Indian Institute of Technology, Delhi

----------------------------------------------------------------------------------------------------------

Pr. No.:

Thermodynamic Modelling of Advanced Power Systems using Gas

Turbine
(MED 411)
Srijan Mishra

2010ME10734

Aalekh Sharan

2010ME10641

Professor P.M.V Subbarao

(Supervisor)

Professor Prabal Talukdar

(Examiner)

Department of Mechanical Engineering

IITD
May 2014

INTRODUCTION
Gas turbine power and thermal efficiency can be augmented by overspray process which
consists of inlet fogging, wet compression and water/steam injection after compressor. In this
study the inlet fogging process is modeled based on the evaporation of droplets and
numerically solved for transient behavior of air and droplet temperatures and droplet
diameter. The wet compression process model is also solved numerically. The transient
behavior of important variables in wet compression such as droplet diameter, air and droplet
temperature, and evaporation rate is investigated. The effects of system parameters on
variables such as droplet evaporation time, compressor outlet temperature are taken up in this
study. Various schemes of steam injection such as steam injected gas turbine(STIG),
recuperative steam injected gas turbine(RSTIG), humidified air turbine(HAT) and
combination of wet compression and steam injection schemes are studied.
In the first part of the B. Tech project, we improved the calculations by using a droplet
size distribution rather than assuming one size for the inlet fogging and wet compression
process. Also, a preliminary investigation into the efficiency drop due to adding an inlet vane
was investigated. Finally, we investigated the amount of green energy that can be added to a
typical power plant by using renewable energy sources such as solar energy. This work would
be continued provisionally in the second part of the project.

NOMENCLATURE
A
droplet surface area (m2)
D
droplet diameter (m)
f
amount of water (kg water/kg dry air)
si
amount of steam injection (kg steam/ kg dry air)
T
air temperature (0C)
Ts
droplet temperature (0C)
h
specific enthalpy (kJ/kg)
cp
specific heat capacity (kJ/kg K)
wc
specific compressor input work (kJ/kg)
wt
specific turbine output work (kJ/kg)
hfg
latent heat (kJ/kg)
n
polytropic index
dw/dT
Evaporative rate (kg vap/ kg dry air/0C)
p
pressure
Q
heat flux (W/m2)
CV
calorific value (kJ/kg)
TIT
turbine inlet temperature
GREEK SYMBOLS

density (kg/m3)

isentropic index

efficiency
p
polytropic efficiency

SUBSCRIPTS
0
compressor duct inlet condition
1
compressor entry
2
compressor exit/ combustion chamber inlet
3
combustion chamber exit/ gas turbine inlet
4
gas turbine exit
ACRONYMS
far
fuel air ratio
CH4
methane
CO2
carbondioxide
H2O
water
O2
oxygen
N2
nitrogen
WET
wet compression
STIG
steam injected gas turbine
RSTIG
recuperative steam injected gas turbine
HAT
humidified air turbine
ORC
organic rankine cycle

CONCEPT
Gas turbines suffer from both decreasing output power and efficiency as the ambient
temperature increases because the air becomes less dense (which results in less mass flow rate),
and the compressor works harder as ambient temperature increases. It has been found that every
1C rise in ambient temperature reduces gas turbine output power by approximately 0.54-0.9%,
Gas turbine inlet air fog cooling is considered a simple and cost-effective method to increase
power output and often also increase thermal efficiency. Fog cooling is done by spraying microscaled water droplets into gas turbine inlet. The air through evaporation absorbs heat until the air
is saturated before entering the compressor; this is called "saturated fogging". If there are water
droplets remaining after the air flow reaching saturation at the wet bulb temperature, the
remaining droplets will enter the compressor as overspray (or high fogging), which can further
cool the compressor. Fog cooling is gaining popularity due to its low initial and maintenance
costs. Steam injection in gas turbine unit is also well proven technique that improves both the
efficiency and power of the gas turbine and also reduces harmful emissions of NOx into the
atmosphere. At present, the steam injection system for power augmentation and reduction of NOx
is offered by dominant gas turbine companies such as General Electric, Rolls-Royce etc.

Most of the power is produced from gas turbine cycle by heat recovery from exhaust gas,
but still there is a potential in the low temperature exhaust gas to produce power. Now a days,
the major thrust is to use organic rankine cycle for low temperature heat recovery. The
organic Rankine cycle technology has many possible applications, and counts more than 250
identified power plants worldwide. Among them, the possible applications are in geothermal
plant, biomass plant, solar thermal plant and in waste heat recovery system. The major
manufactures in ORC technology are Ormat, Enertime, Turboden, GMK etc.
There is very little work done in combined cycle power plant that constitutes gas turbine
cycle and organic rankine cycle. The present work here is to find the possibility of organic
rankine cycle as an alternative to the steam turbine cycle as a bottoming cycle to gas turbine
combined cycle power plant.
The complete block diagram of the present work is shown in figure-1.1. In the advanced gas
turbine cycle, all the processes in the gas turbine cycles are humidified to have higher net power
output and thermal efficiency as compared to simple gas turbine cycle. Most of the thermal
energy of exhaust gas from gas turbine is utilized in gas turbine cycle itself by using recuperator
or HRSG. The remaining exhaust energy at low temperature (200 to 350 0C in the present work)
is recovered in the organic rankine cycle (ORC) as bottoming cycle. The exhaust of the ORC heat
exchanger is cooled to water dew point temperature (55 to 65 0C in the present work), therefore
ORC system can be operated without any regenerative heat exchanger. Further, water

is recovered from the exhaust of ORC heat exchanger in the water recovery system and again
send back to gas turbine cycle for water/steam injection.

Figure-1.1 Block diagram of combined advanced gas turbine and organic rankine cycle

OBJECTIVES

Thermodynamic analysis of advanced gas turbine

cycle (inlet evaporative fog cooling/ wet
compression/ steam injected gas turbine cycle).
Thermodynamic analysis of cycle using a variable droplet size distribution

Thermodynamic
analysis of above system along with hybridization with a renewable energy

source
Incorporating
the above analysis using MATLAB code and NIST-REFPROP software for

ORC.

LITERATURE REVIEW
INLET COOLING/WET COMPRESSION
Chaker et al.[1 ] provides the results of extensive experimental and theoretical studies
conducted over several years, coupled with practical aspects learned in the implementation of
nearly 500 inlet fogging systems on gas turbines ranging in power from 5 to 250 MW. They
treats the practical aspects of fog nozzle droplet sizing, measurement and testing presenting
the information from a gas turbine fogging perspective. This paper describes the different
measurement techniques available, covers design aspects of nozzles, provides experimental
data on different nozzles and provides recommendations for a standardized nozzle testing
method for gas turbine inlet air fogging.
Santos A.P et al. [2] investigated the two cooling methods (absorption chiller and
evaporative cooling) for different inlet conditions (inlet temperature and relative humidity).
Results show that both methods improve the power output and thermal efficiency. The
evaporative cooling method is limited by ambient wet bulb temperature, representing a
suitable solution at low ambient RH inlet conditions. On the other hand, the absorption chiller
reached a larger temperature drop at different inlet conditions. When the ambient temperature
is extremely high with low relative humidity the chiller is most suitable inlet air cooling
system if exhaust energy is available otherwise evaporative cooling method is more
economical for low RH inlet condition.
Araimo L et al. [3] has done the analytical study of wet compression. The results show that
an effective intercooling can be achieved through wet compression. Such a cooling brings the
reduction of compressor work and the increase of compressor efficiency. The wet
compression work can even be lower than that of dry air isentropic compression work. The
wet compression technique has great potential to enhance the performance of the gas turbine.
Bracco S et al. [4] examines the effect of wet compression on gas turbine power plants,
particularly analyzing the influence of ambient condition on plant performance. The water
injection (2% overspray) into the compressor intake air increases the power output by 14%
and this value rises to 17% if the fogging effect is also considered. Wet compression process
causes an efficiency gain from .3% to 1.3% as a function of ambient temperature and the
injected water mass flow rate.
STEAM INJECTED GAS TURBINE CYCLE
Jonsson M et al. [5] reviews all the possible cases of humidified gas turbine. Gas turbine
with air-water mixtures as the working fluid promise high electrical efficiencies and high

specific power output. Different humidified gas turbine cycles have been proposed, for
example direct water injected cycles, steam injected cycles and evaporative cycles with
humidification towers.
Bouam A et al. [6] have studied the influence of steam injection parameters on injected
steam quantity. To maintain the same thermal efficiency (as on ISO condition) the amount of
steam injected is increased when the steam injection pressure and temperature is decreased
and vice versa with varying ambient temperature. The net effect of steam injection on thermal
efficiency and power output is on increasing trend.
Wang F.J et al. [7] studied the steam injected cycle with inlet air cooling. It is found that
STIG and IAC are well suited for retrofitting project without destroying its original integrity.
In this study, an existing simple cycle generation system was considered as the basic system
and converted into the modified system with either IAC or/and STIG features. The results
show that the system with STIG can have the best generation efficiency improved from
29.3% to 39.9% and thus the shortest payback period, while the system with both STIG and
IAC can achieve the greatest power capacity increased from 52.1 MW to 96.8 MW.
Cheng D.Y [8] introduced CLN technology as an emission control system through
improved steam injection in gas turbine cycle. They claimed that this technology will be the
only money saving control technology to be installed. In addition, it is the only proven
greenhouse gas emission reduction technology available now to meet the 20% reduction
before the targeted date of 2020. Such a retrofit retains the fast start up characteristics of the
simple cycle which is typically used as peaking units or mechanical drive applications for
compressors and pumps for the oil and gas industries.
ORGANIC RANKINE CYCLE
Saleh B et al. [9] has done thermodynamic screening of 31 pure component working fluids
for organic Rankine cycles (ORC) is given using BACKONE equation of state. The fluids are
alkanes, fluorinated alkanes, ethers and fluorinated ethers. Thermal efficiencies are presented
for cycles of different types. They used an internal heat exchanger (IHE), in case that the
vapour leaving the turbine is superheated,. The highest thermal efficiencies are obtained for
the high boiling substances with overhanging saturated vapour line in subcritical processes
with an IHE. On the other hand, a pinch analysis for the heat transfer from the heat carrier
with maximum temperature of 120 0C to the working fluid shows that the largest amount of
heat can be transferred to a supercritical fluid and the least to a high-boiling subcritical fluid.
Gao H et al. [10], has done the performance analysis of a supercritical organic Rankine cycle
system driven by exhaust heat using 18 organic working fluids. Several parameters, such as the
net power output, exergy efficiency, expander size parameter (SP), and heat exchanger

requirement of evaporator and the condenser, were used to evaluate the performance of this
recovery cycle and screen the working fluids.
Farnandez F.J et al. [11] used the SpanWagner equation of state for siloxanes, used as
working fluids in high-temperature organic Rankine cycles, is applied in a mathematical
model to solve cycles under several working conditions. The proposed scheme includes a
thermo-oil intermediate heat circuit between the heat source and the organic Rankine cycle to
maintain the thermal stability of organic fluid even if the source temperature is greater than
the thermal stability temperature of the working fluid.The cycle includes an internal heat
exchanger (regenerative cycle), although a non-regenerative scheme is also solved. Simple
linear (MM, MDM) siloxanes in saturated regenerative schemes show good efficiencies and
ensure thermal stability of the working fluid.
Chacartegui R et al. [12] studied low temperature organic rankine cycles as bottoming
cycle in medum and large scale combined cycle power plants. The aim of this work is to use
the alternative cycles with high efficiency heavy duty gas turbine engines with lower exhaust
temperature than in conventional combined cycle gas turbines. Competitive results have been
obtained for toluene and cyclohexane ORC combined cycles, with reasonably high global
efficiencies.
WORK DONE AT IIT DELHI
ADVANCED GAS TURBINE CYCLE
Sairam A [19] has done detailed thermodynamic analysis of wet compression. They
assume the evaporation rate as constant and study the effect of varying evaporation rate on
wet compression process. The adverse effect of water injection in compressor was also
studied. With water injection, polytropic efficiency of the compressor deteriorates compared
to that of dry compression. Steam injected gas turbine is also discussed. They have
emphasized wet compression with steam-injected gas turbine will produce more power output
at hugh thermal efficiency.
ORGANIC RANKINE CYCLE
Hemadri V.B [20] has selected Hexamethyldisiloxane (MM) a linear siloxane as the working
fluid for ORC in their work. The thermal efficiencies of both cycles: simple ORC and
superheated ORC schemes without regeneration are compared with steam cycle for the same
turbine inlet and condenser conditions. The opportunity for regeneration at the turbine exit is
observed and studied. The advantage of ORC with regeneration is emphasized depending upon
the degree of superheat of the vapor at the turbine exit. They have used the house developed
Gibbs energy based property model using generalized PengRobinson (GPR) cubic equation is
used to predict the thermodynamic properties. The results obtained from GPR are compared with
National Institute of Standards and Technology (NIST) developed REFPROP9.0.

Siddiqui, P[21] has combined all of the above research inputs along with their advantages and
disadvantages to create an integrated advanced combine cycle turbine which powers a
secondary ORC turbine. Industry standard software such as REFPROP has been used to
develop a mathematical model on MATLAB which can predict various flow parameters.

COMPRESSOR INLET COOLING - FOGGING

Compressor inlet fogging is a method of cooling intake air by injecting demineralized
water in the duct through the special atomizing nozzles. The evaporation of droplets results in
a cooling of air and a consequent increase of mass flow rate. When the mass fraction of
injected water is less than the water required for complete evaporation within inlet duct then
the process is considered as low fogging. High fogging generally corresponds to larger liquid
mass fractions, in which case droplet evaporation is not completed within the duct but is
extended into the compressor for wet compression. The effect of cooling of air due to droplet
evaporation is the reduction of compression work.
For a given inlet duct length, several factors such as water injection rate, droplet size
distribution, air velocity and intake air condition affect whether injected water droplets
completely evaporate or not. In this study the inlet fogging process is modeled based on the
evaporation of droplets with the help of heat and mass transfer correlations and
thermodynamic relations. The model is solved numerically to evaluate transient behavior of
droplet diameter, air and droplet temperature for different initial diameter of droplets.

MODELING OF COMPRESSOR INLET FOGGING PROCESS

Figure-4.1 Gas Turbine system with inlet fogging process

In this modeling, air is assumed to enter the inlet duct at temperature T 0, pressure P0 and
relative humidity RH0. Water is sprayed at very high pressure (7 to 14 MPa) through low vlume
nozzle, in the form of fine droplets into the inlet air stream, with initial droplet diameter D0 at a
ratio of f0 (kg/kg of dry air). It is assumed that the pressure in the inlet duct is constant.

The initial mass of water vapor per unit mass of dry air (specific humidity) can be expressed

as follows:
0.622 0

w0 =
where,

0 (t0) = vapor pressure.

In this study, water droplets are assumed to be spherical and monodisperse and they do not
interact with each other during the process. Then, the number of droplets per kg of dry air, N,
can be assumed to be constant after injection till the end of evaporation. The number N (per
unit kg of dry air) can
be expressed as
6 0

03

N = constant =
Where, w is the density of water and N is assumed to be constant after injection.

When the droplet diameter is D, the total surface area of the liquid droplets per unit mass of
dry air can be expressed as
2

A=

The conservation of water mass requires that

f + w = constant
where, f = amount of liquid water per unit kg of dry air w
= amount of water vapor per unit kg of dry air

The specific enthalpy of the mixture of dry air, water vapor and liquid water during
compression is expressed as follows
h = ha (T) + whv(T) + f hl(Ts) = constant .. (1) (assuming adiabatic process) where, h a, hv, hl are the
specific enthalpy of air, water vapor and liquid water respectively. Also, liquid water depletion rate = water droplet
evaporation rate
= .

(2)

Rate of increase in the internal energy of water droplets = rate of sensible heat flux rate of
latent heat flux required for vaporization
fcpw dTdts = A(Qs QL ) = A(Qs hfg I) .. (3)

Where, I is the vapor mass flux away from the droplets, QL is the latent heat flux due to
droplet evaporation, hfg is the specific enthalpy of vaporization and Qs is the sensible heat
flux due to diffusion or convection.
We have 3 equations and 3 unknowns (f, T, Ts), that can be solved numerically.
The Stokes model [13] enables formulation of the heat and mass fluxes as follows:
Qs = kNuD (T Ts)

I = Sh Dv (ps pv ) DRv Ts T

Where, Dv is the mass diffusion coefficient of water vapor in air, Rv is the gas constant of
water vapor, T and Ts are the temperatures of humid air and water droplets respectively. p s is
the saturation pressure of water at Ts. pv is the vapor pressure of air- vapor mixture.
Nusselt number, Nu = 2 + .6 .25 .33 2 Sherwood number, Sh = 2 + .6 .25 .33 2

Since, the droplet readily attains the velocity of air on injection in the duct. Therefore, heat
and mass transfer is largely affected by natural convection. Nusselt number and Sherwood
number are taken as constant in this analysis on the basis of small relative velocities for the
small droplet diameter.

FLOW CHART FOR THE MODELING OF COMPRESSOR INLET FOGGING PROCESS

TRANSIENT BEHAVIOR STUDY OF COMPRESSOR INLET FOGGING

The initial state is taken as p0 = 1.01325 bar, T0 = 35 0C, RH0= 60 %. Droplet temperature
is taken to be equal to the ambient temperature at T0. Air and water droplet temperature and
droplet diameter are computed at any given time, for water droplet initial diameter of 20 m
for different water injection ratio ( ratio of water injection rate and critical injection rate ie.
f/fc = 0.8, 0.9, 1, 1.1, 1.2 ) at the inlet duct. f/fc < 1, f/fc = 1 and f/fc > 1 corresponds to low
fogging, critical fogging, high fogging respectively. Critical fogging here signifies the water
injection rate for given ambient condition at which the temperature at the compressor inlet is
the saturation temperature. Low fogging means that the injection rate is lower than critical
injection rate. High fogging means that the injection rate is higher than critical injection rate.
It is inferred from figure-4.2 that the droplet size gradually decreases with time. In case of
low fogging (f < fc), the diameter diminishes to zero after finite time. The dry out time
increases with the amount of initial water injection and grows to infinity as f approaches fc. In
case of high fogging (f > fc), the droplets do not evaporate completely but tend to have small
final diameter. The remaining droplets enter into the compressor of gas turbine system and
then wet compression begins there.
The humid air temperature decreases with time as indicated in Figure-4.3, and the only
way to approach the wet bulb temperature (19 0C) asymptotically is to exceed the critical
injection ratio. In all cases, a rapid temperature decrease is followed by a much slower rate of
decrease, which suggests an optimal duct length for complete evaporation. If complete
evaporation is desired, then the duct must be long enough to accommodate the residence
times for low fogging.

Figure-4.2 Transient behavior of water droplet diameter

Figure-4.3 Transient behavior of humid air temperature

Figure-4.4 Variation of humid air temperature with the amount of liquid water present in
the duct

The humid air temperature is an almost linearly decreasing function of liquid water
depletion as shown in figure-4.4. The slope of temperature change gets steeper as the water
injection ratio f/fc increases. For the low fogging case, the droplets evaporate completely and
outlet humid air temperature is lower for higher f/fc. In contrast there is no complete droplet
evaporation for the high fogging case and the final humid air temperature is almost constant
and independent of water injection ratio.

Issues and Improvements

The issue with this approach is that by taking a uniform droplet size distribution, the end result
can be drastically different than if droplets of various diameters were taken. In this regard, we
were able to use a modified version of the Nukiyama-Tanasawa equation which took only one
parameter(Sauter Mean Diameter) as input in order to create 10 buckets of different drop sizes.
The same simulation was performed for all the buckets in order to get drastically different

results.

PDF of number
of droplets

PDF of volume
of droplets

1Nukiyama-Tanasawa distribution used

WET COMPRESSION PROCESS

MODELING OF THE WET COMPRESSION PROCESS
According to Gibbs equation,
Tds = dh dp

For an ideal wet compression, it is assumed that the evaporative heat equals to the reversible
heat transfer from air to liquid water,
where, hfg = Latent heat of vaporization
-hfgdw = Tds

- hfg dw = dh dp
- hfg dw = dT 1R RT dpp
dpp

= dTT [ 1+ hRfg* dwdT ]

Assuming evaporative rate of water droplets varies linearly with temperature, i.e. dwdT = constant, the isentropic relation is obtained as, p vk = C
k

= [

k1

hfg

dw

dT

If p is the polytropic efficiency of the compressor, it can be shown that

T2
p2
(n1)n
T1 = [ p1 ]

Where,

= p [

n1

hfg

+
1

dw

*
R

]
dT

This equation shows that the increase of evaporation rate decreases the polytropic index
(k) of wet compression from isentropic process towards the isothermal process (k=1) which
results in reduction of compressor power. This can be seen in the p-v diagram in figure-5.1
[17] as the wet compression process 1- 2 is less steeper than dry compression (1-2).
This shows that the work input (i.e. vdp) in infinitesimal stage is lesser in the case of wet
compression process.

Figure-5.1 Fog/overspray cooling process in the compressor (for air only)

Where, Process (1-2) = dry compression when the compressor inlet is at ISO condition.
(1-2) = dry compression when compressor inlet temperature is higher than ISO
condition
(1-1) = inlet fog cooling
(1-2) = dry compression for low fogging and critical fogging cases.
(1-2) = wet compression for high fogging case
In order to evaluate the effect of evaporation rate of water droplet dw/dT (kg water
vapor per unit kg of dry air/ 0C), the study of wet compression process is done by numerical
analysis of heat and mass transfer model as was done in chapter-3 for inlet fogging process.
The only difference in the modeling of wet compression analysis is that there is continuous
pressure variation and enthalpy variation through out wet compression process.
The pressure variation can be expressed by the parameter of compression rate, C defined
by
1 dp

C=

[14]
p dt

It is assumed that C has a constant value that depends on compressor speed, which
implies an exponential form for pressure variation w.r.t time. This relation provides an
acceptable pressure distribution inside a compressor. An average value of C may be used to
characterize actual compressors. Then the pressure varies as
p = p1 exp(C t)

The enthalpy variation can be expressed by using polytropic efficiency of compressor as

follows
cp = vdpdh

Compressor input work per unit kg of dry air:

wc, wet = (ha2 ha1) + w1 (hv2 hv1) + (w2 w1) (hv2 hv1)
When the amount of water completely evaporated, then dry compression begins, and
temperature and work input can be determined for this as usual.
wc, dry = (ha3 ha2) + w2 (hv3 hv2)
wc = wc, wet + wc, dry
where, stae-1 = property at the start of compression
stae-2 = property at the end of evaporation and
state-3 = property at the end of compression

TRANSIENT BEHAVIOR STUDY OF WET COMPRESSION

The behavior of the non-equilibrium wet-compression process depends on operational
parameters such as the water injection ratio f1, the pressure ratio rp, the initial diameter of the
droplets D1, the compression rate C, the polytropic compression efficiency and the ambient
conditions. Parameters are computed at inlet pressure p1 of 1 atm, inlet temperature T1 of
150C, inlet relative humidity RH1 of 60% (ISO condition), compression rate C = 200 s-1, and
polytropic efficiency = 91%.
Transient behavior study w.r.t various droplet diameter:
Figure-5.2 shows the transient behavior of the mass of liquid droplets with time for water
injection ratio of f1= 5%. It can be seen from the figure that once water droplets are injected
at the compressor inlet, the mass of the liquid droplets decreases monotonically as
evaporation continues until the evaporation is completed.
The droplet temperature increases obeying the steady energy balance, after an abrupt initial
transient behavior, which can be seen in Figure-5.3. Figure-5.4 shows that the humid air
temperature increases with time, as pressure increases. The slope increases with droplet
diameter, for the same injection ratio. After complete evaporation, the humid air temperature
increases at a higher rate than during evaporation. Yet, the temperature is lower than that for
dry compression. It can be seen from Figure-5.5 that the temperature difference between the
humid air and droplets decreases as the diameter of the droplets decreases.
Improvements:
We used the Nukiyama-Tanasawa model again to evaluate the drop sizes in 10 buckets. The
resulting graph was an amalgamation of the graphs calculated by P. Siddiqui.

Figure-5.2 Liquid water fraction transient behavior for various droplet diameter (f1 =5 %)

Figure-5.3 Water droplet temperature transient behavior for various droplet diameter (f1
=5 %)

Figure-5.4 Humid air temp transient behavior for various droplet diameter (f1 =5 %)

Figure-5.5 Temperature difference between humid air and water droplet for various droplet
diameter (f1 =5 %)
Transient behaviors w.r.t various water injection rate:
Figure-5.6 shows the transient behavior of the mass of liquid droplets with time for
various water injection rate. The evaporation completion time increases as the water injection
ratio increases. Hence, liquid water may remain at the compressor outlet that can be observed
for high water injection ratios.
Figure-5.7 shows that the humid air temperature increases at a lesser rate during
evaporation and faster after completion of evaporation. The final humid air temperature is
lower for higher injection ratios. Figure-5.8 shows that humid air and droplet temperature
difference increases with time and its magnitude is higher for lower injection ratio values.
This graph then reflects the fact that, whereas humid air temperatures are a strong function of
the injection ratio, liquid temperatures depend only weakly on it as shown in figure-5.9.

Figure-5.6 Liquid water fraction transient behavior for various water injection ratio (D1
=10 micron)

Figure-5.7 Humid air temp transient behavior for various water injection ratio (D1
=10 micron)

Figure-5.8 Temperature difference between humid air and water droplet for various water
injection ratio (D1 =10 micron)

Figure-5.9 water droplet temperature transient behavior for various water injection ratio
(D1 =10 micron)

Compressor outlet temperature and input work:

Figure-5.10 shows the variation of compressor outlet temperature and compressor input
work as a function of compression ratio, for various values of the water injection ratio. It is
evident that compressor outlet temperature increases with compression ratio, however, the
increase rate is lower than for dry compression. When the compression ratio is greater than
the value at which injected water droplets are completely evaporated, the compressor outlet
temperature approaches the same value for a given water injection ratio.

Figure-5.10 compressor exit temperature wrt pressure ratio (D1 =10 micron)
Figure-5.11 shows the variation of compression work as a function of compression ratio,
with injection ratio as parameter. It can be seen from the figure that the compression work
increases with compression ratio, its rate of increase can be lowered by evaporation.

Figure-5.11 compressor input work wrt pressure ratio (D1 =10 micron)

Synthesis of Inlet Fogging and Wet Compression

The two methods used here are synthesized in order to create a complete model. By carefully
adjusting the parameters and the water injection rate, we can arrive at a critical ratio of water
droplets remaining after the end of the wet compression process. Ideally, all the water should
evaporate, but in practice, the larger droplets will remain as it will take much longer to
evaporate them due to a smaller surface area to volume ratio.

Analysis of potential solar energy available at typical power plants

An empirical study was done in order to determine the typical solar energy available at power
plants in India. To do this, a combination of Google maps and an online tool for area
calculation was used. This was combined with the data on solar insolation by latitude to give
an indication of the potential energy augmentation possible by various technologies.

Name of plant
Anpara Thermal Power Station
Faridabad Thermal Power Plant
Guru Nanak Dev Thermal Plant
Kolaghat Power Plant
Korba Power Plant
Kota Power Plant
Narora Power Plant
Panipat Thermal Power Plant
Raichur Power Plant
Rajiv Gandhi Power Plant

Total
Conventional
Total
recoverable Recoverable
annualized power Area(km^2)
area
energy
1385.5
4.916
4.907
215.908
365.5
0.134
0.124
5.456
374
2.38
2.308
101.552
1071
1.728
1.712
75.328
2210
2.5
2.487
109.428
1054
1.715
1.685
74.14
374
5.484
5.211
229.284
374
4.107
3.92
172.48
1462
4.485
4.199
184.756
510
2.349
2.015
88.66

Table 1:Recoverable solar energy of various Indian power plants

The recoverable area was found by subtracting the non recoverable areas from the total power
plant area. For example, the area occupied by cooling towers cannot be used for solar cogeneration.

Energy mix at various powerplants using solar thermal

3000

2500

2000

1500

1000

500

0
Anpara Faridabad Guru Nanak Kolaghat
Korba
Kota Power Narora
Panipat
Raichur Rajiv Gandhi
Thermal Thermal
Dev
Power Plant Power Plant
Plant Power Plant Thermal Power Plant Power Plant
Power Power Plant Thermal
Power Plant
Station
Plant
Non-renewable

Renewable

2Energy mix at various power plants

On an average, it was found that 13.6% of existing capacity can be either augmented or replaced
by solar power.

Project Progress

The simulation of the system was completed using the Nukiyama Tanasawa equation.
Optimization was done to ensure adequate number of water droplets remained for the wet
compression stage.
Also, progress was made into the second phase of the project scheduled for next semester,
i.e., investigating the renewable energy potential of existing conventional power plants.

The methodology for an optimized integration of RE technologies into Indias power plant portfolio is
shown at Figure 1.

In a first step, within a resource and site assessment for utility-scale PV, onshore wind power and
Concentrating Solar Power (CSP), technology specific hot spots for the most promising RE
technologies in India are identified using a Geographic Information System (GIS). Thereby, the
technology specific hot spots are identified by applying a site-ranking process with respect to the
availability of the primary energy source (wind speed, solar radiation) as well as to the distance to
demand centers and infrastructure (substations, transmission lines, major streets, etc.). At each site,
information about the hourly availability of the respective resource, the hourly ambient temperature
and the maximal installable capacity are assessed. With this information representative hourly
generation profiles of each technology at the respective sites are calculated. This information serves,
together with detailed information about power demand and supply in Jordan and related technoeconomic data, as input for the capacity expansion and replacement optimization model ReMixMENA.
The results of the capacity expansion optimization (base case scenario) are shown in Figure 2. As
can be observed, RE technologies are already competitive in the short term in Jordan. Until the end of
the optimization time-frame in the year 2022 about 2200 MW of CSP, 2100 MW of utility-scale PV,
1000 MW of onshore wind power and 1800 MW of conventional capacity is installed. The share of
power generation by RE technologies is increased from about 0.3 % in 2012 to more than 47 % in
2022, whereby Jordan becomes significantly more independent from fossil fuels and the related risk
of cost escalation. In Figure 4, the development of the average specific generation costs of the base
case scenario is compared with a fossil fuel scenario (investments only in
conventionaltechnologies) and a fluctuating RE scenario (no investments in CSP units). As can be
observed, the optimization shows that a well balanced mix of all available RE technologies and some
conventional generation technologies according to the base case scenario is the least cost option for

Jordan to meet future power demand.

The future challenges for Jordans electricity authorities have become obvious in the last years.
Reliable electricity supply at reasonable prices is a key factor for India in order to ensure the future
economic development. The paper at hand describes how a well balanced mix of renewable and
conventional power generation technologies can ensure to keep up with Jordans strong increasing

electricity demand and to get simultaneously more independent from fossil fuel imports whereby the
escalation of future electricity generation costs can be significantly absorbed. It shows that CSP,
utility-scale PV and onshore wind power, due to the excellent solar and wind resources, are already
competitive today in certain load segments of Jordans electricity sector. Each of these technologies
has characteristics which determine the application within the electricity system. PV and wind power
can be used as cheap fossil fuel saver. The CSP technology, as a dispatchable and renewable power
generation technology, can deliver strongly required firm and flexible power generation capacity and due to its constant generation costs - has a significant advantage over fossil-fuel based technologies.
However, even though the analysis has shown that RE technologies are competitive in India in the
short-term and the large-scale introduction is favorable for economic reasons, suitable market
conditions still have to be implemented in order to trigger investments in renewable power generation
projects. Providing security about future revenues is the easiest way to attract private investors and to
bring down generation costs of RE technologies. One possibility is the introduction of technology
specific international insured long-term power purchased agreements whereby project risks can be
brought down to an AAA level.
Several models have been developed to identify the optimal combinations of renewable and
conventional resources on a large scale. Short et al. (2003) divide the United States into 356 wind
regions, and model the cost-e cient installations and operation of wind farms and conventional
generators from 2000 through 2050. DeCarolis and Keith (2006) develop an optimization model for
one investment period in 2020 based on 5 years of hourly wind and load data. Considering the
assumed costs of wind turbines, the simulation indicates that supplying 50 % of the electricity demand
by wind power adds about 1-2 ct/kWh to the costs of electricity generation. Neuho et al. (2008) divide
the United Kingdom into 7 regions and optimize investments and dispatch choices for new and
existing natural gas, coal and wind generators during four 5-year investment periods. The SWITCH
model at the University of California, Berkeley (Fripp, 2008) concentrates on California and
optimizes the combination of more than 229 wind, 464 solar sites and con-ventional resources
considering investment and operational costs. Heide et al. (2010) model the optimal mix of wind and
PV capacities for Europe by minimizing needed storage capacities subject to the constraint that all
renewable energy is used (independent of total system costs). In case of supplying 100 % electricity
by wind and solar technologies, the optimal mix is found to be 55 % wind and 45 % solar power
generation. The DIMENSION model of the Institute of Energy Economics at the University of
Cologne (EWI, 2011) simulates in 5-year time steps the cost-e cient European capacity development
and dispatch for twelve typical days of conventional, renewable and storage technologies until 2050.
Di erent regional conditions for RES-E capacites are considered by modeling 47 wind onshore, 42
wind o shore and 38 solar regions. Due to modeling deterministic feed-in structures and average full
load hours of wind and solar technologies, all of these models neglect the uncertainty about the hourly
availability of renewable energy.
Methodologies incorporating uncertainty in optimization models were developed by Dantzig (1955).
They were applied to electricity generation planning problems to analyze the impact of demand
uncertainty for the rst time in the 1980s (Murphy et al., 1982; Modiano, 1987). A broad overview of
di erent stochastic modeling approaches for electricity markets is given in Most and Keles (2010).

The economic value of wind power, taking into account the volatility of wind velocity, was analyzed
by Beenstock (1995). The method is based on the intuition that one can immunize the output of a
wind turbine against uctuations in wind speed by investing in back-up capacities and the costs of
necessary back-up investments may be regarded as the costs of wind volatility. Papaefthymiou et al.
(2006) present a Monte-Carlo simulation technique to model the extremes of stochastic wind
generation in power systems by sampling wind turbines with similar generation patterns. Swider and
Weber (2006) apply a stochastic fundamental electricity market
model to estimate the integration costs of wind due to the changed system operation and investments
in Germany. The simulation indicates that the value of uctuating renewables is overestimated applying
a static, deterministic model. In particular, investment planning under uncertainty considering power
plant outages and uctuating renewable feed-in was analyzed in Sun et al. (2008). By applying a
stochastic mixed-integer optimization model for power plant investment planning to the German
electricity market, Sun et al. (2008) show how ignoring short term uncertainties signi cantly
undervalues the needed operational exibility and can result in insu cient investments. However, in
these models the deployment of RES-E capacities is not part of the optimization problem and
therefore the optimal mix of conventional, nuclear, storage and renewable technologies in high RES-E
scenarios was not determined.
In this paper, we present a stochastic investment and dispatch optimization model for electricity
markets that accounts for the uncertain feed-in of wind and solar technologies to determine the
optimal mix of conventional, renewable and storage capacities for di erent European RES-E targets.
To our knowledge, a stochastic electricity market model with as much detail concerning the di erent
local RES-E conditions and the uncertain feed-in of uctuating renewables has not appeared before.
The di erence between the stochastic model results and the deterministic solution based on averages in
wind speeds and solar radiation can be interpreted as the impact of the stochastic availability of wind
and solar power.
3. Generation of combined wind and solar feed-in structures
Wind and solar technologies are meant to produce a large share of the future electricity demand.
However, the availability of these technologies depends on local weather conditions and therefore
weather character-istics must be considered when optimizing the future electricity mix. Regional
weather characteristics lead to di erent local RES-E conditions throughout Europe (higher solar
radiation in Southern Europe and stronger winds in Northern Europe), to stochastic amounts of yearly
generated electricity of wind and solar sites as well as to positive or negative correlations between the
availabilty in di erent regions or between technologies. In this section, we describe the characteristics
of wind speeds and solar radiation in Europe (subsection 3.1), a bootstrap approach to create
consistent regional wind and solar feed-in structures and a heuristic to select representative feed-in
structures as input parameters for the stochastic optimization model (subsection 3.2).
3.1. Empirical data for wind speeds and solar radiation in Europe

The description of wind speed (subsection 3.1.1) and solar radiation (subsection 3.1.2) characteristics
for di erent regions throughout Europe is based on hourly wind speed and solar radiation data from
EuroWind for the years 2006-2010 and includes an analysis of the regional correlations between wind
speeds (and solar radiation) in Europe as well as the correlation between wind and solar power
(subsection 3.1.3). The hourly wind speed data in 30 meters above ground and solar radiation for 64
European regions the years 2006-2010 provide a deep insight of the characteristics of regional wind
speed and solar radiation in Europe as well as the correlation between wind speed and solar
radiation.5 In the following the di erent conditions throughout Europe are discussed for some of the
selected regions. The numerical data for all regions can be found in Appendix B.
3.1.1. Characteristics of wind speeds
Wind speed distributions re ect that in most regions strong winds are rare and that moderate winds
occur most often. Due to seasonal characteristics the average wind speed is usually higher in winter
and autumn as in the summer months. Table 2 shows summarizing statistics for some of the selected
wind regions in Europe. As wind speeds are usually higher in Northern Europe, the average wind
speed in 30 meters was 6.74 m/s in Northern Ireland compared to 3.59 m/s in Southern Italy for the
years 2006-2010. Higher wind speeds often result in a higher variance as can be seen by comparing
the variance of the wind speed in the Southern part of the Iberian Peninsula (9.02) and o shore wind in
the United Kingdom (18.81). Due to generally short distances between European regions the same
general weather situations occur. Hence, the hourly wind speeds in Europe are to some extent
correlated. Table 3 shows the Pearson correlation factors for some of the selected wind regions in
Europe. This sample shows that closer regions have a stronger correlation, e.g. 0.587 between on- and
o shore wind in the United Kingdom. However, some wind regions in Europe are not or negatively
correlated (e.g. United Kingdom and Iberian Peninsula).
Table 2: Summarizing statistics for some of the selected wind regions
UK-W (on)

IB-S (on)

Mean [m/s] 6.74

- summer [m/s]
- winter [m/s] 7.65
Median [m/s] 6.28
Variance
10.48
10%-Quantil 2.97
90%-Quantil 11.15

4.80
5.95
5.03
4.12
9.02
1.73
8.90

DE-C (on)

PL-N (on)

IT-S (on)

4.89
4.40
5.47
4.54
5.51
2.18
8.13

3.59
5.49
3.67
3.10
4.34
1.42
6.48

5.03
7.45 4.52
5.30
4.27
10.23
1.71
9.60

6.33
4.38
7.22
5.92
9.24
2.80
10.36

8.82
3.44
10.26
8.28
18.81
3.55
14.85

UK-N (o )

IB-W (o )

Meteorological data for 242 measure stations of the German Weather Service for the years 2000-2010
and the European solar radiation from Satel-Light for the years 1996-2000 con rms the listed
characteristics in the dataset from EuroWind.

Table 3: Correlation matrix for some of the selected wind regions (full table in the Appendix B)
UK-W (on)

IB-S (on)

DE-C (on)

PL-N (on)

IT-S (on)

UK-W (on)
IB-S (on)
DE-C (on)
PL-N (on)
IT-S (on)

1
-0.026
0.204
0.143
0.085

UK-N (o )
IB-W (o )
IT-W (o )

0.587 -0.053 0.298 0.178 -0.002 1

-0.025 0.922 -0.024 -0.006 0.239 -0.039 1
0.027 0.365 0.096 0.003 0.327 0.028 0.303

UK-N (o )

IB-W (o )

1
-0.031 1
-0.014 0.289 1
0.137 0.171 0.029 1

The values in Table 2 and Table 3 represent the average of several years. However, as weather
situations di er between years, the yearly average wind speed varies as well. Table 4 depicts the yearly
average wind speed for the years 2006 to 2010. The average wind speed in the United Kingdom in
2008 was signi cantly higher with 7.26 m/s than the 5.93 m/s in 2010. Even considering the small
dataset, the di erence of more than 1 m/s represents about 20 % of the average over the ve years.
Similar to the yearly average wind speed, the correlation between wind regions di ers as well. The
Pearson correlation factor for wind in the United Kingdom of 0.58 in 2006 indicates a rather strong
correlation but with 0.45 in 2010 the correlation can also be lower. Naturally, data for ve years does
not represent the long term average of wind speeds as it does not capture the variance between years
su cienctly.
Table 4: Di erence between wind years: 2006-2010
UK-W (on)

IB-S (on)

DE-C (on)

PL-N (on)

Mean [m/s]
2006 6.90

4.49

4.86

6.07

3.49

8.80

4.81

2007
2008

4.72
4.94

5.35
5.08

6.74
6.66

3.50
3.63

9.04
9.54

4.95
5.19

6.73
7.26

IT-S (on)

UK-N (o )

IB-W (o )

2009
2010

6.89
5.93

4.75
5.11

4.81
4.34

6.15
6.03

3.69
3.61

8.97
7.74

4.97
5.26

Based on the described wind characteristics three aspects in ucence the optimal electricity mix: First,
from a system perspective it might be cost-e cient to focus on the best European sites i.e. with the
highest full load hours on average. The data suggests that on average more than twice as much
electricity can be produced from the same turbine in Ireland than in Italy. As installation costs are
similar over Europe, levelized electricity costs for wind power are about 50 percent lower in Northern
Europe as in Southern Europe at relatively similar conditions. Second, in particular in electricity
systems with a high share of uctuating RES-E generation a distribution of wind turbines might be
cost-e cient as the hourly European-wide total power generation from wind turbines would be more
stable. Third, the optimal electricity mix has to consider an uncertainty about the yearly availabilty of
wind power - resulting from high as well as low wind years. Hence, there should exist an optimum
between focusing on the best sites and a distribution throughout Europe.
3.1.2. Characteristics of solar radiation
Global radiation depends on the location, daytime, season and local weather conditions. Hence, the
yearly radiation in Southern Europe is higher than in Northern Europe and the average solar radiation
is generally higher in summer than winter. The times of sunrise and sunset also depend on the season
and hence the duration of daily solar radiation varies throughout the year. Regional weather
conditions such as cloudiness or wind signi cantly in uence the solar radiation. Table 5 shows
summarizing statistics for some of the analyzed solar regions in Europe. Due to the same general
weather conditions in Europe, solar radiation in di erent European regions is correlated on an hourly
basis. Table 6 shows the Pearson correlation factors for some of the selected solar regions in Europe
(only daytime hours). Due to the distinguished solar structure with a peak at midday, the Pearson
factors are rather high. This sample shows that some regions have a stronger correlation, e.g. 0.730
between Southern France and Southern Italy compared to 0.643 between Poland and the United
Kingdom.
Table 5: Summarizing statistics for some of the selected solar regions
UK-C IB-SFR-S

DE-C SK-S PL-N

IT-S

Mean [W/m2] 139

- summer [W/m2]
- winter [W/m2]
Maximum [W/m2]

228
231
75
953

138
233
70
909

191
314
172
1,021

138
283
130
997

152
247
61
834

214
250
81
886

Variance
44,884 88,594 68,087 44,537 43,124 48,356 75,138
90%-Quantil 490
746
575
496
497
534
690

309
150
976

Table 6: Correlation matrix for some of the selected solar regions - daytime (full table in Appendix B)
UK-C IB-S

FR-S DE-C SK-S PL-N IT-S

UK-C
IB-S
FR-S
DE-C
SK-S
PL-N
IT-S

1
0.783
0.654
0.646
0.584
0.670

1
0.709
0.717
0.707
0.715
0.643
0.653

1
0.688
0.713
0.603
0.730

1
0.763 1
0.714 0.746 1
0.683 0.728 0.703 1

The values in Table 5 and Table 6 represent the average of several years. However, the yearly average
solar radiation varies between the years. Table 7 depicts the yearly average solar radiation for the
years 2007 to 2010. Average solar radiation of 222 W/m2 in Italy in 2008 was signi cantly higher than
the 206 W/m2 in 2010. Even considering the small dataset, the di erence of more than 16 W/m2
represents about 7 % of the average over the four years. Similar to the yearly average solar radiation,
the correlation between solar regions di ers as well. The Pearson correlation factor between the hourly
solar radiation in Southern France and Southern Italy of 0.86 in 2008 indicates a strong correlation but
with 0.80 in 2007 the correlation can also be lower in a speci c year. Naturally, data for four years
does not represent the long term average of solar radiation as it does not capture the variance between
years.
Table 7: Di erence between solar years: 2007-2010
UK-C IB-S

FR-S DE-C SK-S PL-N IT-S

Mean [W/m2]
2007 134
2008 136
2009 143
2010 144

228
231
231
223

195
185
196
190

133
141
137
143

135
141
141
133

154
149
156
149

213
222
213
206

The optimal regional allocation of solar technologies follows the same concept as for wind turbines.
Due to better conditions solar technologies might be cost-e cient in Southern rather than in Northern
Europe. However, a large deployment of solar technologies in one region might lead to a very
unbalanced availability of solar power in the system. A regional concentration might also need signi
cant grid extensions from solar sites to large load centers.
3.1.3. Correlation of wind speeds and solar radiation
Solar radiation and wind speeds are in uenced by similar local weather characteristics such as air
pressure, sunshine, degree of cloudiness or rain. As higher wind speeds usually occur when the sky is
cloudy and sunshine is low, wind speed and solar radiation are to some extent negatively correlated.
Table 8 shows the correlation factors between wind speed and solar radiation for the years 2006-2010
at daytime. The data re ects that solar radiation and wind speed within the same region are negatively
correlated with a Pearson correlation factor between -0.004 in Iberian Peninsula (north) and -0.231 in
the United Kingdom (central).

Table 8: Correlation matrix of wind and solar radiation for some selected regions - daytime

UK-C IB-N IB-S

Solar UK-C
-0.230
IB-N -0.176
IB-S -0.158
FR-S -0.164
DE-C -0.209
PL-N -0.195
CZ-C -0.196
IT-N -0.189

Wind
FR-S DE-C PL-N CZ-C IT-N
-0.053 -0.187 -0.195 -0.098 -0.137 -0.008 0.065

-0.045
-0.057
-0.107
-0.070
-0.105
-0.086
-0.139

-0.200
-0.140
-0.192
-0.211
-0.182
-0.195
-0.219

-0.163
-0.096
-0.231
-0.232
-0.190
-0.191
-0.248

-0.043
0.018
-0.040
-0.228
-0.124
-0.184
-0.102

-0.090
-0.093
-0.076
-0.150
-0.141
-0.159
-0.104

0.013
0.045
0.026
-0.148
-0.156
-0.198
-0.069

0.069
0.043
0.026
0.011
-0.032
-0.004
-0.147

However, the extent of the negative correlation between the availability of wind and solar power di ers
between years. Table 9 depicts the di erent correlation factors for hourly wind speed and solar
radiation for the years 2007 to 2010. As can be seen for the example of Poland the Pearson correlation
factors vary between -0.077 (2009) and -0.188 (2008) among these years.
Table 9: Extent of the negative correlation between wind and solar for the years 2007-2010 - daytime

UK-C IB-N FR-S DE-C PL-N CZ-C IT-N

2007
0.035 -0.146
-0.186
2008 -0.241 -0.021 -0.214
2009 -0.221 -0.108 -0.290
2010 -0.270 -0.083 -0.284

-0.278 -0.162 -0.233 -0.224

-0.196 -0.188 -0.243 -0.205
-0.215 -0.077 -0.106 -0.289
-0.212 -0.135 -0.206 -0.353

3.2. Extraction of feed-in structures from the data

In subsection 3.1, the characteristics of wind and solar availability for Europe were discussed and their
in uence on the optimal electricity mix indicated. On the one hand long term average wind speed and
solar power as well as average correlations are important for the determination of the optimal
electricity mix. On the other hand characteristics such as the yearly availability or correlations can
signi cantly vary between years and the optimal electricity mix can only be determined by accounting
for these variations.
As the empirical data of combined wind and solar availability is available for ve years for this
analysis, we only have an indication about the variance for yearly full load hours for each region and
for the yearly correlation between regions or technologies. Therefore, we use a bootstrapping
approach to estimate the variance of yearly full load hours as well as the correlations between regions
and technologies. A selection of the created possible feed-in structures are used as input data for the
optimization model. The bootstrap approach is a resampling method which can be used to assess the
properties of a distribution underlying a sample and the parameters of interest that are derived from
this distribution (Efron, 1979). As a necessary condition for the bootstrap method, the original data
needs to re ect the underlying distribution. This leads to two critical assumtions for this analysis: First,
we assume that the hourly data for wind speeds and solar radiation of ve years represents the full
spectrum of possible weather situations. Second, as we create consistent wind and solar structures for
a future year, we need to assume that weather conditions will stay similar as today. It is clear that the
data does not contain all possible weather situations in Europe but it can be assumed that ve years of
hourly wind speed and solar radiation give a broad spectrum. Taking into account the e ects of climate
change on stochastic regional solar and wind availabilities in energy optimization models clearly
remains a challenge, but is beyond the scope of this paper.
As a rst step we generate 2000 di erent feed-in structures based on the provided wind speed and solar
radiation data for the years 2006 to 2010 (subsection 3.2.1). Ideally, all these could be used as input
parameters in the stochastic optimization model considering their relative probability. Due to
computational contraints for the optimization problem we will select representative feed-in structures
for wind and solar technologies throughout Europe (subsection 3.2.2).

3.2.1. Bootstrap approach to generate combined wind and solar feed-in strucures
To account for the above described seasonal characteristics for wind and solar availability, we divide
the dataset in two blocks: months from April to August as spring and summer; months from
September to March as autumn and winter. We randomly pick 30 days of consistent wind and solar
radiation data over all regions in three day-blocks from the dataset and repeat this 2000 times.6 By
taking blocks rather than single hours, typical hourly changes and daily structures of wind speeds and
solar radiation are re ected. Another advantage of picking blocks rather than single days is that
common general weather situations such as a storm traveling from Western to Eastern Europe are to
some extent considered. Naturally, due to picking three day blocks common weather situations which
last for more than three days are not re ected in the bootstrapped data.7 The possible feed-in of wind
power and PV sites in di erent regions in Europe is computed based on the hourly wind speed and
solar radiation of the 30 days (720 hours) as well as the technical parameters of wind and solar
technologies. Future state-of-the-art wind and solar technologies are assumed to have the technical
properties shown in Table 10.
As solar radiation is zero at night, the change from one block to another does not induce an unrealistic
change of solar radiation at midnight. The situation is di erent for wind speeds and therefore we
average wind speeds for the hours between 21 pm to 3 am to smooth the break around midnight. We
nd that taking the moving average of four hours leads to a realistic change of wind speeds.
Table 10: Assumed state-of-the-art wind and solar technologies
Technology

Capacity [MW]

Wind turbine 8
80
PV ground
1
14
PV roof
0.005 14

E ciency [%] Area [km2]

0.423 140

Height [m]

Radius [m]

65

To scale wind speeds from 30 meters to the assumed turbine height, the standard logarithmic
conversion is used. The conversion of wind speeds in reference height to turbine height are
computated by a scaling factor which is a function of turbine height, reference height and the
roughness parameter of the region. The roughness parameter takes the di erent surface conditions into
account.8
The power generation of wind turbines is calculated as a ratio of the installed capacity of the speci c
wind turbine.9 Power output is a function of air density, rotor area, power coe cient, wind speed and e
ciency. A typical power curve for wind turbines (pitch control) is assumed with no generation at wind
speeds lower than 3 m/s and a shutdown at more than 25 m/s to avoid damages. The power generation
by the assumed state-of-the-art photovoltaic system is computed based on the net e ciency, the surface
area and solar radiation. This implies standard con gurations of PV systems directed towards the

South and
with an angle of 30 degrees in order to achieve the highest yearly energy output.
Pel(reg; tech; h; s) = 1=Pnom(tech) 1=2r2(tech) v3(reg; h; s) total
(2)
Pel(reg; tech; h; s) = 1=Pnom(tech) total(tech) A(tech) radiation(reg; h; s)
(3)
The resulting regional wind speed and solar radiation structures have the characteristics shown in
Table 11. When comparing the wind speed and solar radiation to the original data re ected in Table 2
and 5, we nd similar wind speed and solar radiation characteristics. Hence, we argue that this
approach provides consistent feed-in structures of wind and solar technologies for several European
regions.

8Alternatively, the Hellmann height conversion formula could be used to scale wind speeds to di erent
heights:
v(h; s) = vrefh(h; s) [ refhh ] Hell . Typical Hellman coe cients Hell are in the range of 0.06 to 0.60
(Hsu, 1988).
9As we use a linear optimization model, any linear combination of technologies can be realized.
Therefore all capacities are normalized to 1 MW units.
Table 11: Summarizing statistics for created wind speeds and solar radiation for some of the selected
regions
Wind [m/s] UK-W (on)
IB-W (o )

IB-S (on)

DE-C (on)

PL-N (on)

Mean 6.8
Median
Variance

6.6
4.7
5.9

9.0
4.5
8.9

4.3
10.9

4.9
6.3
10.9

5.0
4.2
9.6

5.2
6.2
9.7

5.1
8.5
18.9

Solar [W/m2] UK-C IB-S

FR-S DE-C SK-S PL-N IT-S

Mean 131
Median

222
21

132
38

Variance

41,327 85,413 65,041 42,059 40,716 45,471 72,042

184
55

130
23

146
22

209
33

69

IT-S (on)

UK-N (o )

Figure 1 depicts the distribution of full load hours for two solar (Southern part of the Iberian Peninsula
and Northern Germany) and two wind regions (Central France and Central part of the United
Kingdom) in the 2000 created scenarios. The full load hours of wind as well as solar technologies di
er between the years and are normally distributed. However, the variance is signi cantly larger for
wind than for solar
generation.10
4%
12%
pr obability (wind) [%]
3%
9%
probability (solar) [%]
2%
6%
1%
3%
0%

0%

Wind on - FR-C (left)

Wind on - UK-C (left)

Solar - IB-S (right)

Solar - DE-N (right)

Figure 1: Distribution of full load hours in two wind and two solar regions in the 2000 scenarios

3.2.2. Heuristic to select representative feed-in structures

Due to computational constraints not all 2000 created feed-in structures can be used as input data in
the stochastic electricity market model. Therefore, representative feed-in structures are selected which
are
10As the estimation of yearly full load hours is based on resampling 30 instead of 365 days, it is
possible that the variance of full load hours is overestimated. To account for a possible
overestimation, we exclude the 10 % quantil on each side.
For this purpose, we de ne an indicating value for the yearly availability of wind power and an
indicating value for the yearly availability of solar power in Europe.
The importance of a speci c wind or solar site for an electricity system is mainly de ned by the area
potential and the expected power generation (full load hours). Therefore, we de ne the indicating
values as the average availability of the most important wind (solar) sites in Europe in terms of these
two factors. For wind power, we calculate the average full load hours of onshore wind in the Northern
part of the United Kingdom, Germany, the Iberian Peninsula and Poland and wind sites at the atlantic
coast of France as well as o shore wind at Norway's coastline. For solar power, we select the Southern
part of Italy, the Iberian Peninsula, France and Germany. From the distribution of the indicating
values, we pick ten feed-in structures with the following characteristics: S1 extremly low wind year;
S2 low wind year; S3 average wind year; S4 high wind year; S5 extremly high wind year; S6
extremely low solar year; S7 low solar year; S8 average solar year; S9 high solar year; S10 extremly

high solar year. Table 12 shows the full load hours in the selected scenarios. Apart from the yearly
amount of electricity generation the selected feed-in structures consider di erent hourly correlations
between regions and between technologies (wind and PV).
The bounds (lowest and highest full load hours) for each category are chosen such that the probability
for the extreme scenarios amounts to 2.5 %, for the low and high scenario to 10 % and the average
scenario to 25 %. As the probablility for an extremely high wind year is lower than an average wind
year, the dierent dispatches in the stochastic optimization model are weighted by the specific
probability factor as also shown in Table 12.

References
[1] Chaker M, Homji B.C, Mee III T, Inlet fogging of gas turbine engine- Fog droplet
thermodynamics, heat transfer and practical considerations, ASME Turbo expo, 2002-GT30562, 2002.
[2] Santos A.P, Andrade C.R, Zaparoli E.L, Comparison of different gas turbine inlet air
cooling methods, World academy of science, Engineering and Technology 61, 2012.

[3] Araimo L, Torelli A, Thermodynamic analysis of the wet compression process in heavy
duty gas turbine compressors, Gas turbines department, Ansaldo Energia, 2008.

[4] Bracco S, Pierfederici A, Trucco A, The wet compression technology for gas turbine
power plants: Thermodynamic model, Applied thermal engineering 27 (2007) 699-704.

[5] Jonsson M, Yan J, Humidified gas turbinesa review of proposed and implemented cycles,
Energy 30 (2005) 1013-1078.
[6] Bouam A, Aissani S, Kadi R, Gas turbine performances improvement using steam injection in the

combustion chamber under Sahara conditions, Oil & gas science and Technology Vol 63 (2008),
No.2, pp. 251-261, 2007.

[7] Wang F.J, Chiou J.S, Wu P.C, Economic feasibility of waste heat to power conversion, Applied
energy 84 (2007) 442-454.

[8] Cheng D.Y, CLN Development to Reduce NOx and Greenhouse Gas, Industrial applications
of gas turbines committee Paper No: 07-IAGT-1.2, 2007.

[9] Saleh B, Koglbauer G, Wendland M, Fischer J, Working fluids for low temperature organic
rankine cycle, Energy 32 (2007) 1210-1221.
[10] Gao H, Liu C, Chao H, Xuiaoxiao X, Li W.Y, Performance analysis and working fluid selection

of a supercritical organic rankine cycle for low grade waste heat recovery, energies ISSN 19961073, 2012.

[11] Fernandez F.J, Prieto M.M, Suarez I, Thermodynamic analysis of high temperature regenerative
organic rankine cycles using siloxanes as working fluids, Energy 36 (2011) 5239 - 5249.

[12] Chacartegui R, Sanchez D, Munoz J.M, Sanchez T, Alternative ORC bottoming cycles
for combined cycle power plants, Applied energy 86 (2009) 2162 2170.

[13] Kim K.H, Ko H.J, Park Y.S, Modeling of droplet evaporation for gas turbine inlet
fogging systems with different water injection ratios Vol 1, Numerical analysis and applied
mathematics, International conference 2009.

[14] Kim K.H, Ko H.J, Blanco H.P, Analytical modeling of wet compression of gas turbine
systems, Applied thermal engineering 31 (2011) 834-840.

[15] Kim K.H, Kim G, Thermodynamic performance assessment of steam-injection gas

turbine systems, World academy of science and technology 44, 2010.

[16] Korobitsyn M.A, New and advanced energy conversion technologies. Analysis of
cogeneration, combined and integrated cycles, 1998.

[17] Khan J.R, Fog cooling, wet compression and droplet dynamics in gas turbine
compressors, University of New Orleans Theses and Dissertations, paper 908, 2009.

[18] Higuchi S, Hatamiya S, Seiki N, Marushima S, A study of performance on advanced

humid air turbine systems, Proceedings of the International Gas Turbine Congress 2003
Tokyo November 2-7, 2003.
[19] Sairam. A, Analysis of advanced gas turbine cycle, M.Tech thesis submitted at IIT
Delhi ,2006.

[20] Hemadri V.B, Subbarao P.M.V, An investigation into organic rankine cycle (ORC) with
hexamethyl disiloxane (MM) as working fluid, International Conference on Applied Energy
ICAE 2012, Jul 58, 2012, Suzhou, China.

[21] R. A. Mugele and H. D. Evans, Droplet Size Distribution in Sprays, Industrial &
Engineering Chemistry 1951 43 (6), 1317-1324