Energy Equipment Diffusion & Touristic C ompetitiveness:
Building of an SD Model for the Greek Islands
Paraskevas G ravouniotis, Ausilio Bauen
Imperial College London
RSM Building, Prince Consort Road, London SW7 2BP
T: +44(0)2075947309 / F: +44(0)207594375
p.gravouniotis@imperial.ac.uk
ABSTRACT
The real-world problem the research aims to address is the continuing highly seasonal,
exponential electricity demand growth in the Greek islands that are unconnected to the
national electricity grid over the past decades. This paper presents only part of the on-
going research. It specifically tests an early draft of the sub-model concerned with the
interplay of an island’s tourism volume & attractiveness, local technological learning-
by-using effects and the dynamics of demand-side equipment diffusion. The general
assumption is that a tourist chooses a basket of services received at the place visited,
one of which is cooling comfort. Cooling-comfort eventually translates to installed
cooling capacity and in effect electricity consumption. This paper examines the sub-
model which, based on a figure of cooling comfort per person, constructs an indicator of
competitiveness to similar destinations and relates the flow of tourists to it. Similarly, a
cost comparison incorporating a learning curve between a conventional and an efficient
variant of cooling equipment drives the installation stocks at any time and effectively
alters the efficiency of the overall service across the island. The sub-model is run for a
number of structural and behavioural tests and also assessed for its potential use in
policy making.
Keywords: islands, diffusion, substitution, learning, system dynamics, tourism, Greece, niche markets
1 INTRODUCTION
11. THE BROAD RESEARCH QUESTION & BACKGROUND
The Greek grid-unconnected, or autonomous, islands are almost entirely dependent on
stacks of small to medium size thermal power units. The exponentially growing
consumption trend and great demand variation between the extended off-peak winter
season and the energy-intense but short peak summer season means these already
inefficient engines are running either below the recommended load, if at all, or at peak
emergency rating. These factors drive the unit price of electricity in the islands to costs
up to tenfold the cost in the mainland where the generation is based on cheap local
lignite. The tariffs across the country are uniform thus electricity in the islands is
heavily subsidised leaving the Public Power Corporation (PPC) and its Islands
Directorate with a burgeoning debt. Despite numerous studies on the great potential of
renewable energy technologies on autonomous islands, these have never significantly
picked up due to lack of a consistent support policy, various land use conflicts, a
liberalized but unclear and stagnating energy market and the intertwined interests of the
influential refinery & shipping lobbies.
The intervention envisaged is on the contrary looking on the demand-side of the
problem in a structured cross-disciplinary fashion. How could energy policy makers
foster the great entrepreneurial potential on the islands in order to initiate a self-
propagating demand for energy efficient equipment? How would such a market, on one
hand reduce the costs of power generation in the islands and, on the other hand, not
hinder the local economy heavily dependent on providing a competitive touristic
product? The key aim of the broader research is to assess whether the diffusion of
energy efficient demand-side equipment can significantly reduce the rate of capacity
expansion in the Greek islands beyond a BAU future, look into the right levels of
financial support and comment on paybacks for the policy-makers and adopters. To
date, there has been no concerted and sustained action for DSM in the country or the
islands in particular.
12 OBJECTIVES & SCOPE OF THIS PAPER
The intervention is studied by means of a dynamic non-linear simulation model. This
paper is focussing on the working draft of the sub-model relating tourism growth to
technology diffusion, equipment substitution, destination competitiveness and
technology learning. In the hope of constructive feedback by the readers, what is
presented here are the formulating assumptions of the diffusion sub-model, its operating
principles and analyses of test runs of its behaviour under a number of situations
including the effects of sample policy interventions. This paper is not conclusive to the
broader research question as described previously. Rather, its objective is to produce a
meaningful and island-specific diffusion simulation model that can stand on its own. It
shall be later linked with a bottom-up demand profile generator, a utility costing sub-
model, an intervention scenario selector and an appraisal tool to provide an insight to
the pressing energy problem of the Greek islands as set in the introduction.
The causal loop diagram (CLD) in Figure 1 summarises the approach to the design of
the diffusion sub-model, demonstrating the feedback loops simulated.
There are two dominant loops in the CLD of Figure 1:
LOCAL SERVICE COVERAGE > COMPETITION GAP >... ADOPTIONS... > LOCAL SERVICE COVERAGE
The size of the total adoption of the service provided by the equipment in combination with tourist
arrivals provides a figure of the coverage, i.e. percentage of visitors receiving the service. This figure
compared to an international average determines the urgency to adopt either of the two variants of the
equipment and feeds again back to number of units installed.
Cost DIFFERENCE > EFFICIENT EQUIPMENT ADOPTION > LEARNING > EFF EQUIPMENT COST >COST DIFFERENCE
The older equipment is a mature technology and has a stable cost and price. On the other hand, the more
the installations of the efficient equipment, the more the learning effects. The cost is steadily declining,
reducing the cost difference thus making the efficient technology increasingly competitive. A promotional
scheme as can be seen in the diagram could aim at increasing the learning effect.
4
Tourism Carrying Tourism Arrivals
Capacity
+
Total Adoption Local Service””- Regional Average
v¢ Coverage Coverage
aes
-We
Bi Competition Gap
40“ Older Equipment :
Adoption <%|
Efficient Equipment __-
+ Adoption
+ +
Cost Difference
+
+
Learning
oe Efficient Older Equipment
- Equipment Cost Cost
Promotion Scheme
Figure 1: The approach to diffusion modelling in CLD notation
The next section (2) briefly introduces considerations on tourism, its relevance to the
research question and influence on model design. Section 3 is reviewing the major
parameters used in the model, followed by section 3.5 that illustrates the model’s
mechanics through a simplified model run. Section 4 is testing the structural validity
and robustness of the model’s responses under various test situations. Finally, section 5
is summing up the task undertaken in this paper, draws conclusions on the sub-model
and reveals future steps.
2 TOURISM, ENERGY USE & COMPETITION
2.1 TOURISM IN GREECE
According to a 2004 Financial Times article, tourism is identified as a major income
source of Greece, accounting for about 18% of the national GDP and employing over
15% of the workforce (Hope 2004). The FT reporter finds out in a series of interviews
that apart from better touristic promotion, the country needs to provide better services to
its visitors, as it cannot afford to put that sector in peril. The Greek National Tourist
Authority has drawn up plans for the sustainable growth of the sector (Box 1).
Box 1: GNTO's tourism development plan
A strategic plan for tourism development has been elaborated by GNTO, in the framework of the
“National Plan for Regional Development 2000-2006”. This plan (Operational Program for Tourism)
takes into account all relevant environmental concerns and it enhances specific actions towards a
sustainable tourism development. The main objectives of this plan are:
= Upgrading the quality of tourist services;
+ Elaborating environmental protection projects;
+ Encouraging the wise use of water and energy;
= Modernizing equipment and installations of tourist establishments; and
= Promoting cultural tourism as well as eco-tourism, mountain tourism and other forms of alternative
tourism.
This plan is elaborated through cooperation with regional authorities, local stakeholders and the private
sector.
Source: United Nations Conference on Environment and Development 2002 Country Profiles
2.2 COMFORT, SPACE COOLING & ENERGY DEMAND
The World Trade Organisation (WTO) states that “a key element in leisure travel
demand is the degree of comfort (or discomfort) to be experienced at the traveller’s
destination” (WTO and Todd 2003). It is mentioned that comfort is harder to maintain
at air temperature exceeding 31°C that is the norm in the Greek islands during the peak
summer season. This paper assumes that the ‘comfort factor’ constitutes a major
attractiveness of a Mediterranean destination assuming that they compete on similar
levels of scenery, cultural heritage, bathing facilities and cuisine.
Air-conditioning has also been recognised as a main culprit contributing to extreme
demand peaks in Greece (Daskalaki and Balaras 2004:1091-1105). It does also connect
conceptually and practically to the comfort factor requirement to assess competitiveness
of an island as air-conditioning is the technology of preference providing that service.
2.3. THE COMPETITION HABITAT OF THE ISLANDS
In the simulation, a regional competition indicator on cooling comfort coverage affects
the rate of adoption between the two A/C equipment variants. Among other things the
success of technology diffusion includes keeping as close as possible to a regional
average of cooling service coverage. Is the assumption of a regional competition
average valid? Can one assume all other factors affecting destination attractiveness
constant? What are the geographical boundaries for the aims of the research?
The WTO has reported that there is a mass movement of people with the intrinsic
purpose for travel to visit a sunny seaside destination (WTO and Todd 2003). For 2002,
a total of 133 million arrivals have been registered to the northern coast of the
Mediterranean and the Caribbean. This is clearly the market the Greek islands are
competing in: population flows where the weather is evidently of paramount importance
in much of the leisure travel as opposed to the destinations’ cultural heritage.
Out of the figure quoted above, 116 million arrivals are concerning flows from Northern
Europe to the Mediterranean’s South European coast and islands alone. This refines the
competition environment even further. The activity exhibits a 3% growth rate per
annum and its market worth was US$70 in 2000, projected US$300 by 2050 (WTO and
Todd 2003). Thus, the Mediterranean basin’s boundaries can be adopted as those of the
simulation, which being also an area of homogeneous climatic characteristics enhances
the validity of a regional competition indicator. That is to say:
a) The similar temperature profile and seasonality of the tourism wave indicate an
equally similar visitor expectation and cooling comfort demand in the region.
b) Attractiveness of Mediterranean destinations balance along complementary
elements of hospitality, climate, culture, gastronomy and natural beauty, thus
one can safely assume all other factors constant across that economic habitat
when comparing performance on cooling comfort.
3 ELABORATING ON THE MODEL’S PARAMETERS
3.1 SERVICE BENCHMARKING & COOLING CAPACITY
A standard reporting framework for air-conditioning consumption in tourism does not
seem to exist in international bibliography yet (WTO 2002;WTO 2003;WTO and Todd
2003). Specific statistics on consumption of electricity for air-conditioning on the Greek
islands do not seem to exist either. As a matter of fact, there has been an effort in the
1980s to measure up energy consumption and use patterns in the islands. The
programme reached only the residential sector and was abandoned as early as 1988 with
limited output available. To overcome the lack of statistics on A/C consumption, the
demand profile is build from scratch. Each visiting tourist is ‘credited’ with a peak
demand made up of typical electricity consuming activities for the duration of a visitor’s
stay.
The air-conditioning load of a person comprises of two parts: a) the sensible heat load,
i.e. removing heat to reduce temperature, and b) the latent load that has to do with the
dehumidification necessary when hot air is removed from an enclosed space. Each
person generates 75W of sensible heat and 60W of latent load in sedentary occupation
(Stephenson 1968, CBD-105). In order to confirm that figure as the per person cooling
load suitable for the Mediterranean, papers from France (Cron, Inard, and Belarbi
2003:41-52), Italy (Gugliermetti, Santarpia, and Bisegna 2001) and Turkey (Giirbiiz
2001) were consulted. The warmer climate suggests that the latent load is higher thus an
aggregate of 200W/person is adopted. Based on that, the necessary share of size of
equipment for that load is about 700BTU/person. Assuming each tourist has 10m’ to
move in, a 4,000 BTU unit is suggested making broad assumptions on the number of
windows, orientation of the building and insulation among other parameters'. The
Energy Efficiency Rating (EER) of an A/C unit is its BTU rating over its wattage and is
widely used in the USA. The higher that number is the more efficient the unit.
Assuming an EER of 8, the equipment capacity necessary is 500W/person. A higher
* http://www.purityplanet.com/air-conditioner-sizing.aspx &
http://personal.cityu.edu.hk/~bsapplec/cooling.htm [15/08/2004]
efficiency unit will have a higher EER, i.e. a EER 10 unit will require 400W/person —
20% decrease in required installed capacity per person. On the contrary, an EER 6 unit
requires 670W/person for the same load.
To evaluate a regional average, the coverage of service should also be considered. It
shall be assumed that in the average case there is 40% coverage based on touristically
developed destinations in the Mediterranean where large proportion of visitors reside in
hotel complexes offered through package holidays.
BTU/person + average EER_of equipment installed * (coverage%/100)
In this example: 4,000BTU/p + EER8 * 0.4
Therefore, the average installed capacity of air-conditioning per tourist for the given
cooling load comes to 200W/person. The figure shall be the regional competition
average towards which the island needs to keep as close to as possible in order to be
comparable, in terms of cooling comfort, to other destinations in the region.
In a study of room air-conditioners in Europe by the European Council for an Energy
Efficient Economy (ECEEE), the cooling hours where found to be 1.5 to 2 times more
in Mediterranean countries than the average. All sectors were found to require cooling
above 500 hours/year in all sectors (commercial, domestic, office, hotels). For the study,
the average of the weighted averages of all sectors in Spain, Greece, Portugal and Italy
will be adopted — i.e. 1,023 cooling hours per year.
3.2 THE SIMULATION MODEL & ITS MECHANICS
The methodological approach to capturing the interrelation among tourism arrivals,
cooling service coverage, equipment substitution, and destination competitiveness is
based on two major assumptions”:
a) There is an expected regional, i.e. Mediterranean, cooling comfort average to
visitors (measured in BTU) based on cooling wattage capacity per visitor per
area occupied. The model assumes that all competing destinations must be as
close to that average as possible to attract visitors.
b) The air-conditioning load can be met by two variants of cooling equipment
averaging at distinct values of rating and efficiency. These are assumed to be a
conventional electric A/C unit opposed to a hybrid solar absorption chiller; the
former being more power consuming than the latter albeit cheaper to buy. The
latter has overall more desirable characteristics and the purchase/installation
price is the only deterrent to potential adopters of the technology.
Figure 2 presents a simplified stock-and-flow diagram’ of the diffusion model. What is
not shown in the diagram is the technology-learning loop. This is a separate module that
? The assumptions as well as the simulation model later are built with a single visitor (tourist) as
the base unit of the model metrics. That has been decided to keep in line with WTO practices
that defines a visitor “as a particular type of individual consumption unit, who is distinguished from other
individuals by the fact that he/she is outside his/her usual environment and travels or visits a place for a
purpose other than the ‘exercise of an activity remunerated from within the place visited’’ (WTO 2004).
is not shown here for space economy: the module receives data from the two-adopter
pools, runs a learning algorithm based on the number of installations of each equipment
variant, and then returns this to the cost ratio.
tourism carry ing capacity
2
total adojters
9
|
|
|
rt of realighhg Serv need
potentjal adopters
fr of industry realisiry need
i=
TIME TO REALISE SERVICE NEE!
ELAC discardel
IAC out of business
cris alert end of\iife
out of business
SRAC adoption rt
Figure 2: Diffusion with seasonal tourism and two adopter stocks
A seasonal tourist population appears that has a ramp and random function combined to
generate a fluctuating, seasonal and growing tourist wave. The outcome is shown in
L:-TOURISTS 2: tourism annual peak
6000
109 120.80 240.60 360.40 480.20 600.00
Months
Figure 3: The seasonal tourism wave
Figure 3 where Line | (in red) are
the tourists arriving at any time
given the seasonal pattern while
Line 2 (blue) is effectively the
hosting carrying capacity of the
island. Line 2 represents the stock at
the top-left in Figure 2 and logs
successive peaks of visitors during
the simulation when there is a net
increase. If this year’s visitors are
less than last year’s then the blue
line remains straight. If however,
there are more visitors this year
there is a step increase to accommodate for that demand.
3 In the stock-and-flow notation, the square boxes represent the quantities that accumulate and
perform an integrating function, the valve and double line combination represent flows and rates
of change while the remaining circles are auxiliary functions containing variables and constant
parameters. The single line arrows show where this auxiliaries are used. Valves, auxiliaries and
less commonly stocks, include formulas.
The two variants in the adopter stocks of Figure 2 are ELAC (ELectric Air-
Conditioning) and SRAC (SolaR Assisted Air-Conditioning) whose efficiencies are
relatively related as SRACeq=a*ELACepr where 0<a<1, i.e. the former technology is
more efficient than the latter by a factor of « which can be a constant or a variable.
There are three evident and one concealed stocks on the diagram that share the carrying
capacity at any moment. Referring back to Figure 2 these are the stocks of the “potential
adopters”, the “ELAC adopters” and the “SRAC adopters”. The resulting figure when
subtracting the sum of the three stocks from the hosting carrying capacity of the island
is the concealed stock of visitors that will not experience any space-cooling service
during their stay. The mechanism by which the carrying capacity expands is not
explicitly modelled here as it is of no immediate effect to the modelling exercise and the
research question. It can be assumed to be a forecasting of some sort that allows the
commercial sector to absorb any hosting demands and then maintain that level.
3.3. THE COST COMPARISON
The cost ratio (Co,p /C ygw ) consists of the price paid for the installation per unit of
marginal installation and the running cost aggregated for five years; i.e. a safe payback
period for the Greek islands (Betzios 2003). The ratio does appear and parametrically
affects the flows of population to the adopter pools of the model (Figure 2). A fraction
of the commercial agents servicing the carrying capacity of the island are aware of the
proposed, or improved variant, and are willing to install it if conditions are right
(“fraction of industry realising need” in Figure 2). These agents consequently control a
share of the carrying capacity that ends up in the “potential adopters” pool.
3.4 THE COMPETITION INDICATOR
But do all potential adopters eventually install one of the two variants based just on
relative costs? That exactly is the purpose of the loop that compares the quality of
service to a supposed regional average. A commercial accommodation owner might
wish to upgrade his services to customers but will not do so if the standard is acceptable
for the type of destination and clientele the island appeals to. Furthermore, the model
has that “fraction of industry realising need” variable that is meant to leave out smaller
family-run accommodation owners who only provide basic services, bed and bathroom,
and marginal shops.
When the local cooling capacity per tourist is above the regional average, it is suggested
that it makes the basket of services provided by the island marginally costlier than
competing destinations. Whereas when it is below that figure then the quality is not up
to standard. The competition indicator is based on the amount of service required, i.e.
cooling capacity installed in BTU per person per area quota. There could be a reduction
in power consumption simply by degrading service quality but that is not deemed a
desirable policy or sensible commercial practice. On the contrary, the success would be
to maintain the service standard by reducing the power input required for it. For each
tourist to receive that service there is an installed capacity to generate the cooling load
required. Introducing a new power-saving technology, it would allow the energy
expenditure to drop while maintaining or even increasing service coverage.
3.5
1: Potential Adopters
1500.
2: consumption per tourist
23
4B
2: compettiveness index
1
4c
DEMONSTRATION OF THE MODEL’S OUTPUT
2: ELAC adopters 3:SRAC adopters
y\I—
12080 240.60 7360.40 480.20
Months
2 international average
e000
1280 Pry 350 8020
Months
Unites
e000
~~]
12080 20060 36040 120.20
Month
Figure 4: Test run of extended diffusion
600.00
This paragraph aims to familiarise
the reader with the model’s output
and parameters through a simple run.
Each year a number of visitors arrive
on the island while facilities expand
to accommodate that demand.
However, for illustration purposes
the following assumptions are held
for the model in Figure 2:
+ “Tourists” have a steady flow of
2,000 throughout the year without
any seasonal peaks.
« The “fraction of industry
realising need”, set to 1 signifying
all commercial premises are to be
space-cooled if the competition &
prices permit.
+ The cost ratio Coiy /Cypw is set
at 0.20 defining the relevant flow
from potential adopters to either
of the variant stocks in a constant
manner.
« The consumption of SRAC is
half that of ELAC, i.e. a=0.5.
+ The learning rate in the
background is not affecting the
system since the cost ratio is
stable. The system thus adjusts on
competitiveness performance
alone.
Figure 4 presents the ‘phase diagrams’ for a series of model parameters (indicated at the
top left of each graph). At the start of the simulation, the competition indicator has
maximum value in graph (4.C) since there are no units at all to provide the cooling
service thus the competitive imperative to install is high. On the top graph (4.A), the
arrival of tourists (Line 1) exerts system pressure and eventually the consumption per
tourist overshoots the regional average, illustrated on graph (4.B)/Line 1. Once there is
the need to install, ELAC adoption grows faster since the price favours it on graph
(4.A)/Line 2. By the 5° year of the simulation (60 months), there are 751 tourists in
ELAC and 312 in SRAC. Also, there are 875 people in the “potential adopters” pool
that cannot yet enjoy the service despite their hosts having realised the benefits of the
installation but are yet to install. The three numbers together total 1,938 people.
By year 10 (month 120 in the simulation graphs), these three numbers changed to 388,
312 and 1,300 respectively. The sum now is 2,000, i.e. as much as the carrying capacity.
That is the effect of delays in the system such as the "time to realise need" auxiliary
function that defines the delay the tourist industry service providers need to make up
their minds on the usefulness of the service, and the ‘adoption delay’ in the model that
represents the time it takes for commercial adopters to actually purchase the equipment
once they have justified its superiority.
There are a number of qualitative observations to be made. Once the system overshoots
the international average (4.B) in its momentum to live up to the expectations, there is a
similarly steep decline in a goal seeking type of behaviour that would be expected from
such a system. In the uphill period both adopter pools grow rapidly until roughly month
48, 4 years into the simulation. The system then has to decrease its consumption. The
only outflow from the consuming stocks of the adopters occurs in ELAC and is due to
the equipment reaching the end of its life and being discarded. There is no replacement
as it can be deduced from the flat curve of SRAC (Line 3) in (4.A) for quite some time,
just more people flowing out of ELAC that is steadily reducing.
The curve in (4.C) confirms this through the competition comparison. Around the 4"
year, the indicator becomes negative ceasing any flow to the two stocks. Qualitatively,
this reads as follows: Commercial agents even though realising the need and having
moved the share of carrying capacity they command to the “potential adopters” stock,
do not feel any pressure from competition since a falsely sensed alignment to the
average presides over. Similarly, those that did discard their equipment do not believe
there is a need to replace it immediately; thus the released tourists return to the
“potential adopters” and stay there until the drivers for adoption are ripe again. Policy-
makers can easily misinterpret this situation as equilibrium since it is statistically
observable despite the Greek government not gathering the specific data at present.
Back to graph (4.B), consumption continues to decrease as more equipment is taken out
of the system and consequently the system now undershoots the regional average. It
takes time for the competition indicator to rise again due to sampling and reporting
delays in its estimation. It takes about four years from the first time the local cooling
capacity is found under the average until the indicator gains a positive value (4.C) that
effectively allows adoption to resume (4.A — appr. 180 months). This delay has
accumulated in what can be referred to as pressure to adopt in the system, therefore the
slope increases again, albeit of lesser magnitude, in ELAC and SRAC adoption. That
draws people heavily out of the potential adopters’ pool. The oscillations caused by the
delays in the system ultimately die out and the system reaches equilibrium with the
competition indicator settling close to the average with a healthy replacement rate for
SRAC (4.A).
4 TESTING THE MODEL’S BEHAVIOUR & APPLICATIONS
This section is running a number of scenarios to validate the structural logic and assess
the behavioural sturdiness of the model. The aim here is not to produce conclusive runs
since this is just a sub-model in the overall research but assess the usefulness of the
multi-disciplinary diffusion sub-model and its success in dealing with the required
specs. The testing is performed under three critical headings:
10
1. Validity — aiming to ensure model produces logical behaviours
2. Robustness to confirm it can adjust to possible changes of conditions
3. Policy-making — to evaluate the model as a tool for testing strategies
Each section of the test runs draws policy making hints for decision makers.
Figure 5 is a service sector diagram of the model showing its conceptual components.
For robustness in extreme events, a function that allows closure of commercial premises
in prolonged periods of arrivals being less that the carrying capacity has been added.
That is the only case when the carrying capacity is allowed to reduce. At the same time,
the efficient variant is experiencing learning effects leading to reductions in price as
installations grow (assuming an 82% learning rate). A ‘promotion scheme’ can be seen
that can be designed to provide a single programme of installations over a period of
time or a series of annual installations during the course of many years.
Industry Crisis Monitor S77 PROMOTION SCHEME <7
Tourist Canying Capacty <7 ‘in Adoption Module <7
Uaaring op (con comparison 7 Compativeness Atumement <7)
4
Figure 5: Sector diagram of the diffusion model
11
4.1 VALIDITY : TESTING STRUCTURAL BEHAVIOUR
4.1.1 INFORMATION DELAYS & SYSTEM CONTROL
There is a significant amount of time from the moment a survey is taking place until the
statistics are published and eventually reach decision makers in the private or public
capacity pertourst: 1-2 sectors. In the diffusion model,
“ the existence of a _ regional
average of cooling capacity per
visitor is assumed to require
three years (or touristic seasons)
from the time it is collected
subsequently processed,
checked and published until it
reaches the island’s commercial
“Tho ade Dathoo 3600 78.20 e000 sector through national tourism
: Months PCA : :
institutions and media. The main
tora adopters: 1-2 consequence of delays in case
2000
studies of the System Dynamics
bibliography is the creation of
oscillations in the system
(Forrester 1961;Sterman 2000).
aon This behaviour is met in the
model as can be seen in Figure 6
(sensitivity type graphs, i.e. one
graph depicts a single parameter
Ton wale Tao sao va aoa ON Separate test inputs).
Monts
Figure 6: The oscillations due to information delay Line 1 (blue) in both graphs
represents the case where the
comparison of the local to the regional capacity performance is done almost real time on
a monthly basis. It is noticed the system is very well controlled. On the other hand, Line
2 (red) represents the three-business seasons delay where the momentum gathered is not
properly adapted when the goal is eventually reached and it overshoots the regional
average around which the blue line evolves. The bottom graph of Figure 6 sketches the
total adopters, i.e. the market, and indicates a market with booms and busts for Line 2
(red). A downhill curve signifies a stall in purchases as equipment reaching the end of
serviceable life exits the adopters stock and there is no replenishment.
In the real world, great variations between business cycles will deteriorate efforts of
policy makers to establish an efficient & controllable market. An objective of the policy
makers should therefore be to reduce information delays in critical positions of the
system and be as close to the ideal behaviour of Line | (blue).
12
4.1.2
THE TIME TO RESPOND & OSCILLATIONS
Once the information reaches the commercial operators of an island, there are more
capacity per tourist 1-2-3. 4
400
200:
total adopters
2000.
1000
100 120.80 240.60 360.80 480.20 600.00
Months
1234
00 120.80 240.60 360.40 480.20 600.0
Months
Figure 7: The impacts of operators’ reaction
delays to be faced in their
reaction to the news. In the
model the reflex gain has
been formulated to vary
depending on the extent of
the deviation from the
regional average. The lines 1-
4 in Figure 7 (sensitivity
graph) represent test runs of
an increasing = =minimum
reaction time threshold —
from | to 36 months (Line 1:
1 months, Line 2: 12 months,
Line 3: 24 months & Line 4:
36 months). The impacts are
easily observed despite the
effect dying out in all cases as
the gap narrows and
adjustments become
marginal.
In real life Line 4 (green)
reveals a sector that is very
restrained and cautious in
adopting a new technology.
On the contrary, Line 1 (blue) would represent a risk-taking group of commercial
operators.
Despite the apparent ability to control the oscillations as in the previous case, the access
of policy makers to this parameter is limited. The perceptions of risk in economic
sectors are highly subjective and peer behaviour can only be changed through years of
stable policy aiming to influence those perceptions. The final model will adopt the one-
month response time that characterises risk-taking agents and competitive economies
such as
those of the Greek islands.
13
4.1.3. DIFFUSION & REPLACEMENT OF EQUIPMENT
Graph (8.A) of
Figure 8 illustrates the dynamics of replacement between the two equipment variants in
Line 3 (blue) and Line 4 (green). Initially, there is only the early power-consuming
variant (Line 3). At some point around the course of the pe year a scheme is introduced
that sees the installation of 200 efficient units. The size of the scheme is such that
causes the cost of the new equipment over five years (purchase and operation) to drop
below that of the conventional technology as observed in graph (8.B). As soon as
installations commence, SRAC is the favourable choice therefore its rapid growth
(8.A/Line 4) and the parallel decline of the ELAC stock.
What is the impact to the system however? Graph (8.C — sensitivity) compares the
scenario with (Line 2 — red) and without (Line | — blue) the introduction of the efficient
variant. The number of total adopters, i.e. visitors who enjoy the cooling service, has
increased dramatically in the former case. The improvement in coverage is achieved in
line with the competition requirements of the regional cooling capacity per person as
can be confirmed in graph (8.D - sensitivity) between the two scenarios which has been
estimated to be at 200W (par. 3.1).
30 | =o ot.
‘y
2620 | 100
ee
Too Tia zadlso Seo ery e000 Too Tao Er) ao Tazo 0000
Monte ont
LELAC Sycost capacy pertouret: 1-2
8.D
Too Tae Ero a Tab cor)
Figure 8: Observing the competition of equipment
4
4.1.4
UNIT RATING AND SERVICE COVERAGE
sewice coverage: 1-2-3
0
i
>
ee. IDNVOAIN AA
°
Too Tate roy Thee wn aan
mons
uti cama cap BELAC adore 4 SRAC atone
20 Le
=
2620 ~ om
by?
ae
Loo 120.60 240.60, 360.40 480.20 600.00
vont
LELAC rest 2 5RAC Sr cost
73685
ane.
2000
% Tate ro aie win ain
Months
Figure 9: Assessing the sensitivity of efficiency gain
The simulation runs in the
previous paragraph assumed the
new equipment needed half the
wattage rating to provide the
required cooling load. In Figure
9 (9.A) examines the
sensitivity of service coverage
for a range of SRAC ratings —
namely 4, 2 and % of the
ELAC capacity. A quick look
reveals that the relationship is
not linear; the impacts are
disproportional to the rating
step (Line | for %, Line 2 for 4
and Line 3 for %).
Taking a closer look of the last
run at the % rating in graph
(9.B), it is revealed that the
SRAC support scheme in the 7”
year failed to establish the
market (Line 4 does not pick up
after the intervention). The
reason can be found in graph
(9.C) where the five year cost
of the SRAC for this rating is
still higher that that of the
ELAC units even after the
financed installations.
15
4.2 ROBUSTNESS : TESTING RESPONSE TO EXTREME EVENTS
4.2.1 KEEPING UP WITH IMPOVED COMPETITION
So far it has been assumed that the regional competition average has been fluctuating
randomly between a given set of values. This simulation run explores the situation of a
major and abrupt service upgrade in competing destinations expressed as a step
improvement in the average regional capacity per tourist. The left hand-side column of
graphs in
Figure 10 [(10.A) and (10.B)] explore the sensitivity of the level of competitors’ service
upgrade — scheduled for month 120) to the capacity per tourist and the total adopters of
the island system under examination. The larger the step of improvement in competing
destinations (ascending from lines | to 4) the longer it takes until the local tourist sector
to respond indicating there is a period that the system re-adjusts the relevant shares of
ELAC and SRAC stocks to face improved competition. Line 1 (blue) is the reference
case at 200W per person.
The right hand-side graphs [(10.C) and (10.D)] look into the impacts of the timing of
the step improvement of competitors’ performance. The timing of the step decrease is
not showing any unexpected behaviour. The regional average is accepting the same
service upgrade each time at 100 months interval for each line from | to 5, Line 1 (blue)
being the reference case again. The response is immediate and after few cycles the
relevant values balance around their new reference state. In both cases, it is observed
that the system adjusts its competitiveness by reducing the number of total adopters of
the service. In the real world, that would reduce the coverage of the service and may
alter the island’s attractiveness. This is valid danger for a tourist destination that has not
capacty pertourst: 1-2-3 4 capacty pert: 1-2-3-4
400
20 " | 200 rs oe aad
FREE ivaneal
;
pre)
; Ir
Figure 10: Exploring a changing competitive environment
16
been upgrading its services in line with the competition. Policy-makers being pro-active
on their strategies can alleviate such situations and assist the local economy to be
flexible and agile.
4.2.2 INDUSTRY CRISIS : FACING REDUCTION IN ARRIVALS
The tourism industry is quite volatile (WTO 2003;WTO 2004). Despite the 2004
Olympic Games in Greece, the tourism industry had a very bad year overall due to bad
press due to expectations of preparedness of the country to host the Games prior to the
event, the threat of terrorism and the inflation impacts of the Euro zone (Ktenas 2004).
The model has been design to confront prolonged periods of tourism crisis by allowing
the exit of commercial establishments from the market thus reducing the available
cooling capacity for the tourist population.
Starting from the top left corner of
Figure 11 a sudden crisis occurs in the tenth year (120 months) of the simulation.
Arrivals are kept low for a period of fifteen years after which a new era of touristic
popularity begins as portrayed in graph (11.A). The occasion of an industry crisis is the
only situation when the model allows the carrying capacity to scale down as in graph
(11.B/Line 1). The delay observed until the carrying capacity reduces, in month 180, is
due to an internal loop that first needs to confirm that the crisis is a persistent
phenomenon. Similarly, the relevant stocks of the potential adopters and ELAC/SRAC
adopters adapt to the new market conditions.
As expected, keeping the same facilities for a much smaller tourism population initially
soars the local capacity per tourist availability as Line 2 (red) indicates in graph (11.C) —
Line | (blue) being the reference case. The system responds and eventually stabilises
around the regional values in blue (11.C & 11.D) albeit a long period of abrupt
:TOURIST ARRIVALS capac perturst: 1-2
Fall
term cacy cp. SELAC adopters 4 SRAC adopters sence coverage: 1-2
Le wet
oe
Sa |
7
Figure 11: Confronting a prolonged industry crisis
adjustments (red line). The corrective actions of the model return the system to it
previous competitive state. What is not described here however is the impact of a crisis
in the planned and existent power capacity of the island. Such issues shall be examined
in a follow up paper.
2620.
4: SRAC adopters
3: ELAC adopters
2620:
T2080 240.60 3600
Months
e020 600.00
3: ELAC adopters 4: SRAC adopters
a |
|
aa
2620.
2620.
4 -—— a
360.40
Months
120.80 240.60 480.20 600.00
S:ELAC adopters 4: SRAC adopters
an |
|_|
|
anying cap
T2080 20.60, 36080 8020 00.00
Months
4: SRAC adopters
La |
3: ELAC adopters
alll
fo
jf
La
T2080 20.60 36040
Months
Figure 12: Assessing promotion plans
"8020 00.00
4.3 POLICY-MAKING:
TESTING UTILITY IN
DECISION-MAKING
4.3.1 EVALUATING PROMOTIONAL
STRATEGIES
What is the best promotion strategy
for new equipment? How do policy-
makers know which alternative
scheme will work? The sub-model
gives the ability to design a
promotional policy albeit not yet
containing the costs involved or
potential sources of funding.
Qualitatively though, the demands of
a programme can be observed
through the graphs in Figure 12.
Despite attempting the introduction
of the new equipment in graph (12.B)
as denoted by Line 4 (green), there is
practically no impact compared to the
reference case represented in graph
(12.A). The technology fails to break
into the market, as the cost effects are
not significant enough to establish
the viability of the new equipment
variant despite the planned
installations set to 25% of the
dominant variant at the time.
However, raising the installations to
roughly 28% puts the appropriate
learning effects in motion and the
new technology has a_ startling
development as described graphically
in (12.C).
The final graph (12.D) depicts an
alternative policy design. Instead of a
single large-scale demonstration
scheme, the policy-makers decide on
18
an annual small-scale installation scheme as low as 6% of the installed variant in the
first year. The consideration behind this is the smaller budget required and a fractional
approach that would allow closer monitoring of the procedure. The choice between the
two and a number of other potential schemes will depend on conditions of available
funds, institutional organisation and commitment, and the financial profile of the
commercial actors involved.
The choice will also depend on the rapidness of substitution desired as demonstrated in
the sensitivity graph of Figure 13, where lines | to 4 represent each of the cases from
(13.A) to (13.D) above comparing their service coverage. Line | (blue) is the reference
case of (13.A) and Line 2 (red) is (13.B) in the previous. Line 3 (pink) and Line 4
(green) correspond to graphs (13.C) and (13.D) respectively. Although these two
balance at similar percentages of coverage, their path is different due to the modular
approach of the latter. This shows that a possible saving in funds for the modular
scheme has to be balanced against the time to achieve the result.
service coverage: 1-2-3-4
50:
laa
own
Loo 120-80 240.60 360.40 480.20 600.00
Months
Figure 13: Sensitivity of system response to promotional schemes
4.3.2 RESTRUCTURING THE TARIFFS
This policy choice is one of the most interesting ones despite politically sensitive and
not likely to be a realistic option for the Greek islands where cross-subsidy of their tariff
is considered a social right by island populations. Nevertheless, it is worthwhile
examining how the model reacts to such an intervention. It should be expected that the
peer ern HELAC adopts 4: 5RAC adopts new efficient _—_ technology
ote | would pay back its initial
investment much quicker if
_ ae the tariffs where to reflect the
ss real cost of electricity in the
—a-
—<—| islands. Since it has lower
running costs, its market
potential will be fulfilled
i | sooner than in the reference
case.
Too 120180 240-60 360.40, 480.20 600.04
Months
2620.
Looking back at Graph (12.B)
Figure 14: Restructuring the electricity tariff where the market for the
19
efficient variant failed, a 20% increase of the tariff is introduced. The previous scheme
does now become potent and produces the desired dynamics of substitution (Figure 14).
5 CONCLUDING REMARKS
The paper has worked through the construction of a simulation model for the diffusion
of power-saving equipment in the touristic sector of the Greek islands. Drawing a
conceptual diagram (Figure 1) at the start, the simulation model has been realised
(Figure 5) following very detailed steps of conceptual design and methodology from
relevant bibliography. The model has been sized to the specific aims and objectives of
the general research laid out in the beginning of the paper. It has also been shown that
the model is able to confront a number of disrupting situations likely to arise in the
islands and can have an impact on the diffusion of a technology confirming its validity
and robustness. Finally, the model can be a useful to tool for policy making and
understanding the operation of a system seemingly remote from conventional energy
policy making as it simultaneously relates elements of tourism development, services,
technology diffusion and tariff structure.
It still remains that the diffusion simulation model is linked to a demand-and-supply
model. That will allow the discussion to stretch into utility economics and the financing
of the demonstration and support schemes. These shall be assessed in a follow-up paper
in the near future.
REFERENCES
1. Betzios, G. 2003. Islands Directorate, PPC.
2. Cron, F., Ch. Inard, and R. Belarbi. 2003. Numerical analysis of hybrid ventilation
performance depending on climate characteristics. International J ournal of
Ventilation Volume 1 HybVent - Hybrid Ventilation Special Edition, no. February
2003:41-52.
3. Daskalaki, E. and C. A. Balaras. 2004. XENIOS-a methodology for assessing
refurbishment scenarios and the potential of application of RES and RUE in
hotels. Energy and Buildings, no. 36:1091-1105.
4. Forrester, J. 1961. Industrial Dynamics. Massachusetts: MIT Press.
5. Gugliermetti, F., L. Santarpia, and F. Bisegna. 2001. Integrated energy use
analysis in office spaces. at Rio de Janeiro,Brazil.
6. Giirbiiz, A. Z. 2001. Aquifer thermal storage (ATES) in high heat load
applications. 26 2001-27 2001, at Adana, Turkey.
7. Hope, Kerin. 2004. Athens is hoping the Olympics will deliver a long-term
increase in tourism and investment and secure Greece's position as a regional
business hub. Financial Times.
20
10.
ll.
12.
13.
14.
Ktenas, Sp. 2004. The Waterloo of Greek tourism: The number of stays reduced
by 5-8% nationwide. To Vima, 8 August.
Stephenson, D. G. Heating and Cooling Requirements. CBD-105. 1968.
http://ire.nre-cnre.gc.ca/cbd/cbd-e.html [15/08/2004], National Research Council
of Canada, Institude for Research in Construction. Canadian Building Digest.
Sterman, J D. 2000. Business Dynamics: Systems Thinking and Modelling for a
Complex World.: McGraw-Hill/Irwin.
WTO. Revision of the "international standard industrial classification of all
economic activities (ISIC, Rev.3)". Responses and comment to questionnaire.
2002. Spain, World Trade Organisation.
WTO. Measuring visitor expenditure for inbound tourism (III. Proposal). 2003.
Spain, World Trade Organisation.
WTO. Measuring visitor expenditure for inbound tourism (Appendix 4. Basic
references on visitor consumption and its components). 2004. Spain, World Trade
Organisation.
WTO and Graham Todd. WTO background paper on climate chage and tourism.
2003. Spain, World Trade Organisation.
21
