Energy xxx (2014) 1e7

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Fuzzy logic based modeling and estimation of global solar energy

using meteorological parameters
M. Rizwan a, *, Majid Jamil b, Sheeraz Kirmani b, D.P. Kothari c
a
Department of Electrical Engineering, Delhi Technological University, Bawana Road, Delhi 110042, India
b
Department of Electrical Engineering, Jamia Millia Islamia, Delhi 110025, India
c
J B Group of Institutions, Hyderabad 500075, India

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

Article history: Global solar energy data is considered as the most important parameter in smart grid applications,
Received 19 November 2013 particularly for sizing the photovoltaic system and demand driven supply. However the data of global
Received in revised form solar energy is rarely available on hourly basis, even for those stations where measurement has already
3 March 2014
been done. Due to lack of such measured data, the estimation of global solar energy at the earth’s surface
Accepted 19 April 2014
Available online xxx
is an important study in the present scenario to meet the energy requirement from green energy sources.
This paper is based on fuzzy logic approach for modeling and estimating the global solar energy using
mean duration sunshine per hour, temperature, latitude, longitude, altitude and months of the year as
Keywords:
Fuzzy logic
input parameters. Fortunately, these important for accurate parameters estimation of solar energy are
Meteorological parameter commonly available. Results obtained from fuzzy logic approach are used for the prediction of SPV
Smart grid system output for smart grid application.
Solar energy estimation Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction aforesaid, a thorough literature review is carried out and it is found

that number of mathematical models for the estimation of global
The rapid depletion of fossil fuel reserves and large greenhouse solar energy under cloudless skies is available in the literature [4e
emissions are becoming the major challenge for the today’s power 5]. Regression models and stochastic models for estimating the
sector. Due to increasing population and urbanization, the demand solar energy are also presented by researchers [6e11]. The results
of electricity is increasing exponentially. Under such circumstances obtained using these models were satisfactory, but applicable only
it is highly desirable to adopt the environment friendly energy for clear sky weather condition. In India around 300 days in a year
sources for power generation. In order to achieve the target, are clear sky and remaining days are cloudy, so it is very difficult to
renewable energy sources particularly solar and wind sources can estimate the accurately using mathematical/regression/stochastic
play the important role. In addition, solar energy is one of the most models. Therefore, due to uncertainty in weather conditions, fuzzy
promising and more predictable than other renewable energy logic based models are proposed by researchers to estimate the
sources. Therefore, high penetration of solar technologies into solar energy at a given location using different meteorological pa-
electric power system can meet the desired goals. It is well known rameters [12e18]. The uncertainty in atmosphere may occur due to
fact that, India is located on the equatorial sun belt of the earth has the existence of the following: dust, moisture, aerosols, clouds, or
great potential of solar energy with an average of about 300 solar temperature differences in the lower atmosphere. Among these
days per year. Further, the government is also providing the in- factors, clouds can cause the maximum losses in the extraterrestrial
centives and other benefits for installing the power plants based on solar energy reaching at earth’s surface. The atmosphere causes a
solar energy. Hence, to exploit the environment friendly source of reduction of the extraterrestrial solar input by about 30% on a very
energy, analysis and estimation of solar energy at different stations clear day to nearly 100% on a very cloudy day [13]. Further, the use
of the country is utmost important [1e3]. Keeping in view of the of solar energy estimation in smart grid application is rarely
available in the literature. Keeping in view of aforesaid fuzzy logic
based model for the estimation of global solar energy has been
* Corresponding author. Tel.: þ91 9891558821 (mobile). developed considering latitude, longitude, altitude of the location,
E-mail addresses: rizwaniit@yahoo.co.in, rizwaniitd@gmail.com (M. Rizwan). months of the year, mean duration sunshine per hour (it is the ratio

http://dx.doi.org/10.1016/j.energy.2014.04.057
0360-5442/Ó 2014 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Rizwan M, et al., Fuzzy logic based modeling and estimation of global solar energy using meteorological
parameters, Energy (2014), http://dx.doi.org/10.1016/j.energy.2014.04.057
2 M. Rizwan et al. / Energy xxx (2014) 1e7

Nomenclature T monthly mean hourly temperature ( C)

To monthly mean hourly maximum possible temperature
D day of the year ( C)
H monthly mean daily irradiance on horizontal surface T/To ratio of monthly mean hourly temperature to monthly
H0 mean clear sky daily irradiance mean hourly maximum possible temperature
H/Ho clearness index Ta ambient temperature ( C)
Hmeasured measured value of monthly mean daily irradiance on Tcell cell temperature ( C)
horizontal surface Tstc temperature of PV module at STC
Hestimated predicted value of monthly mean daily irradiance on Voc open circuit voltage of PV module at STC
horizontal surface
Is short circuit current of PV module at STC (A) Greek symbols
Isc solar constant ¼ 1367 W/m2 u hour angle
Pmax maximum power of module at MPP us hour angle at sunset
S monthly mean daily hours of bright sunshine f latitude of the location
So monthly mean of maximum possible daily hours of d declination angle
bright sunshine hc efficiency of the PV module
S/So mean fraction possible sunshine hours

of average daily actual sunshine duration at the location to the final customer. The main advantage of smart grid implementation
theoretical sunshine duration), and temperature as input parame- in terms of utility benefits include reduced perturbations and
ters. Obtained results are further simulated for smart grid appli- outages; minimal power losses and blackout prospects; lower
cations using fuzzy logic approach/toolbox. maintenance and operational cost; lower greenhouse gas emis-
This paper is organized as follows: Section 2 presents the basics sions; increased energy efficiency; increased large scale renew-
of smart grid operation. Section 3 describes the fuzzy logic based able energy and distributed generation integration; enabled
model for solar energy estimation. Fuzzy model for the prediction micro-grid applications and energy management systems envi-
of SPV system output for smart grid application is presented in ronmental benefits and economic growth through clean power
Section 4. Results and discussions are presented in Section 5. A markets [19].
Conclusion followed by the references is discussed in Section 6. In the present grid system generation is following the load, but
in the smart grid system the load will follow the generation. For
2. Smart grid such a system, where load follows the generation the solar energy
estimation at the site is the driving contributor to the power output
The global electricity sector and its customers are faced with a calculations for these systems. If a system is able to predict solar
number of challenges that are unparalleled since the advent of energy with a good accuracy at a particular location then load
widespread electrification. Challenges including climate change, scheduling, economic load dispatch, battery sizing, time of use and
escalating energy prices, energy security and energy efficiency are pricing can be done intelligently and optimally including the supply
converging to drive fundamental change in the way of energy of critical loads for those times when the sufficient amount of po-
produced, delivered and utilized. Keeping in view of aforesaid the wer is available. In addition to this when the power available from
future electricity system must produce and distribute electricity the solar energy based system is less then only critical loads can be
which should be reliable, affordable and clean. To achieve the same, supplied. With the correct data, solar collectors are fairly easy
both the electricity grid and the existing regulatory system must be components to size, install, and begin generating power.
smarter. Hence there arises a need of grid that should be smart i.e.
smart grid. Further, world is venturing into renewable energy re- 3. Fuzzy logic approach for global solar energy estimation
sources like wind and solar. With such unpredictable energy
sources, feeding the grid must be highly adaptive in terms of supply In the global solar energy estimation, geographical features like
and demand. A good electric supply is one of the key infrastructure latitude, longitude, altitude and meteorological parameters like
requirements to support overall development, hence the opportu- mean duration sunshine per hour and temperature play an
nities for building smart grid is immense [3,19e20].
Smart grid is used to predict and intelligently respond to the Table 2
behavior and actions of all users connected to it. Further, it is used Monthly average data of H/H0, S/S0 and T/T0.
to efficiently deliver the reliable, economic, and sustainable
Month New Delhi Jodhpur
electricity services to consumers. In addition, smart grid tech-
nologies are advanced electrical networks that support new H/H0 S/S0 T/T0 H/H0 S/S0 T/T0
generation interactive energy and communication services for the January 0.189 0.836 0.334 0.180 0.890 0.554
February 0.387 0.857 0.509 0.386 0.890 0.612
March 0.657 0.857 0.639 0.675 0.800 0.734
Table 1 April 0.877 0.868 0.809 0.807 0.890 0.858
Geographical features of Indian stations considered. May 0.900 0.879 0.896 0.900 0.890 0.900
June 0.779 0.847 0.900 0.822 0.830 0.883
Station Latitude Longitude Height above the Climate zone
July 0.557 0.697 0.841 0.492 0.720 0.805
( N) ( E) mean sea level (m)
August 0.475 0.697 0.813 0.370 0.726 0.778
New Delhi 28.58 77.20 216 Composite September 0.546 0.847 0.673 0.551 0.810 0.802
Jodhpur 26.30 73.02 224 Hot & dry October 0.412 0.900 0.727 0.505 0.870 0.781
Kolkata 22.65 88.45 6 Warm & humid November 0.241 0.857 0.596 0.254 0.900 0.709
Shillong 25.57 91.88 1600 Cold & cloudy December 0.100 0.804 0.464 0.100 0.900 0.632

Please cite this article in press as: Rizwan M, et al., Fuzzy logic based modeling and estimation of global solar energy using meteorological
parameters, Energy (2014), http://dx.doi.org/10.1016/j.energy.2014.04.057
 M. Rizwan et al. / Energy xxx (2014) 1e7 3

Table 3
Monthly average data of H/H0, S/S0 and T/T0. Low Low Normal Normal High Normal High

Degree of Membership
1
Month Kolkata Shillong
0.8
H/H0 S/S0 T/T0 H/H0 S/S0 T/T0
0.6
January 0.181 0.900 0.645 0.109 0.609 0.100
February 0.386 0.900 0.734 0.398 0.827 0.368 0.4
March 0.702 0.900 0.841 0.690 0.900 0.655
April 0.900 0.889 0.900 0.900 0.706 0.854 0.2
May 0.860 0.877 0.899 0.594 0.585 0.858
June 0.529 0.776 0.867 0.369 0.124 0.845 0
July 0.330 0.641 0.835 0.329 0.100 0.900 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
August 0.376 0.652 0.832 0.201 0.124 0.892 Sunshine
September 0.312 0.754 0.830 0.100 0.270 0.840
October 0.347 0.878 0.821 0.230 0.488 0.700 Fig. 1. Fuzzy subsets membership functions for sunshine duration.
November 0.212 0.900 0.764 0.280 0.512 0.481
December 0.100 0.889 0.680 0.145 0.609 0.280
where

!
important role. There are other meteorological parameters also 248 þ D
which may affects the global solar energy such as relative humidity d ¼ 23:45 sin 360 (3)
365
and wind speed. These parameters do not affect the global solar
energy significantly. Therefore these parameters are not considered
as input parameters in the proposed model. u ¼ cos1 ðtan f tan dÞ (4)
Keeping in mind different climatic conditions, four stations such
as New Delhi, Jodhpur, Kolkata and Shillong are chosen for the
where D is the day of the year, Isc ¼ 1367 Wm2 is the solar con-
present study that cover the different climatic zones and their
stant, f is the latitude of location, d is the declination angle and u is
features are presented in Table 1.
the sunset hour angle [23].
The monthly averaged (1986e2000) hourly meteorological data
To calculate extraterrestrial solar irradiance H0 and relative
such as mean duration sunshine per hour (S/S0), temperature ratio
sunshine duration (S0) at various Indian stations, MATLAB program
(T/T0) and clearness index (H/H0) for these stations is obtained from
has been written.
IMD (Indian Meteorological Department) and MNRE (Ministry of
Data of meteorological parameters like clearness index (H/H0),
New and Renewable Energy) [21].
mean duration sunshine per hour (S/S0) and temperature ratio (T/
The input parameters are latitude, altitude, longitude, months of
T0) for New Delhi and Jodhpur stations are presented in Table 2.
the year, temperature ratio (T/T0) and mean duration sunshine per
Similarly, the data of meteorological parameters mentioned
hour (S/S0). The output parameter is the clearness index (H/H0). The
above for Kolkata and Shillong stations are presented in Table 3.
estimated global solar energy is obtained by multiplying the esti-
The impact of mean duration sunshine per hour and tempera-
mated clearness index by H0. Where H is the monthly mean global
ture is more significant as reported in the literature [11e12,15].
irradiance on horizontal surface, and H0 is the extraterrestrial solar
Therefore, it is necessary to select them as input parameters to get
irradiance on the 15th day of month, S is the monthly mean of the
more accurate output. Hence, the sunshine duration and temper-
daily hours of bright sunshine; S0 is the maximum daily hours of
ature have been selected as input meteorological parameters for
sunshine or day length. The ratio of S/S0 is the fraction of maximum
the estimation of global solar energy. Further, we have also taken
possible numbers of bright sunshine hours and H/H0 is the atmo-
into consideration the effect of geographical parameters like lati-
spheric transmission coefficient, commonly known as clearness
tude, longitude and altitude as inputs, though these parameters
index. The value of S0 can be computed from Cooper’s formula
remain constant for a particular location. Hence these parameters


2 have not been included in defining the rules.
S0 ¼ us (1)
15 The fuzzy system contains a set of rules which are developed
from qualitative descriptions. In fuzzy systems, rules may be fired
The extraterrestrial solar irradiance on a horizontal surface H0 is with some degree using fuzzy inferencing; whereas, in conven-
obtained from the following expression tional expert systems, a rule is either fired or not fired. For the solar
!! energy estimation problem, rules are defined to determine the
24  3:6  103  ISC D accuracy in terms of absolute relative error. For the estimation of
H0 ¼ 1 þ 0:033 Cos 360
p 365 global solar energy at a particular location keeping geographical
 Cos f Cos d Sin u þ u Sin f Sin d parameters as constant, a set of multiple-antecedent fuzzy rules
have been established. The input to the rules is mean sunshine
(2) duration per hour (S/S0) and temperature ratio (T/T0), and the

Table 4
Decision matrix for the estimation of global solar energy.

AND Sunshine duration

Low LoweNor Normal HigheNor High
Temperature Low Low LoweMed Med MedeHigh MedeHigh
LoweMed Low LoweMed Med MedeHigh High
Medium Low LoweMed MedeHigh MedeHigh High
MedeHigh Low LoweMed MedeHigh MedeHigh High
High LoweMed Med MedeHigh High High

Please cite this article in press as: Rizwan M, et al., Fuzzy logic based modeling and estimation of global solar energy using meteorological
parameters, Energy (2014), http://dx.doi.org/10.1016/j.energy.2014.04.057
4 M. Rizwan et al. / Energy xxx (2014) 1e7

Low Low Medium Medium Medium High High Solar Insolation

Degree of Membership

1
PV Power
0.8 Mamdani
0.6
Cell Temperature
0.4

0.2 Fig. 5. Fuzzy logic approach for the prediction of SPV output.

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Temperature Table 5
Decision matrix for the prediction of SPV output.
Fig. 2. Fuzzy subsets membership functions for temperature ratio.
And Solar insolation
Low Med High Very high
Cell temperature Low Low Med Very high Very high
Med Low Med High Very high
Degree of Membership

Low Low Medium Medium Medium High High

1 High Low Med High Very high
Very high Low Low Med High
0.8
0.6
0.4 of the sources particularly wind and solar. This problem can be
0.2 overcome by predicting the renewable energy sources intelligently.
In addition, the generation and load forecasting can provide
0 adequate solutions to face these problems. Depending on the
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Global Solar Energy
availability of power the customers are allowing to decrease their
demand. Keeping in view of aforesaid, an attempt has been made to
Fig. 3. Fuzzy subsets membership functions for global solar energy. predict the solar energy using fuzzy logic approach for smart grid
application.
In this paper, prediction of SPV system output is done on the
output consequent is the clearness index (H/H0). The consequents
basis of data acquired from fuzzy model and cell temperature. This
of the rules are shown in the shaded part of the matrix [22]. Ob-
is done by predicting the output of PV module for a given insolation
tained results are multiplied by the extraterrestrial solar irradiance
level and cell temperature. In this paper, 120 W PV module is
(H0) to get the global solar energy. The rules are summarized in the
selected and operated at maximum power point tracking. The short
fuzzy decision matrix in Table 4.
circuit current and open-circuit voltage of PV module at operating
The fuzzy variables, mean sunshine duration per hour (S/S0),
condition is given by the following equations
temperature ratio (T/T0) and clearness index (H/H0) are described
by the fuzzy terms high, highemedium/normal, medium/normal,
Voc ¼ Voc ðratedÞ  0:0023  hc x ðTcell  Tstc Þ (5)
lowemedium/normal or low. These fuzzy variables described by
linguistic terms are represented by membership functions. The
where
membership functions for sunshine duration, temperature ratio
and global solar energy are shown in Figs. 1e3 respectively.
Tcell ¼ Ta þ 1:25 x ðNOCT  20Þ  H (6)
The rules listed in Table 4 have been implemented using fuzzy
logic toolbox of MATLAB for developing the model to estimate the The fill factor of a PV module is given by
global solar energy shown in Fig. 4.
Pmax ðratedÞ
Fill factor ¼    (7)
4. Fuzzy model for the prediction of SPV system output for Is at 1000W m2  Voc ðratedÞ
smart grid application
Maximum power of a module at operating condition is given by
The integration of renewable energy sources in smart grid is a
challenging task because of intermittent and unpredictable nature Pmax ¼ Is  Voc  Fill factor (8)

Latitude

Longitude

Altitude Global Solar

Mamdani Energy
Months

Sunshine Duration (S/So)

Temperature (T/To)

Fig. 4. Fuzzy logic based model for global solar energy estimation.

Please cite this article in press as: Rizwan M, et al., Fuzzy logic based modeling and estimation of global solar energy using meteorological
parameters, Energy (2014), http://dx.doi.org/10.1016/j.energy.2014.04.057
 M. Rizwan et al. / Energy xxx (2014) 1e7 5

Table 6 Table 9
Estimated monthly mean global solar energy in comparison with measured data for Estimated monthly mean global solar energy in comparison with measured data for
Delhi. Shillong.

Month Measured (MJ/m2) Estimated (MJ/m2) Month Measured (MJ/m2) Estimated (MJ/m2)

Fuzzy logic ANN Fuzzy logic ANN

January 13.32 12.90 12.76 January 14.11 13.26 13.51

February 16.42 15.80 15.69 February 16.67 15.85 15.97
March 20.64 19.71 19.67 March 19.27 18.21 18.33
April 24.07 23.18 23.10 April 21.13 20.10 20.03
May 24.43 23.67 23.25 May 18.41 17.39 17.46
June 22.54 23.20 23.75 June 16.42 17.60 15.41
July 19.07 18.20 20.14 July 16.06 17.10 17.03
August 17.79 16.86 16.93 August 14.93 15.81 15.84
September 18.90 19.95 20.02 September 14.03 13.30 14.78
October 16.80 16.07 15.98 October 15.18 14.23 14.53
November 14.13 13.73 13.42 November 15.63 14.81 14.97
December 11.93 11.56 11.38 December 14.43 13.72 13.85

Table 7 30
Estimated monthly mean global solar energy in comparison with measured data for
Jodhpur. 25

Solar Irradiance
Month Measured (MJ/m2) Estimated (MJ/m2) 20
Fuzzy logic ANN
15
January 15.53 14.92 14.93
February 18.20 17.05 17.43 10
March 21.76 20.69 20.78
April 24.24 23.30 23.08 5
May 25.10 23.56 23.79
June 23.58 22.17 24.90 0
July 19.67 20.93 20.84 1 2 3 4 5 6 7 8 9 10 11 12
August 19.51 20.90 18.33 Months
September 21.06 19.80 19.99
October 19.11 18.27 18.34 Fig. 6. Measured global solar energy (B) with the estimated values () in MJ/m2 for
November 16.17 15.60 15.39 New Delhi.
December 14.84 14.32 14.13

management, optimized resource utilization, grid safety and grid

where Is is the short circuit current at 1000 Wm2, Ta is the ambient stability.
temperature, Tcell is the cell temperature, Tstc is 25  C, NOCT is the
normal operating cell temperature ( C), H is the estimated solar
5. Results and discussions
insolation (Wm2), hc is the efficiency of PV module.
Average power output of SPV system for smart grid application
Fuzzy logic based model is developed and presented for the
can be intelligently predicted using the fuzzy logic based model
estimation of global solar energy using the above mentioned data
presented in Fig. 5. Fuzzy rules for the prediction of SPV system
for New Delhi, Jodhpur, Kolkata and Shillong stations. The output of
output are presented in Table 5. In the proposed model solar
the model is clearness index (H/H0). Obtained results are multiplied
insolation and cell temperature are taken as input parameters and
by the extraterrestrial solar irradiance (H0) to get the global solar
the output of the model is PV power generated by the given
irradiance (H). Results of fuzzy logic based models are compared
module. This data can then be used for variety of purposes which
with the well established REST2 (Reference Evaluation of Solar
among others include reliability of supply, intelligent load
Transmittance, 2 band) model [4] for validation. Further ANN based
model considering the same input parameters is also developed
Table 8
Estimated monthly mean global solar energy in comparison with measured data for
Kolkata. 30
2 2
Month Measured (MJ/m ) Estimated (MJ/m ) 25
Solar Irradiance

Fuzzy logic ANN

20
January 16.15 15.62 15.63
February 19.21 18.15 18.45 15
March 21.93 20.65 20.86
April 23.95 22.45 22.70 10
May 23.59 22.38 22.40
June 18.85 20.20 19.96 5
July 14.81 15.74 15.57
August 14.78 15.62 13.98
0
1 2 3 4 5 6 7 8 9 10 11 12
September 17.54 16.46 18.49
Months
October 18.66 18.07 17.87
November 16.15 15.62 15.63
Fig. 7. Measured global solar energy (B) with the estimated values () in MJ/m2 for
December 19.21 18.15 14.73
Jodhpur.

Please cite this article in press as: Rizwan M, et al., Fuzzy logic based modeling and estimation of global solar energy using meteorological
parameters, Energy (2014), http://dx.doi.org/10.1016/j.energy.2014.04.057
6 M. Rizwan et al. / Energy xxx (2014) 1e7

30 Table 10
Monthly predicted SPV system output in comparison with computed power for New
25 Delhi.
Solar Irradiance

20 Months Solar insolation Cell temp Computed Predicted power

(W/m2) ( C) power (W) (W)
15
January 425.28 28.47 46.12 44.00
10 February 524.26 34.83 55.45 52.80
March 637.03 44.08 64.9 61.70
5 April 689.28 52.94 67.61 64.30
May 699.59 56.95 67.49 63.80
0
1 2 3 4 5 6 7 8 9 10 11 12 June 666.07 55.98 64.49 60.87
Months July 630.62 52.29 62.08 57.85
August 633.54 51.24 62.66 67.07
September 610.46 44.6 62.02 66.06
Fig. 8. Measured global solar energy (B) with the estimated values () in MJ/m2 for
October 486.11 42.6 49.81 52.4
Kolkata.
November 451.14 36 47.5 45.40
December 409.12 29.07 44.28 42.16

and presented in Appendix. The estimated global solar energy us-

ing the above mentioned models is compared with the measured
data for New Delhi, Jodhpur, Kolkata and Shillong stations are very close to each other. It shows that the performance of the model
presented in Tables 6e9 respectively. The fuzzy logic and ANN is satisfactory. Obtained results of solar insolation are further
based models are developed in MATLAB toolboxes. The perfor- simulated along with cell temperature as input parameters for the
mance of the models is evaluated on the basis of mean absolute prediction of output power of the given SPV system. As the cell
relative error in percentage. The mean absolute relative error is temperature of a PV module affects the output of the PV system.
calculated as The predicted output of SPV system in comparison with the
computed power for New Delhi is presented in Table 10. Due to
jHmeasured  Hestimated j paucity of space the results of other stations are not presented in
Absolute relative error ¼  100 (10) this paper. The predicted power of SPV system in comparison with
Hmeasured
computed power is quite accurate. The mean absolute relative error
The mean percentage error using fuzzy logic for New Delhi, for New Delhi, Jodhpur, Kolkata and Shillong is 5.41%, 5.47%, 5.59%
Jodhpur, Kolkata and Shillong are 5.11%. 5.17%, 5.53% and 5.67% and 5.75% respectively. Therefore, the proposed models may be
respectively whereas it is 4.87%, 4.89%, 4.81% and 5.97% respec- used for the prediction of solar energy and PV power generation.
tively using ANN. The results obtained from REST2 model are 5.54%, Further, the obtained results may be used for smart grid
5.70%, 6.26% and 6.47% respectively. As REST2 model is a clear sky application.
model and does not incorporate the uncertainties in the weather,
therefore intelligent models for the estimation of solar energy are
6. Conclusions
required. It is fact that the performance of REST2 model is better as
compared to other models available in the literature. In addition,
In the proposed work, a fuzzy logic based model for esti-
the data of transmittances is not available during monsoon month
mating the global solar energy for a particular station is pre-
(JulyeSeptember). Therefore it is difficult to estimate the solar
sented. The meteorological parameters like mean duration
energy during above mentioned period. Further, the results ob-
sunshine per hour and temperature and the geographical pa-
tained from ANN model are slightly better as compared to the
rameters of the particular location are considered as input pa-
proposed model but it can be compromised with other advantages
rameters. Since there is always ambiguity in climatic conditions
of fuzzy logic like flexibility, tolerant of imprecise data, model for
and it is difficult to predict solar energy accurately by using
nonlinear functions of arbitrary complexity, blended with con-
mathematical formulas and regression techniques at a particular
ventional control techniques, based on natural language apart from
time and at a particular location that is utmost important in the
easy to develop and better understanding. In addition, the esti-
design of solar based power station. However such problems are
mated values obtained from fuzzy logic based model as compared
overcome by fuzzy logic based techniques. Based on the data
to measured data are also represented graphically in Figs. 6e9 for
generated, fuzzy logic is applied to estimate the solar energy at
New Delhi, Jodhpur, Kolkata and Shillong respectively. It is clearly
four Indian stations that covers different climatic as well as
seen from Figs. 6e9 that the measured and estimated values are
geographical conditions.
The results obtained from the proposed fuzzy logic based model
30 are compared with reference data and it is found that the per-
centage error is within permissible limits that are about 5% for all
25 stations considered in this work.
By using proposed model other uncertainties may also be
Solar Irradiance

20
included easily in estimating solar energy more accurate for any
15 location that are required to set up solar based power stations.
Further, the estimation of solar energy and prediction of SPV output
10
would be helpful in smart grid application.
5

0 Appendix
1 2 3 4 5 6 7 8 9 10 11 12
Months
For the estimation of global solar energy at New Delhi, Jodhpur,
Fig. 9. Measured global solar energy (B) with the estimated values () in MJ/m2 for Kolkata and Shillong stations, an ANN based model has been
Shillong. developed and presented in Fig. 10.

Please cite this article in press as: Rizwan M, et al., Fuzzy logic based modeling and estimation of global solar energy using meteorological
parameters, Energy (2014), http://dx.doi.org/10.1016/j.energy.2014.04.057
 M. Rizwan et al. / Energy xxx (2014) 1e7 7

Fig. 10. ANN model for the estimation of global solar energy.

References [12] Sen Z. Fuzzy algorithm for estimation of solar irradiation from sunshine
duration. Sol Energy 1998;63:39e49.
[13] Gautam NK, Kaushika ND. A model for the estimation of solar radiation using
[1] Rizwan M, Jamil M, Kothari DP. Solar energy estimation using REST model for
fuzzy random variables. J Appl Meteorology; 2002:1267e76.
PV-ECS based distributed power generating system. Sol Energy Mater Sol Cells
[14] Paulescu M, Gravila P, Paulescu ET. Fuzzy logic algorithm for atmospheric
2010;94(8):1324e8.
transmittances of use in solar energy estimation. Energy Convers Manag
[2] Ahmed MA, Ahmed F, Akhter MW. Estimation of global and diffuse solar ra-
2008;49:3691e7.
diation for Hyderabad, Pakistan. J Basic Appl Sci 2009;5:73e7.
[15] Paulescu ET, Paulescu M. Fuzzy modeling of solar irradiation using air tem-
[3] Gupta N, Jain A. Smart grids in India. Renew Energy (Akshay Urja) 2011;5:38e41.
perature data. Theor Appl Climatol 2008;91:181e92.
[4] Gueymard CA. REST2: high performance solar radiation model for cloudless
[16] Gomez V, Casanovas A. Fuzzy logic and meteorological variables: a case study
sky irradiance, illuminance, and photosynthetically active radiation-validation
of solar irradiance. Fuzzy Sets Syst 2002;126:121e8.
with benchmark dataset. Sol Energy 2008;82:272e85.
[17] Gomez V, Casanovas A. Fuzzy modeling of solar irradiance on inclined sur-
[5] Gueymard CA. Direct solar transmittance and energy predictions with
faces. Sol Energy 2003;75:307e15.
broadband models. Sol Energy 2003;74:355e79.
[18] Iqdour R, Zeroual A. Prediction of daily global solar radiation using fuzzy
[6] Rizwan M, Jamil M, Kothari DP. Estimation of global solar irradiance using
systems. Int J Sustain Energy 2005;26:19e29.
regression model for Jodhpur, India. In: Proceedings of International Sympo-
[19] Hidayatullah NA, Stojcevski B, Kalam A. Analysis of distributed generation
sium on Photovoltaic Science & Technology, Kanpur; 2010. p. 43.
systems, smart grid technologies and future motivators influencing change in
[7] Kaplanis S, Kaplani E. Stochastic prediction of hourly global solar radiation for
the electricity sector. Smart Grid Renew Energy 2011;2:216e29.
Patra Greece. Appl Energy 2010;87:3748e58.
[20] Kuschke M, Strunz K. Modeling of tidal energy conversion systems for smart
[8] Kaplanis S, Kaplani E. A model to predict expected mean and stochastic hourly
grid operation. In: Proceedings of IEEE Power and Energy Society General
global solar I (h; nj) values. Renew Energy 2007;32:1414e25.
Meeting, Detroit; 2011. pp. 1e2.
[9] Kaplanis SN. New methodologies to estimate the hourly global solar radiation;
[21] Solar radiation handbook. Pune: SEC & IMD; 2008.
comparison with existing models. Renew Energy 2006;31:781e90.
[22] Ng HN, Salama MMA, Chikhani AY. Capacitor allocation by approximate
[10] Falayi EO, Adepitan JO, Rabiu AB. Empirical models for the correlation of
reasoning: fuzzy capacitor placement. IEEE Trans Power Deliv 2000;15:
global solar radiation with meteorological data for Iseyin, Nigeria. Int J Phys
393e8.
Sci 2008;3:210e6.
[23] Duffie JA, Beckmen WA. Solar energy thermal processes. New York: John
[11] Akpabio LE, Etuk SE. Relationship between global solar radiation and sunshine
Wiley & Sons, Inc; 1980.
duration for Onne, Nigeria. J Phys 2003;27:161e7.

Please cite this article in press as: Rizwan M, et al., Fuzzy logic based modeling and estimation of global solar energy using meteorological
parameters, Energy (2014), http://dx.doi.org/10.1016/j.energy.2014.04.057