Energy Conversion and Management 89 (2015) 497–506

Contents lists available at ScienceDirect

Energy Conversion and Management

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

On the fast convergence modeling and accurate calculation of PV output

energy for operation and planning studies
Reza Navabi b, Sajjad Abedi a,⇑, Seyed Hossein Hosseinian b, Ranadip Pal a
a
Electrical and Computer Engineering Department, Texas Tech University, Lubbock, 79409 TX, United States
b
Electrical Engineering Department, Amirkabir University of Technology, Tehran, Iran

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

Article history: Optimal planning of energy systems greatly relies upon the models utilized for system components. In
Received 3 April 2014 this paper, a thorough modeling framework for photovoltaic (PV) power plants is developed for applica-
Accepted 27 September 2014 tion to operation and planning studies. The model is a precise and flexible one that reflects all the effec-
Available online 28 October 2014
tive environmental and weather parameters on the performance of PV module and inverter, as the main
components of a PV power plant. These parameters are surface radiation, ambient temperature and wind
Keywords: speed. The presented model can be used to estimate the plant’s output energy for any time period and
Photovoltaic power plant
operating condition. Using a simple iterative process, the presented method demonstrates fast and accu-
Simulation
Planning
rate convergence by merely using the limited information provided by manufacturers. The results
PV module modeling obtained by the model are verified by the results of System Advisor Model (SAM) and RETScreen in var-
ious operational scenarios. Furthermore, comparison of the simulation results with a real power plant
outputs and the comparative statistical error analysis confirm that our calculation procedure merits over
SAM and RETScreen, as modern and popular commercial PV simulation tools.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction PV modules form the main part of PV power plants. The model
used for PV modules has a substantial influence on the calculation
For years, conventional power plants relying on fossil fuels have of output energy and requires a great deal of investigation [3]. Sev-
been the main source of energy to supply the world’s electricity eral models have been pointed out for the PV modules that are
demand. The environmental concerns regarding carbon emissions known as performance models. The simplest one is the efficiency
of fossil fuels persuade governments to extend the utilization of model that just uses the product of module efficiency and the solar
renewable energy resources, such as wind and solar energy [1]. irradiation to calculate the output energy of the module [4]. It is
Solar energy offers some fascinating features that merits over obvious that this model cannot provide enough accuracy because
other energy resources. The sunlight is ubiquitous and can be of its substantial simplification assumptions [5,6]. Another model
directly converted to electricity by photovoltaic (PV) cells. This is the algebraic model [7]. Although this model is more accurate
technology has many advantages such as absence of moving parts, than the efficiency model, it requires a lot of information not pro-
low maintenance cost, no noise emission, and simple operation vided in manufacturer’s datasheets [6]. The third commonly used
[2]. On the other hand, the drawbacks of the PV system are the high model is the five-parameter model [6,8]. In this model, the PV
initial investment cost and low efficiency of the solar cells. Further- module is characterized by an equivalent circuit with five param-
more, the PV output energy highly depends on solar insolation and eters specifying the circuit components. The output energy can
ambient temperature. Considering fluctuations of the PV output be calculated based on a simple numerical circuit analysis and this
power by variation of the weather conditions, accurate prediction method leads to more accurate results [8–12].
of the output power is a challenging problem. The estimation of out- Several algorithms have been reported for deriving the five
put power is crucial in economic and technical planning of PV sys- parameters model based on the basic information provided in
tems, such as optimal sizing, placement and optimal power manufacturer’s datasheet (i.e. three current–voltage pairs of PV
management of single or hybrid PV-integrated energy systems [3,4]. module at Standard Rating Condition (SRC)). In [9,10], Genetic
Algorithm (GA) is implemented to obtain the parameters of the
PV module but this algorithm uses a complex procedure and needs
⇑ Corresponding and co-first author. large amount of experimental data [13]. A similar method
E-mail address: [email protected] (S. Abedi). based on Differential Evolutionary (DE) optimization algorithm is

http://dx.doi.org/10.1016/j.enconman.2014.09.070
0196-8904/Ó 2014 Elsevier Ltd. All rights reserved.
498 R. Navabi et al. / Energy Conversion and Management 89 (2015) 497–506

introduced in [14]. This method also suffers from unnecessary tools have been developed and utilized. Modeling of the current–
repetitive calculations to obtain the parameters at all operating voltage curve of a photovoltaic (PV) module for outdoor conditions
points. However, in this paper, all the operating point values are has been discussed [21]. However, this model does not account for
determined just by simple modifications to the values calculated exact modeling of incidence solar irradiation. The simulation of a
at Standard Test Condition (STC). grid connected PV system is carried out using an already-devel-
As can be observed, most of these methods are rather compli- oped model, namely TRNSYS tool, and the outputs are compared
cated, usually need a large amount of experimental data, and suffer with measured data [22]. In that work, there is no original model-
from high computational burden [9,13]. Hence, resorting to simpli- ing improvements relative to the utilized software tool.
fying approaches such as decreasing the original model’s degree is This paper proposes a simple and fast method to determine the
sometimes inevitable to gain more simplicity by losing accordance overall PV plant parameters without any modification to the
with practical conditions. For instance, four or three-parameter model. The proposed algorithm is formulated based on the decom-
models can be mentioned [15]. A more effective solution is to position of the equation set into two parts which are solved sepa-
develop an algorithm which can produce accurate results for the rately by an efficient and simple iterative process. The presented
original model in a simple and fast procedure, and without using method demonstrates high accuracy and fast convergence in less
an extra mathematical solution method. than five iterations. Besides, the presented model offers a low com-
Besides the electrical model of PV module, incident solar irradi- plexity calculation procedure to describe the performance of the
ation and cell temperature are two key factors influencing the out- plant, eliminating the need to apply an external mathematical sol-
put energy. The use of sky models is necessary to determine the ver. Taking into account the surface radiation, ambient tempera-
effective incident radiation on the module surface. The total solar ture and wind speed, the model can accurately estimate the
insolation is comprised of two main components including beam plant’s output energy at any operating conditions based on limited
radiation and diffuse radiation. The simplest sky model is the iso- information provided in datasheets. The calculation results are
tropic model which assumes that the diffuse radiation is uniformly compared with two software tools, namely SAM and RETScreen,
distributed across the sky [16]. This model gives acceptable results as well as a real power plants output data. Results obtained by
for overcast skies but underestimates the radiation in clear skies. the presented calculation procedure show more accuracy com-
Two other models have been developed to overcome this problem, pared to the results of SAM and RETScreen, which use different
namely Hey-Davies model [17] and Reindl model [18]. In Hey- models.
Davies model, the circumsolar component of the diffuse radiation The rest of this paper is organized as follows. Sections 2–5
is added to the isotropic radiation to represent the overcast condi- describe the PV module model, surface radiation model (sky
tions. The Reindl model offers more precision by accounting for the model), cell temperature model and inverter model, respectively.
horizon brightening [18,19]. Therefore, Reindl model is chosen as The combination of these models to build up the PV plant model
the default sky model in this study. is illustrated in Section 6. Section 7 provides simulation results
The PV inverter is also another main component of a PV plant. and model verification, as well as error analysis considering
The role of the inverter model in calculation of the output energy measurement data from a real PV power plant.
is negligible [20]. The simplest model of PV inverter is the com-
monly-used efficiency model which estimates the AC output 2. PV module model
energy simply by multiplying the constant inverter efficiency by
the input energy. This model is naive and suffers from lack of accu- The equivalent circuit of PV module with respect to the five-
racy. The second model is the algebraic model which simulates the parameter model is depicted in Fig. 2 [23].
variable efficiency of the inverter under different energy and volt- The nonlinear output characteristic of the PV module at SRC,
age inputs. This model is simple and presents sufficient closeness where the insolation is 1000 W/m2 and the ambient temperature
to realistic conditions for consideration in planning problems. is 25 °C, is expressed by the following equation:
The overall model of the PV plant is obtained by putting
h VþIRs i V þ IR
together the models of sky, PV module and inverter. Fig. 1 depicts s
I ¼ I L  Io e a  1  ð1Þ
the main components of PV power plant model for estimation of its Rsh
output power. Based on the obtained models, several simulation
where I, V, RS and RSH are the output current, output voltage, series
and parallel resistance of the PV module, respectively. IL is the
photocurrent of the solar module and I0 is the reverse saturation
Weather & Location Datasheet information current of the diode. a is the ideality factor of temperature, defined
information by:

Ns nkT c
Cell temperature a¼ ð2Þ
q
model
where q is the electric charge, K is the Boltzmann constant and n is
Electrical model an ideality factor of the diode in the equivalent circuit. The value of
Sky model Inverter
of module n is between 1 and 2. Tc is the module temperature in Celsius.
model

PV module model I
No. of modules and
inverters Rs

IL Io Ish Rsh V

PV plant output power

Fig. 1. Structure of PV power plant model. Fig. 2. Equivalent circuit of PV module.

 R. Navabi et al. / Energy Conversion and Management 89 (2015) 497–506 499

As it can be seen, Rs, Rsh, I0, IL and a are the five characteristic Step (1), Initialization
parameters that describe the performance of the module. For the first iteration, initial values of the model parameters can
be chosen as:
2.1. Calculation of parameters at SRC
V mp
að0Þ  ð12Þ
n  V th
The five-parameter model must be determined at different
operating conditions based on the information provided by manu- bð0Þ ¼ 1 ð13Þ
facturers in the module’s datasheet. Usually, three current–voltage
pairs at SRC are used to calculate the parameters of the module: Rsh ð0Þ ¼ 1 ð14Þ
the short circuit current (Isc), the open circuit voltage (Voc) and
the current and voltage at the maximum power point of the PV In order to obtain an equation for n, it should be noted that in
output curve (Imp, Vmp). Based on these values, Eq. (1) can be the short circuit condition, the diode current can be neglected
rewritten at the three different operating points of the module to compared to the short circuit current. Therefore, Eq. (3) can be
obtain an equation set comprising of five equations, as follows: rewritten as:
h Isc Rs i I R Isc ðRs þ Rsh Þ
sc s
Isc ¼ IL  Io e a  1  ð3Þ IL  ð15Þ
Rsh Rsh

h V oc i V Also, the following equation will be obtained for Io by substitut-

oc
0 ¼ I L  Io e a  1  ð4Þ ing Eq. (15) in Eq. (5):
Rsh    
ðIsc  IMP ÞðRs þ Rsh Þ  V MP V MP þIMP Rs

h i Io ¼ e  a ð16Þ
V mp þImp Rs V mp þ Imp Rs Rsh
Imp ¼ IL  Io e a 1  ð5Þ
Rsh
Using Eqs. (15) and (16) in Eq. (1), n can be given by:

dðIVÞ V oc  V MP  IMP  Rs
¼0 ð6Þ  
dV mp n¼
Isc ðRs þRsh ÞV oc
ð17Þ
V th Ln ðIsc I MP ÞðRs þR ÞV MP
sh

dðIÞ 1 This equation shows that n is related to Rs, while Rs is unknown
¼ ð7Þ
dV sc Rsh in the first iteration. The initial value for the parameter is obtained
In the above equations, all of the parameters are expressed at as n(0) = min {n1(0), n2(0), n3(0)} [8], where:
SRC. Eq. (3) is derived from the short circuit condition, Eq. (4) is 8 V V V
>
> n ð0Þ ¼ oc mp th
written based on open circuit condition, Eq. (5) represents the > 1
> V th Ln I IIsc
>
> sc mp
equivalent circuit at maximum power point of the module, Eq. >
< V V
(6) is the derivative of the output power to yield the maximum n2 ð0Þ ¼ oc  mp  ð18Þ
power point, and the derivative of the current at short circuit con- >
> V th Ln V
V mp
>
> th

dition is related to Rsh by Eq. (7). These five equations must be >
>
>
: n ð0Þ ¼ ðIsc Imp ÞðV mp V th Þ
simultaneously solved to determine the parameters of the PV mod- 3 Isc V th

ule. As can be observed from these equations, the PV module

Given the value of n(0) , the initial value of Rs can be obtained
model has nonlinear characteristics. Hence, numerical methods
by:
in several iterations should be employed. To obtain the model  
parameters, an efficient solution algorithm is presented in this 1 V mp
Rs ð0Þ ¼ V oc  V MP  nð0Þ  V th Ln ð19Þ
paper and described in the following steps: IMP nð0Þ  V th
Step (0), Preparation
Step (2), Iteration
Before solving the equations, some reasonable approximations
In the forthcoming iterations, Rsh and Rs are available from their
can be made that allow a compact and efficient algorithm. The cal-
previous iteration. So n(k) can be given by:
culation process is decomposed into two levels. At the first level, Rs,
Rsh and n are determined by an iterative procedure. Then, IL and Io nðkÞ ¼ min ðn1 ðkÞ; n2 ðkÞ; n3 ðkÞÞ ð20Þ
are directly calculated in the second level.
Using Eq. (17), the three limits for n can be reformulated as
By using Eqs. (4) and (5) and applying some simplifications, Rs
follows:
can be written as:
8
1 >
> n1 ðkÞ ¼  V oc V MP IMP Rs ðk1Þ 
Rs ¼ ½V oc  V MP  n  V th Lnða  bÞ ð8Þ >
> Isc ðRs ðk1ÞþRsh ðk1ÞÞV oc
IMP >
< V th Ln ðIsc IMP ÞðRs ðk1ÞþRsh ðk1ÞÞV MP

V oc V mp ð21Þ
where: >
> n2 ðkÞ ¼ V Lnðaðk1Þbðk1ÞÞ
>
> th
>
: n3 ðkÞ ¼ ðIsc Imp ÞðV mp Imp Rs ðk1ÞÞ
V MP þ nV th  IMP Rs Isc V th
a¼ ð9Þ
nV th
where (k) and (k  1) index the value of parameters in the current
Isc ðRs þ Rsh Þ  V oc and previous iterations, respectively.
b¼ ð10Þ
Isc ðRs þ Rsh Þ  2V MP By using n(k), Rs(k) and Rsh(k) can be calculated by:

Ns kT c 1
V th ¼ ð11Þ Rs ðkÞ ¼ ½V oc  V MP  nðkÞ  V th Lnðaðk  1Þ  bðk  1ÞÞ ð22Þ
q IMP
As it is obvious, a, b and Rs are related to each other; and ðV MP  IMP Rs ðkÞÞV MP  nðkÞV th V MP
thus, an iterative numerical method is necessary to solve these Rsh ðkÞ ¼  Rs ðkÞ ð23Þ
ðV MP  IMP Rs ðkÞÞðIsc  IMP Þ  nðkÞV th IMP
equations.
500 R. Navabi et al. / Energy Conversion and Management 89 (2015) 497–506

Step (2-1), Mutation

Get Vmp, Imp, Voc,
The set of values obtained in the current iteration are slightly Isc from datasheet
altered to generate a new solution set. Checking the new set of values
First iteration
which is close to the solution set in the current iteration can increase
the convergence rate. The mutation step is performed as follows: Set Rsh (0) by (14)
n0 ðkÞ ¼ wn nðkÞ ð24Þ

R0s ðkÞ ¼ ws Rs ðkÞ ð25Þ Calculate n(0) by (18)

R0sh ðkÞ ¼ wsh Rsh ðkÞ ð26Þ

Calculate Rs(0) by (19)
where wn, ws and wsh are the weighting coefficients and randomly
chosen between [0.7, 1.3].
Step (2-2), Selection Other iterations
0
After calculating n (k),R0s ðkÞ and R0sh ðkÞ, the values of I0L ðkÞ and
Calculate n(k) by (21)
I0o ðkÞ are calculated by (14) and (15), respectively. To evaluate the
mutated parameters, the F = [f1, f2,. . ., f5] vector is defined as the
error vector with the following elements:
Calculate Rs(k): by (22)
 0 
Isc Rs Isc R0
f1 ¼ I0L I0o
 e a0  1  0 s  Isc ð27Þ
Rsh
Calculate Rsh (k): by (23)
h V oc
i V oc
f2 ¼ I0L  I0o ea0 1  0 ð28Þ
Rsh
Calculate the mutant parameters
  by (24),(25),(26)
V mp þImp R0s V mp þ Imp R0s
f 3 ¼ I0L  I0o e a0 1   Imp ð29Þ
R0sh
Calculate IL and Io based on the
 mutant parameters by (8), (9)
dðIVÞ
f4 ¼ ð30Þ
dV mp
Calculate vector F: by (27)-(31)

dðIÞ 1
f5 ¼ þ ð31Þ
dV sc R0sh Set the old
parameters equal Yes
Norm(F)<λ
In each iteration, if the norm of the error vector F is smaller than to the new
a predefined value k (Norm(F) < k), the iterative process can be ter- parameters
No
minated and the new values are introduced as the solution. In
other words, a pre-convergence criterion is applied to determine No
|Rs(k)- R s(k-1)|<ε
if the mutation improves the convergence and we can bypass fur-
ther iterations (Fig. 3). The choice of value for k depends on the
Yes
desired convergence accuracy.
Step (2-3) Convergence Calculate IL and Io: by (15), (16)
In this step, a final convergence criterion should be applied to
terminate the iterative process. Since the model is more sensitive
to variations of the series resistance, this parameter is used as con- n, Rs h, Rs, Io, IL
trol signal. The iteration can be terminated if the following condi- are calculated
tion is met:
Fig. 3. Flowchart of parameters calculation.
jjRs ðkÞ  Rs ðk  1Þjj < e ð32Þ
In this case, the final result of Rsh, Rs and n are also set to their variations and make the model comprehensive and practical for
last values before convergence. every environmental condition. If the value of parameters at the
Step (3), Finalization STC is considered as reference, the values at NSRC can be deter-
The calculation process is finalized by calculation of the second mined using modification functions for each of the parameters.
level parameters. Considering Rsh, Rs and n available from results of The ideality factor (a) is a linear function of temperature, as
the previous level, Io and IL are determined by Eqs. (15) and (16). expressed by Eq. (2), so the modification equation for ideality fac-
The summery of this iterative algorithm is depicted in the flow- tor at a new temperature can be expressed by:
chart of Fig. 3.
a Tc
¼ ð33Þ
aref T c;ref
2.2. Parameters in non-standard rating condition (NSRC)
where Tc,ref and aref are the cell temperature and ideality factor at
In the previous section, the five-parameter model was deter- SRC, respectively.
mined at SRC by the illustrated algorithm. Nevertheless, the oper- From the diode theory [6,23], Io is affected by temperature and
ating condition of solar module is not consistent and has its modification equation is given by:
intermittent nature mainly because the insolation level and ambi-  3  
Io Tc 1 Eg Eg
ent temperature vary during a day. So the parameters of equivalent ¼ exp j  jT c ð34Þ
circuit of solar module should be updated to accommodate these
Io;ref T c;ref k T T ref T
 R. Navabi et al. / Energy Conversion and Management 89 (2015) 497–506 501

100 2.3. The maximum power point

95
In the design of PV plants, it is usually supposed that Maximum
Efficiency (%)

90 Power Point Tracker (MPPT) is utilized. Hence, in all operating con-

ditions, the output of the plant is the aggregation of Maximum
85
Power Point (MMP) of the panels. In order to calculate the MPP,
80 Eq. (6) should be solved. In this paper, the Newton–Raphson algo-
rithm [24] is used as the solver. Choosing the short circuit condi-
75 V1
tion for initialization can increase the convergence rate.
V2
V3
70
0 10 20 30 40 50 60 70 80 90 100 3. Surface radiation model
DC input (% of nominal DC input)
The Reindl model is preferred in this paper to calculate the
Fig. 4. Efficiency of the inverter relative to the DC input for three different input
voltages (V1 < V2 < V3).
effective surface radiation [18]. In this model, in addition to isotro-
pic diffuse and circumsolar radiation, the horizon brightening is
also taken into account and the total radiation on a tilted plane
(Gt) is described by the following equation:
where Eg is the material band gap and k is Boltzmann’s constant. Eg
depends on the cell type and its value is available in [6] for different Gt ¼ ðGb þ Gd AÞRbeam þ Gd ð1  AÞ
cell types. Eg has a small dependency on the temperature and " sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi #
neglecting this dependency does not lead to a significant error. So, 1  cosðbÞ Gb 3 b
 1þ sin þ qG G
it is supposed that the values of Eg at SRC and NSRC are equal. 2 Gb þ Gd 2
IL depends on both radiation and temperature and its modifica- 1  cosðbÞ
tion equation is written as:  ð37Þ
2
where Gb is the beam radiation on horizontal plane, Gd is the diffuse
G  
IL ¼ IL;ref þ aIsc T c  T c;ref ð35Þ radiation, G is the total radiation on horizontal plane, b is the tilt
Gref angle of the module, qG is the ground reflectance, Rbeam is a geomet-
ric factor that represents the ratio of beam radiation on a tilted sur-
where Gref is the radiation at SRC and aIsc is the short circuit current face to that on a horizontal surface and A is anisotropy index.
temperature coefficient provided by manufacturers. In order to calculate the total radiation (Gt) by (37), the total
In experimental observations, it is pointed out that the slope of radiation on a horizontal plane (G) and total beam radiation (Gbn)
the I–V curve near the short circuit condition proportionally are taken from the weather database which is available for any
changes with the radiation (G) [6]. In addition, this slope is defined desired location. Other variables in this equation, including Gb, Gd
by Eq. (7), and controlled by Rsh. As a result, the modification equa- and Rbeam should be calculated as discussed in the following
tion for Rsh is written as: paragraphs.
The beam radiation on the horizontal plane is calculated by:
Rsh Gref
¼ ð36Þ Gb ¼ Gbn cos hz ð38Þ
Rsh;ref G
where hz is the zenith angle, i.e., the angle between the vertical and
Finally, the value of Rs can be assumed to be constant in differ- the line to the sun, expressed by:
ent conditions and equal to its value at SRC, because small changes
in Rs show a trivial effect on the characteristic of the model [6]. cos hz ¼ cos / cos d cos x þ sin / sin d ð39Þ

Table 1
List of input parameters for the simulation of the presented PV power plant model.

No. Type Symbol Unit Usage Description

1 Weather info G W/ (42), (52) [Section 3] The total radiation on a horizontal plane
m2
2 Gbn W/ (38), (45) [Section 3] The total beam radiation
m2
3 Ta °C (54) [Section 4] The ambient temperature
4 ws m/s (54) [Section 4] The wind speed
5 Location info b deg (44) [Section 3] The tilt angle of the module
6 c deg (44) [Section 3] The azimuth angle of the module
7 / deg (39), (44) [Section 3] The latitude of the PV plant location
8 qG – (37), (53) [Section 3] The ground reflectance
9 Datasheet info: PV Isc A (3), (10), (15), (16), (18), (21), (27) [Section 2] The module’s short-circuit current at STC
10 module Voc V (4), (8), (10), (17)–(19), (21), (22) [Section 2] The module’s open-circuit voltage at STC
11 Vmp V (5), (6), (8)–(10), (12), (16)–(19), (21)–(23), (29) Voltage at the maximum power point of the PV output
[Section 2] curve at STC
12 Imp A (5), (9), (16)–(19), (21)–(23), (29) [Section 2] Current at the maximum power point of the PV output
curve at STC
13 aIsc – (35) [Section 4] The short circuit current temperature coefficient
14 Datasheet info: C0, C1, C2, C3 – (56)–(58) [Section 5] Coefficients of the inverter DC-AC characteristic curve
inverter
502 R. Navabi et al. / Energy Conversion and Management 89 (2015) 497–506

Get the module and inverter

datasheet parameters and
weather & location information

Calculate the parameters of the

module at SRC based on section 2.1

Set h=1; m=1

Get weather information (G and Gbn)

for hour h

Calculate Gb(h), Gd(h) and Gt(h)

based on section 3

Calculate Tc(h) based on section 4

Calculate Vmp(h), Imp(h) and Pmp(h)

based on sections 2.2 and 2.3

Calculate Pac(h) based on section 5

Did h count all Yes

hours of month m=m+1
m?

No

Pmp(m)=Pmp(m)+Pmp(h)
Pac(m)=Pac(m)+Pac(h)
Gd(m)=Gd(m)+Gd(h)
Gb(m)=Gb(m)+Gb(h)
Gt(m)=Gt(m)+Gt(h)

Yes
h=h+1 Is h < 8760 ?

No

Calculate variables for the first year

based on the month variables

Export output energy

Fig. 5. Flowchart of the hourly PV plant output energy calculation during a year.

where / is the latitude of the PV plant location (north is positive x ¼ ðh  12Þ  15 ð41Þ
and south is negative), d is the declination (angular location) of
where N represents the Nth day of the year and h is the hour of a
the sun at the solar noon, and x is the hour angle that represents
day in 24-h format.
the angular displacement of the sun to east or west of the local
The total radiation on a horizontal plane (G) is sum of diffuse
meridian due to the rotation of the earth on its axis at 15° per hour.
radiation (Gd) and beam radiation on the horizontal plane (Gb), so
d and x are defined by:
Gd is calculated as follows:
284 þ N
d ¼ 23=45 sin 360 ð40Þ Gd ¼ G  Gb ð42Þ
365
 R. Navabi et al. / Energy Conversion and Management 89 (2015) 497–506 503

Table 2
Verification of the module model.

Manufacturer Suntech power Q-cells canadian solar Trina solar Abound solar
Model STP-290 24-vd Q-Smart UFL-95 CS6A 190P TSM 170D AB1 65-A
Cell type Amorphous silicon Amorphous silicon Amorphous silicon Single crystal silicon Amorphous silicon
Voc (V) 45.5 89 29.6 43.6 46.1
Isc (A) 8.42 1.68 8.5 5.25 2.26
Vmp (V) 35.6 66.4 23.9 35.8 35
Imp (A) 8.15 1.43 7.95 4.76 1.88
Pmp at SRC (W) 290.381 94.95 187.83 170.403 65.809
Error at SRC (%) 0.08 0 0.001 0.002 0.001
No. of iterations 3 4 5 5 4
Pmp at NOCT (W) 213 66.1 136.1 118.6 47.4
Error at NOCT (%) 0.9 0.8 0.75 0.2 0.4

Table 3
Parameters for the PV plant model in cases 1 and 2.

Case Module No. of modules Inverter Module Azimuth Ground Location Loss
in array slope (°) angle (°) reflation (%)
Case 1 Suntech power STP 180S-24 100 Satcon Technology PVS 250 20 0 0.2 Arizona – Phoenix 0
Case 2 Grape solar GS-S-250-Fab5 40 Fronius USA, LCC – IG Plus 40 10 0.2 California – Arcata 0
10 uni 240

Another parameter used in (37) is Rbeam, which is defined by: where ns is the reflection index of the module surface. Different val-
ues may be assigned to this index. ns = 1.256 brings about the min-
Gbn cos h cos h imum error [6].
Rbeam ¼ ¼ ð43Þ
Gbn cos hz cos hz The incidence angle modifier is defined by:

where h is the angle of incidence, i.e. the angle between the beam sðhÞ
K sh ¼ ð48Þ
radiation on a surface and the normal to that surface. sð0Þ
where s(h) is called the transmittance through a single cover.
cosðhÞ ¼ sinðdÞ sinð/Þ cosðbÞ  sinðdÞ cosð/Þ sinðbÞ cosðcÞ
" !#
þ cosðdÞ cosð/Þ cosðbÞ cosðxÞ KL
2
1 sin ðhr  hÞ tan2 ðhr  hÞ
sðhÞ ¼ e cos hr 1 þ ð49Þ
þ cosðdÞ sinð/Þ sinðbÞ cosðcÞ cosðxÞ 2 sin2 ðhr þ hÞ tan2 ðhr þ hÞ
þ cosðdÞ sinðbÞ sinðcÞ sinðxÞ ð44Þ
where K is the glazing extinction coefficient and L is the glazing
where c is the azimuth angle of the module. thickness. In this study, K and L are considered 4 m1 and 2 mm,
The last parameter in (37) is the anisotropy index, given by: respectively [5].
Based on Ksh, the modified radiations are as follows:
Gbn
A¼ ð45Þ G0b ¼ K sh Gb ð50Þ
Gon
where Gon is the extraterrestrial radiation that is measured on the
G0d ¼ K sd Gd ð51Þ
plane normal to the radiation.
 
360N qg G0 ¼ K sd qg G ð52Þ
Gon ¼ Gsc 1 þ 0:033 cos ð46Þ
365
where Ksd is the incidence angle modifier at h = 58°. Eventually, the
where Gsc is the solar constant and its ultimate value is 1366.1 W/ total radiation on a tilted plane can be rewritten as:
m2.
G0t ¼ ðG0b þ G0d AÞRbeam þ G0d ð1  AÞ
3.1. Incidence angle modifier " sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi #
1  cosðbÞ G0b 3 b
 1þ sin þ qG G0
As explained in the former section, the incidence angle is the 2 G0b þ G0d 2
angle between the beam radiation on the module and the nor- 1  cosðbÞ
mal to that module. As this angle increases, the amount of  ð53Þ
2
reflection from the surface of the module increases. It is impor-
tant to calculate the amount of radiation which is absorbed by
the module ðG0t Þ. Hence, incidence angle modifiers are defined 4. Cell temperature model
for beam (Gb), diffuse (Gd) and ground-reflected (qgG) radiations.
At first, the reflection angle from the surface of the module The cell temperature is used in the modification equations of
should be calculated by the following equation: the panel ((33)–(36)). However, the available temperature in the
install location of the PV plant is the ambient temperature. Thus,
1 sinðhÞ
hr ¼ sin ð47Þ it is important to find a model which converts the ambient temper-
ns ature to the cell temperature based on the panel and weather
504 R. Navabi et al. / Energy Conversion and Management 89 (2015) 497–506

conditions. The effective radiation on the tilted panel and the wind Table 4
speed are the two most important factors affecting the cell temper- Energy output of the PV plant (kW h) for case 1.

ature. The following temperature model is used in this paper [23]: Month SAM RETScreen Proposed model

0:32 January 2099.77 2151 2079.066

Tc ¼ Ta þ G0 ð54Þ February 2337.21 2309 2264.174
8:91 þ 2ws t March 2892.8 2954 2852.719
April 3308.13 3265 3264.301
where G0t is obtained from (53), Ta is the ambient temperature in °C
May 3535.15 3478 3548.463
and ws is the wind speed in m/s, taken from the weather database of June 3286.48 3328 3406.897
the desired location. July 3279.3 3174 3318.515
August 3169.34 3117 3066.944
September 2937.8 2861 3024.842
5. The inverter model October 2786.09 2727 2747.26
November 2189.58 2234 2175.138
From the viewpoint of the output power, the efficiency of the December 2005.81 2040 1996.598
inverter is the determinative factor which varies with input power Annual 33827.8 33,638 33744.9
and voltage alterations due to different insolation level and tem-
perature. An inverter efficiency model is introduced in [23] that
uses an algebraic equation to relate the AC output of the inverter
to the DC input power and voltage: these modules are various and obtaining the expected results over
  the whole ranges can show the effectiveness of the model for each
Paco
Pac ¼  CðA  BÞ ðPdc  BÞ þ CðP dc  BÞ2 ð55Þ module. Table 2 shows the cell type of each module as well as Voc,
AB
Isc, Vmp and Imp at SRC, provided by its manufacturer. In addition,
this table contains the calculated maximum output power for each
A ¼ Pdco ½1 þ C 1 ðV dc  V dco Þ ð56Þ
module (Pmp) at SRC and Nominal Operating Cell Temperature
(NOCT), where the radiation is 800 W/m2 and the ambient temper-
B ¼ Pso ½1 þ C 2 ðV dc  V dco Þ ð57Þ
ature is 20 °C. The output power of each module at the maximum
power point is calculated by the model and compared to its data-
C ¼ C o ½1 þ C 3 ðV dc  V dco Þ ð58Þ
sheet value. The resultant error is also depicted in Table 2. The
where Pac and Pdc are the AC output and the DC input of the inverter results confirm that the amounts of the errors are too small and
and Paco and Pdco are their nominal values, respectively. Pso is the DC are negligible for a PV plant model. It can be seen that the maxi-
power which is required to start the inversion process (also known mum error is 0.08%.
as self-consumption of the inverter), Vdc is the input DC voltage, Vdco
is the nominal input DC voltage, C0 is a constant defining the curva-
7.1. Comparison with SAM and RETScreen
ture of the relationship between the AC power and DC power at the
nominal operating condition, C1 is a constant that allows Pdco to
As another way to verify the presented PV output energy calcu-
change linearly with the DC input voltage, C2 is a constant that
lation procedure, two different case studies are chosen as pre-
allows Pso to vary linearly with the DC input voltage and C3 is a con-
sented in Table 3. In each case, the monthly output energy of the
stant that allows C0 to vary linearly with the DC input voltage. All of
plant, the monthly total radiation on the tilted module, and the
the mentioned parameters should be provided by manufacturers in
first year output energy of the plant are calculated by the model.
the datasheet.
Results are verified by the output results of System Advisor Model
The effect of the input power and voltage on the inverter effi-
(SAM) [25] and Retscreen [26] softwares, as modern and popular
ciency is depicted in Fig. 4.
commercial PV simulation tools.
Table 4 shows the energy output of the plant in the first year for
6. The overall PV plant system case 1. It can be observed that the output results are close to the
results achieved by the software tools. The existing slight differ-
In previous sections, the detailed models of the system compo- ences are because of differences in the model of components and
nents were discussed. In this section, these models are combined the sky. Fig. 6 shows the total radiation on the tilted module for
together to form the overall model of PV plant. Table 1 tabulates this case.
the list of all input parameters and variables needed in the whole
calculation procedure.
The developed model can provide the following results: the
average value of all the radiation components for a month and
year, the output voltage and current of the modules and the output
power of the plant. These outputs can be accompanied by technical
and economical analyses in optimization problems. The calculation
procedure of the plant model is expressed by a flowchart in Fig. 5.
In this figure, h and m represent the hour and month counters,
respectively. As can be seen in this flowchart, the output power
of modules is calculated based on the weather and datasheet infor-
mation at every hour of a day. The PV plant output is the output of
the inverter which is fed by the PV array.

7. Simulation results

Five different modules are chosen to be simulated by the pro-

posed PV module model. The voltage, current and power range of Fig. 6. Effective radiation (kW h/m2) for case 1.
 R. Navabi et al. / Energy Conversion and Management 89 (2015) 497–506 505

Table 5 Table 7
Energy output of the PV plant (kW h) for case 2. Energy output of the PV plant (kW h) in case 3.

Month SAM RETScreen Proposed model Month Measured output SAM RETScreen Proposed model
January 807.346 836 699.801 January 8740 9950 10,280 9265
February 822.095 873 752.464 February 11,590 13,680 13,921 12,720
March 1169.6 1182 1112.681 March 15,960 17,802 17,751 16,471
April 1418.75 1351 1399.579 April 15,010 15,906 16,160 15,487
May 1507.27 1459 1625.824 May 14,060 15,341 15,052 14,649
June 1403.33 1401 1598.609 June 11,590 10,680 10,825 11,226
July 1517.59 1436 1703.397 July 12,160 11,375 11,698 11,896
August 1356.41 1334 1453.63 August 12,350 11,824 11,367 11,991
September 1289.52 1292 1280.898 September 16,340 15,847 15,975 16,121
October 1078.99 1100 977.905 October 13,490 12,914 12,574 13,056
November 893.036 856 793.32 November 9690 10,526 10,680 9961
December 805.841 788 707.712 December 13,680 13,991 14,121 13,824
Annual 14,069 13,908 14,105 Annual 154,660 159,836 160,404 156,667

Table 8
Statistical error comparison of the employed models.

Month Error (%)

SAM RETScreen Proposed Model
January 17.07 23.03 6.88
February 59.02 56.53 22.61
March 18.82 21.19 6.75
April 2.33 4.39 1.76
May 3.01 1.71 1.41
June 4.99 4.31 2.34
July 5.06 4.25 2.54
August 2.48 3.19 2.50
September 2.81 4.91 1.21
October 3.22 4.60 2.01
November 5.05 6.64 2.54
December 8.16 10.94 5.18
Annual 3.85 4.26 1.35
Fig. 7. Effective radiation (kW h/m2) for case 2. Maximum 59.02 56.53 22.61
Minimum 2.33 1.71 1.21
Mean 11.00 12.14 4.81
Table 5 and Fig. 7 show the energy output of the plant and the
RMSE 495.84 545.14 204.64
total radiation on the tilted module for case 2, respectively. The MBE 0.0360 0.0411 0.0143
accuracy of the proposed model is also noticeable for this case.
The error of the annual energy is very low and equals to 0.2%.
8. Conclusion
7.2. Comparison with measured data from a real power plant
An essential part of economic, forecasting and planning of PV
To examine the accuracy of the proposed method, another case power plants is the assessment of their available power in different
study (case 3) including comparison with measured output from a conditions and time periods. In this paper, a simple, flexible, and
real solar PV power plant in India [27] is carried out and demon- accurate model for the output power of PV power plants is pre-
strated in this section. Table 6 lists the parameters of the real sented. The effect of insolation and temperature are expounded
power plant under study in case 3. The comparative results for case in this paper. Particularly, the availability of all of the model’s
3 are tabulated in Table 7. A comparative statistical error analysis parameters based on the applied algorithm introduces the pre-
is performed using the real data and the simulation results pre- sented model useful in practical planning problems. For model per-
sented in Table 8 which confirms that the most accuracy is formance verification, two different scenarios are simulated and
achieved by using the presented method and our calculation pro- compared with the results of SAM and RETScreen. In addition, a
cedure merits over the methods used in SAM and RETScreen. It real PV power plant is simulated. The mean difference error
can be observed that the least error belongs to the results obtained between the measured output energy data from the real power
by the proposed method in almost all months. The mean error for plant and the simulation results obtained by SAM, RETScreen,
the proposed method is 4.81%, whereas for SAM and RETScreen, it and the proposed method are 11.00%, 12.14%, and 4.81%, respec-
is much more around 11.00% and 12.14%, respectively, which can tively, which unveils more accuracy of the presented calculation
be regarded as a significant improvement. procedure.

Table 6
Parameters for the PV plant model in case 3 [27].

Case Module Number of modules Inverter Module Azimuth Ground Location Loss
in array slope angle reflation
Case 3 TBP1275 M 2620 Normal inverter model 25° 12° 0.2 Khatkar-Kalan, 0%
with 8.7% loss india
506 R. Navabi et al. / Energy Conversion and Management 89 (2015) 497–506

References [13] Carrero C, Ramírez D, Rodríguez J, Platero C. Accurate and fast convergence
method for parameter estimation of PV generators based on three main points
of the I–V curve. Renewable Energy 2011;36:2972–7.
[1] Solangi K, Islam M, Saidur R, Rahim N, Fayaz H. A review on global solar energy
[14] Ishaque K, Salam Z. An improved modeling method to determine the model
policy. Renew Sustain Energy Rev 2011;15:2149–63.
parameters of photovoltaic (PV) modules using differential evolution (DE). Sol
[2] Timilsina GR, Kurdgelashvili L, Narbel PA. Solar energy: markets, economics
Energy 2011;85:2349–59.
and policies. Renew Sustain Energy Rev 2011;16:449–65.
[15] Mahmoud Y, Xiao W, Zeineldin H. A simple approach to modeling and
[3] Abedi S, Alimardani A, Gharehpetian G, Riahy G, Hosseinian S. A
simulation of photovoltaic modules. Sustain Energy IEEE Trans 2012;3:
comprehensive method for optimal power management and design of
185–6.
hybrid RES-based autonomous energy systems. Renew Sustain Energy Rev
[16] Liu BYH, Jordan RC. The interrelationship and characteristic distribution of
2012;16:1577–87.
direct, diffuse and total solar radiation. Sol Energy 1960;4:1–19.
[4] Sajjad Abedi, Hani Gharavi Ahangar, Mostafa Nick, Sayed Hossein Hosseinian.
[17] Hay JE, Davies JA. Calculation of the solar radiation incident on an inclined
Economic and reliable design of a hybrid PV-wind-fuel cell energy system
surface. In: Proceedings first canadian solar radiation workshop; 1980. p. 59–
using differential evolutionary algorithm. In: 19th Iranian conference on
72.
electrical engineering; 2011. p. 1–6.
[18] Reindl D, Beckman W, Duffie J. Diffuse fraction correlations. Sol Energy
[5] Kalogirou S. Solar energy engineering: processes and systems. Massachusetts,
1990;45:1–7.
USA: Academic Press; 2009.
[19] Stanciu, Camelia, Stanciu Dorin. Optimum tilt angle for flat plate collectors
[6] De Soto W, Klein S, Beckman W. Improvement and validation of a model for
all over the World–A declination dependence formula and comparisons of
photovoltaic array performance. Sol Energy 2006;80:78–88.
three solar radiation models. Energy Convers Manage 2014;81:133–43.
[7] King DL, Kratochvil JA, Boyson WE. Photovoltaic array performance
[20] King DL, Gonzalez S, Galbraith GM, Boyson WE. Performance model for grid-
model. United States: Department of Energy; 2004.
connected photovoltaic inverters. Sandia National Laboratories; 2007.
[8] Lo Brano V, Orioli A, Ciulla G, Di Gangi A. An improved five-parameter model
[21] Marion B. A method for modeling the current–voltage curve of a PV module for
for photovoltaic modules. Sol Energy Mater Sol Cells 2010;94:1358–70.
outdoor conditions. Prog Photovoltaics Res Appl 2002;10(3):205–14.
[9] Zagrouba M, Sellami A, Bouaïcha M, Ksouri M. Identification of PV solar cells
[22] Mondol, Deb Jayanta, Yohanis Yigzaw G, Norton Brian. Comparison of
and modules parameters using the genetic algorithms: application to
measured and predicted long term performance of grid a connected
maximum power extraction. Sol Energy 2010;84:860–6.
photovoltaic system. Energy Convers Manage 2007;48(4):1065–80.
[10] Ismail MS, Moghavvemi M, Mahlia TMI. Characterization of PV panel and
[23] Messenger RA, Ventre J. Photovoltaic systems engineering. CRC Press; 2003.
global optimization of its model parameters using genetic algorithm. Energy
[24] Gil A, Segura J, Temme NM. Numerical methods for special functions. Soc Ind
Convers Manage 2013;73:10–25.
Appl Math 2007.
[11] Bai, Jianbo, Liu Sheng, Hao Yuzhe, Zhang Zhen, Jiang Meng, et al. Development
[25] http:// sam.nrel.gov/.
of a new compound method to extract the five parameters of PV modules.
[26] http://www.retscreen.net/.
Energy Convers Manage 2014;79:294–303.
[27] Sharma, Vikrant, Chandel SS. Performance analysis of a 190 kWp grid
[12] Zhu, Xue-Gui, Fu Zhi-Hong, Long Xing-Ming. Sensitivity analysis and more
interactive solar photovoltaic power plant in India. Energy 2013;55:476–85.
accurate solution of photovoltaic solar cell parameters. Sol Energy 2011;85(2):
393–403.