Applied Energy 86 (2009) 163–169

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Optimal design and techno-economic analysis of a hybrid solar–wind power

generation system
Yang Hongxing a,*, Zhou Wei a, Lou Chengzhi b
a
Renewable Energy Research Group, The Hong Kong Polytechnic University, Hong Kong
b
School of Environment Science and Technology, TianJin University, TianJin, China

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

Article history: Solar energy and wind energy are the two most viable renewable energy resources in the world. Good
Received 30 January 2008 compensation characters are usually found between solar energy and wind energy. This paper recom-
Received in revised form 5 March 2008 mend an optimal design model for designing hybrid solar–wind systems employing battery banks for cal-
Accepted 7 March 2008
culating the system optimum configurations and ensuring that the annualized cost of the systems is
Available online 1 May 2008
minimized while satisfying the custom required loss of power supply probability (LPSP). The five decision
variables included in the optimization process are the PV module number, PV module slope angle, wind
Keywords:
turbine number, wind turbine installation height and battery capacity. The proposed method has been
Hybrid solar–wind system
LPSP
applied to design a hybrid system to supply power for a telecommunication relay station along south-
Annualized cost of system east coast of China. The research and project monitoring results of the hybrid project were reported, good
Genetic algorithm complementary characteristics between the solar and wind energy were found, and the hybrid system
. turned out to be able to perform very well as expected throughout the year with the battery over-dis-
charge situations seldom occurred.
Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction
Alternative energy resources, such as solar energy and wind en-
ergy, have attracted energy sectors to generate power on a large
scale. A drawback, common to wind and solar options, is their
unpredictable nature and dependence on weather and climatic
changes. Fortunately, the problems can be partially overcome by
integrating the two resources in a proper combination to form a
hybrid system, using the strengths of one source to overcome the
weakness of the other. However, the complexity, brought about
by using of two different resources together, makes the hybrid sys-
tems more difficult to analysis.
Various optimization techniques, such as graphical construction
method, probabilistic approach and iterative technique, have been
recommended by researchers to guarantee the lowest investment
with full use of the solar system, wind system and battery bank.
A graphical construction technique for figuring the optimum
combination of battery and PV array in a hybrid solar–wind system
has been presented by Borowy and Salameh [1]. This kind of graph-
ical methods, normally only two parameters (PV and wind turbine,
PV and battery, or wind turbine and battery) were included in the
optimization process.
Tina et al. [2] presented a probabilistic approach based on the
convolution technique to incorporate the fluctuating nature of
* Corresponding author. Tel.: +852 2766 5863; fax: +852 2774 6164.
E-mail address: behxyang@polyu.edu.hk (H. Yang).
the resources and the load, thus eliminating the need for time-ser-
ies data, to assess the long-term performance of a hybrid system.
Yang et al. [3,4] have proposed an iterative optimization tech-
nique following the loss of power supply probability (LPSP) model
for a hybrid solar–wind system. From this iterative procedure, sev-
eral possible combinations of solar–wind generation capacities
were obtained. The total annual cost for each configuration is then
calculated and the combination with the lowest cost is selected to
represent the optimal mixture.
In the optimization methods described above, the minimization
of system cost are normally implemented by employing probabil-
ity programming techniques or by linearly changing the values of
corresponding decision variables, resulting in suboptimal solutions
and sometimes increased computational effort requirements. Also,
these sizing methodologies normally do not take into account
some system design characteristics, such as PV modules slope an-
gle and wind turbine installation height, which also highly affect
the resulting energy production and system installation costs.
In this paper, one optimal sizing model for stand-alone hybrid
solar–wind system employing battery banks is developed based
on the concepts of loss of power supply probability (LPSP) and
annualized cost of system (ACS). The decision variables included
in the optimization process are the PV module number, PV module
slope angle, wind turbine number, and also the wind turbine
installation height as well as the battery number. The configuration
that meets the system reliability requirements with minimum cost
can be obtained by an optimization technique – the genetic algo-

0306-2619/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.apenergy.2008.03.008
164 H. Yang et al. / Applied Energy 86 (2009) 163–169

Nomenclature

ACS annualized cost of system, US$ g charging and discharging efficiency

CRF the capital recovery factor f wind speed power law coefficient
f annual inflation rate r hourly self-discharge rate
G solar radiation, W/m2
H height, m Subscripts
I current, A acap annualized capital cost
i and i0 the real and nominal interest rate amin annualized maintenance cost
k Boltzmann constant, 1.38  1023 J/K arep annualized replacement cost
LPSP loss of power supply probability bat battery
n ideality factor, 1 < n < 2 bh beam radiation on horizontal surface
P power, W bt beam radiation on tilt surface
q magnitude of the electron charge C cut-in
R resistance, ohm cap initial capital cost
SFF the sinking fund factor dt diffuse radiation on tilt surface
SOC battery state of charge o parameters under standard conditions
T temperature, K oc open-circuit
t time, h p parallel
V voltage, V proj project
v velocity, m/s PV photovoltaic
Y lifetime, year R rated
re reflected
Greeks rep replacement
a, b, c constant parameters for PV module s series
a0 absorptivity sc short-circuit
b0 PV module slope angle, radians tt total radiation on tilt surface
e emissivity WT wind turbine

rithm (GA), which is generally robust in finding global optimal
solutions, particularly in multi-modal and multi-objective optimi-
zation problems.
2. Model of the hybrid system components
A hybrid solar–wind power generation system consists of PV ar-
ray, wind turbine, battery bank, inverter, controller, and other
accessory devices and cables. In order to predict the hybrid system
performance, individual components need to be modeled first.
the PV array output. If a matrix of Ns  Np PV modules is consid-
ered, the maximum power output of the PV system can be calcu-
lated by
PPV ¼ Np  Ns  Pmodule  gMPPT  goth ð2Þ
where gMPPT is efficiency of the maximum power point tracking,
although it is variable according to different working conditions, a
constant value of 95% is assumed to simplify the calculations. goth
is the factor representing the other losses such as the loss caused
by cable resistance and accumulative dust, etc.

2.1. PV array performance model 2.2. Wind turbine performance model

PV module performance is highly influenced by weather condi- Choosing a suitable model is very important for the wind tur-
tions, especially solar radiation and PV module temperature. A bine power simulations. For a typical wind turbine, the power out-
simplified simulation model proposed by Zhou et al. [5] is used put characteristic can be assumed in such a way that it starts
to estimate the actual performance of PV modules under varying generating at the cut-in wind speed vC, the power output increases
operating conditions. Five parameters (a, b, c, Rs and n) are intro- linearly as the wind speed increases from vC to the rated wind
duced to take account for all the non-linear effect of the environ- speed vR, The rated power PR is produced when the wind speed var-
mental factors on PV module performance. Using the definition ies from vR to the cut-out wind speed vF at which the wind turbine
of fill factor, the maximum power output delivered by the PV mod- will be shut down for safety considerations. Then the wind turbine
ule Pmodule can be written as power output can be simulated by
  8 vvC
V oc
 ln V oc
þ 0:72    a < PR  vR vC    ðvC 6 v 6 vR Þ
>
nKT=q nKT=q Rs G
Pmodule ¼  1  Isco PWT ¼ PR    ðvR 6 v 6 vF Þ ð3Þ
1 þ nKT=q V oc V oc =Isc Go >
:
 c 0    ðv < vC or v > vF Þ
V oco To
  ð1Þ For small-scale wind turbines, the cut-in wind speed is relatively
1 þ b ln GGo T
smaller, and wind turbines can operate easily even when wind
where a is the factor responsible for all the non-linear effects that speed is not very high.
the photocurrent depends on; b is a PV module technology spe-
cific-related dimensionless coefficient [6]; and c is the factor con- 2.3. Battery performance model
sidering all the non-linear temperature–voltage effects.
PV modules represent the fundamental power conversion unit Lead-acid batteries used in hybrid solar–wind systems operate
of a PV system. It is mandatory to connect PV modules in series under very specific conditions, which is often very difficult to pre-
and in parallel in order to scale-up the voltage and current to tailor dict when energy will be extracted from or supplied to the battery.
 H. Yang et al. / Applied Energy 86 (2009) 163–169
Usually, a lead-acid battery is mainly characterized by two indexes,
i.e. the state of charge (SOC) and the floating charge voltage.
2.3.1. Battery state-of-charge (SOC)
For a perfect knowledge of the real SOC of a battery, it is neces-
sary to know the initial SOC, the charge or discharge time and the
current. However, most storage systems are not ideal, losses occur
during charging and discharging and also during storing periods.
Taking these factors into account, the SOC of the battery at time
t + 1 can be simply calculated by
 
r  Dt Ibat ðtÞ  Dt  gbat
SOCðt þ 1Þ ¼ SOCðtÞ  1  þ ð4Þ
24 C 0bat
where r is the self-discharge rate which depends on the accumu-
lated charge and the battery state of health [7]; C 0bat is the nominal
capacity of the battery, Ah. gbat is the battery charging and discharg-
ing efficiency. It is difficult to measure separate charging and dis-
charging efficiencies, so manufacturers usually specify a round-
trip efficiency. In this paper, the battery charge efficiency is set
equal to the round-trip efficiency, and the discharge efficiency is
equal to 1.
2.3.2. Battery floating charge voltage
The battery floating charge voltage response is modeled by the
equation-fit method, which expresses the battery floating charge
voltage by a polynomial in term of battery SOC and battery current
V bat ¼ A  ðSOCÞ3 þ B  ðSOCÞ2 þ C  SOC þ D
where A, B, C and D are functions of the battery current I, and can be
calculated by
0 1 0 1
A a1 a2 a3 0 2 1
B B C Bb I
B C B 1 b2 b3 C
CB C
B C¼B C@ I A
@ C A @ c1 c2 c3 A
1
D d1 d2 d3
By fitting the equations to battery performance data, the parame-
ters a1, a2, a3, . . . , d1, d2, d3 can be calculated using least squares fit-
ting method by fitting the equations to battery performance data
(usually available from the manufacturer).
3. Reliability model based on LPSP concept
Loss of power supply probability (LPSP) is defined as the prob-
ability that an insufficient power supply results when the hybrid
system (PV array, wind turbine and battery storage) is unable to
satisfy the load demand [3].
Two approaches exist for the application of LPSP in designing
stand-alone hybrid system. The first one is based on chronologi-
cal simulation. This approach is computationally burdensome
and requires the availability of data spanning a certain period
of time. The second approach uses probabilistic techniques to
incorporate the fluctuating nature of the resource and the load,
thus eliminating the need for time-series data. Considering the
energy accumulation effect of the battery, to present the system
working conditions more precisely, the chronological method is
employed. The objective function, LPSP, from time 0 to T can
be described by
P
T genetic algorithm (GA), which can dynamically searches for the
Power  failure  time ðPsupplied ðtÞ < Pneeded ðtÞÞ
LPSP ¼ t¼0
N
where N is the number of time intervals, the number of hours in this
study with hourly weather data input. And the power supplied from
the hybrid system can be expressed by
165
Psupplied ðtÞ ¼ PPV þ PWT þ C  V bat
 
0:2C 0bat C 0bat  ðSOCðtÞ  SOCmin Þ
 Min Ibat;max ¼ ; ð8Þ
Dt Dt
where C is a constant, 0 for battery charging process and 1 for bat-
tery discharging process.
4. Economic model based on ACS concept
The economical approach, according to the concept of annual-
ized cost of system (ACS), is developed to be the best benchmark
of system cost analysis. According to the studied hybrid solar–wind
system, the ACS is composed of the annualized capital cost Cacap,
the annualized replacement cost Carep and the annualized mainte-
nance cost Camain. Then the ACS can be expressed accordingly by
(four main parts are considered: PV array, wind turbine, battery
and wind turbine tower):
ACS ¼ C acap ðPV þ Wind þ Bat þ TowerÞ þ C arep ðBatÞ
þ C amain ðPV þ Wind þ Bat þ TowerÞ ð9Þ
4.1. Annualized capital cost
The annualized capital cost of each component (PV array, wind
turbine, battery and wind turbine tower) is
i  ð1 þ iÞY proj
ð5Þ C acap ¼ C cap  CRFði; Y proj Þ ¼ C cap  ð10Þ
ð1 þ iÞY proj  1
where Ccap is the initial capital cost of each component, US$; Yproj is
the component lifetime, year; CRF is the capital recovery factor, a
ratio to calculate the present value of an annuity (a series of equal
annual cash flows). The annual real interest rate i is related to the
ð6Þ
nominal interest rate i0 and the annual inflation rate f.
4.2. Annualized replacement cost
The annualized replacement cost is the annualized value of all
the replacement costs occurring throughout the lifetime of the
project:
i
C arep ¼ C rep  SFFði; Y rep Þ ¼ C rep  ð11Þ
ð1 þ iÞY rep  1
where Crep is the replacement cost of the component (battery), US$;
Yrep is the component (battery) lifetime, year; SFF is the sinking
fund factor, a ratio to calculate the future value of a series of equal
annual cash flows.
4.3. Annualized maintenance cost
The system maintenance cost is deemed to be constant every
year. The configuration with the lowest ACS is taken as the optimal
one from the configurations that can guarantee the required reli-
ability of power supply.
5. System optimization model with genetic algorithm
The minimization of ACS function is implemented employing
optimal configuration. The flow chart of the optimization process
ð7Þ is illustrated in Fig. 1.
The decision variables included in the optimization process are
the PV module number NPV, wind turbine number NWT, battery
number Nbat, PV module slope angle b0 and wind turbine installa-
tion height HWT.
166 H. Yang et al. / Applied Energy 86 (2009) 163–169
Fig. 1. Flow chart of the optimal process by GA.
The hourly data used in the model are the solar radiation on
horizontal surface, ambient air temperature, wind speed and load
power consumption on an annual basis. The year 1989 is chosen
as the Example Weather Year in Hong Kong [8] to represent the
climatic conditions for the studied project design (the studied
project is close to Hong Kong) in the following optimization
process.
The PV array power output is calculated according to the PV
system model by using the specifications of the PV module as well
as the solar radiation data. The wind turbine performance calcula-
tions need to take into account the effects of wind turbine installa-
tion height. The battery bank, with total nominal capacity C 0bat (Ah),
is permitted to discharge up to a limit defined by the maximum
depth of discharge DOD, which is specified by the system designer
at the beginning of the optimal sizing process.
The system configuration will then be optimized by employ-
ing genetic algorithm, which dynamically searches for the opti-
mal configuration to minimize the annualized cost of system
(ACS). For each system configuration, the system’s LPSP will be
examined for whether the load requirement (LPSP target) can
be satisfied. Then, the lower cost load requirement satisfied con-
figurations, will be subject to the following crossover and muta-
tion operations of the GA in order to produce the next generation
population until a pre-specified number of generations has been
reached or when a criterion that determines the convergence is
satisfied.
So, for the desired LPSP value, the optimal configuration can be
identified both technically and economically.
6. Result and discussion
6.1. The hybrid solar–wind project description
The proposed method has been applied to design one hybrid
project, which is built to supply power for a telecommunication re-
lay station on a remote island along south-east coast of China (see
Fig. 2).
The electric use for the normal operation of the telecommunica-
tion station includes 1300 W GSM base station RBS2206 consump-
tion (24 V AC) and 200 W for microwave communication (24 V
DC). According to the project requirement and technical consider-
ations, a continuous 1500 W energy consumption is assumed as
the demand load. The technical characteristics of the PV module
and the wind turbine used in the studied project are listed in Ta-
bles 1 and 2.
The GFM-1000 lead-acid batteries are employed in the project.
They are specially designed for deep cyclic operation in consumer
applications like the hybrid solar–wind energy systems. Each set of
batteries has twelve 2 V/1000 Ah cells which are connected in ser-
ies to give a nominal output voltage of 24 V.
The initial capital cost, replacement cost, maintenance cost and
lifetime of each component are in Table 3.
6.2. Optimal design result and discussion
The optimal sizing results for the desired loss of power supply
probability (LPSP) of 2% are shown in Table 4. The overall optimal
 H. Yang et al. / Applied Energy 86 (2009) 163–169 167

20 m above the ground, which is lower than the optimal sizing re-
sults. To makeup the power deficiency caused by reduced PV mod-
ules, another two 6 kW wind turbines are added. The revised
configuration can also satisfy the reliability requirements with a
LPSP of 1.98, but the annualized cost of system is a higher than
the optimal one as shown in Table 4.
It is noteworthy that, the optimized battery bank has a total
nominal capacity of 5000 Ah (24 V). Since the battery energy
storage capacity is more than three times higher than the daily
energy output of the system, the hourly or even the daily irreg-
ular power supply from the hybrid system can be easily
smoothed away.

6.3. Project performance analysis with field measured data

With the time-series field data of the studied project recorded

from January 2006 to December 2006, the monthly energy contri-
bution of each component (PV module and wind turbine), the bat-
Fig. 2. The hybrid solar–wind power generation project. tery working states and the energy balance are investigated
respectively.

Table 1 6.3.1. Monthly energy contribution of each component

Specifications of the PV module Monthly analysis of the hourly measured field data shows that
Voc (V) Isc (A) Vmax (V) Imax (A) Pmax (W) the energy contributions from photovoltaic and wind generators
vary greatly from one month to the next.
21 6.5 17 5.73 100
Fig. 3 shows the monthly-mean solar and wind energy output
variations of the studied project for the concerned year. Due to
the seasonal wind change, which is consistent northeasterly high
solution is Item 7, resulting in a minimum annualized cost of speed monsoon winds in winter and relatively low wind speed in
9708 US$. summer, more wind energy is generated in months of January
Unfortunately, due to the restriction of limited roof area (see (935 W) and February (986 W); May and August have much less
Fig. 2), instead of using 113 PV modules recommended by the opti- wind power as indicated by smaller wind energy output of 561
mization model, only 78 PV modules were installed. Furthermore, and 600 W respectively. Solar energy has a complementary varia-
due to local terrain constraints and to prevent destructions caused tion with the higher solar energy output from April to October
by severe typhoon in summer season, the wind turbine is installed and lower energy output during other months.

Table 2
Specifications of the wind turbine

Rated power (kW) Cut-in speed VC (m/s) Rated speed VR (m/s) Cut-off speed VF (m/s) Tower Hhigh (m) Tower Hlow (m)
6 2.5 10 None 50 10

Table 3
The initial cost and lifetime of the system components

Initial capital cost Replacement cost Maintenance cost (/year) Lifetime (year) Interest rate i0 (%) Inflation rate f (%)
PV array 6500 US$/kW Null 65 US$/kW 25 3.75 1.5
Wind turbine 3500 US$/kW Null 95 US$/kW 25
Battery 1500 US$/kAh 1500 US$/kAh 50 US$/kAh 5
Tower 250 US$/m Null 6.5 US$/m 25

Table 4
Optimal sizing results for the hybrid solar–wind system
168 H. Yang et al. / Applied Energy 86 (2009) 163–169

Wind Turbine
Dump Load

PV array

1.76 kW
0.73 kW 1.03 kW

Battery 0.94+0.65 kW

0.82 kW
Load

0.65 kW

Fig. 5. Annually average energy balance of the pilot project.

kW out
nbat ¼  100% ð12Þ
Fig. 3. Monthly energy contribution of solar and wind energy. kW in
Calculated from the one-year field data of the hybrid solar–wind
power generation project, the annually average energy balance be-
6.3.2. Probability distribution of the battery state-of-charge
tween the power supplies, the load and the battery are illustrated in
Statistically derived from the one-year battery SOC data of the
Fig. 5.
project, the columns in Fig. 4 indicate the probability, or the frac-
According to Fig. 5, an annually average power of 1.76 kW is
tion of time, when the battery SOC is within the interval given
generated by the PV array and the wind turbine, 0.82 kW of it
by the width of the columns (here 0.1 is used).
was allotted to charge the battery for power shortage period back-
Normally used lead-acid batteries for renewable energy appli-
up, the rest (0.94 kW) was directly supplied to the load. With
cations are characterized to have a limited range of depth of dis-
0.82 kW input to the battery, only 0.65 kW was available from
charge (DOD), being currently limited to a maximum of 80%. If
the battery for the losses occurred during charging, discharging
this value is exceeded, the battery suffers from over-discharge,
and storing periods. So, the battery overall efficiency can be simply
and prolonged over-discharge may result in permanent damage
calculated by Eq. (12) to be about 79%, which is basically consistent
of the battery. Kattakayam and Srinivasan [9] recommended
with the values claimed by others [11].
through trial and error and prolonged experimentations that
0.5 < SOC < 0.8 would be ideal working range for the lead-acid bat-
7. Conclusions
teries. Referring to Fig. 4, the batteries in this case are well con-
trolled and in good working states with nearly 90% opportunities
One optimum match design sizing method for hybrid solar–
for its SOC remaining above 0.5, and the system loss of power sup-
wind system is developed in this paper based on genetic algorithm
ply probability (LPSP) was quite well controlled to be less than 2%
(GA), which has the ability to attain the global optimum with rel-
as required, all these statistical numbers showed the rationality of
ative computational simplicity. This model can be used to calculate
the design and the validity of the load control strategies of the pro-
the system optimum configuration which can achieve the desired
ject. As a result, a long cycle life of the battery can be ensured.
loss of power supply probability (LPSP) with minimum annualized
cost of system. The decision variables included in the optimization
6.3.3. Energy balance between each component
process are the PV module number, PV module slope angle, wind
Most storage systems are not ideal, losses occurs in charging
turbine number, wind turbine installation height and battery
and discharging cycles and also during storing periods. The total
number.
energy efficiency nbat of the battery is the ratio of the energy ob-
The proposed method has been applied to design a hybrid so-
tained during discharging process to that required to restore it to
lar–wind system to supply power for a telecommunication relay
its original condition, and can be expressed by Jossen et al. [10]:
station on a remote island along south-east coast of China. The
algorithm is based upon using the weather data of year 1989 in
Hong Kong as the Example Weather Year for both wind speed
and solar radiation for the site under consideration.
With the one year hourly measured field data of the studied
project, the monthly energy contribution of each component (PV
module and wind turbine), the battery working states and the en-
ergy balance are investigated. The energy contributions from pho-
tovoltaic and wind generators vary greatly from one month to the
next, but good complementary characteristic between solar energy
and wind energy are found. The battery has also been demon-
strated to be in good working states with nearly 90% opportunities
for the battery SOC to remain higher than 0.5, and the over-dis-
charge situations seldom occurred throughout the studied year.

References

[1] Borowy BS, Salameh ZM. Methodology for optimally sizing the combination of
a battery bank and PV array in a wind/PV hybrid system. IEEE Trans Energy
Fig. 4. Battery SOC distribution for the revised configuration. Conver 1996;11(2):367–73.
 H. Yang et al. / Applied Energy 86 (2009) 163–169
[2] Tina G, Gagliano S, Raiti S. Hybrid solar/wind power system probabilistic
modeling for long-term performance assessment. Sol Energy 2006;80:578–88.
[3] Yang HX, Jurnett B, Lu L. Weather data and probability analysis of hybrid
photovoltaic–wind power generation systems in Hongkong. Renew Energy
2003;28:1813–24.
[4] Yang HX, Lu L, Zhou W. A novel optimization sizing model for hybrid solar–
wind power generation system. Sol Energy 2007;81(1):76–84.
[5] Zhou W, Yang HX, Fang ZH. A novel model for photovoltaic array performance
prediction. Appl Energy 2007;84:1187–98.
[6] Van Dyk EE et al. Long-term monitoring of photovoltaic devices. Renew Energy
2002;22:183–97.
169
[7] Guasch D, Silvestre S. Dynamic battery model for photovoltaic applications.
Prog Photovoltaics: Res Appl 2003;11:193–206.
[8] Yang HX, Lu L. Study on typical meteorological years and their effect on
building energy and renewable energy simulations. ASHRAE Trans
2004;110(2):424–31.
[9] Kattakayam TA, Srinivasan K. Lead acid batteries in solar refrigeration systems.
Renew Energy 2004;29(8):1243–50.
[10] Jossen A, Garche J, Sauer DU. Operation conditions of batteries in PV
applications. Sol Energy 2004;76(6):759–69.
[11] Mahmoud MM, Ibrik IH. Field experience on solar electric power systems and
their potential in Palestine. Renew Sust Energy Rev 2003;7(6):531–43.