Renewable and Sustainable Energy Reviews 52 (2015) 757–769

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

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

Reviewing the potential and cost-effectiveness of off-grid PV systems

in Indonesia on a provincial level
A.J. Veldhuis n, A.H.M.E. Reinders
University of Twente, Faculty of CTW, Department of Design, Production and Management, P.O. Box 217, 7500 AE Enschede, The Netherlands

art ic l e i nf o a b s t r a c t

Article history: In this study the amount and costs of off-grid PV systems required to electrify Indonesian rural
Received 15 November 2014 households lacking electricity access are estimated. Due to the Indonesian geography large differences
Received in revised form exist among different provinces, therefore this study evaluates the potential and costs of off-grid PV
23 May 2015
systems for each province as a follow-up of a related study on grid-connected PV systems which we
Accepted 27 July 2015
Available online 25 August 2015
executed in 2012. In this study we calculate cumulative numbers for the nominal power of installed off-
grid PV systems, their LCOE and the relative financial benefits compared with diesel generators, which
Keywords: are a common means for electricity generation in remote areas.
PV systems To determine this potential, a mathematical model has been developed which is based on available data on
Off-grid
among others population density, urbanisation ratio, irradiation, electrification ratio and electricity demand.
Indonesia
We distinguished between stand-alone and hybrid PV systems. Results show that the costs of off-grid
Rural
Potential hybrid PV systems with an average LCOE of 0.38 USD/kWh are 19% cheaper compared with electricity
generation by diesel gensets in most rural parts of Indonesia. Stand-alone PV systems show an average LCOE of
0.76 USD/kWh which is 3% cheaper than stand-alone diesel gensets on average. The potential of off-grid PV
systems is 969 GWh/year, of which 566 GWh/year generated by hybrid PV systems and 403 GWh/year by
stand-alone PV systems.
& 2015 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 758
2. Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 759
2.1. Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 759
2.2. Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 760
2.3. Off-grid electricity demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 761
2.4. Rural population distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 762
2.5. PV system configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 762
2.6. PV system sizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 763
2.7. Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 764
3. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 765
4. Sensitivity analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 766
4.1. Hybrid PV system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 766
4.2. Stand-alone PV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 766
4.3. Overall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 767
5. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 767
6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 768
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 769
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 769

n
Corresponding author. Tel.: þ 31 534893192.
E-mail addresses: a.j.veldhuis@utwente.nl (A.J. Veldhuis), a.h.m.e.reinders@utwente.nl (A.H.M.E. Reinders).

http://dx.doi.org/10.1016/j.rser.2015.07.126
1364-0321/& 2015 Elsevier Ltd. All rights reserved.
758 A.J. Veldhuis, A.H.M.E. Reinders / Renewable and Sustainable Energy Reviews 52 (2015) 757–769
1. Introduction
Indonesia is a large archipelago with over 17,000 islands of
which roughly a half is inhabited, posing huge challenges to the
electricity distribution. In addition, approximately fifty per cent of
the total Indonesian population lives in rural areas [1]. Electrifica-
tion of rural areas remains difficult: The lower population den-
sities in these areas and thus lower electricity demand do not
justify the huge investments associated with the infrastructure
required for grid electricity. Therefore, most rural areas in Indo-
nesia are electrified by diesel generators, which is very expensive
and polluting. Besides, the growing electricity demand, depletion
of oil resources and climate goals require better solutions. Since
Indonesia is a tropical country receiving abundance irradiation
with 4.8 kWh/m2/day on average, PV systems can be an excellent
solution to electrify remote areas on distributed islands. Although
this potential is often recognised in the literature, the actual
potential of off-grid PV systems has not determined in detail so
far, only some rough estimates can be found.
A study from the IEA estimated the technical off-grid PV
potential to be 900 MW [2], based on 50Wp SHS for 65% of
households without electricity in 1999, assuming 5 people per
household. Another study estimates the theoretical off-grid PV
potential for Indonesia to be 1300 MWp [3], based on 50% of the
population without access to electricity in 2005.
The overall electrification ratio of households in Indonesia was
73% by the end of 2012, which varies per province from around
25% in some eastern islands to nearly 100% in Jakarta [4]. In
general, people living in densely populated urban areas have
access to an electricity grid, however most of Indonesia's sparsely
populated rural areas lack grid connection.
Electricity is an important driver for economic activity and
especially in rural areas without grid connection, PV systems can
play an important role to provide power to improve the economics
in these regions [5]. In order to compare the cost-effectiveness of
PV systems with other forms of energy, the levelized cost of
electricity (LCOE) is often applied [5–7]. Apart from the cost-
effectiveness, PV systems can be assessed by their environmental
and energetic performances by using some metrics such as the
Greenhouse Gas Payback Time (GPBT) and the Energy Payback
Time (EPBT) [8].
A study from Akikur et al. [9] reviewed the feasibility of off-grid
rural electrification by hybrid and stand-alone solar PV energy
Fig. 1. Names and average daily irradiation in kWh/m2 for each province in Indonesia. These data were obtained from the NASA Langley Research Center Atmospheric
Science Data Center Surface meteorological and Solar Energy (SSE) web portal supported by the NASA LaRC POWER Project.
systems. Based on various case studies and demonstration projects
throughout the world, they concluded that PV systems can be a
viable solution to cost-effectively provide power in rural areas. To
determine the potential and cost-effectiveness of off-grid PV
systems in rural areas in Indonesia we have first evaluated studies
done for other countries.
A study from Shaahid et al. [10] examined hybrid PV-diesel–
battery systems for five distinct locations in Saudi Arabia using the
software HOMER from NREL (National Renewable Energy Labora-
tory). Simulation results show the diesel fuel consumption, CO2
emissions and the cost of energy for variations in PV penetration
for a given residential load. However, due to differences in climate
conditions and in the costs of diesel fuel in Saudi Arabia compared
with Indonesia, the results will probably be different for Indonesia.
In a study from Breyer et al. [11] a model based on Geographic
Information Systems (GIS) was used to determine the LCOE of
various energy technologies per location. The LCOE of traditional
diesel power was compared with the LCOE of PV and wind energy
systems. The method led to a global map showing the energy
technology with the least cost option per location. However, the
study focused on electrified areas only, excluding non-electrified
settlements from their analysis. Besides, the potential of the
nominal installed capacity of PV systems has not been determined.
A similar spatial-economic analysis has been carried out by Szabo
et al. [12], they studied the least cost option for the rural electrification
for Africa, based on diesel generators, off-grid PV systems and grid
extension. In an updated version, they included micro-hydropower in
their analysis [13]. They found that off-grid PV systems are already
cost-effective in large areas of Africa. However, it is hard to say
whether these findings apply for Indonesia as well, because of
differences in: climate, grid penetration, population densities and
economics for energy supply (e.g. subsidies). Moreover and probably
most important, the many islands of the Indonesian archipelago pose
a special problem on energy distribution, both for fuels and electricity.
For this reason electricity supply in Indonesia has a strong geographic
dependency.
Blum et al. [14] assessed two electricity demand scenarios for
the electrification of villages in Indonesia. The LCOE of off-grid PV
systems, micro-hydropower and diesel generators is evaluated for
both scenarios. However, geographic variations are not taken into
account, so it is difficult to determine the most suitable region for
these technologies. Besides, the study did not focus on the overall
potential of PV systems in Indonesia.
 A.J. Veldhuis, A.H.M.E. Reinders / Renewable and Sustainable Energy Reviews 52 (2015) 757–769 759

Based on the literature review, it seems that the potential
of off-grid PV systems in Indonesia has not been assessed in
detail. Improvements can be made by including (a) areas
without electricity access, (b) the specific situation in Indonesia,
(c) geographic variations of settlements. Therefore, this study will
estimate the potential of off-grid PV systems in Indonesia at a
provincial level as a follow-up of a study on the potential of grid-
connected PV systems in Indonesia, which we executed in 2012
[15]. For this specific study we use the same methodology in a
slightly adapted form, leading to cumulative numbers for nominal
power installed off-grid PV systems, their LCOE and the relative
financial benefits compared with diesel generators. The method is
largely determined by available irradiation on site, population
density, residential electricity demand and grid-penetration. A
preliminary version of this method has been documented in
[16], here we present the full procedure together with a sensitivity
analysis.
In Section 2 the applied methodology will be described in
detail. Section 3 presents the results along with a sensitivity
analysis in Section 4. In Section 5 the results and methodology
will be discussed and finally the conclusions are presented in
Section 6. The next section starts with a description of the context
of this study.
2. Methodology
2.1. Background
In this section several relevant variables for our study, such as
irradiation and electrification ratios, will be presented in a geo-
graphic context to provide insight in quantitatively available data
at our disposal.
The average solar irradiation in Indonesia can be divided into
two regions, one with approximately 4.5 kWh/m2/day with a
monthly deviation around 10% for Western regions and for the
Eastern regions this amount is about 5.1 kWh/m2/day with a
monthly deviation around 9% [17]. In Fig. 1, the various colours
indicate the average daily irradiation in kWh/m2/day per province
ranging from 4.4 to 6.2 kWh/m2/day. These values are based on
satellite derived climatological representative values with a

Fig. 2. The provinces of Indonesia and their population densities.

Fig. 3. The percentage of rural households lacking access to electricity per province in Indonesia. Based on data from 2010 from BPS [1].
760 A.J. Veldhuis, A.H.M.E. Reinders / Renewable and Sustainable Energy Reviews 52 (2015) 757–769
Fig. 4. The geographical distribution of non-electrified rural households per province in Indonesia. Based on data from BPS [1].
temporal coverage of more than 22 year (July 1983 through June
2005) [18].
The population density is an important factor in modelling the
off-grid PV potential. Large differences exist among the provinces
in Indonesia as can be seen in Fig. 2.
The provinces of Java and Bali have high population densities,
these provinces account for roughly 60% of the total Indonesian
population of 238 million people [1], although they only comprise
7% of the land area of Indonesia. This is one of the main reasons
why these provinces are more developed than the other pro-
vinces: In general, economies of scale are a prerequisite for
investments in large infrastructural projects in order to obtain a
reasonable return on investment.
Data about the primary source of lighting per household from
Statistics Indonesia (BPS) [1] is assumed to give a good represen-
tation of the households with access to electricity, this methodol-
ogy is used in a similar study as well [19]. From these data can be
concluded that 17% of homes in rural areas do not have access to
the national grid and 11% lack electricity at all. These are averages
for Indonesia; locally these numbers can vary significantly. As can
be seen in Fig. 3 in which the percentage of rural households
without access to electricity is shown for each province. Numbers
vary from 2% in Java to 80% in Papua.
Most of the rural households without access to electricity live
in the provinces East Nusa Tenggara (ENT) and Papua, with
respectively 63% of the 820 thousand rural homes and 80% of
the 490 thousand rural homes, see Figs. 3 and 4.
In contrast, the provinces at Java show low shares of these rural
households lacking electricity access. Besides, the Special Capital
Region of Jakarta contains even no rural households, although 647
urban households without access to electricity exist.
As can be seen in Fig. 4, roughly a quarter of all rural house-
holds in Indonesia without electricity live in ENT and Papua. So,
even in absolute numbers, most households lacking access to
electricity live in these two provinces. Especially the province of
ENT seems to be a promising location for off-grid PV systems, due
to its large number of rural households without access to elec-
tricity combined with the high level of irradiation received in this
part of Indonesia, as shown in Fig. 1.
Comparing Fig. 3 with Fig. 4 shows the influence of the
population density. Due to the high population density at Java,
the significant lower relative values for Java compared with the
other provinces as shown in Fig. 3, are less pronounced in the
absolute values as shown in Fig. 4.
2.2. Approach
The main research question in this study is: What is the
potential and cost-effectiveness of off-grid PV systems in each of
the provinces of Indonesia? In this section the applied methodol-
ogy to answer this question will be described.
Since large geographical differences exist in Indonesia and
most relevant data are available on a provincial level, this study
examines the potential of off-grid PV systems for each of the 33
provinces1. A mathematical model has been developed to deter-
mine this potential based on among others: population density,
urbanisation ratio, irradiation, electrification ratio and electricity
demand.
Due to data availability we decided to base our methodology on
data from 2010 for ease of comparison between different informa-
tion sources.
It is assumed that urban households which lack access to
electricity will be at PLN's waiting list and will get connected to
the grid in the near future. Besides, in these urban areas grid
extension is assumed to be the most cost-effective electrification
option, therefore the scope of this study is limited to the rural
households without access to electricity. As such, the province of
Jakarta is excluded from this analysis, since it contains no rural
households at all, leaving only 32 provinces left for this study.
Since other renewable energy sources, such as geothermal
plants, large hydropower, wind turbines and modern biomass
installations are mainly grid-connected, it is assumed that they
do not influence the off-grid PV potential. Small-scale wind and
micro-hydro energy systems can be applied in off-grid systems
and can as such influence the economic off-grid potential of PV
systems. Since the potential of wind energy is rather low due to
the prevailing low wind speeds in Indonesia, varying between
2 and 6 m/s on average [20], we assume that the off-grid potential
of wind energy can be neglected.
Micro-hydro is a proven technology which is reliable and
requires little maintenance [14]. When the local resources are
sufficient, the LCOE of micro-hydro is around 0.15 €/kWh [14].
However, micro-hydro has a strong seasonal dependency due to
the monsoon while solar PV is a constant energy source the whole
1
In 2012 the northern part of the province of East Kalimantan has formed its
own province: North Kalimantan. Since we base our study on data from 2010, we
use the former province of East Kalimantan.
 A.J. Veldhuis, A.H.M.E. Reinders / Renewable and Sustainable Energy Reviews 52 (2015) 757–769
Fig. 5. Schematic diagram of the methodology.
year through, besides the potential of micro-hydropower is highly
location dependent, therefore we do not include it in the model.
To calculate the actual potential of off-grid PV systems the
following general assumptions have been made:
(i) Other renewable energy technologies such as hydropower,
geothermal energy, wind energy or biomass are not taken into
account.
(ii) Rural electricity demand is a dominant factor.
(iii) Off-grid PV systems are only viable in rural areas lacking grid-
connection
(iv) The model is based on a provincial level, so regional variations
inside the province (i.e. at a district level) itself are outside the
scope of this study.
In order to determine the potential and costs of off-grid PV
systems, the modelling approach is divided into several steps, as
schematically shown in Fig. 5. The figure shows the different steps
and the sections in which the related assumptions and equations
are described.
Below we will explain the assumptions and equations applied
in the calculations for each step in our method.
761
2.3. Off-grid electricity demand
Detailed data on rural electricity demand in Indonesia are
lacking. Average household's electricity consumption in rural areas
in developing countries vary between 240and 768 kWh/year
[21,22]. Based on these figures in combination of the available
data from the State Electricity Company (PLN) [23] and BPS [1], it
is assumed that an Indonesian household living off-grid would
consume between 250 and 450 kWh/year. Further is assumed that
the rural electricity demand is roughly the same for each province
in Indonesia.
Apart from the residential electricity consumption, there will
be other end-uses in the commercial or public sector, like: local
workshops, offices and schools. Their share would be relatively
low in the agricultural areas which are not connected to the grid
[24]. Although detailed data are lacking, based on rural employ-
ment data [1] and a study from Gibson and Olivia [25], we
estimate that this would not exceed 5% of the connections,
especially in areas lacking electricity. As such they are included
in our study as households with productive use, see Table 1.
For the sizing of off-grid PV systems the daily demand profile is
important. Details are lacking, however most households in rural
areas use their electricity mainly for lighting and in a lesser extent
762 A.J. Veldhuis, A.H.M.E. Reinders / Renewable and Sustainable Energy Reviews 52 (2015) 757–769
for a fan and radio and TV [19,21]. Obviously, the electricity
required for lighting will largely take place between sunset
and sunrise, especially during the time before sunrise from
roughly 5 a.m. – 6 a.m. and between sunset and bedtime, so
roughly from 6 p.m. – 10 p.m.. In contrast, non-residential con-
sumers will consume most of their electricity during daytime.
To be able to model the off-grid PV potential two categories are
proposed (see Table 1). The high demand category extends the low
demand category. It is assumed that the electricity for productive
use is required during working days only and mainly during
daytime, therefore the peak load is only slightly higher compared
with the low demand category.
The off-grid PV potential can then be determined by:
Ep;PV ¼ αh HH p;h EHH;h þ αl HH p;l EHH;l
where Ep,PV is the electricity (kWh/year) generated by PV in
province p, α ( 0 r α r 1) determines the share of the electricity
demand which is generated by PV, HHp is the number of rural
households in province p, EHH is the yearly electricity demand of a
household. The subscripts h and l indicate the high demand or low
demand category, respectively.
We assume that the total rural electricity generation in a
province is based on a combination of PV systems and diesel
generators.
2.4. Rural population distribution
Now the electricity demand of a rural household is modelled, it
is necessary to obtain information about the clustering of these
households in order to determine the energy demand per area.
Therefore the population density distribution of the rural popula-
tion without access to electricity per province, Np,rur,ne, should
be known.
Table 1
Two categories for electricity demand per household.
Category EHH (kWh/year) Lpeak (kW)
Low (mainly lighting, TV & radio) 250 0.128
High (with productive use) 500 0.138
Fig. 6. Schematic diagram of the population density distribution.
Detailed data is not publicly available, estimates from SEDAC
[26] give an indication of the population distribution in Indonesia
on a country level, however a drawback of the estimations from
SEDAC is the skewed population distribution in Indonesia, as can
be seen in Fig. 2. Therefore the national average is not representa-
tive for each of the provinces.
To be able to determine the rural electricity demand per area,
we model the population density distribution in this study.
Based on the assumption of average population densities (ND) for
various areas made in the previous study [15], the population density
distribution can be approximated. The geographical population den-
sity distribution follows often an exponential trend, therefore this
study proposes an exponential function for the modelling of this
distribution. This is schematically shown in Fig. 6, in which the green
ð1Þ
area is equal to the provincial rural population without grid access. The
population densities NDcity, NDsuburbs and NDrur,grid, which are included
in the figure, are based on the previous study on grid-connected PV
systems and represent the average population density in the urban
cores, suburbs and electrified rural villages, respectively [15]. The
urbanisation and electrification ratio of the province, respectively URp
and ERp, are illustrated in Fig. 6 as well.
To calculate the actual population densities per province, some
additional boundary conditions have to be set. Although most
rural people live inside small settlements, it is assumed that in
each province there is at least one area in which the population
density is very low. This is modelled by the boundary condition
that the equal to or less than 10 persons/km2. Further is assumed
that the highest population density in non-electrified rural areas is
smaller than the population density assumed for rural electrified
villages, NDrur,grid, which equals 1000 persons/km2. The rationale
behind this is, that if the population density would be larger, it
would most likely already have been connected to the (local) grid,
because of the favourable conditions for infrastructure invest-
ments in these areas. Based on these aforementioned assumptions,
the rate of diminution, tp, in population density over distance can
be determined according to the flowchart shown in Fig. 7.
Further is assumed that each step i corresponds to a population
density NDi inside an area equal to 1 km2, so that the population
living in this area is numerically equivalent to the population
α density and i is numerically identical to the non-electrified rural
area of province p, Ap,rur,ne, in km2.
1.0
0.6
With the population distribution modelled, the electricity
demand per area can be determined.
 A.J. Veldhuis, A.H.M.E. Reinders / Renewable and Sustainable Energy Reviews 52 (2015) 757–769 763

Fig. 7. Flowchart of the determination of the population density distribution per province.

2.5. PV system configuration Furthermore, when a local grid is possible, the electricity
demand of the settlement is considered as high, otherwise the
There are two PV system configurations evaluated in this study: settlement's electricity demand falls in the low demand category.
(a) hybrid micro-grid consisting of PV systems, batteries and diesel Based on the minimal load requirement, the part of the rural
generators, and (b) stand-alone PV systems with batteries. population without electricity access which could be served by a
When a micro-grid is viable, it is assumed to be the most local grid, Np,lg,L, can be determined as follows:
preferred option. According to a rule of thumb for the feasibility of (  )
AX
p;rur;ne  ND
isolated grids from SWECO [27], the number of connections c in an  i
Np;lg;L ¼ 0:3  NDi  L 410 ð6Þ
area within a radius r of 500 m should be at least 100. We assume N HH;p;rur peak;h
i ¼ Ap;rur;lg
that the connections are equal to the number of households, since
most of the electricity consumers in rural areas are residential. The total population served by a local grid is therefore:
Then the minimum population density in a province for an Np;lg ¼ N p;lg;ND þ N p;lg;L ð7Þ
isolated grid in persons/km2, NDp,min, can be determined as
follows: Subsequently, the population served by stand-alone PV sys-
c tems, Np,sa, will be:
NDp;min ¼  N HH;p;rur ð2Þ
10  6 π r 2 Np;sa ¼ N p;rur;ne N p;lg ð8Þ
where N HH;p;rur is the average household size in rural areas of
where Np,rur,ne is the total rural population in province p with no
province p.
electricity access.
With an average household size of about 4 persons in rural
areas in Indonesia, the average minimum population density for a
local grid in Indonesia is 506 persons/km2. 2.6. PV system sizing
The population living in areas with a population density larger
than NDp,min are assumed to be served with a local grid. This area, System sizing is based on the assumed electricity demand
Ap,rur,lg,ND, is determined by: scenarios (Table 1) and the available irradiation (Fig. 1) based on

 data from the NASA2 [18] which has been averaged over each
1 NDp;min
Ap;rur;lg;ND ¼ ln ð3Þ 11  11 latitude/longitude grid cell inside the bounding box area of
tp NDrur;grid
each province which is obtained by [28].
Subsequently, the population living in this area, Np,lg,ND, is The following formulas related to the system sizing are pre-
obtained by: sented for system configuration a. If not explicitly mentioned, the
same formulas can be applied for configuration b, by replacing the
X
Ap;rur;lg;ND
N p;lg;ND ¼ NDi ð4Þ subscripts a and h by b and l, respectively.
i¼1 The daily electricity delivered by the PV system for a house-
where hold, EPV,HH,a (kWh/day), is determined as follows:

NDi ¼ NDrur;grid e  tp i ð5Þ

αh EHH;h
EPV ;HH;a ¼ ð9Þ
365 PRa
Besides, if the population density is less than NDp,min, different
factors, e.g. the load size and willingness to pay, determine where PRa is the performance ratio of the PV system in configura-
whether a local grid is viable or not [27]. To meet the criteria for tion a.
a local grid the minimum load L should be roughly 10 kW [27].
When this criterion has been met, it is assumed that in 30% of 2
Data based on monthly averaged values per latitude/longitude grid cell for a
these cases a local grid is feasible. 22-year period (July 1983 – June 2005).
764 A.J. Veldhuis, A.H.M.E. Reinders / Renewable and Sustainable Energy Reviews 52 (2015) 757–769

Table 2 Table 3
Assumptions for energy system sizing. Assumptions for the calculation of the LCOE.

Variable Symbol Value Unit Variable Value Unit

Diesel generator efficiency (a) ηDG,a 30 % Investment costs

Diesel generator efficiency (aref) ηDG,a,ref 25 % PV system (PV modules, mounting & installation) (IPV,p) 2500 $/kWp
Diesel generator efficiency (bref) ηDG,b,ref 20 % Inverter 500 $/kW
Diesel generator best efficiency point DGBEP 80 % Battery cost 150 $/kWh
Wires efficiency ηwires 97 % Diesel generator 650 $/kW
Battery charge controller efficiency ηctrl 98 % O&M costs
Inverter efficiency ηinv 95 % Annual operations PV 13 $/kW/year
Inverter capactiy factor finv 0.83 – Fixed O&M costs diesel 15 $/kW/year
Battery (dis)charge efficiency ηbatt 85 % Variable O&M costs diesel 0.03 $/kWh
Depth of discharge DOD 50 % Fuel
Autonomy (a) βa 60 % Energy density diesel 35.86 MJ/l
Autonomy (b) βb 200 % Additional transport costs (a) 10 %
Additional transport costs (b) 50 %
n
Different values for the efficiency of diesel generators are given, corresponding to Lifetime
system configuration (a) and (b). The subscript ref indicates the efficiency used in Diesel 20 years
the reference scenario in which only diesel generators are used to supply electricity. PV (N) 25 years
Inverter 12 years
Battery 5 years
The peak power of the PV modules for a household in province LCOE
Residual value (RV) 0.5 % of IPV,p
p, PPV,HH,a,p (kWp), is determined as follows:
Discount rate (DR) 5.75 %
EPV ;HH;a System degradation rate (SDR) 0.5 %
P PV ;HH;a;p ¼ ð10Þ
Hp

where Hp is the daily average horizontal irradiation (kWh/m2/day) [29,31]. Other values are based on own assumptions. The inverter
in province p, which is numerically identical to the peak sun capacity factor is chosen to be less than 1, since the power output
hours3. It is assumed that the yearly difference in horizontal vs. of PV modules is often less than the peak power rating due to
tilted irradiation in Indonesia is negligible, due to the location higher operating temperatures. In this way, the inverter can be
close to the equator. sized smaller which improves the operating efficiency and
The inverter capacity for a household in province p, Cinv,HH,a,p decreases the investment.
(kW), is chosen to be finv times the total peak power of the PV The battery self-discharge is assumed to be negligible due to
modules if this is larger than the peak load, otherwise the inverter the short term storage. For standalone PV systems the perfor-
capacity is chosen to be equal to the peak load: mance ratio, PRb, is assumed to be 55%, due to (i) mismatch
( between time of electricity use and production, (ii) shadow losses
f inv P PV;HH;a;p ; f inv P PV;HH;a;p 4 Lpeak;h
C inv;HH;a;p ¼ ð11Þ and (iii) due to higher PV module temperatures in these tropical
Lpeak;h ; f inv P PV ;HH;a;p rLpeak;h
areas. For PV systems in an isolated grid a performance ratio, PRa,
of 70% is assumed.
The battery capacity for a household Cbatt,HH,a (kWh) is deter-
mined as follows:
EPV ;HH;a β a 2.7. Costs
C batt;HH;a ¼ ð12Þ
ηwires ηctrl ηbatt DOD
To compare the costs of the different configurations the
where βa is the number of days of autonomy, ηwires, ηctrl and ηbatt levelized cost of electricity (LCOE) is used, which is calculated as
are the efficiencies of the wires, charge controller and battery follows:
charge, respectively, and DOD is the battery's depth of discharge.
P
N
AC p RV p
In the hybrid configuration, the diesel genset will mainly I PV ;p þ  ð1 þ DRÞ
ð1 þ DRÞn n

produce the electricity during the peak hours in the evening. In
this case, the genset can be sized to run efficiently during these
peak hours, the moments of lower power demand during the day
and night will be covered by the PV-battery system. The nominal
power of the diesel genset for a household, PDG,HH (kW), is
determined as follows:
Lpeak;h
P DG;HH ¼
DGBEP
where DGBEP is the best efficiency point (BEP) of the diesel genset.
The assumed values used to size the PV systems, batteries and
diesel generators are presented in Table 2. The efficiency of the
diesel generator is based on the range of values found in the
literature [14,29]. The depth of discharge (DOD) of the battery is
chosen to be 50%, which is lower compared with a DOD of 80%
which is often reported [14,30], but in this way the operating
conditions of the batteries are improved by minimising the time
between full charge which will increase the battery life time
3
A peak sun hour is defined as 1 kW/m2.
n¼1
LCOEp ¼ ð14Þ
P
N
EPV;p ð1  SDRÞn þ Ed;p
ð1 þ DRÞn
n¼1
where LCOEp is the total LCOE in province p, IPV is the initial
investment for the PV system, AC is the annual costs, RV is the
residual value, DR is the discount rate, EPV is the annual electricity
production by PV, SDR is the system degradation rate, Ed is the
ð13Þ
electricity production by diesel gensets and N is the project's
lifetime in years.
The costs of the PV system are included in the initial invest-
ment IPV, the costs of the other components (e.g. batteries,
inverters, fuel) are included in the annual costs AC, which include
maintenance and replacement costs as well. To determine the
LCOE a number of cost assumptions have been made, see Table 3.
Values related to the LCOE are in great extent the same as applied
in the previous study [15], which are in the range of values found
in literature [6]. The PV system price is slightly cheaper, because of
the cost reductions of PV modules in recent years [5]. Besides, the
cost of labour in Europe and the US is higher compared with
Indonesia, which will lower the balance of system (BOS) costs,
 A.J. Veldhuis, A.H.M.E. Reinders / Renewable and Sustainable Energy Reviews 52 (2015) 757–769
Fig. 8. The difference in LCOE between 100% diesel based electricity and hybrid PV system configuration a. The colours indicate the difference in costs in US$cents/kWh,
positive values show a lower LCOE for the PV-battery-diesel option. The circles indicate the potential amount of PV in GWh/year.
which justifies a slightly lower turn-key PV system price compared
to some other values found in literature [6,7]. The lifetime of a
battery is chosen to be five years [32].
The cost of diesel is based on the fuel costs from PLN per
province from 2010 [23] and adjusted for price increase4 and
exchange rate5, resulting in an average cost of diesel in Indonesia
of 0.91 $/l. The fuel costs from PLN include transportation costs
already; however these costs represent only the average fuel costs
in the areas in which PLN operates. Since these are mainly urban
areas, the fuel costs for rural areas would be larger; therefore extra
costs are taken into account for the transportation of fuel, as
shown in Table 3. For the hybrid system configuration a the
influence of transportation costs is assumed to be small, since a
micro-grid is only viable in more populated areas. In contrast, for
the stand-alone PV system configuration b, the influence is
assumed to be higher, since these off-grid systems are mainly
located in sparsely populated remote areas. The additional trans-
portation costs are modelled quite modest, in the study of Blum
et al. [14], factors of 2.0 and 2.73 are used for the medium and high
cost scenario, respectively.
3. Results
In Fig. 8 the differences in LCOE for PV system configuration a
compared with the LCOE based on diesel gensets only are shown
for each province. The figure also includes the potential annual
electricity generation by PV systems in GWh. For Indonesia in
total, 943 GWh net electricity can be produced by hybrid PV
systems inside local micro-grids. PV systems contribute 566 GWh
and the remaining 377 GWh is supplied by diesel generators. This
amount can be produced by a total capacity of 427 MWp PV
systems, 321 MW diesel gensets and 2.5 GWh battery capacity.
The total cost over the complete lifetime of 25 years is estimated
to be 7.4 billion USD6, which includes among others fuel and
battery replacements, but excludes expected increasing fuel prices
4
The average diesel price over 2010 has increased with a factor 1.70 compared
to the price in April 2014, source: www.indexmundi.com.
5
The exchange rate of 11,546 IDR/USD as of 30-4-2014 is used in this study,
source: www.xe.com.
6 7
Based on the LCOE calculation, which includes a discount rate and residual
value, the total cost is estimated to be 4.6 billion USD.
765
and decreasing PV system prices. This results in an average LCOE of
0.38 USD/kWh, which is lower than the average LCOE of 0.47
$/kWh for the reference scenario based on diesel gensets only.
As can be seen, for all provinces in Indonesia the LCOE of hybrid
PV systems is lower than that of conventional electricity produc-
tion by diesel gensets. Only in the province of West-Papua hybrid
PV is less attractive, mainly due to lower cost of diesel in this
province according to data from PLN [23], which does not
necessarily have to be true.
In Fig. 9 the differences in LCOE for stand-alone PV system
configuration b compared with the LCOE based on diesel gensets
only are shown for each province. The total annual net amount of
electricity which can be produced by PV-battery-systems in
Indonesia is 403 GWh, of which 339 GWh is cost-effective. The
total amount can be produced by a total of 389 MWp of PV and
6.0 GWh battery capacity. The total cost for these stand-alone PV
systems over the complete lifetime of 25 years is 5.9 billion USD7,
resulting in an average LCOE of 0.76 USD/kWh.
As can be seen, in most provinces in Indonesia the stand-alone
PV system is a cheaper solution compared with the LCOE of
electricity generation by diesel. In some provinces, namely South
and East Kalimantan, Maluku, West Sulawesi, Lampung, Jambi and
West-Papua it is not cost-effective yet, although the LCOE of diesel
is only significant lower (40.01 USD/kWh) in the provinces
Maluku (0.01 USD/kWh) and West-Papua (0.09 kWh). Besides, in
some other provinces, namely Banten, West and Central Kaliman-
tan, North and South Sumatra, Bangka-Belitung, Bengkulu and
Riau, the difference in LCOE is relative low compared with the
diesel-only scenario (o0.01 USD/kWh).
The LCOEs found in this study are compared with those found
in the study of Blum et al. [14]. Since they distinguished between
subsidised Indonesian and world fuel prices, we compare the
LCOEs based on the unsubsidised world fuel prices, because we are
evaluating the real cost of electricity.
The LCOE of both configurations are in line with the calculated
LCOE in the study of Blum et al., however the average LCOE of the
stand-alone PV option in this study is slightly higher. Besides, the
average LCOE of the hybrid PV system found in this study
corresponds to the lower range of LCOE for the hybrid solar
Based on the LCOE calculation, which includes a discount rate and residual
value, the total cost is estimated to be 3.8 billion USD.
766 A.J. Veldhuis, A.H.M.E. Reinders / Renewable and Sustainable Energy Reviews 52 (2015) 757–769
Fig. 9. The difference in LCOE between 100% diesel based electricity and stand-alone PV system configuration b. The colours indicate the difference in costs in US$cents/kWh,
positive values show a lower LCOE for the stand-alone PV system. The circles indicate the potential amount of PV in GWh/year.
PV/battery/diesel system. The LCOE of electricity generated by
diesel in the hybrid reference scenario corresponds to the lower
range of LCOE for diesel found by Blum et al. Although a larger
range in additional transportation costs is assumed in the study of
Blum et al., the LCOE of the stand-alone reference scenario in our
study matches with the higher range of LCOE. This can be
explained by the higher cost of fuel applied in this study, which
is 0.91 USD/l on average, compared with the approximated
0.77 USD/kWh8 applied by Blum et al., based on values from 2012.
From Figs. 8 and 9 it can be concluded that in particular the
provinces with the largest population without electricity – Papua
and ENT – are one the most attractive provinces for PV systems
based on their LCOE.
Based on the reference scenarios, 9.8 Gl and 5.2 Gl would be
necessary for the diesel generators to power all rural non-
electrified households in Indonesia over the complete lifetime of
25 years, for respectively the hybrid and stand-alone category.
Although diesel is still used in the hybrid configuration, the PV
system saves a significant amount of diesel, it is estimated that
over the complete lifetime 6.5 Gl can be saved.
To determine the CO2 reduction potential, the GHG emissions
of the different components have to be known. Based on a study of
Fleck and Huot, the total GHG emissions of diesel fuel is 3.15 kg
CO2-eq/l [33], which corresponds to the value of 1.27 kg CO2-eq/
kWh as applied in [34] when assuming an efficiency of the diesel
generator of 25%. For PV systems GHG emissions of 45 g CO2-eq/
kWh is assumed [35]. For the batteries the GHG emissions are
based on a study by García-Valverde et al., based on the energy
requirements to produce a lead-acid battery with new materials, a
battery corresponds to GHG emissions of 31.0 kg CO2-eq/kWh [34].
The emissions depend on a large number of assumptions and a
range of values can be found in literature, besides other factors
such as transportation and end-of-life usage of the various
components can be taken into account as well [8]. However, since
this is outside the scope of this paper, the above-mentioned values
will be used to get an idea of the CO2 reduction potential of the
modelled off-grid PV systems.
For the PV systems in the hybrid category, the total emissions
which can be saved during the complete lifetime of 25 years
correspond to 20.4 Mton CO2-eq, compared with the reference
8
0.61 €/litre in 2012.
scenario. For the PV systems in the stand-alone category, this
reduction potential is 16.2 Mton CO2-eq.
4. Sensitivity analysis
Since it is difficult to validate the results, a sensitivity analysis can
give insight in the importance of the model parameters and can as
such determine which parameters require additional research in
order to reduce the uncertainty of the model output. Besides, the
LCOE is sensitive to the input assumptions, therefore a sensitivity
analysis can give insight in these uncertainties as well [6].
An initial one-at-a-time sensitivity analysis was carried out to
indicate the most sensitive model parameters. This has been done
by varying the following parameters by 50% and 150%: PV module
lifetime (N), PV system investment costs (IPV), electricity demand
(EHH), peak load (Lpeak), additional transportation costs, battery's
depth of discharge (DOD), battery's days of autonomy, residual
value (RV), discount rate (DR), system degradation rate (SDR),
diesel genset lifetime, battery lifetime, performance ratio (PR),
annual costs (AC), battery investment costs, diesel genset invest-
ment costs, fixed O&M costs diesel, variable O&M costs diesel and
the fuel price. The parameters which influence the LCOE of the
hybrid or stand-alone system with more than 10% for either a
decrease or increase with 50% of its original value have been
selected for the further analysis. The following parameters are
found to be important: N, IPV, DOD, battery's days of autonomy,
battery lifetime, PR, battery investment costs, DR and the fuel cost.
The energy demand itself has minor influence on the LCOE of
stand-alone PV. For the hybrid configuration a decrease of energy
demand by 50%, increased the average LCOE with 9%, where an
increase of 50% in energy demand caused the LCOE to decrease by
2% only.
The input values for the sensitivity analysis are relative to the
values applied in this study as shown in Tables 2 and 3 and are
assumed to represent a realistic range of values for the particular
parameters in Indonesia. These variations are shown as percen-
tages in Tables 4 and 5.
4.1. Hybrid PV system
In Table 4 the results of the sensitivity analysis are shown for
PV system configuration a. The average LCOE for all provinces in
Indonesia is in the range of 336–668 $/MWh.
 A.J. Veldhuis, A.H.M.E. Reinders / Renewable and Sustainable Energy Reviews 52 (2015) 757–769 767

Table 4
Results Sensitivity Analysis LCOE PV System Configuration A.

Uncertain system parameters Lowa Moderate High Very high

PV module lifetime (50%) 424 (80%) 391 (100%) 381 (120%) 376
PV system investment costs (50%) 340 (80%) 365 (100%) 381 (150%) 423
Performance ratio (PR) (60%) 456 (90%) 394 (100%) 381 (120%) 363
Depth of discharge (DOD) (50%) 473 (80%) 404 (100%) 381 (150%) 351
Battery days of autonomy (100%) 381 (150%) 427 (200%) 473 (300%) 565
Battery lifetime (40%) 501 (60%) 434 (80%) 401 (100%) 381
Battery investment costs (50%) 336 (80%) 363 (100%) 381 (150%) 427
Fuel cost (100%) 381 (150%) 453 (200%) 525 (300%) 668
Discount rate (50%) 349 (80%) 368 (100%) 381 (150%) 418

a
The percentages are relative to the values used in this study as described before. The LCOE values are in $/MWh, the base case corresponds to an average LCOE
of 381 $/MWh.

Table 5
Results Sensitivity Analysis LCOE PV System Configuration B.

Uncertain system parameters Lowa Moderate High Very high

PV module lifetime (50%) 860 (80%) 779 (100%) 763 (120%) 753
PV system investment costs (50%) 671 (80%) 726 (100%) 763 (150%) 854
Performance ratio (PR) (60%) 923 (90%) 789 (100%) 763 (120%) 723
Depth of discharge (DOD) (50%) 1282 (80%) 893 (100%) 763 (150%) 589
Battery days of autonomy (100%) 763 (150%) 1022 (200%) 1282 (300%) 1,82
Battery lifetime (40%) 1440 (60%) 1063 (80%) 874 (100%) 763
Battery investment costs (50%) 503 (80%) 659 (100%) 763 (150%) 1022
Discount rate (50%) 684 (80%) 730 (100%) 763 (150%) 848

a
The percentages are relative to the values used in this study as described before. The LCOE values are in $/MWh, the base case corresponds to an average LCOE
of 763 $/MWh.

Changes in the fuel cost will affect the reference scenario as
well. For the normal reference scenario (diesel only), the average
LCOE is 469 $/MWh. If the fuel cost is increased with 50%, 100%
and 200%, this average LCOE will become 678, 886 and 1,303
$/MWh, respectively. Since the additional transportation costs are
modelled as an increase in the fuel costs, the LCOE values for an
increase in fuel cost apply to an equal increase in these transpor-
tation costs as well.
4.2. Stand-alone PV
In Table 5 the average LCOE for all provinces of Indonesia based
on the sensitivity analysis are shown for PV system configuration
b. The results show a LCOE in the range of 503–1802 $/MWh.
Changes in the fuel cost will affect the reference scenario only.
For the normal reference scenario (diesel only), the average LCOE
is 783 $/MWh. If the fuel cost is increased with 50%, 100% and
200%, this average LCOE will become 1138, 1494 and 2205 $/MWh,
respectively.
4.3. Overall
In Fig. 10 box charts show the range of possible LCOEs for the
different system configurations and reference scenarios for all
provinces, based on the values for the various input parameters. As
can be seen, the LCOE of the hybrid PV system are within the range
of 0.31–0.70 $/kWh. The stand-alone PV system shows a larger
range of LCOE, with a range of 0.46–1.83 $/kWh. This difference
can be mainly explained by the model's sensitivity to the battery
related parameters, which is more pronounced in the stand-alone
PV system.
The boxplots of the reference scenarios are relatively small,
because most model parameters do not influence the LCOE of the
reference models; the LCOE of the reference model is mainly
influenced by the cost of fuel. However, the large range of LCOE for
these reference scenarios indicates that the LCOE is highly sensi-
tive to the cost of fuel.
5. Discussion
The LCOE of the off-grid PV systems are compared with the
LCOE of diesel generators based on the costs for PLN, which
represents the real costs. However, since fuel subsidies exist, the
LCOE of the electricity generated by diesel is much less for
individual households or companies, making off-grid PV systems
financially less attractive for these groups.
The potential of off-grid PV systems is based on the rural
households which are categorised as lacking access electricity
based on the data from BPS. Another part of the rural households
is classified as “other”, indicating that they have access to
electricity which is not supplied by PLN. These households could
be served by stand-alone diesel gensets, in this case the potential
of off-grid PV systems could be higher.
For the determination of the LCOE of the hybrid PV systems –
which is assumed to be operated inside a micro-grid – the added
costs of the infrastructure of the grid itself have been neglected.
Including investments in the required infrastructure will increase
the final LCOE of this configuration. However, at the other hand,
this study neglects economies of scale as well, which could lower
the overall cost of the hybrid PV system since one large central PV
system could be used instead of interconnected PV systems per
household. Besides, from the sensitivity analysis can be seen that
even if the investments of the hybrid PV system would be 50%
higher, this would still be cheaper than the reference scenario
using diesel only. Moreover, for the reference scenario the grid
infrastructure is neglected as well, as such the difference in LCOE
will remain approximately the same.
Future studies could extend this study to include the modelling
of the power distribution line lengths. A methodology similar to
768 A.J. Veldhuis, A.H.M.E. Reinders / Renewable and Sustainable Energy Reviews 52 (2015) 757–769
Fig. 10. Box chart of the overall results of the sensitivity analysis showing the range of possible LCOEs for the different scenarios for all provinces.
that discussed in the study of Ruijven et al. [24] could be used for
this purpose.
The additional transportation costs for the fuel use is an
important variable for the determination of the LCOE of diesel.
The final LCOE of electricity generated by diesel is almost linearly
related to these fuel costs. The actual difference in LCOE in many
remote areas could be significant larger, because these transporta-
tion costs are modelled relatively modest.
Besides, the efficiency of diesel generators is assumed relatively
high with a range of 20–30%, especially for times in which the
generator is lightly loaded, resulting in lower efficiencies and
consequently higher operational costs.
The average battery storage size for the stand-alone PV systems
is found to be 5.4 times the daily load, which is in the range of 3–
10 of typical stand-alone systems [30]. Since the LCOE of the
stand-alone system is sensitive to the battery capacity, a larger
storage size will increase the LCOE significantly. The same counts
for the battery lifetime, which is modelled as five years, but which
could be considerably shorter due to the higher operating tem-
peratures in Indonesia [30]. Besides, no or wrong maintenance can
lead to shorter battery lifetimes. Since the battery related costs
form a large part of the final LCOE for stand-alone PV systems and
since the LCOE is sensitive to these costs, accurate lifetime
prediction is important for the estimation of the real life cycle
cost. Batteries are often found one of the key weak links in off-grid
energy systems [31]. However, the life time depends on a large
number of factors such as charge factor, discharge rate, time
between full charge and time at low state of charge (SOC) [31].
Therefore, it is recommended for real projects to model the stand-
alone system with a high time resolution to simulate the (dis)
charge rates and times of full charge and low SOC. In this study
there is strived to use most common values for each of the related
battery parameters in order to estimate the average capacity and
LCOE for such systems.
For simplicity reasons, some parameters which are intercon-
nected are not modelled as such. For instance, the battery lifetime
is dependent on the DOD: Increasing the DOD, can lead to a
reduction of the battery's lifetime. This relationship is not included
in the model, however a DOD less than 80% is assumed to avoid
lifetime reduction in lead-acid batteries [30].
Another example is the battery days of autonomy which is
modelled such that it influences the battery capacity only, how-
ever, in practice, the PV system should be designed slightly larger
to be able to charge the larger battery capacity.
Additional transport costs are modelled for the fuel consump-
tion only, since this is required during the whole project's lifetime
and these transportation costs will therefore have the highest
impact on the cost of fuel. These extra transportation costs will
play a role in the construction phase as well adding to the PV
system investment costs, however this impact would be smaller
compared with the costs of fuel transportation, evaluated over the
whole lifetime. The sensitivity analysis shows that the LCOE will
increase with 11% and 12% for an increase in PV system investment
costs with 50%, for the hybrid and stand-alone configuration,
respectively. So, dependent on the location, this could influence
the cost-effectiveness.
Overall, the sensitivity analysis shows that the LCOE is sensitive to
various parameters. It is recommended to determine the related
parameters for a particular project as accurate as possible in order to
be able to make a well-based decision, since in some provinces the
difference in LCOE between stand-alone PV and diesel gensets is small.
6. Conclusions
In this paper the potential and cost-effectiveness of off-grid PV
systems have been determined for Indonesia, such a detailed
analysis for Indonesia has not been found in the existing literature.
 A.J. Veldhuis, A.H.M.E. Reinders / Renewable and Sustainable Energy Reviews 52 (2015) 757–769
Since PV systems can play an important role to overcome chal-
lenges related to the electricity supply which Indonesia faces, this
study shows in which provinces these off-grid PV systems are
most cost-effective and as such give the most return on
investment.
In most rural parts of Indonesia the LCOE of off-grid PV systems
is lower compared with the electricity generation based on diesel
gensets with unsubsidised fuel costs. Especially, the hybrid PV
configuration shows a significant lower LCOE compared with the
diesel generated electricity for all provinces. The LCOE of electricity
of the stand-alone PV configuration is lower in 25 of the 32
provinces9 of Indonesia, of which eight provinces show a LCOE
which is at least 0.05 USD/kWh cheaper, including the two
provinces with the highest share of households without access to
electricity – East Nusa Tenggara and Papua. Besides, in 13 of the 32
provinces the LCOE is only 70.01 USD/kWh cheaper (eight pro-
vinces) or costlier (five provinces) than electricity generated by
diesel, indicating that PV systems are on the edge of becoming cost-
effective on large scale. Since it is expected that fuel prices will
increase and PV systems will become cheaper, it is expected that
this difference in LCOE will change in favour of the PV systems, so
that the PV systems will become cost-effective in all rural areas.
The potential of off-grid PV systems is projected to correspond
to an annual electricity supply of 969 GWh, of which 566 GWh is
generated with hybrid PV systems inside local grids and 403 GWh
in the form of stand-alone PV-battery systems. A total nominal
power of 816 MWp PV systems, 321 MW diesel gensets and
8.5 GWh battery capacity is estimated to be required to achieve
100% electrification in rural areas. To electrify all rural areas in
Indonesia by the combination of the proposed hybrid PV micro-
grids and stand-alone PV systems, the total cost over 25 years is
estimated to be roughly 13 billion USD.
On average the LCOE for hybrid PV is 0.38 USD/kWh, for the
stand-alone PV system this is 0.76 USD/kWh. Both configurations are
able to supply electricity cost-effectively in large parts of Indonesia.
Compared with the estimated LCOE of grid-connected PV for
rural areas ranging from 0.17 to 0.24 USD/kWh [15], off-grid PV
systems are significant more expensive.
The hybrid PV system shows the lowest LCOE with the smallest
range, but requires a certain population density in order to be
feasible. Overall, PV systems are cost-effective in large rural parts
in Indonesia already and it is highly likely that this will become
the case for all rural areas of Indonesia in the coming years. At the
same time, in regions where PV systems are already cost-effective,
they will become financially more attractive.
Acknowledgements
This work was made possible through the INDF Grant INDF10/
RI/12 ‘Joint Development of a Knowledge Centre on Solar Energy’.
The Indonesia Facility (INDF) was part of the Netherlands Enter-
prise Agency (former NL Agency) which acts under the Ministry of
Economic Affairs of the Dutch government.
References
[1] BPS. Population Census Data - Statistics Indonesia [Online].
[2] IEA, “IEA PVPS Task III Stand-alone PV systems,” 1999.
[3] Developing Renewables, “Renewable energy potential - Country report Indo-
nesia,” 2006.
[4] PT PLN, “PLN STATISTICS 2012,” ed. Jakarta: PLN, 2013.
[5] Sener C, Fthenakis V. Energy policy and financing options to achieve solar
energy grid penetration targets: accounting for external costs. Renew Sustain
Energy Rev 2014;32:854–68.
9
Excluding the province of Jakarta, which lacks rural areas
769
[6] Branker K, Pathak M, Pearce JM. A review of solar photovoltaic levelized cost of
electricity. Renew Sustain Energy Rev 2011;15:4470–82.
[7] Hernández-Moro J, Martínez-Duart J. Analytical model for solar PV and CSP
electricity costs: Present LCOE values and their future evolution. Renew
Sustain Energy Rev 2013;20:119–32.
[8] Cucchiella F, D’Adamo I. Estimation of the energetic and environmental
impacts of a roof-mounted building-integrated photovoltaic systems. Renew
Sustain Energy Rev 2012;16:5245–59.
[9] Akikur R, Saidur R, Ping H, Ullah K. Comparative study of stand-alone and
hybrid solar energy systems suitable for off-grid rural electrification: a review.
Renew Sustain Energy Rev 2013;27:738–52.
[10] Shaahid S, Al-Hadhrami L, Rahman M. Review of economic assessment of hybrid
photovoltaic-diesel-battery power systems for residential loads for different
provinces of Saudi Arabia. Renew Sustain Energy Rev 2014;31:174–81.
[11] C. Breyer, “Identifying off-grid Diesel-Grids on a global scale for economic
advantageous upgrading with PV and wind power,” In: Proceedings of the 5th
Alliance for Rural Electrification Workshop on Joint research activities
between the academia and the industry: bringing new opportunities to the
renewables sector, Frankfurt, September, 2012.
[12] Szabo S, Bodis K, Huld T, Moner-Girona M. Energy solutions in rural Africa:
mapping electrification costs of distributed solar and diesel generation versus
grid extension. Environ Res Lett 2011;6:034002.
[13] Szabó S, Bódis K, Huld T, Moner-Girona M. Sustainable energy planning:
leapfrogging the energy poverty gap in Africa. Renew Sustain Energy Rev
2013;28:500–9.
[14] Blum NU, Sryantoro Wakeling R, Schmidt TS. Rural electrification through
village grids—assessing the cost competitiveness of isolated renewable energy
technologies in Indonesia. Renew Sustain Energy Rev 2013;22:482–96.
[15] Veldhuis AJ, Reinders AHME. Reviewing the potential and cost-effectiveness of
grid-connected solar PV in indonesia on a provincial level. Renew Sustain
Energy Rev 2013;27:315–24.
[16] A. J. Veldhuis and A. H. M. E. Reinders, “Potential and cost-effectiveness of off-
grid PV systems in Indonesia – An evaluation on a provincial level,” In:
Proceedings of the Photovoltaic Specialist Conference (PVSC), 2014 IEEE 40th,
2014, pp. 3562–3567.
[17] Ministry of Energy and Mineral Resources. (2010). Indonesian Solar Energy
Utilization. Available: 〈http://www.esdm.go.id/news/54-articles/3354-indone
sian-solar-energy-utilization.html〉.
[18] NASA. Atmospheric Science Data Center [Online]. Available: 〈http://eosweb.
larc.nasa.gov/sse/〉.
[19] van Ruijven BJ, van Vuuren DP, de Vries BJ, Isaac M, van der Sluijs JP, Lucas PL,
et al. Model projections for household energy use in India. Energy Policy
2011;39:7747–61.
[20] Purwanto WW, Nugroho YS, Dalimi R, Soepardjo H, Wahid A, Supramono D,
et al. Indonesia energy outlook & statistics 2006 2006.
[21] IEG. The Welfare Impact of Rural Electrification: A Reassessment of the Costs
and Benefits. An IEG impact evaluation. Washington, D.C.: World Bank; 2008.
[22] Clean Development Mechanism. (2012), Annex 5 – Rationale for default
factors used in the proposed methodology SSC-I.L “Electrification of rural
communities using renewable energy”. Available: 〈https://cdm.unfccc.int/
Panels/ssc_wg/meetings/035/ssc_035_an05.pdf〉.
[23] PT PLN, “PLN STATISTICS 2010,” ed. Jakarta: PLN, 2011.
[24] van Ruijven BJ, Schers J, van Vuuren DP. Model-based scenarios for rural
electrification in developing countries. Energy 2012;38:386–97.
[25] Gibson J, Olivia S. The effect of infrastructure access and quality on non-farm
enterprises in rural Indonesia. World Development 2010;38:717–26.
[26] Center for International Earth Science Information Network – CIESIN –
Columbia University. “National aggregates of geospatial data collection:
population, landscape, and climate estimates, Version 3 (PLACE III)”. Palisades,
NY: NASA Socioeconomic Data and Applications Center (SEDAC); 2012.
[27] SWECO, “Assessing technology options for rural electrification. Guidelines for
project development,” 2009.
[28] Klokan Technologies GmbH. (19-03-2014). Bounding Box Tool: Metadata Enrich-
ment for Catalogue Records by Visually Selecting Geographic Coordinates (Latitude
/ Longitude) for Maps. Available: 〈http://boundingbox.klokantech.com/〉.
[29] Dufo-López R, Bernal-Agustín JL, Yusta-Loyo JM, Domínguez-Navarro JA,
Ramírez-Rosado IJ, Lujano J, et al. Multi-objective optimization minimising
cost and life cycle emissions of stand-alone PV–wind–diesel systems with
batteries storage. Appl Energy 2011;88:4033–41.
[30] Jossen A, Garche J, Sauer DU. Operation conditions of batteries in PV
applications. Sol Energy 2004;76:759–69.
[31] Svoboda V, Wenzl H, Kaiser R, Jossen A, Baring-Gould I, Manwell J, et al.
Operating conditions of batteries in off-grid renewable energy systems. Sol
Energy 2007;81:1409–25.
[32] Ghafoor A, Munir A. Design and economics analysis of an off-grid PV system
for household electrification. Renew Sustain Energy Rev 2015;42:496–502.
[33] Fleck B, Huot M. Comparative life-cycle assessment of a small wind turbine for
residential off-grid use. Renew Energy 2009;34:2688–96.
[34] García-Valverde R, Miguel C, Martínez-Béjar R, Urbina A. Life cycle assessment
study of a 4.2 kW p stand-alone photovoltaic system. Sol Energy 2009;83:1434–45.
[35] Fthenakis VM, Kim HC, Alsema E. Emissions from photovoltaic life cycles.
Environ Sci Technol 2008;42:2168–742008/03/01 2008;42.