Anticipation for Efficient Electricity Transmission Network Investments

Abstract

This paper proposes a model and preliminary results to evaluate the efficiency of anticipating the connection of power plants with shorter construction duration than the time needed to obtain the right to upgrade the network and finally to do this reinforcement. This evaluation is made in presence of a cost of anticipation related to the study of the project of network investment and to the administrative procedures needed to obtain the building agreement. This model compares a proactive TSO that anticipates the connection of new generators and then the required network reinforcement, with a reactive one that does not make any anticipation but that may then face greater congestion while the network is being reinforced. The efficiency of these behaviors is measured in terms of social cost. We find out that there exists a limit of probability for the connection of generators beyond which a proactive TSO is more efficient than a reactive one. Evaluated on a realistic case of connection, this limit of probability is found quite low, which indicates that the proactive behavior for a TSO shall generally be the optimal one.

Full text

Page 1

Abstract—This paper proposes a model and preliminary
results to evaluate the efficiency of anticipating the connection of
power plants with shorter construction duration than the time
needed to obtain the right to upgrade the network and finally to
do this reinforcement. This evaluation is made in presence of a
cost of anticipation related to the study of the project of network
investment and to the administrative procedures needed to
obtain the building agreement. This model compares a proactive
TSO that anticipates the connection of new generators and then
the required network reinforcement, with a reactive one that
does not make any anticipation but that may then face greater
congestion while the network is being reinforced. The efficiency
of these behaviors is measured in terms of social cost. We find
out that there exists a limit of probability for the connection of
generators beyond which a proactive TSO is more efficient than
a reactive one. Evaluated on a realistic case of connection, this
limit of probability is found quite low, which indicates that the
proactive behavior for a TSO shall generally be the optimal one.

I. INTRODUCTION
OWER generation and transmission are complementary
activities that need to be coordinated, to the short-run and
to the long term to ensure an optimal use and development of
the transmission network.
The coordination between generation and transmission is
more difficult in a liberalised power system, not only because
these activities are unbundled but also because of the
investors’ choice for generation technology. The power
reform has prompted the generation investors to build mainly
power plants that can be built quickly, such as Combined
Cycle Gas Turbines or wind farms [1] [2]. At the same time,
the right of way of powerlines faces increasing oppositions
[3]. These movements increase the time needed to build
transmission lines.
These differences in construction duration create
uncertainty for the network planning. Indeed, these
differences in construction time are all the more detrimental
that the generation capacities of these new plants are

Manuscript received April 19, 2008. Vincent Rious gratefully
acknowledges the financial and scientific support he received from the
French TSO RTE for this work while he was a Ph. D student under the
supervision of Prof. Jean-Michel Glachant at both the GRJM Team at the
Research Center ADIS – University Paris XI, and at the Energy department at
Supélec.
V. Rious and P. Dessante are with Supelec, Gif-sur-Yvette, 91190 France
(corresponding author: V. Rious; phone: +33 (0)169 851 531; fax:
+33 (0)169 851 539; e-mail: vincent.rious@supelec.fr).
J.-M. Glachant is with University Paris-Sud XI, Fontenay-aux-Roses,
92260 France.
significant compared to the transmission lines capacities. The
connection of these power plants can thus create congestion.
A solution to this problem could be that the Transmission
and System Operator (TSO) anticipates the connection of
these new generation plants and the congestions that they may
create. By anticipating the connection of generation plants, the
TSO can adapt the network planning. To implement this
process, the TSO must anticipate the administrative
procedures required before the network upgrading. But if the
network is not eventually upgraded, this anticipation is costly
because of these administrative procedures. The efficiency of
anticipating the generation connection and of the required
transmission investment thus depends on this cost, on the
uncertainty on the effective generation connection and on the
required transmission investment.
This paper evaluates the efficiency of the strategy of
anticipating the connection of power plants for the TSO in
terms of the minimization of the network cost. The question is
then to know if it is efficient for such a TSO to plan the
development of its network in advance of the request of
connection so that there is sufficient planned transmission
capacity to accommodate these new generation investments.
This problem has been poorly addressed [4].
A model is built in this paper in order to measure the
efficiency of anticipating the connection of generation. In this
model, the connection of a generator is a probabilistic event
and the TSO can choose to anticipate the connections or not,
despite this anticipation may be costly if the anticipated event
does not eventually occur.
The paper is then organized as follow. In section II, we
show that the need to coordinate generation and transmission
varies with the considered generation technology. In section
III, we present our model to evaluate the efficiency of
anticipating the generation connection and the required
transmission investment depending on the cost of this
anticipation, on the probability of the power plants to connect
and on the difference in construction duration between the
transmission and generation investments. In section IV, we
made a preliminary measure of efficiency of anticipating
connection on a realistic example.

II. A NEED TO COORDINATE TRANSMISSION AND GENERATION
THAT VARIES WITH THE GENERATION TECHNOLOGY
In a liberalized power system where generation and
transmission are generally unbundled, the need to coordinate
Anticipation for Efficient Electricity Transmission Network
Investments
Vincent Rious, Philippe Dessante, Jean-Michel Glachant
P
hal-00339254, version 1 - 1 Dec 2008
Author manuscript, published in "1st International scientific conference "Building Networks for a Brighter Future", Rotterdam :
Pays-Bas (2008)"

Page 2

these activities varies with the generation technology. Indeed,
the time needed to build powerlines can be longer than the
time needed to build some generation technologies.
It takes at least five years to build a powerline and on
average seven to ten years in Europe [3]. There are two steps
to build a powerline. Firstly, the TSO must fulfil the
administrative procedures to have the right to build the line.
This step to obtain the administrative agreements lasts at least
three years. But in practice, it can last five years on average.
The second step consists in building the line. This step is quite
short, about two years only, and faces few uncertainties. The
step to obtain the administrative agreements is then the crucial
step for the time between the investment decision and the
completion of the project. The uncertainty on building the
powerline comes from this period because of the local
oppositions to the right of way of the transmission lines, which
can result in postponing the line project.
The generation technology induces an important variation
of the difference in investment time between transmission and
generation. Besides, some generation technologies have an
important notional size while they can be quickly built. The
connection of the power plants can then create congestion
while the TSO has not yet upgraded the network to evacuate
this power. This can make the accommodation of these
generators more difficult. This difficulty is evaluated in table
1 by the third column that gives the notional size of an
installation divided by the time to build it.

TABLE I
TIME NEEDED TO BUILD DIFFERENT GENERATION TECHNOLOGIES
[6] [7](RAE2004, DGEMP2003)
Time
needed to
build (year)
Combustion turbine
1
Coal
4-5
CCG
2
Nuclear
5-7
Hydro
-
Wind onshore
2
offshore
2
*Depending on technologies

The investment dynamics of the nuclear and coal
technologies are similar to the investment dynamics of
network investments. The TSO can then deal with their
connection when required. To the contrary, the CCGT and the
wind farms can be built and connected faster than the network
can be modified to accommodate them. Therefore, these new
generation units can create important congestion before the
TSO can upgrade the network.
This phenomenon becomes very important because these
technologies, CCGT and wind farms, are the fastest to be built
and are actually the preferred ones in Europe and in the USA
(see [1] for CCGT and [2] for wind power).
The time to build CCGT is quite short since it is only about
two to three years [5] [6] The notional size of CCGT is 800
Generation
technology
Notional size
(MW)
Size divided by
time to build
(MW/year)
40
30 to 400*
400
200 to 300
-
12.5
50
40
150 to 1600*
800
1600
-
25
100
MW. It cannot be neglected compared to the transmission
capacity of powerlines, between 1000 and 2500 MW for the
voltage level where they connect (400 kV or 225 kV).
The massive connection of wind power to the transmission
network is also problematic for two reasons. Firstly, compared
to the time to upgrade the network, the time to build wind
farms is quite short, since it is about two to three years.
Secondly, the network must adapt to the massive connection
of intermittent power plants. Besides, their massive
connection can then require upgrading the transmission
network to evacuate the generated power.
Therefore, the problem is the following one: some
generation technologies can require important network
upgrading whose time to build is quite longer than the time to
build these power plants. There may then be congestion
between the moment when these generators connect to the
network and the moment when the TSO upgrades the
facilities.
The TSO can anticipate the connection of these plants and
consequently plan the network investment to avoid these
congestions. Then it can better deal with the uncertainty
coming from the difference in time to build power plants and
time to upgrade the network. But this work of anticipation is
costly. As a consequence, it is necessary to evaluate the
efficiency of anticipating the network investment to
accommodate new generators.

III. A MODEL TO MEASURE THE EFFICIENCY OF ANTICIPATING
NETWORK INVESTMENTS
In order not to limit and even in order to facilitate the
development of the types of generators with the shortest time
to build, it can be efficient to anticipate their connections. The
network can then have sufficient capacity to accommodate
them. The TSO can anticipate the connection of these
generators and can so study in advance the opportunity of
upgrading the network.
Such anticipation does not compel the TSO to invest if it
eventually happens unnecessary. Because, once the TSO has
obtained the administrative agreements required before
building the power line, the TSO can decide to upgrade the
network effectively only after the relevant assumptions of the
investment project has come true or extremely certain. To the
contrary, a TSO whose objective is to maximise the social
welfare can decide to cancel a planned investment if the
relevant conditions does not eventually happen. The
administrative steps needed before the building of the
transmission line have then an appreciable option value if the
TSO can implement various planning strategies to invest [7].
The TSO can implement two strategies to anticipate
transmission reinforcements. 1° The TSO can be proactive
and anticipates the change in the generation mix and location.
2° The TSO can be reactive and upgrades the network only
once he knows where and when the power plants connect. In
the reference [4], Sauma and Oren show that, from the point of
hal-00339254, version 1 - 1 Dec 2008

Page 3

view of the minimization of the expected social cost, the
proactive TSO is always more efficient than the reactive one
in an uncertain environment. But they implicitly assume that
anticipation has no cost while it is costly in reality. And it
could be quite expensive [8].
The essential parameters to evaluate the efficiency of the
anticipating the generation and transmission investment are
then the three following ones: 1° the cost of anticipating
investments, 2° the difference between the time to build power
plants and the time to upgrade the network, and 3° the
probability of connection of generators. Our model will allow
measuring the influence of these different elements on the
opportunity for the TSO to be proactive. We describe this
model below. Next, we will illustrate our results on a
representative case of connection.

A. A necessary condition for anticipation to be optimal
We present here a simplified model where the congestion
cost is assumed given and sufficient to require a transmission
investment. The problem of calculating the congestion cost is
dealt with further detail in section IV of this paper in the
realistic case of connection of a power plant. In our modelling,
we search for the conditions when it is efficient, from the point
of view of the minimization of the expected social cost, to
anticipate the connection of power plants whose time to build
is shorter that the time to upgrade the network.
We consider the two proactive and reactive behaviors of the
TSO that we described just above. To anticipate the
connection of new power plants, a proactive TSO does in
advance the study of the transmission investment project and
the administrative procedures that are required to have the
agreements to build the powerline.
A reactive TSO does not study the project nor does he make
the administrative procedures in advance. He does these steps
only once the generators have effectively asked to connect to
the network.
We assume that in a step before the application of our
modelling, an expert has highlighted the nodes or areas where
generators are more likely to connect and the lines that may
experience congestion. It is similar to the approach used in the
National Electric Transmission Congestion Study of the U.S.
Department of Energy [9]. This phase consists in determining
where it will be needed to upgrade the network taking into
account the primary energy sources and the areas where the
generators will be able to locate. The goal of our modelling is
then not to find where to upgrade the network. Our modelling
finds the lines whose forecasted constraints are such that it is
efficient to anticipate its upgrading and especially to
anticipate the long administrative procedures. Anticipating
can result in accommodating the considered generation
technologies more efficiently and more quickly.

B. Definitions and assumptions
For each year y, we define two types of congestion cost that
we note CUy and CWy respectively with and without network
reinforcement. Then, whatever the year y, the congestion cost
without reinforcement is greater than the congestion cost with
reinforcement, other saying CWy ≥ CUy. For a year y, the
congestion cost CWy or CUy depends only on reinforcing the
network, and not on the moment when the network is
upgraded. We define two functions of discounted and
cumulative congestion cost over several years d with a
discount rate a. The first function, CW(d) is the total
congestion cost discounted during d years before upgrading
the network. The second function, CU(d) is the total
congestion cost discounted during ten years after the network
being upgraded the year d. The avoided costs thanks to the
reinforcement of the network are evaluated only over ten years
for two reasons. Firstly it is difficult to know accurately the
state of a power system beyond ten years. Secondly
discounting considerably decreases the costs beyond this
temporal horizon. CW(d) and CU(d) can be expressed as
functions of CWy and CUy as follow:

CW
d CW
1 1

Fig. 1 exemplifies the two sums CW(d) and CU(d). For
illustrative reasons, we assume that the terms CWy / (1 + a)y
and CUy / (1 + a)y increase linearly with time.CW(d)
corresponds to the grey trapezoid and CU(d) corresponds to
the black trapezoid.

( )
()
∑
=
y
+
=
d
y
y
a
and
( )
d
()
∑
=
d
+
+
+
=
10
1 1
d
y
y
y
a
CU
CU
(1)

Fig. 1. Definition of CW(d) and CU(d).

We assume that the network must be reinforced as soon as a
power plant connects. That is to say that the cost saved by the
network upgrading as soon as the generator connects is greater
than the related transmission investment cost. Fig. 2 illustrates
the cost saved by upgrading the network as a function of
CW(10) and CU(0).
The equation (2) links CW(10), CU(0) and the investment
hal-00339254, version 1 - 1 Dec 2008

Page 4

cost I.

CW

( )
10
( )
0
I CU
≥−
(2)
With
( )
10
( )
0
()
∑
=
i
+
−
=−
10
1
1
i
ii
a
CU CW
CUCW



Fig. 2. Cost saved by upgrading the network as a function of CW(10) and
CU(0). The greyed area stands for this saved cost.

The instant of reference for discounting the cost over time is
chosen such that the most conservative condition about the
efficiency of a proactive TSO is obtained. This instant is then
the moment when the generation investment begins to produce
power. This convention degrades the advantage of
discounting for the strategy of anticipating compared to the
strategy of not anticipating. Indeed, if the beginning of the
administrative procedures has been chosen as the instant of
reference for discounting, the discounting of cost would have
mechanically decreased the cost of network investment. By
taking the beginning of production of the power plant as a time
reference for discounting, this effect is avoided.

C. Expected cost of the network for a reactive TSO
A reactive TSO does not anticipate the connection of
generators. It studies the network upgrading only once the
generator has invested. The network investment to evacuate
this power is ready to serve only d years after the connection
of the generation unit, where d is the difference between the
time to build a power plant and the time to upgrade the
network. Fig. 3 exemplifies this sequence of the generation
and transmission investments.
The generator can connect to the network with a probability
p (and so does not connect with a probability of 1-p). This
uncertainty is not intrinsically quantifiable. However it is
possible to attribute it a subjective value to evaluate the
robustness of the study framework. This approach stimulates a
dialog with the other stakeholders of the power system and
creates a shared anticipation of the evolution of the system
[10]. Besides, the sooner the residents are involved in the
transmission investment process, the easier the powerline
would be built [11] [12].
If the generator connects, the system must successively
support:
•
CW(d), the total congestion cost discounted for d
years, while the TSO is upgrading the network
•
CU(d), the residual total congestion cost discounted
for ten years after upgrading the network, that is to say
d years after the connection of the generator,
•
I/(1+a)d, the discounted cost of this upgrading d years
after the beginning of our study (corresponding to the
moment when the generator is ready to generate
power).


Fig. 3. Sequence of the generation and transmission investment with a
reactive TSO. The timeframe of generation is in grey while the timeframe of
transmission is in black. While the power plant has already connected and the
network has no yet been upgraded, there is congestion for d years.

If the generator does not connect, the network investment
cost and the congestion cost to the TSO is null. Table 2
summarises these two cases.

TABLE II
THE COSTS FACED BY A REACTIVE TSO
Generator

TSO
waits for the
connection of
the power
plant before
studying and
upgrading the
network

invests
Probability p
does not invest
Probability 1-p
Expected social
cost
CS(d) + CR(d)
+ I/(1+a)d
0 + 0
( )
+
[]
( )
(
1
( )
d
)




++
=
d
TSO
d
active
CS
aI
CR
p
pCE
re

D. Expected cost of the network for a proactive TSO
A proactive TSO anticipates the connection of the
generator. He studies the network upgrading and asks for the
administrative agreements to build the powerline (without
building it) before the connection of the power station. The
network is upgraded only once the power plant is being
effectively connected. Fig. 4 illustrates the sequence of the
generation and transmission investments.
The generator can connect with a probability p (and so does
hal-00339254, version 1 - 1 Dec 2008

Page 5

not connect with a probability 1-p). In case the generator
connects to the network, the system must successively
support:
•
CU(0), the residual total congestion cost discounted
for ten years after the power plant connecting and the
network upgrading
•
and I, the upgrading cost.


Fig. 4 Sequence of the generation and transmission investment with a
proactive TSO. The use of the new transmission line and of the new power
plant begins at the same time because the TSO has anticipated the network
upgrading.

In case of the generator not connecting, the congestion cost
is null and the cost to anticipate linked to the non realization of
the anticipated event is a share α of the total investment
transmission cost. Indeed, the transmission investment is not
done but the preliminary steps are however realized. The cost
α.I includes not only the cost to anticipate to the TSO but also
the cost born by the local authorities involved in the process of
administrative agreements. Moreover, the cost α.I takes into
account the a posteriori discounting of the cost to anticipate.
We assume the cost to anticipate α.I is proportional to the
investment cost because a power line faces all the more
oppositions that it is longer and goes through a wider area.
Table 3 summarises these cost.

TABLE III
THE COSTS FACED BY A PROACTIVE TSO
Generator
TSO
Probability p
Studies beforehand
the installation of
the power plant
and invests at the
same time as the
generator

Invests Does not invest
Probability 1-p Expected social cost
I + CR(0)
α.I + 0
( )
( )
0

[]
] (
+
[
)
α
pCRIp
pCE
=
TSOproactive
+−
1
E. Condition for a proactive TSO to be efficient
We are searching for the necessary and sufficient condition
for the proactive TSO to be more efficient than the reactive
one from the point of view of the minimization of the expected
social cost.
This condition links the cost α to anticipate, the probability
p to connect a power plant and the difference d in time to build
a power plant and a power line such that the equation 3 is
respected.

( )
CEpCE
reactiveTS SOproactiveT
≤

The equation 4 equivalently expresses this relation.

(
d CWap
11
+−++≤
αα

To interpret this formula, we consider the case of equality
of the equation 4 and then define the limit of probability or
“probability limit” plim to connect a power plant.

α
[]
( )
p
[]
O
(3)
)( )( )
d
( )
0
()
[] ICWCU
d
−+
−
(4)
() ( )
d
( )
d
( )
0
[]
[] ICWCUCWa
p
d
11
lim
−++−++
=
−
α
(5)

If the probability to connect a power plant is greater than
plim, then the proactive TSO is more efficient than the reactive
one. To the contrary, if the probability of connection p is less
than plim, then the reactive TSO is more efficient than the
proactive TSO. The equivalence between the equations 3 and
4 then shows that the strategy of anticipation is all the more
efficient that the probability plim is small. The interpretation of
the equation 4 also consists in evaluating how the probability
limit plim varies with the cost α for anticipating and the
difference d in time to build a generation investment and a
transmission one. For a given cost α for anticipating, plim
decreases when the difference d between the time to build a
power plant and a power line increases. This is because the
congestion cost generally increases more quickly than the gain
from postponing the network investment and its discounting.
For a given difference d, plim increases when the cost α to
anticipate increases.

IV. ILLUSTRATION OF PROACTIVE AND REACTIVE TSOS
Now we illustrate the equation 4 on the concrete case of
connection of a power plant. We can determine the
combinations of parameters where the proactive TSO is more
efficient than the reactive one in this case.
We apply the criterion of the equation 4 on the simple case
of the connection of a power plant on a two node network. The
figure 5 is realistic because the connection of a power plant
(CCGT or windfarm) raises problems mainly because these
power plants are far from the load centres.


Fig. 5 System to test the strategy to anticipate the connection of a power
plant. The load and most of generation are to the east. These power stations to
the east are quite expensive, 40 €/MWh for the first 3000 MW and
hal-00339254, version 1 - 1 Dec 2008

Page 6

100 €/MWh after. The power plants to the west are less expensive, only
35 €/MWh.

A new generator wish to connect a 800 MW power plant to
the west. To evacuate the power of this new power plant, it is
necessary to add a 1000 MW network upgrading that costs
100 million euros. We assume that the situation of Fig. 5 is
representative of the 8760 hours of the year. The investments
are dashed on this figure.
First of all, we draw plim the “probability limit” to connect a
power plant. When the effective probability is above the
probability, the proactive TSO is more efficient for a given
cost α to anticipate (equal to 10% of the considered network
investment). The “probability limit” can be defined as a
function of the difference d between the time of building a
power plant and the time of building a powerline. Fig. 6
illustrates this function.


Fig. 6. “Probability limit” of connection beyond which the TSO must be
proactive depending on the difference of temporal dynamics d between
generation and transmission investment

Above the curve associating the probability limit with the
difference in dynamics of investments, the TSO being
proactive is the optimal behavior. Under this curve, the TSO
being reactive is the optimal behavior. On Fig. 6, we notice
that for the considered case, the probability limit decreases as
the difference in dynamics increases. We find again that the
TSO has little interest in anticipating the connection if the
temporal rhythms of these two complementary investments
are near each other. That is to say that the time of building a
network investment is short or if the time of building a
generation investment is long. For the connection of a CCGT
or a windfarm, this difference in time is at best of three to four
years (see section II). For such a difference, the probability
limit of connection is around 15 to 20%. This weak value is
already significant to justify the anticipation of the network
reinforcement.

V. CONCLUSION
This paper presents a model to evaluate the efficiency of the
strategy of anticipating the connection of power plants to the
network for the TSO in terms of the minimization of the
network cost. This model has permitted to see on a simple but
realistic example that it can be quite efficient for a TSO to
anticipate the connection of power plants. He can then plan
the reinforcement of the network in advance and reduce
congestion on the network. Taking into account the interest of
investors for generation technology with short time to build,
the proactive behavior of the TSO can facilitate the
connection of these types of power plant and increase the
market entries.
However, the study proposed in here is only a preliminary
one. We assume that the cost of anticipating is know whereas
the planner does not know in advance the cost of anticipating
the reinforcement, that is to say the cost of the administrative
agreements needed to build the reinforcement. It is indeed
difficult to rely on historical data for the cost of anticipation,
because the powerlines face more and local oppositions. A
sensitivity analysis of the probability limit to the cost of
anticipating of the investment should then be one of the future
researches of this work. Besides, we could duplicate this study
for a windfarm with a given generation duration curve.
Eventually, regulatory actions should be studied to incentivise
the TSO to anticipate investment if it is efficient. This is still
an opened question that is still to be answered.
REFERENCES
[1] JM. Glachant, “Generation technology mix in competitive electricity
markets” in Investments frame and incentives in the electrical
industry, Lévêque F. (ed), 2006. Kluwer
[2] ETSO, “European Wind Integration Study (EWIS). Towards a
Successful Integration of Wind Power into European Electricity Grids”.
Final Reports, 15th January 2007. www.etso-net.org
[3] ETSO, “An Overview of Current Cross-border Congestion
Management Methods in Europe”. May, 2006. www.etso-net.org
[4] E. Sauma and S. Oren, “Proactive planning and valuation of
transmission investments in restructured electricity markets”, Journal
of Regulatory Economics, 30 (3), 2007, 261-290
[5] Royal Academy of Engineering (RAE), The cost of generating
electricity, London, March 2004
[6] DGEMP, Coûts de référence de la production électrique, Ministère de
l’Économie, des Finances et de l’Industrie, Paris, December 2003
[7] G. Boyle, G. Guthrie and R. Meade, “Real options and transmission
investment: the New Zealand Grid Investment Test”, ISCR working
paper, 2006, www.iscr.org.nz
[8] G. Christiner, “VERBUND-Austrian Power Grid. Austria –
Completing the 380kV-Ring. Authorization Procedures and Major
Problems”, CEER Electricity Infrastructure Workshop, 13 February
2007, Bruxelles
[9] U.S. Department of Energy (USDoE), National Electric Transmission
Congestion Study,
www.nietc.anl.gov/documents/docs/Congestion_Study_2006-9MB.pd
f
[10] J. Bråten, “Estimating benefits. Uncertainty, scenario design and the
analysis of benefits”, European Transmission Investment Conference,
18-19th October 2004, Copenhague, Danemark
[11] R. Hughes, “Environmental Impact Assessment and Stakeholder
Involvement”,
International
Development, 2000, www.iied.org/pubs/pdf/full/7789IIED.pdf
[12] J. MacLaren Loring, “Wind energy planning in England, Wales and
Denmark: factors influencing project success”, Energy Policy, 35(4),
2007, 2648-2660
August 2006,
Institute for Environment and
hal-00339254, version 1 - 1 Dec 2008