You are on page 1of 15

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/267594671

Optimization of Thermal Power Plants Operation in the German De-Regulated


Electricity Market Using Dynamic Programming

Conference Paper · November 2012


DOI: 10.1115/IMECE2012-86113

CITATION READS
1 852

5 authors, including:

Sergio Rech Andrea Lazzaretto


University of Padova University of Padova
14 PUBLICATIONS   88 CITATIONS    142 PUBLICATIONS   2,724 CITATIONS   

SEE PROFILE SEE PROFILE

Andreas Christidis George Tsatsaronis


Technische Universität Berlin Technische Universität Berlin
16 PUBLICATIONS   109 CITATIONS    257 PUBLICATIONS   6,783 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Combustion View project

Other Papers View project

All content following this page was uploaded by George Tsatsaronis on 17 November 2014.

The user has requested enhancement of the downloaded file.


Proceedings of the ASME 2012 International Mechanical Engineering Congress & Exposition
IMECE2012
November 9-15, 2012, Houston, Texas, USA

IMECE2012-86113

OPTIMIZATION OF THERMAL POWER PLANTS OPERATION


IN THE GERMAN DE-REGULATED ELECTRICITY MARKET USING DYNAMIC PROGRAMMING
G. SCARABELLO S. RECH* A. LAZZARETTO
Department of Industrial Engineering Department of Industrial Engineering Department of Industrial Engineering
University of Padova, Italy University of Padova University of Padova, Italy
via Venezia, 1 - 35151 Padova, Italy
Email: sergio.rech@studenti.unipd.it

A. CHRISTIDIS G. TSATSARONIS
Institute for Energy Engineering, Institute for Energy Engineering,
Technische Universität Berlin, Germany Technische Universität Berlin, Germany

ABSTRACT stable electricity supply in the whole system. Since the


The prospect of clean electrical energy generation has introduction of the European Union Greenhouse Gas Emission
recently driven to massive investments on renewable energies, Trading Scheme (EU ETS) in 2005, also CO2 allowance prices
which in turn has affected operation and profits of existing have been participating in the electricity price formation as an
traditional thermal power plants. essential part of specific generation costs. For all these reasons
In this work several coal-fired and combined cycle power uncertainty on fossil fuel utilities is great (lower prices reduce
units are simulated under design and off-design conditions to full load hours and earnings, while volatility requires higher
adequately represent the behavior of all modern thermal units flexibility [1]), and investments on these plants need to be
included in the German power system. A dynamic optimization supported by accurate economical estimations, which require:
problem is then solved to estimate the short-run profits obtained Knowledge of thermodynamic models of the plants.
by these units using the spot prices of the German electricity General principles of energy system modeling at design
market (EEX) in years 2007-2010. The optimization model is (see, e.g., [2-4]) and off-design (see e.g., [5, 6]) conditions
developed using a Mixed Integer Linear Programming approach are widely discussed in the literature. Several paper deal
to take the on-off status into account and reduce computational with the application of these principles to traditional thermal
effort. New market scenarios with increasing renewable shares power plant simulations. Interesting applications to coal-
(and consequently different spot prices) are finally simulated to fired power plants are shown in [7-9]. In particular [7]
analyze the consequences of a larger capacity of renewable compares two different control strategies in terms of part
energies on the optimal operation of traditional thermal power load efficiency. Other significant applications on combined
plants. cycle power plants are presented in [10, 11] where part load
performance is evaluated using different control strategies.
INTRODUCTION Definition of the plant optimization strategy.
In the last years environmental concerns and high oil prices General principles of energy system design and operation
has led in Germany to a relevant growth of renewable energy optimization are presented in [12, 13]. A review of a large
capacity. So, power utilities are facing new market conditions: number of modeling methods is shown in [14]. Searching
lower spot electricity prices and higher volatility, which can the optimal operation strategy of power plants under
greatly affect operation and earnings of traditional power units. conditions of deregulated market is usually a dynamic
Moreover, the continuous fluctuation in the renewable energy problem. In order to reduce the computational effort the
supply not only complicates the unit commitment planning, but optimization model is generally simplified by using only
also considerably increases the reserve requirement for the linear relationships, and the on-off status of each plant is

1 Copyright © 2012 by ASME


modeled by binary variables. Accordingly, a Mixed Integer Letter symbols
Linear Programming (MILP) is to be solved. Several papers a generic constant [adimensional]
describe MILP models for short-term scheduling b generic constant [MWh]
optimization. In particular [15] proposes an interesting c specific price/cost [€/MWh or €/t]
approach to model the plant behavior during operation C cost per unit [€/#]
(including load ramps and start-ups behavior). [16] and [17] D electric demand [MWh]
offer significant examples of various constraints in MILP F fuel consumption [MW]
models. m number of period required to a warm start-up [#]
Analysis of different market scenarios with a growth of n number of period required to a cold start-up [#]
renewable energy capacity. P electric power [MW]
Renewable Energy Sources (RES) generation strongly ∆P load ramp rates [MW/h]
influences market prices. Within the wide literature on this RL residual load [MWh]
issues [18] and [19] study the effect of wind capacity in the t time set [#]
German electricity market (EEX). T1, T2 (T2 > T1) time periods that define hot, warm or cold
start-up [h]
In this work the German electricity generation system is ∆Τ time interval [h]
analyzed by an economical point of view. The goal consists in z short-run profit [€]
evaluating profits generated by thermal power plants in years
2007-2010, and in foreseeing how these profits may change in Greek symbols
different market conditions, in particular under a growth of α specific emission of CO2 [t/MWh]
energy production from RES. To this end we first consider δ binary variable [adimensional]
several coal-fired and combined cycle power units, which are η efficiency [adimensional]
representative of the German power units system. These power Subscripts
plants are modeled using a commercial simulation software to av average
obtain their characteristic curves (power vs. fuel). Thereafter, CO2 carbon dioxide
these curves are used in a MILP model to optimize the short-run CS cold start-up
profits obtained using the spot prices of the EEX. Finally, the DT down time
electricity market system is modeled and the share of electric e electricity
energy produced by wind and PV plants is increased to show el electric
how profits deriving from thermal power plants are affected by f fuel
different amounts of electricity generated by RES. HS hot start-up
max maximum
NOMENCLATURE min minimum
OT operating time
Acronyms WS warm start-up
CC Combines Cycles power plant
CC3P CC based on a three pressure levels steam cycle Superscripts
CC1P CC based on a single pressure levels steam cycle min minimum
CCHC high performing CC
CCLC low performing CC THE GERMAN POWER SYSTEM
EEX German electricity market In 2010 the German electricity production amounted to
EU ETS European Union Greenhouse Gas Emission Trading 621·106 MWh. The production was shared between four main
Scheme sources: coal, nuclear, gas and renewable energies. The installed
GT Gas Turbine capacity at the end of 2010 amounted to 165859 MW: 53944
MILP Mixed Integer Linear Programming MW from renewables, 48225 MW from coal, 20490 MW from
PV Photo Voltaic nuclear, 25500 MW from gas and 17700 MW from other
RES Renewable Energy Sources sources.
ST coal-fired Steam Turbine power plant Figure 1 shows the share of electricity production in 2010:
STHC ST based on a subcritical steam cycle with advanced among RES [20]. Wind is the main source with 6% of total
characteristics electricity production, followed by biomass (5%), hydro (3%),
STLC ST based on subcritical steam cycle with low PV (2%) and wastes (1%) [20]. The power plants capacity
characteristics factors were calculated (Tab. 1) as total electricity production
STSC ST based on supercritical steam cycle divided by maximum electricity production (number of hours in
one year multiplied by installed capacity) using data taken from
[20].

2 Copyright © 2012 by ASME


Merit order curve
Electricity can be produced in many ways using a variety of
fuels and applying different technologies. On competitive short
term markets, prices are set by the short-run marginal cost of
the plant producing the last unit of electricity required to meet
demand (Fig. 2). The last, or marginal, unit needed to meet
demand is also the one with the highest short-run marginal cost
at a given time. This process ensures that power plants in
operation are those with lower marginal costs among all power
plants available to operate. So, it can be expected that nuclear
and lignite-fired power plants will be dispatched continuously,
i.e. they serve as base load units. These plants (together with
hydroelectric plants) are listed on top of the so-called merit
Fig. 1: Electricity production by source in 2010 in order list, which means their marginal costs are low compared
Germany [20] to those of the other plants. [21].
Tab. 1: Average power plants capacity factors by source
in 2010 in Germany.
CAPACITY FACTOR
Coal 0.57
Nuclear 0.74
Gas 0.37
Renewables 0.21
Wind 0.15
Biomass 0.67
Hydro 0.42
PV 0.08
Wastes 0.70
Oil and Others 0.18

This picture is not stable: in 2011 after Fukushima disaster


five nuclear power plants were shut-down, and this source is Fig. 2: Merit order curve and marginal costs of different
expected to be totally abandoned by 2022. Moreover, in the last power plants: hydroelectric, nuclear, lignite-fired, coal-
years the capacity of renewable energies in Germany has greatly fired, combined cycles (CC), gas turbine (GT)
increased. At the end of 2010 the installed wind capacity was
27 GW, while the PV capacity amounted to 17.3 GW. The All plants receive the marginal price of the system. If the
growth of installed capacities in the last years is impressive: the market is competitive, with a sufficiently large number of
PV installed capacity has almost doubled in 2010, while the generators, each one knows that it cannot affect the price by
wind capacity has shown a constant growth of 2 GW per year in changing its own bid. In that case, its profit-maximizing strategy
the last 4 years. consists in submitting a bid equal to its own running costs: the
These data show how the German production system has plant must run every time running is profitable, that is when
been changing, as well as in most of the European countries. market price is higher than running costs. A higher bid imposes
Investments on new power plants must be supported by careful the risk that the station will not be scheduled to run during some
analyses: the paper aims at evaluating the influence of these periods when the market price is below this bid, but greater than
changes on thermal power plants operation. the station’s costs, implying that running would be profitable. A
lower bid imposes the risk that the station would be required to
DE-REGULATED ELECTRICITY MARKET run when the price was below its costs, incurring losses.
Due to technological and economical problems associated
with electricity storage, electricity generation and demand must THERMAL POWER PLANTS MODELING
be matched continuously in the electricity market. This requires A system model is the set of mathematical relationships that
every unit to follow the operating instructions of a central defines all the interactions between the quantities required to
dispatcher, and to modify its load according to its schedule. The describe the real system behavior with the desired degree of
dispatchers decide on the hourly operational status of every approximation. Basically, these relationships include mass and
single unit of their power plant fleet depending on the energy balances, component characteristic curves and equations
generation costs of the units and the expected electricity market of states [6].
price, determined by the merit order and the electricity demand.

3 Copyright © 2012 by ASME


Typically the design model aims at evaluating the “design Thermal power plants chosen as representative of the
point”, i.e. the size of the system and its behavior at nominal German traditional electricity generation system
load, being technological and environmental constraints Steam and combined cycle power plants were chosen as
assigned. Starting from the design point, the off-design model representative categories of the “traditional” German power
describes the system behavior at partial load. generation system. Three coal-fired and four combined cycle
power plants were included in each category. Off-design models
Control Strategy of these plants were built using the commercial software
In building off-design models, attention has to be paid on Ebsilon®Professional to obtain reliable characteristic curves
the control strategy. In particular, in the steam cycle power (fuel consumption vs. output power).
plant models the two main control strategies were used [7]: The three coal-fired power plants are based on: a
Fixed pressure: a nozzle governing stage (see Fig. 3) is supercritical steam cycle (STSC, Pmax=500MW), a subcritical
placed between the steam generator and the high pressure steam cycle with advanced characteristics (STHC,
turbine: it consists of a small number of nozzles (mostly Pmax=500MW) and a subcritical steam cycle with low
between three and five). When the plant is working at characteristics (STLC, Pmax=420MW). STSC and STHC
nominal load, all nozzles are open. On the other hand, when represent the state of the art of the coal-fired power generation,
the plant is working at partial load, some nozzles might be instead STLC represents out-of-date and small plants. Design
closed, some might be still completely open and one might features of the three plants are based on [9]. Sliding pressure
be partially open, introducing a throttling. In this way the control is utilized for all models. As an example, the
steam pressure at the exit of the boiler is constant. The goal Ebsilon®Professional model (at design conditions) of STHC is
of this control strategy consists in having the maximum shown in Annex A. Design and off-design simulations allows
amount of fluid working at design conditions. An interesting the three characteristic curves in Fig. 4 (in the form thermal
study on the operation of steam turbines which are efficiency vs. load) to be obtained.
controlled by a governing stage is shown in [8].
0,60
η el

STSC
0,55
STHC

0,50
STLC
DesignPoint
0.4605
0,45
0.4300
0,40
0.3727
0,35

0,30
30% 40% 50% 60% 70% 80% 90% 100%
Fig. 3: Simple scheme of a steam turbine governing P/P max
stage
Fig. 4: Characteristic curves of the three considered
Sliding pressure control: the turbine operates with fully coal-fired power plants
opened governing valves in every load condition, and the
pressure at the outlet of the steam generator equals the inlet Two of the four combined cycle power plants are based on a
pressure of the high pressure turbine, which is determined three pressure levels (CC3P, Pmax=400MW) and a single
by the actual process conditions in the steam turbine pressure level (CC1P, Pmax=385MW) steam cycles. The other
(turbine inlet temperature, mass flow rate, and condenser two are a high performing (CCHC, Pmax=416MW) and a low
pressure [22]). performing (CCLC, Pmax=350MW) combined cycles. CCHC
and CC3P represent the state of the art of the combined cycles
Moreover, the precise knowledge of the off-design behavior power generation. Instead, CC1P and CCLC represent small
of each plant component is required. This behavior is defined power plants.
by the so called “characteristic curves”, which link component As an example, the Ebsilon®Professional model (at design
performance and efficiency to load. When not available from conditions) of CC3P is shown in Annex B. Again, the
the manufacturers, these curves can be obtained from the characteristic curves (Fig. 5) are obtained by running the
literature or using experimental data (e.g., heat transfer simulation models.
coefficient vs. load for the heat exchangers).

4 Copyright © 2012 by ASME


0,60 under the condition:
0.5900
(δ ( t + 1 ) − δ ( t )) ⋅ (∆TDT ( t ) − ∆TDTmin ( t )) ≥ 0
η el

0.5662 (3)
0,55 0.5428
were δ(t) is the same variable that appears in (1), ∆TDT is the
0,50 0.5000
elapsed time from the latest plant shut-down and ∆TDTmin is
0,45 CCHC the minimum downtime imposed.
CCLC According to (2) the value of ∆TDT is increased by one hour
0,40 CC3P for each hour in which the system remains off (δ=0).
CC1P Equation (3) states that the plant can be turned up only if the
0,35 DesignPoint minimum downtime has been reached. Equations (2) and (3)
need to be linearized because they include a product
0,30
between two time dependent variables (δ and ∆TDT). To this
30% 40% 50% 60% 70% 80% 90% 100%
P/Pmax end the Glover’s linearization method [23] is used.
Constraint on the minimum operating time of each plant.
Fig. 5: Characteristic curves of combined cycle power This is not a technological constraint but it is a “good
plants practice”:

OPTIMIZATION METHODOLOGY ∆TOT ( t + 1 ) = (∆TOT ( t ) + 1) ⋅ (δ ( t + 1 )) (4)


The operation of the plants analyzed in the previous section
under the condition:
is optimized in the period 2007-2010 by building a dynamic
optimization model. To adequately model the behavior of the (δ ( t + 1 ) − δ ( t )) ⋅ (∆TOT ( t ) − ∆TOTmin ( t )) ≥ 0 (5)
plans and to simplify the problem a MILP approach is used. The
three main features of the MILP model are: where ∆TOT is the elapsed time from the latest start-up of the
1. Binary variables are introduced to take the on-off status into plant and ∆TOTmin is the minimum operating time imposed.
account. Again (4) and (5) are linearized using the Glover’s method
2. Only linear relationships are included in the model in order [23]. The detailed reformulation of (2) to (5) can be found
to reduce the computational effort derived from the very in the appendix of [24]
large number of time intervals that are considered. Counting the start-ups to evaluate the associated costs. A hot
3. The period 2007-2010 is subdivided into one-hour time start-up is defined when the time period between shut-down
intervals, so that the total number of time intervals is and start-up is shorter then T1, a warm start-up when it is
4·8760=35040. Accordingly, any model variable assumes a shorter than T2 and longer than T1 (T2 > T1), a cold start-up
discrete value in each interval t. when it is longer than T2. T1 and T2 are specified periods of
time which depend on the plant being considered.
The fundamental equations of the model consist in: Three inequalities are used to identify hot, warm and cold
Characteristic curves of each plant. The characteristics in start-ups:
Figs. 4 and 5 can be represented also by replacing the
δ ( t ) − δ ( t + 1 ) + δ HS ( t + 1 ) ≥ 0
efficiency with fuel consumption F(t), as shown in (1). A
binary variable δ(t) is introduced in (1) to take into account δ ( t ) − ... − δ ( t + T 1 ) − δ ( t + T 1 + 1 ) + δWS ( t + T 1 + 1 ) ≥ 0 (6)
the on-off status of each unit: δ ( t ) − ... − δ ( t + T 2 ) − δ ( t + T 2 + 1 ) + δ CS ( t + T 2 + 1 ) ≥ 0
F ( t ) = a ⋅ P( t ) + b ⋅ δ ( t )
Inequalities (6) force the three binary variables δHS, δWS and
P( t ) ≤ Pmax ⋅ δ ( t ) (1)
δCS to turn their value to 1 when a hot, warm or cold start-up
P( t ) ≥ Pmin ⋅ δ ( t ) occurs, respectively. So:
- in the first inequality, the hot start-up implies δ(t)=0 and
In fact, when δ(t) is set to 0 the output power production δ(t+1)=1 so that δHS(t+1)=1);
P(t) and the fuel consumption F(t) of the plant are zero (i.e. - in the second inequality, the warm start-up implies
the plant is turned off). On the other hand, when δ(t) is set to δ(t)=0,…,δ(t+T1)=0, and δ(t+T1+1)=1 so that
1, P(t) can vary from the minimum value Pmin to the δWS(t+T1+1)=1);
maximum value Pmax (i.e., the plant is running). - in the third inequality, the cold start-up implies
Constraint on the minimum downtime ∆TDT of each plant. δ(t)=0,…,δ(t+T2)=0, and δ(t+T2+1)=1 so that
This is introduced to avoid excessive thermal stress of plant δCS(t+T2+1)=1);
components.
∆TDT (t + 1) = (∆TDT (t ) + 1) ⋅ (1 − δ (t + 1) ) (2) If no start-up takes place δHS, δWS and δCS are set to 0.

5 Copyright © 2012 by ASME


Maximum start-up and load ramp rates: four different values where ce is the specific electricity price (€/MWh); cf is the
are imposed to the maximum load ramp rates depending on specific fuel cost (€/MWh); α is the emission of CO2 per unit of
the working conditions (i.e., hot start-up, warm start-up, fuel (t/MW); c*CO2 is the specific cost of the CO2 emission (€/t);
cold start-up and normal operation). The binary variables CHS, CWS and CCS are the cost (€) of a single hot, warm and cold
δHS, δWS and δCS introduced in (6) are used again to identify start-up, respectively. It is worth noting that a colder start-up
the working condition and the respective maximum load implies a warmer one. Thus, in (6) when a hot start-up occurs
ramp rate: only δHS turns to 1, when a warm start-up occurs both δHS and
δWS turn to 1 and when a cold start-up occurs all the three
P( t + 1 ) − P( t ) ≤ ∆Pmax +
variables turn to 1. This is the reason why in (9) the costs of a
+ δ HS ( t + 1 ) ⋅ [∆Pmax,HS − ∆Pmax ] + warmer start-up are subtracted from the costs of a colder one
+ δ WS ( t + 1 ) ⋅ [∆Pmax,WS − ∆Pmax ] + (7) (so, CHS is subtracted from CWS and CWS is subtracted from CCS).
+ δ CS ( t + 1 ) ⋅ [∆Pmax,CS − ∆Pmax ] +
The decision variables of the problem are the binary variables
δ(t) and the load P(t) in all time intervals of the period 2007-
t −m
+ ∑ δ WS ( i ) ⋅[∆Pmax,WS − ∆Pmax ] +
2010 (the size of the plants is assigned).
i =t
t −m
OPTIMIZATION OF THERMAL POWER PLANTS
+ ∑ δ CS ( i ) ⋅[∆Pmax,CS − ∆Pmax,WS ]+
i =t + 1
OPERATION IN THE PERIOD 2007-2010
The optimization methodology described in the previous
t −m
+ ∑ δ CS ( i ) ⋅[∆Pmax,CS − ∆Pmax ]
section is applied to build and solve seven MILP optimization
models (one for each power plant) using the commercial
i =t
software GAMS® and the solver CPLEX.
where ∆Pmax, ∆Pmax,HS, ∆Pmax,WS and ∆Pmax,CS are the
maximum load ramp rates in normal operation or during a Input data
hot, warm or cold start-up respectively, m is the number of Input data used in the optimization model include (these
time intervals required for a warm start-up and n is the data were taken from the references indicated within square
number of time intervals required for a cold start-up (it is brackets):
assumed that a hot start-up requires only one time interval). • Thermodynamic parameters (a and b in (1)) describing the
In (7) the last three summations are introduced to fix the plant characteristic curves (Figs. 4 and 5).
correct value of the maximum load ramp during the start- • Data associated with the dynamic plant behavior (Pmin, T1,
ups. It is worth noting that the constraints on load ramp rates T2, ∆TDTmin and ∆TOTmin) [25,26] (Tab. 2). Note that the
link every time interval with the others, i.e. makes the warm start-up is not defined for CC plants (T1=T2, see (6)).
problem dynamic. • Start-up costs (Tab. 3) [25]. Warm start-up and hot start-up
The variables Pav(t) and Fav(t) are used to evaluate correctly costs are obtained by multiplying the cold start-up costs by
the energy production and the fuel consumption during the the factors indicated in the Tab. 2.
load ramps. These two variables are calculated as arithmetic • Specific costs of fuel [27,28] and CO2 emissions [29]
average between the values of P and F at t and t-1: (Tab. 4).
P( t ) + P( t − 1 ) F( t ) + F( t − 1 ) • Features of the hourly price of the electricity (average value
Pav ( t ) = ; Fav ( t ) = (8) and standard deviation from the average, see Tab. 5).
2 2
Performance degradation between overhaul and maintenance Tab. 2: Dynamic constraints for the MILP model
interventions is not considered for simplicity. ST PLANT CC PLANT
The objective function z to be maximized is the short-run Pmin 0,35 Pmax 0,4 Pmax
profit of each power plant (considered separately). It is T1 3h 2h
computed as the difference between the revenues deriving from T2 5h 2h
selling electricity and the operating costs due to fuel ∆TDTmin 5h 5h
consumption, CO2 emissions, hot, warm and cold start- ups: ∆TOTmin 5h 5h

z = ∑[ Pav ( t ) ⋅ ∆T ⋅ ce ( t )] −∑[ Fav ( t ) ⋅ ∆T ⋅ ( c f + α ⋅ c *CO2 )] +


t t Tab. 3: Start-up costs
(9)
− ∑[ δ HS ( t ) ⋅ CHS ] − ∑[ δWS ( t ) ⋅ (CWS − CHS ) ] + ST CC
t t SPECIFIC COSTS
PLANT PLANT
− ∑[ δ CS ( t ) ⋅ (CCS − CWS ) ] Fuel additional cost [€/MW] 35.30 56.00
t CCS
Maintenance additional cost [€/MW] 4.80 10.0
CWS Warm start-up factor 0.50 0.50
CHS Hot start-up factor 0.30 0.30

6 Copyright © 2012 by ASME


Tab. 4: Fuel and CO2 emission costs In 2007 the very low fuel and CO2 costs resulted in high
profits. In 2008 the EEX prices increased more than coal and
2007 2008 2009 2010
CO2 costs yielding high profits for all plants. Instead, in 2009
COAL [€/MWh] 9.23 14.01 12.04 12.50
GAS [€/MWh] 17.19 23.15 17.95 17.73
and 2010 coal and CO2 costs decreased less than EEX prices so
CO2 [€/t] 0.66 17.38 13.00 14.32 that profits decreased as well.
Table 7 and Fig. 7 show the optimization results of the
Tab. 5: Summary of EEX prices in 2007-2010 period combined cycle power plants operation. Total profits in the four
years amount to 465.77 €/kW for CCHC (ηel=59.00%), 426.91
ce,av Standard deviation €/kW for CC3P (ηel=56.62%), 387.82 €/kW for CC1P
[€/MWh] [€/MWh]
(ηel=54.26%) and 314.71 €/kW for CCLC (ηel=50.00%).
2007 38.11 28.51
2008 65.90 30.32
As for the coal-fired plants, 2008 turned out to be a very
2009 38.86 19.40 profitable year. Profits in 2007 are much closer to 2010 and
2010 44.49 13.98 2009 than for coal-fired plants: CO2 costs are less relevant in
combined cycles because of lower carbon emission from natural
Results gas and higher efficiency of these plants. On the average,
The main results of the seven optimization problems are specific profits increased by 17 €/kW per each efficiency point
summarized in the following. between 50% and 59% during the period 2007-2010.
Table 6 and Fig. 6 show the optimization results of the coal- Figure 8 shows the average production costs of each power
fired power plants operation. Total profits in the four years plant, including fuel consumption, CO2 emissions and start-up
depend on plant efficiency. So, higher profits are obtained with costs. It is worth noting how the merit order has changed from
the STSC (ηel=46.05%, 513.53 €/kW), lower profits for STHC 2007 to 2010. In 2007 the lowest costs are obtained by coal-
(ηel=43.10%, 457.71 €/kW) and STLC (ηel=37.27%, 341.02 fired plant configurations (see Fig. 8 (a)). In the following
€/kW). On the average, specific profits increased by 20 €/kW years, the importance of CO2 costs increased so that combined
per each efficiency point between 37% and 46% during the cycle became progressively more convenient in spite of the
period 2007-2010. increase in natural gas cost (see Fig. 8 (b) to (d)).

Tab. 6: Specific profits of coal-fired power plants Tab. 7: Specific profits of CC power plants

TOTAL PROFITS 2007-2010 TOTAL PROFITS 2007-2010


Power plant STSC STHC STLC Power plant CCHC CC3P CC1P CCLC
Total specific profits z/Pmax Total specific profits
513.53 457.71 341.02 465.77 426.91 387.82 314.71
[€/kW] z/Pmax [€/kW]
∆ ( z / Pmax ) ∆ ( z / Pmax )
[ €/kW ] 19.64 [ €/kW ] 16.78
∆ η el % ∆ η el %

600 600
z/Pmax [€/kW]
z/P max [€/kW]

2007 2008
2007 2008
500 500 2009 2010
2009 2010

400 400

300 300

200 200

100 100

0 0
STSC STHC STLC CCHC CC3P CC1P CCLC
(ηel=46.05%) (ηel=43.10%) (ηel=37.27%) (ηel=59.00%) (ηel=56.62%) (ηel=54.28%) (ηel=50.00%)

Fig. 6: Profits of coal-fired power plants in 2007-2010 Fig. 7: Profits of CC power plants in 2007-2010

7 Copyright © 2012 by ASME


OPTIMIZATION OF THERMAL POWER PLANTS
production cost [€/MWh]

60
2007 OPERATION FOR INCREASING MARKET SHARE OF
50
RENEWABLE ENERGIES
40
Different scenarios of market share increase of renewable
30 energies are analyzed in the following to forecast the changes in
20 the optimal operation strategy of tradition thermal power plants
10 operation in the German electricity market.
0
Effect of renewable energy in the merit order
(ηel=46.05%)

(ηel=43.10%)

(ηel=37.27%)

(ηel=59.00%)

(ηel=56.62%)

(ηel=54.28%)

(ηel=50.00%)
RES fuelled power plants have priority on traditional power
CCHC
STHC

CCLC
CC3P

CC1P
STSC

STLC

plants in the German electricity market. So, the market share of


the latter is reduced depending on the amount of RES generated
(a) electricity [30].
production cost [€/MWh]

60
2008
50

40

30

20

10

0
(ηel=46.05%)

(ηel=59.00%)

(ηel=43.10%)

(ηel=56.62%)

(ηel=54.28%)

(ηel=37.27%)

(ηel=50.00%)
CCHC

STHC

CCLC
CC3P

CC1P
STSC

STLC

(b)
production cost [€/MWh]

60
2008 Fig. 9: Merit order effect of RES
50

40
This concept is shown in the cost vs capacity curve (Fig. 9).
30
The so-called residual load is the power that must be supplied
20 by thermal power plants:
10

0
RL(t ) = D(t ) − PRES (t ) (10)
(ηel=46.05%)

(ηel=59.00%)

(ηel=43.10%)

(ηel=56.62%)

(ηel=54.28%)

(ηel=37.27%)

(ηel=50.00%)
CCHC

STHC

CCLC
CC3P

CC1P
STSC

STLC

where D(t) is the electricity demand and PRES(t) is the


production from RES at time t, which is considered at zero
variable production costs in this representation. This implies a
(c) shift of the merit order curve to the right, which in turn results
60
production cost [€/MWh]

in a decrease in the electricity price ce determined by the


2010
50
intersection between demand and supply at time t. The
40 consequence is the exclusion of high production cost traditional
30
power plants from the market.
20
Model for the evaluation of the new price patterns
10 We consider here that the contribution for RES to electricity
0
production may vary while both the number and type of
traditional power plants remain the same.
(ηel=59.00%)

(ηel=46.05%)

(ηel=56.62%)

(ηel=43.10%)

(ηel=54.28%)

(ηel=50.00%)

(ηel=37.27%)
CCHC

STHC

CCLC

Starting from available hourly data of electricity demand


CC3P

CC1P
STSC

STLC

and hydroelectric, wind, biomass and photovoltaic production


estimates, the actual residual loads (RL(t)) can be evaluated. For
(d) simplicity we assume that the shape of the merit curve remains
the same during the period in which the contribution of RES to
Fig. 8: Average production costs of the power plants electricity production is varied. Each point of this curve is

8 Copyright © 2012 by ASME


obtained using the electricity price (ce(t)) and the corresponding Accordingly, the following four scenarios are considered, which
value of RL(t), which in turn is evaluated by subtracting PRES(t) differ for PV and wind installed capacities only:
to D(t). This assumption of constant shape of the merit curve
• Base: it includes the 2010 photovoltaic and wind installed
would actually apply only during a short period of time in
capacities.
which bids of traditional power plants remain the same due to
approximately constant electricity production costs. • PV3: the photovoltaic installed capacity is three times as
The merit order curve is finally used to calculate “new” much the 2010 one.
electricity prices corresponding to different renewable shares, • W2: the wind installed capacity is twice as much the 2010
which create a shift of the curve to the right depending on the one.
considered values of RES production. • PV3W2: the photovoltaic and the wind installed capacities
are three times and twice as much the 2010 ones,
A model is created in Matlab Simulink® environment (Fig. 10) respectively.
to evaluate the effects of increasing RES generation.
The Base scenario reproduces the actual 2010 power system,
Inputs of the model are:
but market prices of this scenario are not exactly corresponding
• The actual hourly pattern of electricity demand D(t).
to the actual ones as they are given by the merit order curve that
• The actual hourly patterns of electricity production by PV,
was built as described above (basically, price peaks are
wind and biomass.
smoothed). These prices are used as reference in the
• The merit order curve (constant in time) that was built as comparison with other scenarios. PV3W2 scenario represents
described above. the installed capacity expected in 2020 [31]. Three test-months
The RES electricity production (PRES(t)) in each hour t is are considered as representative of Summer, Mid-season and
obtained by summing the electricity productions by PV, wind Winter conditions: August 2010, October 2010 and December
and biomass, which are first amplified (multiplying by a 2010. A single month is a long enough period to cover a quite
constant) to model a capacity increase of each RES. The wide spectrum of price values, and sufficiently short at the same
residual loads (RL(t)) in each hour are then obtained by time to consider a merit order curve “constant”, as explained
subtracting PRES(t) from the demand D(t). Finally, RL(t) is used above. Table 8 shows the prices obtained from running the three
as an input in the merit order curve to evaluate the new hourly Simulink® models (one for each test-month).
pattern of electricity price (ce(t)).
Tab. 8: Electricity prices in the different scenarios

AUGUST OCTOBER DECEMBER


ce,av price ce,av price ce,av price
[€/MWh] volatility [€/MWh] volatility [€/MWh] volatility
Base 39.43 12.42 48.80 13.91 54.21 18.57
PV3 35.57 13.45 46.22 13.51 53.49 18.17
W2 34.35 14.57 40.55 16.93 45.48 21.58
PV3W2 30.56 14.79 38.03 16.42 44.85 21.17

It is worth noting that there are two effects of the renewable


energy share on the electricity price: the first effect is on the
average price ce,av, the second one is on price volatility. Solar
and wind sources have different frequencies and amplitudes
during days and seasons, therefore, their influence on the prices
is different:
®
Solar source has daily and seasonal periodic trend. Thus, the
Fig. 10: Simulink model of the electricity market in the solar power source reduces prices only during the day and,
PV3W2 scenario in particular during peak hours. This effect is higher in
Summer. In fact in December and October, a higher PV
The main assumptions of the model are: feed-in (PV3 scenario) involves a lower volatility than the
Bids of traditional power plants match exactly their short- Base case. On the other hand, in August the high PV feed-in
run marginal costs; involves a higher price difference between day and night
Biomass production is considered constant; (i.e. the volatility is higher). In addition, during weekends
While being a RES, hydroelectric production is not the prices of electricity can drop to very low values.
considered in PRES(t). This simplification is justified by the Wind source is characterized by a frequency that not
fact that hydroelectric production is not expected to vary necessarily follows a daily trend. A windy period can last
significantly in future. some consecutive days, having a bigger effect on electricity

9 Copyright © 2012 by ASME


price than solar source. As a consequence, in the scenario fact, when one of the two RES additional contribution implies
W2 prices fall when a windy period occurs and they can the shut down of a traditional power plant, the effect of the
decrease significantly also in low demand periods (like night other RES is null being the plant already off.
or weekends). For this reason wind energy has a remarkable
effect both on mean price and price volatility. The effect on October and December
volatility is similar in all months considered. In October and December the profits reduction deriving by an
When generation by both these sources increase (PV3W2 increase of RES electricity production is lower than in summer
scenario) the two effects are achieved together: in December because of the higher demand. In addition, the contribution by
and October the volatility is lower than in W2 thanks to the PV is generally lower, so that PV3 profit reduction is limited.
higher photovoltaic feed-in. On the other hand, in August On the other hand, the high wind production in October had a
the volatility increases due to both energy sources. significant effect on W2 profits, which fell by 37% and 35% for
STHC and CC3P, respectively. PV3W2 profit reduction is
Optimization of thermal power plants operation lower than in August because of the lower overall RES
The maximum short-run profits (9) of two power plants contribution.
(STHC, CC3P) showing average performance within the Steam
and Combined Cycle categories are evaluated using the new Note that all profit reductions are expected to increase if
hourly pattern of electricity price (ce(t)) calculated in the plant efficiencies are lower than the considered ones, and may
previous section for different RES shares. The goal consists in lead to exclusion of older plants from the market. Note also that
foreseeing how RES generation influences the optimal the analysis took into account the variation of the real installed
operation of these traditional power plants. capacity in Germany in August, October and December 2010,
Tables 9 and 10 show the STHC and CC3P CC3P maximum and the consequent effects on the electricity price.
achievable profits and number of full load hours in the three
scenarios (PV3, W2 and PV3W2) compared to the Base one. It CONCLUSIONS
appears clearly that profit decreases similarly for both plants in The economic performance of traditional thermal power
each scenario, as explained in detail in the following. plants in Germany was analyzed in this paper, both using
historical data of the period 2007-2010 and foreseeing new
Tab. 9: Optimization results for the coal-fired power plant market scenarios characterized by additional renewable
AUGUST OCTOBER DECEMBER
electricity production in the next future.
full load full load full load 1. The significant increase of CO2 costs in the period 2007-
profit hours profit hours profit hours
PV3 -37% -16% -17% -7% -4% -1% 2010 resulted in a strong reduction of profits for all
W2 -29% -19% -37% -25% -29% -23% traditional power plants, and lead to a reverse merit order of
PV3W2 -62% -37% -50% -35% -32% -23% steam coal Rankine and natural gas fired combined power
plants. The latter replaced the former as to lowest operation
costs in spite of the stronger increase in natural gas cost with
Tab. 10: Optimization for the combined cycle power plant
respect to coal.
AUGUST OCTOBER DECEMBER
full load full load full load 2. Additional contributions to electricity production from
profit hours profit hours profit hours renewable sources demonstrated to decrease the profits from
PV3 -30% -14% -17% -9% -5% -1%
W2 -23% -23% -35% -30% -28% -24% traditional power plants dramatically (up to 68% when
PV3W2 -53% -32% -48% -33% -31% -24% photovoltaic is three times as much and wind is doubled
compared to the real installed capacities of 2010). Profit
August reduction resulted to be only slightly lower for natural gas
The most relevant profit reductions occur in August because of fired combined cycle power plants than for coal-fired steam
the combined effect of low electricity demand and significant plants.
contribution by RES. In particular, in PV3 STHC and CC3P
profits fall by 37% and 30%, respectively, due to the high solar 3. Profit reduction at point 2 may lead to the exclusion from
irradiation. In W2 profits fall similarly because of the relatively the electricity market of less performing traditional power
high load factor of wind energy. On the other hand, full load plants, with a positive effect on the average efficiency of the
hours are lower in W2 than in PV3 because of the higher price German power generation system.
volatility generated by possible wind energy production both
during the day and the night. The combined effect of PV and 4. The presented study demonstrates why investments in
wind capacities of the PV3W2 scenario results in a stronger thermal power plants are being held back nowadays in
profit reduction (up to 62% for STHC), which does not Germany due to the inclusion of electricity market rules
correspond to the sum of profit reductions in PV3 and W2. In favorable to renewable sources within the demand/supply

10 Copyright © 2012 by ASME


frame, basically not suitable to energy sources of very [11] Kim, T.S., 2004, “Comparative Analysis on the Part Load
different nature such as fossil and renewable ones. Performance of Combined Cycle Plants Considering
Design Performance and Power Control Strategy”, Energy,
5. Detailed models of both steam coal Rankine and natural gas
29, pp. 71-85.
fired combined power plants showed an off-design behavior
that can be well approximated by mixed-integer linear
[12] Reklaitis, G.V., Ravindran, A., Ragsdell, K.M., 1983,
characteristic equations.
Engineering Optimization, Methods and Applications, J.
Wiley and Sons, New York.
REFERENCES
[1] Lima, J., Jerónimo, R., Silva, T., 2008, “Impact of the
[13] Rao, S., 1996, Engineering Optimization, Theory and
Renewable Energy Generation on the Iberian Electricity
Practice, J. Wiley and Sons, New York.
Market”, Proceedings of the 5th International Conference
of European, Electricity Market EEM 2008, may 28-30,
[14] Yemin, H.Y., 2004, “Review on Methods of Generation
2009, Lisbon, Portugal.
Scheduling in Electric Power System”, Electric Power
System Research, 69, pp. 227-248
[2] Stoecker, W.F., 1989, Design of Thermal Systems,
McGraw-Hill, New York.
[15] Arroyo, J.M., and Conejo, A.J., 2004, “Modeling of Start-
up and Shut-down Power Trajectories of Thermal Units”,
[3] Boehm, R.F., 1987, Design Analysis of Thermal Systems,
IEEE Transactions on Power Systems, 19, pp. 1562-1568.
J. Wiley and Sons, New York
[16] Sun, N., and Ellersdorfer, I., 2008, “Typical Hour Based
[4] Bejan, A., Tsatsaronis, G., and Moran, M., 1996, Thermal
Modeling of the Power Generation System”, Proceedings
Design and Optimization, John Wiley & Sons, Inc., New
of the 6th International Conference of European,
York.
Electricity Market EEM 2009, may 27-29, 2009, Leuven,
Belgium.
[5] Walsh, P., and Fletcher, P., 2004, Gas Turbine
Performance, second edition, Blackwell Publishing.
[17] Carrión, M, and Arroyo, J.M., 2006, “A Computationally
Efficient Mixed-Integer Linear Formulation for the
[6] Rech, S., and Lazzaretto, A., 2010, “From Component to
Thermal Unit Commitment Problem”, IEEE Transactions
Macro Energy Systems: a Common Design and Off-
on Power Systems, 21,pp. 1371-1378.
Design Modeling Approach”, Proceedings of the ASME
2011 International Mechanical Engineering Congress &
[18] Weigt, H., 2009, “Germany’s Wind Energy: the Potential
Exposition IMECE2011, November 11-17, 2011, Denver,
for Fossil Capacity Replacement and Cost Saving”,
Colorado, USA.
Applied Energy, 86, pp. 1857-1863.
[7] Darie, G., et al., 2007, “Sliding Pressure Operation of
[19] Rosen, J., Tietze-Stoeckinger, I., and Rentz, O., 2007,
Large Conventional Steam Power Units”, Proceedings of
“Model-Based Analysis of Effects from Large-Scale Wind
the 5th IASME/WSEAS International Conference on Heat
Power Production”, Energy, 32, pp. 575-583.
Transfer, Thermal Engineering and Environment, August
25-27, Athens, Greece.
[20] Energie für Deutschland 2011, Word Energy Council,
Weltenergierat, Deutschland.
[8] Jesionek, K.J., 2001, “Method of Increasing Steam
Turbine Control Stage Efficiency”, Journal of
[21] Biernatzki, R. et al., 2006, “Modified Pricing Model for
Computational and Applied Mechanics, 2, pp. 37-43.
Energy Exchangers”, Proceedings of the 41st International
Universities Power Engineering Conference UPEC ’06,
[9] Bernero, Y.C., 2008, “Comparative Evolution of Advanced
September 6-8, Newcastle upon Tyne, UK
Coal-Based Plants”, PhD thesis at the Technical University
of Berlin.
[22] Cooke, D.H., 1985, “On Prediction of Off-Design
Multistage Turbine Pressure by Stodola’s Ellipse”.
[10] Topolski, J., and Badur, J., 2002, “Comparison of the
Transactions of the ASME, 107, pp. 595-606.
Combined Cycle Efficiency with Different Heat Recovery
Steam Generators”, Transactions of the Institute of Fluid-
[23] Glover, F., 1975, "Improved Linear Integer Programming
Flow Machinery, 111, pp. 5-16.
Formulations of Nonlinear Integer Problems",
Management Science, 22, pp. 455-460.

11 Copyright © 2012 by ASME


[24] Christidis, A., Koch, C., Pottel, L., Tsatsaronis, G., 2012, [29] Matthes, F., et al., 2011, Untersuchung im Auftrag der.
“The Contribution of Heat Storage to the Profitable Umweltstiftung WWF Deutschland, Berlin
Operation of Combined Heat and Power Plants in
Liberalized Electricity Markets”, Energy, 41, pp. 75-82. [30] Sensuß, F., Ragwitz, M., and Genoese, M., 2007, “The
Merit-Order Effect: A Detailed Analysis of the Price Effect
[25] Grimm, V., 2007, “Einbindung von Speichern für of Renewable Electricity Generation on Spot Market Price
Erneuerbare Energien in die Kraftwerkseinsatzplanung in Germany”, Working Paper Sustainability and
(Integration of Storages for the Renewable Energies in the Innovation, 7, Fraunhofer Institute Systems and Innovation
Power Plant Schedule)”, PhD Thesis at Ruhr-Universität Research.
Bochum.
[31] Nitsch, J., et al., 2010, Langfristszenarien und Strategien
[26] Wick, J., 2006, “Advanced Gas Turbine Technology für den Ausbau der erneuerbaren Energien in Deutschland
GT26” Proceeding of Presentation at the Jornada bei Berücksichtigung der Entwicklung in Europa und
Tecnológica en Madrid, October 26, Madrid, Spain. global „Leitstudie 2010“,Deutsches Zentrum für Luft- und
Raumfahrt (DLR), Stuttgart Institut für Technische
[27] Genesis-Online Datenbank, Statistisches Bundesamt Thermodynamik, Abt. Systemanalyse und
Deutschland, Einfuhr von Steinkohle: Deutschland, Technikbewertung Fraunhofer Institut für Windenergie
Monate, Ursprungsland, Ergebnis – 43511-0001. und Energiesystemtechnik (IWES), Kassel Ingenieurbüro
für neue Energien (IFNE), Teltow
[28] Bp, Statistical Review of World Energy 2011,
http://www.bp.com/sectionbodycopy.do?categoryld=7500
&contentld=7068481

12 Copyright © 2012 by ASME


ANNEX A
®
EBSILON PROFESSIONAL MODEL OF A POWER PLANT BASED ON A SUBCRITICAL RANKINE CYCLE WITH
ADVANCED CHARACTERISTICS

13 Copyright © 2012 by ASME


ANNEX B
®
EBSILON PROFESSIONAL MODEL OF A THREE PRESSURE LEVEL COMBINED CYCLE POWER PLANT

14 Copyright © 2012 by ASME

View publication stats

You might also like