.

Transient Performance of
Siemens SGT-750 and SGT-
800
Modeling and Simulations of
Industrial Gas Turbines on Island
Grids
Alexander Raddum

Master thesis 30 hp
Master programme of Energy engineering 300 hp
Department of applied physics and electronics. Spring 2020
Sammanfattning
Med en ökande mängd distribuerad energiproduktion, främst i form av förnyelsebara
energikällor såsom vind och solkraft krävs ökat fokus på nätstabilitet. Ett atraktivt
alternativ för att åstadkomma denna stabilitet är gasturbiner, mycket tack vare deras
förmåga att arbeta snabbt och flexibelt. Detta examensarbete, utfört för Siemens In-
dustrial Turbomachinery AB, syftar till att utvärdera transientkapaciteten hos deras
gasturbinmodeller SGT-750 och SGT-800. I arbetet har modeller för ett önät utveck-
lats i modellerings och simuleringsmiljön Dymola. Modellerna har verifierats mot data
för verkliga provkörningar och visat sig överensstämma väl med verkligheten. Model-
lerna har sedan körts för att utvärdera kapaciteten hos gasturbinerna i specika scenarion.
Utöver detta har simuleringar körts med alternativa bränslen innehållande varierande
mängd vätgas: 25, 50, 75 och 100 volymprocent vätgas. Resultaten visar att SGT-750
och SGT-800 klarar stora lastpåslag motsvarande 50% av märkeffekt i varierande omgivn-
ingstemperaturer (-30, 15 och 30o C). Den dubbelaxliga SGT-750 påvisade något större
frekvenstapp i dessa simuleringar. Vidare visade simuleringar med vätgasbränslen inga
större avvikelser vid lastpåslag med undantaget 75% vilken innehöll en stor mängd in-
ert gas. Utöver dessa resultat diskuteras även föreslagen κ-parameter för att kvantifiera
stabilitet och kapacitet vid större lastpåslag och effekterna av tröghetsmoment vid tran-
sienter. De begränsande faktorerna utreds till att vara gasgeneratorns rotationshastighet
och turbinernas inloppstemperatur. Vätgassimuleringarna gav slutsaterna att körningar
är fullt möjliga men modellen för brännkammare bör ses över för att få representiva
resultat. Vid körningar med gaser som uppvisar låg volymetrisk energidensitet bör hän-
syn tas till volymer och tryck i gassystem. Slutligen ges rekommendationer för framtida
arbeten på området.
Abstract
Distributed energy production in the form of renewable energy sources are expected to
increase in the coming years, a consequence of this is instability of the power grids due to
the stochastic nature and lack of inertia of renewable energy sources. In addition, small
and local, so called island grids, are on the rise and these system may present an even
higher sensitivity to frequency fluctuations. In these applications gas turbines are an
attractive option owing to the quick start capabilities, flexible fuel options and reliable
operation.
The aim of this thesis is to evaluate the transient capabilities of the Siemens SGT-750
double shaft and SGT-800 single shaft industrial gas turbines in island grid settings,
through simulations of substantial load increases in varying ambient settings. Further-
more the possibility of using hydrogen fuel as a renewable option to the standard natural
gas will be evaluated.
This thesis provides a model of a simple island grid for load sharing between two
or three turbines. The model was tuned to real life test data for the two gas turbines
considered. In order to evaluate the capabilities of the turbines simulations were run in
cold (-30o C), hot (30o C) and ISO (15o C) conditions, evaluating the maximum instant
load increase capabilities. Case studies were also run on island grids containing two or
three turbines in order to determine the frequency response in case of an event. Case A
regarded a scenario in which two turbines ran on 50 % of rated power and one tripped,
case B regarded three turbines working on 33 % of rated power and one tripped out.
Lastly, the maximum load increase cases with hydrogen fuel mixes (25, 50, 75 and 100%
hydrogen by volume) were considered.
The results suggest that the SGT-750 and SGT-800 gas turbines are capable of hand-
ling scenarios on reasonably dimensioned power systems, with both machines capable of
recovering instant load increases of over 50 % of the rated power. The findings shows that
short periods (<10 s.) of allowed overfiring temperatures are necessary for the transient
performance for the most extreme scenarios of high ambient temperatures and large load
increases (around 50% of rated power). Furthermore an empirical κ parameter, related
to inertia and operational stability is discussed in order to compare GT load increase
capability. The relevance of inertia and dynamic response is discussed and conceptually
simulated to highlight the their role in gas turbine transient response.
The hydrogen simulations, aside from the 75% case, showed little difference from
natural gas in transient scenarios. The 75% hydrogen fuel consisting of high amounts of
inert gas however, rendered the turbine not able to withstand substantial load increases.
The hydrogen simulation results are suggested to be accounted for by the rather simple
combustion system and the energy densities of the gases.
Acknowledgements
I would like to thank and acknowledge the support, patience and help provided by my
supervisors at Siemens industrial turbomachinery AB Anna Sjunnesson and John Svens-
son. Without your tremendous expertise and input this would not have been possible,
you have taught me a lot. Thank you for staying in contact every day throughout the
Covid pandemic of 2020. Futhermore I would like to thank the performance department
at Siemens Industrial Turbomachinery for their hospitality and helping efforts. Lastly I
would like to thank my supervisor, Dr. Gireesh Nair at the department of applied physics
and electronics at Umea University. For his support, patience and meticulous review of
the following work, and a lot of other work throughout my time as a student.
Nomenclature

List of Abbreviations
CCGT Combined cycle gas turbine RES Renewable energy source
CT Compressor turbine ROCOF Rate of change of frequency
DG Distributed Generation TIT Turbine inlet temperature
GT Gas turbine TOT Turbine outlet temperature
HHV Higher heating value VGV Variable guide vane
IGV Inlet guide vane
ISO Standard conditions
OCGT Open cycle gas turbine
PID Proportional integral derivative
ppmvd parts per million by volume

Symbols unit Greek symbols unit

cp Heat Capacity at constant p J·(kg·K)−1 γ heat capacity ratio
h Specific enthalpy kJ·kg−1 δ Angle
f Frequency Hz ∆ change
J Moment of inertia kg·m2 φ relative humidity
ṁ mass flow rate kg·s−1 ϕ power angle
p pressure Pa η efficiency
P power W π pressure ratio
q specific heat J·kg−1 ρ density kg·m−3
Q heat J τ torque Nm
R gas constant J·K−1 ·mol−1 ω angular velocity rad·s−1
T temperature oC / K

U internal energy J
V mean velocity m·s−1
w specific work J·kg−1
W work J
Contents
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Project description 2
2.1 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.3 Siemens SGT-750 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.4 Siemens SGT-800 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

3 Literature review 4

4 Theory 9
4.1 Gas turbine principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.2 Combustion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.3 Other factors effecting performance . . . . . . . . . . . . . . . . . . . . . . 15
4.4 Mechanical and electrical torque balance . . . . . . . . . . . . . . . . . . . 16
4.5 Gas turbine control overview . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.6 SGT-750 and SGT-800 control systems . . . . . . . . . . . . . . . . . . . . 20
4.6.1 SGT-750 control system . . . . . . . . . . . . . . . . . . . . . . . . 20
4.6.2 SGT-800 control system . . . . . . . . . . . . . . . . . . . . . . . . 23

5 Method 25
5.1 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.2 Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.3 Dymola model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5.4 Validation method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.5 Simulation case studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

6 Results 34
6.1 Results of validation cases . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.2 Case studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.2.1 Instant load increases under different ambient conditions. . . . . . 36
6.2.2 Island grid cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6.2.3 Hydrogen fuel cases . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6.3 Remarks on the results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6.3.1 Limiting factors for the SGT-750 . . . . . . . . . . . . . . . . . . . 41
6.3.2 Limiting factors for the SGT-800 maximum instant load increase . 42
6.3.3 Hydrogen simulation results . . . . . . . . . . . . . . . . . . . . . . 42
6.4 On the differences of the machines . . . . . . . . . . . . . . . . . . . . . . 43
7 Discussion 45
7.1 The island grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
7.2 Maximum capability simulations . . . . . . . . . . . . . . . . . . . . . . . 45
7.3 Hydrogen simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7.4 Reflections and areas of improvement . . . . . . . . . . . . . . . . . . . . . 47

8 Conclusions 49

9 Future work 50
1 Introduction
1.1 Background
The power grids of the future will possibly consist of a large number of producers util-
izing renewable energy sources (RES). The International Energy Agency (IAE) predicts
renewable-based capacity will increase 50% globally between 2019 and 2024[1]. The in-
crease in RESs seen today mainly consists of solar and wind which, due to their stochastic
nature caused by their dependency of weather conditions, are unreliable and hard to pre-
dict [2]. In addition to the increased share of RES microgrids are also on the rise, these
are localized electricity grids which can work in synchronous with larger (national) grids
but also separately (island mode)[3]. No strict definition applies to microgrids, they are
vaguely defined as smaller grids able to run independently of national grids. Some of the
advantages of microgrids are (but not limited to) reduced line losses, resilience to large
scale disruptions, since they can work parallel to national grids in cases of emergency,
and easier RES integration[3]. RESs present problems when implemented in larger scale
to existing grids, and microgrids can be designed in order to handle variable generation
through storage and power balancing [3].
In order to secure a consistent power supply several technical solutions for energy
storage exist depending on the time frame (e.g. flywheels, batteries and pumped hydro
power)[4]. The most common methods of energy storage as of 2016 were pumped hydro
power and battery storage[5], these methods, however, presents some issues respectively.
Pumped hydro power requires the right geographical circumstances (i.e. considerable
space and elevated bodies of water) [4], while batteries, depending on chemical compos-
ition, experience low energy density and economical and safety concerns[6].
Other solutions to the problem of fast transients, oscillations of the grid frequency,
due to increased shares of RESs involves highly flexible conventional power plants such
as gas turbines (GT) or steam plants (SP) [7]. Gas turbines in particular are faster than
SP (with a load ramp capability of 20% of maximum load per minute as opposed to
about 6% of maximum load per minute for SP)[7] but slower than batteries, however,
they are able supply more power to the grid and provide large quantities of power[8].
Gas turbines could be a great option for microgrids due to their flexibility, in terms
of swiftness, as mentioned earlier [7], fuel capabilities and lower emissions as compared
to SPs [9]. Furthermore GTs do not require require large boilers (as SPs do), specific
weather conditions (as RESs do) or particular settings, as they can be delivered as
packages fitting within a single container. In order to further reduce the environmental
impact hydrogen fueled GTs may also constitute an alternative. It is of importance to
evaluate the capability of the GTs to quickly and forcefully respond to transients caused
by load increases or frequency drops via increases in power output, but also fuel flexibility,
before implementation.

1
2 Project description
2.1 Objectives
This thesis will consider the performance of Siemens latest single and double shaft indus-
trial GTs: SGT-750 and SGT-800 (57 MW) under different operational conditions. The
core models of the engines and control systems that will be used are developed by Siemens
industrial turbomachinery AB in the modeling and simulation environment Dymola. In
addition to the existing models models for simulations of island grids are to be developed.
The developed models are to be tuned to real on site data and Simulink models in order
to produce representative results. The models will then be used to predict the behaviour
of the SGT-750 and SGT-800 under different, predetermined, transient events and fuel
compositions.
The goals of this master thesis is to evaluate and quantify the transient capabilities
of the SGT-750 and SGT-800 in terms of maximum transient load capabilities under
different ambient conditions and scenarios. It will also investigate the prospects and
implications of implementing hydrogen gas mixtures as fuel to reduce the environmental
impact.
The purpose of this work is to investigate, in order to ensure, Siemens Industrial
Turbomachinery products are competitive in future energy systems consisting of large
numbers of distributed producers and possibly grids subjected to an increase of transients.
The project will be done at the SIT AB Offices in Finspang, Sweden for the Performance
R&D group.

2.2 Limitations
This thesis is limited to the evaluation of the Siemens SGT-750 and SGT-800 gas turbines.
In addition, the following limitations applies: No major changes of the gas turbines will
be implemented (if not stated otherwise), no specific power system implication analysis
will be undertaken. The focus will be on the overall performance of the existing engines
and the operational aspects causing them. Thus excluding component design aspects,
aerodynamics and exact heat transfer analysis.

2.3 Siemens SGT-750

The first GT considered in this thesis is the SGT-750, a twin shaft industrial gas turbine
for mechanical drive or power generation applications launched in November 2010 [10].
General power generation performance data of the SGT-750 data is shown in tab. 1. The
SGT-750 is a flexible machine capable of of liquid and gaseous fuels, fast start capability
(full load in <10 min) and best on the market NOx emissions. It offers high reliability
with only 17 scheduled maintenance days in 17 years. A cut through of the SGT-750
showing its main parts can be seen in fig. 1. The SGT-750 has two VGVs and four
compressor bleed valves.

2
 Table 1 – SGT-750 simple cycle power generation data [11]

Power 39.8 MWe

Efficiency 40,3 %
Shaft speed 6100 rpm
π 24.3:1
Exhaust gas flow 115.4 kg/s
Exhaust temperature 468 o C
NOx emissions <15 ppmvd

Figure 1 – The Siemens SGT-750 cut through showing the compressor (13 stages) on the
left, the burner cans, CT (two stages) and power turbine (two stages).

2.4 Siemens SGT-800

The second GT considered is the single shaft SGT-800, originally launched in as GTX100
capable of 45 MW output [12]. The later, updated versions of the SGT-800 are capable
of up to 62 MWe . The data for the B5 version, considered in this thesis, is found in
tab. 2. As its twin shaft counterpart the SGT-800 is capable of low NOx emissions and
high reliability, proven over more than 7 million operating hours, from the more than 370
units sold [13]. The SGT-800 has three rows of VGVs and two compressor bleed valves.

3
 Table 2 – SGT-800 simple cycle power generation data [13]

Power: 57 MWe
Efficiency 40,1 %
Shaft speed 6600 rpm
π 21.6:1
Exhaust gas flow 136.6 kg/s
Exhaust temperature 565 C
NOx emissions <15 ppmvd

Figure 2 – The Siemens SGT-800 cut through showing the compressor (15 stages) on the
left, the circumferential burners, and turbine (three stages).

3 Literature review
Transients and disturbances affecting the quality of the power delivered to the electrical
grid can be a result of external events (e.g. lightning) or internal issues, such as utilities
connected to the grid or producers delivering uneven frequencies or voltages. Issues
may also arise when large inductive loads (e.g. inductive motors) are introduced due to
reactive power imbalance. The increasing integration of distributed generation (DG) from
RESs contributes to these issues and can cause dips and steady state voltage rise, voltage
flicker and harmonics(which in turn increases the current in the system by propagating
fluctuations)[14]. Fluctuations of frequency in power grids are countered by the inertia

4
of the power grid caused by heavy rotational masses (turbines and generators) connected
synchronously to the grid and capacitors, the increase of RESs may lead to a low-inertia
power grid since they do not posses these qualities in general [15]. Wind turbines, while
rotating, are in general decoupled from the grid through converters while photovoltaics
experiences virtually zero inertia, these problems are amplified on island grids that are
inherently sensitive due to the size and potential lack of backup power as required [15].
The problems experienced by RESs may be mitigated in several ways, such as energy
storage and inertia emulation discussed in [15, 16]. With inertial contributions and on
demand power supply the capabilities of GTs for the purpose of power balancing and
grid support will be evaluated.
In order to evaluate the capability of GTs as a tool for balancing power and transients,
a simulation approach will be undertaken. Investigations of load sharing of one and
several GTs and load distribution in the case of failure or trip (emergency shutdown) will
also be undertaken. For the applications regarded in this thesis, transient behaviour is
to be simulated, hence a review of previous work on the subject will be undertaken in
order to provide insight and a foundation to work from. The review will, in addition,
investigate factors affecting GT performance.
Firstly, a brief review of the main factors influencing thermodynamic performance
of GTs are regarded, these factors were investigated by Rahman et al. [17]. In [17]
the thermodynamic relations are studied, and parametric simulations were undertaken.
The findings suggests that compression ratios, ambient temperature, air to fuel ratio and
isentropic efficiency strongly influences the thermal efficiency of a gas turbine power plant.
The results shows that thermal efficiency is increased mainly by: an increase in turbine
inlet temperature, a decrease in ambient temperature and an increase in compression ratio
[17]. These finding were also substatiated by Saif and Tariq in [18]. From parametric
calculations, it was concluded that an increase in ambient temperature will yield lower
efficiency and specific work output while lower temperatures leads to the contrary.
These findings regard thermodynamics, the influence of control systems are discussed
in later segments.
Asgari et al. released a comprehensive paper on the matter of transient simulations
of single shaft industrial GTs in 2014 [19]. In said paper two approaches to transient
GT simulations are reviewed: a) Neural networks (black box approach) and b) A Model
based on physical and thermodynamic properties (white box approach). In [19] four
distinct data sets were considered for simulation and validation. The sets were defined
by four binary parameters1 : I. Starter is on or off: 1 or 0. II. The GT is connected to
the grid or not: 1 or 0. III. Customer trip happens or not: 0 or 1. IV. The flame is
on or off: 1 or 0. The following data sets were considered: [0 0 0 1], [0 1 0 1], [0 1 0 1
a] and [0 1 0 1 b], where the first two sets were used for tuning of models and the last
two were used for validation [19]. The simulations carried out all considered cold start
of the GT in addition to the previously mentioned factors. The simulations carried out
were able to predict rotational speed (N ), compressor pressure ratio (πc ), compressor
1
A data set [0 1 0 1] will, by this definition, mean that starter is off, the GT is connected to the grid,
customer trip does not happen and the flame is on.

5
outlet temperature (T02 ) and turbine outlet temperature (T04 ) for all data sets with both
models. The results presented RMSE of 0-4% for both models, they presented larger
errors at the beginning of the simulations until they stabilized to the response [19].
Furthermore, a component based Simulink library GasTurboLib has been developed
by Panov [20]. In the publication a meticulous description of the component modeling
is undertaken, discussing the physical and thermodynamic assumptions as well as the
control system layout. The paper proceeds by verifying the model for a single shaft
turbine (Siemens SGT-100-1S) and a twin shaft model (Siemens SGT-400), the results
presented are in agreement with engine test data for load acceptance and startup [20].
In a paper by Bahrami et al. the transient performance (i.e. short term response),
in particular during frequency drops, of GTs are discussed [21]. In [21] the differences
of single- and double-shaft GTs are discussed, it is pointed out that while the single
shaft turbine inherently has greater inertia and less lag, the mass flow rate of air will
decrease during frequency drops, due to the fact that the turbine provides torque directly
to the compressor, thus leading to potential unit instability. The control system of GTs
is evaluated and it is stated that the temperature controller (an overview of a generic GT
control system presented in [22]) , struggles in the case of fast load changes. The governor
adjusts fuel flow in to the combustion chamber during normal operation in order to match
the required power output, the temperature controller maintains the temperature of the
components by controlling fuel flow and IGVs. It is worth noting that the control system
selects the lowest value of the temperature control signal and the speed governor signal in
order to get smoother transitions [21] (also discussed in [23]). The control system could
be limited by the speed of the IGV actuators, thus rendering air flow the limiting factor
for fast response to load changes [21].
In order to improve the transient performance of GTs a separate steam injection
system which works on a separate controller (event-based controller) was suggested in
[21]. Steam injection will allow for higher power outputs due to an increase in mass
flow and lowered risk of overheating and may as such improve performance. For small
amounts of steam injections there is no need for modifications of the gas turbine or the
combustion chamber [24]. The system was simulated for a 160 MW heavy duty GT. The
results of the simulations indicated that steam injection improved the response to a step
change in GT load while slightly reducing the compressor surge margin. The results also
showed a higher allowed change load, in particular at full load conditions, this is due to
the reduced temperature which allows the GT to continue running without triggering
the temperature controller [21].
In a publication by Meegahapola the dynamics of gas turbines during frequency
variations in power networks were characterized [25]. The model used for simulation
is based on the model by Rowen [26]. The scenarios from [25] that were considered
were operation below rated power output, load reduction and generator outage/load
increase. Simulations were conducted on combined cycle GTs (CCGTs) and open cycle
GTs (OCGTs), the rated power output was 400 MW, in the case of OCGTs two 200 MW
units in parallel were considered. Furthermore, the system considered operated with an
1800 MW load, the CCGT had an inertial constant of 8 s and the OCGT had an inertial

6
constant of 4 s, both setups were rated at 500 MVA. The test network had an inertia
constant of 6 s and short-circut capacity of 5000 MVA and a frequency droop of 4 %
was set. Simulations of load increased considered an additional 200 MW of load to the
system. The CCGT presented a frequency drop to 49.18 Hz after a few seconds (from
50 Hz), the OCGT dropped to 49.22 Hz minimum [25]. Furthermore the power output
of the CCGT reduced beyond the OCGT after 6.8 seconds of simulation worsening the
frequency stability of the system [25]. The causes of the phenomena found was explained
by lowering of compressor speed (single shaft OCGT), and for the CCGT an increase
in the exhaust temperature finally leading to an override by temperature controller (as
discussed in[21, 25]) causing altering of the IGVs2 resulting in a power reduction. The
temperature spread in the load increase scenario was however low, likely not leading to
lean blow out (LBO). The GTs were also simulated operating below rated power (380
MW). When the load was reduced, the exhaust temperature did not rise as rapidly as
when it was increased, this in turn led to to power increase. Both GTs frequencies
decreased to about 49.3 Hz (from 50) and none of them experienced temperature spreads
causing concerns for LBO [25]. The simulation regarding load reduction assumed an
initial load of 2000 MW reduced to 1800 MW and both GTs operating at rated power
output. Both systems showed an increase in frequency to about 50.8 Hz, the temperature
controller never over rode the governor leading to both GTs reducing power output
by about 24% of initial output [25]. The CCGT presented a high spreads of exhaust
temperatures due mainly due to increased air to fuel ratio due to inertia causing continued
air pumping when the fuel injection is reduced [25]. Finally, [25], concludes that the
temperature control system (also discussed in [21]), in this paper considered for the
CCGT, may lead to increased turbine power output in the case of frequency changes.
When considering frequency response of systems the rate of change of frequency
(ROCOF) is a recurring measure, especially in cases considering RES [25, 27, 28]. As
previously mentioned it is of importance to avoid frequency fluctuations in order to avoid
interruption, additional generator trip or even blackouts, for these purposes ROCOF can
serve as an indicator [30]. In large grids ROCOF is also an important measure to avoid
unwanted islanding, which also may pose risks [27]. In the case of GTs specifically,
ROCOF can lead to LBO [25].
In a paper by Kakimoto [29] a parametric study of GT power plants (CCGT) during
frequency drops is conducted. The authors uses a variation of the Rowen ([26]) model
to investigate factors influencing the behaviour of the GT. The findings suggest that the
temperature control overrode speed control in the case of frequency drops and restricts
the fuel flow about the initial value [29]. It is also found that if the IGVs open fully,
the temperature control and the frequency determines the fuel flow and thus the power
output [29].
In order to reduce greenhouse gas emissions several parameters may be considered,
including but not limited to overall efficiency and fuel choice. Hydrogen may be a fuel
2
In the simulations carried out the OCGT considered was not equipped with controlled IGVs [25].

7
of the future since the combustion of hydrogen
1
H2 + O2 = H2 O − ∆H (1)
2
releases no greenhouse gases. In addition to the environmental benefits hydrogen has
more than double the energy density of conventional fuels at around 120 MJ/kg, how-
ever the volumetric energy density is less than half that of conventional fuels [31]. For
the time being, the volumetric energy density poses practical problems with regards to
transporting and storing hydrogen [32]. Some of the solutions proposed for the problems
of storage and transportation of hydrogen includes compression, liquid storage and solid
storage all with their potential difficulties and issues [32]. Several ways of producing hy-
drogen have been proposed including reforming of fossil fuels, production through RESs
(biofuels or via electrolysis), gasification and thermocemical pyrolysis (water splitting
through heat) [33]. In [33] it is concluded that thermochemical pyrolysis and gasification
are economically viable options, furthermore it is pointed out that the issues of storage
and safety concerns need to be evaluated further.
In [34] the prospects of hydrogen as a fuel in heavy-duty GTs designed for natural gas
is discussed. The authors runs simulations considering a CCGT consisting of a single shaft
GT capable of 300 MW power output and steam turbine capable of of over 130 MW power
output. The effects of hydrogen combustion on turbomachinery is summarized by three
effects: variation of enthalpy drop in the expander, variation in flow rate affecting the
turbine-compressor matching and variations in heat transfer in turbine blades affecting
the cooling system performance [34]. Since the volume flow rate will differ for different
fuel compositions the compressor and turbine will land on different operating points
in their characteristics (further described in later sections). The authors carries out
calculations through their developed code (from [35]) for three strategies of operation:
increased π pressure ratio) and reduced TIT, re-engineering of the machine (increasing
flow area and changing blade cooling) and VGV operation with lowered TIT. The variable
VGV case shows slight differense for no-diluted hydrogen fuel with slightly lower TIT and
TOT, and marginally increased efficiency. Similar results go for the diluted cases with
steam (6.78 dil./fuel mass ratio) and nitrogen (14.44 dil./fuel mass ratio) [34]. While
hydrogen reduces CO, CO2 and HC emissions a potential drawback of hydrogen fuels
may be an increase in NOx with increased hydrogen content [36].

8
4 Theory
4.1 Gas turbine principles
A GT consists of three main components: compressor, combustion chamber and turbine.
The two types of GTs considered in this paper, single shaft and double shaft, can be
seen in Fig. 3. The single shaft GT, as the name suggests, dispenses all work through
one shaft, driving both the compressor and the load (e.g. generator). The double shaft
GT utilizes a second turbine for the load thus allowing the gas generator (GG) to work
on separate speeds independent of the power turbine (PT). The double shaft GT is able
to produce higher torques at lower rpm, but may also "rush" (overspeed) when load is
decreased, which is not the case for single shaft GTs due to higher inertia [37]. Moreover,
single shaft GTs may experience an increase in TOT (thus decreasing η 3 ) and at lower
frequencies close in on the surge margin while the double shaft GT is free to work at its
design point regardless of PT operation [37].

Figure 3 – Schematic of single shaft (upper) and double shaft (lower) gas turbines, where
the arrows represents flow of air, gas and exhaust.

In order to describe the GT one may start with the well established theory of conservation
of energy [38]
δU = δQ + δW, (2)
stating the internal energy of a system δU is the sum of work done δW (by or on the
system) and the heat, δQ, entering or leaving the system. The GT works on an open
power cycle, the Brayton cycle proposed by George Brayton in 1870 [39], and consists of 4
main stages shown in fig. 4. The cycle displayed however is ideal (∆scomp = ∆sexp = 0),
since in a real cycle entropy will increase and the vertical lines will be slightly tilted. The
offset from vertical in the T-s diagram is due to the isentropic efficiency ηis .
3
This may not be true for CCGT operation where an increase in TOT increases heat transfer in the
boiler.

9
 Figure 4 – Entropy (s) and temperature (T) diagram of the ideal Brayton cycle including:
(1-2) compression, (2-3) combustion and (3-4) expansion.

Using the relation for conservation of energy and neglecting changes in kinetic and po-
tential energy, the Brayton cycle in steady state operation is described (on a unit mass
basis) by
(qin − qout ) + (win − wout ) = hin − hout , (3)
where q is heat, w is work and h is enthalpy. From eq.3 the a relation for each main
component can now be written
w12 = −(h2 − h1 ) = −cp (T2 − T1 )
q23 = h3 − h2 = cp (T3 − T2 ) (4)
w34 = (h4 − h3 ) = cp (T4 − T3 )
∂h
where cp is heat capacity (cp = ∂T |p ). Still considering the assumption of neglecting
kinetic and potential energy the cycle efficiency is described as the work output divided
by the heat supplied, i.e.
cp (T4 − T3 ) − cp (T2 − T1 )
ηBrayton = . (5)
cp (T3 − T2 )
Considering the isentropic p − T relation (no heat transer to the surroundings and no
irreversibilities [39])
T2 γ−1 T3
=π γ = , (6)
T1 T4
where π is the pressure ratio (π = p2 /p1 = p3 /p4 ) and γ is the pressure ratio of the gas,
the efficiency can now be expressed as
1
γ−1
ηBrayton = 1 − γ
, (7)
π
giving a description of the efficiency only depending on the pressure ratio and nature of
the gas [40]. Since, however, the specific work output can be expressed as
w = cp (T3 − T4 ) − cp (T2 − T1 ), (8)

10
the expression  
w 1 γ−1

= t 1 − γ−1 − π γ − 1 , (9)
cp T1 π γ
where t = T3 /T1 , shows the dependency of temperatures and the pressure ratio [40].
The presented equations (2-7), as stated, are not representative of real life GTs, as such
further theory on methods to account for losses follows. To include the neglected terms
of velocity, stagnation quantities (subscript 0) may be considered as

V2
h0 = h +
2
V2
T0 = T + (10)
2cp
  γ−1
T0 γ
p0 = p
T

based on the assumption of a perfect gas (h = cp T ), where V is velocity [40]. Furthermore

considerations of irreversibilities should be made to calculate the actual work needed
and delivered by the components. for this reason the isentropic efficiency is introduced,
expressed for the compressor and the turbine as [39]

ws h02s − h1 T02s − T01

ηc = = = .
w h02 − h1 T02 − T01
(11)
w h03 − h04s T03 − T04
ηt = = = .
ws h03 − h04 T03 − T04s
where subscript s represents the isentropic state (states shown in fig. 4). Eq.11 represents
the ratio of the real and ideal performance. Temperature correlations can be written as
   γ−1 
T01 p02 γ (12)
T02 = T01 + 1− ,
ηc p01

and
  γ−1 
1 γ
(13)
T04 = T03 − ηt T03 −1 .
p03 /p04

For calculations the isentropic efficiencies may be assumed, however, when designing
GTs, or calculating performance at different pressure ratios, these are not sufficient[40].
For the aforementioned purposes polytropic efficiency is introduced. The case for the
introduction of polytropic efficiency can be made based on eq. 6. Considering the
isentropic efficiency of one stage of the compressor ησ to be constant it follows that
N
1 X
T2 − T1 = (Ti − Ti−1s ), (14)
ησ
0

11
where N is number of stages. As seen from eq. 11 (T2s − T1 ) = (T2 − T1 )/ηc thus
PN
ησ 0 (Ti − Ti−1s )
= , (15)
ηc (T2 − T1s )

here it is clear that the numerator will be of greater magnitude since it the distance
between pressure lines will increase as entropy increases, and it will increase the more
stages that are present [40]. In practice this means that the isentropic efficiency of the
compressor will decrease as pressure ratio increases [39], since some of the work input
will result in heating of the gas. A similar reasoning goes for the turbine resulting in an
increased efficiency as the pressure ratio, and thus the temperature difference, increases.
Temperature correlations for using polytropic efficiencies are [40], similar to eqs. 16-17,

p02 X
   
T02 = T01 + T01 1 − , (16)
p01

and
 Y 
1 (17)
T04 = T03 − T03 −1 .
p03 /p04

where
ln(p2 /p1 )(γ−1)/γ

X = (γ − 1) γ (18)
ln(T2 /T2 )
and
ln(T3 /T4 )γ/(γ−1)
Y = ((γ − 1)/γ). (19)
ln(p3 /p4 )
Continuing with the assumption of an ideal gas, gives ways to express the power output

Pt = ṁ∆ho =
= ṁcp ∆T0 =
 1−γ
 (20)
γ
= ṁcp T0,in ηt 1 − πt

similarly for the compressor

 γ−1 
ṁcp T0,in γ
Pc = πc − 1 . (21)
ηc

The overall characteristics of a given compressor can be represented in a compressor

map also known as a characteristics diagram. An arbitrary compressor diagram can
be seen in fig.5, where the quantities displayed are normalized as to be representative at
arbitrary ambient conditions. It is of essence that the compressor works to the right of the
"surge line" in order to avoid damage to the compressor and performance deterioration.
Surge is a complex phenomenon, which, in short, causes separation from the compressor

12
blades causing reversed discharge through the compressor potentially causing damage
[45].

Figure 5 – Typical compressor characteristic.

As with the compressor, the turbine characteristics can be mapped, this can be useful
for off-design performance calculations [45]. An arbitrary turbine efficiency map is shown
in fig 6 in which the axis values are normalized and the different lines represents different
normalized speeds. The non dimensional quantities are implemented for generalization
purposes and can be found by the Buckingham Pi theorem

Π1 = ψ(Π2 , Π3 , ..., Πk−r ), (22)

which states that a function of k variables in r dimensions can be expressed in k − r
groups (Π-terms) [43]. For a compressor these are
√
p02 T02 m RT01 N D
, , ,√ , (23)
p01 T01 D2 p01 RT01
these terms may be implemented to represent performance as in fig.5 and for a given
machine at constant size working on the same fluid R, A and D can be omitted [40].

13
 Figure 6 – Typical turbine efficiency characteristic.

Similar maps as described above of the SGT-750 and SGT-800 are implemented in the
simulation program used in this thesis.

4.2 Combustion
The addition of energy in the GT takes place in the combustion chamber where fuel,
typically natural gas or liquid hydrocarbons, are burned. The combustion process is
one of high complexity involving fluid dynamics and chemistry outside the scope of this
thesis. However, the basics will be touched upon in order to get a holistic view of the GT.
In the combustion chamber fuel is sprayed and mixed with air with the goal of highly
efficient combustion and even temperature distribution. To achieve this the aerodynamics
of the combustion chamber must be consider so as to ensure complete combustion and
sufficient mixing of fuel and air. In order to increase the overall efficiency of the GT it is
desirable to reduce the pressure drops occurring in the CC as well. The pressure losses
arise as a result of two causes: turbulence and skin friction and temperature increases
[40].The stagnation pressure loss due to temperature increase can be described (assuming
incompressible) as [40]
p01 − p02  T2 
= − 1 , (24)
ρ1 V12 /2 T1
from the momentum equation. The losses form friction can be adequately be described
by the pressure loss factor [40]

∆po
P LF = , (25)
m2 /(2ρ 2
1 Am )

where Am is the maximum cross sectional area of the chamber. The combustion reaction
of hydrocarbons in excess air is described as

Cx Hy + m(O2 + 3.76N2 ) =
y (26)
= xCO2 + H2 O + (m − x − y/4)O2 + 3.76N2 ,
2

14
here m denotes number of moles of air and x and y the number of carbon and hydrogen
atoms in the fuel(e.g. methane CH4 ). The energy released is the difference of the energy
in the reactants and the products. Eq. 26 does not take into consideration hydrocarbons
(e.g. CO) formed from incomplete combustion and nitrous oxides (NOx ). NOx pollution
cause the formation of ground Ozone (O3 ) and leads to acid rain thus explaining the
need to reduce formation and release [46]. CO is an odorless gas which may cause several
health problems (comprehensive summary in [47]). The formation of CO can be reduced
by increasing the available air in the combustion process allowing complete combustion
[45]. Nitrogen does not take part in the combustion process, however nitrogen from the
fuel and air may react to form NOx due to the conditions in the combustion chamber.
NOx formation increases with an increase in pressure and temperature, while the inverse
is true for CO [45]. In order to compare fuels in a useful way the Wobbe Index is
introduced
HHV
Iw = p , (27)
ρgas /ρair
which is a measurement of the volume of fuel required for a certain amount of energy,
in units MJ/Nm3 . For a fuel mixture containing inert gases (e.g. CO2 ) some energy is
required to heat these, which in turn means the required volume is not only depending
on the HHV [37].

4.3 Other factors effecting performance

Apart from thermodynamics the performance of a GT is, of course, depending on the
design of the machine. In some ways the physical properties of the machine indirectly
effects the thermodynamics discussed previously. Ambient conditions also influences the
performance which can be seen from eq.4 which can be written as

W12 = ṁ∆h, (28)

for the compressor. Since the density of air is [39]

p φp
ρ= + , (29)
Ra T Rs T
where R are specific gas constants and φ relative humidity, it is clear from eq.28 that
ambient conditions directly effects the compressor power required, a similar reasoning
goes for the turbine. The work output of a GT will, as seen in eq. 20, depend on the
TIT and is thus limited by the materials. The performance of the compressor can be
changed using VGVs, which in practice changes the geometry of the compressor as well
as the angles and swirl of the airflow. VGVs help in ensuring performance at sub-rated
operation by controlling airflow. At the same time VGVs impact the TOT temperature.
Opening of VGVs will serve to change the compressor characteristics in fig. 5 and allow
for a greater surge margin [45]
πs
SM = − 1, (30)
πw

15
where s denotes πc at surge and w the working point, describing the proximity to the
surge line. The so called velocity triangles for a stage of rotors and stators, often used
([40],[45], [37]) is shown in fig.7. In fig.7 subscripts a and w represent axial (vertical) and
tangential velocities respectively, U is the velocity of the rotors α represents the angle
att which the flow enters the rotors and stators and β are the rotor blade air angles.

Figure 7 – Velocity triangles of one stage of of rotors and stators.

Once again, utilizing conservation of momentum (and simple trigonometry4 ) the power
input can be expressed as

W = mU Ca (tanβ1 − tanβ2 ), (31)

so closing the VGVs reduces stage loading on the effected stage [45]. In turn the VGV
angle will affect flow and thus temperatures by, in effect, changing the geometry of the
compressor.

4.4 Mechanical and electrical torque balance

GTs are interconnected with the grid via synchronous generators, meaning their frequency
is equal to that of the grid. The electrical frequency is described as a function rotor speed
and number of poles of the generator by the relationship
p
fe = nm , (32)
120
where p is number of poles, and nm is the rotational velocity of the machine in rpm. The
equation of motion for a synchronous machine, based on Newtons II:nd law is written as

dωm
J = τm − τe , (33)
dt
4
Works under the assumption Ca1 = Ca2 .

16
where J is the total moment of inertia (i.e. from turbine and generator rotor), ω is the
angular velocity of the rotor axis relative to a reference, τe and τm are electrical and
mechanical torque respectively[41]. Eq. 33 can be normalized by introducing the inertia
constant defined as
2
Jω0m
H= , (34)
2V A0
where ω0m is the nominal rotational angular velocity, and V A0 is the rated power output
VA (Volt Ampere) of the machine. H can simply be seen as normalized kinetic energy
2
since Ek = Jω2 for rotation round a fixed axis [42]. Solving for J from eq. 34 and
substituting into eq. 33 yields
2HV A0 dωm
2 = τm − τe . (35)
ω0m dt

By rearranging and utilizing the fact that V A0 /ω0m (since τe = P/ω) gives eq.33 in per
unit form as
dω˜r
2H = τ˜m − τ˜e , (36)
dt
where ω˜r = ωm /ω0m = ωr /ω0 (where ωr is the instant rotor electrical angular velocity
and ω0 is rated electrical angular velocity, based on eq. 32). By introducing the terms

δ = ωr t − ω0 t + δo
dδ
= ωr − ω0 = ∆ωr (37)
dt
d2 δ d(∆ωr ) d(∆ω˜r )
2
= = ω0
dt dt dt
representing angular position of the rotor with respect to a synchronous rotating reference
and where δ0 is δ at t=0 we arrive at

2H d2 δ
= τ˜m − τ˜e , (38)
ω0 dt2
describing the rate of change of deviations of the rotor rotation from the synchronous
electrical rotational reference, the quantities are visualized in fig. 8.

17
 Figure 8 – Simplified illustration of the quantities describing the operation of a synchron-
ous generator, including a rotor and an electrical rotation frame.

Furthermore, eq. 38 can, when considering small deviations (∆) from nominal values
(subscript 0)

P = P0 + ∆P,
τ = τ0 + ∆τ, (39)
ωr = ω0 + ∆ωr ,

be written as
d∆ωr ∆Pm − ∆Pe
= , (40)
dt 2H
giving a relationship between power deviations and deviations (from synchronous rota-
tional speed) in rotor rotational speed [41]. Eq. 40 is represented visually in fig 9 to
show the impact of inertia on frequency response.

18
 Figure 9 – Visualisation of the frequency response as a function of inertia and power
deviations.

4.5 Gas turbine control overview

The GT control system is responsible for maintaining desired operation of the machine
through fuel and air flow. The control system needs to monitor and adjust temperatures
and speed in order to deliver the required power but also reduce the wear of parts.
The control system of a GT consists of several sensors, controllers and actuators in
order to achieve optimal operation. The main sensor inputs are temperatures, pressures
throughout different locations of the GT and turbine shaft rotatinal speed [45]. The
signals are compared to reference signals (set points) and the result is fed through a low
signal selector and the actuators controlling fuel valves and VGVs.
The output of the closed loop control system is achieved by proportional (P), integral
(I) or derivative (D) or a combination of the aforementioned. The controller compares
the sensor value to a reference (e.g. maximum temperature) and the outputs (u) are
described by the following [48]

up = Kp  + u0 ,
1 t
Z
ui = dt, (41)
Ti 0
d
ud = Td
dt
where subscripts p, i and d represents the type of controller. In eqs. 41,  represents
the error, uo the default output signal and Kp , Ti and Td constants. It can be noted
from the equation for the P controller that using this type only will result in a residual
error since the output signal for a given error will be constant. In order to eliminate the
error the I part is included, while the P part is proportional only to the rate of change
of error, thus increasing output if the error is increasing rapidly. In order to achieve the
correct governor output signal all measurements are compared to their set point and sent

19
through a minimum value selector. The principle is shown as a block diagram in fig. 10,
where three controllers each responsible for different quantities (e.g temperature, speed
and power) are compared and sent to the GT depending on the magnitude of deviation
from the set point.

Figure 10 – Block diagram of the min value selection principle, here SP refers to set
points of a given quantity and the GT output represents the corresponding quantity for
each controller.

4.6 SGT-750 and SGT-800 control systems

The layouts of the SGT-750 and SGT-800 governors differ slightly, the following sections
will serve to describe the basics of the implemented control systems, since these, besides
the physical properties of the GT have a great impact on the GT dynamic response. Since
the control systems are proprietary, only the overarching principles will be discussed and
as such some details are left out.

4.6.1 SGT-750 control system

The SGT-750 governor is comprised of several controllers, a brief description of their
function follows.
Starting with the FLC (frequency load controller) controller block two main con-
trol methods are used for active power and frequency control, isochronous (constant
frequency) and droop control, the principle is shown in fig. 11. The frequency control
mode allows a turbine to adapt the power to the frequency present, thus the load on

20
each present generating unit is determined by the maximum power output and the droop
[37]. As a consequence, assuming two or more machines have the same droop setting,
the larger machine will deliver more power as is apparent in fig.11.
Isochronous control (shown in fig. 11), on the other hand, will deliver the set point
power independent of the frequency. In turn, isochronous control may present problems
when used on several units connected to the same system since they will compete to get
system frequency to its own setting, requiring precisely the same measured frequency
for stable operation [41]. In order to operate units in parallel governors are provided
with droop control. The droop characteristics are shown in fig. 11, as shown, this
characteristic makes the speed droops as load is increased and vice versa. On smaller
local grids (island grids) frequency needs to be sustained by the present machines, and
the power output is decided by the present loads.

Figure 11 – Characteristics of isochronous and droop control.

The regulator droop in percent is defined as

∆f
Droop ≡ δρ = [%], (42)
∆P
here ∆f is interchangeable with ∆ω. In practice the droop value dictates the change in
power output (or frequency) at a given frequency (or load), this allows turbines working
in parallel to find a stable working frequency as they share a load. The droop setting
on gas turbines working on island grids as sole power producers will, however, cause the
frequency to change depending on the present load. Since synchronous machines are

21
forced to work on the same frequency, machines of different sizes and/or with different
droop values may share loads unevenly. It is clear that not all machines can work in
isochrnous (frequency control) mode, due to the previously mentioned reasons, i.e. one
machine might end up producing all power, while the other starts slowing down. In order
to operate several machines on a smaller grid the speed setting of the controller can be
changed thus moving the power-frequency equilibrium [41]. The speed change setting
influence is shown in fig. 12 representing an increase in power output and a change in
speed setting. In fig 12 it is shown how the frequency can be maintained at nominal
value while using droop control.

Figure 12 – Speed change in droop control mode, line A represents the initial droop,
line B represents the new speed setting resulting in the same frequency at higher power
output.

The load controller simply adjust the speed through feedback of the power set point. If
Yt represents the output signal (in the time domain), or the deviation, then5

Yt = (fsp − f ) + δρ δf (Psp − P ), (43)

where subscript sp represents the set point value and δf is the droop factor representing

f0
δf = . (44)
Pmax
From this it is clear that when the generator via the turbine delivers the required power
Psp the droop will not restrict the output signal, thus allowing the turbine to return to
nominal frequency. Eq. 43 also shows the impact of droop when the power set point is
5
This formulation does not express response time or any delays.

22
unchanged causing additional influence on the output signal. The signals entering the
FLC are sensor values and restrictions calculated by other controller blocks. In the FLC
the fuel flow set point is determined by a PI-controller depending on the deviations of
f and/or P . The deviation itself, devi , and the hf svi is fed to the selector unit which
selects the hf svi with the lowest corresponding devi .
The start control, STC , is responsible for reliable starts and to avoid damage. The
STC limits acceleration of the machine and hence limits the thermal stress. The inputs of
the STC are monitored and kept within limits of the specific machine until the required
speed of the gas generator is achieved, once this criteria is achieved the NGGL takes over
the operation.
The gas generator speed limiter NGGL, as the name suggests, is responsible for
limiting the speed of the gas generator shaft. The STC unit sets the reference value of
ngg and based on the deviation a PID controller outputs the fuel flow set point.
The temperature limiter T800L is fed values of the TIT and the exhaust temperature.
The output signal (set point) is based on the deviation from the temperature set points.
Inside of the temperature limiter are two PI controllers, for each of the temperatures,
the deviations from both are compared and the smaller is passed out of the block to the
main min value selector.
The pressure ratio controller, πC , takes in measurements of the pressure ratio π and
the normalized speed, s
Tref
nggnorm = ngg , (45)
(Tin,comp + Ck )
where Tref is a reference temperature and Ck = 273.15 K. The normalized gas generator
speed is transformed through interpolation to the corresponding pressure ratio value and
compared to measured pressure ratio value through a PI controller. In addition the
nngnorm value affects the variable guide vanes in the VGVC by interpolation.
Furthermore the load loss detection block, LLD, measures rate of change of power
and signals the other controllers in order to bypass them and reduce the fuel flow to lean
flame sustain level. The block measures the rate of change of fuel flow and at a critical
value activates the output signal. The signal will also manipulate the bleed valves and
the gas servos in order to avoid surge. The Selector unit, described previously, passes
its output to the flame sustain control, FSC , which, in normal operation passes this
signal to the fuel valves. The FSC is also responsible for setting the lower limit on the
fuel flow set point based on an interpolation of nngnorm . which block decides the fuel
flow.

4.6.2 SGT-800 control system

The control system of the SGT-800 GT shares many similarities with that of the SGT-
750, however, the differences warranties a separate section in order to differentiate. The
differences between the two arise, mainly from the fact that the SGT-800 consists of
a single shaft only. Some of the obvious consequences of the single shaft is that the

23
GT cannot deliver any torque at zero speed, neither can it change the speed of the gas
generator, since it also is the power turbine and needs to be kept within a strict frequency.
First of is the T52 limiter , this unit serves to limit the TIT. The TIT is compared
to the measured value and the deviation and calculated required heat flow (from a PID
controller) is sent to the selector as in the SGT-750. The required heat flow is bound
by a lower and upper limit calculated based on several of the conditions measured. The
calculation of TIT reference is done through interpolation of a predetermined table. The
T52 limiter will serve to protect the turbine from over heating.
The tCON , temperature controller, block adjusts heat flow based on exhaust tem-
perature and TIT. The block contains a minimum value selector in which the measured
value is compared to the set point for both. The tCON is responsible in adjusting VGV
positioning to affect the TIT. The set point value for exhaust temperature is based on
several parameters. At idle and and part load the output will be based on the exhaust
temperature deviation until the the value has reached the set point, when the tCON will
adjust the turbine inlet temperature by adjusting the VGVs.
The BleedC controls the bleed valves to prevent choking of the later compressor
stages. At lower rpm the bleed valves are open, and then closes as the rpm approaches
nominal value.
The MLC , maximum load controller, limits the output power of the GT and works on
the same deviation to selector principle as many of the other controllers. The maximum
allowed power output differs slightly between gas and liquid fuels.
Depending on the inlet temperature and pressure the normalized electrical power is
calculated in the Pel Norm. block.
The FLC , FSC and LLD works very similar to those of the SGT-750 with slight
differences due to the difference in single and dual shaft design.
In the SGT-800 the upper and lower fuel flow limit is set by the tCON and the FSC.

24
5 Method
5.1 Procedure
The investigation of GT performance in island grid situations commenced with the design
of a fitting model. The modeling was done in the modelling and simulation environment
Dymola, where in house models for the SGT-750 and SGT-800 designed by Siemens were
available. Simulations of the system were compared to data from the factory test rig and
on site data and the models were altered to reproduce the test results. The alterations
were simply adjustments of certain values recorded by the sensors at the time of the
measurement, apart from those corrections a heat flow limiter, discussed in detail in
later sections and a rate limiter of the power set point also discussed in coming sections
were added.

5.2 Modelling
In order to simulate the turbine response an island system capable of connecting the
auxiliary turbine models was designed. The system designed included generators, a load
table and active load sharing logic. An overview schematic is shown in fig 13. The blocks
shown in fig. 13 are mathematical operations of the main quantities entering and leaving.
The island setup, includes blocks 1 through 4 in fig. 13, blocks 1-3 were constructed as
generic components, while block 4 needs to be adjusted depending on the gearbox and
generator present. The quantities that was adjusted in order to match measurement data
were: losses in the generator and gears, fuel system pressure (in accordance with those
measured in the test rig data), inertia of the generator and gears and ambient conditions.
The adjusted values were all obtained from measurement data and data sheets, the gear
losses were accounted for as

Wgearloss = α + βN γ + δWin , (46)

where α, β, γ, δ are constants for polynomial fitting, N is rpm and Win is the shaft
power. The model can be explained as a mechanical torque calculated based on a user
defined active load which is shared between two or more turbines equally. The torque
produced, which is divided provides the electrical torque τe from the torque balance
equation discussed earlier.

25
 Figure 13 – Simplified schematic of the turbine island setup and the main quantities
entering and leaving each block, not including logical signals.

Since the focus of the simulations were turbine performance a simple generator model
was implemented, disregarding voltage control. The control system of the turbines were
modeled to receive the load power value as the power set point in order to be able to
operate with droop, without a residual frequency drop. The generator model was set to
a power factor of cosϕ = 0.9 and the losses was extrapolated from a map of values for
the generator. The inertia of the machine, the gear losses and the generator loss map
was based on the generator used in each respective case.

5.3 Dymola model

The Dymola model used consisted of four main blocks as touched upon previously. The
active load on the can easily be set via a time table or ramp function. The torque that
arises from the given load is calculated by the island torque block seen in fig. 14.

26
 Figure 14 – The island torque block calculating the resulting torque from a given load.

The total torque given the load is then shared equally between the turbines connected via
the load share block depicted in fig. 15. Here we note that pink lines represent boolean
signals (1 or 0). In the load sharing block switches, 1 through 3, are activated as soon
as the turbines are ready to sync and then delivers three equal torque values. Within
the load sharing block there are also three switches for tripping the turbines. The trip
switches are governed by timers and once one of the turbines trip the integer value of
the denominator governing the fraction of torque to each turbine is increased evenly by
the amount of power the tripping unit was delivering.

27
 Figure 15 – The load sharing block for three turbines containing trip switches, fractions
of torque to each turbine and a load set point output fet to the control systems of the
turbines.

The fraction of torque to acting on each turbine is then routed to the generator block
shown in fig.16 in which the numerical values for torque are connected to a flange ad-
apter which in turn is connected to each of the generators. The generator flanges are
then externally connected to the turbine shafts. The generators, as described earlier,
are simple models taking into account losses regarding cosϕ, gears and mechanical, e.g
friction. The power delivered and the frequency of the generators are fed back to the GT
control systems as measurement values.

28
 Figure 16 – The generator block for a triple GT island grid where the torque values from
the load share block is converter through flanges to produce a mechanical torque on the
generators.

All blocks described are contained within the island grid block seen in fig.17. This
block is connected to the GTs via shaft flanges and it feeds the turbine governors the
power output, the frequency and the power set point.

29
 Figure 17 – Island Grid Dymola diagram, consisting all components but the GTs which
are connected from without.

5.4 Validation method

The constructed model was, in order to verify its reliability, compared to aforementioned
data. Ambient conditions, relative humidity, pressure ,temperature, and fuel pressure
was matched. Data for the present generator and gearbox was, from measurements and
data sheets, fed to the model as was the gas system pressure and volumes of the gas
system. The available measurement data was limited and scenarios had to be matched
closely in order to be comparable. The available data contained heat flow (hf) in MJ/s,
valve opening in % of max, velocity in RPM and controller setting amongst more. The

30
controller data settings from the test runs indicated that, in one of the validation cases,
the SGT-750 load increase run, the maximum load controller limited the heat flow, thus
altering the transient response. This sudden controller discrepancy, was accounted for
simulated by adding a heat flow limiter block as seen in fig. 18.

Figure 18 – Dymola diagram of the heat flow limiter chocking the heat flow set point in
a given time interval to duplicate the available data.

The heat flow limiter was tuned based on the heat flow in the test rig data. The slew
rate limit (SRL) was calculated as

∆hf
SRL = , (47)
∆t

in the time interval ∆t where the hf increases by ∆y shown in fig.19

Since grid frequency stability is of essence in island grid configurations the normalized

31
 Figure 19 – Heatflow of the SGT-750 when load is increased from 11-18 MW.

rpm (npu ) of the simulations

n
npu = , (48)
n0
were bench marked to the test rig data. The Dymola models were also compared to
MATLAB Simulink models, which contained more sophisticated generator models (but
less detailed GT models) in order to evaluate potential differences. Measurement data
for heat flow, initial rotational speed and loads were closely matched in order to ensure
correct response to input of the model.

5.5 Simulation case studies

Several cases were simulated to determine the SGT-750 and SGT-800s potential for island
operation. The cases considered consisted of no additional producers, hence no additional
inertia. This in term is an extreme case where the GTs present are the only means of
frequency and load control. The main Island grid simulations cases are presented in tab.
3. In addition to those presented in tab. 3 simulations were carried out to determine the
maximum instant load increase the GTs were capable of. The transient occurrence in each
simulated case was set to take place once steady state operation was reached for a given
initial load. All case study simulations were tested at three different ambient setting:
hot, where Tambient =30o C, ISO, where Tambient =15o C and cold, where Tambient =-30o C.
In all ambiant cases the relative humidity was set to 60 % and ambient pressure to 101.3
kPa.
The implementation of the case studies, including the maximum load capability tests,
consisted of numerous simulations and monitoring of engine parameters. Once the sim-
ulations were completed the limitations and capabilities were determined.
Apart from the main cases, simulations of different fuel compositions were to be eval-
uated through simulations using the same models. No modifications were made to the
gas turbine or auxiliary systems aside from the addition of a fuel and chemical reaction
function taking place in the combustion chamber. A second chemical reactions function

32
 Table 3 – Main island grid cases simulated, the final power output was Pmax at ISO, if
not otherwise stated.

Case Pinit [MW] Descpription

A-750 0.5Pmax Two SGT-750 one trips
B-750 0.33Pmax Three SGT-750 one trips
A-800 0.5Pmax Two SGT-800 one trips
B-800 0.33Pmax Three SGT-800 one trips

was added for the reactants and the products of hydrogen fuels working on the same
stochiometric matrix principle as the the regular NG-function, the function is simply cal-
culating chemical balance of reactants and products. The combustion, in short, calculates
the energy of the reactants and products to deliver power to the turbine. Simulations
were conducted with fuels consisting of 25, 50, 75 and 100% hydrogen by volume, the ex-
act fuel compositions are presented in tab. AI.1 in appendix. The simulations hydrogen
fuel regarded maximum instant load capability.

33
6 Results
6.1 Results of validation cases
The described procedures were conducted in order to validate the models, the cases are
presented in tab. 4.

Table 4 – List of validation cases carried out.

GT model Validation
Model ∆ P MW Comment Figure
SGT-750 11-18 n0 = 5900, /w hf limiter 20
39-18 n0 = 6100 21
SGT-800 15-24-38 n0 = 6610 22
24-15 n0 = 6610 23

Since the frequency of the turbine will reflect the overall dynamic response of the
GT, and is of outmost important for stable grid operation, this quantity serves as the
benchmark for correlation. As is apparent in figs. 20-23, the simulations, while corres-
ponding in time, slightly under predicts the magnitudes of the responses. In fig. 20 e.g.
the maximum frequency drop of the simulation is about 0.975n0 while the test rig data
shows a minimum frequency of about 0.965n0 . The maximum frequency drops, however,
occur after the same amount of time has passed since the load was increased, this goes
for all validation cases.

Figure 20 – Dynamic response to load increase 11-19 MW of the SGT-750 with heat flow
limiting using the test rig gearbox and generator.

The validation cases of the SGT-750 were compared to data from the test rig, in which
the gearbox and generators are slightly heavier than the standard package. The case
depicted in fig. 20 utilizes the heat flow limited discussed in previous section to replicate
the conditions of the test. The results for the SGT-750 simulations shows an under
prediction of magnitude < 1% of f0 at the point of greatest deviation.

34
 Figure 21 – Dynamic response to load decrease 39-18 MW of the SGT-750, highlighting
the point of shutdown of the test rig GT, explaining the large deviation after the fact.

The validations of the SGT-800 (figs.22 and 23) were carried out with data where stand-
ard gears and generators were used. Once more there were slight under predictions by
the Dymola model. Here the greatest discrepancy between the model and simulations
were of the magnitude < 1 % of f0 .

Figure 22 – Dynamic response to load increase 15-24-38 MW of the SGT-800.

35
 Figure 23 – Dynamic response to load decrease 24-15 MW of the SGT-800.

Validations carried out comparing the Dymola model to the MATLAB Simulink
model (see appendix II fig. AI.1) showed that the Simulink model under predicted
instant load increase for the SGT-750. However, it marginally over predicted frequency
drop from a load increase of the SGT-800.

6.2 Case studies

6.2.1 Instant load increases under different ambient conditions.
In order to evaluate the maximum cable load increase on the GTs, simulation were run
under three different ambient settings, all at an initial power output of 10 MW, the
results are displayed in fig. 24.
For the SGT-750 the maximum capability is at -30 o C with a capability of a 30
MW instant load increase resulting in a minimum frequency of 0.87f0 , at higher ambient
setting the capability decreases somewhat, resulting in a minimum frequency of 0.89 f0
for a 25 MW increase at -15 o C and a minimum frequency of 0.91 f0 for a 20 MW increase
at 30 o C.
The identical, ambient condition procedure, was carried out for the SGT-800. The
results shown in fig. 24 shows very slight differences between the different ambient
conditions, although we see that the capability is highest at -30 o C where the SGT-800 is
capable of a 50 MW increase resulting in a minimum frequency of 0.96 f0 . The other two
cases results in a minimum frequency of 0.95 f0 at the maximum instant load increase
of 40 MW.
For capability comparisons the maximum frequency drop at a given fraction of rated
power of both GTs is shown in fig. 25. Within the displayed intervals ambient temper-
ature has little impact on the maximum frequency drop, note however that the SGT-750
data at ISO and 30o C conditions only contain two and three data points respectively.

36
 Figure 24 – Maximum instantaneous load increases for both GTs at different ambient
settings.

Figure 25 – Maximum frequency drop at instantaneous load increases at different ambient

conditions with Pi =10 MW corresponding to 0.25Prated for the SGT-750 and 0.17Prated
for the SGT-800.

From fig. 25 a linear fit can be formed as

∆nmax = κ∆P (pu) + m, (49)
where κ = ∆n/∆P (pu), empirically, can describe the sensitivity to load increases so that
∆n(pu)max = fκ (P (pu)). Based on the linear fit the SGT-750 shows κ750 =0.19, m750 =
-0.0084 (r2 > 0.99) and the SGT-800 κ800 =0.052, m800 =-0.0017 (r2 > 0.99).

37
 Figure 26 – Frequency response of all the grid-trip cases, here c* denotes the maximum
power for SGT-750 at -30o C, i.e. not at ISO conditions.

6.2.2 Island grid cases

Island grids of two and three turbines as specified in 3 were simulated, the results in the
form of frequency responses are presented in fig. 26. For the SGT-750 we see that at
ISO condition the minimum frequency is 0.919f0 and at -30o C the minimum frequency
0.91f0 . In fig.26 the line denoted with c* represents operation at higher output, since
the power rating is inversely proportional to the ambient temperature. This phenomena
goes for all GTs but the special notation was only used since the SGT-750 at -30o C
was also ran at power rating at ISO conditions power rating, i.e slightly lower. The
fκ,750 -linear fit for the SGT-750 -30o C case predicts fmin = 0.91. At 30 o C we see that
the SGT-750 is able to recover a load increase of 0.5Prated in the process dropping to
0.85f0 . In addition the recovery requires a longer time span (20 s.), relative to the other
conditions. The B-750 case is managed at all ambient conditions with fmin ≈ 0.97f0 , the
fκ,750 predicts fmin ≈ 0.98f0 . The island grid cases for the SGT-800 (26 A-800 and B-
800), shows that it is capable of prevailing through both cases at all the ambient setting
considered. The A-800 shows that the a load increase of 0.5Prated results in a similar
response at all ambient temperatures where the frequency drops to 0.975f0 , predicted by
fκ,800 to be fmin ≈ 0.97f0 . the B-800 case shows fmin ≈ 0.992f0 predicted by fκ,800 to
be fmin ≈ 0.99f0 .

6.2.3 Hydrogen fuel cases

Simulations regarding maximum load increase capability were repeated using different
fuel mixes containing hydrogen. A total number of 48 simulations were carried out for
each GT, the complete set of results are displayed in figs.27-28. The cases using 25, 50

38
and 100% H2 by volume showed similar results, in terms of frequency response, as the
NG cases. However, the 75% by volume (BV) case was only able to handle a 10 MW
increase in the SGT-750 simulations

Figure 27 – SGT-750 hydrogen mix fuels at different ambient conditions at instant load
increases.

Figure 28 – SGT-800 hydrogen mix fuels at different ambient conditions at instant load
increases.

39
6.3 Remarks on the results
We start the analysis of the results by looking at the maximum load cases, where the
SGT-750 showed a maximum instant load increase capability of 20 MW (>0.5Prated ) in
all conditions considered, and higher capabilities in lower ambient temperatures, see fig.
25. The SGT-800 was capable of recovering from a 40 MW (>0.7Prated ) instant load
increase in all ambient settings as seen in fig. 25. In the simulations, as in practical island
operations settings, violations of the maximum temperatures were allowed over short
periods of time. The most extreme cases are shown in fig29 (the A-750 and A-800 at
30o C), where we see short periods in which temperatures over the maximum operational
value (in the figure 1 pu) occur. In fig. 29 it is apparent that the gas generator speed is
kept below or at max at all time, while the TIT slightly overshoots the operation limit
for the SGT-750. In the SGT-800 the TIT and TOT overshoots for a time period of little
less than 10 seconds before settling the allowed operational temperature as the frequency
recovers.

40
 Figure 29 – GT quantities at island case where one unit trips at half power.

6.3.1 Limiting factors for the SGT-750

In the the ISO case of the SGT-750 (fig. 26), when instantaneously loading 25 MW
(or >0.65%Prated) the GG shaft reaches its normalized speed limit while at the same
time the TIT (which can be overlooked in these timescales) reaches its maximum allowed
value this in turn causes the longer and deeper frequency dip. At -30o C when adding
30 MW of load the speed limiter is engaged for a brief period of time, preventing over
speeding and then recovering without remarks. In the +30o C case however the nggl
system limits power output for a longer period of time which is apparent from the long
and deep frequency dip in fig. 24.

41
6.3.2 Limiting factors for the SGT-800 maximum instant load increase
The results of the SGT-800 simulations show lesser frequency dips but similar recovery
times, the recovery from the local minimum is slightly more level. The ambient temper-
ature does not impact the magnitude nor the recovery time substantially, this is in part
due to the tolerance imposed on short bursts of over temperatures. Slight oscillations
occur at -30o C, possibly due to the slight oscillations of the VGVs shown in fig.30.

Figure 30 – Per unit VGV opening at different ambient conditions for the SGT-800 in
scenario A-800.

6.3.3 Hydrogen simulation results

All but the 75%-H2 mix showed similar results as the NG simulations. The short answer
to why this is can probably be found in the similarities of Iw allowing the same amount
of energy flow through the constant dimension fuel system, valves and CC. This is em-
phasized by the fact that the 75%-H2 is not capable of larger load increases. The cause is
the high wt% of inert gas (N2 ) through an undersized and under pressurized fuel system.
The temperatures were slightly lower when the amount of inert gases were high. The
results of the hydrogen simulations can be summarized as follows: the 25, 50 and 100%
hydrogen fuels presented similar results in regards to frequency response, especially for
the SGT-800. We can see from fig. 31 that the main fuel valve opening is capable of
counteracting the lower energy density, by allowing a greater volume flow of fuel gas and
higher rate of valve opening. The simulations suggest that the fuel flow is the limiting
factor for the SGT-800.

42
 Figure 31 – Fuel valve opening comparison for NG and 75% H2 (by volume) for the
SGT-800 at instant load increase (10-40MW).

This was further investigated by increasing the fuel pressure by 15 %. The results
in fig. 32 showed that the pressure increase to some extent could mitigate the problems
with a low energy density fuel.

Figure 32 – Fuel valve opening and frequency response (10.50 MW load increase) for
nominal and 15 % increased fuel pressure.

The SGT-750 had a more violent frequency response for the 100% H2 fuel at 30o C
ambient temperature at 10-30 and 10-35 MW increases, in these cases the speed limiter
is in operation which decreases the output causing a larger frequency dip and a longer
recovery time. No significant changes in temperatures were noted when using hydrogen
fuels.

6.4 On the differences of the machines

From the simulations we have seen that the SGT-750, in terms of magnitude of frequency
drop, is more sensitive to load increases, even when the rated power is taken into consid-
eration. From sec. 4.4, we have seen that the change in frequency is dependent on the
inertia of the machine. A simple simulation,using the block diagram seen in fig.33, of the
swing equation (eq. 40) can give indication of the influence of inertia on the frequency

43
drop. In the swing equation simulations the GT power output is simplified as a first
1
order transfer function on the form G(s) = 1+T ts
where Tt is a time constant.

Figure 33 – Simulink block diagram for simulating the swing equation.

In the simple simulations we can vary the inertial constant H (1<H<10) and the time
constant Tt (1<Tt <5) which yields the results in fig. 34.

Figure 34 – Swing equation simulation results when varying inertial constant H (upper)
and time constant of the turbine Tt (lower).

44
7 Discussion
The goal of this thesis was to develop a model for island grid scenario simulations, validate
the simulations with the help of measurement data and further investigate the capabilities
of the Siemens SGT-750 and SGT-800 in such settings. The models implemented and
simulations carried has proven to be an easy, rather quick, cheap and effective way to
predict GT performance of the specified cases. The results adds to the body of, and
attests to the capability of simulations as previous publications have shown [19, 20, 22,
26].

7.1 The island grid

An island grid system was modelled and four basic cases were validated to real life test
data showing similar results in frequency response (figs. 20-23). Since the main purpose
was to evaluate the dynamics of the gas turbines, the grid designed (figs.14-17) is not
fitting for a deeper analysis of the power system dynamics, including active power and
voltage control since these factors have been neglected. An additional weakness of the
model limiting its applicability, is the fact that no additional rotational machinery, e.g
synchronous motors, are present. This may have caused greater frequency dips than what
would have occurred if additional rotational machines were present contributing inertia.
For e.g in [25] the power system tested had an inertial constant (H) of 6 s, which could
be argued as reasonable, however in this thesis it could detract from the goal of solely
evaluating the GTs performance. The results from the simulations carried out in this
thesis will be representative of smaller grids consisting solely of GTs supplying power, such
as off shore applications and industry power supply. For local power grids, specifics of
additional suppliers, predicted demand, start time and more need be considered, however,
the model can provide useful predictions of the limits of the GTs for such applications.
The complete model for the simulation, however, included highly specific GT models
which were able to predict real life behaviour of a myriad of parameters. The frequency
response as an indicator was studied, since it will be of utmost importance for end
users and will be a result of the total dynamics of the GT. The fact that the Dymola
and Simulink results presented similar results despite the simple generator models may
further serve to justify the reliability of the model. Specifically the similarities between
the Dymola and Simulink results are shown in fig. AI.1 where it is seen that the differences
in frequency response are of the magnitude 0.01n0 .

7.2 Maximum capability simulations

Some of the main consequences and causes of the results were mentioned in the result
section (sec. 6.3), a discussion on the results will follow.
We start by noting that this work has focused on two specific GT models, it has
successfully evaluated their potential in an island grid setting. As a consequence of
using highly specific models not all results are generalizable. It is, however, clear that:
The inertia of the GT has a substantial impact on maximum frequency drop and that

45
temperatures, in particular TIT, are limiting factors when increasing loads. It is not
surprising that this is the case, since GTs are designed for high temperatures in order
to maximize efficiency and power output. Were the turbines to accept much higher
temperatures, their rated power outputs would be higher. In an attempt to quantify the
frequency sensitivity to load increases we introduced the machine specific κ-parameter
which was empirically calculated. The κ-parameter can be seen, merely, as an uncertain
predictor of frequency drop of a GT for a given fraction of maximal power. As soon as
events occur that evokes sudden responses in the control system or the turbine leaves
stable operation this a linear κ function will not be a reliable predictor. For that reason
we can speculate that a κ presenting a high linearity (e.g r2 ≈ 1) is indicative of operation
within a region stable for the machine. In particular κ was linear within regions where
only the FLC was governing the engine. This and and the following discussion points to
the fact that κ ∝ H.
Moreover, the bare bones model shown in fig.33 was presented in order to get a general
understanding of the impact of rotor inertia and a turbine time constant Tt on frequency
response. The swing equation simulations (fig.34), assumes the behaviour of the GT can
be represented by a first order transfer function of one time constant Tt , although this is a
simplification that was not derived, the results gives indications of the impacts of inertia
and GT dynamic response time. We see that an increase in inertia only decreases the
magnitude of the frequency drop at constant Tt with diminishing returns. The value of
Tt however has an impact on the timescale of recovery and oscillations. These properties
ties into the simulation results in that the time of recovery is rather similar for both
machines indicating advantageous control response and machine dynamics. In the cases
where the governor limits the energy flow, however, the recovery time is effected (eg. fig.
24).
The results overall, implies that the SGT-750 and SGT-800 are well suited for island
operation and provides resilience to large power fluctuations. Their resilience is shown
by their instant load increase capabilities which are over 50% of their rated power re-
spectively, and as seen in fig.25 they both recover from these load increases in around
10 s. The SGT-800 allows for smooth operation over a wide range conditions. Although
the SGT-750 does not provide the same load increase capability and frequency stability,
in absolute terms, as the SGT-800 the relative capabilities are rather similar. The res-
ults would indicate that both the SGT-750 and SGT-800, on a reasonably dimensioned
power grid, would persevere through heavy transients. This is evident from thre results
in fig.26. We note that the line marked T?-30o c*, is an out liar. This simulation took
into consideration the increased capability at lower ambient temperatures, for the SGT-
750 approximately 15 % and the GT was not able to recover. We can argue, however,
that this event only would occur if the system was under dimensioned. The logic goes
that a power system of two GTs, designed to handle a trip out of half the power pro-
duction without a total blackout, would not calculate the rated power output based on
arctic conditions. This is due to the fact that such a power system only would be able
to deliver full rated power at temperatures below -30o C, which would seem unlikely.
The intrinsic differences between single and double shaft GTs, such as the torque/speed

46
relationship [37] and efficiency at lower loads and, again, inertia must also be taken into
consideration when considering the application [21]. Furthermore we see that the limiting
factor of load increase will ultimately be the temperature limit, this fact has been pointed
out in the literature ([22, 23, 25, 29]), and implies the model acts realistically. One of
the main limitations of GTs are the turbine materials resistance to extreme temperatures
[40, 45], thus GTs are designed to run as hot as possible without causing unnecessary
stress. High TIT increases the efficiency ([17, 40, 45]) and therefore it is desirable that the
operational temperature is kept high, for this reason it is easy to see why the temperature
controller will be the limiting factor in violent transient events.

7.3 Hydrogen simulation results

The hydrogen simulation cases (results in figs. 27, 28) showed overall acceptable opera-
tion. The lower fraction hydrogen mixes behaved similarly to the NG operation, which
can be traced back to the energy density and Wobbe index. In the simulations the CC
was simply fed the required energy rate resulting in greater opening of the fuel valve and
thus a higher volume rate. This fact was emphasized when increasing the fuel pressure
(fig. 32), the higher pressure allowed for a quicker response, in turn delivering a higher
energy rate. The higher degree of opening of the fuel valves may also make controlling
the flow accurately difficult due to the characteristics of valves. Operation on low Wobbe
index gasses may as such require special valves and tuning of the control system.
The rather simple combustion chamber of the model does not take into consideration
intricate physics of combustion e.g heat transfer, laminar flame speed etc. and the
analysis of such mechanisms are beyond the scope of this thesis.
Lower concentration hydrogen fuel tests have been successfully conducted at Siemens
and are being successfully investigated, higher amounts H2 fuels still require compre-
hensive research and tests however. The results produced summarily points out the
requirements for higher pressure systems and/or larger volume components for operation
on high inert gas content fuels. These procedures may be costly, since larger volumes
may require redesign of some components,high pressure tanks and compressors. Further
investigations regarding the formation of NOx should also be considered in order to get
a complete outlook on the matter [36].
Aside from the possible challenges in GT design, the problems of production, storage
and transportation remains [32, 33]. For hydrogen to become a realistic option capable
of replacing fossil fuels, these problems need to be solved.

7.4 Reflections and areas of improvement

Allowing temperatures that exceeds the maximum operational is common practice in
certain cases during transients. This, however, may cause uncertainty on the implications
of the results concerning the higher load increases (in particular the high temperature
grid trip cases A-800 in fig. 26). The control systems may be tuned in order to allow
such events, however, in this thesis no additional tuning of the control systems were
performed. In fig. 30, we see some leeway after the initial peak and although the VGV

47
position is a function of several variables, this may give us a clue that it is not perfectly
tuned, since a more aggressive opening of the VGVs might have lowered the temperature,
which is desirable for longevity and stable operation. Although these events only took
place for short periods of time (<10 s.), a majority of which was in single digit percent
above the operational limit. They should be accounted for and quantified in the control
system model more meticulously in order to give a definite limit in terms of time and
temperature. This could help in approximating the additional wear on the components.
Although the validation scenarios of the model showed good correlation (sec. 6.1), both to
actual site data and previous Simulink models, a larger set of data would be preferred to
investigate how more extreme cases correlate to the simulations. While a GT is a complex
and highly dynamic system, the dynamics are well understood and in detail described at
length in the literature (eg. [40, 45]) and mathematical simulation models have proven
time again successful at predicting transient behaviour (e.g [20, 21, 23, 26, 29]). From
sec. 6.4, we saw that the the time constant Tt had substantial impact of the time frame of
frequency fluctuations. The fact that the comparisons in figs. 20-23, corresponds closely
in time to the test rig data (i.e peaks and valleys in frequency occur at similar times)
indicates that the models are successful in capturing the fundamental dynamics of the
GTs they represent. The high specificity of the SGT-750 and SGT-800 Dymola models
and the results presented in conjunction with the theoretical and practical knowledge
on GTs in general speaks to the validity of the results. Taking all discussed points into
consideration, the results are representative, although a deeper analysis of the controls
such as the response time of the TIT controller would be preferred.

48
8 Conclusions
The work presented in this thesis set out to construct an island grid model to evaluate
gas turbine performance during transients. Investigations of the effects of hydrogen fuel
implementation in transient events were also to be covered. The model was built, suc-
cessfully tested and validated against on site recorded data and other GT models. Several
specified simulation cases were carried out and the results can be summarized by the fol-
lowing conclusions: (i) The Siemens SGT-750 and SGT-800 gas turbines are, according
to the finding, capable of handling sudden power increases corresponding to >50 % of the
rated power.(ii) The factors eventually limiting the SGT-750 from responding quicker to
load increases are the speed of the gas generator and TIT, for the SGT-800 the TIT and
TOT will eventually cap out the rate of power increase. This can somewhat be mitigated
by allowing short periods of temperature above maximum operational at transients. (iii)
Hydrogen fuels consisting of 25%, 50% and 100% H2 showed no significant differences
in transient events, however, high contents of inert gas would require higher pressures
and/or enlarged gas system and valves.
The results should be viewed as indications of the capabilities of the SGT-750 and
SGT-800 when subjected to large transients but also a reference to areas of improvement.

49
9 Future work
The vast and complex topic that is gas turbine performance has, in this thesis, been con-
densed and applied to produce concrete results and predictions. This work, nevertheless,
opens the door to new, interesting and important questions. We have provided a general
outlook on the performance of the SGT-750 and SGT-800, future work should immerse
in the specifics left out of this work and and continue developments of expansive project
similar to this one. Material, heat transfer, combustion and CFD research will likely
continue pushing the physical limits of GTs which is crucial to improve performance.
Aside from the vast research going on on these subjects future work on GT performance
are suggested to include: A Dymola implementation and validation of a leading governor
unit. This unit should be capable of alerting several units if instability or trips occur in
order to prime the other units and possibly avoid violent loading. A validation of hydro-
gen combustion would be suggested in order to be able to quickly test hydrogen transient
operations in Dymola, this would possibly include a review of the combustion system of
the model. The relative simplicity of the model could also allow an optimization study
of transient handling, this could include VGV operation water or steam injection and
control system parametric studies.

50
References
[1] IEA (2019), "Renewables 2019", IEA, Paris https://www.iea.org/reports/
renewables-2019

[2] Khare, V. Nema, S. Baredar, P., Solar-wind hybrid renewable energy system: A
review, Renewable and Sustainable Energy Reviews 58 23-33, (2016)

[3] Hirsch, A. Parag, Y. Guerrero, J. Microgrids: A review of technologies, key drivers,

and outstanding issues, Renewable and Sustainable Energy Reviews 90, (2018)

[4] Carrasco, J.M. Franquelo, L.G. Bialasiewicz, J.T. Galvan, E. Portillo, R. Prats, M.M.
León, J.I. Moreno, N., Power Electronic Systems for the Grid Integration of Re-
newable Energy Sources: a Survey, IEEE Transactions on Industrial Electronics 53,
(2006)

[5] Aneke, M. Wang, M., Energy storage technologies and real life applications – A state
of the art review, Applied Energy 179, (2016)

[6] Poullikkas, A., A comparative overview of large-scale battery systems for electricity
storage, Renewable and Sustainable Energy Reviews 27, (2013)

[7] Angel Gonzales-Salazar, M. Kirsten, T. Prchlik, L., Review of the operational flex-
ibility and emissions of gas- and coal-fired power plants in a future with growing
renewables, Renewable and Sustainable Energy Reviews 82, (2018)

[8] Nagahara, M. et al., (2012). H Control of Microgrids Involving Gas Turbine Engines
and Batteries. In IEEE 51st IEEE Conference on Decision and Control (CDC). Maui,
HI, USA, 10 December. Maui, HI, USA: IEEE.

[9] Cai H, Wang M, Elgowainy A, Han J. Updated Greenhouse gas and criteria air
pollutant emission factors of the U.S. Electric generating units in 2010. Chicago,IL:
Argonne National Laboratory; 2013

[10] Hellberg, A. Häggmark, A. Junsten, A. Andersson, T. Troger, C. (Siemens Industrial

Turbomachinery AB) SGT-750 - 37 MW GAS TURBINE, 19TH SYMPOSIUM OF
THE INDUSTRIAL APPLICATION OF GAS TURBINES COMMITTEE BANFF,
ALBERTA, CANADA OCTOBER 17-19, (2011)

[11] Siemens, SGT-750 product information. Avaliable: https://new.siemens.com/

global/en/products/energy/power-generation/gas-turbines/sgt-750.html,
retrieved: 2020-04-27

[12] Shukin, S. Annerfeldt, M. Bjorkman, M. Siemens Industrial Turbomachinery AB,

Proceedings of ASME Turbo Expo 2008: Power for Land, Sea and Air GT2008,
Berlin, Germany, June 9-13 (2008).

51
[13] Siemens, SGT-800 product information. Avaliable: https://new.siemens.com/
global/en/products/energy/power-generation/gas-turbines/sgt-800.html,
retrieved: 2020-04-27

[14] Sandhu, Er.M. Thakur, Dr.T., Issues, Challenges, Causes, Impacts and Utilization
of Renewable Energy Sources - Grid Integration, Int. Journal of Engineering Research
and Applications, vol 4, (2014)

[15] Tielens, P. Van Hertem, D., The Relevance of Inertia in Power Systems, Renewable
and Sustainable Energy Reviews, Vol. 55, (2016)

[16] Morren, J. Sjoerd, W.H de Haan. Wind Turbines Emulating Inertia and Supporting
Primary Frequency Control, IEEE transactions on power systems, VOL. 21, NO. 1,
(2006)

[17] Rahman, M.M, Ibrahim, T.K. Abdalla, N., Thermodynamic performance analysis
of gas-turbine power-plant, International Journal of the Physical Sciences Vol. 6(14),
pp. 3539-3550, (2011)

[18] Saif, Er. Mod. Mohammad, Dr. Tariq. PERFORMANCE ANALYSIS OF GAS
TURBINE AT VARYING AMBIENT TEMPERATURE, International Journal of
Mechanical Engineering and Technology (IJMET) Volume 8, Issue 1, January (2017)

[19] Asgari, H. Venturini, M. Chen, X.Q. Sainudiin, R., Modeling and Simulation of the
Transient Behavior of an Industrial Power Plant Gas Turbine, Journal of Engineering
for Gas Turbines and Power Vol 136, (2014)

[20] Panov, V., GASTURBOLIB – SIMULINK LIBRARY FOR GAS TURBINE EN-
GINE MODELLING Proceedings of ASME Turbo Expo 2009: Power for Land, Sea
and Air, June 8-12, Orlando, Florida, USA, (2009)

[21] Bahrami, S. Ghaffari, A. Thern, M., Improving the Transient Performance of the
Gas Turbine by Steam Injection during Frequency Dips, Energies Vol 6, (2013).

[22] Iliescu, S.St. Fagarasan, I. Popescu, V, Soare, C., Gas turbine modelling for load-
frequency control U.P.B. Sci. Bull., Series C, Vol. 70, Iss. 4, (2008)

[23] Hajagos, L.M. Bérubé, G.R UTILITY EXPERIENCE WITH GAS TURBINE
TESTING AND MODELING IEEE Power Engineering Society Winter Meeting. Con-
ference Proceedings, Columbus, OH, USA, 28 Jan - 1 Feb. (2001)

[24] Haselbacher, H. Performance of water/steam injected gas turbine power plants con-
sisting of standard gas turbines and turbo expanders. Int. J. Energy Technol. Pilicy
2005, 3, 12-23, (2005)

[25] Meegahapola, L. Characterisation of gas turbine dynamics during frequency excur-

sions in power networks, Transmission and Distribution, vol. 8, (10) pp. 1733-1743,
(2014),

52
[26] Rowen, W.I. ‘Simplified mathematical representations of heavy-duty gas turbines.
J. Eng. Power, 105, (1), pp. 865–869, (1983)

[27] Ten, C.F. Crossley, P.A., EVALUATION OF ROCOF RELAY PERFORMANCES

ON NETWORKS WITH DISTRIBUTED GENERATION, IET 9th International
Conference on Developments in Power Systems Protection, Glasgow, UK, march,
(2008)

[28] Chown, G.A. Wright, J.G, Van Heerden, R. Coker, M. System inertia and Rate
of Change of Frequency (RoCoF) with increasing non-synchronous renewable energy
penetration, Cigré, 8th Southern Africa Regional Conference, November, (2017)

[29] Kakimoto, N. Baba, K. Performance of Gas Turbine-Based Plants During Frequency

Drops, IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 18, NO. 3, AUGUST
(2003)

[30] O’Sullivan, J. Rogers. A., Studying the Maximum Instantaneous Non-Synchronous

Generation in an Island System – Frequency Stability Challenges in Ireland, IEEE
Transactions on Power Systems, 29 (6), (2014)

[31] Juste, G.L. Hydrogen injection as additional fuel in gas turbine combustor. Evalu-
ation of effects, International Journal of Hydrogen Energy 31 (2006)

[32] Makridis, S. Methane and Hydrogen for Energy Storage, Edition: 1, Chapter: 1,
Publisher: IET Digital Library, Editors: Rupp Carriveau, David S-K. Ting, pp.1-28
(2016).

[33] Nikolaidis, P. Poullikkas, A. A comparative overview of hydrogen production pro-

cesses, Renewable and sustainable Energy Reviews 67 (2017)

[34] Chiesa, P. Lozza, G. Mazzocchi. Using Hydrogen as Gas Turbine Fuel, Journal of
Engineering for Gas Turbines and Power, Vol. 127, January (2005)

[35] Chiesa, P., and Macchi, E., “A Thermodynamic Analysis of Different Options to
Break 60% Electric Efficiency in Combined Cycle Power Plants,”ASME paper GT-
2002-30663 (2002)

[36] Akansu, S.O. Dulger, Z. Kahraman, N. Veziroglu, T.N. Internal combustion engines
fueled by natural gas—hydrogen mixtures, International Journal of Hydrogen Energy
29 (2004)

[37] Genrup, M. Thern, M. Gasturbinteknik Årsrapport 2019 (Eng. Gas turbine techno-
logy annual report 2019), Energiforsk (2019)

[38] Clausius R. "On the Motive Power of Heat, and on the Laws which may be de-
duced from it for the Theory of Heat", Communicated in the Academy of Berlin,
Feb.; Published in Poggendorff ’s Annalen der Physick, March-April. LXXIX, 368,
500.(1850)

53
[39] Cengel, Yunus A., Boles, Michael A. 2014. Thermodynamics: An engineering ap-
proach. Eighth Edition. New York: McGraw-Hill Education.

[40] Saravanamuttoo, H.I.H. Rogers, G.F.C. Cohen, H. Straznicky, P.V. Nix, A.C.Gas
Turbine Theory, Seventh Edition. Pearson education Limited, (2017)

[41] Kundur, P. Power System Stability and Control, McGraw-Hill, Inc. (1994)

[42] Nordling, C. Österman, J. Physics Handbook for science and engineering. Edition
8:14, Studentlitteratur, (2006)

[43] B.R, Munson, D.F, Young, T.H, Okiishi, W.W, Huebsch Fundamentals of fluid
dynamics sixth Ed., pp. 514-515, John Wiley Sons, Pte Ltd (2010), ISBN: 978-0-470-
39881-4

[44] Porretta, B. Porretta, S. Calculation of power systems inertia and frequency re-
sponse, 2018 IEEE Texas Power and Energy Conference (TPEC), Texas, USA, Feb
2018. Frequency response can be

[45] Razak, A.M.Y. Industrial Gas Turbines Performance and operability, Woodhead
Publishing Limited and CRC Press LLC (2007)

[46] EPA (Environmental Protection Agency of the US), Clean Air Technology Cen-
ter (MD-12) Information Transfer and Program Integration Division Office of Air
Quality Planning and Standards U.S. Environmental Protection Agency Research
Triangle Park, North Carolina 27711 Nitrogen Oxides (NOx),Why and How They Are
Controlled, November (1999)

[47] World Health Organization (WHO), Environmental Health criteria, Carbon Monox-
ide (Second Edition), Geneva (1999)

[48] Thomas, B. Modern Reglerteknik (eng. Modern control technology). 5th ed. Liber
AB Stockholm (2016).

54
Appendix I.

Table AI.1 – Fuel compositions and Wobbe index for the fuel gasses considered.

Fuel NG H2 (25) H2 (50) H2 (75) H2 (100)

CH4 wt% 92 84 78 - -
C2 H6 wt% 4.8 7 6 - -
C3 H8 wt% 0.5 5 5 - -
n-C4 H1 0 wt% 0.2 - - - -
N2 wt% 1.5 - - 82 -
CO2 wt% 1 - - - -
H2 wt% - 3 11 18 100
Iw (MJ/Nm3 ) 53 51.2 47.7 17.6 48.3

Figure AI.1 – Comparisons of Simulink and Dymola simulations of instant load increases
(one turbine trips), SGT-800 20-40 MW (upper, SGT-750 15-30 MW (lower), both at ISO
conditions.