You are on page 1of 10

Energy Policy 133 (2019) 110895

Contents lists available at ScienceDirect

Energy Policy
journal homepage: www.elsevier.com/locate/enpol

T
Cost minimization for fully renewable electricity systems: A Mauritius case
study
D. Timmonsa,∗, A.Z. Dhunnyb, K. Elaheec, B. Havumakia, M. Howellsd, A. Khoodaruthc,
A.K. Lema-Driscolla, M.R. Lollchundb, Y.K. Ramgolame, S.D.D.V. Rughooputhb, D. Surroopf
a
Department of Economics, University of Massachusetts Boston, USA
b
Department of Physics, Faculty of Science, The University of Mauritius, Reduit, Mauritius
c
Mechanical and Production Engineering Department, The University of Mauritius, Reduit, Mauritius
d
Department of Energy Technology, Royal Institute of Technology (KTH), Stockholm, Sweden
e
Electrical and Electronic Engineering Department, The University of Mauritius, Reduit, Mauritius
f
Chemical and Environmental Engineering, The University of Mauritius, Reduit, Mauritius

A R TICL E INFO A BSTR A CT

Keywords: Climate change is the world's most prominent environmental problem, and fossil-fuel combustion is its primary
Climate change cause. To set carbon reduction goals, policy makers require information on feasibility and cost of renewable
Decarbonization energy systems. In this study, we describe an economic approach to modeling a national electricity system based
Renewable energy economics entirely on renewable sources, using the island-nation of Mauritius as a case study.
Energy storage
Many renewable electricity studies consider levelized costs of electricity (LCOE), which represents average
OSeMOSYS
cost of electricity, but cost minimization requires equalizing marginal costs. With variable sources, marginal
costs change over time. Minimizing cost thus requires a model incorporating variability at high time resolution
(daily or hourly), with the objective of minimizing the levelized cost of electricity for an entire electricity system
(LCOES).
In Mauritius, the minimum-cost renewable electricity portfolio includes roughly equal proportions of solar,
wind, and biomass electricity, along with electricity storage. Policy issues include setting renewable energy
targets, selecting policy instruments to incentivize optimum renewable energy portfolios, incorporating energy
efficiency, determining appropriate discount rates, ensuring land availability, and accounting for non-cost
considerations. Many of the economic and policy issues identified apply universally, and methods demonstrated
in this study could be used anywhere in the world.

1. Introduction energy adoption, they often have little information on the feasibility,
composition, and cost of renewable energy systems, or on important
1.1. Motivation and objectives implications for issues such as land and water use.
In this study, we contribute to the literature on the economics of
One of the greatest environmental threats faced by the world today fully renewable electricity systems dominated by variable sources such
is the changing climate caused by the accumulation of heat-trapping as wind and solar electricity. Like fossil sources, costs of renewable
gases in the atmosphere. Fossil-fuel combustion is the largest producer sources have often been expressed as levelized cost of electricity
of greenhouse gases and must therefore be curtailed. To limit global (LCOE), representing average cost over the life of a resource. Unlike
warming to 1.5 °C, the Intergovernmental Panel on Climate Change fossil sources, the variability of many renewable sources implies that
(IPCC) estimates that world emissions must be reduced by about 45% the marginal cost of obtaining electricity varies with ambient condi-
from 2010 levels by 2030, and the world must achieve net-zero emis- tions. For example, generating more power from solar electricity on a
sions around midcentury (IPCC, 2018). To achieve such targets, more cloudy day requires more solar panels, hence increasing the marginal
economic studies of energy systems dominated by variable renewable cost of electricity at that time. LCOE for individual electricity sources
sources are needed. While policy makers must set targets for renewable obscures such critical information. As we demonstrate in section 2.1,

*
Corresponding author.
E-mail address: david.timmons@umb.edu (D. Timmons).

https://doi.org/10.1016/j.enpol.2019.110895

0301-4215/ © 2019 Elsevier Ltd. All rights reserved.


Received 28 March 2019; Received in revised form 18 July 2019; Accepted 20 July 2019
D. Timmons, et al. Energy Policy 133 (2019) 110895

accommodating varying marginal costs requires simultaneously mod- whole faces the same question.
eling all candidate generation technologies and electricity storage as an
integrated energy system, with the objective of minimizing LCOE for 1.3. Relevant literature
the energy system as a whole (LCOES). This implies that a region
transitioning to renewable electricity must have explicit knowledge of A number of previous studies use temporally explicit models to si-
all candidate sources, of their variability, and of their costs, as we de- mulate fully renewable electricity systems and estimate their costs
monstrate in a case study of Mauritius. We use current estimates of (Elliston et al., 2013; Pleßmann et al., 2014 Jacobson et al., 2015;
capital costs as an upper-bound estimate of future costs (since costs will Lenzen et al., 2016; Child et al., 2018). Diesendorf and Elliston (2018)
likely decline in the future). identify 44 studies at national and international levels, and using
Results from the case study also suggest policy questions that will broader criteria, Hansen et al. (2019) find 181 articles since 2004 re-
likely occur wherever renewable energy is developed. For example, if lated to 95% or greater renewable energy penetration. Europe is the
wind turbine development is limited for aesthetic or other reasons, the most studied region, with the United States and Australia also having a
resulting renewable energy portfolio may include a greater proportion number of studies. Most studies are national in scope (Hansen et al.,
of solar photovoltaic (PV) electricity, increasing land demand and re- 2019). Some of these studies use simulation models (proving a parti-
quired electricity storage (given the day-night variation of solar power). cular renewable portfolio is feasible), while others include optimization
Furthermore, technologies such as pumped-hydroelectric storage may (finding the least-cost renewable portfolio). Most optimization studies
require different institutions for water management. Such considera- find that a mixture of renewable resources minimizes cost because of
tions represent critical information for policy makers responsible for resource complementarity over time (Solomon et al., 2016). To our
implementing measures to mitigate climate change. Other major policy knowledge, the economic principles that support the use of simulation-
issues include what proportion of renewable energy to obtain by when optimization approaches have not been articulated prior to the current
(e.g. 100% of electricity by 2040 as modeled in this study); what dis- study.
count rate to use for capital investments; where to get land, which can Small islands are interesting cases for transitions to greater renew-
be a substantial requirement in a renewable energy system; and which able energy use. The components of energy systems in islands tend to be
policy instruments most effectively promote optimal energy portfolios. less complex than larger continental systems and islands may be used as
We adopt a cost-effectiveness approach, which seeks only to identify blueprints for transitions to sustainable energy systems (Child et al.,
the least expensive way to obtain electricity from renewable sources. 2017). Franzen et al. (2013) suggest that small islands provide an op-
We do not attempt to measure the benefit of controlling climate portunity for quicker transitions to renewable energy, given their lower
change—economically, socially, environmentally, or otherwise—or energy requirements. Many islands also have high costs for diesel
conduct benefit-cost analysis by comparing the cost of renewable generation, and thus have greater financial incentives to adopt re-
sources to fossil fuels. There are now many potential sources of re- newable electricity than mainland systems.
newable energy; our question is how to combine sources in a system A critical decision for renewable electricity modeling is the degree
that minimizes the total cost of electricity for society, with little or no of temporal resolution. Hourly models are often preferred for accuracy,
reliance on fossil fuels. Mauritius provides a useful case study to de- typically for periods of one year (Hansen et al., 2019). For an optimi-
monstrate cost minimization for renewable electricity, and represents zation model, the objective is to identify the set of feasible renewable
an example of an electricity transition that must ultimately occur in resources that always meets demand and has the minimum total cost.
more complex electric grids around the world. Optimizing an annual model with 8760 hourly time periods is a com-
putational challenge; several approaches have been used to address this
1.2. Mauritius case study problem, and to optimize periods longer than one year (Pfenninger,
2017). In this study we balance the accuracy and tractability criteria by
Mauritius is a small island developing state (SIDS) in the southwest using day and night time slices (730 per year) rather than hours. We
Indian Ocean with a population of approximately 1.3 million people discuss possible limitations for results in section 3.1.
(Statistics Statistics Mauritius, 2017). Island area is about 2040 km2, Another important modeling question is what prices to use for fu-
including the main island of Mauritius and Rodrigues. The country ture renewable energy technologies. Prices are typically assumed to fall
currently has a goal to obtain 35% of electricity from renewable sources in the future, but how much and when are uncertain. For capital cost of
by 2025 (Ministry of Renewable Energy and Public Utilities, 2009). In the core utility-scale photovoltaic (PV) technology, Elliston et al.
this study we describe a hypothetical 100% renewable portfolio for the (2013) use a 2030 cost range of $1.42-$1.87/W, Pleßmann et al. (2014)
year 2040 to illustrate its composition and cost for policy makers. use a 2020 cost of $1.13/W, Jacobson et al. (2015) use $1.70/W, and
Electricity distribution, transmission, and supply are solely con- Child et al. (2018) assume cost reductions from $1.15/W in 2015 to
trolled by the Central Electricity Board (CEB). In 2016, electricity $0.35/W in 2050. There is clearly no single correct price for solar PV, or
generation was dominated by fossil fuels, with 36.6% of generation any reliable price forecast for the future. Assessing renewable energy
from oil products and 41.6% from coal. Renewables accounted for based on current prices demonstrates to policy makers that renewable
21.8% of total electricity production, with 16.3% from sugarcane ba- energy is feasible without depending on uncertain future price reduc-
gasse (available only during the 6-month crop season), 3.3% hydro- tions. We thus use current technology prices, e.g. $1.24/W for PV
electricity, 1% solar electricity, 0.6% wind electricity, and 0.6% landfill (Lazard, 2017).
gas (Statistics Statistics Mauritius, 2017). Most existing world electricity systems that use primarily renewable
Mauritius is a useful location to study fully renewable electricity. sources rely on hydroelectricity (Diesendorf and Elliston, 2018); few
Being an island, the electricity grid is isolated from other countries, working examples exist of systems based on variable wind and solar.
simplifying modeling. Mauritius already has most of the major renew- Godina et al. (2015) describe the electricity system in El Hierro, one of
able electricity sources (except geothermal) on a limited scale, in- the Canary Islands of Spain, which utilizes mostly wind power and
cluding wind, solar, hydroelectricity, biomass from sugarcane bagasse, pumped-hydroelectric storage. Today, this is one of the world's few
landfill gas to electricity, and bioethanol from sugarcane (Statistics working examples of an electricity grid powered almost entirely from
Mauritius, 2017). Island land area is finite and land-use questions are variable sources.
important. And since Mauritius produces no fossil fuels, economic and In Mauritius, Khoodaruth et al. (2017) evaluate various routes to a
security issues related to energy imports are prominent (Timilsina and carbon-free energy future. Surroop and Raghoo (2017; 2018), and
Shah, 2016; Wolf et al., 2016). Given feasible but finite and costly re- Elahee (2010) survey the current energy landscape, supporting a larger
newable resources, which ones should be developed? The world as a role for renewables. Recent papers on Mauritius provide useful

2
D. Timmons, et al. Energy Policy 133 (2019) 110895

background on peak demand (Elahee, 2011; Adam et al., 2017), daily last), and (2) with ambient energy availability, which depends on ever-
insolation (Ramgolam and Soyjaudah, 2015), annual precipitation changing weather. This changing marginal cost is the most prominent
(Staub et al., 2014), and hydropower potential (Sahai, 2004; Beeharry challenge in renewable energy economics. There are nearly infinite
et al., 2009; Elahee, 2013). A recent LCOE analysis provides a useful combinations of possible energy sources and ambient conditions. A
first approximation of costs to achieve 35% of electricity production computer model that captures daily or hourly variation in ambient
from renewable sources, but the study does not explicitly model elec- energy availability is thus used to capture changing marginal costs, and
tricity production variability or include the cost of electricity storage to identify an optimum renewable energy system.
(Shea and Ramgolam, 2019). An earlier study of the Mauritian island of Within an electricity system, average costs can vary greatly. For
Rodrigues also considered only renewable-electricity LCOE and possible example, a bioethanol power plant used at times of low ambient energy
electricity price paths (Weisser, 2004). may run only a few hours per year, and because of low capacity factor
have high LCOE. Yet, this dispatchable electricity source may be a key
2. Methods and data enabling technology for multiple variable resources in an electricity
system. Minimizing the levelized cost of an integrated electricity system
2.1. Economics of cost minimization for renewable energy (LCOES) is therefore the only relevant objective. And minimizing LCOES
ensures the equality of LCOET at critical times (Timmons, 2018), so a
If climate change mitigation requires obtaining virtually all energy model need not calculate LCOET explicitly.
from renewable sources, then the primary economic question is how to Economic theory thus proves that a specific modeling strategy is
minimize total energy-system cost. To determine this, we must first needed to identify a minimum-cost renewable energy portfolio: we
express one-time capital and on-going operating expenses in compar- must model temporal variation in energy sources and energy demand,
able terms. Levelized Cost of Electricity (LCOE) is often used to express we must consider all candidate energy sources and energy storage si-
total cost per unit of electricity. Assuming for simplicity that annual multaneously, and our objective must be to minimize LCOES, the
operating costs and electricity production are constant for the life of a average cost of energy for an entire energy system.
project, one formula for LCOE is:
2.2. OSeMOSYS model
C+K
LCOE =
annual energy production (1)
2.2.1. OSeMOSYS background and purpose
where C is annual fixed and variable operating cost, K is capital cost, The Open Source energy Modeling System (OSeMOSYS.org) was
and α represents a capital recovery factor, which amortizes the capital developed with several goals in mind: removing financial barriers for
expense over the life of a project: initial uptake and deployment of an optimizing model, allowing a
network of researchers to collaborate without requiring license fees
r (1 + r )T
= (ETSAP, 2018); providing a code base that can be reformulated with
(1 + r )T 1 (2)
new representations of an evolving energy system (Howells et al.,
where r is a discount rate and T is the life of the project. Note that the α 2011); injecting ‘peer-reviewed’ transparency allowing for a full audit
factor—and discount rate r in particular—reflect the crucial compro- by policy makers and experimental repeatability by scientists
mise between capital and operating cost. This affects, for example, the (Morrison, 2018); and finally, supporting and enlarging the community
relative desirability of a wind turbine (with high investment and low of energy-systems analysts (Gardumi et al., 2018) by providing infra-
operating cost), as compared to a gas turbine (with lower investment structure that is easily accessible and replicated. The resulting char-
and on-going fuel costs). acteristics of OSeMOSYS are critical for the current study, for example,
The economic objective is to minimize cost for the desired quantity the flexibility to adapt formulations to assess variability, and the
of energy, subject to constraints on availability of energy sources. At the transparency needed to inform Mauritian policy dialogue about re-
minimum cost, the equimarginal principle applies: the marginal cost of newable energy. Code, documentation, discussion fora, and teaching
obtaining another unit of energy must be equal for each energy source material are freely available online (KTH - DESA, 2018). OSeMOSYS
used (or only one energy source is used). This is a substitution principle; has been applied to a wide array of questions, and it provides the op-
if a less expensive energy substitute is available, total cost can be fur- timization routines of the world's most used energy model, LEAP (Long-
ther reduced. At the minimum total cost, no cost-reducing substitutions range Energy Alternatives Planning system; Heaps, 2016).
are available, so all marginal costs must be equal.
Unlike dispatchable generation sources, with variable sources the 2.2.2. Mauritius renewable electricity model in OSeMOSYS
marginal cost of obtaining energy changes through time with ambient We use OSeMOSYS to model a fully renewable electricity grid for
conditions. For example, capturing a given quantity of wind energy Mauritius in the year 2040, when existing infrastructure will be fully
requires fewer turbines on a windier day. LCOE—based on average depreciated (or nearly so). The model thus represents a tabula rasa
generation over a year (equation (1))—does not capture the marginal electricity system. Though the model does not describe how that system
cost of delivering electricity at a specific time. For this we modify should be developed over time, the existence of an optimal solution
equation (1) slightly: suggests that there is also an optimal path.
The OSeMOSYS linear programming solver identifies the cost-
C+K
LCOET = minimizing quantities of generating and storage resources for which
(time period energy production )(time periods per year ) (3)
electricity supply equals or exceeds demand in each time period, subject
LCOET captures the marginal cost of electricity for the ambient to constraints on land, fuels, etc. as described below. For this one-year
conditions at a particular time. Timmons (2018) demonstrates that the model, we annualize all capital costs as in equation (1); costs thus re-
equimarginal principle holds for variable renewable energy sources, but flect a representative year in the life of an electricity system. For each
only for LCOET, and only at critical times when the supply of energy is generation source we include both fixed operating costs per kW that are
limited. LCOET when a demand constraint is binding—when supply just incurred regardless of production level, and variable costs per kWh that
equals demand—determines the optimum composition and cost of a depend on use, such as fuel.
renewable energy portfolio. The discount rate (r in equation (1)) reflects assumptions about risk,
For variable renewable energy, marginal cost thus changes in two about financing with debt vs. equity, and about private vs. public fi-
ways: (1) in the familiar way, as the quantity supplied increases (e.g. nancing. For new generating sources in the United States, the Energy
since the next unit of hydroelectricity may be more expensive than the Information Administration uses an after-tax weighted cost of capital of

3
D. Timmons, et al. Energy Policy 133 (2019) 110895

7.0% (EIA, 2019), while Lazard (2017) uses a before-tax rate of 9.8%. demands and extreme weather conditions, we include an additional
Much higher rates can be found in an African context (Ekholm et al., hypothetical model day (day 366). We define this day as having the
2013). Discount rates have a large effect on renewable energy costs, estimated 2040 peak demand occurring over the entire day and night
since most technologies are capital intensive, but we find that the dis- time slices. In addition, this day is assumed to have no solar, wind, or
count rate has little effect on the optimum mix of sources, with most net hydroelectricity. The model solution must provide sufficient dis-
sources being similarly capital intensive. We assume an 8% rate in the patchable capacity and energy storage to accommodate these hy-
model; achieving this rate in Mauritius may require public-sector in- pothetical extreme conditions for at least one day (or similar but less
volvement in renewable energy investments. extreme conditions for multiple days). For scenario 5, we remove day
The model includes time slices for each day and night of the year,1 366, to estimate the potential benefit of meeting peak demand and
or 730 time slices (365 x 2), plus two additional time slices (discussed extreme weather conditions with only demand-side management.
below). We include separate time slices for days and nights, antici-
pating that solar electricity will be a significant source on this tropical 2.3. Electricity and cost data
island, and that electricity storage and/or non-solar generation will be
needed at night. Time slices for days represent the hours 8:00–16:00, In the following sections we describe potential Mauritian renewable
when insolation is potentially sufficient to meet demand. Using 8 and electricity sources and their costs. Costs include only generation capa-
16-h time slices is a simplification of reality used for modeling tract- city and operations, not other new infrastructure which may be needed,
ability, but it ignores variability within time slices. We discuss effects such as roads, grid extensions, etc. We do not include potential future
on results in section 3.1. technologies for which there are not yet reliable production and cost
Available generating technologies include solar PV, wind turbines, data, including wave energy, tidal energy, and offshore wind turbines
hydroelectric turbines, a plant burning municipal solid waste, bioe- suitable for the deep waters around Mauritius. A recent report on
thanol power plants (a renewable equivalent of existing diesel gen- geothermal energy in Mauritius finds it unlikely (ELC Electroconsult,
eration), and solid-fuel plants burning sugarcane bagasse plus a variety 2015), so this is also excluded. However, should any of these sources
of possible biomass fuels. Assumed performance and constraints for prove to have costs or characteristics that warrant their use, this would
each potential generating resource are described in the following sec- reduce the cost of renewable electricity that we estimate. Unlike fossil
tions. fuel prices, renewable energy prices should decrease over time, and
Given data limitations, performance data for the variable renewable today's costs are thus an upper bound on future costs (while today's
resources are drawn from various sources and years and combined into fossil fuel costs are a lower bound on future costs).
one composite year. Generating performance enters the model as ca- To our knowledge, there is no single source for ground observations
pacity factors (CF) for each time slice in the model: of ambient energy conditions in Mauritius. Satellite data can be gen-
erated for specific time intervals and contain a wide range of environ-
MWht mental data; however, satellite-based data have their own sources of
CF =
ht (4) error. Ernst et al. (2016) and Ramgolam and Soyjaudah (2015) have
reported that satellite data overestimate conditions actually measured
where MWht is electricity production estimated for 1 MW of capacity in at ground level, with overestimations varying between 9% and 40%. In
a particular time slice, and ht is hours in the time slice. The capacity order to rely exclusively on ground observations, we combine data from
factor thus reflects how much electricity would be obtained from a different sources and years, as described below. For example, solar data
particular generating source in a particular time slice, if OSeMOSYS are from 2017 to 2018, wind data are from 2010 to 2011, river-flow
includes that generating source in the solution. Solar and wind capacity data are from 2009 to 2010, and demand data are from 2015. We
factors are calculated from observed ambient conditions in each model construct a composite year with consistent days, i.e. solar, wind, hydro,
time period. Some authors have observed differences between calcu- and demand conditions are all observed on the same calendar days,
lated and actual capacity factors (Jefferson, 2012); should any such albeit in different years.
discrepancies arise from operational issues, they are not reflected in the The resulting data capture seasonal relationships between important
model. variables such as sun and wind, but daily covariances between the
OSeMOSYS also models electricity storage investments and perfor- sources are missing. A number of studies have investigated correlations
mance. For each time slice where electricity generation exceeds de- between environmental parameters such as wind and solar irradiance.
mand, electricity can be stored, subject to available charging capacity Bett and Thornton (2016) investigated the daily covariability of wind
and the available quantity of electricity storage. Similarly, electricity and solar irradiance in Britain. The authors report that irradiance is
can be removed from storage if needed to meet demand in a subsequent weakly anticorrelated with wind speed throughout the year and there is
time slice. Like generating capacity, OSeMOSYS chooses storage to a weak tendency for windy days to be cloudier (at least in Britain). Any
ensure that electricity supply equals or exceeds demand in each time such weak daily correlations are lost in our model, which would have
slice and to minimize cost. In this model we include pumped-hydro- the effect of increasing our estimated energy storage requirements and
electric storage, the least-cost option for long-term applications, as well costs.
as lithium batteries. Cost of storage charge-discharge capacity in MW is
specified separately from cost of stored energy potential in MWh. 2.3.1. Solar electricity
A limitation of the modeling strategy is that the time-slice incre- Mauritius already uses solar PV electricity on a limited scale. An
ments (day and night) are too long to capture peak demand, which initial program in 2010 and 2011 facilitated the connection of 4.94 MW
occurs over much shorter intervals of minutes or hours. In addition, the of capacity. The first 15 MW PV park of Mauritius was commissioned in
model data reflect only one composite year of weather conditions, and 2014. Net metering schemes introduced in 2015, 2016 and 2017 en-
other years will have different conditions, potentially including pro- abled the connection of 17 MW of small and medium-scale PV dis-
blematic days when little electricity is available from any of the vari- tributed generators. A home solar project launched by the CEB in 2017
able sources. For sufficient dispatchable capacity to meet both peak allows 2000 PV connections of 1 kW each for five years. Aided by these
policies, PV installed capacity is almost 40 MW, or about 4.5% of in-
1
Some other simulation studies model variability at higher resolution, for stalled capacity in Mauritius.
example in 1-h or 5-min time slices. However for a cost optimization model as According to ground-based measurement data obtained using c-Si
used in this study, a suitable compromise must be made between time resolu- reference cells, Ramgolam and Soyjaudah (2014) report spatial varia-
tion, the number of technologies considered, and computing power and time. tion of average daily global horizontal radiation between 3.91 and

4
D. Timmons, et al. Energy Policy 133 (2019) 110895

5.09 kW h/m2 and sky clearness index variation between 0.31 and 0.61 slated for development of a new wind farm. The model turbine is a
across the island. Under clear sky conditions GHI values of above Suzlon 2.1 MW turbine, which is proposed for the Plaine Sophie project,
7.5 kW h/m2 are also reported. and is thought to have favorable characteristics including cyclone re-
With the financial support of the Global Environment Facility and sistance (SRM, 2012). Data observed at 60 m are adjusted for the Suzlon
United Nations Development Programme, research-grade equipment 90 m hub height. Time-slice capacity factors range from 0.0 to 1.0, with
was installed across the island in 2016 to measure global horizontal a mean value of 0.29. Capital cost and operating costs for wind turbines
irradiance (GHI) and plane of array (POA) irradiance and hence de- are shown in Table 1. Capital cost is based on an announced cost for the
velop solar maps for Mauritius. Temporal variations of radiation can be Plaine Sophie project (François, 2014), adjusted to 2017 dollars and
obtained from a GIS solar map website for Mauritius (Solarmap, 2017), further adjusted to reflect turbine price decreases since 2014 based on
and are used in this study. The northern part of Mauritius receives the Lazard (2017).
highest level of insolation, followed by the west, south, and center of
the island, while the eastern part receives the least insolation. Ac- 2.3.3. Hydroelectricity and electricity storage
cording to extended research performed by Ramgolam (2018) mean Hydropower was the original electricity source in Mauritius, and
annual insolation levels vary from 1237 kW h/m2 to 1901 kW h/m2 still contributes 3.3% of supply (Statistics Mauritius, 2017). But po-
over the island. These values correspond to PV capacity factors of be- tential for additional development is clearly limited, as expressed by
tween 0.141 and 0.217. Elahee, who writes that “… more than 90% of the technical potential of
Solar data for the OSeMOSYS model is based on observations from hydroelectricity has been fully tapped,” and that “large scale hydro has
August 2017 to July 2018 (Solarmap, 2017), for stations at Goodlands no potential in Mauritius” (Elahee, 2013, p. 922). Yet Elahee (2013) is
(north), La Mivoie (west), and Souillac (south). Daily capacity factors optimistic about additional small-scale hydroelectricity; we assume that
range from 0.03 to 0.29, with mean capacity factors of 0.173, 0.169, a 12% increase in net hydroelectricity is feasible (Elahee, 2013), with
0.147, for the north, south, and west respectively. Table 1 provides new capacity in pico- or micro-hydro plants.
capital and operating cost assumptions for PV panels in ground-mount, Pumped-hydroelectric storage is different, since it entails little or no
utility-scale plants; for PV floating on inland water reservoirs; and for water consumption. While no sites have yet been developed in
commercial and industrial rooftop collectors. All costs are in US dollars Mauritius, Elahee (2013) notes positive prospects for future im-
and based on current US prices, which may be an underestimate of plementation, given existing water resources and topography.
prices in Mauritius. Havumaki (2018) estimates costs and technical feasibility for two
prospective pumped-hydroelectric storage sites. Each scenario com-
2.3.2. Wind electricity bines existing reservoir storage with a hypothetical new reservoir. As-
There is currently one commercial-scale wind farm of 9.35 MW at suming use of just solar PV and hydroelectricity, Havumaki finds that a
Plaine des Roches on the main island of Mauritius, and several addi- single pumped-hydroelectric storage installation in the area of Ta-
tional turbines on the island of Rodrigues. Dhunny et al. (2014, 2015) marind Falls could provide all required electricity storage, including all
develop and test different probability densities for selected locations in nighttime electricity and approximately 1,300 GW h out of the total
Mauritius. The island's relatively complex orography requires modeling 2,982 GW h consumed in the scenario year. It is clear from Havumaki
with computational fluid dynamics (CFD), along with in-situ measure- that significant storage potential resides in the existing reservoir net-
ments of wind speed (Dhunny et al., 2017). WindSim CFD software was work, given the favorable elevation profile of these reservoirs in the
employed to generate yearly wind and power maps for Mauritius at a Mauritian Central Plateau.
resolution of 500 m, identifying the best locations for wind harvesting. In OSeMOSYS we model hydroelectric water as an aggregate flow
Data for the OSeMOSYS model are from measured wind speed in into a single reservoir. River flows are based on 2009–2010 observed
2009–2010 at the Plaine Sophie site (Suzlon-Padgreen, 2012), a site data from a representative gauging station on the Central Plateau

Table 1
Cost parameters for renewable electricity model.
Capital cost, USD Fixed operating cost, USD/kW Mean annual capacity factor

Photovoltaic panels, ground-mount utility scale, per kW $1238 $10.50 0.173i


Photovoltaic panels, floating, per kW $1547a $10.50g 0.163j
Photovoltaic panels, roof mount, per kW $2875 $17.50 0.147k
Wind turbines, per kW $2066 $35.00 0.289
Hydroelectric and pumped storage plants, per kW $2743b $57.14b nal
Pumped storage reservoir, per potential kWh $12c na na
Lithium battery AC charge/discharge power, per kW $107d $2.30 nal
Lithium battery energy storage, per potential kWh $376d $5.17 na
Solid fuel plants, per kW $2850 $50.00 nal
Bioethanol power plants, per kW $650e $10.00 nal
Municipal solid waste plant, per kW $9250f $50.00h 0.950

Costs from Lazard (2017) except as noted.


a
Assumed 25% greater than ground mount, based on Sahu et al. (2016).
b
Kantor Management Consultants (2008), adjusted for inflation.
c
Havumaki (2018).
d
Lazard (2018b).
e
Assumed same cost as diesel internal-combustion generator.
f
Carnegie Clean Energy (2017)
g
Assumed same as ground mount, based on Sahu et al. (2016).
h
Assumed same as solid fuel plant.
i
Based on best site monitored at Goodlands (Solarmap, 2017).
j
Mean of 3 sites monitored (Solarmap, 2017).
k
Assumed 90% of 3-site mean (Solarmap, 2017), due to potential shading.
l
Capacity factors for dispatchable sources determined within model.

5
D. Timmons, et al. Energy Policy 133 (2019) 110895

(station J04; Ministry of Energy and Public Utilities, 2011). Total hy- Finally, sugarcane is a useful energy crop both because its waste can
droelectricity from 2015 is assigned to time slices based on the pro- be burned as biomass, and because its molasses and sugar products can
portion of annual flow in each time slice. be used to make bioethanol fuel. Costs of bioethanol are included in the
Excess electricity may be used to pump water from a lower reservoir OSeMOSYS model, and are shown in Table 2. Bioethanol power plants
to an upper one. Reservoirs are thus charged by both river flow and are a dispatchable electricity source, and unlike solid-fuel plants, they
storage pumps. Electricity may be dispatched immediately or in a can be started and stopped quickly to match changes in electricity de-
subsequent time slice, as determined in the model, and depending on mand and/or changes in variable supply. Either turbines or internal
charge/discharge capacity and reservoir volume, also determined in the combustion engines could be used in bioethanol power plants; more
model. Since large reservoirs already exist at high elevation in research on optimal plant configurations is needed.
Mauritius, we assume construction of only lower reservoirs. We also
assume that a single turbine-pump-generator set is used to both charge 2.3.6. Electricity total demand and peak demand
and discharge reservoirs, using a cost of USD 2,743/kW, based on an Daily electricity demand data for the 2015 data year are provided
earlier consultant's report for hydroelectricity in Mauritius (Kantor by the Central Electricity Board (CEB). We estimate day and night time-
Management Consultants, 2008), and adjusted to 2017 dollars. The slice demand from typical hourly data also from the CEB. To reflect
International Renewable Energy Agency (IRENA, 2018) notes that a possible demand in the 2040 model year, we use demand growth as-
typical range for hydroelectricity capital costs is USD 500–4500/kW, sumptions from the Mauritius Renewable Energy Agency (MARENA) of
with a global weighted mean cost of USD 1,558/kW in 2017. We thus approximately 4.1% annual growth until 2020, 1.8% from 2021 to
use a relatively high cost assumption for constructing new hydroelec- 2025, 1.6% from 2026 to 2030, and subsequently 1.5% annual growth.
tricity and pumped-hydroelectric storage facilities in Mauritius; more This pattern reflects an assumption of increasing energy efficiency over
accurate estimates require engineering evaluation. Reservoir cost is time, though in this study we do not model efficiency as a specific
based on recent construction costs for the Bagatelle reservoir in technology option. Peak demand from 2015 is scaled to 2040 to
Mauritius (Havumaki, 2018). All capital and operating cost estimates maintain the same ratio of peak to total demand. Peak and total elec-
are shown in Table 1. tricity demand assumptions are shown in Table 3.
We include lithium batteries as another electricity storage option.
Compared to pumped hydroelectric storage, batteries have much lower 2.4. Scenarios
power charge-discharge costs and much higher costs for energy storage
potential (Table 1), making them more economical when only short- To test the sensitivity of results to different assumptions, we model
term storage is required. five different scenarios: (1) a base scenario; (2) a scenario where only
80% of electricity generation is renewable, with the balance from fossil
2.3.4. Waste to electricity fuels; (3) a low-wind scenario, reflecting possible aesthetic or other
There is currently a 3.3 MW landfill gas generating plant in concerns with further wind power development; (4) a scenario where
Mauritius, but at this writing a new plant burning municipal solid waste no land on this small island is used for solar arrays, relying instead on
appears likely. We include a waste-to-electricity plant of 30 MW and more-expensive rooftop panels and floating panels on reservoirs; and
force this plant into the model solution. Costs are high compared to (5) a demand-side management (DSM) approach to meeting peak de-
other generating sources, as shown in Table 1, but the primary purpose mand and atypical weather conditions (with little solar, wind, or hydro
of the plant is waste disposal. We omit landfill gas, since a shift to solid availability). All scenarios include the total costs of electricity genera-
waste combustion will eventually curtail landfill gas production. tion, but not other utility costs such as transmission and distribution.

2.3.5. Biomass 3. Results and discussion


Mauritius currently has several biomass electricity plants, all col-
located with sugar processing facilities. During the six-month sugar The cost-minimizing solution for each scenario is shown in Table 4.
harvest season (approximately July–December), these plants burn pri- In general, we find that optimal renewable portfolios combine all the
marily sugarcane bagasse (crushed cane residue), and to a lesser extent, available resources to different extents, including solar, wind, hydro-
sugarcane tops and leaves, and coal. During the six-month non-harvest electricity, biomass, and electricity storage. The variable nature of solar
season, most of these plants burn coal exclusively. Unlike variable wind and wind electricity is mitigated partly by a diversity of sources, partly
and solar, biomass energy is a dispatchable source, and can serve as an by coupling with dispatchable biomass and hydroelectricity, and fur-
important complement to variable sources. ther mitigated with electricity storage. Though none of the solutions
There are several options to avoid burning coal in biomass/coal relies on bioethanol power plants for a large portion of electricity,
plants (Lema-Driscoll, 2018). These include shutting down in the non- bioethanol is used to meet demand when other sources are unavailable.
harvest season, storing bagasse during the harvest season for use in the
off season, producing additional biomass crops or timber residues to 3.1. Scenario 1: base
burn in the off season, importing biomass from other countries, or a
combination of these options. Our OSeMOSYS model includes para- In the base scenario, wind and solar PV supply the same amounts of
meter values and temporal availability for all of these options (Table 2). electricity, at 27% of total electricity each. As in the current Mauritian
Purchase et al. (2014) explored options for storing bagasse to make electricity grid, solid-fuel plants play a prominent role with 35% of total
it available in the non-harvest season. Economic issues include costs for electricity produced, but in addition to sugarcane bagasse they now
placing bagasse in storage and later removing it, as well as losses due to burn a variety of biomass feedstocks to replace coal. In particular, 15%
degradation in storage. We follow Purchase et al. (2014) and their re- of total electricity is provided by imported wood chips. Bagasse storage
commended methods for storing bagasse in open-air piles. We assume is not utilized due to costs being greater than for imported wood chips.
that 25% of the net bagasse energy is lost in storage. Power plants using bioethanol provide 334 MW of installed capacity
We include miscanthus as a possible off-season biomass crop. (17% of total capacity) but provide only 1.5% of total electricity: these
Potential cropland includes marginal agricultural lands that once plants seldom operate, but are occasionally needed when variable
served the sugar industry and other marginal agricultural land. In ad- supply plus pumped-hydroelectric storage is insufficient to meet de-
dition, we consider a portion of net forest growth as a possible biomass mand.
source, and consider the option of importing biomass to burn at the In the base results, hydropower provides 3% of net electricity, si-
biomass/bagasse plants. Values used are shown in Table 2. milar to its current production. Capacity grows from 56 MW currently

6
D. Timmons, et al. Energy Policy 133 (2019) 110895

Table 2
Potential fuels for electricity generation in Mauritius.a.
assumed moisture gross energy at moisture content, fuel cost per kWh generated maximum electrical energy production,
content MJ/kg GWhc

Sugarcane bagasse, as produced 50% 8.2 nab 452.6


Sugarcane bagasse, stored 50% 6.2 $0.067 339.5
Sugarcane tops and leaves 14% 13.6 $0.009 147.4
Miscanthus biomass crop, domestic 15% 15.3 $0.069 77.7
Wood chips, domestic 30% 11.8 $0.056 123.3
Wood chips, imported 30% 11.8 $0.140 na
Coal na 25.1 $0.065 na
Bioethanol na 26.9 $0.324 na
Diesel fuel na 46.9 $0.057 na

a
Numerous assumptions are included in these estimates; a full set of assumptions is available on request from the corresponding author.
b
Assumes bagasse is a costless byproduct of sugar refining.
c
Based on physical constraints in Mauritius.

Table 3 2.0%. This is an additional cost not reflected in model results, though
Mauritius electricity demand and growth assumptions. some cost reductions would also be likely in an optimized hourly sce-
2016 2040 growth rate nario.
Total curtailment in the base model is 7.2%, i.e., this amount of
Annual electricity demand, MWh 2,559 4,011 1.8% electricity is produced by the variable PV and wind sources at times
Peak power demand, MW 460 737 1.9%
when it is not needed and cannot be stored. Some level of curtailment is
a common result for cost-optimized models (Solomon et al., 2019); as
described in section 2.1, the most difficult days of the model year drive
installed effective capacity to 114 MW of reversible pumped storage.
system design, and under favorable sun and wind conditions extra
New lower reservoirs with a total volume of 7.3 Mm3 are included in
electricity is available. Given the day-night structure of the model for
the solution. With an assumed 335 m of head, this provides 5.1 GW h of
this study, actual curtailment is likely underestimated.
storage, or 0.47 days of average electricity consumption. Estimated cost
System levelized cost of generating electricity (LCOES) in scenario 1
of this is USD 0.139/kWh stored, not including the cost of electricity to
is estimated to be USD 0.130/kWh. While total cost of current gen-
charge the storage.
eration for the CEB is not publicly available, USD 0.130/kWh appears to
None of the model results include lithium battery storage, given its
be a plausible cost of electricity given current CEB retail tariffs. For
current costs. However, in testing results on an hourly basis, we find
example, residential rates are stepped, starting at a rate of USD 0.093/
that the day-night model structure underestimates storage charging
kWh for the first 25 kW h per month, and rising to USD 0.258/kWh for
power required, particularly during sunny days: the model is based on
usage in excess of 300 kW h per month (CEB, 2018). Total CEB revenue
daily average solar radiation, while in fact, there is a large insolation
in 2015 amounted to USD 0.172/kWh sold (CEB, 2016), though the
peak at midday. Lithium batteries would be the least expensive solution
utility obviously incurs expenses beyond obtaining electricity.
to this problem. For example, using lithium batteries to increase storage
With 785 MW of installed capacity in the base scenario, solar PV
power by 50% with 4 h of energy potential would increase LCOES by
farms would require an area of 17.7 km2, or 1.0% of the main island

Table 4
Results: Mauritius renewable electricity generating costs.
Scenario 1: base 100% Scenario 2: 80% renewable Scenario 3: low wind Scenario 4: no ground- Scenario 5: DSM to meet peak
renewable mount PV demands

LCOES, $/kWh $ 0.130 $ 0.105 $ 0.136 $ 0.136 $ 0.124

MW GWh % GWh MW GWh % GWh MW GWh % GWh MW GWh % GWh MW GWh % GWh

Wind turbines 475 1112 27% 486 1137 28% 100 234 6% 578 1341 33% 481 1126 28%
Solar PV, utility scale 785 1120 27% 687 963 24% 1079 1557 38% 0 0 0% 762 1079 27%
Solar PV, floating 0 0 0% 0 0 0% 0 0 0% 620 842 21% 0 0 0%
Solar PV, roof mounted 0 0 0% 0 0 0% 0 0 0% 0 0 0% 0 0 0%
Hydro and pumped hydro 114 130 3% 60 128 3% 87 130 3% 78 130 3% 108 130 3%
Lithium batteries 0 na na 0 na na 0 na na 0 na na 0 na na
Municipal solid waste plant 30 250 6% 30 250 6% 30 250 6% 30 250 6% 30 250 6%
Solid fuel plant 280 201 336 298 280
fuel: sugarcane bagasse 453 11% 453 11% 453 11% 453 11% 453 11%
fuel: cane tops and leaves 147 4% 147 4% 147 4% 147 4% 147 4%
fuel: domestic wood chips 120 3% 120 3% 120 3% 120 3% 120 3%
fuel: biomass crop 75 2% 75 2% 75 2% 75 2% 75 2%
fuel: imported wood chips 626 15% 0 0% 1077 26% 678 17% 619 15%
fuel: coal (nonrenewable) 0 0% 418 10% 0 0% 0 0% 0 0%
Liquid fuel plant 334 467 290 353 102
fuel: bioethanol 61 1.5% 0 0% 41 1% 53 1% 53 1%
fuel: diesel 0 0% 388 10% 0 0% 0 0% 0 0%
(nonrenewable)
Total 2019 4095 100% 1932 4080 100% 1973 4084 100% 1956 4089 79% 1764 4051 100%

7
D. Timmons, et al. Energy Policy 133 (2019) 110895

land area.2 Preferred areas for PV farms would combine high insolation described in section 2.1. At the minimum total cost, the marginal costs
with low-value land. The base result also includes 475 MW of wind of all generating sources must be equal on critical design days. If it
capacity, or 226 turbine towers at the 2.1 MW turbine size. Assuming should also be the case the marginal costs for different generating
for illustration that turbines were arranged on a grid with spaces of 5.0 sources are relatively constant, then there may be a variety of renew-
rotor-diameters (485 m for the turbine modeled), 53.2 km2 or 2.8% of able electricity portfolios with similar performance and costs. This
main-island land area would be needed for wind farms. Unlike solar PV, implies some flexibility for policy makers in establishing renewable
other activities can take place beneath wind turbines, and the best wind energy systems, at least in some cases. But for systems with variable
farm locations are often on high mountain ridges with fewer conflicting sources, the degree of this flexibility cannot be known without mod-
land uses than for solar PV. Important policy questions are whether and eling of variability, as demonstrated in this study.
how such quantities of land could be made available, and at what cost. In scenario 2, which requires only 80% renewable electricity, LCOES
Offshore wind may also be possible, but this is currently more ex- falls by 19% as compared to the base case. This suggests that obtaining
pensive than onshore (Lazard, 2018a), and the deep waters around the final increments of electricity from renewable sources may be re-
Mauritius would increase costs. latively expensive in the case of Mauritius. Here, coal is less expensive
than the renewable alternative of imported wood chips, and diesel fuel
3.2. Other scenarios is less expensive than renewable bioethanol, at least at current market
prices, and without considering the external costs of fossil fuels.
Scenario 2 reduces the renewable portion of electricity to 80% Continued reliance on imported diesel rather than switching to do-
(Table 4), to determine the cost of the final renewable increments, but mestic ethanol is another cost of this scenario. This suggests that strong
without fully addressing the costs of climate change and other fossil fuel policy measures would be needed to attain 100% renewable electricity,
externalities. The 80% goal causes a number of changes in results. In e.g. limits on fossil fuel use and/or subsidies for renewable fuels.
this scenario, wind generation increases slightly, and solar decreases by
12%. Coal takes the place of the imported wood chips used in the base 3.4. Roles of efficiency and demand-side management
scenario. Diesel oil replaces bioethanol for fuel, and liquid-fuel gen-
eration capacity increases to 467 MW. With greater capacity from dis- In theory, the equimarginal principle of section 2.1 should also
patchable fuels, pumped-hydroelectric storage capacity is reduced to apply to energy efficiency options, that is, the marginal cost of reducing
60 MW. LCOSS is USD 0.105/kWh, a reduction of 19.1% from the base energy consumption with greater efficiency should equal the marginal
scenario. cost of producing energy (Timmons et al., 2016). If the marginal cost of
In scenario 3 we test an arbitrary limit of 100 MW on wind gen- efficiency were less than generation, total cost could be reduced with
eration capacity, to see whether any obstacles to wind power devel- greater efficiency investments. However, it is difficult to assess effi-
opment (lack of accessible sites, aesthetics, public opposition, etc.) ciency opportunities and their marginal costs at a national scale, and
would greatly affect costs of achieving fully renewable electricity. The efficiency considerations are thus omitted from this study. We assume,
result is much greater production of solar electricity, which accounts for however, that a large and as yet unidentified portion of demand for
38% of total electricity in this scenario. Solid-fuel plant capacity in- electricity services can be met through better energy efficiency rather
creases 336 MW to accommodate the greater day-night variability of than electricity production. For example in Mauritius, there is potential
solar as compared to wind, and imported wood chips account for 26% for cooling buildings using water from the surrounding deep ocean,
of total electricity production. LCOES rises modestly to USD 0.136, an potentially reducing the electrical load at less cost than generating
increase of 4.2% over the base case. electricity for cooling (Francois, 2013).
Since the model minimizes cost, the base scenario utilizes only Similarly, the opportunity to manage electricity demand at times of
ground-mount PV, the least expensive solar option (Table 1). In sce- high marginal cost could be valuable in an electricity system based on
nario 4 we exclude ground-mount solar, forcing the model to use variable sources. Babonneau et al. (2017) consider a system with flex-
floating PV, the next most expensive option. Solar generation falls to ible loads responding to varying price signals, for example, charging
21% of the total, while wind generation rises to 33%. Pumped-hydro- electric vehicles at different times of day or night. Economic efficiency
electric storage capacity falls to 78 MW. At USD 0.136, LCOES is 4.6% requires that the marginal cost of providing a service equals its mar-
greater than in the base scenario. Rooftop solar is still not utilized, since ginal benefit. Where marginal cost of energy varies over time, accep-
based on our cost assumptions, floating solar is less expensive, and the table uses of energy should vary correspondingly, diminishing the need
projected area available for floating solar is not exceeded. to provide energy when it is expensive to do so.
Finally, scenario 5 tests the result of relaxing the requirement to
meet peak demand for a full day with no wind, solar, or hydroelectricity 3.5. Management of water for energy and consumption
available (the extreme scenario modeled as day 366). Bioethanol
power-plant capacity falls to 102 MW from 334 MW in the base sce- Hydroelectricity and pumped-hydroelectric storage in Mauritius are
nario, demonstrating that these bioethanol plants are used primarily to but two dimensions of a complex water management system whose
accommodate extreme events. LCOES falls slightly to USD 0.124/kWh, primary function is to ensure that domestic and agricultural water
a 4.3% reduction. However, the efficiency and demand-side manage- needs are satisfied. A new network of lower reservoirs constructed for
ment measures that would be required to implement this scenario pumped-hydroelectric storage might address both requirements—ena-
would result in additional cost savings not modeled here, as we discuss bling grid managers to dispatch electricity during times of high demand
in section 3.5. but capturing water flows in lower reservoirs before they are lost to the
sea. And new lower reservoirs would increase total water storage vo-
lume on the island, potentially of value during water shortages.
3.3. Range of options and cost sensitivity
Hydroelectric production in Mauritius is currently overseen by the
CEB, while other water functions are managed by the Central Water
As demonstrated by results shown in Table 4, electricity costs for
Authority, the Irrigation Authority, and the Wastewater Management
Mauritius vary somewhat by scenario, but costs do not necessarily vary
Authority (Proag, 2006). Using water for energy storage as well as
greatly. This outcome can be traced to the equimarginal principle
domestic and agricultural water consumption would require a high
level of coordination in managing water to meet multiple objectives.
2
Assumes solar farm density of 2.27 ha/MW, as found at the 15 MW Sarako Policy makers will need to consider whether the existing institutional
Solar farm at La Ferme, Mauritius. structures are compatible with managing resources for a fully

8
D. Timmons, et al. Energy Policy 133 (2019) 110895

renewable electricity system. Welsch et al. (2014) study the inter- time, renewable energy prices will likely decrease; the scenarios de-
relationships of climate, land use, energy, and water systems (CLEWS) scribed in this study thus provide an upper-bound cost of generating
in Mauritius, an approach that will likely be needed to assess the full renewable electricity in Mauritius. At the same time, nonrenewable
impacts of large-scale pumped-hydroelectric storage. fossil fuels must eventually become more expensive with scarcity, and
there is already a large environmental cost for burning fossil fuels. The
4. Conclusions and additional policy implications case for renewable energy is clear. Yet fossil-fuel resources remain
abundant, and a near-term transition to renewable energy sources is not
Historically, the sugarcane monocrop played a key role in the de- assured without policy intervention.
velopment of colonial Mauritius, long being the country's economic
backbone and providing much of its electricity. More than three cen- Funding
turies since sugarcane was introduced, this crop may be the basis of a
new bagasse-bioethanol-solar-wind renewable energy industry. This This project was supported by a Fulbright Program grant sponsored
study demonstrates that such a system can provide sufficient electricity by the Bureau of Educational and Cultural Affairs of the United States
to meet demand in each day and night of the year, at a price similar to Department of State and administered by the Institute of International
current electricity generation. Education.
Demand-side management with smart energy systems would also
allow the engagement of domestic, industrial and service-sector stake- References
holders in this energy model. Social harmony can be enhanced by such
a transformation, along with economic and environmental bene- Adam, N.-R.B., Dauhoo, M.Z., Elahee, M.K., 2017. A simulated-based genetic algorithm
fits—demand-side management has the potential to reduce both total for the forecasting of monthly peak electricity demand. Int. J. Oper. Res. 28 (2),
164–182.
costs and total land requirements. Babonneau, F., Caramanis, M., Haurie, A., 2017. ETEM-SG: optimizing regional smart
Policy implementation in a small-island economy like Mauritius energy system with power distribution constraints and options. Environ. Model.
should benefit from less resistance since it urgently concerns a country Assess. 22 (5), 411–430.
Beeharry, G.K., Sookdeb, M., Dookhun, V., 2009. Hydro energy for maurice ile durable.
vulnerable to climate change. The policy challenge lies in integrating a In: Paper Presented at University of Mauritius Research Week 2009-2010.
balanced, holistic approach combining energy issues transversely with Bett, P.E., Thornton, H.E., 2016. The climatological relationships between wind and solar
engineering, economic, environmental and ethical considerations. At energy supply in Britain. Renew. Energy 87, 96–110.
Carnegie Clean Energy, 2017. High Penetration Renewable Energy Roadmap for the
the same time, many aspects of the energy transition need to be better Republic of Mauritius. Mauritius Research Council. http://www.mrc.org.mu/
understood, for example: linkages with land planning and development English/Pages/High-Penetration-.aspx, Accessed date: 11 November 2018.
in a densely-populated island; sustainable agriculture particularly re- CEB, 2016. Annual Report 2015: Optimizing Our Renewable Energy Potential. Central
Electricity Board, Ebene, Mauritius. . http://ceb.intnet.mu/CorporateInfo/ar2015.
lated to the sugarcane industry; water management; transport and
pdf, Accessed date: 11 November 2018.
mobility; climate change impacts on building and infrastructure; waste CEB, 2018. Domestic Tariff." Central Electricity Board. http://ceb.intnet.mu/tariffs/
management; potential for use of ocean energy resources; and con- domestic.asp, Accessed date: 7 October 2018.
sumer behavior. But undeniably, the question of energy governance has Child, M., Nordling, A., Breyer, C., 2017. Scenarios for a sustainable energy system in the
Åland Islands in 2030. Energy Convers. Manag. 137, 49–60.
to be addressed through a transformation of the current institutional Child, M., Bogdanov, D., Breyer, C., 2018. The role of storage technologies for the tran-
and regulatory framework. sition to a 100% renewable energy system in Europe. Energy Procedia 155, 44–60.
Similar policy considerations will apply in many places where re- Dhunny, A., Lollchund, R.M., Boojhawon, R., Rughooputh, S.D., 2014. Statistical mod-
elling of wind speed data for Mauritius. Int. J. Renew. Energy Resour. 4 (4),
newable energy is developed, and some policy questions are universal. 1056–1064.
For example, the cost of capital-intensive renewable energy resources Dhunny, A., Lollchund, M., Rughooputh, S., 2015. Long-term wind characteristics at se-
depends greatly on the discount rate used: a fundamental policy ques- lected locations in Mauritius for power generation. J. Wind. Energy 2015.
Dhunny, A., Lollchund, M., Rughooputh, S., 2017. Wind energy evaluation for a highly
tion is whether renewable energy development will be undertaken with complex terrain using Computational Fluid Dynamics (CFD). Renew. Energy
public or private capital, and what rates of return will be provided to 101, 1–9.
investors. Also, non-cost considerations such as land use (solar PV, Diesendorf, M., Elliston, B., 2018. The feasibility of 100% renewable electricity systems: a
response to critics. Renew. Sustain. Energy Rev. 93, 318–330.
wind, and biomass), aesthetics (wind), and water management (hy-
EIA, 2019. Levelized Cost and Levelized Avoided Cost of New Generation Resources in the
droelectricity) greatly affect a determination of which portfolio is op- Annual Energy Outlook 2019. U.S. Energy Information Administration, Washington,
timal, and policy makers must appropriately weigh these issues. D.C. http://www.eia.gov/forecasts/aeo/pdf/electricity_generation.pdf, Accessed
date: 10 July 2019.
As shown in this study, obtaining a cost-minimizing renewable-
Ekholm, T., Ghoddusi, H., Krey, V., Riahi, K., 2013. The effect of financial constraints on
electricity generating portfolio implies constructing specific technolo- energy-climate scenarios. Energy Policy 59, 562–572.
gies, albeit with some degree of flexibility around the proportions of Elahee, K., 2010. Energy management challenges in a small island economy: the case of
those technologies. Another important question is thus which policy Mauritius. Energy Environ. 21 (7), 803–813.
Elahee, K., 2011. The challenges and potential options to meet the peak electricity de-
instruments can most effectively incentivize the desired generating mand in Mauritius. J. Energy South. Afr. 22 (3), 8–15.
portfolio. Any approach must recognize the inherent variability of Elahee, K., 2013. Potential of hydropower in Mauritius: myth or reality? Energy Sources,
many renewable energy sources, along with the resulting variability in Part A Recovery, Util. Environ. Eff. 35 (10), 921–925.
ELC Electroconsult, 2015. Opportunity Assessment for the Development of Geothermal
marginal costs and marginal values of electricity generated or stored. Energy in Mauritius. Ministry of Energy and Public Utilities, Milan, Italy. http://
Fixed subsidies or feed-in tariffs are therefore unlikely to provide an publicutilities.govmu.org/English/Documents/030316/Mauritius_Final_Report.pdf,
optimal generating portfolio or to minimize total cost. For policy ma- Accessed date: 7 October 2018.
Elliston, B., Macgill, I., Diesendorf, M., 2013. Least cost 100% renewable electricity
kers, mandating a specific portion of energy to be obtained from all scenarios in the Australian National Electricity Market. Energy Policy 59, 270–282.
renewable sources collectively (a renewable portfolio standard) is Ernst, M., Thomson, A., Haedrich, I., Blakers, A., 2016. Comparison of ground-based and
perhaps the simplest approach, with a combination of utility planning satellite-based irradiance data for photovoltaic yield estimation. Energy Procedia 92,
546–553.
and market mechanisms used to allocate funding to specific technolo-
ETSAP, 2018. Energy technology systems analysis Programme: MARKAL FAQ's.
gies. Available at: http://iea-etsap.org/markal/faq.html, Accessed date: 9 September
Perhaps most importantly, policy makers must establish targets and 2018.
Francois, F., 2013. Mauritius: A Project to Cool the Buildings in Port Louis with Sea
dates for an energy transition. Renewable energy is available, feasible,
Water. Indian Ocean Times. https://en.indian-ocean-times.com/Mauritius-A-project-
and potentially affordable. As we demonstrate in this study, an opti- to-cool-the-buildings-in-Port-Louis-with-sea-water_a1800.html, Accessed date: 11
mization model can be used to approximate the generation capacities of November 2018.
technologies needed to meet electricity demand at all times in any Franzen, E., Strauch, N., Triebel, C., Bosch, E., Richards, B., 2013. Switching off the
generator: technical and economic approach for storage based renewable energy
given area, and to estimate the cost of that generating portfolio. Over

9
D. Timmons, et al. Energy Policy 133 (2019) 110895

systems for islands. Energy. Sustain. 176, 41. Pfenninger, S., 2017. Dealing with multiple decades of hourly wind and PV time series in
François, F., 2014. Mauritius: a wind farm soon under construction in the region of Mare - energy models: a comparison of methods to reduce time resolution and the planning
aux-Vacoas. Indian Ocean Times. Available at: https://en.indian-ocean-times.com/ implications of inter-annual variability. Appl. Energy 197, 1–13.
Mauritius-A-wind-farm-soon-under-construction-in-the-region-of-Mare-aux-Vacoas_ Pleßmann, G., Erdmann, M., Hlusiak, M., Breyer, C., 2014. Global energy storage demand
a4591.html, Accessed date: 11 November 2018. for a 100% renewable electricity supply. Energy Procedia 46, 22–31.
Gardumi, F., Shivakumar, A., Morrison, R., Taliotis, C., Broad, O., Beltramo, A., Proag, V., 2006. Water resources management in Mauritius. Eur. Water 15 (16), 45–57.
Sridharan, V., Howells, M., Hörsch, J., Niet, T., 2018. From the development of an Purchase, B., Rosettenstein, S., Bezuidenhoudt, D., 2014. Challenges and potential solu-
open-source energy modelling tool to its application and the creation of communities tions for storage of large quantities of bagasse for power generation. Int. Sugar J. 116,
of practice: the example of OSeMOSYS. Energy Strategy Reviews 20, 209–228. 592–602.
Godina, R., Rodrigues, E., Matias, J., Catalão, J., 2015. Sustainable energy System of El Ramgolam, Y.K., 2018. Modelling and Evaluation of Photovoltaic System through Holistic
Hierro island. In: International conference on renewable energies and power quality Resource Assessment. Ph.D. dissertation, Faculty of Engineering. University of
(ICREPQ’15) Renew. Energy. Power Qual.J(RE&PQJ) Vol 13. pp. 46–51. Mauritius.
Hansen, K., Breyer, C., Lund, H., 2019. Status and perspectives on 100% renewable en- Ramgolam, Y., Soyjaudah, K., 2014. Enhanced Insolation and Global Irradiance in Near-
ergy systems. Energy 175, 471–480. Tropic Region. EuroSun 2014, Aix-les-Bains, France.
Havumaki, B.M., 2018. Hydropower in the Decarbonized Mauritian Grid: A Prospective Ramgolam, Y.K., Soyjaudah, K., 2015. Unveiling the solar resource potential for photo-
Study. Department of Economics, University of Massachusetts, Boston. voltaic applications in Mauritius. Renew. Energy 77, 94–100.
Heaps, C.G., 2016. LEAP: Introduction to Optimization. Stockholm Environment Institute. Sahai, I., 2004. Power plans in Mauritius. Int. Water Power Dam Constr. 56 (9), 10–13.
https://www.energycommunity.org/Help/Optimization/OptimizationIntroduction. Sahu, A., Yadav, N., Sudhakar, K., 2016. Floating photovoltaic power plant: a review.
htm, Accessed date: 11 November 2018. Renew. Sustain. Energy Rev. 66, 815–824.
Howells, M., Rogner, H., Strachan, N., Heaps, C., Huntington, H., Kypreos, S., Hughes, A., Shea, R.P., Ramgolam, Y.K., 2019. Applied levelized cost of electricity for energy tech-
Silveira, S., DeCarolis, J., Bazillian, M., 2011. OSeMOSYS: the open source energy nologies in a small island developing state: a case study in Mauritius. Renew. Energy
modeling system: an introduction to its ethos, structure and development. Energy 132, 1415–1424.
Policy 39 (10), 5850–5870. Solarmap, 2017. Solar map of Mauritius, Rodrigues, and agalega. Available at: http://
IPCC, 2018. Global Warming of 1.5°C. Intergovernmental panel on climate change. solarmap.uom.ac.mu/, Accessed date: 7 October 2018.
Available at: https://www.ipcc.ch/pdf/assessment-report/ar5/syr/AR5_SYR_FINAL_ Solomon, A., Kammen, D.M., Callaway, D., 2016. Investigating the impact of wind–solar
SPM.pdf, Accessed date: 10 July 2019. complementarities on energy storage requirement and the corresponding supply re-
IRENA, 2018. Renewable Power Generation Costs in 2017. International Renewable liability criteria. Appl. Energy 168, 130–145.
Energy Agency. http://www.irena.org/publications/2018/Jan/Renewable-power- Solomon, A., Bogdanov, D., Breyer, C., 2019. Curtailment-storage-penetration nexus in
generation-costs-in-2017, Accessed date: 11 November 2018. the energy transition. Appl. Energy 235, 1351–1368.
Jacobson, M.Z., Delucchi, M.A., Cameron, M.A., Frew, B.A., 2015. Low-cost solution to SRM, 2012. Proposed Construction and Operation of a Wind Farm Comprising 14 Wind
the grid reliability problem with 100% penetration of intermittent wind, water, and Turbines with a Total Installed Capacity of 29.4 MW. National Solidarity, and
solar for all purposes. In: Proceedings of the National Academy of Sciences, vol.112. Environment and Sustainable Development (Environment and Sustainable
pp. 15060–15065 49. Development Division), Port Louis, Mauritius.
Jefferson, M., 2012. HC 517, the Economics of Wind Power. UK Parliament, London. Statistics Mauritius, 2017. Energy and Water Statistics - Year 2016. Ministry of Finance
https://publications.parliament.uk/pa/cm201213/cmselect/cmenergy/writev/517/ and Economic Development, Port Louis, Mauritius.
m26.htm, Accessed date: 10 July 2019. Staub, C.G., Stevens, F.R., Waylen, P.R., 2014. The geography of rainfall in Mauritius:
Kantor Management Consultants, 2008. Energy Policy for the Republic of Mauritius: Draft modelling the relationship between annual and monthly rainfall and landscape
Final Energy Policy: TASK C: Electricity Sector. Ministry of Energy and Public characteristics on a small volcanic island. Appl. Geogr. 54, 222–234.
Utilities, Ebene, Mauritius. Surroop, D., Raghoo, P., 2017. Energy landscape in Mauritius. Renew. Sustain. Energy
Khoodaruth, A., Oree, V., Elahee, M., Clark II, W.W., 2017. Exploring options for a 100% Rev. 73, 688–694.
renewable energy system in Mauritius by 2050. Util. Policy 44, 38–49. Surroop, D., Raghoo, P., 2018. Renewable energy to improve energy situation in African
KTH - DESA, 2018. OSeMOSYS: open source energy modeling system. Available at: www. island states. Renew. Sustain. Energy Rev. 88, 176–183.
osemosys.org, Accessed date: 7 October 2018. Suzlon-Padgreen, 2012. Environmental Impact Assessment (EIA) Reports: Proposed
Lazard, 2017. Lazard's Levelized Cost of Energy Analysis-Version 11.0. Available at: Construction and Operation of a Wind Farm Comprising 14 Wind Turbines with a
https://www.lazard.com/perspective/levelized-cost-of-energy-2017/, Accessed date: Total Installed Capacity of 29.4 Mw by Consortium Suzlon-Padgreen Co Ltd, Annex 6.
1 February 2018. Ministry of Social Security, National Solidarity, and Environment and Sustainable
Lazard, 2018a. Lazard's Levelized Cost of Energy Analysis-Vversion 12.0. https://www. Development. (Environment and Sustainable Development Division). http://
lazard.com/media/450784/lazards-levelized-cost-of-energy-version-120-vfinal.pdf, environment.govmu.org/English/eia/Pages/EIA-Reports.aspx, Accessed date: 11
Accessed date: 1 February 2019. November 2018.
Lazard, 2018b. Lazard's Levelized Cost of Storage Analysis-Vversion 4.0. https://www. Timilsina, G.R., Shah, K.U., 2016. Filling the gaps: policy supports and interventions for
lazard.com/media/450774/lazards-levelized-cost-of-storage-version-40-vfinal.pdf, scaling up renewable energy development in Small Island Developing States. Energy
Accessed date: 1 February 2019. Policy 98, 653–662.
Lema-Driscoll, A.K., 2018. Renewable Biomass Energy Utilization in Mauritius: A Cost- Timmons, D., 2018. Optimal renewable energy systems: minimizing the cost of inter-
Minimization Approach. Department of Economics, University of Massachusetts mittent sources and energy storage. In: Letcher, T.M., Fthenakis, V.M. (Eds.), A
Boston. Comprehensive Guide to Solar Energy Systems. Elsevier, pp. 485–504.
Lenzen, M., McBain, B., Trainer, T., Jütte, S., Rey-Lescure, O., Huang, J., 2016. Simulating Timmons, D., Konstantinidis, C., Shapiro, A.M., Wilson, A., 2016. Decarbonizing re-
low-carbon electricity supply for Australia. Appl. Energy 179, 553–564. sidential building energy: a cost-effective approach. Energy Policy 92, 382–392.
Ministry of Energy and Public Utilities, 2011. Hydrology Data Book 2006-2010. Weisser, D., 2004. Costing electricity supply scenarios: a case study of promoting re-
Available at: http://publicutilities.govmu.org/English/Pages/Hydrology-Data-Book. newable energy technologies on Rodriguez, Mauritius. Renew. Energy 29 (8),
aspx, Accessed date: 11 November 2018. 1319–1347.
Ministry of Renewable Energy and Public Utilities, 2009. Republic of Mauritius, Long-Term Welsch, M., Hermann, S., Howells, M., Rogner, H.H., Young, C., Ramma, I., Bazilian, M.,
Energy Strategy 2009-2025, Port Louis, Mauritius. https://sustainabledevelopment. Fischer, G., Alfstad, T., Gielen, D., 2014. Adding value with CLEWS–Modelling the
un.org/content/documents/1245mauritiusEnergy%20Strategy.pdf, Accessed date: energy system and its interdependencies for Mauritius. Appl. Energy 113,
11 November 2018. 1434–1445.
Morrison, R., 2018. Energy system modeling: public transparency, scientific reproduci- Wolf, F., Surroop, D., Singh, A., Leal, W., 2016. Energy access and security strategies in
bility, and open development. Energy Strategy Reviews 20, 49–63. small island developing states. Energy Policy 98, 663–673.

10

You might also like