© 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

(chsimoglou@ee.auth.gr, pbiskas@auth.gr, stelvag@gmail.com, bakiana@eng.auth.gr)

Corresponding author: Anastasios G. Bakirtzis (bakiana@eng.auth.gr)

Power Systems Laboratory, Division of Electrical Energy, Department of Electrical and

Computer Engineering, Aristotle University of Thessaloniki, 54124, Thessaloniki, Greece, E-

mail: bakiana@eng.auth.gr, 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