Documente Academic
Documente Profesional
Documente Cultură
( )
(5)
where o is the speed of reversion to the mean,
ln ( )
F
g
p t
o is the volatility of natural logarithmic of fuel prices, and
F
g
p is the normal level of the natural logarithmic of fuel price ( )
F
g
p t , i.e. the level to which
F
g
p tends to
revert
3
.
The historical as well as the forecast data [15] on costs and prices have been used to estimate the numerical
parameters of eq. (5). These parameters are listed in Table III. In the simulations, nuclear fuel cost prices are
assumed constant over the time horizon. The main fundamental of this assumption is based on the fact that the
uranium cost is only a small fraction of the total variable cost (around 5%) in nuclear plants and the deviations
around the expected value are quite narrow in comparison to the fossil fuel price fluctuations [16].
TABLE III.
MEAN REVERSION PROCESS PARAMETERS
Fuel
Type
(0)
F
g
p
/MW
F
g
p
/MW
ln ( )
F
g
p t
o
%
Gas 12.46 17.94 0.129
Oil 20.99 28.21 0.3
Coal 5.51 6.64 0.14
2.2.3 Weighting of the regarded wind scenarios
In order to reduce the number of calculated situations for each realization and each year, two load situations
(base load and peak load) and three wind situations are taken into consideration. The probability of occurrence of
each wind situation is determined according to the empirical histogram shown in Figure 2.
The underlying data are actual values of the wind feed-in in Germany during the year 2006 in a 15 minute
resolution. The overall installed capacity in Germany is around 19.5 GW. The histogram is divided in three
3
Instead of formulating the mean-reversion process in the prices themselves, it should be formulated in the logarithms of prices for ensuring
that electricity prices are always positive [17].
- 9 -
sections. The first section on the left side, low wind, comprises 50% of all values. The next 30% of the values are
in the second section, medium wind, and the remaining 20% represent a high wind condition.
Fig. 2. Weighting of the calculations based on the frequency distribution of the wind feed-in.
The actual wind feed-in that is used in the calculations is defined as the median in these three sections. With
the assumption that 70% of the year can be represented by a base load situation, the matrix shown in Table IV of
the six possible combinations is obtained. Therefore, for each realization and each year six situations are
calculated and weighted according to their probability of occurrence to get representative results of one year.
TABLE IV.
WEIGHTING OF THE WIND FEED-IN SCENARIOS.
low
wind
(lw)
medium
wind
(mw)
high
wind
(hw)
peak load
15 % 9 % 6 % 30 %
base load
35 % 21 % 14 % 70 %
50 % 30 % 20 % 100 %
wind power
feed-in
6 % 19 % 46 %
2.2.4 Modelling of pumping storage units
The complexity in the modelling of pumping storage units results from the interdependency of the pumping
and generating process. Units without any natural inflow can only generate that amount of electricity that was
stored before, taking into account the limited process efficiency. In the presented approach, this problem is
solved by a sequential simulation of the base load situation first, followed by the peak load situation. During the
- 10 -
different base load situations, the pumping storage units are considered as dispatchable loads in the OPF.
Depending on the price, a certain amount of electricity is stored in the reservoir. The assumed size of the
reservoir results from the assumption that every unit is able to generate with maximum power during all peak
load hours. According to the previous section, 30% of the time, 7.2 hours per day respectively, is considered as
peak load situation. Therefore, the reservoir has to be filled completely during the 16.8 base load hours. As a
dispatchable load is modelled with negative values in the OPF calculation, this results in the following two
constraints.
,max
,
0
i
g base
P = (6)
,
,min
,
7.2
16.8
g inst
i
g base
g
P
h
P
h q
= (7)
with the installed generating capacity
, g inst
P and the total process efficiency
g
q . The maximum price for
storing electricity is also linked with the efficiency as the expected profit during peak load hours has to cover the
expenses for the pumped energy. The expected profit is calculated on the basis of the mean of the peak prices of
the previously simulated year. For the first year an assumption has to be set. If the resulting price at node i
where unit g
is installed is below the value ( )
base
g
p t in the base load situation, then electricity is stored in the
reservoir. For the calculation of the value also the cost for operation and maintenance are considered.
( ) ( 1) O&M
base
peak
i i g
p t p t q = + (8)
In the subsequent simulation of the peak load situation, the maximum generating capacity
,max
,
i
g peak
P is set
according to the stored energy; the minimum capacity is set to zero.
,min
,
0
i
g peak
P = (9)
,max
,
( )
7.2
g
i lw lw mw mw hw hw
g peak
P l dur l dur l dur
h
q
= + + (10)
where ,
lw mw
l l , and
hw
l represent the dispatch of the pumping storage unit during base load, multiplied with
the respective duration of each wind situation.
The price or marginal cost of the unit results from the weighted sum of the stored energy
( )
( )
, , ,
( ) ( ) ( ) ( ) ( ) ( )
( ) O&M
( ) ( ) ( )
base lw lw lw base mw mw mw base lw hw hw
peak i i i
i
lw lw mw mw hw hw
g
p t l t dur p t l t dur p t l t dur
p t
l t dur l t dur l t dur q
+ +
= +
+ +
(11)
2.2.5 FACTS Devices modelling
The development of flexible controllers of transmission systems based on the progress carried out in power
electronics is essential to efficiently utilize the existing grid and generation infrastructure. Flexible Alternating
- 11 -
Current Transmission System (FACTS) devices have been largely investigated during the last decades [18]. The
impacts of these power electronic-based controllers on the power systems are well established; numerous
investigations have analyzed FACTS on congestion management, as well as their ability to improve
controllability and the security of interconnected systems [13], [18]. These are key features in a competitive
environment, where the fast-reacting FACTS controllers can properly help to avoid or relieve bottlenecks,
delaying the execution of large transmission line investment projects.
As mentioned, maybe the outstanding feature, still barely explored, that FACTS devices offer to manage
network investments under uncertainty is a set of options that enhance the transmission investment flexibility.
Those options, such as: relocation, abandonment, operational flexibility, expansion and contraction, offer a
substantial additional value to these investments and consequently, must be fairly evaluated.
Within this research, the Thyristors Controlled Series Compensator (TCSC) is analyzed under steady state
operation. This FACTS was chosen since it is most appropriate for controlling power flows in the new power
markets. In comparison to more conventional compensation strategies (e.g. fixed series compensation), it has the
advantage of a very fast output reaction to changes in control values. In fact, given that this work tries to analyze
the electric system operation under uncertainty; a continuous and flexible compensation according to the
unfolding of the uncertain variables is preferred rather than fixed and discrete series compensation [11].
The mathematical model of this device modifies the reactance of the transmission line, allowing the TCSC to
operate as capacitive or inductive compensation respectively (Fig. 3 a). The rating of TCSC is defined by the
reactance and current capacity of the transmission link where the TCSC is installed:
1
;
ij TCSC TCSC Line
Line TCSC
b X r X
X X
= =
(12)
where
Line
X is the reactance of the transmission line and
TCSC
r is the coefficient which represents the degree
of compensation by TCSC.
For static implementations, FACTS devices can be modelled by power injection models (PIM) . The PIM
model depicts FACTS as devices that inject a certain amount of active and reactive power into its nodal
connections; so that this controller operation is replicated by this injection flows. The advantages of the PIM are
that it does not lose the symmetrical structure of the admittance matrix and allows efficient and convenient
integration of FACTS devices into existing power system analytical tools [19].
On the basis of a DC power flow model, the active power flow along the line i-j with a TCSC can be
formulated as:
- 12 -
( )
ij ij i j
P b 0 0 = (13)
Therefore, the equivalent circuit of the general PIM of a TCSC under the DC approach is shown in Fig. 3(b),
in which the operation of the FACTS device is represented by two power injections on both sides of the
transmission line. These power injections are the followings:
(a) (b)
Fig. 3. (a) Equivalent circuit of TCSC, and (b) power injection model of TCSC.
, ,
( )
( )
TCSC
F i F j i j
ij ij TCSC
X
P P
X X X
u u = =
(14)
These power injections must be added to the nodal power balance constraints into the classic DC-OPF
problem (eq. (1)). Moreover, the line transfer capacity, where the TCSC is installed, may increase (10-12%) due
to the stability improvement [20]. Therefore, the constraints related to the power transfer limit in the
compensated branch must be also modified. Then, the improved transfer limit is equal to the algebraic sum of the
power and the power injection.
Due to the TCSC power injections are a function of voltage angles as well, which are state variables in DC-
OPF linear programming, there are no fixed limits for these power injections (although there are fixed limits for
XTCSC). Therefore, its operating constraints can be formulated with two additional inequality constraints, which
have to be added to the DC-OPF problem (eq. (1))
max min
, ,
max min
0; 0
( ) ( )
TCSC TCSC
F i F i
ij ij TCSC ij ij TCSC
X X
P P
X X X X X X
_ _
(15)
To avoid overcompensation, the working range of the TCSC is chosen between 0.7
Line
X and 0.2
Line
X
[21]. The DC-OPF estimates the optimal power flow injections of TCSC devices in order to minimize the
production costs. From these values, the optimal compensation level can be obtained so that:
2 *
,
*
* * *
,
( )
ij F i
TCSC
i j ij F i
X P
X
X P 0 0
=
(16)
3 Financial Evaluation
- 13 -
3.1 Decision Making of Investment Portfolios in Transmission System
By means of the Monte Carlo simulation, the cost savings (CS) for each realization along the investment
horizon are estimated. Thus, the stochastic project cash flow is defined by these cash flows and the capital
expenditures of the expansion project. Fig. 4 illustrates the resulting cash flow of a Monte Carlo realization,
which is composed by the annual
,
s
i
CS
.
, investments costs (
,
n
s t
I ) and operation cost (
,
n
s t
OC ).
Fig.4. Resulting Cash Flow for the transmission investment.
The cash flow, thus originated, is discounted by the financial cost of the investment () in order to obtain the
present value of the Incremental Social Welfare (ISW), according to the following expressions:
( )
,
, ,
;
(1 )
n
n
s
M
i
s t i
i t
CS
PV ISW
.
.
j
=
1
( )
_
(17)
( )
, , ,
, ,
;
(1 )
n
n
s
M
i s i s i
s t i
i t
CS I OC
NPV ISW
.
.
j
=
1
( )
_
(18)
( ) ( )
, , , ,
1
1
s t s t
n n
E NPV ISW NPV ISW
. .
.
\
=
l 1
l =
l
( ) \
l
_
(19)
where
,
s
i
CS
.
and
, s i
I are the generation cost savings and the investment cost respectively in the realization,
( )
,
j
s k
PV ISW and
( )
,
j
s k
NPV ISW are the Present Value (PV) and Net Present Value (NPV) of the ISW by
executing the investment strategy s in the year
n
t and M the investment horizon, finally, ( )
, ,
n
s t
E NPV ISW
.
l
l
l
l
is its expected value for Monte Carlo samples. In each case, the subscripts correspond to the h-th hour, i-th
year, -th realization of the Monte Carlo power system simulation.
If the decision was made under the traditional NPV approach, the decision making rule would dictate: When
the NPV is positive, the investment should be executed [22]. Nevertheless, this decision rule often overlooks
some implicit assumptions. Most important, it is assumed that the investment decision are of kind now-or-never
when the investment is irreversible, that is, the only decisions which the investor could make are either execute
the investment or discard the project. Although some investments meet these conditions, most do not. Nowadays,
irreversibility and the chance of postponing are quite relevant characteristics of most investments. In fact, the
- 14 -
ability to delay irreversible investment expenditure can deeply impact on the decision to invest [22]. These
features often undermine the simple NPV rule, and thus the theoretical foundation of the standard neoclassical
investment models.
The basic theory of irreversible investment under uncertainty, developed in [22], [23], emphasizes the option-
like features of the investment opportunities and proposes to develop decision rules based on methods for pricing
option in financial. Thus, to stress the analogy with the financial options, the opportunities to take investment
decisions are called real options. Besides the deferral option, there are many contingent claims (e.g. abandon,
expansion, relocation, etc.), which act as a hedge of the investments in case unfavourable market conditions
should turn out to be worse than anticipated. These options enhance the flexibility of investments, and represent
an opportunity cost that must be included as part of the investment. This opportunity cost of investing can be
large and highly sensitive to the uncertainties over the future value of the project, and investment rules that
ignore it can lead to gross errors. Within the transmission network, investments are not normally able to be
undone and the expenditures recovered. Therefore, it is somehow necessary to redefine the NPV decision rule by
including the opportunity cost of the flexibility options.
3.2 Valuing Flexible Investment Portfolios in Transmission System
Flexibility can be defined as the attribute that allows investment project adapt itself when external changes
occur. Real option (RO) is a novel investment valuation technique for valuing it, which applies concepts derived
from finance theory to the valuation of capital investments. It refers to choices on whether and how to proceed
with investment projects. In the first RO applications, valuations were normally bounded to options for which
appraisal of the financial could directly be applied. Afterwards, the introduction of multiple interacting options
into the RO models increases the problem complexity, making these traditional numerical approaches
impracticable.
In the recent years, simulation procedures for solving multiple American options have been successfully
proposed. One of the most promising approaches is the Least Square Monte Carlo (LSM) method proposed by
Longstaff and Schwartz [12]. The LSM approach is based on a Monte Carlo simulation algorithm to value the
financial American options. More recently, Gamba [24] presented an extension of this approach for valuing a
wide set of investment problems with many embedded ROs taking into account the interaction and strategic
interdependence among the options. Given an American option, with a state variable X
t
, payoff ( )
, X
t
t H that
can be exercised from
0
t until T, its valuation function can be expressed as follows :
- 15 -
( )
( )
*
( , ) max , (1 )
t
t
T
F t X X
t
t t
t
t j
' ' l 1 1
1 1
l = H
! !
l 1 1
l 1 1 + +
(20)
where t is the optimal stopping time ( [ , ]) t T t
and the operator
*
t
[.] represents the risk neutral expectation
conditional on the information available at
n
t , which could also be expressed by
*
t
l
l
. The discount factor
between any two periods is
1
(1 ) df j
= .
Under this approach, the underlying asset evolution is simulated through a Monte Carlo simulation. Then, the
optimal stopping policy for each simulated path is obtained using Bellmans principle of optimality: An optimal
policy has the property that, whatever the initial action, the remaining choices constitute an optimal policy with
respect to the sub-problem starting at the state that results from the initial action [22]. This means, it is possible
to split the decision sequence into two parts, the immediate period and the whole continuation beyond that:
( ) ( )
1
*
1
( , ) max , , ,
n n n n
n t n t t n t
F t X t X F t X df
' ' l 1 1
1 1
= H l ! !
1 1 l
l 1 1 + +
(21)
At this point the LSM makes its key contribution. This approach proposes to compute the continuation for all
previous time-stages by regressing from the discounted future option values on a linear combination of
functional forms of current state variables [25]. Working recursively until
0
0 t = , the optimal decision policy
on each path choosing the largest between the immediate exercise and the expected continuation value can be
computed. Finally, by applying the following equation the American option value is calculated.
( )
( )
( )
1
1
(0) ( ), ( ) F X df
t .
t .
.
t .
\
=
= H
\
_
(22)
As pointed out in [11], traditional investment appraisal methods are typically inappropriate when assessing
transmission investments, since the presence of uncertainties dramatically increases the risk involved in
irreversible large-scale decisions. Moreover, these investments have embedded multiple flexible options which
are difficult to assess but significantly impact on the optimal decision making.
It has been considered as investment alternatives: 1. a FACTS device and 2. a transmission line (TL).
Therefore, the available investments strategies are either investing in the FACTS first, in the line first or jointly
in the FACTS and the line. The strategic flexibility of postponing both investments as well as abandoning or
relocating the FACTS device are compounded options. Hence, these available options are valued by means of
the LSM method, by applying the following Bellmans equations [11]:
1. Option to invest first in the FACTS:
( ) ( ) ( ) ( )
( )
1 1 1
1
1 1 1
*
1
, max , ; , ; , ;
( , ) max
,
n n n n
n
n n
F
F n t R n t A n t TL n t
F n t
t F n t
t X F t X F t X F t X df
F t X
F t X df
' ' 1 1 1
1 1
H
1 1
( ) 1 1 1 1
=
! !
l
1 1
1 1 l
1 1
l 1 1
l 1 1 + +
(23)
- 16 -
2. Option to invest first in the line:
( ) ( )
1 1
*
1 1
( , ) max , ( , ) ; ,
n n n n n
TL
TL n t TL n t F n t t TL n t
F t X t X F t X df F t X df
' ' l 1 1
1 1
= H l ! !
1 1 l
l 1 1 + +
(24)
3. Option to invest in the FACTS and the line jointly:
( ) ( )
( )
1 1
1
& &
& 1 1
&
*
& 1
, max ( , ); ( , ) ;
( , ) max
,
n n n
n
n n
TL F TL F
TL F n t R n t A n t
TL F n t
t TL F n t
t X F t X F t X df
F t X
F t X df
' '
1 1
1 1 H
1 1
1 1
=
! !
l
1 1
1 1 l
1 1
l
1 1 l + +
(25)
where ( , )
n
n
m n t
F t X is the option value and ( , )
n
n
m n t
t X H the profit value, both for the option m (F: FACTS, L:
transmission line, R: FACTS relocation, A: FACTS abandon) at the state n (F: FACTS investment done, L: line
investment done, Ab: FACTS abandon done. Expanding the equation (23):
( ) ( )
( )
1
1
1
*
1
, max ( , ); ( , ) ;
( , ) max
,
n n n
n
n n
R
R n t TL n t A n t
R n t
t R n t
t X F t X F t X df
F t X
F t X df
' '
1 1
1 1 H
1 1
1 1
=
! !
l
1 1
1 1 l
1 1
l
1 1 l + +
(26)
( ) ( )
1
*
1
( , ) max , ( , ); ,
n n n n n
A
A n t A n t TL n t t A n t
F t X t X F t X F t X df
' ' l 1 1
1 1
= H l ! !
1 1 l
l 1 1 + +
(27)
( ) ( )
( )
1 1
1
& &
1 1
*
1
, max ( , ); ( , ) ;
( , ) max
,
n n n
n
n n
F TL F TL F
TL n t R n t Ab n t
F
TL n t
F
t TL n t
t X F t X F t X df
F t X
F t X df
' '
1 1
1 1 H
1 1
1 1
=
! !
l
1 1
1 1 l
1 1
l
1 1 l + +
(28)
In the same way, developing the equations (24) and (25):
( ) ( ) ( )
( )
1 1
1
& &
1 1
*
, max , ; , ;
( , ) max
,
n n n
n
n n
TL TL F TL F
F n t R n t Ab n t
TL
F n t
TL
t F n t
t X F t X F t X df
F t X
F t X df
' ' 1 1 1
1 1
H
1 1
( ) 1 1 1 1
=
! !
l
1 1
1 1 l
1 1
l 1 1
l 1 1 + +
(29)
( ) ( ) ( )
1 1
& , & & * &
1 1
( , ) max , , ; ,
n n n n n
TL F R TL F TL F TL F
R n t R n t Ab n t t R n t
F t X t X F t X df F t X df
' ' l 1 1
1 1
= H l ! !
1 1 l
l 1 1 + +
(30)
( ) ( )
1 1
& & * &
1
( , ) max , ; ,
n n n n
TL F TL F TL F
Ab n t Ab n t t Ab n t
F t X t X F t X df
' ' l 1 1
1 1
= H l ! !
1 1 l
l 1 1 + +
(31)
In the investment cases:
( )
( )
, ,
, ,
, ( )
n n
n
n
m n t s t
s t
t X PV ISW I
.
.
. H = (32)
where
, ,
n
s t
I
.
is the investment cost of the s-th investment strategy at the
n
t -th year. On the other hand, in the
relocation and abandon cases:
( ) ( )
, , ,
, ( )
n n n
n
R n t R t t
t X PV ISW CR
. .
. H = (33)
- 17 -
( )
( )
,
, ,
, ( )
n n
n
n
A n t t
s t
t X SV PV ISW
.
.
. H = (34)
where
,
n
t
CR
.
is the relocation cost and
,
n
t
SV
.
is the scrap value of the FACTS devices at the
n
t -th year.
The present values, expressed in equations (32) to (34), are calculated according to equation (17) and
regarded as underlying assets for the RO valuation. Based on each available investment strategy, the resulting
cash flow is composed. An illustrative example is shown in Figure 5, where one strategy, invest first in the
FACTS, one year later in the line and finally during the third year abandon the FACTS device, is exposed. In all
cases, the investment project begins to operate one year after the investment execution. It is important to notice
that all possible investment strategies and their intrinsic real options have been exhaustively evaluated, that is all
possible combinations among the available flexibility options are assessed.
By applying this procedure, the option values for each strategy are calculated. Hence, the optimal investment
strategy is the one with the highest value. It is important to notice that the optimal decision policy obtained by
the LSM approach is not a deterministic value. Actually, there is an optimal policy for each simulated path.
Therefore, it is possible to determine a probability density function of option values.
Fig.5. Cash flow composition for a transmission investment strategy.
Accuracy of the estimates of the value of the option values can be improved by increasing the number of time
steps N, the number of simulated path . Hence, the Monte Carlo stop criterion applied is the control of the
relative error [26]. Setting 10% c = entails demanding a confidence level in the attributes assessment of 95%.
1
(0)
(0)
1
2
(0),
(0)
m
n
m
n
F
m
s n
F m
n
F
F
c
c o
o
1 ( )
=
( )
\
(35)
- 18 -
where
1
c
is the inverse of the Standard Normal Distribution (SND), ( ) 1 / 2 c the confidence level specified,
( )
1
1 / 2 c c