagangai, India. Besides
performance evaluation it differs from other literatures as cited
above, by employing a simulation software called renewable
energy and energy efficient technologies, RETscreen for perfor-
mance validation. RETscreen developed by the Government of
Canada is used to validate the ongoing energy performance of
energy systems. As the concentration on solar power projects
brings out considerable support by the Government, a 5 MWp solar
plant was set up at Sivagangai by Moserbaer Corporation with the
aid of Government of Tamilnadu in 2010 in order to export the
power generated by the solar PV panels to the southern grid.
2. System, measurements and application
The inputs for the performance analysis is derived from the
operational condition of the 5 MWp solar photovoltaic plant situ-
ated at about 8 km from Sivagangai in Rettaipalyam village with
latitudinal and longitudinal ranges of 9.47°–9.48°N and 78.26°E–
78.27°E with an altitude of 102 m above the sea level. Sivagangai
is located in the southeastern coast and is bounded by Madurai dis-
trict on western side, Pudukkottai district on northern side and
Ramanathapuram district on southern side. The annual average
in-plane solar insolation for the 5 MWp site at Sivagangai is around
5.4149 kW h/m2/day which is measured by employing CMP11
pyranometer which has the sensitivity ranging from 7 to
14 lV/W/m2. The balance of system and its single line schematic
is shown in Fig. 2.
2.1. PV array
The modules installed were thin film modules manufactured by
Moserbaer from an indigenous technology. A total of 61,020 PV
modules of varying peak power capacity ranging from 76 Wp to
86 Wp were arranged amounting to peak power rating of 5 MWp.
The power generation is segmented into 5 sections each contributing
1 MWp. Each MWp was formed from 226 array junction boxes (AJB),
33 sub main junction box (SMJB) and 8 module junction box (MJB).
This arrangement repeats consecutively to form 5 MWp. There were
5 control rooms for monitoring the power delivered by the set of
module junction boxes. The modules are predominantly south ori-
ented, tiltled at an angle of 10°. The ambient and module tempera-
ture are monitored by Vaisala weather transmitter.
2.2. Power conditioning unit
There were 10 numbers of inverter each of 500 kW capacity
ensuring the conversion of 5 MWp DC to AC. Each power condition-
ing unit is supplied by 4 module junction boxes capable of gener-
ating power of 500 kWp. Thus a set of 2 power conditioning units
form a segment of 1 MWp power generation. The energy generated
by the inverter is measured by the energy guard sensor manufac-
tured from Skytron and the data is transmitted through a RS485
data bus.
2.3. Power evacuation
The inverter output is fed to 1250 kVA, 270/11 KV, 50 Hz trans-
former for stepping up the voltage to 11 kV. There were 5 trans-
formers of 1250 kVA which was capable of handling the generated
power. The total power output from the five transformers are fed
to a main transformer of 6.3 MVA, 11 kV/110 kV, 50 Hz which
further steps up of voltage for the Tamilnadu Electricity Board
(TNEB) grid export. The power exported to the grid is monitored
by a metering cubicle supplied by the TNEB.
2.4. Data monitoring system
The solar photovoltaic plant runs for all working days except
for exceptional technical faults interrupted the system. The data
 S. Sundaram, J.S.C. Babu / Energy Conversion and Management 100 (2015) 429–439 431

acquisition system was designed as per IEC61724. The parameters
such as global horizontal irradiance (GHI), Direct normal irradiance
(DNI), ambient and module temperature (Ta, Tm), wind speed (Vs),
DC and AC energy generated (Edc and Eac) are measured instanta-
neously for every 5 min duration in a day. The parameters of wind
speed and relative humidity are measured employing skyCONNI
universal weather sensor manufactured by Skytron and are trans-
mitted through METEON dataloggers through Canbus communica-
tion to the server which updates the information graphically. The
ambient and module temperature are transmitted by RS485 USB
cable which is connected to the server equipped with the Vaisala
configuration software for storing the received data. The average
of the observed instantaneous readings for a day is calculated
and continued for the rest of the working days yielding the
monthly measured parameter values for all working days. The ser-
ver thus works on the principle of supervisory control and data
acquisition (SCADA) for assessment of the monitored data.
3. Monitored input results
The monthly average daily variation of in-plane solar insola-
tion and wind speed are shown in Fig. 3 below. The monthly
average daily solar insolation varies from a minimum of
4.38 kW h/m2/day in December to 5.98 kW h/m2/day in
September and the average wind speed varies from a minimum
of 1.5 m/s to 4.2 m/s.
The monthly average daily measured Ta and Tm for the moni-
tored period vary as shown in Fig. 4. The module temperature is
always found higher than the ambient due to the generation of
thermal losses which occurs evidently on power generation. The
ambient temperature found to vary from 25.7 °C to 35.4 °C per
day. Better system performance is provided with higher wind
speeds and lower ambient temperature. Higher the wind speed,
lower is the module temperature which further reduces the heat
losses providing betterment in the system performance. As the
wind speed is found higher which is 4.1 m/s for the month of
December the corresponding module temperature is lower for
the same measuring to 31 °C.
The ambient temperature and module temperature are seen to
increase as level of radiation increases. The value of ambient and
module temperature for the irradiance of 359.1 W/m2 correspond
to 26.9 °C and 31.4 °C respectively. But the value of the same for
the irradiance of 474 W/m2 correspond to 29.6 °C and 36.9 °C
respectively. The maximum temperature difference between the
PV module and the ambient is 9.3 °C which is less by 17 °C as in
Ayompe, Duffy (2011) [9].

226 AJB 226 AJB

33 SMJB 33 SMJB

4 MJB 4 MJB 4 MJB 4 MJB

DC distribution board DC distribution

board (1000V)
1000V

DC DC
DC DC
AC AC
AC AC

Transformer -1/ 1250KVA Transformer -5 /1250KVA

(270V/11KV) (270V/11KV)

Transformer -6 /6.3MVA ( 11KV/110KV)

TNEB Metering / Substation

TNEB Grid

Fig. 2. Single line diagram for the 5 MWp grid connected system.
432 S. Sundaram, J.S.C. Babu / Energy Conversion and Management 100 (2015) 429–439

t¼T
Xrp X
N
Edc;d ¼ V dc  Idc  T r ; Edc;m ¼ Edc;d ð1Þ
t¼1 d¼1

Tr is the recording time interval and Trp is the reporting period and N
is the number of operating days of plant in a month.

4.3. Energy output or Energy fed to utility grid (Eac)

The energy generated by the PV system is the measure of same

across the inverter output terminals for every 5-min duration by
the data logger. The total daily monitored value of AC power out-
put and the monthly AC energy generated is given by
t¼T
Xrp X
N
Eac;d ¼ V ac  Iac  T r ; Eac;m ¼ Eac;d ð2Þ
t¼1 d¼1

where Pac = Vac * Iac is the AC power recorded as specified above.

4.4. Array yield (Ya)

Fig. 3. Variation of monthly average daily solar insolation and wind speed for the
monitored period. Array yield is time taken by the PV to operate at nominal power
generating Edc. Hence it is the ratio of the daily or monthly average
DC energy generated by the PV system to the rated PV power and is
given by Eq. (3). The daily array yield are calculated by employing
Eq. (3) as stated below

Y aðdailyÞ ¼ Edc;d =PpvðratedÞ ½h=d ð3Þ

The monthly average daily array yield (Ya,m) is given by

  X N
1
Y a;m ¼  Y aðdailyÞ ð4Þ
N d¼1

The monthly average array yield varies from a minimum of

4.194 h/day to 6.5647 h/day. Array yield represents the actual
operation of the photovoltaic generator.

4.5. Final yield (Yf)

The term final yield represents the time taken by the PV to gen-
erate Eac with respect to its nominal power capacity. Hence it
becomes the ratio of final output power generated (Eac) to the rated
PV power as specified by the manufacturer at standard tempera-
ture conditions. As array yield, final yield can also be calculated
daily and as a monthly average. It is dependent on the mounting
Fig. 4. Monthly average variation of module (Tm) and ambient temperature (Ta) for
structure and on location [11]. The daily final yield is given by
the annual period.

Y f ðdailyÞ ¼ Eac;d =PpvðratedÞ ð5Þ

  X N
1
4. Performance analysis of grid connected PV systems
Y a;m ¼  Y f ðdailyÞ ð6Þ
N d¼1

4.1. Performance indicators for grid connected PV system

Performance parameters for grid connected photovoltaic
systems are established by International energy agency which
are described in IEC standard 611724. As described in literatures
[7–14] the most appropriate ones include energy output, array
yield, final yield, reference yield, PV module efficiency, inverter
efficiency, system efficiency, energy loss which comprises of array
capture loss and system loss, performance ratio and capacity
factor. These normalized indicators act as key comparators for com-
paring the performance of the existing grid connected PV systems.
4.2. Energy generated by the PV array system (Edc)
The total daily monitored value of DC power output and the
monthly DC energy generated is given by
4.6. Reference yield (Yr)
Reference yield is the total in-plane solar insolation or global in
plane horizontal insolation divided by the reference irradiance
under standard temperature conditions which is 1 kW/m2.
P
day Gi
Y r;daily ¼ T r  ð7Þ
GSTC
The reference yield depends on the daily in-plane solar radi-
ance. It is a measure of theoretical energy available at a given loca-
tion [6]. Reference yield is termed as corrected reference yield
when it is corrected by the effect of module and the ambient tem-
perature. Corrected reference yield is given by Eq. (8)
Y cr ¼ Y r ð1  C t ðT m  T STC ÞÞ ð8Þ
 S. Sundaram, J.S.C. Babu / Energy Conversion and Management 100 (2015) 429–439
where Yr, Ct, Tm and TSTC represent the reference yield, temperature
coefficient (% °C1), module and ambient temperature respectively.
4.7. PV Module efficiency or Energy efficiency (gpv)
It represents the effective energy generated by the module with
respect to the available radiation. The instantaneous PV array effi-
ciency is given by
gpv ¼ P dc =Gi  Am
where P(dc) is the DC power generated by the PV array system, Gi
represents the global solar irradiation and Am represents the area
of the PV module.
The monthly average PV module efficiency is calculated as
gpv;m ¼ Edc;d =ðGi  Am Þ  100%
where Edc,d represents the total daily DC energy output.
4.8. Inverter efficiency (ginv)
Inverter efficiency presumes to be the highest of module and
system efficiency. The inverter efficiency appropriately called as
conversion efficiency is given by the ratio of AC power generated
by the inverter to the DC power generated by the PV array system.
The instantaneous inverter efficiency is given by
ginv ¼ Pac =Pdc
The monthly inverter efficiency is calculated by equation as
follows
Eac;d
ginv;m ¼  100%
Edc;d
where Eac,d represents the total daily AC energy output.
4.9. System efficiency (gsys)
The photovoltaic system efficiency is associated with the
balance of systems comprising the PV generator and the inverter
module. The instantaneous system efficiency can be calculated
by applying Eq. (13)
gsys ¼ gpv  ginv
4.10. Performance ratio (PR)
Performance ratio represents the effect of losses (which occur
due to the effect of temperature, inverter, wiring loss, mismatch
loss and loss across the bypass diodes) on the performance or
the output delivered by the photovoltaic system and pictures
the incomplete utilization of incoming solar radiation as PR is
a normalization factor with respect to incident solar insolation
[10]. It is the true efficiency of the system and measures the
closeness to the ideal efficiency. It acts as a key comparator
for comparing the grid connected PV system irrespective of the
location, mounting structure and their power capacity [11]. It
also describes the energy transformation in a grid connected
PV system.
It is thus defined as the ratio of final yield to the array yield as
given by
Performance ratio; PR ¼ Y f =Y r :
It can also be expressed as the product of efficiencies repre-
sented below
PR ¼ gdegr  gtemp  gsoil  ginv
433
where gdegr represents degradation efficiency; represents gtemp
temperature efficiency; gsoil represents soiling efficiency and ginv
represents inverter efficiency.
4.11. Capacity factor (CF)
Capacity factor is a methodology for presenting the energy
delivered by an electrical power distribution system. If the system
delivers full rated power continuously its CF will be unity. It is
ð9Þ
defined as the ratio of actual annual energy output to the amount
of energy the PV system can deliver at its rated capacity for 24 h
per day for a year [7–14].
Y FðannualÞ Eac;annual
CF ¼ ¼ ð16Þ
ð24  365Þ P ðpvÞrated8760
ð10Þ
The capacity factor of a system can also be calculated as
h
day
of the peak sun
CF ¼ ð17Þ
24 h=day
The capacity factor varies in proportion to the variation in final
yield.
4.12. Energy loss
The operation of photovoltaic cell to generate power involves
ð11Þ
heat transfer through convection and radiation modes resulting
in losses which further reduces the system performance. The losses
which are more pronounced are the array capture loss, system loss
and cell temperature loss. The cell temperature loss is negligible in
ð12Þ comparison with the other two and hence neglected in this study.
As a thumb rule, the peak power of the PV module decreases by
0.3–0.4% for every 1 °C rise in temperature above standard temper-
ature conditions [9].
Under actual operating condition the following losses add up to
the system performance:
1. Thermal losses due to elevated cell temperatures.
2. Optical reflection loss due to non-perpendicular irradiance and
losses due to low irradiance levels.
3. Effects of shadowing (partial shadow) due to the objects sur-
ð13Þ rounding the mounted PV structure.
4. The effect of conversion efficiency due to the decrease in irradi-
ance and temperature reducing the performance ratio.
5. Non-continuous inverter operation which includes tripping of
inverter and its failure.
4.12.1. Array capture loss (Lc)
Array capture loss is given by the difference between the array
yield and final yield. The loss which occur due to the variation of
actual irradiance from the reference or theoretical irradiance.
Lc ¼ Y r  Y a ð18Þ
where Lc represents the array capture loss in (h/day).
There also occur losses which are constituents of the capture
loss termed as thermal capture loss and miscellaneous capture
loss. Thermal capture losses (Lct) are associated with the thermal
energy loss which are evident due to increase in the module tem-
perature of above 25 °C. It is given by the difference between refer-
ence and the corrected reference yield.
ð14Þ
Lct ¼ Y r  Y cr ð19Þ
Miscellaneous capture losses are embedded with multiple
causes such as wiring, diode loss, shading effects, low irradiance,
ð15Þ dust accumulation over the module, mismatch losses and losses
434 S. Sundaram, J.S.C. Babu / Energy Conversion and Management 100 (2015) 429–439
due to maximum power point tracking [11]. This is given by the
difference between corrected reference yield and array yield.
Lcm ¼ Y cr  Y a
4.12.2. System loss (Ls)
These loss in (h/day) occur due to discontinuous operation of
inverter over the monitored period and is given by [11]
Ls ¼ Y a  Y f
4.13. Energy and exergy analysis of 5 MWp grid connected PV system
Exergy analysis acts as an efficient performance assessment tool
for determining the true performance of the system close to its
ideal working condition. It thus becomes essential to know the
maximum amount of useful work for an operating system by
applying the above rational analysis. Among existing energy con-
servatory techniques exergy analysis terms to be a simper, fruitful
and enlightening technique for performance prediction. Exergy
analysis plays a decisive role in analysis, improvement, design,
assessment and optimization of the energy system [15]. The main
key features are to provide a true measure of actual plant perfor-
mance and to identify the types, causes and location of thermody-
namic losses in the system.
4.13.1. Energy efficiency of solar photovoltaic systems
Energy efficiency of solar photovoltaic systems is defined as the
ratio of DC power output generated by the PV or the electrical
power generated to the energy input [16], which is the product
of the solar array area and the insolation incident on the PV
surface.
P pv
Energy efficiency of a photovoltaic system ¼
EðsolarÞ
Energy of the solar radiation is given by EðsolarÞ ¼ Apv  G
where Apv and G represents the area of PV array and instantaneous
solar radiation.
4.13.2. Exergy efficiency of solar photovoltaic systems
Exergy efficiency depends on the second law of thermodynam-
ics where it is the ratio of output exergy to input system exergy.
Exout
Exergy efficiency of photovoltaic system ¼
Exin
ExðoutÞ ¼ Electrical exergyðExele Þ þ Thermal exergyðExth Þ
þ destruction
The input irradiation received by the solar cell are converted
into electrical energy which constitutes the electrical exergy.
During the process of power generation there occurs heat loss or
thermal loss in the system which corresponds to the thermal
destruction. Thus exergy analysis accounts for the thermal destruc-
tion by the system.
Exergy input or the exergy of the solar radiation is given by
[17,18]
   
4 Ta
ExðinÞ ¼ 1  GA
3 Ts
  
Ta
ExðoutÞ ¼ P pv  1  ½ð5:7 þ 3:8  tÞ  A  ðT m  T a Þ
Tm
where t represents wind speed (m/s).
5. Results and discussion
Monitoring the monthly average daily energy generated is
ð20Þ
responsible for calculating the final yield which largely depends
on the performance ratio of the plant. The monthly average daily
generated energy varies from a minimum of 19413.1 kW h/
kWp/day (December) to 27482.8 kW h/kWp/day (September). The
average energy generated collectively for summer(March–May),
winter(January–February), monsoon(June–September) and post
ð21Þ monsoon months (October–December) is 25483.9 kW h/day,
24927.95 kW h/day, 24765.5 kW h/day and 21097.8 kW h/day
respectively. Thus the energy output vary in accordance to the sea-
sonal weather change which further depends on the bright sun
shine hours.
The monthly average final yield varies from a minimum of
3.882 h/day in December to maximum of 5.496 h/day in
September as seen in Fig. 5. This is due to the fact that the monthly
average in-plane solar insolation is minimum for the month of
December (4.261 kW h/m2/day) to maximum for September
(5.53 kWh/m2/day). The reference yield depends on the daily
in-plane solar radiance. The higher the in-plane solar insolation
the higher is the reference yield. The corrected reference yield is
slightly lower than the reference yield due to the effect of differ-
ence between the module and the ambient temperature with the
difference varying from a minimum of 0.127 h/day to a maximum
of 0.524 h/day.
It can be seen from Fig. 5 that the reference yield and final yield
show a similarity in the nature of variation among the annual mon-
itored period. They are found directly proportional to the in-plane
solar insolation. The monthly average PV module efficiency varies
from a minimum of 5.45% to a maximum of 7.06%.
The monthly average inverter and system efficiency varies from
79.2% in December to 97.8% in September and 4.688% to 5.282%
respectively as described in Fig. 6. The seasonal average inverter
efficiency for the season of summer, winter, monsoon and
ð22Þ
post-monsoon is found to be 92.10%, 90.78%, 86.36% and 84.90%
ð23Þ corresponding to the average in-plane solar insolation which is
5.891 kW h/m2/day for summer, 5.68 kW h/m2/day for winter,
5.465 kW h/m2/day for monsoon and 4.688 kW h/m2/day for
post-monsoon respectively. This is due to the dependency of inver-
ter efficiency over in-plane insolation which is yet proved by
RETscreen through regression analysis in Table 4. The performance
ð24Þ
ð25Þ
ð26Þ
Fig. 5. Monthly average final and reference yield over the monitored period.
 S. Sundaram, J.S.C. Babu / Energy Conversion and Management 100 (2015) 429–439
Fig. 6. Monthly average daily efficiencies for the monitored period.
Fig. 7. Monthly average performance ratio and capacity factor over the monitored
period.
ratio is fairly maintained a constant measuring an annual average
of 89.15% as seen in Fig. 7 varying with minimum of 85.46–92.34%.
The performance ratio for Summer (March–May) and Winter
(January–February) is 86.67% and 87.86% respectively with the
performance of the plant slightly higher in winter than in summer.
This performance variation is decisively due to module tempera-
ture. The capacity factor varies in accordance with the final yield
which ultimately varies with the AC energy generated. The
monthly average capacity factor is high for September amounting
to 22.90% where the final yield is also high as 5.496 (h/day) and
is less for December yielding 16.17% where the final yield is also
consecutively low to 3.882 (h/day).
Fig. 8 shows the distribution of array capture and system losses
during the monitored operational performance of the plant. The
monthly average absolute capture loss varies from a minimum of
0.032 (h/day) to 0.758 (h/day) and system loss varies from 0.123
(h/day) to 1.174 (h/day). The negative the capture loss the less is
the time taken by the photovoltaic system to produce DC energy
at its nominal power capacity. Thus summarizing the above perfor-
mance values the plant generates good amount of energy with
fairly good performance ratio of 89.15% and final yield of
435
Fig. 8. Monthly average daily array loss over the monitored period.
Fig. 9. Monthly average thermal exergy destruction and temperature difference
evaluated for the 5 MWp system.
4.8106 h/day. The annual average of PV module efficiency is
around 6.08% which is low as maximum power point tracking
function which aids in ensuring maximum power extraction is
absent.
The variation of thermal exergy of 5 MWp photovoltaic system
over a monitored period depends on the difference between mod-
ule and ambient temperature. It is represented in Fig. 9 as follows.
As seen from Fig. 9 the thermal exergy loss or destruction is
higher for the month of May 2011 and least for the month of
November 2011 as the difference between Tm and Ta is higher for
May which is 9.3°°C and least for November which is 4.5 °C. The
thermal exergy loss and exergy efficiency are inversely propor-
tional. There occurs higher exergy efficiency with least thermal
loss. This infact is justified in this study as shown in Fig. 10 which
shows the variation of monthly average energy and exergic effi-
ciency over the monitored period. The exergy efficiency is higher
for November amounting to 5.66% as thermal loss is found to be
least to 1.15 MW and least for May amounting to 2.25% as the aver-
age thermal loss for May is found to be higher to 2.78 MW.
The highest energy efficiency in Fig. 10 occurs for the month of
November 2011 evaluating to 7.06% and the least for September
2011 evaluating to 5.45%. This is because the difference between
the energy produced and the maximum available energy is least
436 S. Sundaram, J.S.C. Babu / Energy Conversion and Management 100 (2015) 429–439

The above Tables 1 and 2 shows a comparison of kWp and MWp

plants separately to have a clear picture on the performance of the
present plant with that of others available. The comparative results
show that the actual energy generated, final yield and the perfor-
mance ratio is higher for the 5 MWp Sivagangai plant as
represented.

7. Validation of the system performance employing RETscreen

RETscreen plus is employed for predicting the monthly average

daily DC Energy generated by the PV array system, AC energy gen-
erated, final yield, array yield, PV module efficiency, inverter effi-
ciency and PV module efficiency by regression analysis. It is a
unique decision support tool for monitoring, analyzing and evalu-
ating the operational performance of renewable energy systems
[20].
The percentage difference between the average predicted and
actual values of Edc and Eac are graphically represented in Fig. 11.
The percentage difference between the DC energy values range
from 3.15% to 2.59%. Similarly the lowest and highest percentage
difference for AC energy values occur for the month of November
Fig. 10. Monthly average exergy and energy efficiency for the monitored period.
2011 and April 2012 levelling to 0.04% and 2.74% respectively.
A graphical comparison of the evaluated %difference for the pre-
for November which is 37.22 MW h and higher for September dicted and actual values of PV module efficiency, inverter effi-
which amounts to 49.42 MW h. Also as inferred from the formula ciency and system efficiency are shown in Fig. 12 below. As
(23), the energy efficiency and solar insolation are inversely pro- predicted from the graph below there occur a very less difference
portional. This also holds good for the justification that the solar between the actual and the predicted values of efficiencies justify-
insolation is higher for September and hence least energy ing the actual performance of the 5 MWp solar photovoltaic plant
efficiency. considered.
It is also inferred that the module performance or efficiency is
good for winter season (October, November and December) though 7.1. Statistical comparative analysis for validation
the insolation is less as the module temperature goes low in winter
33.3 °C and higher for Summer which is 43.76 °C. Increase in mod- The accuracy or the closeness in the prediction of performance
ule temperature will further lead to increased thermal exergy indicators with the actual values are evaluated by enumerating the
destruction thereby reducing the performance of the module. The statistical indicators such as mean bias error (MBE), root mean
overall annual average energy and exergy efficiency counts to square error (RMSE), mean percentage error (MPE) and t-static
3.54% and 6.08%. As thermal energy destruction is incorporated value commonly available in literatures [21,22].
exergy efficiency is less than energy efficiency. Further, the energy
efficiency is relatively equal to the PV module efficiency. As seen
the reduced energy efficiency occurs due to the generation of high
thermal loss which is dependent on the fabrication technology
Table 2
adopted for PV production and module temperature. The thermal Performance comparison of MW scale plants in India.
loss can be reduced by reducing the band gap width between the
Reference Capacity Actual PV type Final Monitoring
p-layer and the n-layer during fabrication or by controlling the
(kWp) generation used yield period
module temperature which is done by surface cooling. in MW h (h/day)
[19] 5000 7473 Crystalline-Si, 4.24 352 days
6. Performance comparison thin film and
CPV
[6] 3056 4204 Crystalline-Si 3.77 365 days
A comparison of performance indicators for the present system
[19] 1000 1130 Crystalline-Si 3.09 365 days
with some of the existing grid connected PV plants reported in lit- Present study 5000 8495.29 Thin film 4.81 Annual
eratures are tabulated below.

Table 1
Performance comparison of the present system with certain reported grid connected PV system.

Ref no. Capacity (kWp) Total annual Yf (h) Total annual Yr (h) g(pv) Annual daily (%) ginv (%) PV type PR Monitored period
[8] 67.84 Spain 1000.1 1558.5 9.21 87.82 – 0.65 Three years
[8] 67.84 Spain 846.8 1463.6 7.50 95.88 – 0.58 Annually
[9] 1.72 Ireland 868.7 1029.3 13.52 89.6 MC–A-Si 0.82 Annually
[10] 190 Khatar khatar 905.2 2.99 – – PC 0.73 Annual
[13] 13 Northern Ireland 620.5 – 7.5–10 87 MC-Si 0.62 Three years
[14] Poland 839.5 – 13.7 89.5 A-Si 0.69 Annually
Present study Sivagangai 1752.3 1976.4 6.08 88.2 Thin Film 0.89 Annually
amorphous Si

MC-Si: Monocrystalline silicon.

A-Si: Amorphous silicon.
PC-Si: Polycrystalline.
 S. Sundaram, J.S.C. Babu / Energy Conversion and Management 100 (2015) 429–439 437

  X k
1
MBE ¼  ððpredÞ  ðmeasÞÞ ð27Þ
N i¼1
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  X ffi
1 k ^
RMSE ¼  ððpredÞ  ðmeasÞÞ 2 ð28Þ
N i¼1

 X  
1 k ððpredÞ  ðmeasÞÞ
MPE ¼  100% ð29Þ
N i¼1 ðmeasÞ
!2
ðN  1ÞðMBEÞ2
t-static ¼ ð30Þ
ðRMSEÞ2  ðMBEÞ2

The mean bias error gives accurate information on the long

term performance of the model. A low value of MBE is always
desired for better accuracy of the proposed model. A positive value
of MBE shows an over-estimate while a negative value an
under-estimate by the model. The RMSE test gives the information
on the short-term performance of the proposed model by allowing
a term-by-term comparison of the actual deviation between the
predicted and measured GHI values. Although MBE and RMSE pro-
Fig. 11. Percentage difference between the actual and predicted Edc and Eac over the vide a reasonable methodology to compare models, they do not
monitored period. objectively indicate the model’s statistical significance from the
measured counterpart [21]. Thus this study also includes the eval-
uation of t-static which additionally indicates whether the model
estimates are statistically significant at a particular confidence
value. The smaller the static t-value the better is the performatric
accuracy of the predicted model. In order to determine whether
the model’s estimates are statistically significant, one has to esti-
mate a critical t-value from standard statistical table that is ta/2
at a level of significance and (n  1) degrees of freedom which
should be always greater than the calculated t-value [22].
A comparison of best regression fit values generated by
RETsceen for validation the real time performance of 5 MWp plant
is tabulated below in Table 3. The more the regression coefficient
approaches unity the lesser is the difference between the predicted
and actual estimates. RETscreen encompasses wide range of trend-
lines numbering 200 for choosing the best fit for a particular
parameter.
Furthermore from RETscreen plus, the dependence of an output
performance indicator over the input factors are clearly inferred
from the value of regression coefficient of best fit as seen in
Table 4. For instance, Edc is considered to be dependent on the
Fig. 12. Difference between the actual and predicted gpv, ginv and gsys over the
in-plane solar insolation, ambient temperature and the module
monitored period. temperature. The coefficient of best fit for the variation of DC
energy generated with respect to ambient temperature is 0.9442
as seen in Fig. 13, where as the coefficient of best fit for the varia-
Table 3 tion of the same with the solar irradiance and module temperature
Statistical indicators for the predicted and actual values of performance indicators. bears a coefficient of 0.7082 and 0.4791 which predicts that the
change in Edc is mostly affected by the change in ambient temper-
Estimated MBE RMSE MPE (%) Regression
indicator coefficient (R2) ature for the present system.
Similarly the interactional effect of change in responses or out-
Edc 1.63 4.571 0.0291 0.977
Eac 61.369 3.552 0.2425 0.997
put performance indicators Eac, g(pv), g(inv) and g(sys) with respect to
gpv 8.3e5 0.0487 0.00617 0.994 inputs are performed and the best value of regression coefficient
ginv 0.0175 0.336 0.0747 0.902 generated with the variation of the same reflects high level inter-
gsys 7.4e17 0.0456 0.0013 0.977 action. Thus the effect of change in Eac, g(pv), g(inv) and g(sys) with

Table 4
Dependency plot of input factors over the responses as generated by RETscreen.

Estimated performance Plot of X1 Vs Y Plot of X2 Vs Y Possible plot of X3 Vs Y R2 for X1 Vs Y R2 for X2 Vs Y R2 for X3 Vs Y Influential factor
indicator
Edc Gi Vs Edc Ta Vs Edc Tm Vs Edc 0.7082 0.9442 0.4791 (Edc, Ta)
Eac Gi Vs Eac Ta Vs Eac Tm Vs Eac 0.9827 0.8452 0.7840 (Eac, Gi)
PV module efficiency Gi Vs gpv Ta Vs gpv Tm Vs gpv 0.9901 0.6617 0.7418 (gpv, Gi)
Inverter efficiency Gi Vs ginv Ta Vs ginv Tm Vs ginv 0.7841 0.7521 0.6641 (ginv, Gi)
System efficiency Gi Vs gsys Ta Vs gsys Tm Vs gsys 0.9301 0.8649 0.7697 gsys is more dependent on Gi
438 S. Sundaram, J.S.C. Babu / Energy Conversion and Management 100 (2015) 429–439
Fig. 13. Plot of monthly average daily solar insolation Vs Edc as generated by RETscreen.
Table 5
Summary statistics of the above dependency plot stated in Table 4.
Dependency plot No. of observation No. of iteration Sum of residual
(Edc, Ta) 12 251 0.4137
(Eac, Gi) 12 9 0.6211
(gpv, Gi) 12 17 3.529e6
(ginv, Gi) 12 12 1.894e6
(gsys, Gi) 12 13 2.2181e6
respect to in-plane solar insolation is more significant as the
regression coefficients for the variation of the same with respect
to in-plane solar insolation are high amounting to 0.9827,
0.9901, 0.7841 and 0.9301 respectively. Thus the significant inter-
action between the input factors and responses paves way for the
justification of selected factors resulting in mathematical predic-
tion model for responses.
Table 5 presents the summary of the interaction plot statistics
as computed by RETscreen. Durbin Watson test is used in identify-
ing autocorrelation in regression models [23]. By default the value
of the same varies between 0 and 4. A value 2 of Durbin Watson
constant indicates the absence of autocorrelation in the samples.
Value approaching toward 0 indicate positive autocorrelation and
if the same approaches toward 4 a space for negative autocorrela-
tion is left [19].
8. Conclusion
A 5 MWp grid connected PV system located at Sivagangai was
monitored annually between May 2011 to April 2012 and its per-
formance were evaluated on monthly average daily basis.
Depthful energy and exergy analysis were also carried out for the
same. The significant conclusions of the analysis are as listed
below.
 The annual average of in-plane solar insolation, ambient and
module temperature and wind speed were 5.414 kW h/
m2/day, 30.76 °C, 37.90 °C and 3.191 m/s respectively. The
importance and the effect of solar insolation over energy gener-
ation is emphasized. In addition, the importance of maximum
power point tracking is clearly brought out as the module effi-
ciency is found to be less at annual average of 6.08% though the
performance ratio of the plant is found to be high at 0.8915
indicating its absence. In connection to the above, the inverter
efficiency also reduces due to the fact of PV system’s ineffi-
ciency to track the instantaneous maximum power which aids
maximum DC generation. Inference of highest annual energy
output is made from 5 MWp plant as inferred from Table 2 as
Best eight order
polynomial fit
Average of residual Standard error of estimate R2 Durbin Watson static
0.0345 712.46 0.9442 2.208
0.0518 577.57 0.9827 1.941
2.941e7 0.0011 0.9901 1.926
1.578e7 0.0531 0.7841 1.918
1.848e7 0.0007 0.9301 1.970
the module titlt angle is approximately equal to the latitude
of the site which will typically result in maximum energy
extraction for grid tied installations which can be adopted [24].
 Thus an improved performance than reported can be realized if
the inverter employed is used for multi-MPPT function rather
than master–slave operation as the occurrence of inverter fail-
ure is less than MPPT which is most essential for intermittent
renewable energy sources such as solar. Improved inverter
efficiency also predicts an increased power delivery. Also, the
module efficiency is low than other evaluated system as seen
in Table 1 because of the PV technology adopted. The present
system employs thin film amorphous silicon PV cells for har-
nessing PV power which results in less PV module efficiency
typically in the range of 4–9% only. Thin film CiGs and thin film
CdTe posses typical module efficiencies in the range of 8–13.5%
and 9–11% which indicated the usage of them. But the main dis-
advantage lying with the thin film CiGs and CdTe are fabrication
complexity and raw material availability for production. Thus
thin film silicon cells and monocrystalline cells (typical effi-
ciency of 13–17%) together as a whole can be combined for
power generation rather adopting same type of cells. The
advantage behind thin film lies in its production or fabrication
cost, weight and occupancy area. Furthermore, hybrid cell
topology will also yield better system efficiency than the pre-
sent, paving way for betterment. Thus to resolve the cited prob-
lem a combination of thin film with mono or polycrystalline
would improve the PV harnessing capacity indigenously with
a trade off met with the cost and efficiency. The energy analysis
also suggest the technology improvement or module tempera-
ture control for improved system performance.
 The significance of the real time PV system is yet validated by
RETscreen plus where a overall coefficient of best fit for the
selected key performance indicators varied from a minimum
of 0.7841 to a maximum of 0.9901 which proves remarkably
good agreement between the actual and the predicted values
of the same. In addition, statistical validators such as MBE,
RMSE, MPE were evaluated for the same and the overall annual
average value of RMSE and MPE between 0.04 to 4.57 and
 S. Sundaram, J.S.C. Babu / Energy Conversion and Management 100 (2015) 429–439
0.0013 to 0.2425 respectively. The lesser the value of RMSE
and MPE, the better is the operational accuracy of the plant.
 The dependency of input factors over the output is essential for
formulating theoretical models such as regression, time-series
and neural network for the prediction of daily average of
monthly based Eac where selection of input factors for output
prediction forms the most important part. This would be the
future scope of the present study.
 Finally a comparison of the performance results obtained from
this study to the other reported results as in Tables 1 and 2
reveal that the PV system’s annual average of the final yield
and reference yield is higher than plants at Spain, Poland,
Northern Ireland, Dublin and Khatar-khatar marking to 4.801
(h/day) and 5.4149 (h/day) thus indicating the higher annual
average performance ratio of 89.15%. The PV module efficiency
of 6.08% is least of all cited literature in Table 1 and inverter effi-
ciency of the system tend to be the lowest of 88.2% compared to
the Spain, Poland and Ireland sites. Collectively, Sivagangai
shows vast solar energy potential which are utilized effectively
and can still be made more efficient by inclusion of MPPT
equipped inverter and employing a combination of thin film
with mono or polycrystalline cells(hybrid technology) for PV
power extraction.
Acknowledgement
The authors acknowledge the assistance given by M/s Sapphire
private limited for accessing the input data of the plant which was
required for the above analysis.
References
[1] Energy Statistics, Central statistics office, 21st Issue; 2014. <www.mospi.gov.
in>.
[2] Goura R. Analyzing the on-field performance of a 1-megawatt-grid -tied PV
system in South India. Int J Sustain Energ 2015;34:1–9.
[3] Government of India. New and renewable energy policy statement 2005.
Ministry of non-conventional energy sources, Government of India; 2005.
[4] SERC (Solar energy corporation of India). <www.seci.gov.in/content/
innerinitiative/jnnsm.php>.
439
[5] IEA (International Energy agency) Country report on analytical monitoring of
photovoltaic systems IEA PVPS Task13; 2014. <www.iea-pvps.org>.
[6] Padmavathi K, Arul Daniel S. Performance analysis of a 3 MWp grid connected
solar photovoltaic power plant in India. Energy Sustain Dev 2013;17:615–25.
[7] Makrides G, Zinsser B, Norton M, Georghiou GE, Schubert M, Werner JH.
Potential of photovoltaic systems in Countries with high solar irradiation.
Renew Sustain Energy Rev 2010;14:754–62.
[8] Drif M, Perez PJ, Aguileria J, Almonacid G, Gomez P, De la Casa J, et al. A grid
connected photovoltaic system of 200 kWp at Jean University. In: Overview
and performance analysis, vol. 91; 2007. p. 670–83.
[9] Ayompe LM, Duffy A, McCormack SJ, Conlon M. Measured performance of a
1.72 kW roof top grid connected photovoltaic system in Ireland. Energy
Convers Manage 2011;52:816–25.
[10] Vikrant S, Chandel SS. Performance analysis of a 190 kWp grid interactive solar
photovoltaic power plant in India. Energy 2013;55:476–85.
[11] Trilo-Montero D, Santiago I, Luna-Rodriguez JJ, Real-Calvo R. Development of a
software application to evaluate the performance and energy losses of grid
connected photovoltaic systems. Energy Convers Manage 2014;81:144–59.
[12] Micheli D, Alessandrini S, Radu R, Casula L. Analysis of the outdoor
performance and efficiency of two grid connected photovoltaic systems in
northern Italy. Energy Convers Manage 2014;80:436–45.
[13] Mondol JD, Yohanis Y, Smith M, Norton B. Large Long term performance
analysis of a grid connected photovoltaic system in Northern Ireland. Energy
Convers Manage 2006;47:2925–74.
[14] Pietruszko SM, Gradzki M. Performance of grid connected small PV system in
Poland. Appl Energy 2003;74:177–84.
[15] Dincer I, Ratlamwala TAH. Importance of exergy for analysis, improvement,
design, and assessment. Wiley Interdiscipl Rev – Energy Environ
2013;2:335–49.
[16] Saber S, Mehran A. Exergy analysis of photovoltaic panels-coupled solid oxide
fuel cell and gas turbine-electrolyzer hybrid system. J Energy Res Technol
2014;136:1201–9.
[17] Spanner DC. Introduction to thermodynamics. London: Academic Press; 1964.
[18] Saitoh H, Hamada Y, Kubota H, Nakamura M, Ochifuji K, Yokoyama S, Nagano
K. Field experiments and analysis on a hybrid solar collector. Appl Therm Eng
2003;23:2089–105.
[19] MNRE (Ministry of New and Renewable Energy. Performance of grid solar PV
power plants under demonstration programme, Report: 32/54/2011-12/PVSE;
2011. <www.mnre.gov.in/file manager/user files/Grid_Solar_Demo_
Performance.pdf>.
[20] RETscreen International Empowering cleaner energy decisions <http://www.
retscreen.net/ang/performance_analysis.php>.
[21] Tarhan S, Ahmed S. Model selection for global and diffuse radiation over the
Central Black Sea (CBS) region of Turkey. Energy Convers Manage
2005;46:605–13.
[22] Wanxiang Y, Zhengrong L, Yuyan W, Fujian J, Lingzhou H. Evaluation of global
solar radiation models for Shanghai, China. Energy Convers Manage
2014;84:597–612.
[23] Durbin Watson static <http://www.investopedia.com/terms/d/durbin-
watson-statistic.asp>.
[24] Amorphous silicon thin film Solar PV module installation manual. 2011.
Moserbaer solar limited. Document No. PV/TF1/ENG/WI/036.