Table of Contents
Go Back
A SYSTEM DYNAMICS VIEW OF ENERGY DEMAND
AND SUPPLY IN SMALL ISLANDS
Paraskevas Gravouniotis
Imperial College London
Centre for Energy Policy & Technology
Department of Environmental Science and Technology
Prince Consort Road, London SW7 2BP, United Kingdom
W: http://www. iccept.ic.ac.uk, T: +44 (0)20 75947309, F: +44 (0)20 75949304
p.gravouniotis@imperial.ac.uk
ABSTRACT
The paper summarises preliminary research on feedback relationships and structures
among stakeholders that make up the unique energy profile of an autonomous island.
The model builds on data from the islands of the Greek archipelago. Despite being a
small-scale application, there is a complex socio-economic structure with convoluted
stakeholder interests in place that can provide insights into the current debate on
technology choice in energy policy. The real world problem addressed is the great
financial cost of electrifying island’s not connected to a mainland grid in conjunction
with a state policy of uniform subsidised tariffs. The objective is to design demand-side
interventions that are able to identify appropriate technologies, early adopters,
potential niche markets and amalgamate this into an optimal integrated policy to
mitigate that cost.
Keywords: islands, demand-side
1 INTRODUCTION TO THE UNOCONNECTED GREEK ISLANDS
The energy systems on the Greek islands of the Aegean exemplify five main
characteristics in varying scales across the archipelago that also underlie their
problematic behaviour:
1. High costs of electricity generation that is subsequently sold at subsidised rates
to end-users.
2. Distribution level local grids and limited interconnections to neighbouring
islands.
3. A growing residential consumption mainly attributed to rising income and
readily available electric appliances.
4. A growing, highly seasonal and energy-intensive tourism sector.
5. A limited range of currently utilised supply options resulting in heavy reliance
on imported fuel oil for the generation of electricity and heating.
The Public Power Corporation (PPC) solely operates the autonomous power systems of
the Greek unconnected islands through its Island Directorate. The cost of generating one
kWh of electricity on the islands ranges from €0.11 to €1.28 (PPC 2002). The retail
price for the household sector is about €0.08/kWh (the standard household tariff is
uniform across the country for social equity purposes). This price difference amasses
each year to an unavoidable gap in the finances of the corporation. For 1998 it has been
reported in the press to stand at €205 million (£133m) and rising.
The introduction of more efficient technologies to provide the same energy services to
customers can potentially yield great economic benefits for consumers and the Public
Power Corporation (PPC) alike. The benefits of such kind of demand-side management
to the operator can accrue through investment deferral and optimal plant loading, i.e.
reduced fuel consumption. The current paper centres on setting a simulation exercise to
capture the agents and their decisions, which is the necessary step to assess the
aforementioned impact on costs. This latter element is a major objective of the ongoing
research. Namely, the model sets up a demand/supply structure on basic assumptions,
rules for capacity expansion that were devised from literature and interviews and,
finally, it touches on substitution dynamics among energy service appliances.
The broader aim is to prove efficient demand-side technologies can halt the rate of
capacity expansion and thus save money compared to a BAU case. It is then a question
of policy making to decide how much of potential savings can be fed back into support
for efficient technologies, in what form and for how long. The key objective is to meet
the load at all times for all users’ needs by utilising market forces to kick of an energy-
efficient technology market which can show convincing benefits to both supplier and
consumer and thus overcome the barriers of commercial deployment.
2 REVIEW OF THE CURRENT SITUATION
2.1 Energy Use
Fuel oil, diesel and petrol are the predominant fuels in the islands for electricity
generation, transportation and the agricultural sector. Figure / below shows the
relative distribution of fuels in energy supply typically found across the Aegean.
Gasoline is consumed almost entirely in transportation. Diesel is principally consumed
for heating and electricity generation by the Public Power Corporation (PPC). Most of
the heavy fuel oil is used for power generation too (Balaras et al. 1999, 24:335-
350;Mihalakakou et al. 2002, 26:1-19).
Hotels Public ~~ Commer
% \. 3% % Agric
fo Biomass oe Oe
‘ 3% \ a,
Sol < indust
gag Residentl eo
7 5%
18%
LPG
1%
Gasoline Diesel
21% 34% PF Transp
57%
Figure 1: Distribution of energy supply per Figure 2: Useful energy consumption per
fuel in the Dodecanese. economy sector for the Dodecanese.
Generation sets present on the islands are in the range of 80-1000kW.
2.2 Electricity
The pie chart in _ Figure 2 is based on data from the Dodecanese and is a pattern met
across the smaller islands of the Aegean Sea. The residential sector and hotels together
account for 26% of the consumption and the energy vector concerned is predominantly
electricity. The hotel and commercial sector weigh much more in relevant terms, as they
are operational half of the year (Balaras, Santamouris, Asimakopoulos, Argiriou,
Paparsenos, and Gaglia 1999, 24:335-350;Mihalakakou, Psiloglou, Santamouris, and
Nomidis 2002, 26:1-19).
Electricity consumption by sector reveals a particular increase for the domestic and
commercial sectors. There could be many reasons behind these increases and there have
been no conclusive studies on the issue to date. The most likely reasons are tourism -
domestic as much as international and rising standard of living. Domestic electricity
consumption between 1992 and 1996 increased by 25% in the Cyclades. Increase in the
tertiary sector was 60% while both industrial and agricultural electricity use in the
Cyclades increased by 30% for the same period (Balaras, Santamouris, Asimakopoulos,
Argiriou, Paparsenos, and Gaglia 1999, 24:335-350;Mihalakakou, Psiloglou,
Santamouris, and Nomidis 2002, 26:1-19).
Net electricity production and peak power demand between 1996 and 2001 show an
average increase 40-45%. Figure 3 presents the case for a selection of islands across the
Aegean. Peak power demand has been observed to occur in August that is the most
touristic month of the year and there is increased air-conditioning, lighting and hot
water demand. For the rest of the year demand lies substantially lower than that
(Balaras, Santamouris, Asimakopoulos, Argiriou, Paparsenos, and Gaglia 1999, 24:335-
350;Mihalakakou, Psiloglou, Santamouris, and Nomidis 2002, 26:1-19).
Percentage Increase of Generation and Peak Demand 1996-2001
O Generation OPeak
©
a
2
s
PS
ro
rs
S
8
=
0%
SPE ELE EL - ES » g
SE OES CEES ~ SS BS ES SEES &, Ss LE SLES
& oe eS Pot aS
we Islands
Figure 3: Power generation and peak demand increase for selected islands.
2.3 The Domestic Sector
Table 1 shows the great dependence on modern energy services, which in turn indicates
rigid energy consumer behaviour. As long as the real costs of electricity production are
not reflected on island consumers’ bills, it is difficult to prove the conservation benefits
to customers given that the study quoted in the table below showed the majority seemed
indifferent to energy conservation practices, did not keep sufficient records of energy
bills and were not informed on the energy efficiency of appliances (Haralampopoulos et
al. 2001, 26:187-196).
Table 1: Electric appliance penetration in 571 households of Mytilene.
Typeof appliniies Households by no. of appliances Total Penetration
I 2 3 4 appliances | (%)
Oven 20 54417 - - 558 97.7
Refrigerator 3 529 | 35 3 1 583 102.1
Electric water heater | 203 | 365 | 3 - - 371 65.0
Kitchen el wir heater | 396 | 173 | 2 : . 173 30.3
Washing machine 63 502 | 6 - - 5l4 90.0
Dish washer 41s | 152 | 1 - : 154 27.0
Iron 36 525 | 5 4 1 551 96.5
Television 12 406 | 128 | 22 3 740 129.6
Freezer 489 | 22 = : . 22 39
Microwave oven 549 | 15 - - - 15 2.6
Toaster 560_| Il : : UW 19
2.4 The Tertiary Sector
The sector consists mainly of hotels, public sector (schools, hospitals etc.) and
commercial (e.g. restaurants, shops etc.). The hotel sub-sector spends most of the
energy available for the sector. In the case of the Dodecanese the relative share is 50%,
16% and 34% respectively (Mihalakakou, Psiloglou, Santamouris, and Nomidis 2002,
26:1-19). Taking under account the hotel sector is operational for half of the year; its
consumption has a great impact on the energy supply of an island. For the islands of the
Aegean, the hotel sub-sector seems a candidate segment of the appliances market to
introduce technologies that cut down on energy bills. Hotels account for 9% of total
energy consumption in the Dodecanese. The energy is mainly used in water heating, air
conditioning and lighting (ALTENERII 2001). The simulation model however only
examines adoption of these technologies in the domestic sector as its smooth
consumption pattern allows better study of the suggested substitution algorithm.
The main driver behind the increasing energy demand of hotels is tourism arrivals.
Tourism has a great impact on seasonal population that in consequence stresses capacity
levels to meet base and peak loads that are physically impossible to meet unless there is
generation increase on the island. There are rare and limited interconnections among
islands that would allow extended load management options. There has been no
restriction on the efficiency range of electric appliances so far by the government.
Increased generation has an immediate impact on the costs of the generated power, as
either the gensets need to run close to rated output, the operator has to utilise older
stand-by and inefficient gensets or eventually invest in expansion of the power plant.
The first two occasions add to the running costs of the operator whereas the latter comes
in as a fixed cost. In addition, the cost of the outages and the lost revenue of foregone
sales should be included in the evaluation of the cost of not meeting the load. That
figure can then provide the basis for comparing alternative interventions.
2.5 Case Study: Serifos island
The small island of Serifos has been chosen as a representative of an unconnected island
of the Aegean Sea in order to demonstrate more detailed data on the underlying
characteristics of the problem.
Generation and Peak Demand From 1977 to 2001
—*— Peak (kW) —8— Generation (MWh)
6,000
5,000
4,000
3,000
2,000
1,000
Gross Generation & Peak
Demand Values
© A MS & © A @ NS O SN S oP
CELE IEFESEPEFEESPEESES
PP F
Year
Figure 4: Historical generation & peak demand for Serifos.
The island has been facing increasing demand growth over the past 25 years with no
sign of stabilisation (Figure 4). The clear exponential growth exemplified suggests there
is a dominant reinforcing loop at play.
Similar graphs are found in the official annual summary of island electrification of the
unconnected islands. Future consumption and peak demand growth are forecasted
simply by drawing a line assuming about the same growth rate (PPC 2002). This,
creates a situation of a ‘self-fulfilling prophesy’ as the future growth of any exponential
growth requires multiplication of effort to mitigate its effects. As a result all attention is
drawn to fulfilling the projected demand as required by law.
Fuel Consumption & Cost of kWh From 1977 to 2001
—*—LFO (tonnes) +—=—Cost (€/kWh)
2,000 0.35
Fuel oil
consumption
(tonnes)
3
8
8
°
8
8
Price of kWh
Generated (€/kWh)
0+
CM LOO EAPO SDS HAP OS
LL LPP PPE PEPE FE EE ES SL
Year
Figure 5: Historical fuel consumption & cost of kWh for Serifos.
Looking at the fuel consumption and generation cost of kWh data for the same time
series the pattern is repeating albeit with periodical fluctuations in the cost of electricity.
These oscillations are assumed to be due to the operating costs based on loading of the
plant throughout the year (summer/winter) and the fluctuations of tourism arrivals.
Causes in the price of fuel are ignored.
3 BUILDING THE SIMULATION EXERCISE
It is quite apparent from the case of Serifos that peak demand and consumption over the
past three decades cannot incessantly grow. One would expect the system will
eventually reach saturation and balance in either of two ways:
i) Energy consumption rate slows down significantly when households reach
the required comfort level with respect to electricity appliances and,
furthermore, tourism arrivals stabilise or,
ii) The island reaches it natural carrying capacity for power plant additions to
accommodate tourist and inhabitants when further expansions degrade the
natural environment so that the flow of inhabitants and visitors halts.
However, given that costs already outweigh revenues and the foreseeable future is likely
to bring more wealth and thus energy spending, as it has been suggested in section 2.3,
the system would likely collapse financially before anything else. One can now define
the horizon of our simulation exercise to the short- to medium- term. Thirty years
should be enough to allow the full effect of a demand-side intervention to settle.
3.1 The Dynamic Hypothesis
Drawing together the elements presented in section 3 above, the following Causal Loop
Diagram (CLD) in Figure 6 elaborates further on the proposed dynamic hypothesis
responsible for the problematic behaviour (high costs) on the Greek islands.
Tourism
Tourist Arrival
Demand
“y Peak
<P Expected Capacity
Base Load “=
& Cap
12) 7
Income
Ng al il. lhc
ilectrcity °
Spending A) Capacity stan 2
s
Figure 6: Telling the story with a Causal Loop Diagram.
The CLD consists of two balancing (B1 & B2). All the elements are interlinked
demonstrating the endogenous reasoning of the dynamic behaviour whereas there are
two exogenous elements, household income and tourist arrivals. Going around B1, an
increase in household (hhs) consumption leads to an opposite direction change, i.e.
decrease, in the capacity margin. Moving around the loop, a decreased margin will
likely cause reduced electricity expenditure subject to income, as the possibility of
outages is higher. Eventually, the loop closes by forcing hhs consumption to decrease.
Assuming a similar initial change but following loop R1, one ends up generating a
further increase in hhs consumption. This reinforcing effect is assumed to be at play, for
example in the case of Serifos, being dominant and responsible for the exponential
growth pattern. The disrupting role of tourism is central to the dynamics of the system
and is effectively acting as an agitator once a year peaking in August. On one hand, it
intensifies the base load profile, which distorts the forecasting data and amplifies
response, i.e. greater one-off capacity investments. On the other hand, it initiates peaks
of random nature, duration and timing that can cause extremely fuel consuming
operating conditions, outages or even propagate into the forecasting horizons
amplifying even further any investment imperatives.
The former effect is captured within R1 by being added directly to the base load. This
indicates that it is assumed largely predictable. The latter impact of peaks forms itself a
balancing loop that feeds into an adjusted capacity expansion suggesting that power
shedding is unacceptable due to the social remit of the PPC and the decision-making in
place acts to mitigate any gap.
To summarise, the more the foreign and domestic tourists, the higher the energy
requirements during the summer. As soon as there are repeated losses of load in the
summer, the PPC will have to add capacity to stick to its public service obligations.
However, during the winter demand is roughly 20% lower than in summer
(Haralampopoulos, Fappas, Safos, and Kovras 2001, 26:187-196). As incomes grow
(also due to tourism) and electricity comfort is lacking, this capacity will be absorbed by
the households. The next power shortage is not likely to arise until a few summer
seasons depending on tourists arrivals which is exogenous to the system (bound to the
international economy and competition). This circle is repeated over the years leading to
the exponential growth seen for Serifos since the dominant loop is R1. As long as
measures of energy efficiency are not institutionally initiated the demand will always
rise due to the intricate dynamics of the island economies (high building rate, increasing
volumes of tourism and rising standard of living).
3.2 The Residents and Tourism
The model assumes an island of 1,000 households with the following appliances:
Table 2: The typical household's pool of appliances in the Average island.
Description of appliance Number Rating Cons each Operation details
(Ww) forlhr (kWh)
5 hours each a day. Winter
Incandescent light bulbs 5 100 0.10 darkness substituted by
summer late nights.
Space heater (electric) 1 2000 2.00 2 Hours Bee day’ fon +6
months each year.
Water heater (electric)
[50°C for 8Olit] 1 4000 2.60 2 hours per day.
Refrigeration [131 lit] 1 90 0.50 24hrs figure. All year.
Air Conditioning N/A - - No A/C yet.
That amounts to an average consumption 310 kWh a month. The peaks of the household
sector are assumed to be covered by capacity expansion to meet summer peaks therefore
have not been explicitly modelled. Indeed, there is hardly any occasion where out-of-
season peak demand exceeds installed capacity (PPC 2002).
Also, assume that each tourist comes on a two-week holiday package and receives the
following energy services:
Table 3: Services to a typical tourist on a two-week holiday package.
Description of service Number of | Rating | Cons each for Service use details
appliances (W) 1hr (kWh)
Lighting 4 100 0.10 8hrs per day (assume 1 bulb per
commercial place visited)
‘ Beit g 2hrs in room, Ihrs in shops &
Air Conditioning: 1/2 1000 1 2hrs in entertainment. Shrs in
[per 15sq.m.]
total.
., 1 shower per day. Equivalent to
Hot watts. 1 2000 | 0.33 2hrs use as service not
[50°C for 10lit] :
customised (losses).
Dining (refr’n) [13 1lit] 1/2 90 0.50 24hrs operation. Everyday.
Cleaning services 1/2 2800 1.30 Laundry: 2hrs/week
(laundry+dishwasher) 1/2 3200 1.30 Dishwasher: 2hrs/day
Each tourist loads the system by 245 kWh on a monthly
demand of 2.5 kW.
average and carries a peak
3.3. The Technology Substitution Sub-model
The critical component to the objectives of this paper is the introduction of an additional
sub-model that basically isolates a consuming technology out of the electricity services
basket of households.
saprgete AC cons 2
PRODUCT INTRODUENON 2pergate AC cons
Potential Adopts
aortas
Adopters 2
[ADVERTISING —sdopton tom
stvetting
adoption
Figure 7: The technology spread and substitution curve.
It is assumed this technology is an air-conditioning units that was not originally
included in the list of services in Table 2. The concept is that a market for the service
develops until some point that a more efficient mark of the unit is available.
The timing and size of a supposed efficient units demonstration scheme decided by the
policy-makers, as well as parameters relating to word-of-mouth effect, will define the
substitution function. The policy-maker has to be able to assess the cost of the scheme
with respect to the expected benefits but also intervene in the word-of-mouth dynamics
of the population. It should be noted that this component has been parameterised to
demonstrate its potential impact during simulation and does not constitute a culmination
in research on substitution dynamics in the context of the Greek islands or the
assessment of costs and benefits to producer and consumers.
4 OVERVIEW OF SIMULATION RESULTS
The initial condition for which the model is tested simulates a situation where there is
only increasing household consumption. After an initial investment the system stabilises
at 755 kW installed capacity. During the simulation there are two occasions where the
consumption rate (forecasted) comes close to available supply and the expansion
mechanism is put at work to mitigate that (Figure 8). At year 15, installed capacity is up
50% to 1,100kW and ten years later it has to increase by another 400kW.
q
Naer ?
Figure 8: Max generation and consumption rates. No efficient technology or tourism.
The following test (Figure 9) is assuming a new energy service (i.e. electric air-
conditioning units) is available to the households from the start of the simulation and
then eventually replaced with its efficient (at 1/3 rating) version available at month 100.
oo ————1
Neer 2
Figure 9: Replacing with more efficient equipment.
Line -1- represents the original situation as in Figure 8. Line -2- is the capacity
expansion with the new energy service technology becoming dominant and available in
the original consumption rating only. There is additional capacity investment and also
change in the timing and amplitude of all additions likely to incur relative additional
costs to the system (i.e. the time value of money having to invest now rather than later).
The calculation of that cost is beyond the reach of this paper but forms an integral part
of the broader research topic. The installed capacity for Line -2- capacity stands at about
470kW more than Line -1- by the end of the simulation. Lines -3- and -4- indicate a
‘softer’ and ‘harder’ institutional promotion of the more efficient version respectively. It
is noted that the impact of these two modes is not as significant as the introduction
itself. This can be attributed to the nature of the service (air-conditioning) designed to
be in high demand, only available to households and reaching saturation within the
simulation horizon. However, the major impacts of efficiency are investment deferral
and size of additions as can be noted in both lines. By the end of the simulation,
installed capacity is as low as 20% compared to Line -2-. Figure 10 shows the
substitution curves at play behind the impact of efficiency.
sss
4 a
Figure 10: The substitution curves.
The final test introduces a wave of tourists.
Neer 2
Figure 11; Impact of tourism on capacity expansion.
In Figure 11 above Line -1- represents the original no A/C unit and no tourism
condition for comparison. Then, a seasonal, varying but steadily increasing wave of
tourists is introduced at the beginning of the simulation along with the new energy
service (in its original rating mark). This situation is represented in Line -2-. The
situation seems to be spiralling out of control with very short and erratic capacity
additions. Allowing the intervention of more efficient A/C units to take effect, there is a
noticeable impact on timing of the expansion (i.e. deferral) and less on scale as sketched
in Line -3-. However, the system still expands well beyond what can be intuitively
considered an economic sense in comparison to the initial situation (Line -1-). That
confirms the design of the simulation model, which had to capture the very disrupting
role of uncontrolled tourism demand, and indicates towards the justification of great
generation costs referred to at the start of this paper. One can envisage the
mismanagement of resources being due to this inability to control such a situation,
especially when the operator’s basic institutional lever seems to be building more
capacity as in the case of the PPC!
5 CONCLUSIONS
The preliminary simulation exercise undertaken in this paper proves the great utility of
System Dynamics methodology to analyse and understand a complex management
structure with cause-effect relationships and a mixture of physical and policy-making
arrangements. The benefit of looking into islands is that simulation is attempted on an
energy system in its most basic configuration but still found in the real world. Among
the benefits of examining autonomous island energy systems are the traceable demand
and supply characteristics, the availability of historical data in a comparable format and
a straightforward local economy structure with recognisable links outside their physical
boundary.
Despite its limitations, the model raises questions and dialogue about the future of
energy policy in the Greek islands. Most significantly, it sketches a micro-world where
the links among sectors of the economy, stakeholders and exogenous factors although
simplified for modelling purposes still maintain significant validity to the real case.
Eventually, this exercise unveils the structures that can turn the islands’ energy system
into a healthy and even profit-making organism. Further research on the topic aims to
unveil a viable set of options to make that change possible to a policy-maker.
On the actual lessons learnt, despite the efficiency gains of a technology being
exogenous to an island, once such an improvement is commercially available, policy-
makers and the system operator can provide incentives to bring about the expected
benefits. Determination of the nature, timing, size and relative advantage of candidate
strategies is possible though a tool such as proposed here. One could intervene in
boosting retail entrepreneurship, draft regulations for replacing old equipment or
introduce a subsidy for substitution — a dynamic cross-sectoral model can relatively
assess the results of such a variety of measures. The tourism sector has proved to be a
major energy consumer with great impact on the size and cost of the system despite its
periodical appearance.
REFERENCES
1. ALTENERII. 2001. Interim Report on the Sustainable Energy Project for the
Economic Development of Remote and Isolated Island Communities.
2. Balaras, C. A., M. Santamouris, D. N. Asimakopoulos, A. A. Argiriou, G.
Paparsenos, and A. G. Gaglia. 1999. Energy policy and an option plan for
renewable energy sources (RES) for the Hellenic islands of the North
Aegean region. Energy 24, no. 1999:335-350.
3. Haralampopoulos, D. A., P. Fappas, M. Safos, and H. Kovras. 2001. The structure
of residential energy use on a North Aegean island: the town of Mytilene.
Energy 26, no. 2001:187-196.
4. Mihalakakou, G, B Psiloglou, M Santamouris, and D Nomidis. 2002. Application
of renewable energy sources in the Greek islands of the South Aegean Sea.
Renewable Energy 26, no. 2002:1-19.
5. PPC. 2002. Annual review programme of the island autonomous power plants
(2001). Athens: PPC.
Back to the Top
