Upload
buikien
View
223
Download
0
Embed Size (px)
Citation preview
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability
in deregulated power markets
Anlisis de metodologas de clculo de potencia firme
como mecanismos para asegurar la fiabilidad a largo plazo en mercados elctricos liberalizados
Proyecto fin de carrera Escuela Tcnica Superior de Ingeniera (ICAI)
Universidad Pontificia Comillas Madrid
Autor: Carla Monge Liao
Directores: Ernesto Parrilla, Carlos Batlle y Carlos Vzquez
Madrid, junio de 2005
1
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
Summary
The National Electric Reliability Council of the United States (NERC)
defines reliability as: the degree to which the performance of the elements
of the electrical system results in power being delivered to consumers
within accepted standards and in the amounts desired.
The concept of reliability can be broken down into short and long-
term aspects, system security the short-term ability of the system to
withstand deviations between supply and demand in real time, which is
not the object of study in this project, and system adequacy the long-
term capacity of the system to be able to meet the fluctuating energy
demand requirements throughout a broad time-span.
In the liberalized electric market framework, generation planning has
become a decentralised activity, whilst both transmission and distribution
networks remain regulated. The systems long-term reliability is now
determined by market interactions between consumers and generators.
So far no consensus has been reached on how to deal with the
adequacy problem. The debate is focused mainly on two aspects: on the
one hand, on the suitability of settling incentives to enhance investment in
order to guarantee a sufficient margin -it remains uncertain whether an
acceptable level of adequacy will be obtained through these market
mechanisms alone- and on the other, on the way to evaluate the actual
contribution to the system adequacy of the different generators.
The regulator can either choose to intervene indirectly in the market
to promote an adequate level of system reliability, or not to intervene and
to trust in market mechanisms to promote suitable system adequacy.
Interventions can range from establishing payments for available capacity
to incentive additional investments, to requiring consumers to contract a
certain level of firm power, promoting a capacity market.
2
In some of the alternatives proposed, regulators need to be able to
evaluate system generator firm capacity values -the degree to which each
generator effectively contributes to the system reliability-. This project is
going to focus on analysing, comparing and develop some of the most
commonly used firm capacity calculation methodologies.
The four methodologies evaluated in this project are: both Spanish
and US heuristic methods, the convolution model and the cooperative
game model.
The first two evaluate firm capacity through heuristic formulae.
These methods are simple to calculate, but they do not correctly reflect
each generators contribution to the systems adequacy. This is due to the
fact that the contributions do not depend solely on each individual
generators generation capacity, but also on various other factors which
depend on the rest of the generators present in the system.
The convolution model is based on a mathematical technique which
enables the probabilistic dispatch of all the units in the system. This
method carries out a marginal analysis of how each unit contributes to the
system in very specific conditions. Firm capacity values are obtained from
the results of this dispatch, according to given reliability criteria. The
advantage of this method in comparison with the previous ones is that it
takes the full power system into account when calculating individual
generator contributions. The drawback of this method however is the fact
that it does not take into account the cooperative game theory.
The cooperative game model seems to be the most complete of the
four methods analysed. Based on probabilistic production costing and on
the cooperative game theory, it focuses on the fact that the contribution to
the system reliability of the group of generators as a whole is greater than
the sum of the generators individual contributions -it implies that the
combination of all the generators has a synergic effect on the system
reliability-. The excess system reliability generated is distributed
appropriately amongst the contributing generators through linear
3
optimisation techniques. The drawback of this method is that it is too
complex to be able to simulate large-scale power systems without facing a
combinatory explosion problem.
Simulation models using Matlab have been developed in this
project to implement the previously presented methodologies. The
differences that have arisen when calculating generator firm capacities
through the different methods have been analysed and evaluated, as have
their strengths and weaknesses, the key aspects they concern, those they
have in common and those in which they differ.
The cooperative game model has been developed on the basis of an
already existing model. The efficiency of this model has been improved;
enabling it to be run entirely on Matlab, incorporating an optimisation
function and a final result ponderation using both Loss of Load
Probability (LOLP) and Non Served Energy (NSE) criteria.
In order to be able to adapt the cooperative game model to large-
scale hydrothermal power system simulations, model simplifications have
been tested in search of solutions for the combinatory explosion problem
which continue to provide accurate simulation results.
In conclusion, in order for the regulator to obtain reliable firm
capacity values for each system generator in order to be able to use these
values as the basis of a remuneration criterion with the aim to ensure
adequacy in liberalised power markets, the cooperative game model
seems to be the most complete of the four models analysed in this project,
were it not for the simulation problems it entails for large-scale power
systems. The results obtained in this project suggest that the US heuristic
method could be an interesting alternative on which approximations of
the cooperative game model could be based, due to the similarity of the
firm capacity values each of the methods calculate. Further developments
beyond the scope of this project would be to determine the precise power
system conditions under which these approximations are valid.
Analysis of firm capacity calculation methods
as mechanisms to ensure long-term reliability
in deregulated power markets
1. Introduction 5
1.1 Defining power system reliability 6
1.2 Where the adequacy problem arises: Power system
evolution 8
1.3 How to define the adequacy problem 13
1.4 Project aims 16
1.5 Document structure 17
2. State of the art 19
2.1 Regulatory measures to deal with power system reliability 19
2.2 Firm capacity calculation methods 27
3. Convolution models 31
3.1 Mathematical concept of convolution 31
3.2 Convolution applied to power systems 33
3.3 Dispatch of different types of generators 37
3.4 Problematic convolution of hydro groups 43
3.5 Traditional convolution method to determine generator
firm capacity 48
3.6 Convolution of other generators 50
4. The cooperative game model 51
4.1 Motivation 51
4.2 Aims 63
6
4.3 How the cooperative game method works 65
4.4 Additional considerations 74
4.5 The advantages of the cooperative game model 78
4.6 The disadvantages of the cooperative game model 80
5. Results 84
5.1 Convolution model 84
5.2 Restricted Cooperative game model 91
5.3 Full cooperative game model 98
5.4 Comparison of the results obtained 102
6. Conclusions 107
7. Further developments 112
8. References 114
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
7
Figure index
Fig. 1.1 Illustration of LOLP and NSE values 14
Fig. 3.1. From LDC to ILDC 33
Fig. 3.2. Inverted load duration curves illustrating the effect of
successive generators being dispatched in the system 35
Fig. 3.3. Reading data on ILDC curves 36
Fig. 3.4. Scenarios A and B with G1 active and inactive respectively 38
Fig. 3.5. Convolution of scenarios A and B to obtain a new ILDC 39
Fig. 3.6. LDC 39
Fig. 3.7. ILDC 39
Fig. 3.8. Active scenario, p = 0.85 40
Fig. 3.9. Inactive scenario, p = 0.15 40
Fig. 3.10. Convolution results 40
Fig. 3.11. Hydro plant dispatch mechanism 42
Fig. 3.12. Load-duration curve for a period of three hours 43
Fig. 3.13. Inverted load-duration curve for a period of three hours 44
Fig. 3.14. Dispatch of T1 first 45
Fig. 3.15. Dispatch of T2 first 45
Fig. 3.16. Dispatch of all units: T1, T2 and H1 47
Fig. 3.17. Convolution method algorithm 49
8
Fig. 4.1. Scenario where no generators have been dispatched yet 54
Fig. 4.2. Scenario where generators have been dispatched 54
Fig. 4.3. Scenario with a high proportion of thermal units 56
Fig. 4.4. Scenario with a low proportion of thermal units 56
Fig. 4.5. Probability of scenario occurring = 0,095 57
Fig. 4.6. Probability of scenario occurring = 0,045 58
Fig. 4.7. Ideas behind the cooperative game 65
Fig. 4.8. Probabilistic dispatch algorithm. 69
Fig. 4.9. Cooperative game model 79
Fig. 5.1. Relative errors for each unit in each restricted combination 93
Fig. 5.2. Maximum percentage errors in the different restriction
combinations 93
Fig. 5.3. Medium-sized power system results for the restricted
cooperative game model 94
Fig. 5.4. Proportion of firm capacity distributions made by each model
106
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
9
Table index
Table 3.1. Thermal and hydro unit data 43
Table 3.2. Dispatch results 45
Table 3.3. Dispatch parameters: LOLP, NSE and ERC 46
Table 3.4. Dispatch parameters for the T1, T2 and H1 dispatch 47
Table 3.5. Full dispatch results 47
Table 4.1. Generator data 57
Table 5.1. Small-sized power system input data for the convolution
model 85
Table 5.2. Small-sized power system absolute results for the
convolution model 85
Table 5.3. Small-sized power system relative results for the convolution
model 85
Table 5.4. Medium-sized power system input data for the convolution
model 89
Table 5.5 Medium-sized power system absolute results for the
convolution model 89
Table 5.6. Medium-sized power system relative results for the
convolution model 90
Table 5.7. Small-sized power system results for the restricted
cooperative game model 92
Table 5.8. Spanish power system input data 96
10
Table 5.9.Spanish power system results for the restricted cooperative
game model 97
Table 5.10. ERC cooperative game results for a small-sized power
system 99
Table 5.11. ERC cooperative game model results for a medium-sized
power system 100
Table 5.12. Comparative table of absolute results for the small case
example 102
Table 5.13. Comparative table of relative results for the small case
example 102
Table 6.1. Table comparing model characteristics 107
Table 6.2. Firm capacity evaluation criteria for each method studied 111
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
5
1. Introduction
Electricity is an asset which modern society has come to depend
upon entirely. The repercussions of a lack of electricity stretch deep into
the social, economic and political domains of any country. Hence it is of
primordial importance for politicians, system regulators and operators
that power systems function correctly, avoiding emergency situations
or blackouts. One of the fundamental measures of the correct
functioning of a power system is its reliability, provided through the
interaction between its principal activities: generation, transport and
distribution systems.
In the liberalized electric market framework generation planning is
decentralised, whilst both transmission and distribution networks
remain regulated. As far as electricity generation is concerned, the
search for an adequate regulatory measure to ensure long-term
reliability in liberalised power markets remains an open issue to the
present date. The regulators concern for the long-term reliability issue
is reflected through the continuous search for adequate mechanisms to
ensure the long-term availability of sufficient generation capacity in
power systems. [Prez-Arriaga 1999] The regulator can either choose to
intervene indirectly in the market to promote an adequate level of long-
term system reliability, or not to intervene and to trust in market
mechanisms alone to promote adequate long-term system reliability,
intervening only in cases of emergency.
Some of the proposed alternatives to deal with the long-term system
reliability issue require the regulator to evaluate the individual
contributions of each generator to the system reliability as a whole with
three main aims. The first aim is to be able to supervise the fulfilment of
Chapter 1. Introduction
6
these capacity obligations, in order to be able to establish an adequate
degree of short-term reliability in the system. The second aim is to be
able to make coherent predictions on the need to expand the power
systems generation capacity according to demand evolution, and hence
be able to maintain an adequate long-term level of system reliability.
The third aim is to be able to provide fair economic incentives, for each
generator in the system and for incoming generators, on the basis of
their individual contributions to the system in order that they continue
contributing optimally to cover the required demand. Not only will
these three aims enable the regulator to establish an adequate degree of
short-term reliability in the power system and to maintain it in the long-
term, but just as importantly, to make the system generation capacity
evolution and the subsequent adequate system reliability evolution a
sustainable reality.
The object of study of this project is generator firm capacity or in
other words, the individual contributions of each generator to the
system reliability as a whole. These firm capacity values constitute
one of the basic sets of data the regulator needs in order to be able to
comprehend and evaluate the power system to be dealt with. As there
is no consensus as of yet about how these firm capacity values should
be calculated, this project is going to focus on analysing and comparing
some of the most commonly used firm capacity calculation
methodologies.
1.1 Defining power system reliability
The National Electric Reliability Council of the United States
(NERC) defines reliability as: the degree to which the performance of
the elements of the electrical system results in power being delivered to
consumers within accepted standards and in the amounts desired. In
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
7
other words, a reliable power market is that which provides a quantity
and quality of electricity supply to satisfy all its consumers.
Furthermore, the concept of reliability can be broken down into short
and long-term aspects [Prez-Arriaga 2001].
System security is the short-term ability of the system to withstand
deviations between supply and demand in real time. It is dealt with
by the System Operator (SO) via protection devices, ancillary service
procedures and ad hoc markets for operating reserves which stabilise
the system by matching electricity production and demand
instantaneously. As a consensus has been reached regarding the
measures to provide security in deregulated power systems, and for
the time being these are functioning satisfactorily, this aspect of
reliability will not be an object of study in this project.
System adequacy is the long-term capacity of the system to be able
to meet the fluctuating energy demand requirements throughout a
broad time-span. Quantitative analysis of this variable is complex
for various reasons. Amongst others, electricity cannot be stored and
neither supply availability nor demand can be fully predicted. This
makes it complicated to always have enough generation resources
available to cover the constantly fluctuating demand requirements.
In addition, power generation plants require large investments, at
least a two to three year lead-time period to be built and operative,
and then have an average life-span of thirty years or more. Hence it
is necessary to make predictions of generation requirements well in
advance to cover future demand as accurately as possible, avoiding
capacity shortages due to underinvestment on one hand and
capacity surplus due to overinvestment on the other [Ford 2000].
Chapter 1. Introduction
8
As so far no consensus has been reached on how to deal with
adequacy, this project is going to focus on analysing methods to
calculate firm capacity as the basis of various potential regulatory
measures aimed to maintain an appropriate level of generation
investment to ensure power system adequacy.
1.2 Where the adequacy problem arises: Power system evolution
Traditionally the electric power industry did not have great
adequacy problems. Long-term guarantee of supply was ensured by
regulators through a central planning system. In view of their
predictions of future demand requirements, they made a trade-off
between generator investment costs and a socially acceptable degree of
system reliability. In order to avoid the social and political problems
that power shortages entail, they tended to overweight the reliability
factor, programming capacity reserves as a large percentage surplus of
estimated peak demand. As a result there was a high degree of system
adequacy, but average electricity prices for consumers were also higher
than they are in present day power markets.
With the restructuring of electric power systems, although the
transmission and distribution systems have remained as regulated
businesses, the power generation system has been liberalised. Central
planning has been removed and the generators themselves make their
own investment decisions. Therefore the systems adequacy level is
now determined by market interactions between consumers and
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
9
generators. It is not clear however that an acceptable adequacy level
will be obtained through market mechanisms alone1.
1.2.1 Power market characteristics
Relying only on market mechanisms implies that all generators
whose investment is well adapted to the existing demand and to the
existing generator collective as a whole will be able to recover their total
costs through electricity market prices alone. [Schweppe 1988], [Prez-
Arriaga 1997] This idea is based on classic economic theory, where
equilibrium is maintained between the availability of an asset, its price,
and the demands response to both these factors. In our particular case,
the unexpected outage of a generating unit or a sudden demand
increase could cause a shortage of electricity leading to price spikes.
This would cause new generation investments to rise on one hand, and
the demand to fall on the other until a new equilibrium between supply
and demand was reached once again.
In theory this would account even for peaking units with the
highest variable costs to recover their total costs. They would be able to
do so when they are the only units supplying power at prices as high as
the demand side will accept, until supply and demand equilibrium is
re-established. As it has already been described however, the long lead-
1 This new market environment not only affects adequacy by putting investment
decisions in players hands, but also affects the management of generation units.
This is the case of hydro plant reservoir management. Traditionally, the central
planning regarding these hydro plants was much more oriented to assure system
security. However, it is being observed all around the world that private utilities,
more short-term oriented, tend to diminish the weight of the system security
variable in their hydro management models.
Chapter 1. Introduction
10
times necessary for capacity expansion in power systems makes the
natural economic cycles of power supply lag behind the demand, so
leading to the known boom and bust cycles [Ford 1999].
This ideal market behaviour therefore only applies to a scenario of
perfectly adapted production units, which in practice is never the case.
In addition, electricity markets have various characteristics which make
them react differently from classic economic theory.
To begin with, it is very difficult to maintain equilibrium between
supply and demand in a system where the demand side is inelastic,
passive and immature. In electricity markets most consumers pay
regulated electricity tariffs independently from power market prices.
They are unaware of real-time price fluctuations and therefore cannot
react to them. In addition, the demand side tends to assume that
electricity is an asset which is always available; it is unaware of the
existence of the adequacy problem and of the need to pay for a service
to guarantee supply availability2.
Even if the demand side did have the choice to pay for a desired
level of system adequacy, the nature of electricity distribution as a
network that supplies all standard consumers indifferently makes it
impossible to differentiate the consumers that have paid for higher
levels of adequacy from those who have not. Although it could be
physically possible to disconnect standard consumers selectively from
the network, economically it is unviable and this leads to the problem
of free riding where consumers with lower electricity tariffs, paying
2 Besides, at least in a relevant number of countries, the consumers trust on the State
providence remains fully active.
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
11
for a lower degree of system adequacy, could benefit from those paying
higher tariffs, accounting for a higher degree of system adequacy.
Finally there is a third problem, probably the most relevant of all,
that is: generator investment risk aversion. Due to the fact that peaking
units are only operative during very specific periods of time, their
income is very volatile and hence investors do not feel the security that
they will be able to recover their costs through electricity market prices
alone. This problem is particularly serious for peaking units which have
to recover their fixed costs in only a few hours throughout the year. As
these units lack the suitable economic incentives to expand the
production system as demand grows, they demand an additional
remuneration, independently of their expected benefits, in order to be
able to build new generation units with lower investment risks.
1.2.2 Regulatory measures
This leads us to have to determine what role the regulator should
play to ensure the correct functioning of power markets. There are
currently two main lines of action that have been adopted by power
market regulators throughout the world.
The first is an approach of minimal intervention where the regulator
only intervenes in cases where the market mechanisms diverge
significantly from the natural market equilibrium. In other words, when
the situation becomes critical, rationing is near or price spikes are
continuous [Prez-Arriaga 1997].
The second is an approach of active regulatory intervention where
the regulator establishes fixed rules and measures to guarantee the
functioning of the power market at socially acceptable standards
[Prez-Arriaga 2001].
Chapter 1. Introduction
12
The implementation of each regulatory approach depends greatly
on the characteristics of each power system and on the political
preferences of each country. In practice however, the regulator usually
implements some form of regulatory mechanism to provide an
additional remuneration for generators in order to ensure a minimum
level of system adequacy. Since the main aim of these payments is to
enhance system adequacy, the idea is that they should be based on each
generators actual contribution to the systems reliability, but in practice
they are made in very different ways.
Several regulatory mechanisms have been implemented worldwide
to enhance generation adequacy in deregulated power markets, each
with different levels of success. To the present date however, none of
them has managed to fully resolve the long-term reliability problem in
power markets.
Initially Chile and Argentina implemented different forms of
capacity payments, followed by most other Latin American countries.
Energy-only markets were initially adopted by Nordic European
countries and California, but currently only Australian power markets
rely on market mechanisms alone. Capacity markets constitute an
intermediate measure between the more interventionist capacity
payment methods and the less interventionist energy-only market
methods. This method has been adopted in places such as the pools of
PJM (Pennsylvania, New Jersey and Maryland), New England and New
York in the USA [Creti 2003]. Specific analysis of these methods escapes
the scope of this project, however, their characteristics will be outlined
in section 2.1. The different methods used to evaluate firm capacity
how each generator contributes to the systems adequacy is the main
aim of this project and will be analysed in section 2.2.
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
13
Definite conclusions cannot be drawn as of yet due to the limited
time these measures have been active, but recent electricity shortages
that have occurred for example in Chile, California and Brazil, or
serious problems that have arisen in the Norwegian and Australian
systems are evidence enough that supply adequacy is a variable that
should not be taken lightly in deregulated power markets.
Nevertheless, so far in regional markets such as the Internal Electricity
Market of the European Union, Mercosur in South America, the Central
American Electric Market or the Regional Transmission Groups in the
USA, the aspects regarding system adequacy have seldom been dealt
with. The issue was first addressed relatively recently be the Central
American Market and by the Spanish and Portuguese regulators when
defining the Iberian Electricity Market. Only after the California crisis
has the issue been addressed in the European Union and in the Green
Book on supply security [Prez-Arriaga 2001].
1.3 How to define the adequacy problem
Having established the precedents of how the adequacy problem
arises, we are now faced with three fundamental problems when
dealing with adequacy in deregulated power systems. Firstly, a
measure of adequacy must be defined for the system as a whole.
Secondly, according to this measure, each generators contribution to
the systems adequacy must be evaluated. Finally, a payment method
must be determined on the basis of each generators contribution in
order to maintain appropriate generator investment incentives to
ensure system adequacy.
In order to tackle the first problem of defining a measure of
adequacy for the system, different criteria can be adopted. The first
Chapter 1. Introduction
14
adequacy measure is the loss of load probability (LOLP). This
probability indicates the expected amount of hours in an established
period of time where the demand cannot be covered by the generators.
For example, US power systems establish an overall system LOLP at
one day of system failure in ten years [PJM 2004]. The second measure
is non-served energy (NSE). This is the expected amount of energy out
of a total predicted demand requirement that generators will be unable
to cover. Thirdly, the two principal methods described previously can
be combined with different weighting factors to obtain yet other
measures of the systems adequacy. The concepts of loss of load
probability and non-served energy are illustrated in Fig. 1.1
0
0,2
0,4
0,6
0,8
1
1,2
0 50 100 150 200 250 300 350 400
Probability [h]
Pow
er [M
W]
Original ILDC curve T1 T2
LOLP = 0,5
NSE = area under thelast ILDC curve
0
0,2
0,4
0,6
0,8
1
1,2
0 50 100 150 200 250 300 350 400
Probability [h]
Pow
er [M
W]
Original ILDC curve T1 T2
LOLP = 0,5
NSE = area under thelast ILDC curve
Fig. 1.1 Illustration of LOLP and NSE values
The second problem concerns the evaluation of each generators
contribution to the systems adequacy on the basis of one of the
methods described above. The individual contribution of each
generator is known as firm capacity. It is a measure of the power a
generator can assure it will be able to provide at any given moment for
the consumer. There are various possible methods to evaluate firm
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
15
capacity and each provides different results concerning each
generators contribution to the systems adequacy (see chapter 2).
In addition, different calculation methods must be determined for
different types of generators. For example, the relatively constant
reliability of thermal units is very different to the reliability of hydraulic
units which depends greatly on external climatic factors and moreover,
on the way the utility that owns the reservoir manages its water
reserves. This project is going to focus on how to calculate firm capacity
for the different types of generators that could appear in deregulated
power systems: thermal units and hydro units with energy limitations.
It is also going to focus on the analysis of the differences that arise when
using different methods to calculate each generators firm capacity
contributions to the systems adequacy.
Eventually these calculations could be made by the generators
themselves, deciding the capacity they are willing to offer the regulator
to support system adequacy -such as the reliability options mechanism
[Vzquez 2002]- but in all the other cases it is the regulator who
administratively determines the amount of capacity he will admit as
firm capacity.
To tackle the third problem of how to calculate the total
remuneration of the generators in order to obtain an acceptable level of
system adequacy there are also various possible solutions which
provide us with different results. One option is to distribute payments
from a fixed sum of money destined to cover capacity reserves. This
implies that the money received will depend on the number of
generators in the system and on the initial sum of money allocated.
Chapter 1. Introduction
16
Another option is to pay each unit independently, on the basis of its
contribution to the systems adequacy3.
This project is going to focus on the first and second of these issues
related to the adequacy problem.
1.4 Project aims
Parting from the basis that in order for regulators to be able to
establish, control and sustain an adequate degree of reliability in power
systems they need to be able to evaluate system generator firm capacity
values or the degree to which each generator effectively contributes
to the system reliability this project sets out with the following aims:
To evaluate the different measures which can be used to define and
determine system reliability the differences that arise when using
LOLP or NSE criteria and the differences that arise when calculating
generator firm capacities through different methods.
To analyse and compare three methodologies to evaluate firm
capacity: both Spanish and US heuristic methods, the convolution
method and the cooperative game method.
To develop simulation models using Matlab to implement the
previously presented methodologies to evaluate firm capacity in
order to be able to study their strengths and weaknesses, the aspects
they have in common and those in which they differ.
3 This option would probably imply the need to determine which is the maximum
amount of capacity the regulator is willing to pay.
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
17
The last of these models will be developed on the basis of an already
existing cooperative game model. The aim is to update and improve
the efficiency of this model; enabling it to be run entirely on
Matlab, incorporating an optimisation function and a final result
ponderation using both LOLP and NSE criteria.
To adapt the cooperative game model so that it is able to simulate
large-scale hydrothermal power systems. Due to the known fact that
the cooperative game model is too complex to be used for large
system simulations because of the combinatorial explosion of
variables to evaluate, the aim is to develop a simplification of the
model which continues to provide accurate simulation results.
1.5 Document structure
In the introduction of this project the concept of power system
reliability has been presented. The reasons that make it a particularly
problematic issue in the present deregulated power market framework
have been tackled. Stressing the fact that no consensus has been reached
regarding the definition of adequate regulatory measures to ensure
long-term system reliability, the need to calculate individual generator
contributions to system reliability through various possible firm
capacity calculation methodologies has been presented. Finally, the
aims of this project have been outlined.
Chapter 2 is subdivided into two sections. The first describes
international regulatory measures which have been applied to deal with
power system adequacy. The second describes the most common
methodologies which have been approached in different countries so
far to calculate firm capacity.
Chapter 1. Introduction
18
Chapter 3 describes the first of these methodologies in detail: the
convolution method. The general functioning of the method is
described together with a description of how the model has been
developed in this project, both for thermal and for hydro units. A
fundamental flaw in hydro unit dispatch is also described as an
important conclusion of this thesis.
Chapter 4 describes a second methodology which also uses
convolution techniques: the cooperative game method. This model was
initially presented in [Batlle 2000] and has been further developed in
this project.
In chapter 5, all the results obtained from the simulation models
developed throughout the project are displayed and commented,
together with a comparative analysis of all of them.
Chapter 6 finally wraps up all the conclusions and new insights
obtained from this project on the subject of firm capacity calculations
and long-term power system reliability.
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
19
2. State of the art Equation Chapter 2 Section 1
2.1 Regulatory measures to deal with power system reliability
Regulatory approaches to enhance generation adequacy in
deregulated power systems can be classified into eight broad categories.
[Prez-Arriaga 2001]
2.1.1 Energy only markets
In energy only markets, following classic economic theory, there is
no regulatory intervention in the market [Ford 1995]. The drawback of
this approach is that it ignores the existence of power market
characteristics such as the passivity of a significant part of the demand,
investor risk aversion or free riding which have already been
described in section 1.2.1: Power market characteristics. These
distinguishing characteristics suggest that electric power markets will
not behave in the same way as classic markets, possibly resulting in
undesirable adequacy levels with episodes of electricity shortages and
high market prices.
This approach has been followed in Nordic European countries
the NORDEL group. However, Finland and Sweden have already
begun to apply the approaches considered in sections 2.1.4: Purchases
of peaking units by the system operator and 2.1.5: Purchases of reserves
by the system operator. Norway is also beginning to consider these
options due to the growing concern that the large surplus of generation
capacity experienced in previous years has not been following the same
trend in recent years. Energy only markets have also been used in
California where they have resulted in serious generation shortages
coupled together with market design flaws and other problematic
Chapter 2. State of the art
20
circumstances [Joksow 2001]. The only country where an energy only
market is implemented at the present date is Australia, where
generation shortages have been avoided so far.
2.1.2 Long term capacity payments
Capacity payments are issued to each system generator on the basis
of an administratively determined amount of money, proportional to
the estimated contribution of each generator to the systems reliability
[Sanz 1999]. The amount of money to be received by each generator is
computed in advance annually, but the payments can be performed on
a monthly or daily basis. In practice generators only receive capacity
payments for the periods of time in which they are available.
These estimated contributions are defined as firm capacity values
and can be calculated through different methods such as mathematical
modelling, historical operation data or heuristic procedures. These
different firm capacity calculation methodologies are the object of study
of this project and will be described in section 2.2: Firm capacity
calculation methods.
The long term capacity payment approach has a theoretical
justification. It has been demonstrated that market prices alone can fully
account for the total (fixed and variable) costs of a well adapted
generation mix in perfectly competitive market conditions. In the same
way, it can be proved that any given amount of additional generation
adequacy desired by the regulator, above that provided by the market
alone, requires a capacity payment for each generator on the basis of
each of their theoretical contributions to the system reliability in order
to maintain the systems economic viability [Prez-
Arriaga&Messeguer 1996].
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
21
The main advantages of the long-term capacity payment approach
are the reduction of peak unit income volatility, the reduction of
investor risk aversion, the fact that it provides incentives for mature
generation units to remain in the system and the fact that it hardly
interferes with the market, allowing the demand side to mature.
The approach has its disadvantages however, as adequacy is not an
identifiable or tangible commercial product which can be clearly
defined or calculated in order that generators can be remunerated for it.
The term firm capacity attempts to provide a measure for each
generators contribution to system adequacy, but a credible and feasible
definition for firm capacity values is complex, and no consensus has
been reached as of yet as far as this concept is concerned. The
calculation methodologies for firm capacity values become even more
complicated in hydrothermal generation scenarios due to the intrinsic
differences between these two types of plants. In addition, this method
does not encourage the availability of generation units when the system
is short of reserves, and hence there are no incentives for units to
increase system reliability during short margin seasons. Finally, this
method can also induce a perverse incentive for generators to lie in
their availability status in order to receive more capacity payments.
This approach was implemented in Chile for the first time in 1981,
and with different variations has also been used in various Latin
American countries such as Argentina, Peru, Colombia, Brazil, Central
American countries and Spain.
2.1.3 Short term capacity payments
This approach, like the previous one consists in paying an
administratively determined amount of money to each generator for
Chapter 2. State of the art
22
their reliability contributions to the system. The difference in this case is
that the amount to be paid is calculated daily by overestimating the loss
of load probability for the following day [Saunders 1999].
The drawbacks of this method are basically the same as those for
the previous one. This approach however is not based on a firm
capacity criterion but on a loss of load probability criterion which can
lead to undesirable consequences. The reasons for this are that loss of
load probability can be manipulated by suppliers to increase
production plant revenues, giving way to perverse operating
incentives.
This mechanism has been implemented in the initial pool
mechanism of England and Wales, where the ex ante method of
computation of market prices was used. Due to the large interference of
this mechanism in the day-ahead market, which is extremely
undesirable to preserve system adequacy, the pool mechanism was
substituted by NETA in 2000.
2.1.4 Purchases of peaking units by the system operator
In this approach the System Operator, following the instructions of
the regulator, purchases a certain number or all of the peaking units of
the system. The incurred costs are charged on the consumers as an extra
cost associated to reliability enhancement. These peaking units are
under the System Operators absolute control, and will be used to cover
any generation deficits that may appear whilst trying to meet the
system demand.
The aim of this approach is to avoid generation units which provide
capacity at the margin from leaving the market when their revenues are
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
23
too volatile or insufficient to cover their total operating costs. The units
concerned would typically be peaking units, mature or inefficient units.
The drawbacks of this approach are that it is strongly
interventionist and may seriously interfere with proper market
functioning. Generators not belonging to the System Operator, as well
as potential new entrants, may judge that market prices depend too
much on the purchasing decisions dictated by the regulator. Another
inefficiency of this method is that it only remunerates part of the system
units, although all of them provide adequacy to the system. In this
sense, it doesnt give any operation incentives to non-ISO units in order
to enhance system adequacy, and hence it unfairly reduces the
remuneration for their investment.
Purchasing of peaking units is currently being used in Finland and
Italy.
2.1.5 Purchases of reserves by the system operator
This idea is based on the allocation by means of an ad hoc auction
of some capacity payments to a subset of generation units. These are
known as strategic capacity reserves. The chosen units usually have
to meet some specific requirements such as: to belong to the kind of
technology that provides capacity at the margin. These units would
typically include peaking units, mature or inefficient units as in the
previous method. In the case of the NETA pool they are the most
reliable system units. As it can be observed, this method is closely
related to the purchase of peaking units explained previously, the
difference being in this case that the regulator purchases capacity
reserves a given volume of a units production for a defined time
Chapter 2. State of the art
24
scope, typically a year or a season rather than purchasing the whole
unit.
By imposing this mechanism it is possible to make an impact on
generation adequacy. However, besides the drawbacks presented in
section 2.1.4, an additional concern that arises with this method is the
breakdown of the market into two different parts: the competitive
market and the units in control of the System Operator as strategic
reserve units. This approach may be suitable for a limited amount of
time for instance whilst some initial surplus of capacity disappears
because of natural demand growth but this market division may
result in undesirable generator behaviour patterns.
In England and Wales, under NETA, the System Operator can
purchase any amount of operating reserves in advance. Norway is also
presently securing long-term operating reserves, but only with a yearly
anticipation [Wolfgang 2004].
2.1.6 Regulatory determined competitive bidding
In this case, although free entrance of new generation is allowed
into the system, the regulator or some other administrative authority
supervises that there is no threat of insufficient generation adequacy
according to some pre-established criterion. If it considered that there
was a lack of entry of new generation, this authority could start a
competitive bidding process for the addition of the required extra
generation.
This approach was adopted by countries such as France and
Portugal in the implementation of the European Directive 96/92/EC on
the Internal Electricity Market. This was later amended by a new
European Directive presented in Stockholm by the European
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
25
Commission in April 2001, which explicitly forbade the tendering
procedure to acquire new generation, except when used as an
exceptional measure for reasons of supply security.
The evaluations for methods 2.1.4: Purchases of peaking units by
the system operator and 2.1.5: Purchases of reserves by the system
operator are also valid here. This approach is strongly interventionist
and may seriously interfere with the proper functioning of the market.
Market agents, as well as potential new entrants, may judge that market
prices depend too much on the purchasing decisions dictated by the
regulator. The most important drawback of this mechanism however is
that no new generation can be built outside the auction. Investors will
hence have to wait for the auction in order to obtain extra
remuneration.
2.1.7 Capacity markets
The motivation behind this approach is to guarantee a regulated
generation adequacy level for the system. This is carried out by
imposing specific purchasing commitments of firm production capacity
on all the consuming entities, usually known as Load Serving Entities
(LSEs). The approach specifies the commitment that each generation
unit must maintain in real time, i.e. the commercial product associated
to the concept of generation adequacy.
The advantage of capacity markets is that there exists an identifiable
commercial product associated to generation adequacy, as well as a
commitment by the agents to purchase the product. Despite this, the
concept of firm capacity and the conditions of delivery remain
somewhat ambiguous, particularly when hydro units are involved. The
regulator determines both the total amount of desired firm capacity and
Chapter 2. State of the art
26
the rules by which to calculate the firm capacity provided by each unit.
There is no doubt that these rules could easily be contested in the case
of coexistence of thermal and hydro units of varied reservoir capacities.
In downsides of this method are that consumers remain fully
exposed to the potential high prices in the energy market. In addition,
the generators commitment to supply power in times of need is not
precisely defined i.e. it is complicated to define the availability of a
hydro unit with limited storage capacity.
Market mechanisms determine the price of capacity, which may be
very volatile depending on the tightness of the margins of installed
capacity over the system peak load and the anticipation of the auctions
with respect to real time. The volatility may be reduced however, by
increasing the time horizon of the auctions. Otherwise, the uncertainty
in generator remuneration may not gain much in terms of stability
using this approach. Thus, investors risk aversion will not be minimised
and the system could fail in providing adequacy.
This approach has been adopted in the PJM power market and in
other regional entities of eastern USA such as New York and New
England.
2.1.8 Reliability contracts
This method was initially proposed and implemented in the
Colombian power industry by the research group of the IIT. The idea
has been further developed [Vzquez 2003, Vzquez 2002] as a way to
solve the adequacy problem in deregulated power systems.
This approach proposes to establish an organised market in which
the regulator requires the Market or System Operators to buy a
prescribed volume of reliability contracts from generators on behalf of
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
27
the whole demand. These reliability contracts allow consumers to
obtain a price cap on the market price in exchange for a fixed
remuneration for the generators. Additionally, consumers obtain a
satisfactory guarantee that there will be enough available generation
capacity whenever it may be needed. Otherwise the generators will be
penalised. The generators are also compensated economically for this
service; the higher the contribution to the reliability of the system, the
higher their compensation will be.
2.2 Firm capacity calculation methods
Three main categories of methods to calculate the firm capacity of a
generator can be found throughout the world: methods based on
heuristic procedures, methods based on convolution techniques and
methods based on optimisation techniques.
2.2.1 Methods based on heuristic procedures
These methods have the advantage of being simple to calculate, but
they do not accurately reflect each generators contribution to the
systems adequacy. This is mainly because each generators
contribution does not depend solely on its own generation capacity but
also on that of the rest of the system generators. There are many factors
in play when calculating each generators contribution as a function of
the rest of the system generators contributions, such as the time of
dispatch, considerations on when to dispatch hydraulic units, cost
considerations when dispatching etc.
Examples of these methods are the firm capacity in the PJM capacity
market which is defined as a function of each generators maximum
power output and its outage rate [PJM 2004]:
Chapter 2. State of the art
28
For thermal units:
max (1 )FirmCapacity P FOR= (Eq. 2.1)
For hydro units:
dispatchFirmCapacity P= (Eq. 2.2)
In Spain firm capacity is defined as a function of each generators
net power output and its average power output limited by resource
availability [OMEL 2001].
The general equation is:
12
resource limitationsnetFirmCapacity P P = + (Eq. 2.3)
For thermal units:
1 (1 )2
resource limitationsmaxFirmCapacity P FOR P = + (Eq. 2.4)
For hydro units:
[ ]1 0,35 ( )2 max max min
FirmCapacity P P P= + + (Eq. 2.5)
2.2.2 Methods based on convolution techniques
The advantage of these methods in comparison with the previous
ones is that they take the full power system into account when
calculating individual generator contributions.
In these methods the load-duration curve is convoluted, followed
by the convolution of each of the systems generators. In this manner all
the generators are dispatched depending on their nature (thermal,
hydro, etc.). Then each of the generators is removed from the system by
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
29
deconvolution, either individually or in different combinations with the
other generators from the system. The absence of each generator or
group of generators leaves different amounts of non-served energy in
the system. There are two possible variations of this method:
The first the standard convolution method defines firm
capacity as the power equivalent necessary to cover the non-served
energy remaining in the system after the deconvolution of each
individual generator on its own. This method was initially
implemented in Peru and has been proposed in Panama. Its main
disadvantages are that it doesnt take into account the different
dispatch variations that could occur within the system. In addition,
it doesnt take into account the fact that the sum of the generators
individual contributions to the system reliability is less than the
contribution of the group of generators as a whole, and therefore it
doesnt distribute this excess reliability amongst the contributing
generators. This important aspect is the basis of the cooperative
game theory and is taken into account in the cooperative game
method described next. It will be commented in more depth in
chapter 3.
The second the cooperative game method described in
[Batlle 2000] defines firm capacity as the solution of a cooperative
game. It has the advantage that it does take into account all possible
generator combinations in the system, unlike the previous method,
and it also takes into account the distribution of the excess
reliability provided by the group of generators as a whole.
However, it appears to be unviable for large systems where all the
possible generator combinations would be far too many and would
saturate the system.
Chapter 2. State of the art
30
The use of these methods still leaves us with problematic
considerations such as when to dispatch hydraulic units and cost
considerations when dispatching, which will have to be studied.
2.2.3 Methods based on dispatch simulations under critical
conditions
This method relies on the simulation of how all the generators are
going to be dispatched in a system under critical hydro conditions. On
the basis of the simulation results each generators firm capacity is
determined as the average power provided by each unit in every critical
hour previously defined. The disadvantages of this method are that
once again it does not take into account all possible dispatch
combinations, and that all of the groups contribute to the system
reliability. Hence, it doesnt remunerate all the system generators, and it
uses economic criteria to establish the remuneration process which can
lead to unfair remuneration mechanisms. In addition, in many cases the
simulations results are very sensitive to the input data. Examples of this
method implementation are the present and the future method in
Colombia [CREG 2000], and the current procedures used in Argentina,
Peru, Bolivia or Chile.
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
31
3. Convolution models Equation Chapter 3 Section 3.1
Convolution is a mathematical, probabilistic technique which has
been widely used to model power system scenarios and generator
dispatch as it enables the handling of historical data through a
probabilistic approach. Amongst the countries which have used
different convolution techniques to model their power systems are
Peru, Panama, Bolivia and Chile.
3.1 Mathematical concept of convolution
The main problem this power system modelling aims to resolve is
whether there is sufficient capacity in the system to cover predicted
demand. Clearly, the total capacity provided by the system generators
must exceed the peak demand. However, as each group has an outage
probability, a certain degree of excess capacity must be considered in
order to be able to adequately fit the system demand requirements. In
other words, a value of loss of load probability must be considered for
the system, and the aim is to minimise this value.
Each generator is therefore modelled as a random variable where
if the group is available it can contribute a given power output to the
system, and if it isnt available then its contribution will be zero.
Hence, the capacity all the generators can provide to the system is the
sum of a set of random variables.
Let us consider the set of random variables of three generators X ,
Y and Z . Each variable can be associated to a density function ( )xf x ,
( )yf y and ( )zf z . The probability that for example generator X will be
unable to supply a power load greater than x will be:
Chapter 3. Convolution models
32
[ ] ( )( )x
xP X x F x f x dx
= (Eq. 3.1)
The density function therefore indicates the probability that an
event will occur in a given interval:
[ ][ ] [ ]
[ ] ( ) ( )x dx x
x x
P x X x dx P X x dx P X x
P x X x dx f x dx f x dx+
+ = +
+ = (Eq. 3.2)
We now take two independent random variables X and Y .
Consider the sum of these variables to take on a defined value z . If in
addition Y takes on a value y , X will then have to be equal to z y .
The probability that Z is equal to z and Y is equal to y must therefore
be proportional to:
( ) ( )x yP Z z Y y f z y f y = = (Eq. 3.3)
Integrating this expression for all possible values of Y we obtain
what is known as the convolution integral:
( ) ( ) ( )z x yf z f z y f y dy
= (Eq. 3.4)
This formula expresses the density function of a random variable Z
as the convolution of the density functions of two other independent,
random variables X and Y .
z x y= (Eq. 3.5)
Convolution also has an inverse function known as deconvolution:
1x z y= (Eq. 3.6)
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
33
3.2 Convolution applied to power systems
For power simulations it is more common to work with the inverted
load-duration curve (ILDC) rather that with the load density function
(LD). This first function is simply the complementary function of the
latter one, and tells us the load that will go unserved in the system
rather than the load that will be served:
( ) 1 ( )ILDC LDF d F d= (Eq. 3.7)
The ILDC curve (with probability on the y -axis and unserved load
in MW on the x -axis), can be easily obtained from the usual load-
duration curve (with load in MW on the y -axis and time in hours on
the x -axis), as is shown in Fig. 3.1.
Fig. 3.1. From LDC to ILDC
As a starting point for the convolution model, the predicted ILDC
does not necessarily have to come from only one LDC. A probabilistic
ILDC can be obtained by integrating historical data from different
Chapter 3. Convolution models
34
demand scenarios, with different probabilities of occurring, by
convolution.
If we define the non-served demand remaining in the system after
all the generators have been dispatched (ILDCN) as the difference
between the total system demand (ILDC) and the sum of all the system
generator capacities ( iC ):
1
N
N ii
ILDC ILDC C=
= (Eq. 3.8)
We can obtain the ILDCN of this non-served demand as the
subtraction of each generators capacity distribution functions from the
full system ILDC by convolution as follows:
1
...N NILDC ILDC C C
f f f f = (Eq. 3.9)
After the convolution of all the system generators we will be left
with a series of ILDC curves similar to the ones shown in Fig. 3.2. The
first curve is the ILDC of the full system demand when no generator
has been dispatched yet. The successive ILDCs with decreasing
probabilities are the result of dispatching each generator in the system,
so reducing the system demand to be covered each time. The last curve,
ILDCN represents the demand left uncovered once all the system
generators have been dispatched.
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
35
LOLP
NSE ILDCN
ILDC3
ILDC2
ILDC1
ILDC
Prob
abili
ty
Load [MW]
Fig. 3.2. Inverted load duration curves illustrating the effect of successive
generators being dispatched in the system
From the last ILDC curve obtained (ILDCN) we can extract the two
following valuable pieces of information:
The loss of load probability (LOLP) is the point where this last curve
cuts the y -axis. It represents the probability that there will still be
capacity demand left in the system after all the generators have been
dispatched.
The non-served energy (NSE) is the area beneath this last curve. It
represents the amount of energy in MWh that is left unsupplied in
the system after all the generators have been dispatched.
Convolution therefore enables us to add two probability functions,
in this particular case with probability on the y -axis and unserved load
in MW on the x -axis. In this manner, the coordinates of each point on
Chapter 3. Convolution models
36
the curve represent the probability ( y -coordinate) that a load bigger
than the specified load ( x -coordinate) will go unserved in the system.
(See Fig. 3.3)
[ ][ ]
, 600 0,6
, 600 1 0,6 0,4
P UnservedCapacity x
P UnservedCapacity x
> =
= = (Eq. 3.10)
In addition, the area under the curve represents the energy which
will go unserved for power demands above or below a given point on
the curve. (See Fig. 3.3). This can be justified as we are evaluating
probability in hours.
600
0 600
( , 600) ( )
(0 , 600) ( ) ( )
NSE UnservedCapacity x ILDC x dx
NSE UnservedCapacity x ILDC x dx ILDC x dx
> =
> > =
(Eq. 3.11)
0
0,2
0,4
0,6
0,8
1
1,2
0 200 400 600 800 1000 1200
Load [MW]
Prob
abili
ty
Probability of 0,6 that a load bigger
than 600MW will go unserved
NSE due to unserved loads
greater than 600MW
NSE due to unservedloads smaller than
600MW
0
0,2
0,4
0,6
0,8
1
1,2
0 200 400 600 800 1000 1200
Load [MW]
Prob
abili
ty
Probability of 0,6 that a load bigger
than 600MW will go unserved
NSE due to unserved loads
greater than 600MW
NSE due to unservedloads smaller than
600MW
Fig. 3.3. Reading data on ILDC curves
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
37
3.3 Dispatch of different types of generators
When dealing with different types of generators such as thermal
and hydro units, it is important to take into account that some of them
have power and energy restrictions which condition the way in which
they can be dispatched optimally in power systems. In the following
sections different types of unit dispatch will be analysed.
3.3.1 Thermal units
In the case of thermal units, the input parameters for the model are:
maximum power output and forced outage rate.
Thermal units are dispatched using the following convolution
formula:
( ) ( ) (1 ) ( )n 1 i n i n iILDC x FOR ILDC x FOR ILDC x C+ = + + (Eq. 3.12)
Where: nILDC = inverted load-duration curve
1nILDC + = the new inverted load-duration curve after the
dispatch
iFOR = forced outage rate for generator i
x = the unserved load at each point of the curve
iC = maximum capacity for generator i
The meaning of this equation in words is that the probability of
unserved load in the system after the dispatch of generator i , ( nILDC ),
is equal to the sum of two terms. The first term: ( )i nFOR ILDC x , is the
probability that generator i is unavailable when the system demand
requires x MW of power which generator i could have been able to
supply. The second term: (1 ) ( )i iFOR ILDC x C + , is the probability that
Chapter 3. Convolution models
38
generator i is available when the system demand requires x C+ MW of
power, which generator i is unable to supply as the amount required
exceeds the generators maximum capacity iC . In other words, the
equation adds the two probabilistic scenarios where a given generator
cannot cover the required demand at a given moment.
The following example illustrates this process more visually. Let us
consider a system with one unit, 1G , a scenario in which the unit is
active with a probability 1p of occurring and a scenario in which the
unit is inactive with a probability 1(1 )p of occurring. Fig. 3.4 and Fig.
3.5 show how these scenarios are integrated by convolution to obtain a
new ILDC after the dispatch of 1G .
Fig. 3.4. Scenarios A and B with G1 active and inactive respectively
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
39
Fig. 3.5. Convolution of scenarios A and B to obtain a new ILDC
Another discrete example can be set to illustrate more clearly how
the calculations are carried out. Let us consider the following LDC and
corresponding ILDC:
LDCILDC
MW
MW
h
h
LDCILDC
MW
MW
h
h
Fig. 3.6. LDC Fig. 3.7. ILDC
If we now add a generator to the system with a maximum power
output of 150MW and a FOR of 0.15, we are left with the two following
scenarios: Fig. 3.8 where the generator is active with a probability of
Chapter 3. Convolution models
40
0.85 of occurring and Fig. 3.9 where the generator is inactive with a
probability of 0.15 of occurring.
3760
MW
Expected hours peryear
450
350
760
Group is activeprob=0.85
8760
MW
Expected hours peryear
100
600
500
3760
760
Group is inactiveprob=0.15
Scenario 1 Scenario 2
3760
MW
Expected hours peryear
450
350
760
Group is activeprob=0.85
8760
MW
Expected hours peryear
100
600
500
3760
760
Group is inactiveprob=0.15
8760
MW
Expected hours peryear
100
600
500
3760
760
8760
MW
Expected hours peryear
100
600
500
3760
760
Group is inactiveprob=0.15
Scenario 1 Scenario 2
Fig. 3.8. Active scenario, p = 0.85 Fig. 3.9. Inactive scenario, p = 0.15
Integrating both scenarios by convolution can be reduced in this
discrete example to pondering each scenario by its probability of
occurring and adding them. The result obtained is shown in Fig. 3.10.
8760*0.15+3760*0.85 = 4510
3760*0.15+3760*0.85 = 3760
3760*0.15+760*0.85 = 1210
MW
Expected hours per year
450350 600
760*0.15 = 114
100
8760*0.15+3760*0.85 = 4510
3760*0.15+3760*0.85 = 3760
3760*0.15+760*0.85 = 1210
MW
Expected hours per year
450350 600
760*0.15 = 114
100
Fig. 3.10. Convolution results
The convolution technique then simply consists on repeating this
process for each of the generators in the system one by one. As units are
Analysis of firm capacity calculation methods as mechanisms to ensure long-term reliability in deregulated power markets
41
successively dispatched in the system, obviously the values of LOLP
and NSE decrease, but they never become zero as there is always a
small probability that all the generators in the system could fail at the
same time.
It is important to bear in mind that because thermal plants tend to
have the same characteristic parameters throughout the different
periods to be simulated, the convolutions of their distribution functions
are commutative:
x y y x = (Eq. 3.13)
Dispatch order is therefore irrelevant in the case of thermal units,
but we shall see in the next section how this is not the case for hydro
units.
3.3.2 Hydro units
In the case of hydro units, the input parameters for the model are:
maximum power output, minimum power output (known as run-of-
the-river capacity) and total available energy (a measure of their
reservoir capacity management).
Hydro unit dispatch is not carried out by convolution as hydro
units forced outage rate is considered to be insignificant in comparison
with their energy restrictions, and hence, in order to simplify the model,
no probability-related factor in their defining characteristics. In section
3.4 of this chapter we shall see how this causes an inherent problem in
hydro plant dispatch to arise.
Usually hydro unit dispatch is carried out in two steps:
Chapter 3. Convolution models
42
Firstly the run-of-the-river capacity is dispatched by convolution as
though it where a thermal plant with 0FOR = . This is due to the fact
that it is a constant power output which is always supplied by the
hydro units.
Secondly, the rest of the hydro capacity, which is not constant, is
dispatched by moving the y -axis rightwards step by step
(1MW/step) checking in each step that either the plants energy
restriction or maximum power restriction is not violated. As soon as
either of these limiting factors is reached, the hydro plant is either
dispatched at its maximum power output if its energy restriction
was not violated before reaching it, or at a power value inferior to its
maximum power output if its energy restriction was violated before.
(See Fig. 3.11)
eH
PgE e
HgEm=
= eHg
PmU
11
MW
Prob.
1 MW 1 MW1 MW
P
p
eH
pgE
eH
PgE LOLP
eHgi
-1LOLPi
P
eH
pgE e
HgEm