www.ietdl.

org
Published in IET Renewable Power Generation
Received on 27th July 2011
Revised on 19th April 2012
doi: 10.1049/iet-rpg.2011.0204

ISSN 1752-1416

Eco-design optimisation of an autonomous hybrid

wind – photovoltaic system with battery storage
D. Abbes1 A. Martinez1 G. Champenois2
1
EIGSI, 26 rue de vaux de Foletier, 17041 La Rochelle, Cedex, France
2
LIAS - University of Poitiers, Bât B25, 2 rue Pierre Brousse, BP 633, 86022 Poitiers, Cedex, France
E-mail: dhaker.abbes@hei.fr

Abstract: In this study, a new approach to design an autonomous hybrid wind – photovoltaic (PV)-batteries system is presented in
order to assist the designers to take into consideration both the economic and ecological aspects. Primary embodied energy (EE)
has been introduced as a new criterion for hybrid systems, designing with the objective to minimise loss of power supply
probability (LPSP). For a location, meteorological and load data have been collected and assessed. Then, modelling and
primary energy analysis have been achieved for all components of the hybrid system. Finally, an optimal configuration has
been carried out using a dynamic model and applying two different algorithms with single and multi-objective optimisation.
The methodology has been performed successfully for the sizing of a wind – PV-batteries system to supply at least 95% of
the yearly total electric demand of a residential house. The simulation results show that the optimal configuration meets the
desired system reliability requirements (LPSP ,5%) with the lowest EE. A life cycle cost analysis has been established at the
end of the study to demonstrate the importance of economic considerations.

SOCmin minimum allowable battery storage capacity, %

Nomenclature
Ta ambient temperature, 8C
Apv surface area of PV panels, m2 Tc photovoltaic cell temperature, 8C
Apv_max maximum surface area of PV panels, m2 Ubus nominal DC bus voltage, V
Apv_min minimum surface area of PV panels, m2 V wind speed, m/s
Awt wind turbine swept area, m2 Vbat nominal voltage of each individual battery, V
Awt_max maximum wind turbine swept area, m2 Z elevation, m
Awt_min minimum wind turbine swept area, m2 B generator efficiency temperature, W/8C
Cn nominal capacity of the battery bank, Ah hacdc AC/DC converter efficiency
Cbat nominal capacity of each battery, Ah hdcdc DC/DC converter efficiency
Cn_max maximum nominal capacity of the battery bank, hg wind turbine generator efficiency
Ah hgb wind turbine gearbox efficiency
Cn_min minimum nominal capacity of the battery bank, hinv DC/AC inverter efficiency
Ah
hpc PV module power conditioning efficiency
Cp wind turbine efficiency
hpv power conversion efficiency of a PV module
Ir solar radiance, W/m2
hr reference PV module efficiency
LPSP_max maximum allowable LPSP, %
ht wind turbine overall efficiency factor
Nbat total number of installed batteries
hwr wires’ losses factor
Nbatp number of batteries strings connected in parallel
r air density, kg/m3
Nbats number of batteries connected in series in every
string
NOCT normal operating photovoltaic cell temperature,
8C 1 Introduction
Ppv output electric power from the PV generator, W
Energy from renewable sources is being considered as a
Pwg electrical power output of a wind generator, W viable alternative to fossil fuels. Among them, wind and
SOCmax maximum allowable battery storage capacity, % solar energies have made a fast and significant breakthrough

358 IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358 –371
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-rpg.2011.0204
 www.ietdl.org
in the past 10 years. In addition, they can be consumed To overcome this lack, we suggest a relevant approach that
locally, hence reducing both impacts from high-voltage consists in minimising both systems embodied energy [(EE)
transmission lines through rural and urban landscapes and is the energy necessary for components manufacturing
power losses. However, neither a solar nor a wind energy process] and LPSP [the LPSP defines the likelihood
stand-alone system can fully satisfy load consumption (probability) that a system demand will exceed the
because of seasonal and periodic climatic variations. generating capacity during a given period]. In addition, to
Therefore it is more reliable and efficient to install a hybrid simplify the procedure and make it as affordable as
energy system with storage because of renewable energy possible, simple but accurate models have been performed.
sources intermittent character. Photovoltaic panels are often Moreover, photovoltaic modules’ installed area (m2), wind
combined with wind turbines or diesel generators and turbine swept area (m2) and battery bank capacity (Ah)
batteries. Evidently, an optimal configuration must be found have been considered as decision variables for design
to build the final system, taking particularly into optimisation. This concept is often used for solar PV
consideration consumer energy demand and energy generation systems but less frequent for hybrid wind – PV-
resources of the site. For this purpose, sizing and batteries systems.
optimising stand-alone hybrid renewable energy systems First, meteorological and load data are collected and
has been carried out by a few researchers and studies [1]. assessed for a residential house in order to determine load
For instance, [2, 3] propose some methodologies for the consumption. Then, the hybrid wind – PV system with
design and evaluation of small hybrid power systems. Also, batteries is modelled. This phase includes: mathematical
Borowy and Salameh [4], explicitly describe a method to models and components embodying primary energy
optimise the size of a photovoltaic (PV) – wind-batteries analysis. Optimisation programming is executed in the final
system: the desired unmet load is achieved by modifying step. It is carried out by using a dynamic model of the
the number of photovoltaic panels and batteries. Wind global system under Matlab/Simulink and optimal
turbine, panels’ type and battery technology are fixed. As configuration is obtained by applying two different
there is more than one technically feasible solution, they algorithms for non-linear mono- and multi-objective
select the less expensive one. optimisation.
A methodology has been developed by Chedid and Thus, the paper is organised as follows. In Section 2,
Rahman [5] that determines the optimal design of a hybrid renewable energy sources (wind and solar) data and load
wind – solar power system for either autonomous or grid- profiles are assessed. Section 3 describes the hybrid system
connected applications. The proposed analysis employs models. Next, respectively, in Sections 4 and 5,
simple linear programming techniques to minimise the optimisation objectives (design and evaluation criteria) and
average production cost of electricity, and takes the computational method are explained. Finally, results and
environmental factors into consideration both in design and conclusions are presented in Section 6.
operation phases. Nevertheless, components manufacture is
not discussed in the paper. Only environmental impacts
(such as CO2 , NOx and SO2 emissions because of diesel 2 Data collection: renewable energy sources
generator) are assessed. and load consumption profile
An iterative optimisation method for PV – wind-batteries
The first step consists in collecting data. In this section,
systems has been presented in [6]. It has been used to
renewable energy sources have been assessed for a long
determine the optimum generation capacity and storage
period (8 years) and load consumption has been acquired
needed for a stand-alone, wind, PV and hybrid wind –PV
with a power analyser. The data must be accurately
system for an experimental site in a remote area in Montana
acquired to reach a suitable sizing of the hybrid system.
with a typical residential load. Generation and storage units
for each system are properly sized in order to meet the
annual load and minimise the total annual cost to the 2.1 Data sources
customer.
Yang et al. [7] present a method for a hybrid PV – wind- Wind speed and solar radiance data with a small sampling
batteries system that minimises the levelised cost of energy interval period are generally not available for public users.
(LCE). The optimisation is made by changing many To overcome this barrier, data were acquired from the
parameters: components’ combinations, orientation of PV National Wind Technology Centre – Colorado (latitude:
modules, rated power, tower height of the wind turbine and 398 north, longitude: 1058 west, elevation: 1855 m) web
the battery bank capacity. site [10]. Data were available with a high accuracy and the
Diaf et al. [8] have made the optimisation of a hybrid PV – periodicity can be defined by the user. The choice of
wind-battery system in different locations in Corsica in sampling period has been already discussed before [11]; it
France. Their procedure takes into consideration loss of is demonstrated that it is not necessary to choose a very low
power supply probability (LPSP) and LCE. sampling period to estimate correctly produced energy. Half
In 2007, Shi et al. [9] proposed a robust design method for hourly data are sufficient to make a suitable energy
an autonomous PV – wind hybrid power system that could estimation.
achieve an optimum system configuration insensitive to Renewable energy potentials are given for each year in
design variable variations. It was based on a constraint Fig. 1. They are calculated using numerical integration of
multi-objective optimisation problem solved by a multi- the power.
objective genetic algorithm, Non-dominated Sorting Genetic
Algorithm (NSGA-II). 
8760×2
1
To summarise, most of the aforementioned studies mainly Energy(T ) = (Power(i) + Power(i + 1)) × T (1)
focus on economic cost minimisation. A few of them consider i=1
2
the minimisation of primary energy cost tackling reduction of
the ecological impact. where T corresponds to the sampling interval.

IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358– 371 359
doi: 10.1049/iet-rpg.2011.0204 & The Institution of Engineering and Technology 2012
www.ietdl.org

Fig. 1 Renewable energy potentials of the National Wind Technology Centre, Colorado
∗
Calculated using wind data speed between 3.5 and 25 m/s

According to the graph shown in Fig. 1, photovoltaic potential These profiles have been extrapolated for the whole year
is prevailing on wind potential for this location. Therefore data respecting real electrical load seasonal variation as depicted
for the year 2009 with the lowest photovoltaic potential have in Fig. 3. Annual consumption of the residence is then
been employed. The probability that design meets the estimated at 2199 kWh/year by using numerical integration
optimisation criteria for all the years, especially in terms of of the power according to (1).
unmet load, has been verified by many simulations.
3 Hybrid system models
2.2 Load profile
Modelling is an essential step before starting any phase of
Real consumption data were acquired for a residential home optimal sizing. A hybrid energy system is implemented as a
of four occupants, except for cooking, heating and hot combination of three power sources: wind turbine,
water production. The residence is occupied 7 days per photovoltaic generator and batteries as shown in Fig. 4. A
week all year long. All the loads are supplied with 230 V methodology for modelling each component is described in
AC. A half-an-hour period has been programmed in order the following paragraphs.
to be in concordance with renewable energy resources
acquisition. Fig. 2 shows a week and weekend day 3.1 Wind turbine model
consumption acquired in the ‘mid-term season’, in order to
neutralise the influence of heating in electrical energy Since most home-sized wind turbines tend to operate towards
consumption. During a weekday, the home is less occupied the same efficiency level, rotor diameter or swept area rather
by its residents. The energy consumption has been than generator size makes far better criteria in order to
registered at 3.787 kWh/day. On a weekend day, the compare various models from different manufacturers. For
electricity demand increases significantly and reaches this reason, swept area (Awt) has been chosen as the
8.32 kWh/day. decision variable for wind turbine sizing. Consequently, the

Fig. 2 Half-hourly load profile for a week and weekend day recorded in June

360 IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358 –371
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-rpg.2011.0204
 www.ietdl.org
As most wind turbines available today are three-bladed
horizontal axis, this technology will be considered with an
overall efficiency factor ht ¼ 35% [13].

3.2 Photovoltaic generator model

Output electric power from the photovoltaic generator is

given by the following equation [14]

Ppv = hpv × Apv × Ir (3)

Fig. 3 Monthly load energy changes recorded in the residence where hpv is the power conversion efficiency of the module
(power output from the system divided by power input
electrical power output of a wind generator is given as follows from the sun);
[12] Apv (m2), is the surface area of PV panels;
Ir (W/m2), is the solar radiance. PV generator efficiency is
given by [15]
1
Pwg = Cp × hgb × hg × × r × Awt × V 3
2 hpv = hr × hpc × [1 − b(Tc − NOCT)] (4)
1
= ht × × Awt × V 3 (2)
2
with hr is the reference module efficiency; it depends on cell
with material. A polycrystalline silicon technology has been used
with 13% of efficiency [16].
hpcis the power conditioning efficiency. hpc is equal to 0.9
353.049 (−0.034(Z/Ta )) with a perfect maximum point tracker [17].
r[kg/m3 ] = Air density = e
b is the generator efficiency temperature coefficient, ranging
Ta
from 0.004 to 0.006 /8C.
and ht = Cp × hgb × hg r Tc is the cell temperature (8C). For a PV module of
polycrystalline silicon solar cells, it can be estimated from
the ambient temperature Ta (8C) and the solar irradiation Ir
All the parameters are described below:
as follows [18]
Z (m) is the elevation and Ta (8C), is the temperature;
V (m/s): is the wind speed; Tc = 30 + 0.0175 × (Ir − 300) + 1.14 × (Ta − 25) (5)
Awt (m2): is the wind turbine swept area;
Cp is the turbine efficiency;
hgb is the gearbox efficiency; The normal operating cell temperature (NOCT) (8C) can be
hg is the generator efficiency. calculated when the cells operate under standard operating
conditions: irradiance of 800 W/m2, 208C ambient
Equation (2) supposes that the power rating and efficiency temperature, average wind speed of 1 m/s, module in an
of the wind turbine have a linear relation with swept area. This electrically open-circuit state, wind oriented parallel to
linearisation is acceptable for small wind turbines with close array’s plane and all sides of the array fully exposed to wind.
optimal power factors and swept area ranging from 0 to After consultation of several different polycrystalline
25 m2. Maximum power is reached at 15 m/s wind speed silicon manufacturers (such as Evergreen ES-A210 or Trina
and is assumed constant from this threshold. Solar TSM-PA05), a typical value of NOCT equal to 458C

Fig. 4 Schematic diagram of wind–PV hybrid system with battery storage

IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358– 371 361
doi: 10.1049/iet-rpg.2011.0204 & The Institution of Engineering and Technology 2012
www.ietdl.org
and a typical value of b approximated to 0.0045/8C has been SOCmax corresponds to the nominal capacity of the battery
considered. bank, Cn [21]
For sizing the optimisation procedure, the effective area of  
photovoltaic generator (Apv) is defined as the decision Nbat
variable: if Apv is measured in m2, Ppv is numerically equal Cn = × Cbat = Nbatp × Cbat (9)
Nbats
to the peak power rating of the array.

3.3 Battery storage model where Nbat¼ total number of batteries;

Nbats¼ number of batteries connected in series in every string;
Most batteries used are lead-acid type in hybrid systems, for Nbatp¼ number of batteries strings connected in parallel;
many reasons. Discharge and recharge cycles do not Cbat (Ah) ¼ nominal capacity of each battery;
significantly affect energy storage capacity. It also offers SOCmin corresponds to the lower limit that the battery bank
safety margin against a wet acid battery, since acid is should not exceed when discharging. It is determined by
contained in a gel and cannot be easily spilled [19]. A the maximum allowable battery depth of discharge (DOD)
simple but efficient model has been implemented for the as follows
optimisation method. An ideal battery model is depicted in
Fig. 5.
Charging or discharging phases are controlled considering SOCmin = (1 − DOD) × SOCmax (10)
the difference between resources production (Pres) and load
consumption (Pload). In the charging phase, state of charge The maximum allowable DOD depends on battery type and
(SOC) can be calculated as load profile. 80% DOD can be tolerated [22, 23]. Batteries
 are connected in series in order to reach the DC bus
nominal voltage (48 V) or in parallel to obtain the desired
SOC(t) = SOC(t − 1) + Ppv (t) × hdcdc + Pwg (t) storage capacity.
 An intelligent switcher manages batteries connection,
Pload (t) h according to four phases:
× hacdc − × cha × Dt (6)
hwr × hinv Ubus
† When charge state of batteries is below SOCmax (100% Cn)
In the discharging phase, SOC is given by and Pload , Pres , excess of energy [(Pres 2 Pload). DT ] is
stored in batteries. DT corresponds to the period of this
 excess state.
SOC(t) = SOC(t − 1) + Ppv (t) × hdcdc + Pwg (t)) † When it is above SOCmin (20% Cn) and Pload . Pres ,
Pload  1 energy previously stored is used to support the lack of
× hacdc − × × Dt (7) energy [(Pload 2 Pres).DT ] (battery discharge).
hwr × hinv hdis × Ubus † When the SOC is equal to SOCmax (100% Cn) and
Pload , Pres , energy is lost [(Pres 2 Pload).DT ] during the
Pload (t) is the power consumed by the load at a given time. Dt corresponding period DT.
is the simulation time step (Dt ¼ half-an-hour). hdcdc , hacdc † When the SOC is equal to SOCmin (20% Cn) and
and hinv are, respectively, DC/DC, AC/DC and DC/AC Pload . Pres , an unmet load occurs. In this case, Pload must
converter efficiencies (hdcdc ¼ hacdc ¼ hinv ¼ 0.95 have be equal to Pres by load shedding.
been assumed in the literature [3, 17]). hcha and hdis are
battery efficiency during charging and discharging phases Finally, the design variable that needs optimisation is the
(hcha ¼ 0.85, hdis ¼ 1 [20]). hwr is introduced to consider total storage capacity of battery bank Cn (Ah): it is
wire losses (hwr ¼ 0.98 [17]) and Ubus is the nominal DC proportional to the number of batteries in parallel Nbatp . The
bus voltage (48 V). number of batteries to be connected in series Nbats is not
For longevity consideration, SOC is delimited by the subject to optimisation but is a straightforward calculation.
following constraints
4 Objectives: design and evaluation criteria
SOCmin ≤ SOC(t) ≤ SOCmax (8)
In our study, the main objective is to satisfy the energy
requirements with lowest environmental impact in terms of
EE. We are focusing on finding the best combination
between multiple sources for a hybrid system, that is,
photovoltaic array area (Apv), wind turbine swept area (Awt)
and battery bank nominal storage capacity (Cn) according to
the ecological criterion. Economic aspects are also
important and will be considered a posteriori in order to
evaluate the final configuration with all components. Hence,
system reliability model, primary energy analyses and cost
considerations are considered in the following sections.

4.1 Reliability requirements: minimising LPSP

Owing to the intermittence of renewable energy sources,

power system reliability is considered as an important step
Fig. 5 Ideal lead-acid battery model in the hybrid system design process. Throughout this

362 IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358 –371
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-rpg.2011.0204
 www.ietdl.org

Fig. 6 Wind– PV-Battery hybrid system Simulink mode

investigation, reliability of the system has been expressed in account each manufacturing step of the considered element.
terms of LPSP [24 – 26]. It represents the ratio of all energy It is not uncommon to find a great disparity in the literature
deficits to the total load demand during the Dt considered regarding these costs. Thus, in [28], EE corresponding to
period. It can be defined as [20] the technology polycrystalline silicium, has been calculated
to a value of 2300 MJ/m2, or 54% of the value proposed
T in [27]. This may be because of the changing technologies
t=1 DE(t, Apv , Awt , Cn ) × Dt
LPSP(Dt, Apv , Awt , Cn ) = T of production wafers and manufacturing processes.
t=1 Pload (t) × Dt Similarly, in [29, 30], by considering the case where the
(11) reference PV panels do not come from recycled materials,
the EE is estimated at about 4600 MJ/m2, which is similar
Deficit energy (DE) (t, Apv , Awt , Cn) corresponds to energy to the value proposed by [27]. We must therefore keep a
deficit at t hour. It is considered when the total energy critical view on the results, knowing that our work is
available within a time interval [(t 2 1), t] and methodological and can be applied to the desired data.
energy stored in batteries at the beginning of this interval For a lead-acid battery replaced every five years, it has been
are insufficient to satisfy the load demand during that time evaluated at 50 MJ/Ah (for 25 years). This value assumes an
period. One relevant objective of the optimisation procedure optimistic 90% recycling rate for scrap batteries. At 0%
is to find combinations between elements that will give an recycling, the energy requirement of the battery increases to
LPSP lower than the maximum allowed by the user. 86 MJ/Ah [31].
For wind turbines, data are not available in the literature,
4.2 Environmental impact optimisation criterion: particularly for small ones; hence we have conducted our
minimising system EE own analysis. Table 2 lists the assumptions and data used
for small wind turbines EE assessment. Information is
In order to reduce environmental impact, EE (energy required drawn from a few references: a fact sheet prepared for the
by all the activities associated with a production process, Office of Industrial Technologies, US Department of
expressed in MJ or kWh) must be minimised. For Energy by Princeton Energy Resources International, LLC
photovoltaic panels, batteries and additional components [33], a study conducted by Fleck and Huot [34], and an
[from ‘balance-of-system (BOS)], data and assumptions are example of 2-MW land-based turbine life cycle assessment
based on the Markvart and Castañer book [27]. They are presented in the book ‘Materials and the Environment: Eco-
summarised in Table 1. EE of 1 m2 solar panels is Informed Material Choice’ [35]. Some approximations are
estimated at 5300 MJ/m2 (for 25 years). This value includes based on a scaling data according to weight and power.
BOS components. We also have to consider that energy Finally, the amount of net energy in different small wind
cost data on life cycle has to be handled with great caution. turbines has been determined. Calculations are based on the
Indeed, they often result from complex analysis taking into intrinsic energy of each material [36] and energy necessary

IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358– 371 363
doi: 10.1049/iet-rpg.2011.0204 & The Institution of Engineering and Technology 2012
www.ietdl.org
Table 1 Assumptions and data used for energy requirements assessment for a typical installed multicrystalline silicon module and
batteries

Assumptions

1 Data concerns a typical multicrystalline silicon PV module with aluminium frame

2 We assume an optimistic 90% recycling rate for scrap batteries [31]
3 BOS components like cables and charge controllers contribute relatively little to the energy requirement of a solar home system (, 10%)
4 We assume an annual efficiency improvement of 1% for materials processing and manufacturing [32]

Data
PV module
Process Energy requirements (MJ/m2 module) Energy requirements, kWha/m2
silicon winning and purification 2200 611.6
silicon wafer production 1000 278
cell /module processing 300 83.4
module encapsulation materials 200 55.6
overhead operations and equipment manufacture 500 139
total module without frame 4200 1167.6
module frame (aluminium) 400 111.2
total module (framed) 4600 1278.8
BOS 700 194.6
total module (installed) 5300 1473.4
Batteries
State Energy requirements (MJ/Ah storage) Energy requirements (kWh/Ah storage)
battery (initial) 11 3.058
battery (after 5 years) 10.46 2.907
battery (after 10 years) 9.95 2.766
battery (after 15 years) 9.46 2.63
battery (after 20 years) 9 2.5
total batteryb 50 14

a
1 KWh ¼ 3.6 MJ or 1 MJ ¼ 0.278 kW
b
Round number

in the manufacturing process. The results are presented in The LCC is then given by
Table 3.
According to these results, 1631 MJ/m2 (swept area) (for LCC = Co + Cinst + PWmaint + PWreplace (14)
25 years) is an appropriate assessment of energy
requirements for small wind turbines. It is also interesting
to note that wind turbines are more energy efficient than with Cinst for the installation cost; PWmaint for the present
photovoltaic panels (1631 MJ/m2 against 5300 MJ/m2). worth of maintenance cost (Cmaint) and is calculated using
Thereafter, another objective of the optimisation procedure the cumulative present worth factor as follows
is to minimise the primary energy cost in terms of EE as a
function of PV size (Apv), wind turbine rotor swept area 1 − xn
(Awt) and battery capacity (Cn). PWmaint = (Cmaint ) × x × (15)
1−x
4.3 Cost considerations: life cycle cost as PWreplace is the present worth of the component taking
evaluation criterion into consideration the fact that it has to be periodically
The economic aspect is considered a posteriori to evaluate the replaced.
total life cycle cost (LCC) of the hybrid system. The LCC is In our case, we assume that total wind turbine initial cost is
calculated taking into account the present worth of all composed of about 55% for turbine and tower purchasing,
components. This is calculated as follows [17] 25% for installation and 20% for ‘BOS’ costs [37]. Current
costs only concern yearly maintenance and represent about
2.5% of the whole initial investment cost.
PW = Pr × Co (12)
For photovoltaic modules, we have assumed that the
purchasing price represents 50% of the initial investment
where Co is the component cost and Pr represents the present cost. The remaining 50% is divided between BOS (30%)
worth factor for an item that will be purchased ‘n ’ years later and installation (20%). In addition, the yearly photovoltaic
(in our case 25 years), and is given by panels’ maintenance represents about 2% of the whole
 n initial investment cost.
1+i For batteries, we have chosen deep cycle valve-regulated
Pr = = xn (13)
1+d lead-acid types for cost and reliability reasons. They do not
need maintenance practically and are typically suitable
where i represents the inflation rate (supposed 5%) and d because of superior high-current rate performance.
corresponds to the discount rate (supposed 3%). However, given the often-unpredictable nature of the

364 IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358 –371
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-rpg.2011.0204
 www.ietdl.org
Table 2 Assumptions and data used for energy requirements assessment for small wind turbines

Assumptions

1 Rotor blades are either glass reinforced plastic, wood-epoxy or injection moulded plastic with carbon fibres. For calculations’
standardisation, we have considered moulded fibre glass as we know its EE
2 Owing to lack of data, an assumption was made to include permanent magnets in the category of aluminium because of its high EE

Wind turbine component weight

Turbine
Component % by turbine weight Materials’ repartition
rotor
hub 4% 95% steel
5% aluminium
blades 10% 100% moulded fibre glass
nacelle
gearbox 6% 100% steel
generator 15% 50% magnets
20% steel
30% copper
others
frame, machinery and shell 35% 30% aluminium
12% copper
5% glass reinforced plastic
53% steel
cables, internal supports, electronic components 30% 80% steel
20% copper
Tower
tower 30 kg/m2 98% steel
swept areaa 2 % aluminium
foundations
pile and platform 65 kg/m2 97% concrete
swept areaa 3% steel
Materials’ EE
material EE, MJ/kg EE, kWh/kg
steel 24.4 6.783
aluminium and magnets (2) 155 43.09
copper 48 13.344
moulded fibre glass 28 7.784
glass reinforced plastic 100 27.8
concreteb 0.9 0.25
Required energy for wind turbine manufacturing
manufacturing process 100 MJ/m2 27.8 kWh/m2
swept areaa swept area

a
Approximation based on scaling data according to weight and power
b
Concrete type RC30 adequate for foundations (25% cement replacement fly ash)

sources, it can be very difficult to estimate their life span. In Subject to restriction in terms of demand satisfaction
our case, we have supposed a replacement every 5 years.
LPSP(Dt, Apv , Awt , Cn ) ≤ LPSP max (17)
5 Optimisation procedure
Constraints: feasibility for an autonomous residence
5.1 Optimisation problem formulation

The hybrid system optimisation problem can be perceived as Apv min ≤ Apv ≤ Apv max (18)
a typical mono-objective minimisation problem, because of,
minimum primary EE cost. Formulation (I) presents the Awt ≤ Awt ≤ Awt (19)
min max
mathematical programme of this case.
Formulation (I):
Object function (Optimum): Cn min ≤ Cn ≤ Cn max (20)
Minimum  Hybrid system primary
EE cost: where Apv , Awt and Cn are the decision variables. LPSP_max
is the maximum LPSP allowed by the designer for the
EE(Apv , Awt , Cn ) = 5300 × Apv + 1631 × Awt + 50 considered period Dt (Dt: one year for example). Apv_min ,
Apv_max , Awt_min , Awt_max , Cn_min and Cn_max are lower and
× Cn (16) upper decision variables bounds specified by the user.

IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358– 371 365
doi: 10.1049/iet-rpg.2011.0204 & The Institution of Engineering and Technology 2012
www.ietdl.org
Table 3 Different small wind turbines EE

Product Swept area, m2 Wind turbine weight, kg Global EE, MJ Global EE, MJ/m2

southwest (Air X) 1.020 5.850 1346.909 1320.499

southwest (Whisper 100) 3.460 21.000 4630.560 1338.312
southwest (Whisper 200) 5.725 30.000 7495.722 1309.2
southwest (Skystream 3,7) 10.870 77.000 15 135.374 1392.399
Aeromax engineering
(Lacota S, SC) 3.430 16.000 4333.512 1263.415
Bergey (BWC 1500) 7.070 76.000 11 226.249 1587.871
Bergey (BWC XL, 1) 4.910 34.000 6795.031 1383.917
Bergey (BWC Excel-R) 35.260 477.000 61 212.101 1736.021
Bornay (Inclin 6000) 12.570 107.000 18 459.980 1468.574
abundant renewable energy (ARE110) 10.180 143.000 17 954.454 1763.699
abundant renewable energy (ARE442) 40.720 612.000 73 950.670 1816.077
Kestrel wind (800) 3.460 45.000 5910.273 1708.171
Kestrel wind (1000) 7.070 75.000 11 172.928 1580.329
Kestrel wind (3000) 11.340 150.000 19 504.738 1719.995
Proven WT 0,6 5.100 70.000 8907.392 1746.547
Proven WT 2,5 9.000 190.000 19 263.231 2140.359
Proven WT 6 23.600 500.000 50 607.266 2144.376
Proven WT 15 63.500 1100.000 1 23 086.038 1938.363
Mean (round number) 1631

The same problem can be treated as a multi-objective The system must ensure the desired LPSP. This reliability
design, minimising, simultaneously, EE and LPSP. In this objective is included in the block, ‘LPSP’.
case, it is presented by formulation (II). The design EE of the system has to be minimised. This
Formulation (II): objective is included in the block, ‘Primary EE.
Object functions:
Minimum  Hybrid system primary 5.3 Choice of the optimisation algorithm
EE cost (EE)
and The hybrid system optimisation problem is a constrained non-
Minimum  LPSP (Dt, Apv , Awt , Cn) linear multivariable programming model. The EE object
Subject to feasibility constraints (15) – (17). function is linear and the LPSP subject/object function is
With this formulation, it will be for the designer to choose non-linear as confirmed by Fig. 7: no explicit representation
the solution that responds to its criteria in terms of LPSP. of this function is available. Its different values are
calculated using dynamic simulation. Thus, a mathematical
5.2 Computational method programming optimisation method has been used.
Generally, for mono-objective optimisation (formulation
Optimisation has been carried out using Simulink and Matlab I), a conventional optimisation method can be appropriated
tools. Thus, all the mathematical models of the hybrid wind – because of efficiency reasons (the number of calculations
PV-batteries system have been implemented in the Matlab/ are limited). Thus, the fmincon function of Matlab for non-
Simulink environment as shown in Fig. 6. The main linear programming has been used. This function uses an
subsystems are wind turbine model, photovoltaic array, algorithm Sequential Quadratic Programming (SQP) [38].
ideal battery bank and battery controller. The stop criteria are: average change in objective and

Fig. 7 LPSP function non-linearity identification

a LPSP as a function of photovoltaic array area (Apv) and wind turbine rotor swept area (Awt)
b LPSP as a function of battery nominal capacity (Cn)

366 IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358 –371
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-rpg.2011.0204
 www.ietdl.org

Fig. 8 SQP non-linear programming mono-objective optimisation results

constraint function is less than functions tolerance which is total primary EE cost as small as possible. The optimisation
equal to 1E-6. This value is small enough to well approach has been done twice according to problem formulation (I)
the global optimum. The optimisation routine in this or (II).
case runs for 1 h in order to converge [with an Intel Core In the first case, the optimisation routine was repeated ten
Duo processor (CPU T6500 @2.1 GHz) with parallel times with different initial conditions and decision variables
processing]. bounds. The final solution for problem formulation (I) is
Subsequently, to solve the formulation (II) sizing problem, described in Fig. 8. It has been obtained in 26 iterations and
we have chosen a controlled elitist genetic algorithm (a after 242 evaluations of the objective function starting from
variant of NSGA-II [39, 40]) implemented in Matlab as the initial conditions [Apv_in ¼ 10 m2, Awt_in ¼ 10 m2,
‘gamultiobj’ command because it provides a good Cn_in ¼ 200 Ah]. Decision variables bounds are [Apv_min ¼
distribution of solutions on the compromise surface, without 6 m2, Apv_max ¼ 15 m2, Awt_min ¼ 6 m2 Awt_max ¼ 12 m2,
requiring too complex settings. In fact, this algorithm Cn_min ¼ 150 Ah, Cn_max ¼ 300 Ah]. The design
always favours individuals with better fitness value (rank) configuration of the hybrid system is obtained by:
that can help to increase the diversity of the population photovoltaic modules’ installed area Apv ¼ 13.349 m2, wind
even if they have a lower fitness value. The parameters turbine swept area Awt ¼ 11.421 m2 and battery bank storage
used in our case are the following: population size: 100, capacity gives Cn ¼ 240.57 Ah.
crossing fraction: 0.8, mutation rate: 0.2, Pareto front In the second case of multi-objective optimisation (MOP)
fraction: 0.8 and stop criterion is maximum number of according to problem formulation (II), the solution is not a
generations ¼5 to accelerate the optimisation routine. The single solution but a set of solutions known as a Pareto
results are obtained after 2 h simulation. optimal set, named a Pareto border or Pareto front. Any
solution of this set is optimal in the sense that no
6 Results and discussions improvement can be made on one of the two objective
functions without worsening at least one of them. Fig. 9
6.1 Results shows the obtained Pareto front. The Pareto set is
composed of 80 solutions. This method allows the user to
A few residence electric loads are not ‘critical’ in a house and choose between different solutions and gives an idea about
do not require continuous quality electric power. Their the evolution of EE according to LPSP. We remark that EE
momentary outage in case of power lack can be tolerated is very sensitive to LPSP. In fact, the more the consumer
by the residents. Therefore 5% of shedding can be admitted tolerates load shedding, the greater the hybrid system is
to avoid the addition of a diesel generator or to not oversize undersized and therefore cheaper in terms of the primary
the hybrid system. Therefore the optimisation procedure energy requirements. Indeed, there is a considerable
aims to find the wind turbine rotor swept area (Awt), variation of EE (up to 30%) between sizing a system that
photovoltaic array area (Apv) and battery bank provides 95 and 100% of system electrical needs.
nominal storage capacity (Cn) to meet at least 95% Finally, the solution that satisfies maximum 5% LPSP
(LPSP_max ¼,5%) of electrical energy demand with a criterion is chosen:

IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358– 371 367
doi: 10.1049/iet-rpg.2011.0204 & The Institution of Engineering and Technology 2012
www.ietdl.org

Fig. 9 Pareto front

† Photovoltaic modules’ installed area Apv ¼ 13.06845 m2; In comparison with the results from the mono-objective
† Wind turbine swept area Awt ¼ 11.6112 m2; optimisation, this solution provides a lower LPSP (4.881%
† Battery bank storage capacity gives Cn ¼ 249.9426 Ah. ,5%) but a higher total primary system EE

Fig. 10 Battery bank SOC over the year

Table 4 Solution evaluations

Year Load, kWh PV, kWh Wind, kWh Excess, kWh Unmet load, kWh %Unmet load (LPSP)

2002 2204 3170 3395 3808 59.65 2.706

2003 2203 3045 3763 4047 72.58 3.294
2004 2193 3046 3234 3549 74.06 3.377
2005 2195 3068 3324 3676 91.89 4.186
2006 2190 3150 4481 4829 38.61 1.763
2007 2198 3057 3619 3938 85.88 3.908
2008 2209 3135 4674 4978 51.44 2.328
2009 2199 3013 3702 3976 96.32 4.38

368 IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358 –371
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-rpg.2011.0204
 www.ietdl.org
Table 5 Hybrid system LCC analyses (138189.14 MJ . 13749.3 MJ). The choice of optimisation
method depends on user requirements in terms of rapidity/
Item LCC, $ Present % Total LCC
precision. The SQP method is rapid but it does not always
worth, $
guarantee finding a good optimum. Those routines need
capital costs various repetitions to get an a posteriori idea about the
array 2370 2370 6.76473 adequate solution. The MOP is slower but has the
wind turbine 5400 5400 15.41331 advantage of giving a multitude of solutions.
batteries 1343 1343 3.83334 In our case, the first solution: Apv ¼ 13.349 m2,
tower (10 m) 1300 1300 3.71061 Awt ¼ 11.421 m2 and Cn ¼ 240.57 Ah ensure that the
PV charge controller mppt 500 500 1.42715 lowest EE is retained.
inverter 2159 2159 6.16247
BOS (PV and WT) 3858.4 3858.4 11.0130 6.2 Solution evaluation
installation 4000 4000 11.4172
recurring costs This evaluation aims to verify the reliability of the proposed
maintenance (PV and WT) 400 7863 22.44349 methodology and the compliance with design criteria
replacement (Table 4).
batteries 5 years 1343 1219.87 3.48189 The optimal configuration for hybrid PV – wind batteries
batteries 10 years 1343 1108.04 3.16269 system is composed as follows:
batteries 15 years 1343 1006.46 2.87275
batteries 20 years 1343 914.18 2.60935 † Photovoltaic modules’ area Apv ¼ 13.349 m2; six CHSM
PV charge controller 500 374.7 1.06951 6610P-235 235 W panels [41] for instance can be installed;
mppt 15 years † Wind turbine swept area Awt ¼ 11.421 m2; a Kestrel
inverter 15 years 2159 1618 4.61828 (3000 W) wind turbine has a close swept area of 11.34 m2 [42];
total 35 034.65 $ 100 † Batteries bank storage capacity gives Cn ¼ 240.57 Ah,
eight Deep Cycle Numax SLC120 – 12 12 V 120 Ah battery
[43], to form two parallel strings of four batteries, each one
can be installed.

6.2.1 Reliability evaluation: Table 5 validates the

adequacy of the proposed sizing method. For all the years,
lack of production does not exceed 5%. However, a
considerable excess is denoted. Its exploitation depends on
the system supervision technique, hence it can be used
for cooking or heating water or it is possible to avoid its
production by generators maximum power point degradation.
Concerning battery bank SOC, it is important to note that
throughout the year it is always between 20% Cn and 100%
Cn as shown in Fig. 10 and Fig. 11.
However, installed batteries should be able to support rare
power picks because of high wind speed that can reach
7712.6 W, in fact an occasional maximum current to 80 A
Fig. 11 Battery bank DOD frequency distribution over the year

Fig. 12 Battery bank power evolution over the year

IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358– 371 369
doi: 10.1049/iet-rpg.2011.0204 & The Institution of Engineering and Technology 2012
www.ietdl.org
under 48 V occurred in the case of two battery strings in 2 Gado, A., El-Zeftawy, A.: ‘Design and economy of renewable energy
parallel (see Fig. 12). sources to supply isolated loads at rural and remote areas of Egypt’.
Twentieth Int. Conf. Exhibition on Electricity Distribution, Prague,
Hence globally, the results prove the reliability of the Czech Republic, 8 –11 June 2009, (CIRED 2009), IET Conf. Pub.
proposed methodology for sizing autonomous hybrid Issue CP550, pp. 404
wind – PV systems with batteries storage. 3 Lee, T.-Y., Chen, C.-L.: ‘Wind-photovoltaic capacity coordination for a
time-of-use rate industrial user’, IET Renew. Power Gener., 2009, 3, (2),
pp. 152–167
6.2.2 LCC evaluation: The resulting LCC analyses based 4 Borowy, B., Salameh, Z.M.: ‘Methodology for optimally sizing the
on components prices approximations is shown in Table 5. combination of a battery bank and PV array in wind/PV Hybrid
An LCC of about 35,000 $ is obtained at the end of the system’, IEEE Trans. Energy Convers., 1995, 11, (2), pp. 367– 375
process. This considerable amount proves the important 5 Chedid, R., Rahman, S.: ‘Unit sizing and control of hybrid wind-
solar power systems’, IEEE Trans. Energy Convers., 1997, 12, (1),
weight of economic considerations. It is recommended to pp. 79–85
integer it to the design stage. Therefore it would be very 6 Kellogg, W.D., Nehrir, M.H., Venkataramanan, G., Gerez, V.:
interesting to make a triple MOP, minimising ‘Generation unit sizing and cost analysis for stand-alone wind,
simultaneously LCC, EE and LPSP. photovoltaic, and hybrid wind/PV systems’, IEEE Trans. Energy
Convers., 1998, 13, (1), pp. 70–75
7 Yang, H., Zhou, W., Lou, C.: ‘Optimal design and techno-economic
7 Conclusions analysis of a hybrid solar-wind power generation system’, Appl.
Energy, 2009, 86, (2), pp. 163– 169
In this paper, a methodology for optimising the size of a 8 Diaf, S., Notton, G., Belhamel, M., Haddadi, M., Louche, A.: ‘Design
hybrid wind –PV batteries system has been successfully and techno-economical optimization for hybrid PV/Wind system
under various meteorological conditions’, Appl. Energy, 2008, 85,
implemented. The optimisation procedure consists of (10), pp. 968– 987
minimising, both system primary EE and LPSP. The 9 Shi, J.-h., Zhong, Z.-d., Zhu, X.-j., Cao, G.-y.: ‘Robust design and
optimisation is insured by a dynamic model of the system optimization for autonomous PV-wind hybrid power systems’,
using the Matlab/Simulink environment. Optimal J. Zhejiang Univ. SCIENCEA, 2008, 9, (3), pp. 401– 409
configuration is obtained by applying a non-linear SQP 10 Meteorological Data source: Available at http://www.nrel.gov/midc/
nwtc_m2/
algorithm for mono-objective optimisation and genetic 11 Abbes, D., Martinez, A., Champenois, G., Gaubert, J.P., Kadri, R.:
algorithm for multi-objective configuration. ‘Estimation of wind turbine and solar photovoltaic energy using
The methodology has been applied successfully for the sizing variant sampling intervals’. Fourteenth Int. Power Electronics and
of a wind–PV-batteries system with the main goal to supply at Motion Control Conf. (EPE/PEMC), Ohrid, Republic of Macedonia,
6 –8 September 2010, ID 151, T12– 28 T12–34
least 95% in electric energy for a residential house. The results 12 Shepherd, W., Shepherd, D.W.: ‘Energy studies’ (Imperial College
show that the optimal configuration which meets the desired Press, 2003, 2nd edn.), pp. 306– 311
system reliability requirements (LPSP ¼ ,5%) with the 13 Tong, W.: ‘Wind power generation and wind turbine design’ (WIT
lowest primary EE is obtained for a system composed of six Press, 2010), p. 20
235 W photovoltaic modules, one wind generator (3 kW) and 14 Hofierka, J., Kaňuk, J.: ‘Assessment of photovoltaic potential in urban
areas using open-source solar radiation tools’, Renew. Energy, 2009,
batteries storage with 240 Ah capacity storage. 34, (10), pp. 2206– 2214
EE has been considered as a criterion that denotes the 15 Liu, B., Duan, S., Cai, T.: ‘Photovoltaic DC building module based
innovative aspect of this work among previous BIPV system: concept and design considerations’, IEEE Trans. Power
investigations. The analysis of primary energy requirements Electron., 2011, 26, (5) (99), pp. 1418– 1429
has been performed for each component. Owing to poor 16 Kalogirou, S.: ‘Solar energy engineering processes and systems’
(Academic Press, 2009), Ch. 9, photovoltaic systems, pp. 480
contributions in the literature, a primary energy assessment 17 Messenger, R.A., Ventre, J.: ‘Photovoltaic systems engineering’ (CRC
for small wind turbines has been conducted directly by the Press, 2005, 2nd edn.), Ch. 7: Stand-alone PV systems, p. 209
authors. The results show that the batteries are greedy in 18 Lasnier, F., Ang, T.G.: ‘Photovoltaic engineering handbook’ (published
primary energy requirements. Their manufacturing must be by Taylor & Francis, 1990, 1st edn.), Ch. 11, p. 258
significantly improved, for example by introducing better 19 Ekren, O., Ozerdem, B.: ‘Hybrid combination of photovoltaic and wind
energy conversion system, sizing and optimization of hybrid energy
energy management techniques. Moreover, our analysis systems’ (Lambert Academic Publishing, 2010), p. 44
confirms that small wind turbines are more energy efficient 20 Hongxing, Y., Lu, L., Burnett, L.J.: ‘Weather data and probability
than photovoltaic panels. analysis of hybrid photovoltaic –wind power generation systems in
In the end of the paper, a reliability and economic Hong Kong’, Renew. Energy, 2003, 28, (11), pp. 1813–1824
21 Koutroulis, E., Kolokotsa, D., Potirakis, A., Kalaitzakis, K.:
evaluation has been carried out. This one shows the efficacy ‘Methodology for optimal sizing of stand-alone photovoltaic/wind-
of the proposed method. This procedure can help designers generator systems using genetic algorithms’, Sol. Energy, 2006, 80,
of hybrid wind – PV-batteries systems to take into (9), pp. 1072– 1088
consideration the environmental impact in addition to 22 Kimball, J.W., Kuhn, B.T., Balog, R.S.: ‘A system design approach for
economic evaluation. This study can also contribute to a unattended solar energy harvesting supply’, IEEE Trans. Power
Electron., 2009, 24, (4), pp. 952–962
triple MOP considering LCC, EE and LPSP for power 23 Luque, A., Hegedus, S.: ‘Handbook of photovoltaic science and
hybrid systems design. engineering’ (John Wiley& Sons Ltd, 2003), pp. 812–849
24 Abouzahr, I., Ramakumar, R.: ‘Loss of power supply probability of
standalone photovoltaic systems: a closed form solution approach’,
8 Acknowledgment IEEE Trans. Energy Convers., 1991, 6, (1), pp. 1 –11
25 Bin, A., Hongxing, Y., Hui, S., Xianbo, L.: ‘Computer aided design
The authors would like to thank Region Poitou-Charentes for PV/wind hybrid system’, Renew. Energy, 2003, 28, (10),
(Convention de recherche GERENER N8 08/RPC-R-003) and pp. 1491–1512
Conseil General Charente Maritime for their financial support. 26 Kornelakis, A., Koutroulis, E.: ‘Methodology for the design
optimisation and the economic analysis of grid-connected photovoltaic
systems’, IET Renew. Power Gener., 2009, 3, (4), pp. 476– 492
9 References 27 Markvart, T., Castañer, L.: ‘Practical handbook of photovoltaic:
fundamentals and applications’ (Elsevier, 2003), Part V, Ch. 2,
1 Bernal-Agustin, J.L., Dufo-Lopez, R.: ‘Simulation and optimization of pp. 870–884
stand-alone hybrid renewable energy systems’, Renew. Sustain. 28 Stoppato, A.: ‘Life cycle assessment of photovoltaic electricity
Energy Rev., 2009, 13, (8), pp. 2111–2118 generation’, Energy, 2008, 33, (2), pp. 224– 232

370 IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358 –371
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-rpg.2011.0204
 www.ietdl.org
29 Alsema, E.A., Nieuwlaar, E.: ‘Energy viability of photovoltaic systems’, 35 Ashby, M.F.: ‘Materials and the environment: eco-informed material
Energy Policy, 2000, 28, (14), pp. 999 –1010 choice’ (Elsevier, 2009), Ch. 7, pp. 146– 148
30 Koroneos, C., Stylos, N., Moussiopoulos, N.: ‘LCA of multicrystalline 36 Hammond, G., Jones, G.: ‘Inventory of carbon and energy (ICE)’
silicon photovoltaic systems- part 1: present situation and future (University of Bath, 2008), Version 1.6a
perspectives’, Int. J. Lifecyc. Assess., 2030, 11, (2), pp. 129 –136 37 Johanson, T., Burnham, L.: ‘Renewable energy: sources of fuels &
31 Alsema, E.A.: ‘Environmental life cycle assessment of solar electricity’ (Island Press, 1992, 1st edn.), p. 147
home systems’, Report NWS-E-2000– 15, Department of 38 Nocedal, J., Wright, S.J.: ‘Numerical optimization’ (Springer Verlag,
Science, Technology and Society, Utrecht University, Utrecht, 2002, 2006, 2nd edn.), Ch. 18, Springer series in operations research
Available at http://www.projects.science.uu.nl/nws/publica/Publicaties 39 Katsigiannis, Y.A., Georgilakis, P.S., Karapidakis, E.S.: ‘Multiobjective
2000/e2000-15.pdf genetic algorithm solution to the optimum economic and
32 De Beet, J.: ‘Potential for industrial energy-efficiency improvement environmental performance problem of small autonomous hybrid
in the long term’ (Kluwer Academic Publishers, The Netherlands, power systems with renewables’, IET Renew. Power Gener., 2010, 4,
2000) (5), pp. 404– 419
33 Ancona, D., McVeigh, J.: ‘Wind turbine – materials and manufacturing 40 Deb, K., Agrawal, S.A., Pratap, T.: ‘Meyarivan, a fast and elitist multi-
fact sheet’, pdf prepared for the Office of Industrial Technologies, US objective genetic algorithm: NSGA-II’, IEEE Trans. Evol. Comput.,
Department of Energy by Princeton Energy Resources International, 2002, 6, (2), pp. 182 –197
LLC, 29 August 2001 41 http://www.wholesalesolar.com/products.folder/module-folder/
34 Fleck, B., Huot, M.: ‘Comparative life-cycle assessment of a small wind Astronergy/Astronergy235.html
turbine for residential off-grid use’, Renew. Energy, 2009, 34, (12), 42 http://www.renewableintegrator.com/Kestrel%203000.htm
pp. 2688– 2696 43 http://www.tayna.co.uk/Numax-SLC120-12-P7628.html

IET Renew. Power Gener., 2012, Vol. 6, Iss. 5, pp. 358– 371 371
doi: 10.1049/iet-rpg.2011.0204 & The Institution of Engineering and Technology 2012