Energy Policy 60 (2013) 142–154

Contents lists available at SciVerse ScienceDirect

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

A system dynamics approach for the photovoltaic energy market

in Spain¤
Santiago Movilla a, Luis J. Miguel b, L. Felipe Blázquez c,n
a
Department of Geography, University of Bergen, Faculty of Social Sciences, PO Box 7802, 5020 Bergen, Norway
b
Department of Automatic Control and Systems Engineering, University of Valladolid, E.I.I., Paseo del Cauce s/n, 47011 Valladolid, Spain
c
Department of Electrical and Automatic Control and Systems Engineering, University of León, E.II.I.I., Campus de Vegazana s/n, 24071 León, Spain

H I G H L I G H T S

The paper describes a simulation model of the photovoltaic energy sector in Spain.

The model allows to analyse dynamic behaviour of PV sector under different scenarios.

The work is directed to assist policymakers in designing energy policies.

Results show that PV energy could start to be competitive in the short-medium term.

Model structure could be applied to other countries with other parameters.

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

Article history: The goal of this paper is to contribute to understanding the behaviour of the photovoltaic (PV) sector in
Received 19 April 2012 Spain and its expectations under possible scenarios. Currently, PV solar energy is not a profitable sector
Accepted 29 April 2013 by itself. Therefore, the Spanish government, like the governments of other countries, has stimulated
Available online 23 May 2013
investment with subsidies. The spectacular increase of PV facilities exceeded all forecasts and the
Keywords: government decided to curb the trend. The present hypothesis is that continuing with this support to PV
Photovoltaic energy energy, the technological advances and the economy generated from the production of panels would be
System dynamics able to make the sector profitable in the future without the necessity of subventions. Based on this
Computer simulation hypothesis, a computer simulation model was built using the system dynamics methodology. To test its
utility, the model was challenged to fit the historical data and to explore several futures over the next few
years. The model allows an understanding of the sector's behaviour under the latest policies of the
Spanish government, thus helping to design future public policies. The simulation results are different
depending on the adopted policy and the scenario. Therefore, these factors will determine the success or
failure of the investments in this type of energy.
& 2013 Elsevier Ltd. All rights reserved.

1. Introduction 2011). This work analyses the sensitivity of different variables on

Green energy policies have experienced continually accelerat-
ing growth in the past several years. In the framework of these
green energy policies, PV electricity generation is receiving
increasing support worldwide from public authorities due to its
environmental benefits in comparison with fossil energy sources.
However, this emergence of public support has been subject of a
relatively small number of research works focused on the analysis
and assessment of these widespread and varied policies (Avril
et al., 2012; Carvalho et al., 2011; Zhang et al., 2011; Zhao et al.,
n
Corresponding author. Tel.: +34 987 293 471; fax: +34 987 291 790.
E-mail address: diefbq@unileon.es (L.F. Blázquez).
¤
This work has been developed within the project CGL2009-14268 funded by
Spanish Ministry of Science and Innovation (MICINN).
the development of the PV market in Spain, with special emphasis
on policy decisions. The method allows an understanding of the
dynamic relation between the different variables, such as public
subsidies or electricity price, and the number of installed PV
panels. The understanding of these dynamic relations is a power-
ful tool to design PV energy policies under different scenarios.
PV energy consists of the direct conversion of sunlight into
electricity through an electronic device called “solar cell”. The solar
radiation is captured in the PV panels to generate electricity (PV
effect) in the form of direct current. The performance of the panels
is the fraction of the solar radiation which is converted into
electricity. In facilities connected to the grid, this energy is
transformed into alternating current by an electronic device
known as an inverter, and discharged to the grid distribution at
the point of connection. At present, the performance of these

0301-4215/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.enpol.2013.04.072
 S. Movilla et al. / Energy Policy 60 (2013) 142–154
panels is very low (around 15%), which means that energy
obtained in this way is not cost-effective as the cost of producing
the panels is far greater than the value of their energy production
throughout their useful lives.
With the idea of developing this technology so that in the future it
can be profitable, the Spanish government, like those of other
nations, is betting strongly on this type of energy. The government
is making it artificially profitable through the use of large subsidies
paid according to the energy they produce and through encouraging
the research and development of the sector in ways that someday,
whether through the yield being higher or the costs of the compo-
nents being lower, the sector can be competitive. Furthermore, the
use of PV energy avoids the discharge of greenhouse gases to the
atmosphere. As is well known, greenhouse gases can cause global
warming and a rise of the planet's temperature with possible
catastrophic consequences for human beings. In this respect, the
Kyoto Protocol, in which Spain participates, sets emission limits for
some member countries, which must pay fines for emitting excess
tonnes. As for the emission of such gases, the energy sector occupies
a very important place, because in Spain this sector is primarily
responsible for emissions into the atmosphere, mainly due to the
plants producing from thermal energy, such as coal, oil or natural gas.
However, to account for the uncertainty and variability from solar
energy, it is necessary to have additional flexible conventional
generation capacity available in a power system with solar energy,
to provide fast ramping capability. Hence, the production of elec-
tricity from PV panels not only does not pollute, but it also replaces
some thermal energy, thus mitigating the discharge of CO2 to the
atmosphere and avoiding the payments or fines for such emissions,
since Spain is one of the countries that exceed emissions agreed by
the Kyoto Protocol (Martínez et al., 2009).
It is necessary to bear in mind that the energy required to
construct a PV panel is very large because they are made of pure
silicon, which is produced in ovens at high temperatures. How-
ever, a silicon solar panel with a regular purity can produce nearly
ten times more energy than the energy used in its manufacture.
This is because the energy needed to obtain silicon of regular
purity is much lower with respect to that needed for a silicon cell
of high purity, but the differences in PV energy production are not
very significant. As estimated, a silicon solar cell has to be
functioning between two and three years in order to “return” the
energy used for its manufacture. This is assuming that the panel is
installed in a relatively sunlit country.
The research method to be used in this context is a computer
simulation model of the main variables that have an influence on the
PV market. The simulation model is based on the system dynamics
approach (Forrester, 1961; Sterman, 2000). The method has been
previously used to model other energy systems with the purpose of
being used for decision-making tools (Castro et al., 2009; Mediavilla
et al., 2013). The main idea consists of building a computer simulation
model to reproduce the behaviour of such variables as the PV panel
price, the evolution of the panel efficiency, the PV power installed, the
subsidy value, the electricity price, the investments in PV energy, the
payback period on investments, or the costs and yield of panels in
general, over the last few years. The structure of the model is built
according to the observed relations between the involved variables.
The model has to replicate the historical data, explain the causes of the
behaviour and how it is reproduced.
The article is organised as follows: In Section 2 a literature
review and the problem description are presented, where neces-
sary background information to understand this problem is given.
Section 3 describes the research method used in this work, where
the process of using system dynamics is shown. Section 4 explains
the conceptual model description. In Section 5 the simulation
results are presented. Finally, Section 6 shows the conclusions of
the work and some references are presented.
143
2. Renewable energy policy in Spain: The photovoltaic sector
The key points of the energy policy of the Spanish government
are written in the Renewable Energy Plans (IDAE, 1999, 2005,
2011). The subsidy value and the conditions of every new PV
installation are collected in a Royal Decree (RD) that can be
updated and reviewed depending on the fulfilled goals or new
situations. This subsidy value, which is also called a feed-in tariff,
is the price per unit of electricity that a utility or supplier has to
pay for PV electricity from private generators. Once an installation
is hosted in a decree, the grants received during the whole life of
the installation are the subsidies associated to this decree, even
after its expiration. From 1998 up to the present, there have been
four Royal Decrees in total (ME, 2004; MIE, 1998; MITC, 2007,
2008).
The Renewable Energy Furtherance Plan, PFER, 2000–2010
(IDAE, 1999) included a program of financial aids for renewable
energies, including photovoltaic. From 1998, Spain began to
support PV energy and, until the year 2000, PV energy in Spain
was limited to a few experimental facilities (IDAE, 2007). In 2002
the degree of fulfilment was only 19.2% with respect to the
energetic goal of the PFER.
The PER 2005–2010 was designed with the following two main
objectives: 12% of primary energy consumption would be supplied
in 2010 by renewable energy; and for PV energy, a goal of 371 MW
in 2010 was set. In September 2007, 85% of the second target was
reached, and the government established a new temporal decree
of one year before its deadline to allow the construction of PV
facilities according to the conditions of the previous subsidy
program. That decree expired in September 2008. At the beginning
of 2008 the accumulated electric power was higher than 506 MW
(IDAE, 2008), surpassing the total forecasts for 2010. Spain's
photovoltaic capacity rose to 3404.8 MW, amounting to a newly
installed capacity of 2670.9 MW in 2008 (including 1 MW off-
grid). This dizzy growth (352% up on 2007) is explained by the
reduction in the photovoltaic energy feed-in tariff from the end of
September 2008, which prompted a rush for installations. One
installation hosted in RD 661/2007 can receive 0.44€/kW h, when
the electricity price (TMR: Mean reference tariff) was around 0.075
€/kW h, but a similar new installation hosted in the RD 1578/2008
receives only 0.32€/kW h. Moreover, this RD 1578/2008 has
limited the construction of new PV facilities with a ceiling of
500 MW installed in 2009 (including 233 MW from ground-based
power plants), 502 MW in 2010 (including 207 MW from ground-
based power plants) and 488 MW in 2011 (including 162 MW
from ground-based power plants). This new legislation effectively
curbed the growth of Spain's capacity in 2009. However, the
additional PV capacity was still increasing over the following
years, because with the subsidies of the RD 1578/2008, the PV
energy was still a profitable sector in Spain, even though less
profitable than with the subsidies of the RD 661/2007. These
additional PV capacities were 145 MW in 2009, 370 MW in 2010
and 400 MW in 2011 (EPIA, 2012; REN21, 2011).
Once a policy has been established by the government, it
should be known if this policy is suitable for the purpose for
which it was designed. If we were capable of seeing all the factors
involved in the development and evolution in the implementation
of solar energy, it would be possible to assess the decision of the
rulers who justify it, by means of a dynamic model to monitor
whether the policy has taken advantage of the chances of success
or is only a failure and a waste of money that serves only good
intentions or vested interests.
On the one hand, the development of cheaper PV generators in
the near future is a good line of research to invest in. But on the
other hand, it is also good to look at creating a PV market by
increasing the amount of PV installations, because this growth is
144 S. Movilla et al. / Energy Policy 60 (2013) 142–154
accompanied by many small innovations that, in the end, will also
produce cheaper PV generators. Therefore, the research question is
whether the Spanish government support to PV energy will be
cost-effective or not, in other words, whether to invest or not to
invest and how. At this point, the suitability of the policies taken is
essential to achieve the desired goals. The success or not of the
investments after subsidies will depend on the price of the panels,
and the price of the panels will depend on the previous invest-
ments, since investments encourage investigation in the market.
The literature review has also revealed previous research into
PV energy. Some of these documents offer information about the
market situation and the policy designs according to the country
(EIA, 2011; REN21, 2011; Teske, 2010; Zhao et al., 2011). There are
also catalogues providing technical information concerning the
different PV panels and the types of installations. In any case, no
research has so far appeared where a PV sector is modelled by the
use of system dynamics (Algieri et al., 2011; Leloux et al., 2012;
Zhang et al., 2011). Other PV publications talk about hypotheses
and forecasts where possible developments of PV energy are
shown (Carvalho et al., 2011; EPIA, 2011a, 2011b; García and
Cantero, 2005; IDAE, 2008; Jager, 2006), but they do not mention
which modelling tool has been used to simulate the behaviour, or
if they simply estimate the evolution by analysing the trend of the
curves from the historical data. Normally, these estimations
establish an optimistic and a pessimistic forecast within a range
of error, but in most cases the predictions are biased depending on
the source. However, in this research, all the collected data have
been properly treated, creating a simulation model that repro-
duces the historical information by itself. Unlike the analysis based
on inertias, where a trend alteration means a different prediction,
system dynamics uses the concept of feedback, where the para-
meters and variables have self control through a dynamic model.
Among the studies that deal with the possible evolution of PV
energy in Spain (Martínez et al., 2009; Salas and Olias, 2009), the
question about the underlying structure causing this evolution has
not been addressed yet, and it is the question that this work aims
to answer. The present research contains a dynamic model of PV
energy growth in Spain. The photovoltaic technology is in a
process of development and the potential advances will determine
the future of this energy. The framework in which these advances
are made is the key to understanding the situation.
3. Modelling and analysis method
The research method used in this work consists of building a
simulation model by a system dynamics approach. System
dynamics has proven to be very suitable when dealing with
complex systems with feedback and long time horizons. To test
its validity or utility, the model has been challenged to reproduce
historical data. Moreover, the model reproduces different policies
that simulate alternative futures that can be expected. The process
of using system dynamics follows the steps described below
(Sterman, 2000).

First it is essential to define the problem. Knowing that the PV
sector currently needs subsidies to survive, it is necessary to
analyse the profitability of the grants in order to develop
technological advances. These advances would make the con-
struction of PV installations possible in the future without
economic help. Moreover, it is necessary to analyse whether
the chosen subsidy policy is optimal for the government's
purposes and what the result could be if the policy were
modified.

Once the problem is known, it is time to define a hypothesis
about the model. System dynamics uses stocks and flows to
represent the system's evolution, it is therefore necessary to
identify the key stocks and flows and sketch the influences
between these elements. With the linked main elements, a
synoptic diagram displaying the hypothesis is generated. Start-
ing from the hypothesis, and depending on the influences, it is
possible to develop a formal model through a computing
program. The generic behaviours associated with the generic
structures of feedback loops are defined within the computer
model. In the application, the model contains the most impor-
tant variables involved in the behaviour of the Spanish PV
market. Hence, the structure links the parameters with the
appropriate formulation affecting the evolution of the stocks
and flows.

The next step is to test the validity or utility of the simulation
model. The model is challenged to reproduce the historical
data, but sometimes it does not match with the data and a
readjustment is needed. Firstly, the modeller must be sure that
the built structure is the right one and no other important
factors are forgotten. Later, it is possible that some parts of the
formulation may need a recalculation in order to be validated
once more. This process is repeated again and again in a loop,
which is an essential part of the modelling process. The loop
ends when the simulation fits the historical data and, conse-
quently, the model is validated. For this task, it is helpful when
the model is divided into subsystems, where the associated
variables are contained in each subsystem. In this case, it is
easier to identify the different structures or behaviours and the
modelling process can be readily modulated.

Once the model works properly in an equilibrium point, the
simulation is tested under extreme conditions to check the
structure's behaviour. Moreover, the sensitivity of the key
parameters is analysed to know how they affect the system.
When it is considered that the modelling process is finished,
the result is a run model which is able to understand the
development of PV energy in Spain. At this point, different policies
can be applied to check the suitability or not of the designed
strategy, depending on the simulated results. Thus, the policy test
helps the design purposes, since it gives the possibility of running
different scenarios and seeing what happens. Therefore, the model
can be used to project the behaviour of alternative future situa-
tions through the testing of different parameters: the development
of cheaper panels, the reduction of the subsidies given from the
government, the evolution of polluting taxes, energy prices, etc. In
this sense, the model is aimed towards policy modelling.
The results of a single country will evidently depend on the
feed-in tariff granted by the government of that same country. But
besides this, the global PV cell market can provide technical
advances or cheaper PV cells. In this sense, the subsidies of the
different countries are encouraging the purchases, which in turn
are helping the development of better panels. This situation could
produce a progressive reduction of the necessary grants in order to
convert a PV facility into a profitable investment. The problem is
that PV improvements are not due to a single country, but to the
global PV market in general, where there are many participants on
both sides; energy producers and cell producers. In consequence,
the study of the PV market is not an isolated problem. The results
for the analysis in Spain depend not only on the Spanish market,
but also on the interaction with other markets. It is true that, while
PV energy needs the subsidies to amortise the disbursements, the
development depends on government decisions, but the global
development is influenced by the PV cell industry and the other
countries with subsidy policies which are betting on encouraging
the sector. In the present case, the simulation model is focused on
the particular situation in Spain, where energy production is
rather more outstanding than the production of PV panels or cells.
 S. Movilla et al. / Energy Policy 60 (2013) 142–154 145

Within the model, there are several scenarios where the simula- 4. Conceptual model description
tions present different behaviours. There are different stories with
two possible conclusions. The only difference between these The hypothesis of the model is represented by a major feedback
conclusions is to see whether the investments will be worthwhile loop, as shown in Fig. 1, which contains the principal blocks
or not. The future depends on the PV investments, but it also involved in the PV market. The construction of the hypothesis
depends on how these investments influence the technological model starts with the block of ‘Photovoltaic panels’. The panels
upgrades. produce an energy flow represented in the block of ‘Photovoltaic
energy’ and this energy is based on the number of panels installed
and the installation's performance. In turn, the performance
depends on the solar radiation and the technical properties of
the panels and the facility.
Solar The PV energy produced has two directions that affect the rest
Radiation of the model: On the one hand, it produces the energy that avoids
a percentage of CO2 emissions, and on the other, it informs about
- Payback period the nominal power achieved in Spain. Although CO2 emissions do
- Installation rate - Power capacity not affect the subsidies directly, Spain is hosted in the Kyoto
- Maintenance Photovoltaic - Efficiency
protocol and the avoided emissions can produce a reduction in
Panels taxes which will influence the balance of subsidies paid by the
government.
The goals of the government subsidies have three directions:
Market & Photovoltaic The development of the PV sector until it becomes profitable; the
Technology Energy reduction of CO2 emissions into the atmosphere; and the reduc-
tion of energy dependence from abroad. The way in which the
government meets its purposes is by establishing goals in terms of
- Subsidies Government - Goals
installed power. The government decisions basically consist of
Decisions
setting up the policy of subsidies which influence the market
conditions. Obviously, the subsidies have a direct effect on the
CO2 prevented demand for installations within the market, which is ultimately
Kyoto taxes what makes the number of installed watts grow. A good way of
considering the demand growth rate is to calculate the payback
Fig. 1. Hypothesis model. period. Given the cost of the initial investment, the payback period

Fig. 2. A causal loop diagram from the ‘Government Decisions’ block.

146 S. Movilla et al. / Energy Policy 60 (2013) 142–154

is the time taken to repay the PV facility. This time will in fact be
smaller thanks to government subsidies. Keeping in mind that the
use of these aids propels the technological research, results are
expected in two other directions:

The lowering of costs. On the one hand, it is important to obtain
PV panels at a lower price as is the case, for instance, with
organic PV cells built on the basis of polymers.

The increase of panel efficiency. On the other hand, scientists
believe it is possible to enhance the efficiency of the modules in
order to exploit the solar radiation as much as possible.
‘B’ sector shows that the real subsidy granted by the government is
the difference between the subsidy and the electricity price. The
name of this variable is ‘net subsidy value’, as Formula (1) shows.
In order to ensure that the value of the subsidy is not below the
value of the electricity price, the subsidy value is the maximum
value between the established subsidy and the electricity price, as
Formula (2) shows.
Net subsidy value ¼ subsidy value−electricity price in Spain ð1Þ
Subsidy value ¼ MAXðestablished subsidy value;
electricity price in SpainÞ ð2Þ

The model was developed using the theory of system dynamics

through the software Vensims. This program is able to simulate
the system calculating all the variables and its evolution over time.
Inside the program, the model is divided into five windows
(Movilla, 2009). Four of them are subsystems that explain the
behaviour of the variables in the PV market, which are ‘Photo-
voltaic Panels’, ‘Photovoltaic Energy’, ‘Government Decisions’ and
‘Photovoltaic Market’. The fifth, ‘Historical Data’, contains the
historical data needed to validate the simulation. Each subsystem
contains the appropriate structures with stocks and flows and the
related variables. The blocks are connected to each other with the
dependence of their parameters, shaping feedback loops in
many cases.
The complete model is described by 28 causal loop diagrams
and 257 equations, and it can be found in Movilla (2009). An
example of a causal loop diagram is given in Fig. 2. It presents a
causal loop diagram from the ‘Government Decisions’ block and
shows all the variables involved in obtaining the subsidy value.
In the ‘A’ sector, the established subsidy value is shown and the
5. Results
First in this section, some of the most important variables in
the model are validated with historical data, where the time
horizon is 2009. Later, a possible projection to the future, depend-
ing on the scenario, is presented. In order to implement policies
and scenarios, the time horizon for the following runs is 2020,
because this year is also the time horizon of the current Spanish
Renewable Energy Plan PER 2011–2020. Hence, depending on the
simulated scenarios, it is possible to take optimal or suitable
decisions in reference to the PV sector of such a plan.
The validated variables are the accumulated installed power,
the annual growth rate and the established subsidy value, as Fig. 3
shows. The upper graph of Fig. 3 represents together the historical
Installed Power and the simulated growth of Power Capacity from
the model. It can be seen that both curves match with a reasonable
accuracy until 2009. Consequently, the model simulation provides
the same result as the real development. The drastic trend change

4,000
Installed power:
Installed power historical data:
3,000
MW

2,000

1,000

0
4,000
Nominal installation rate:
Nominal installation rate historical data:
3,000
MW/year

2,000

1,000

0
0.6
Established subsidy value:
Eur/kW·hour

0.5 Established subsidy value historical data:

0.4
0.3
0.2
2004 2005 2006 2007 2008 2009 2010
Time (Year)
Fig. 3. Model validation.
 S. Movilla et al. / Energy Policy 60 (2013) 142–154 147

10,000
Installed power:
Installed power historical data:
7,500

MW
5,000

2,500

0
6,000
Nominal installation rate:
4,500 Nominal installation rate historical data:
MW/year

3,000

1,500

0
20
Perception of amortising period:
Nº of years

17 Perception of amortising period historical data:

14
11
8
0.6
Established subsidy value:
Eur/kW·hour

0.5 Established subsidy value historical data:

0.4
0.3
0.2
0.6
Electricity price evolution:
Cost per kW hour of photovoltaic energy including interest rate:
Established subsidy value:
0.45
Eur/kW·hour

0.3

0.15

0
2004 2006 2008 2010 2012 2014 2016 2018 2020
Time (Year)
Fig. 4. Same scenario and same policy. Default case.

from 2009 is due to the new RD, where the feed-in tariffs are
capped at approximately 500 MW per year. This annual ceiling is
achieved through a linear increase. The deviations between the
two graphs are because the simulation works in a continuous
process, whereas the historical data is the connection of data using
straight lines with a time step of one year in between.
The graph in the middle of Fig. 3 corresponds to the annual
growth rate. The historical data of this variable is grouped in a
time step function, where a constant flow value is taken for each
year. It can be deduced that the continuous curve from the
simulation is more accurate than the historical steps, since the
flow of installations cannot be a constant value during one year
and grow sharply at the end of the year. It is logical to think that
the variation is gradual. However, both flow curves give almost the
same values of stock, since the areas below the curves are almost
the same. The lower graph of Fig. 3 represents the evolution of the
average subsidy value from 2004 till July 2009. The feed-in tariff
was relatively constant from 2004, with a value of 0.44€ per kW h.
148 S. Movilla et al. / Energy Policy 60 (2013) 142–154
This value was reduced to 0.32€ per kW h in September 2008. The
model structure successfully simulates the implementation of the
last RD, especially the RD 661/2007. This part of the structure was
essential to obtain the historical results. Future policies will
probably not use the same method, but the structure worked in
accordance with the government decisions until 2009. In fact, the
real shortfall of the feed-in tariff in September 2008 matches the
simulated behaviour of the subsidy value perfectly.
The following simulation results are classified depending on
the PV general situation and the decisions. Consequently, there are
four cases:

Same scenario and same policy.

Same scenario and different policy.

Different scenario and same policy.

Different scenario and different policy.
It must be noticed that the results are focused on the case of
Spain, which is mainly an energy producer. In consequence, the
situation concerning panel manufacturing or cell production is in
global terms, treated in a general way. In order to differentiate
between scenario and policy, the market of PV cells and the energy
price are managed as a scenario, while the feed-in tariffs are
derived from the government decisions.
5.1. Same scenario and same policy
Until now, the Spanish government has used two main policies:
the first was the establishment of power goals, and the second was
the establishment of annual ceilings, which is the policy in force.
With the first policy system, the subsidy value was revised
downwards after fulfilling the goal. Concretely, when the goal
was about to be surpassed, the changes in feed-in tariff rates were
produced at the end of the year, and this also became a powerful
incentive to “buy now”. The second system apparently maintains
the subsidy value, but limits the growth. However, the develop-
ment is more continuous.
The continuation of the current RD is supposed to maintain the
annual ceilings and a constant value of feed-in tariff. Besides this,
the energy price evolves following the historical trend and the
price of the panel develops according to the Spanish demand.
As for the power capacity, it is in fact the prolongation of the
validation, with very predictable behaviour, as the first graph of
Fig. 4 shows. The historical data graph into the future is a constant,
because the installation growth rate was defined as null from
2009. Hence, the stock of accumulated capacity conserves the level
for the rest of the simulation.
The installation rate increased enormously until 2009, but with
the current policy, the flow experienced a sudden collapse after
the historical overshoot, as the second graph of Fig. 4 shows. With
the actual scenario, the trend for the future is a constant growth
rate of 500 MW per year in accordance with the present RD.
From the year 2006, the payback period started to decrease
because of the improvements in price, which in turn made the
facility construction cheaper, especially due to the panel price
reduction, as the third graph of Fig. 4 shows. With the shortfall of
the feed-in tariff at the end of 2008, the payback period was
MW
momentarily increased, but the generated market made the
continuation of a gradual price reduction possible, which is
improving the payback period. The reduction of subsidies was
not a problem for maintaining the profitability. Even now, the
payback period with the current feed-in tariff is reasonably
attractive. The essential condition to invest is that the subsidy
value is higher than the production cost, as the lower graph of
Fig. 4 shows.
The default scenario indicates that the subsidy value, which is
the upper line, is higher than the production cost, which is the line
in the middle. Therefore, the conditions are favourable to invest.
But the problem of the default policy is that the growth is limited
by caps of installed power per year.
5.2. Same scenario and different policy
In this case, there will be different simulations with different
policies regarding the feed-in tariff. The purpose is to analyse what
subsidy system could be optimal to maximise the utility. The first
policy consists of the worst case, which is the absence of a policy.
The following situation assumes the extreme test condition in
which the government stops the support and the PV should
continue selling the electricity at the same price as the market
does. For this reason there will be no new PV facilities, because
they will not be profitable. Furthermore the old PV facilities, which
finish their life cycle, will be retired. So, the installed power will be
decreasing over the next few years, as is shown in Fig. 5.
The second policy is based on the costs of producing PV energy.
These costs are the key, because nobody invests unless the income
per kW h is higher than the cost of producing one kW h from PV
energy. The goal of the investment is to achieve a competitive PV
energy. This energy will be competitive when the production cost
is similar to the electricity price. Therefore, if the line of electricity
rates crosses the production cost curve, the target will be accom-
plished. Fig. 6 considers a policy where the subsidy value is equal
to the production cost. This policy is considered as going for the
most efficient places, subsidising with the minimal profit margin,
which only pays the production costs. With this situation, a big
growth is expected during the next two years due to the industry's
inertia, but after the inertia the development will experience
stagnation due to the difficulty of obtaining profitability. The
potential benefits are only available to large investors.
The decrease in subsidies gives a higher payback period, which
in turn produces decreasing interest in investing. The result of this
effect is a decreasing amount of investments for new installations
while trying to preserve the old ones. In this case, there is no
annual growth ceiling, but if the feed-in tariff is equal to the
production cost, the growth from 2012 is even lower than that
established for the current RD.
The next implemented policy consists of a bigger feed-in tariff.
In this case, the subsidy also decreases gradually, but the tariff is
higher than the production cost. Concretely, the tariff is 20% higher
than the production cost. Under this simulation, the production
cost curve will reach the electricity rate in approximately 2020, as
Fig. 7 shows. Therefore, PV energy could be profitable without
subsidies from that year onwards. The PV capacity growth shows a
reasonably good development for the following years, but without
achieving the goal of a profitable energy source without subsidies
in the short term.
6,000
Installed power:
Installed power historical data:
4,500
3,000
1,500
0
2004 2006 2008 2010 2012 2014 2016 2018 2020
Time (Year)
Fig. 5. The government stops the support
 S. Movilla et al. / Energy Policy 60 (2013) 142–154 149

20,000
Installed power:
Installed power historical data:
15,000

MW
10,000

5,000

0
20

17
Nº of years

14

11
Perception of amortising period:
8
0.6
Electricity price evolution:
Cost per kW hour of photovoltaic energy including interest rate:
Subsidy value:
0.45
Eur/kW·hour

0.3

0.15

0
2004 2006 2008 2010 2012 2014 2016 2018 2020
Time (Year)
Fig. 6. The subsidy value is equal to the production cost.

At this point, it would be interesting to see the variation range
of the system's behaviour for different margins of profitability. This
part of the results is obtained through the sensitivity analysis.
With this tool, it is possible to simulate different evolutions across
the variation of one or more variables. In this case, the sensitivity
analysis considers the variation of the feed-in tariff factor. Assum-
ing that the minimum feed-in tariff is the production cost, this
factor is the percentage of the production cost which is added to
the subsidy value, giving a margin of profits equivalent to the
factor itself. The variation range of the factor is from zero to one.
If the value is zero, the profit margin is null, representing the
minimal subsidy. If the value is 1, it represents the other extreme
situation where the subsidy is double the production cost.
The sensitivity analysis is used to represent the installed power
and the PV energy cost from 2009 till 2020 depending on the factor
variation, as Fig. 8 shows. The lower limit curve represents the
installed power in the default scenario where the factor is null. The
upper limit curve represents the most favourable case where the
factor is one. The coloured sectors in between represent all the
factor values between zero and one. The first point is to see if the
production cost can achieve the electricity rate during the next few
years. In this case, the PV energy could start to be profitable without
subsidies, and the logical decision is to stop the subsidies from then
on. Therefore, when the PV energy is cheaper than electricity,
subsidies will end. It is necessary to take into consideration the
fact that the subsidies include the payment of electricity; hence the
net subsidy given by the government is the value of the subsidy
minus the electricity rate. In consequence, the subsidy value cannot
be under the electricity rate. Fig. 9 shows the simulation where the
PV energy is subsidised under the last policy, consisting of a tariff
which is double that of the PV energy cost.
Nevertheless, once the PV energy is cheaper than electricity for
more than one year, the subsidy value descends to the electricity
rate value. In this case, the subsidy value is big enough to
encourage a large development, which in turn provides a fast
decrease of the PV energy cost. When the PV energy cost is
cheaper than the electricity, PV energy starts to be competitive
and its survival no longer depends on the economic support. In
this sense, the main target of the government is accomplished, and
the predictable policy is to stop the subsidies regarding this kind of
energy. If, before crossing the benchmark, the net government
expenditure was the difference between the subsidy value and the
electricity rate, the net subsidy or benefit is now the difference
between the electricity rate and the PV energy cost. Fig. 9 also
shows the net subsidy calculation when the government retires
the support once the PV energy cost crosses the electricity price
150 S. Movilla et al. / Energy Policy 60 (2013) 142–154

40,000
Installed power:
Installed power historical data:
30,000

MW
20,000

10,000

0
0.6
Electricity price evolution:
Cost per kW hour of photovoltaic energy including interest rate:
Subsidy value:
0.45
Eur/kW·hour

0.3

0.15

0
2004 2006 2008 2010 2012 2014 2016 2018 2020
Time (Year)
Fig. 7. The tariff is higher than the production costs.

300,000
50% 75% 95% 100%

200,000
MW

100,000

0
2004 2008 2012 2016 2020
Time (Year)
Fig. 8. Sensitivity of installed power to the feed-in tariff factor. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of
this article.)

curve. The simulation corresponds with a feed-in tariff factor of
0.5. The accumulated capacity until 2020, assuming the previous
policy, observes the reduction in the growth trend when the
shortfall of subsidies is produced, as Fig. 10 shows.
Assuming the achievement of a profitable PV energy, the
question is what policy could be optimal to maximise the utility.
The utility is considered as the historical balance of net subsidies
over time. Fig. 11 shows the utility of the policy in which the
subsidies are in accordance with the PV energy costs. This case
represents the optimal case with a feed-in factor of 0.4. The
positive balance at the beginning of the simulation is due to the
big impulse generated by the construction of new installations,
which in turn provide taxes for the government. Later, all these PV
facilities accumulate subsidies paid per year during their contract
lifetime, and the balance becomes negative.
Finally, when the PV energy starts to be profitable, the net
subsidy value is favourable to the government and the balance
evolves towards positive values. Analysing the results in which PV
energy becomes profitable, the best policy is to pay a subsidy
according to the PV energy cost.
5.3. Different scenario and same policy
PV energy will become competitive if the PV production cost is
similar to the electricity price. Hence, there are two ways of
obtaining this achievement. Either the electricity price goes up,
or production costs come down. The possible production cost
reduction was analysed in the previous subsection as a conse-
quence of PV development. Hence, the overgrowth of the elec-
tricity rate is contemplated in this case as a different scenario.
Fig. 12 shows a possible situation in which the electricity rate
grows from 2010 until reaching double its value in approximately
5 years. The increase in the electricity rate negatively affects the
panel price reduction, producing a delay in it. This delay decele-
rates the PV energy cost reduction.
By contrast, the electricity rate crosses the PV cost curve ahead
of schedule, and then, PV energy starts to be competitive without
subsidies, since the profitability is obtained from selling the PV
energy at the same price as the electricity rate. Therefore, the
electricity rate overgrowth generates a favourable situation for the
PV sector, which could experiment a premature success. Fig. 13
 S. Movilla et al. / Energy Policy 60 (2013) 142–154 151

0.6

0.45

Eur/kW·hour 0.3

0.15

0
0.6

0.45
Eur/kW·hour

0.3

0.15

0
2004 2008 2012 2016 2020 2024 2028 2032 2036 2040
Time (Year)
Fig. 9. Net subsidy calculation with a feed-in tariff factor¼0.5.

100,000
Installed power:
Installed power historical data:
75,000
MW

50,000

25,000

0
2004 2006 2008 2010 2012 2014 2016 2018 2020
Time (Year)
Fig. 10. Accumulated capacity when the government retires the support once the PV cost crosses the electricity price curve.

shows the sensitivity of the installed power to the electricity rate
increase, (increase factor from 0 to 1). The big differences are
perceived when the subsidy value reaches the electricity price curve,
and in consequence, the profit margin increases significantly.
The next scenario assumes a fast PV panel development which
starts from 2010. The development consists of obtaining a revolu-
tionary panel, which is 10 times cheaper than the current PV
module. The achievement happens within 5 years. The panel price
is 10 times cheaper if the starting price is multiplied by 0.1.
Therefore, the decrease in the panel price is obtained through a
dimensionless price factor which is multiplied by the panel price,
as the upper graph of Fig. 14 shows. The consequences can be
observed in that the PV energy cost experiences a premature
decrease, helped by the panel price reduction, as the lower graph
of Fig. 14 shows. Although the subsidy value also decreases
according to the PV energy cost, PV development is affected
favourably. A cheaper PV panel reduces the PV energy cost. But
since the subsidy is proportional to the PV energy cost, the PV
energy cost reduction makes the subsidy value decrease in same
proportion, maintaining the profit margin constant, which in turn
provides the same payback period. A constant payback period
provides a constant amount of investments, but since the PV
panels are cheaper, the investments have a higher purchasing
power that allows a higher amount of acquired panels for PV
facilities. Finally, this increase of installed panels provides a higher
PV power capacity.
152 S. Movilla et al. / Energy Policy 60 (2013) 142–154

200 B

135 B

Eur 70 B

5B

-60 B
2004 2006 2008 2010 2012 2014 2016 2018 2020
Time (Year)
Fig. 11. Utility.

0.6 Electricity price evolution:

Cost per kW hour of photovoltaic energy including interest rate:
Subsidy value:
0.45
Eur/kW·hour

0.3

0.15

0
2004 2006 2008 2010 2012 2014 2016 2018 2020
Time (Year)
Fig. 12. The electricity rate grows from 2010 until reaching double value in 5 years.

300,000
50% 75% 95% 100%
150,000
MW

100,000
50,000
0
2004 2008 2012 2016 2020
Time (Year)
Fig. 13. Sensitivity of installed power to the electricity rate increase.

The upper graph of Fig. 15 represents the sensitivity of the
installed power to the panel price reduction. The reduction range
is between 0% and 90%. The variation range is not very wide, since
the feed-in tariff consists of a decrease according to the PV price
reduction. On the contrary, if the subsidy policy consists of
maintaining the feed-in tariffs, the effect of cheaper panels could
be more outstanding. The policy of gradual reduction optimises
the utility, instead of the PV power capacity. All the factors used
for sensitivity analysis are designed to see future effects; therefore,
they are activated from 2010 till the end of the simulation. The
upper limit curve represents the possible growth if the improve-
ments are enough to manufacture panels 10 times cheaper, and
the lower limit curve represents the default scenario.
The lower graph of Fig. 15 shows the sensitivity of the utility to
the panel price reduction. In this case, the variation range is wider,
since a cheaper panel makes the government support less costly.
In this scenario, the sensitivity analysis has considered the
possible effect of PV improvements, which could provide cheaper
modules. There is not enough data to calculate this variable
endogenously. In consequence, this advance could represent a
technological breakthrough or how the accumulation of purchases
can improve the manufacturing performance.
5.4. Different scenario and different policy
There are many possible policies and scenarios that can be
simulated. The sum of all possibilities provides an area of con-
vergence in which the PV energy will start to be competitive
sooner or later. Fig. 16 shows a combined sensitivity analysis. The
graph contains the possible evolution of electricity rates in a small
variation range. The other range of curves is the possible decrease
of the average PV energy cost. The intersection also provides a
 S. Movilla et al. / Energy Policy 60 (2013) 142–154 153

1
Panel price factor:

0.75

0.5

0.25

0
0.6
Electricity price evolution:
Cost per kW hour of photovoltaic energy including interest rate:
Subsidy value:
0.45
Eur/kW·hour

0.3

0.15

0
2004 2006 2008 2010 2012 2014 2016 2018 2020
Time (Year)
Fig. 14. From 2010 to 2015 the PV panel is 10 times cheaper.

150,000
50% 75% 95% 100%

100,000
MW

50,000

0
6B

-5.5 B
Eur

-17 B

-28.5 B
50% 75% 95% 100%
-40 B
2004 2008 2012 2016 2020
Time (Year)
Fig. 15. Sensitivity of the installed power (up) and of the utility (down) to the panel price reduction.

convergence area where the PV energy supposedly starts to be
profitable.
6. Conclusions
Although sustainable development policies have promoted PV
energy in Europe, the behaviour of the PV energy market is
affected by each national policy and other economic and
technological variables. The analysis of the Spanish case may help
to understand the dynamic behaviour of the PV energy market
and, therefore, to design optimal policies. System dynamics have
been shown to be an adequate tool to perform the modelling and
simulation of this kind of problems to help decision-makers. The
proposed model and its simulations contribute to a better under-
standing of the behaviour of the main variables of the PV energy
market in Spain. The Spanish experience makes it possible to learn
some lessons to be applied in other countries. The most important
154 S. Movilla et al. / Energy Policy 60 (2013) 142–154
0.3
Eur/kW·hour
0.2
0.1
50% 75% 95% 100%
0
2004 2008
Time (Year)
Fig. 16. Combined sensitivity analysis.
weakness of the sector is that, nowadays, the PV market still
depends on subsidies to be profitable. In this sense, if the
governments retire the PV support, the expected results are not
very promising for this type of energy. In consequence, the success
of the PV investments will depend on panel prices, but panel
prices evolve according to technological development, which in
turn is encouraged by PV purchases. However, if someone man-
ages to find a breakthrough to overcome the technological
barriers, PV energy could have a chance to play an important role
in the electricity market in sunlit countries.
The present research concludes that, considering current
trends, PV energy could start to be competitive in the short term.
Nevertheless, this waiting period is very sensitive to the adopted
policy. Best results were obtained for continuous support with a
limited profit margin, and gradual reduction according to PV
development. Although the model was tested, some parts and
parameters could be improved, it being the modeller himself who
decides if the approximation is good enough according to the
necessities of the project. In this sense, future lines of research
could expand the model by obtaining new data.
References
Avril, S., Mansilla, C., Busson, M., Lemaire, T., 2012. Photovoltaic energy policy:
financial estimation and performance comparison of the public support in five
representative countries. Energy Policy 51 (12), 244–258.
Algieri, B., Aquino, A., Succurro, M., 2011. Going “green”: trade specialization
dynamics in the solar photovoltaic sector. Energy Policy 39 (11), 7275–7283.
Carvalho, D., Wemans, J., Lima, J., Malico, I., 2011. Photovoltaic energy mini-
generation: future perspectives for Portugal. Energy Policy 39 (9), 5465–5473.
Castro, C., Miguel, L.J., Mediavilla, M., 2009. The role of non conventional oil in the
attenuation of peak oil. Energy Policy 37 (5), 1825–1833.
EIA, 2011. International Energy Outlook 2011. U.S. Energy Information Administra-
tion, Washington DC.
EPIA, 2011a. Solar Generation 6: Solar Photovoltaic Electricity Empowering the
World. European Photovoltaic Industry Association, Brussels, Belgium.
EPIA, 2011b. Global Market Outlook for Photovoltaics Until 2015. European
Photovoltaic Industry Association, Brussels, Belgium.
EPIA, 2012. European Photovoltaic Industry Association, Brussels, Belgium, Market
Report 2011.
Forrester, J.W., 1961. Industrial Dynamics. The MIT Press, Cambridge, MA, Reprinted
by Pegasus Communications, Waltham, MA.
2012 2016 2020
García, J.L., Cantero, A., 2005. Renewable 2050, A Report About the Potential of
Renewable Energies in Peninsular Spain. Greenpeace.
IDAE, 1999. Plan de Fomento de las Energías Renovables en España (PFER) 2000–
2010. Instituto para la Diversificación y Ahorro de la Energía, Madrid. Spain.
IDAE, 2005. Plan de Energías Renovables en España (PER) 2005–2010. Instituto para
la Diversificación y Ahorro de la Energía, Madrid. Spain.
IDAE, 2007. Solar Energy in Spain 2007. Current State and Prospects. Instituto para
la Diversificación y Ahorro de la Energía, Madrid. Spain.
IDAE, 2008. El sol puede ser tuyo: Respuestas a todas las preguntas clave sobre
energía solar fotovoltaica. Instituto para la Diversificación y Ahorro de la
Energía, Madrid. Spain.
IDAE, 2011. Plan de Energías Renovables (PER) 2011–2020. Instituto para la
Diversificación y Ahorro de la Energía, Madrid. Spain.
Jager, W., 2006. Stimulating the diffusion of photovoltaic systems: a behavioural
perspective. Energy Policy 34 (14), 1935–1943.
Leloux, J., Narvarte, L., Trebosc, D., 2012. Review of the performance of residential
PV systems in France. Renewable and Sustainable Energy Reviews 16 (2),
1369–1376.
Martínez, I., Díaz, P., Martínez, I., Ruiz, P., 2009. European Union's renewable energy
sources and energy efficiency policy review: the Spanish perspective. Renew-
able and Sustainable Energy Reviews 13 (1), 100–114.
ME, 2004. RD 436/2004. Ministerio de Economía. Boletín Oficial del Estado 75,
13217–13238.
Mediavilla, M., de Castro, C., Capellán, I., Miguel, L.J., Arto, I., Frechoso, F., 2013. The
transition towards renewable energies: physical limits and temporal condi-
tions. Energy Policy 52 (1), 297–311.
MIE, 1998. RD 2818/1998. Ministerio de Industria y Energía. Boletín Oficial del
Estado 312, 44077–44089.
MITC, 2007. RD 661/2007. Ministerio de Industria Turismo y Comercio. Boletín
Oficial del Estado 126, 22846–22886.
MITC, 2008. RD 1578/2008. Ministerio de Industria Turismo y Comercio. Boletín
Oficial del Estado 234, 39117–39125.
Movilla, S., 2009. Photovoltaic Energy Market in Spain: A Systems Dynamic
Approach for Policy Development. University of Bergen, Norway.
REN21, 2011. Renewables 2011. Renewable Energy Policy Network for the 21st
Century, Paris, France, Global Status Report.
Salas, V., Olias, E., 2009. Overview of the photovoltaic technology status and
perspective in Spain. Renewable and Sustainable Energy Reviews 13 (5),
1049–1057.
Sterman, J.D., 2000. Business dynamics. Systems Thinking and Modelling for a
Complex World. McGraw-Hill Higher Education, NY.
Teske, S., 2010. Energy [r]evolution: A Sustainable World Energy Outlook. Green-
peace International, European Renewable Energy Council EREC.
Zhang, Y., Song, J., Hamori, S., 2011. Impact of subsidy policies on diffusion of
photovoltaic power generation. Energy Policy 39 (4), 1958–1964.
Zhao, R., Shi, G., Chen, H., Ren, A., Finlow, D., 2011. Present status and prospects of
photovoltaic market in China. Energy Policy 39 (4), 2204–2207.