© 2014 elsevier. - aristotle university of...
TRANSCRIPT
© 2014 Elsevier.
Personal use of this material is permitted. Permission from
Elsevier must be obtained for all other users, including
reprinting/ republishing this material for advertising or
promotional purposes, creating new collective works for resale
or redistribution to servers or lists, or reuse of any copyrighted
components of this work in other works.
Digital Object Identifier: 10.1016/j.enpol.2013.11.065
Electricity Market Models and RES Integration:
The Greek Case
Christos K. Simoglou, Pandelis N. Biskas, Stylianos I. Vagropoulos,
and Anastasios G. Bakirtzis
Power Systems Laboratory, Department of Electrical and Computer Engineering, Aristotle University
of Thessaloniki, GR 54124, Thessaloniki, Greece
([email protected], [email protected], [email protected], [email protected])
Corresponding author: Anastasios G. Bakirtzis ([email protected])
Power Systems Laboratory, Division of Electrical Energy, Department of Electrical and
Computer Engineering, Aristotle University of Thessaloniki, 54124, Thessaloniki, Greece, E-
mail: [email protected], Tel.: +30 2310 996383, Fax: +30 2310 996302
Abstract
This paper presents an extensive analysis of the Greek electricity market for the next seven-
year period (2014-2020) based on an hour-by-hour simulation considering five different RES
technologies, namely wind, PV, small hydro, biomass and CHP with emphasis on PV
integration. The impact of RES penetration on the electricity market operation is evaluated
under two different models regarding the organization of the Greek wholesale day-ahead
electricity market: a mandatory power pool for year 2014 (current market design) and a power
exchange for the period 2015-2020 (Target Model). An integrated software tool is used for
the simulation of the current and the future day-ahead market clearing algorithm of the Greek
wholesale electricity market. Simulation results indicate the impact of the anticipated large-
scale RES integration, in conjunction with each market model, on specific indicators of the
Greek electricity market in the long-term.
Keywords: Day-ahead market, RES integration, wholesale electricity market.
1. Introduction
In the past twenty-five years, wholesale electricity markets have emerged in several
countries to facilitate the competitive operation of the electricity sector (Stoft, 2002). In this
2
context, large volumes of electric energy are usually traded in centrally organized day-ahead
markets, which may take the form of either power exchanges or power pools (Stoft, 2002;
Kardakos et al., 2013).
In parallel with the electricity market restructuring, the increasing environmental concerns
for global warming have promoted the use of Renewable Energy Sources (RES) worldwide.
The European Union (EU) has established the common target "20-20-20" (European
Commission, 2009a) among all Member States aiming to combat climate change and increase
the EU's energy security while strengthening its competitiveness. The share of RES in the
electricity production is expected to increase to 30-35% by 2020 (European Renewable
Energy Council, 2004). This goal motivated the EU countries to provide incentives for the
increase of RES installed capacity (especially wind and PV) across Europe.
Besides the obvious environmental benefits, the increasing share of intermittent RES
electricity production poses new challenges to the efficient management of the power systems
and the operation of the electricity markets, since it can seriously affect the short-term
scheduling of the conventional generating units, the maintenance scheduling of generating
plants and transmission lines, the required levels of system reserves, the wholesale electricity
prices, etc.
In this context, the research community has focused lately on the development of new
methods and tools to tackle these problems and provide effective solutions towards the large-
scale RES integration. An approximation algorithm for the unit commitment problem
incorporating wind generation forecasts in the dispatch decisions for the investigation of the
wind power effect on CO2, SO2 and NOx emissions is presented in Denny and O’ Malley
(2006). Li and Kury (2009) presented a dynamic programming approach for the solution of
the weekly unit commitment problem at hourly resolution, in order to evaluate the impact of
wind generation on the generation schedule, particularly the overall fuel cost, emissions and
system security. Ummels et al. (2007) proposed an adapted unit commitment and economic
3
dispatch tool for system operations in systems with large-scale wind power penetration,
taking into account cost, reliability and environmental factors. A stochastic unit commitment
problem is solved in Tuohy et al. (2009), where a mixed-integer stochastic optimization
model (developed in WILMAR project (WILMAR, 2006)) is used on a "rolling planning"
basis for generation rescheduling, as more precise wind and load forecasts are made available.
In a more general framework, emissions are incorporated in the dispatch scheduling using
multi-objective optimization (Muslu, 2004; Catalão et al., 2008), where different values are
given to a weighting factor in the objective function in order to enforce the trade-off between
fuel cost and emissions minimization. Several simulation models have also been developed
for the qualitative and quantitative assessment of the effect of increased RES penetration on
market prices (Green and Vasilakos, 2010; Morales et al., 2011; Sensfuß et al., 2008;
Jacobsen and Zvingilaite, 2010), system operation costs and operation patterns of generating
units (Jacobsen and Zvingilaite, 2010; Delarue et al., 2009; Maddaloni et al., 2009; Troy et
al., 2010).
In all these works, a single market model (i.e. power pool or power exchange) as well as a
single renewable technology (i.e. wind resources) is considered. In addition, all studies refer
to a short- or mid-term scheduling period (one-day to one-year ahead).
In this paper, an extensive simulation analysis of the Greek electricity market for the next
seven-year period (2014-2020) on an hour-by-hour basis considering five different RES
technologies, namely wind, PV, small hydro, biomass and CHP, is presented. The impact of
RES penetration on the electricity market operation is evaluated under two different models,
regarding the organization of the Greek wholesale day-ahead electricity market: a mandatory
power pool for year 2014 (current market design) and a power exchange for the period 2015-
2020 (harmonization with the European “Target Model”). A detailed description of the
characteristics of each market model is presented in the following section. An integrated
software tool is used for the simulation of the current and the future day-ahead market
4
clearing algorithm of the Greek wholesale electricity market.
In each case, the impact of the RES integration on the Greek electricity market is analyzed
in terms of:
a) the wholesale electricity market prices (System Marginal Prices, SMPs),
b) the total CO2 emissions and generation cost,
c) the number of start-ups and shut-downs ("cycling") of the conventional units,
d) the profitability of all generating units (conventional and RES plants) and the total
payment of the consumers for the energy withdrawal, and
e) the RES uplift charge, imposed by the Greek State for the payment of the (usually
higher than the SMP) feed-in tariffs to the RES producers.
An initial study on the subject was carried out in Simoglou et al. (2011). In that work, a
yearly simulation analysis of the Greek electricity market under a single market model (i.e.
the mandatory power pool) and ten different scenarios regarding the RES installed capacity
was performed. In the present study, a comparison of the impact that each market model (i.e.
power pool vs power exchange) in conjunction with the anticipated large-scale RES
integration has on specific indicators of the Greek electricity market in the long-term is
presented. This work also provides insight on the specific measures enacted lately by the
Greek State in order to alleviate the additional economic burden undertaken by the electricity
consumers (through the imposed RES uplift charge in their electricity bills), so that RES
producers are fully compensated according to the agreed feed-in-tariffs.
A concise description of the two electricity market models adopted in this study is
presented in Section 2. Section 3 describes the software tool as well as the mathematical
optimization models used for the multi-year simulation analysis. Section 4 provides relevant
test results, while valuable conclusions are drawn in Section 5.
5
2. Wholesale Electricity Market Models
The Greek wholesale day-ahead electricity market is currently organized as a centralized
mandatory pool, where each generating unit submits a complex offer comprising an energy
offer, a reserve offer and a declaration with its techno-economic data, including start-up and
shut-down procedures and costs. The Market Operator (MO) solves a short-term unit
commitment problem on a daily basis (also known as “Day-Ahead Scheduling” or DAS),
where a simultaneous 24-hour co-optimization of energy and reserve resources is performed
under a large set of unit and system constraints (e.g., unit start-up and shut-down procedures,
minimum-up/down time constraints, min/max power output restrictions, ramp-rate limits,
system reserve requirements, transmission limits, etc.). The DAS objective is the
maximization of the social welfare (or equivalently the minimization of the total production
cost minus the load utility) within the 24-hour period of the next day. In fact, DAS is a non-
convex optimization problem that is formulated and solved as a Mixed-Integer Linear
Program (MILP) and yields the SMP and the units' energy production and reserves
contribution schedule for each hour of the following day. In this model, additional payments
(usually called “make-whole side payments”) are also provided for the full recovery of the
units’ total operating (variable) cost, in case this is not covered through the day-ahead and
imbalances settlement (cost recovery mechanism) (Stoft, 2002).
However, by 2015 the Greek wholesale electricity market is expected to be transformed to
a decentralized market, based on bilateral trading and the operation of a voluntary day-ahead
Power Exchange, in order to become compliant with the European Target Model (European
Commission, 2009b). Following the current trend in most European countries, in the future
Greek power exchange the participants (producers/suppliers) will submit simple energy
quantity (MWh)-price (€/MWh) offers/bids for the energy they wish to sell/buy at every hour
of the next day. The Market Operator (MO) will create the aggregated supply and demand
curves and clear the day-ahead market on an hour-by-hour basis by solving a convex Linear
6
Programming (LP) problem, without considering any unit operating constraints or system
operation constraints.
In the power exchange case, the solution of the aforementioned day-ahead market clearing
algorithm will provide only the SMP for each hour of the next day. In order to yield feasible
commitment and dispatch schedules for all generating units, the MO (or the SO) will solve
again the day-ahead scheduling taking into account all system and unit operating constraints
already described, in the framework of a “day-ahead balancing market”, in compliance with
the current market structure of the most European countries and the future European Target
model. Therefore, the solution of the day-ahead balancing market as an MILP model yields
the units’ energy production as well as all other operating results (e.g. CO2 emissions, unit
operating costs, number of unit start-ups, etc.). The involvement of each producer/consumer
in the feasible schedule derived by the solution of the day-ahead balancing market is, in
general, different as compared to the respective initial positions assigned according to the
clearing of the power exchange. There are various mechanisms applied in the modern
electricity markets to clear these differences, usually under a Pay-As-Bid (PAB) scheme.
However, the financial settlement of the wholesale market participants (conventional and RES
units) is outside the scope of this paper.
Finally, it is noted that, in this paper, the term “Day-Ahead Scheduling (DAS)” will be
used interchangeably for the day-ahead market clearing or the day-ahead balancing market
clearing of both wholesale market models.
3. Methodology
3.1 Simulation Tool
The simulation tool used for the analysis already described is an integrated software called
“Long-Term Scheduling” (LTS). LTS was developed by the Power Systems Lab, Aristotle
University of Thessaloniki, Greece, for the solution of the mid-/long-term scheduling problem
of the Greek wholesale electricity market. The mid-/long-term scheduling problem can be
7
formulated and solved either as an LP problem (for the daily simulation of the wholesale
market operation as a power exchange) or an MILP problem, for the daily simulation of: a)
the wholesale market operation as a mandatory power pool (e.g. year 2014) and b) the day-
ahead balancing market of the power exchange (e.g. period 2015-2020), in accordance with
the market framework studied. Both models are analyzed in Section 3.4.
LTS can simulate the operation of the system for many years on a year-by-year basis. In
this context, LTS can be appropriately parameterized and used effectively for the mid-/long-
term simulation of different wholesale electricity markets operating in other countries, in case
they follow, in general, the structure of the aforementioned market models. It is clarified that
LTS does not decide on the commitment of new generation resources. New unit commitments
and old unit withdrawals are inserted as input data in LTS.
The optimization component used for the solution of the mid-/long-term scheduling
problems is the Generic Algebraic Modeling System (GAMS) along with the CPLEX solver
(Brooke et al., 1990).
3.2 Model Assumptions
There are two main approaches to simulate the optimal yearly operation of a power system
(or, equivalently, a wholesale electricity market) comprising thermal, hydro and RES units on
a cost minimization (or, equivalently, social welfare maximization) basis, as follows:
The first approach is to formulate and solve a unique MILP problem for the entire year,
modeling both medium-term aspects, such as the hydro resources management, and short-
term decisions, such as the generating units’ commitment and dispatch, in a unified
framework. This one-shot solution approach is advantageous in terms of optimality but the
computational burden of the respective MILP model is notably high, especially when a
detailed representation of the generating unit operating constraints (i.e. start-up and shut-
down procedures, minimum-up/down time constraints, power output constraints, etc.) is
adopted in the problem formulation (Baslis et al., 2009).
8
The second approach is to decompose the optimal yearly operation problem. In this case, a
Peak-Shaving (PS) problem is first solved to simulate the twelve-month usage of hydro
resources, followed by a day-by-day successive solution of the day-ahead market clearing
problem (DAS problem) for the entire year. The mandatory hourly hydro injections computed
by the solution of the PS problem are introduced as input parameters in the DAS problem of
each day of the year. This is the approach followed in the simulation analysis of this paper
and, although less computationally intensive, it naturally leads to suboptimal solutions (e.g.
higher total system cost) as compared to the first approach. It should be noted that the sub-
optimality created by the choice of the second approach is acceptable for a long-term analysis.
This is due to the fact that the main difference in the constraints included in the PS and DAS
models lies in that the thermal units are included in PS without binary variables, so their
technical minimums are ignored in the PS (for computational performance reasons).
However, this “omission” leads to negligible errors in the unit scheduling and dispatch in the
PS, since for the hours of hydro peak-shaving (peak-load hours) most available thermal units
are usually dispatched at their maximum output. Therefore, the sub-optimality introduced in
the distribution of the hourly mandatory injections (PS results, which are then transferred to
DAS as input data) is negligible for a long-term analysis.
Another significant reason we selected the second approach in our simulation is in order to
more accurately reproduce the reality of the Greek wholesale market: According to the
provisions of the Greek Power Exchange Code (PEC), the hydro producers submit yearly,
monthly and weekly hydro release schedules first, and they are obliged to follow them in the
day-ahead market, unless otherwise justified.
An additional modeling aspect regards the handling of RES production. The uncertain
character of wind and solar energy production may affect the behavior of the power system in
terms of the scheduling of the conventional units, the system production cost, the emissions,
the need for reserve capacity, etc. (Doherty and O’ Malley, 2005; Morales et al. 2009;
9
Meibom et al. 2009). However, the LTS software does not currently provide the opportunity
to account for the uncertainty of RES production e.g. through a Monte Carlo simulation or in
a probabilistic or stochastic programming framework. In other words, a perfect knowledge of
the future is assumed. The extension of LTS software to incorporate, besides the thermal unit
availability, additional stochastic parameters related to the RES production is the goal of our
future work.
With regard to the transmission network modeling, recent studies dealing with the large-
scale RES integration in power systems have shown that the transmission network modeling
may have a critical impact on the simulation results (Morales et al., 2009; Fink et al., 2009).
However, in the solution algorithm of the Greek day-ahead wholesale electricity market, the
Greek power system is divided into two zones (North/South) and only a single transmission
corridor flow constraint that limits the active power flow between the system zones has been
provisioned for each dispatch period. However, according to the enacted regulatory
framework of the Greek wholesale market, this constraint is inactive, and, in fact, not
included in the market clearing problem formulation. Therefore, since the aim of this paper is
to accurately represent the regulatory framework of the Greek electricity market as well as to
provide specific quantitative results based on the realistic day-ahead market clearing
formulation, in the simulation runs performed with the LTS software, the transmission
network modeling is neglected.
As already mentioned, for the accurate simulation of the wholesale electricity market, LTS
formulates and solves consecutively two distinct scheduling sub-problems, namely the “Peak-
Shaving” (PS) problem (Simopoulos et al., 2007) and the “Mid-Term Scheduling” (MTS)
problem, further described in the following subsections.
10
3.3 Peak-Shaving (PS) Problem
The energy-limited nature of storage hydroelectric power plants introduces time-coupling
in the hydro production decisions within the (usually annual) scheduling horizon. The aim of
the PS problem is the computation of the mandatory hourly hydro injections that are
subsequently used as non-priced energy offers in the day-by-day successive solution of the
day-ahead market clearing model that follows (MTS problem). In fact, the PS problem
simulates the twelve-month usage of hydro resources that is performed by the hydro
producers (or the System Operator, SO) annually.
The desired hydro production is inserted as input (in MWh) and is included in the problem
mathematical formulation as either a single equality constraint covering the entire year or
separated in twelve monthly hydro injection equality constraints to enforce monthly water
release intervals.
3.4 Mid-Term Scheduling (MTS) Problem
Once the PS problem is solved for each year of the scheduling horizon, LTS proceeds with
the formulation and solution of the MTS model. MTS model consists on the successive
formulation and solution of the day-ahead market clearing algorithm of the Greek wholesale
electricity market (DAS algorithm) operating either as a power pool or a power exchange, on
a daily basis for each entire year; the DAS solution for day N-1 provides the initial conditions
of the DAS problem for day N.
The basic assumptions of the MTS problem are presented in the following subsections.
3.4.1 Energy and reserve offers of units
In the MTS model formulation, the priced energy offer consists of a monotonically
increasing, step-wise function of up to ten (10) quantity (MWh) - price (€/MWh) pairs (steps).
In general, in day-ahead markets, the generating companies are allowed to submit energy
offers whose offer prices do not necessarily reflect the units’ marginal costs. In the Greek
11
wholesale electricity market, the dominant power company (ex-monopolist) owns the
majority of the conventional thermal generating units and the entire hydroelectric production.
Moreover, it controls almost the entire retail sector (around 99%). Overall, it is a net buyer in
the wholesale market in order to fulfill its retail supply obligations. Therefore, it has an
incentive to keep the wholesale market clearing prices as low as possible by offering its
generating capacity in the wholesale market at marginal cost or even at lower prices. To avoid
strategic behavior (dumping) that would result in barrier to entry, the Greek Regulator
currently imposes a floor to the units’ energy offer prices, according to the provisions of the
Power Exchange Code (PEC) (Greek RAE, 2013). This floor price is the Minimum Variable
Cost (MVC) of the unit and is crucial in the offering strategy of the dominant power
company. Therefore, in both market models, all generating units belonging to the dominant
power company are considered to submit offers at their MVC, fully incorporating the total
emissions cost in their offer price.
On the contrary, in the case of the mandatory power pool model, all private producers
owning CCGT units that are more costly than the dominant company’s units (mainly lignite
units) take advantage of specific enacted market rules (Greek RAE, 2013) allowing them to
implement a different bidding strategy. According to these rules, the first step corresponding
up to the 30% of the unit available capacity can be fixed well below the unit's marginal cost
(i.e. starting from a value 0.001 €/MWh), whereas the remaining unit available capacity
should be assigned appropriate offer prices so that the weighted average energy offer price
remains above the MVC. The private producers are likely to follow that strategy, since it
allows them to secure their dispatch in the day-ahead market and consequently in real-time.
However, in the case of the power exchange model, the above rules will no longer be in force,
and, therefore, for the period 2015-2020 all private producers are also considered to submit
offers at their MVC.
12
In the case of power pool, generating units are considered to submit reserve offers also.
Specifically, they submit one-step priced reserve offers for primary reserve and secondary
range, in cohesion with the provisions of the PEC. The quantity involved in these offers is the
maximum capability of each unit to provide each type of reserve. There are no priced reserve
offers for tertiary reserve (and consequently no remuneration for its provision). Tertiary
reserve requirements are defined per dispatch period and are inserted as constraints in the
MTS model.
3.4.2 RES injection data
The RES injection data considered in this study are derived on the basis of the real or
estimated hourly RES injections of the period 2009-2012 for all RES technologies, as follows:
Wind real hourly production data are made publicly available by the Greek Independent
Power Transmission Operator (Greek IPTO, 2013).
For the PV, small hydro, biomass and decentralized CHP plants, since no real hourly
injection data but only total monthly statistics (i.e. total installed capacity (MW) and total
injected energy (MWh)) are available in Greek IPTO (2013), different assumptions are made
for the calculation of their respective hourly energy injections, as follows:
Small hydro, biomass and decentralized CHP plants are considered to have a flat hourly
production profile within each month, since their production is mainly subject to seasonal
rather than daily variations. For these technologies, a mean hourly energy injection is
considered for each month, calculated as the ratio of the total monthly energy injection to
the total number of hours of the month.
For the calculation of the hourly energy injection of the ground-mounted PV plants,
typical monthly injection profiles are used (Greek IPTO, 2009).
In all cases above, the past real or estimated hourly energy injection profiles of the period
2009-2012 are calculated in per unit values (i.e. injected MWh/installed MW). In this sense,
Fig. 1 illustrates the mean hourly energy injection per month for each RES technology except
13
for PV for the period 2009-2012, while Fig. 2 presents the typical monthly injection profiles
of the PV plants used in the study. This set of per unit values (injection profiles) is repeated to
cover the period 2014-2020 in order to capture the usual seasonal variation of RES
production. The forecasted hourly RES injections (in MWh) for the entire study period are
easily computed by multiplying the respective per unit values with the target installed
capacity per RES technology. In this context, Fig. 3 summarizes the estimated evolution of
the installed capacity of all RES technologies for the study period with the presented values
corresponding to the target capacity at the end of each year.
<INSERT FIGURE 1 HERE>
<INSERT FIGURE 2 HERE>
<INSERT FIGURE 3 HERE>
3.4.3 Monte Carlo simulation
Thermal unit availability is the only stochastic parameter modeled in the MTS using Monte
Carlo simulation. Random sequences of availability / non-availability states for each unit are
created using the two-state Markov model. Each thermal generating unit is characterized by a
failure rate λ (times per year), according to which it leaves the state "Up" (Available) and
enters the state "Down" (Unavailable). Similarly, it leaves the state “Down” with a repair rate
μ (times per year), according to which it enters again the state "Up". The unit failure and
repair rates result through the assessment of the unit availability during a long study period in
the past and, in general, increase with the age of the units. However, in this study, the
simplifying assumption of considering both above reliability metrics constant throughout the
seven-year study period was made. In this context, next-failure and next-repair times are
random variables assumed to follow an exponential distribution and their corresponding
random values are sampled using the inverse transform sampling method (Devroye, 1986).
Each unit availability scenario comprises a list of values representing the hourly
availability/non-availability (in MW) of each thermal unit for all years of the study period
14
taking also into account the units' maintenance periods. For each unit availability scenario,
LTS software formulates and solves the PS problem for each year, followed by the solution of
the MTS problem. In this sense, the simulation procedure is highly dependent on the number
of the selected unit availability scenarios. In the present study, 100 unit availability scenarios
were simulated.
For the reader’s convenience, a flowchart explaining the entire methodology followed step-
by-step in the LTS simulation runs is illustrated in Fig. 4. In this flowchart, N denotes the
number of unit availability scenarios, YN denotes the number of years under study, and yDN
denotes the number of days of each year under study.
<INSERT FIGURE 4 HERE>
3.4.4 Mathematical formulation
In this section, the mathematical formulation of both day-ahead market clearing algorithms
is presented. At first, the DAS formulation of the power pool is described, while the
description of the simplified formulation of the power exchange follows.
The goal of the MTS model solution is the maximization of the total social welfare (or
equivalently the minimization of the total production cost minus the load utility). The load
utility refers to the priced-load declarations of demand (if any) and exports.
3.4.4.1 Power Pool
In the case of the power pool, the objective function of the day-ahead market clearing
algorithm formulated and solved successively on a daily basis in the framework of the MTS
model is described as follows:
15
1 1 2 2 2+π πi i Rft ft it it it it it i it i it
it i f
imp imp exp exp d dft ftft ft ft ft
imp exp dimp exp df f f
Min C b r r r SUC y SDC z
C b C b C b
T I F
Imp Exp DF F F
(1)
where eftC is the price of step f of the priced energy offer of market entity e (i.e. e=i:
generating unit, e=imp: imports) or the priced load bid of market entity e (i.e. e=exp: exports,
and e=d: loads) during hour t, in €/MWh, eftb is the cleared quantity of step f of the priced
energy offer or the priced load bid of the respective market entity e during hour t, in MWh,
1πit and 2π Rit is the price of the primary reserve offer and the secondary range offer,
respectively, of unit i during hour t, in €/MW, mitr is the contribution of generating unit i in
reserve type m (m=1: primary reserve , m=2+: secondary-up reserve, m=2-: secondary-down
reserve) during hour t, in MW, /i iSUC SDC is the start-up/shut-down cost of unit i, and
/it ity z is the binary variable that is equal to 1 if unit i is started-up/shut-down during hour t,
respectively.
The objective function to be minimized (1) includes: a) the energy production cost i ift ftC b
, as reflected on the respective units' energy functions, b) the primary reserve provision cost,
1 1π it itr , and the secondary reserve provision cost, 2 2 2π Rit it itr r , as reflected on the
respective units' reserves offer functions, c) the units' start-up cost, i itSUC y , and shut-down
cost, i itSDC z , d) the cost functions of the import agents imp impft ftC b , based on the respective
offers, e) the load utility functions d dft ftC b and the export utility function
exp expft ftC b , based on
the load and export bids submitted to the MO, respectively. The RES energy injection term is
not present in (1), since it is considered as non-priced energy injection, according to the PEC.
The aim of the present model is the minimization of the value of the objective function over
16
all hours of the scheduling horizon (i.e. 24 hours of the next day) subject to a large set of
problem constraints, which comprise various system and unit operating constraints.
In this model, the system constraints (2)-(6) are valid for each hour of the dispatch day, as
follows:
imp expr dit t ftft ft
imp exp di imp r exp df f f
p b E b b t
I Imp R Exp DF F F
T (2)
where itp is the energy injection of unit i during hour t, in MWh, rtE is the energy injection
of Renewable Energy Source (RES) r during hour t, in MWh, and eftb is the cleared quantity
of step f of the priced energy offer or the priced load bid of the respective market entity e
during hour t, in MWh.
1 1it t
i
r RR t
I
T (3)
2 2it t
i
r RR t
I
T (4)
2 2it t
i
r RR t
I
T (5)
3 3 3S NSit it t
i i
r r RR t
I I
T (6)
where mitr is the contribution of unit i in reserve type m (i.e. m=1: primary, m=2+: secondary-
up, m=2-: secondary-down, m=3: tertiary (spinning - 3S and non-spinning - 3NS) during hour
t, in MW, and mtRR is the corresponding system requirement in reserve type m during hour t,
in MW.
Constraints (2) enforce the power balance equation for each dispatch period t. This means
that in each dispatch period, the energy injection in the power system (left-hand side of the
equation) must be equal to the energy withdrawal from the system (right-hand side of the
equation). Constraints (3)-(6) enforce the deterministic system requirements for the provision
of the primary, secondary-up, secondary-down and tertiary (spinning and non-spinning)
17
reserves, respectively. In all cases, the sum of all conventional units’ contribution to a single
reserve type must be at least equal to the respective system reserve requirement.
The unit operating constraints comprise (7) and (8), further explained in the following:
,fix iit ftit
if
p P b i t
F
I T (7)
0 , ,i i ift ftb B i f t I F T (8)
Equation (7) enforces that, for each generating unit i, the power injection itp can be
divided into two components:
The first term,fix
itP , represents the non-priced component of the energy offer function of the
unit, including the mandatory hydro energy injection and the energy production of units in
commissioning tests. This component may follow a constant and pre-specified schedule
during the dispatch day.
The second term represents the priced component of the energy offer function of the unit.
This component is equal to the sum of the cleared blocks of energy iftb .
Constraints (8) denote that the portion of step b of the i-th generating unit energy offer
function that is cleared in hour t, iftb , cannot exceed the size of the corresponding step, i
ftB .
Apart from constraints (7)-(8), there also exist various unit operating constraints (i.e. simple
and inter-temporal constraints, which model the power output constraints, the minimum
up/down constraints, the ramp-rate limits, initial conditions, etc). All these constraints are
analytically described in Simoglou et al. (2011).
3.4.4.2 Power Exchange
In the case of the power exchange, the objective function of the initial MTS problem (pure
economic solution) is a simplification of the respective function of the power pool, as follows:
18
imp impi ift ft ft ft
i impt i impf f
exp exp d dft ftft ft
exp dexp df f
Min C b C b
C b C b
T I ImpF F
Exp DF F
(9)
The objective function in (9) is differentiated from (1) only in that the cost components
associated with a) the primary reserve provision cost, 1 1π it itr , b) the secondary reserve
provision cost, 2 2 2π Rit it itr r , and c) the units' start-up cost, i itSUC y , and shut-down
cost, i itSDC z , included in (1) are neglected.
In this case, the optimization model is subject only to constraints (2), (7) and (8), since all
unit operating constraints and the system reserve requirements described above are ignored.
The formulation of the respective day-ahead balancing market problem subsequently
solved as an MILP model is identical with the formulation of the power pool model described
above.
4. Test Results
4.1 Input Data
In the present case study, the long-term scheduling problem is solved for the Greek Power
System, which at the end of 2013 (Dec. 2013) will comprise 32 thermal units and 14
hydroplants, with a total installed thermal and hydro capacity of 9,170 MW and 3,200 MW,
respectively. Numerous RES plants (mainly wind plants and dispersed PV installations) that
are going to be connected in the Greek power system are expected to raise the total RES
installed capacity to 4,498 MW by the end of 2013 (see Table 1). In this study the category
“RES plants” includes also both the existing CHP plants and those to be installed in the
following years.
<INSERT TABLE 1 HERE>
19
For the study period 2014-2020, the system load demand is determined on an hourly basis
through a long-term load forecast method. For this purpose, the hourly system load of the year
2013 along with an annual increase/decrease rate estimated on the basis of the mid-/long-term
perspectives of the Greek economy for the period 2014-2020 is considered. As regards the
hourly values of the system load for 2013, the real values already made publicly available
were used for the first semester (Greek IPTO, 2013), while for the second semester forecasted
values obtained on the basis of the respective values of 2012 were used. The total electricity
consumption (in TWh) and the peak load demand (in MW) for the seven-year period are
presented in Fig. 5. Figure 6 illustrates the projected natural gas cost and the estimated cost of
the CO2 emission rights for the examined period.
<INSERT FIGURE 5 HERE>
<INSERT FIGURE 6 HERE>
In this study, three different scenarios regarding the simulation of the Greek wholesale
electricity market were considered, which are differentiated only in terms of the evolution of
the installed capacity of the ground-mounted PV systems in the period 2015-2020; the total
PV capacity at the end of 2020 is considered equal to 3,770 MW (S.1) or 4,520 MW (S.2) or
5,270 MW (S.3), respectively, of which 675 MW account for the rooftop PV systems in all
scenarios.
<INSERT TABLE 2 HERE>
The new thermal units that will start their commercial operation in the period 2014-2020
have already been defined. Similarly, the old thermal (mainly lignite) units to be withdrawn
from the Greek power system within the study period have already been announced. Thus, for
the purposes of this study, the units' construction and withdrawal plan is considered
deterministically known and is summarized in Table 2.
4.2 Simulation Results
4.2.1 Wholesale market operation
20
In this section, simulation results regarding specific indices of the wholesale electricity
market operation for the three scenarios studied over the seven-year study period are
presented.
<INSERT FIGURE 7 HERE>
Figure 7 shows the average yearly system marginal price (SMP). It is clear that the model
transition from the pool to the power exchange in 2015 will lead to a considerable increase in
the resulting SMP despite the continuous increase of the expected RES injection (see Fig. 3).
This is due to the fact that in the current mandatory pool the privately-owned CCGT units
take advantage of the enacted market rules (already discussed in Section 3.4.1) and secure
their commitment and dispatch all day long in the day-ahead market, usually operating at their
minimum power output due to their high variable operating cost. As a result, the market
clearing algorithm forces the low-cost lignite units to act as marginal units very often during
the year, thus resulting in considerably lower SMP.
On the contrary, in the case of the power exchange, those rules will no longer be in force
and the commitment and dispatch of all conventional units will follow a pattern which is
closer to the expected "merit-order". In this case, the CCGT units are expected to act as
marginal units more often and, as a result, the evolution of the SMP will follow the respective
trend of the forecasted natural gas cost (see also Fig. 6). Finally, it is shown that the increase
of the total PV installed capacity among the three scenarios results in a slight decrease of the
SMP for the period 2015-2020. This is due to the fact that the energy production from
renewable sources is considered as non-priced injection and, therefore, substitutes the energy
injection of the medium-cost CCGT units, which will usually define the SMP in the power
exchange. In quantitative terms among all scenarios, the reduction in the SMP is estimated, on
average, at 1.9 €/MWh per GW of PV installed capacity or alternatively 1.4 €/MWh per TWh
produced by PV systems.
21
In this context, Fig. 8 illustrates the total number of start-ups of the lignite and CCGT units,
while Fig. 9 shows the total operation hours per thermal unit technology. The number of start-
ups includes: a) the units start-ups following a forced outage or scheduled maintenance and b)
the start-ups that occur as a result of the day-ahead market clearing algorithm solution.
<INSERT FIGURE 8 HERE>
<INSERT FIGURE 9 HERE>
Since the number of start-ups belonging in the first category is rather constant per year, it is
clear that the transition from the mandatory pool to the power exchange model will lead to a
significant increase of the start-ups of the CCGT units (see Fig. 8), while at the same time the
respective total operation hours decrease (see Fig. 9), due to the wholesale market
participation strategy followed by these units. In addition, it is apparent that the withdrawal of
six old lignite units during 2019 in conjunction with the large increase of the PV capacity
result in an increasing number of unit start-ups from S.1 to S.3 during the years 2019-2020,
while the respective total operation hours decrease. This is due to the fact that the CCGT units
are forced to cycle more often and remain on-line for restricted number of hours each time in
order to compensate for the large intermittent energy injection coming from renewable
sources.
<INSERT FIGURE 10 HERE>
The evolution of the annual thermal production per generation technology and the total CO2
emissions for the period 2014-2020 are shown in Fig. 10. The abolition of the aforementioned
favorable market rules for CCGT units from 2015 onwards in combination with the increased
RES installed capacity leads to a significant decrease of their total energy production as
compared to 2014. In contrast, the lignite units' energy production remains rather constant
among all scenarios up to 2019, since their dispatch is not seriously affected by the increased
RES injection, given that the latter substitutes mainly the energy production from the
intermediate-load CCGT units. However, the notable change in the thermal generation mix in
2015 due to the market model transition leads to a respective slight increase in the CO2
22
emissions, whereas in the following years the total CO2 emissions follow a continuous
reduction up to 2020, given that the conventional generation is mainly substituted by the
increased RES production. During 2019-2020, the withdrawal of six pollutant lignite units
results in a steep reduction in the total CO2 emissions, since the CCGT units replacing in part
the lignite production are characterized by lower emissions rates than the base-load lignite
units (see also Table 1).
In quantitative terms, the avoided CO2 emissions owing to the increase of PV production
among the three scenarios are, on average, equal to 550 ton CO2 per MW of PV installed
capacity per year or alternatively 400 kgr CO2 per MWh of energy produced by PV systems.
This follows because a single MW of PV installed capacity in Greece produces on average
approx. 1,360 MWh per year, while each MWh produced by PV systems substitutes a MWh
that would be alternatively produced mainly by CCGT units, which are characterized by a
mean emission rate of 0.4 T/MWh (see also Table 1).
<INSERT FIGURE 11 HERE>
Similar conclusions are drawn regarding the thermal production cost. Figure 11 illustrates
the apparent decrease observed in the yearly thermal production cost from 2014 to 2015
(around 11%) due to the market model transition, since the energy production of the medium-
cost CCGT units is mainly substituted by the low-cost lignite units. From 2015 onwards, the
continuous reduction in the thermal production cost is mainly due to the increased RES
injection, except for 2020, where the change in the generation mix owing to massive lignite
unit withdrawal (see also Fig. 10) causes a slight increase in total thermal production cost
with respect to 2019. In this context, the reduction of the thermal production cost among the
three PV penetration scenarios is, on average, equal to 58.5 € per MWh produced by PV
systems.
To conclude, it should be noted that the substantial changes to appear in the wholesale
market indicators, such as the clearing prices, the number of units start-ups, the CO2
23
emissions, etc. between 2014 and 2015, as described above, are owing to both the abolition of
the favorable market rules for CCGT units and the transition from the power pool to the
power exchange itself.
4.2.2 Retail market operation
In this section the effects of large RES penetration along with the anticipated regulatory
framework changes on the operation of the retail electricity market and the total payments of
the electricity consumers over the seven-year study period are investigated.
In Greece, the energy injection from RES plants is compensated under a fixed feed-in tariff
(FiT) (Couture and Cagnon, 2010; Dusonchet and Telaretti, 2010), which is among the most
common RES support schemes across Europe.
<INSERT TABLE 3 HERE>
<INSERT FIGURE 12 HERE>
Table 3 shows the expected share (in terms of installed capacity) and the average FiTs for
the five RES technologies considered in this study at the end of 2013.
According to the current Greek regulatory framework, the average FiTs for all RES
technologies except for PV systems will remain constant throughout the seven-year study
period and equal to those shown in Table 3. However, due to the very high PV penetration
already achieved, PV tariffs are going to decrease significantly for both the new ground-
mounted and rooftop PV systems as of 2014. In this context, the FiTs of the rooftop PV
systems are going to follow a predetermined yearly decrease from 2014 to 2020, which has
already been announced and shown in Fig. 12.
Regarding the FiTs of the new ground-mounted PV systems, two different pricing schemes
are examined in this study, as follows:
Wholesale market SMP-based pricing scheme (Scheme A): According to this scheme,
which is currently provisioned in the Greek energy market legislation, from 2015 onwards
24
the FiTs of the ground-mounted PV parks for year y will be computed on the basis of the
average SMP of the wholesale electricity market of the previous year, y-1, as follows:
11.15 , 2015,...,2020PVy yFIT SMP y
(10)
"Market-corridor" pricing scheme (Scheme B): This is a volume-responsive degression
pricing scheme introduced in 2009 in Germany, under which FiT rates decline based on the
amount of capacity installed during prior periods (Dusonchet and Telaretti, 2010; Fulton
and Capalino, 2011). This allows FiT to be adjusted to maintain growth within predefined
limits. In this study, it is considered that the FiT implemented in year y is defined indirectly
by the growth of the PV market (new PV installations) during year y-1, as follows:
1 1
1 1
, 250 W
0.95 , 250 W
PV PVy y y
PV PVy y y
FIT FIT C
FIT FIT C
(11)
where 2015,...,2020y and PVyC is the total installed capacity of the new ground-
mounted PV installations during the year y.
Figure 12 shows the future trajectory of the FiTs for the new rooftop PV systems and the
resulting FiTs for both pricing schemes and the three scenarios studied regarding the new
ground-mounted PV installations. These FiTs were derived on the basis of: a) the simulation
results of the wholesale market regarding the expected SMP for the period 2014-2020 (Scen.
A.1-A.3, see also Fig. 7) and b) the estimated evolution of the PV installed capacity (Scen.
B.1-B.3, see also Fig. 3).
<INSERT FIGURE 13 HERE>
Since all transactions between the MO and all energy producers and suppliers in the
framework of the wholesale day-ahead market are settled with the SMP (except for the RES
producers that are compensated under the FiT), there is a deficit on the payments balance of
the MO. Therefore, the MO maintains a special account known as “RES & Cogeneration
Special Account” (henceforth, “Special Account”), in which different revenue sources
25
contribute so that the RES producers are fully compensated (see Fig. 13).
Specifically, the main components of this account currently comprise the following:
Income from the day-ahead market, calculated as the product of the hourly total RES
injection times the hourly SMP.
Income from auctions for available CO2 emission rights. This income is computed as the
product of the total yearly CO2 emissions times the average CO2 emissions rights cost (see
Fig. 6). For the period 2014-2015, this income will be fully utilized (100%) as additional
resource of the Special Account, while for the period 2016-2020 only 50% of this income
will be credited to the Special Account.
Income from special levy on lignite production, which is set equal to 2 € per MWh
produced by lignite units.
Income from special levy on public television revenues, which is estimated at approx. 75
M€/year.
Income from special levy on RES producers. This income is calculated on the basis of a flat
rate imposed on the gross annual revenues of all RES producers. This levy will expire in
June 2015 and is currently estimated at approx. 370 M€/year.
Income from direct tax on the electricity consumers’ electricity bill. This income is
calculated as the product of the total energy withdrawal of the consumers times a special
regulated charge, also known as “RES uplift charge”. Mathematically, the average yearly
RES uplift charge [in €/MWh] is defined as follows:
yCharge
y
RevRES , 2014,..., 2020
Consumers
zty zt
yD
ZT
(12)
The sole expenses of the Special Account are the payments to all RES and CHP producers
on the basis of their agreed FiTs.
All above components except for the income from the direct tax on the electricity
26
consumers are calculated based on the simulation results of the day-ahead market. The deficit
of the Special Account to be covered directly by the electricity consumers through the
imposed RES uplift charge is defined as the difference of the total expenses minus the
summation of the revenues of the additional available resources. The relevant calculations for
scenarios A.1 - A.3 are summarized in Table 4.
<INSERT TABLE 4 HERE>
<INSERT FIGURE 14 HERE>
Figure 14 illustrates the required RES uplift charge to be imposed directly on the electricity
consumers for both pricing schemes and the three scenarios studied, so that the Special
Account of the MO remains balanced during the seven-year study period. Regarding the
current regulatory framework (Scenarios A.1 - A.3), it is clear that the increased PV
penetration in the long-term will not affect substantially the RES uplift charge, since the FiTs
of the new ground-mounted PV parks for years 2015-2020 are comparable to the resulting
day-ahead SMP (see Fig. 7 and 12).
In fact, the relevant increase in the RES uplift charge from Scen. A.1 to Scen. A.3 is lower
than 0.35 €/MWh among all years. In addition, the implementation of the alternative "market
corridor" pricing scheme for the new ground-mounted PV installations (Scenarios B.1 - B.3)
has marginal effect on the required RES uplift charge, as compared to Scheme A (the
maximum difference is equal to 0.7 €/MWh for the respective PV penetration scenarios
among all years).
Finally, it is clearly shown that the full implementation of the temporary additional
measures for the period 2014-2015 (i.e. revenues from CO2 emissions rights auctions and the
special levy on RES producers) mitigate the economic impact of the imposed RES uplift
charge on the electricity consumers by around 35-40% with respect to the period 2016-2020.
27
5. Conclusion
In this paper a simulation analysis of the Greek electricity market for the next seven-year
period (2014-2020) on an hour-by-hour basis considering five different RES technologies was
presented. Two distinct wholesale market models and three different scenarios regarding the
evolution of the installed capacity of the ground-mounted PV systems in the period 2015-
2020 were considered.
Simulation results showed that the wholesale market model transition from the mandatory
pool to the power exchange in 2015 along with the abolition of the current favorable market
rules for merchant CCGTs from 2015 onwards is expected to lead to: a) an unambiguous
increase in the resulting day-ahead SMP, b) a significant decrease of the CCGT total
operating hours and energy production and c) a notable increase of the CCGTs cycling in
order to compensate for the large intermittent energy injection coming from RES plants. In
addition, the continuous increase of RES installed capacity will result in a notable decrease in
SMP and CO2 emissions from conventional thermal units.
The aforementioned forthcoming changes in the wholesale market regulatory framework
and the future generation mix call into question the economic viability of CCGTs under high
renewable penetration, since the profits from their participation in the day-ahead market are
expected to decrease dramatically. Therefore, alternative measures should be promoted to
ensure the continuing profitable operation of these units. One such measure would be the
possible increase in the capacity availability tickets (CAT) prices, according to which energy
producers are remunerated for the availability of their generating units in the framework of
the capacity assurance market.
Regarding the retail market operation, the effects of the recently introduced in the German
PV market "market-corridor" pricing scheme for the compensation of the new ground-
mounted PV installations from 2015 onwards were investigated. Test results indicated that
this alternative to the currently provisioned wholesale market SMP-based FiT pricing scheme
28
would have little effect on the RES uplift charge undertaken by the Greek electricity
consumers, while at the same time the domestic PV industry would continue to grow in the
following years with the associated positive effects for the environment and the Greek
national economy.
References
Baslis, C.G., Papadakis S.E., Bakirtzis, A.G., 2009. Simulation of optimal medium-term hydro-thermal system
operation by grid computing. IEEE Trans. on Power Syst. 24 (3), 1208-1217.
Brooke, A., Kendrick, D., Meeraus, A., 1990. GAMS User’s Guide, Redwood City, CA: The Scientific Press.
<http://www.gams.com/docs/document.htm>
Catalão, J.P.S., Mariano, S.J.P.S., Mendes, V.M.F., Ferreira, L.A.F.M., 2008. Short-term scheduling of thermal
units: emission constraints and trade-off curves. Euro. Trans. on Electr. Power. 18 (1), 1-14.
Couture, T., Gagnon, Y., 2010. An analysis of feed-in tariff remuneration models: Implications for renewable
energy investment. Energy Policy 38 (2), 955-965.
Delarue, E. D., Luickx, P. L., D'haeseleer, W. D., 2009. The actual effect of wind power on overall electricity
generation costs and CO2 emissions. Energy Convers. Manage. 50 (6), 1450-1456.
Denny, E., O' Malley, M., 2006. Wind generation, power system operation and emissions reduction. IEEE Trans.
on Power Syst. 21 (1), 341-347.
Devroye, L.,1986. Non-Uniform Random Variate Generation, Springer-Verlag, New York.
Doherty, R., O' Malley, M., 2005. New approach to quantify reserve demand in systems with significant
installed wind capacity. IEEE Trans. on Power Syst. 20 (2), 587-595.
Dusonchet, L., Telaretti, E., 2010. Economic analysis of different supporting policies for the production of
electrical energy by solar photovoltaics in western European Union countries. Energy Policy 38 (7), 3207-3308.
European Commission, 2009a. Directive 2009/82/EC of the European Parliament and of the Council of 23 April
2009: Promotion of the use of energy from renewable sources.
<http://eurlex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2009:140:0016:0062:en:PDF>
European Commission, 2009b. Directive 2009/72/EC of the European Parliament and of the Council of 13 July
2009 concerning common rules for the internal market in electricity.
29
<http://eurlex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2009:211:0055:0093:EN:PDF>
European Renewable Energy Council, 2004. Renewable energy target for Europe: 20% by 2020.
Fink, S., Mudd, C., Porter, K., Morgenstern, B., 2009. Wind Energy Curtailment: Case Studies. Tech. Rep.
NREL/SR-550-46716. <http://www.nrel.gov/docs/fy10osti/46716.pdf>
Fulton, M., Capalino, R., 2011. The German feed-in tariff for PV: Managing volume success with price
response. Deutsche Bank Group – DB Climate Change Advisors.
Greek Independent Power Transmission Operator (IPTO), 2009. Wind energy withdrawal capability in
Peloponnese, Greece. Tech. Rep. (in Greek). <http://www.rae.gr/site/file/system/docs/misc/24022011_1>
Greek Independent Power Transmission Operator (IPTO), 2013. <www.admie.gr>
Greek Regulatory Authority for Energy (RAE), 2013. Power Exchange Code.
<http://www.rae.gr/en/codes/main.htm>
Green, R., Vasilakos, N., 2010. Market behaviour with large amounts of intermittent generation. Energy Policy
38 (7), 3211-3220.
Jacobsen, H. K., Zvingilaite, E., 2010. Reducing the market impact of large shares of intermittent energy in
Denmark. Energy Policy 38 (7), 3403-3413.
Kardakos, E.G., Simoglou, C.K., Bakirtzis, A.G., 2013. Short-term electricity market simulation for pool-based
multi-period auctions. IEEE Trans. on Power Syst. 28 (3), 2526-2535.
Li, F., Kuri, B., 2005. Generation scheduling in a system with wind power. In: Proc. IEEE/PES Transmission
and Distribution Conference & Exhibition: Asia and Pacific, Dalian, China.
Maddaloni, J.D., Rowe, A.M., Cornelis van Kooten, G., 2009. Wind integration into various generation
mixtures. Energy Policy 34 (3), 807-814.
Meibom, P., Weber, C., Barth, R., Brand, H., 2009. Operational costs induced by fluctuating wind power
production in Germany and Scandinavia. IET Renew. Power Gen. 3 (1), 75-83.
Morales, J.M., Conejo, A.J., Perez-Ruiz, J., 2009. Economic valuation of reserves in power systems with high
penetration of wind power. IEEE Trans. on Power Syst. 24(2), 900-910.
Morales, J.M., Conejo, A.J., Perez-Ruiz, J., 2011. Simulating the impact of wind production on locational
marginal prices. IEEE Trans. on Power Syst. 26 (2), 820-828.
Muslu, M., 2004. Economic dispatch with environmental considerations: trade-off curves and emission reduction
rates. Elec. Power Syst. Res. 71 (2), 153-158.
30
Sensfuß, F., Ragwitza, M., Genoese, M., 2008. The merit-order effect: a detailed analysis of the price effect of
renewable electricity generation on spot market prices in Germany. Energy Policy 36 (8), 3086-3094.
Simoglou, C.K., Biskas, P.N., Zoumas, C.E., Bakirtzis, A.G., 2011. Evaluation of the impact of RES integration
on the Greek electricity market by mid-term simulation. In: Proc. IEEE/PES PowerTech Conference, Trondheim,
Norway.
Simopoulos, D.N., Kavatza, S.D., Vournas, C.D., 2007. An enhanced peak shaving method for short-term
hydrothermal scheduling. Energy Convers. Manage. 48 (11), 3018-3024.
Stoft, S., 2002. Power Systems Economics: Designing Markets for Electricity, IEEE Press, John Wiley and Sons.
Troy, N., Denny, E., O’ Malley, M., 2010. Base-load cycling on a system with significant wind penetration.
IEEE Trans. on Power Syst. 25 (2), 1088-1097.
Tuohy, A., Meibom, P., Denny, E., O’Malley, M., 2009. Unit commitment for systems with significant wind
penetration. IEEE Trans. on Power Syst. 24 (2), 592601.
Ummels, B.C., Gibescu, M., Pelgrum, E., Kling, W.L., Brand, A.J., 2007. Impacts of wind power on thermal
generation unit commitment and dispatch. IEEE Trans. on Power Syst. 22 (1), 4451.
WILMAR, 2006. Wind power integration in liberalised electricity markets. <http://www.wilmar.risoe.dk/>
Acknowledgment
This work was supported in part by the General Secretariat of Research and Technology
(GSRT), Hellenic Ministry of Education and Religious Affairs, Culture and Sports, in the
context of the Action “ARISTEIA” (Project Code: 1522), in part by the Hellenic Association
of Photovoltaic Companies and in part by the Research Committee of Aristotle University of
Thessaloniki, Greece, in the context of a Post-Doctoral Fellowship.
31
Table 1. Greek Power System Data (Dec. 2013)
Unit Type
Number
of
Units
Installed
Capacity
[MW]
Emissions Rate
Range
[T/MWh]
Base-Load (Lignite-fired) 16 4,302 1.04 - 1.96
Intermediate-Load (CCGTs) 10 4,209 0.38 - 0.49
Peak-Load (OCGTs,
Steam Gas units & Oil units) 6 659 0.41 - 0.60
Hydro Units 14 3,200 -
RES plants (incl. CHP) 1 - 4,498 -
Total 46 16,868
Table 2. Construction (Withdrawal) of New (Old) Thermal Units
Construction of new units Withdrawal of old units
Unit Capacity
[MW]
Commercial
Operation
Date
Time
Period
Number
of
units
Capacity
[MW]
Megalopoli 5
(CCGT) 811 07/2015 2014-2016 5 902
Ptolemaida 5
(Lignite-fired) 600 01/2019 2019 6 1,656
Total 1,411 11 2,558
Table 3. RES Average Feed-in-Tariffs (Dec. 2013)
RES Technology Share
(%)
Average FiT
[€/MWh]
Wind 34.7 91.4
Small Hydro 4.9 90.1
Biomass 1.3 106.6
CHP Decentralized units 2.1 183.2
Large CCGT 2 2.4 110.0
PV
Systems
Ground-mounted 46.3 396.5
Rooftop 8.2 500.7
32
Table 4. RES & Cogeneration Special Account Balance (Scen. A.1 - A.3)
Expenses
(A) Revenues (Other) (B)
Income from
Electricity Consumers
(C) = (A) - (B) [M€]
RES Uplift Charge
[€/MWh] Scen. Year
RES
FiTs
[M€]
Day-Ahead
Market
[M€]
CO2 Emissions
Rights
Auctions [M€]
Special Levy
on Lignite Production
[M€]
Special Levy
on Public Television
[M€]
Special Levy
on RES producers
[M€]
2014 1,946.6 552.5 147.9 52.3 75.0 367.2 751.7 14.14
A.1
2015 2,039.5 693.6 155.9 54.7 75.0 174.7 885.7 16.61
2016 2,171.7 774.4 78.5 52.9 75.0 - 1,190.9 22.20
2017 2,290.7 914.1 76.4 51.9 75.0 - 1,173.4 21.61
2018 2,425.6 1,035.2 75.3 51.8 75.0 - 1,188.3 21.57
2019 2,543.8 1,131.7 70.9 50.7 75.0 - 1,215.5 21.74
2020 2,698.2 1,312.0 60.0 41.3 75.0 - 1,210.0 21.32
A.2
2015 2,042.2 695.8 155.8 54.7 75.0 174.7 886.2 16.62
2016 2,182.9 782.7 78.4 52.9 75.0 - 1,194.0 22.26
2017 2,312.5 932.1 76.2 51.8 75.0 - 1,177.4 21.68
2018 2,460.4 1,059.7 75.0 51.7 75.0 - 1,199.0 21.76
2019 2,594.8 1,179.5 70.3 50.5 75.0 - 1,219.6 21.81
2020 2,768.8 1,379.3 59.3 41.2 75.0 - 1,214.1 21.39
A.3
2015 2,044.9 698.5 155.8 54.7 75.0 174.7 886.3 16.62
2016 2,194.1 794.8 78.2 52.8 75.0 - 1,193.2 22.24
2017 2,334.0 952.3 75.9 51.7 75.0 - 1,179.1 21.71
2018 2,495.0 1,086.2 74.5 51.6 75.0 - 1,207.7 21.92
2019 2,645.2 1,219.1 69.6 50.2 75.0 - 1,231.2 22.02
2020 2,838.6 1,437.8 58.5 40.9 75.0 - 1,226.4 21.61
33
Fig. 1 Mean hourly energy injection per month for each RES technology
Fig. 2 Mean hourly energy injection per month for PV plants
Fig. 3 RES installed capacity per technology
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24PV
Mea
n H
ourl
y E
ner
gy I
nje
ctio
n [
MW
h/M
W]
Hours
Jan-Feb-Dec Mar-Nov Apr-May-Sep
Oct Jun-Aug Jul
0
1000
2000
3000
4000
5000
6000
2014 2015 2016 2017 2018 2019 2020
Inst
alled
Ca
pa
cit
y [M
W]
Wind Biomass sHydro Cogen
PV - S.1 PV - S.2 PV - S.3
34
Fig. 4 Wholesale electricity market simulation algorithm
35
Fig. 5 Load Demand of the Greek Interconnected Power System
Fig. 6 Natural gas and CO2 emissions cost
Fig. 7 Average yearly system marginal price
10000
10200
10400
10600
10800
11000
50
51
52
53
54
55
56
57
58
59
60
2013 2014 2015 2016 2017 2018 2019 2020
Pea
k L
oa
d D
em
an
d [M
W]
Ele
ctr
icit
y C
on
sum
pti
on
[T
Wh
]
Electricity Consumption [TWh] Peak Load Demand [MW]
0.0
1.0
2.0
3.0
4.0
5.0
0.37
0.38
0.39
0.40
0.41
0.42
0.43
0.44
0.45
2014 2015 2016 2017 2018 2019 2020
CO
2 E
mis
sio
ns
Co
st [€
/T]
Na
tura
l Ga
s C
ost
[€
/Nm
3]
Natural Gas CO2
45
50
55
60
65
70
75
80
2014 2015 2016 2017 2018 2019 2020
Sy
stem
Ma
rgin
al P
rice [
€/M
Wh
]
S.1 S.2 S.3
36
Fig. 8 Number of start-ups of thermal units
Fig. 9 Operation hours of thermal units
Fig.10 Thermal production and CO2 emissions
0
100
200
300
400
500
600
700
800
900
1000
2014 2015 2016 2017 2018 2019 2020
Nu
mb
er o
f S
tart
-up
s
Lignite - S.1 Lignite - S.2 Lignite - S.3
CCGT - S.1 CCGT - S.2 CCGT - S.3
0
20000
40000
60000
80000
100000
120000
2014 2015 2016 2017 2018 2019 2020
Un
its
Op
era
tio
n h
ou
rs
Lignite - S.1 Lignite - S.2 Lignite - S.3
CCGT - S.1 CCGT - S.2 CCGT - S.3
26
28
30
32
34
36
38
40
42
0
5
10
15
20
25
30
2014 2015 2016 2017 2018 2019 2020
CO
2 E
mis
sio
ns
[MT
on
]
Th
erm
al P
rod
ucti
on
[T
Wh
]
Lignite - S.1 Lignite - S.2 Lignite - S.3
CCGT - S.1 CCGT - S.2 CCGT - S.3
CO2 - S.1 CO2 - S.2 CO2 - S.3
37
Fig. 11 Yearly thermal production cost
Fig. 12 FiTs of PV systems
Fig. 13 Special RES & Cogeneration Account balance
0
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
2014 2015 2016 2017 2018 2019 2020
Th
erm
al P
rod
ucti
on
Co
st [
M€
]
S.1 S.2 S.3
50
60
70
80
90
100
110
120
130
2014 2015 2016 2017 2018 2019 2020
PV
Fed
d-I
n-T
ari
ffs
[€/M
Wh
]
PV Roofs Scen. A.1 Scen. A.2 Scen. A.3
Scen. B.1 Scen. B.2 Scen. B.3
38
Fig.14 RES uplift charge
Footnotes
1 See Table 3 for share of different RES technologies
2 According to the current Greek regulatory framework, a portion of the nominal capacity of a large
CCGT plant (Alouminio) (i.e. 110 MW out of 310 MW) operates as a CHP unit and is
compensated on the basis of a regulated FIT.
0
5
10
15
20
25
2014 2015 2016 2017 2018 2019 2020
RE
S U
plift
Ch
arg
e [
€/M
Wh
]
Scen. A.1 Scen. A.2 Scen. A.3Scen. B.1 Scen. B.2 Scen. B.3