2017 Fourth International Conference on Mathematics and Computers in Sciences and in Industry

Generator Model Extension for Higher Accuracy

Simulation of Power System Transients in
OpenModelica

Michael Kyesswa, Hüseyin K. Çakmak, Uwe Kühnapfel, Veit Hagenmeyer

Institute for Applied Computer Science (IAI)
Karlsruhe Institute of Technology (KIT)
P.O. Box 3640, 76021 Karlsruhe, Germany
{michael.kyesswa, hueseyin.cakmak, uwe.kuehnapfel, veit.hagenmeyer}@kit.edu

Abstract— Mathematical modeling is a key step in the analysis response of the system during the transient period following a
of power system stability. In view of model accuracy, this process disturbance.
should include all relevant details that affect the transient
behavior of the power system and hence influence the electrical Dynamic simulations are therefore essential in view of
and mechanical torques in the system. To this aim, this paper power system operation as they aid in analysis of the impact of
describes the extension of the generator subsystem model details potential network contingencies and provide information on the
in OpenModelica by adding a turbine-governor model to the dynamic variations in system operation after occurrence of such
existing generator subsystem. Taking advantage of the multi- disturbances. Utility companies rely on real-time, fast and
domain modeling property of OpenModelica, two libraries are accurate dynamic studies to train network operators, assess the
coupled to interconnect the mechanical and electrical domains of real-time dynamic security of the network and plan short-term
the generator subsystem. The new generator subsystem is operations. Dynamic simulations are also necessary for the
beneficially tested using the standard IEEE 9-Bus test feeder by design of future power systems to evaluate the impact of
comparing its response to network disturbances to that of the proposed system changes.
previously existing OpenModelica generator model. The
simulation results in OpenModelica are furthermore successfully
A. State of the art: dynamic simulation tools
compared to respective results in DIgSILENT PowerFactory.
Hence, the new generator subsystem model is integrated into the A number of open source and commercial simulation tools
validated 20 kV power grid model of Campus North of Karlsruhe exist for studying the dynamic behavior of power systems.
Institute of Technology and respective results are presented for Commercial software packages such as SimPowerSystems [2],
abrupt load changes. PSS/E [3], DIgSILENT PowerFactory [4] and NEPLAN [5],
among others, are widely used to analyze system transients [6].
Keywords— generator subsystem model, modeling and Whereas these tools are computationally efficient and
simulation, OpenModelica, power system dynamics reasonably user-friendly, they have a closed architecture with
an application-programming interface (API). This implies that
I. INTRODUCTION it is not possible for users to access the source code in order to
modify and extend the functionality of the tools.
The power system is frequently considered to be operating
in steady state for the sake of convenience. However, due to On the other hand, open source tools provide users with full
random changes in loads, which take place at all time with access to source code allowing modeling flexibility and the
subsequent adjustment in generation, such a state never exists ability to contribute to the further development of the tools [7].
in the practical sense. Other occurrences, which result into Milano and Vanfretti [8] show a comparison of some open
major changes, include network faults, equipment failure, source software packages with different complexity levels.
sudden load application, or line or generation unit failure. A Among the studied packages, Modelica is selected for closer
significant imbalance between generation and demand results studies due to the following properties: 1) Object-oriented
into a change in system operation from one equilibrium state to modeling; which makes it possible to create physically relevant
another [1]. and easy-to-use model components. 2) Non-causal modeling
based on differential and algebraic equations. This enables a
Using power flow analysis, the equilibrium operating states
faster modeling process and results into development of
of the system can be determined [1]. However, this kind of
reusable models. 3) Multi-domain modeling capability, which
analysis ignores the transition period from one state to another,
allows combining electrical, mechanical, thermodynamic and
during which the dynamics may result into loss of synchronism
hydraulic model components within the same application model
or growing oscillations in the entire system. Dynamic
for a future co-simulation [9].
simulations are applied in order to determine the time-series

978-1-5386-2820-1/17 $31.00 © 2017 IEEE 44

DOI 10.1109/MCSI.2017.17
 Several Modelica related tools have been implemented. The Table I. Modelica libraries for the respective components
commercial tools include Dymola, MathModelica, MOSILAB, Subsystem Model Library
and SimulationX, whereas OpenModelica [10] and Scicos [11] 4th order Generator iPSL
are the available open source tools. In Modelica based tools, the Generator Steam Turbine Thermopower
models are decoupled from the mathematical solver. This subsystem Governor Standard Modelica
enables modifications in model equations and parameters, and Excitation system iPSL
thus provides unambiguous simulation results among different Transformer iPSL
Network
Transmission Line iPSL
simulation tools. These modeling characteristics are attractive subsystem
Load iPSL
for power system simulations and allow model information
exchange between different simulation tools [12], [13]. II. MATHEMATICAL MODELS
OpenModelica applies the phasor time domain mode of
B. Reasons for revising mathematical models
transient simulation. In this mode, the following simplifying
Mathematical modeling of power system components assumptions are considered during component modeling:
should include all relevant elements influencing the electrical
and mechanical torques of the machines in the system. The x Changes in network voltages and currents are
complexity of the models depends upon the type of transient instantaneous and hence a lumped transmission line
and the system under investigation. However, the power system model can be used.
analysis problem is increasing in dimension and complexity due x Voltages and currents are represented by fundamental
to the growth in electricity demand, integration of renewable phasor quantities.
energy sources and network expansions in form of large x The frequency remains nearly constant, which allows
network interconnections. This necessitates the reconsideration representing transmission lines using equivalent line
of the models used in system analysis in order to account for the impedances and admittances.
changes in network complexity. x Balanced network conditions and operations are
considered. This reduces the system representation to a
From a time scale analysis point of view, a disturbance in
single phase positive sequence network equivalent.
interconnected systems can cause slow power oscillations,
which can result into loss of synchronism within a short period With the above assumptions, the system is generically
of time. This network behavior is usually considered during represented by a set of differential algebraic equations
transient stability studies in the order of a few seconds.
However, the slow power oscillations are caused by the slow ẋ = f(x,y,u)
action of power and frequency governing systems, which are 0 = g(x,y,u) (1)
considered in long term dynamic stability studies in the order
of a few minutes. This therefore necessitates the reconsideration
where x are dynamic state variables, y are algebraic variables,
of the means and methods of analysis through revising and
and u are system parameters.
expanding mathematical modeling in order to improve
simulation and analysis accuracy. The nonlinear dynamic behavior of the generator and its
controllers is described by differential equations, while the
C. Component definition in OpenModelica generator-stator and the network are represented by algebraic
Several libraries have been developed in Modelica for equations. The generator equations are developed in the rotor
modeling and simulation in different domains due to the multi- d-q reference frame whereas the network is modelled using the
domain modeling property. In this paper, a steam turbine- common system reference frame (D-Q). Fig. 1 shows the
governor model is coupled to an existing generator model in relationship between the two reference systems, where rotor
OpenModelica. This involves coupling the mechanical and angle δ defines the position of the rotor q-axis relative to the
electrical domains using the ThermoPower library [14] and stator reference axis.
iPSL (iTesla Power Systems Library) [15], [16] respectively. The transformation between generator quantities (vd and vq )
The governor model is developed using standard Modelica and the network quantities (UD and UQ ) is given by (2) [17].
library components.
An additional component to connect the turbine system to
the generator is implemented using the textual editor to directly
implement the model equations. The libraries used for the
respective power system components are shown in Table I. The
next section describes the mathematical models of the power
system components used for a phasor time domain simulation
in OpenModelica.
Fig. 1. Reference frame transformation [17]

45
 vd sinδ – cosδ UD 1
v  =    (2) ̇R = K V – VF – VC +VS  – VR 
q cosδ sinδ UQ TA A Ref (9)

The generator subsystem used in this paper is composed of 1 KA

̇F =  (V – VX – KE EFD ) – VF  (10)
a 4th order generator, an excitation system, a steam turbine, and TF TE R
a governor model. The mathematical models of the components
will be described in the following subsections. The saturation function  is represented in exponential form as
SE = Aex exp(Bex EFD ). This leads to
A. Generator model VX = EFD Aex exp(Bex EFD ) (11)
The generator model is adapted from the existing iPSL 1
fourth order model [15,16]. The simplifying assumption in this Ė FD = [V – KE EFD – VX ] (12)
model is that the effects of the rotor damper windings are TE R
neglected. This results into a generator model represented by
transient emfs (E'd and 
) behind transient reactances (X'd The parameters used in the model (9) – (12) are given in [19].
and X'q ). The resulting stator algebraic equations are C. Turbine-Governor
Vd E'd R X'q Id The governor is used to control the valve positioning (y) of
V  =  '  –  '  I  (3) the steam turbine in order to vary the amount of steam flowing
q Eq – Xd R q
into the turbine. Fig. 3 shows the block diagram of the speed
The corresponding changes in emfs, speed deviation and rotor governor sub-block using a proportional-integral (PI)
angle are described by controller. The turbine-governor model is developed using
standard Modelica library components. The dynamics of the
1 governor are described as
ω̇ = (P – D(ω – ω0 ) – Pe ) (4)
2H m
y1 = Pref – KR ω – ωref  – Pm (13)
δ ̇ = ωb (ω – ω0 ) (5)
1
T'do E'q̇ = Ef – E'q + (Xd – X'd )Id (6) ̇ 2 = y
Tp 1
(14)

T'qo E'ḋ = – E'd – (Xq – X'q )Iq (7) y = Kp y1 + y2 (15)

The air gap power of the generator is where y is the valve position, Pref is reference power, Pm is
Pe = (E'd Id + E'q Iq ) + (X'q – X'q )Id Iq (8) mechanical power, ωref is reference speed, ω rotor speed,
KR =1/R, R is the droop constant, Kp and Tp are the gain and
The changes in mechanical power Pm and emf Ef are time constant of the PI controller. The variables y1 and y2 are
computed as described from the turbine-governor and exciter the integrator input and output signals, respectively.
mathematical descriptions. All the parameters of the model (3)
– (8) are taken from [18]. The steam turbine is modelled using the ThermoPower
library turbine model [14], which is an aggregated cascade of
B. Exciter turbine sections and described by the algebraic equations (16) –
(19) [20]. The steam mass flow rate w is given by the Stodola
The generator excitation system consists of an exciter and law as
an automatic voltage regulator (AVR) and is necessary to
supply the generator with DC field current. The applied exciter
model is obtained from the iPSL library [16]. Fig. 2 shows the ws = KT ρi Pi 1-PR 2 (16)
representation of a field-controlled DC commutator exciter with
a continuously acting voltage regulator. The dynamics of the where Pi is the inlet pressure, ρi is the inlet steam density,
exciter are hence modelled by (9) – (12). PR =Po /Pi , with Po being the outlet steam pressure. The

Fig. 3. Turbine – governor block diagram

Fig. 2. Type DC1 - DC Commutator Exciter [19]

46
constant KT can be determined from design geometry on
experimental data. The energy equation is given by
hi – ho = (hi – hiso )ηiso (17)
where hi and ho are the inlet and outlet enthalpies, hiso is the
outlet enthalpy if the expansion were isentropic and ηiso is the
isentropic turbine section efficiency. Fig. 5. Transmission line pi-model
The mechanical power Pm,turb transferred from the steam to the series admittance, Ys =1/Zs = Gs +jωBs . The mathematical
the shaft and the mechanical torque τm at the turbine output shaft description of the network is given by
are given by
I Y +Y –Ys V
Pm,turb = ηm ws (hi – ho ) (18)  s  =  s sh   s (20)
Ir –Ys Ys +Ysh Vr
p
τm = m (19) where Is , Ir and Vs , Vr are the sending and receiving end
ωm currents and voltages.
where ηm is the mechanical efficiency of the turbine and ωm is
the mechanical angular velocity of the turbine shaft. F. Transformer
The transformer is represented by an ideal transformer with
D. The new generator subsystem model turns ratio t sr in series with an impedance Zsr which represents
The developed governor (13) – (15) is combined with the the resistive (load dependent) losses and leakage reactance as
steam turbine (16) – (19) from the ThermoPower library to form shown in Fig. 6. The no-load and magnetization losses of the
the turbine-governor system. The turbine-governor and the transformer are neglected.
excitation systems (9) – (12) are connected to the iPSL fourth-
order generator model (3) – (8) to form the complete generator
subsystem. Due to the signal mismatch between ThermoPower
and iPSL library components, it is not possible to directly
connect components from the two libraries. A new Fig. 6. In-phase transformer representation
OpenModelica connector is developed to interconnect the
turbine-governor output to the generator input. The component The transformer ratio is given by tsr = aejθsr where a= Vp /Vs
converts the turbine-mechanical domain output signal into an and θsr = θp – θs . Hereby (Vs , θs ) and (Vr , θr ) are the sending
electrical domain input signal for the generator. Fig. 4 illustrates and receiving end voltage magnitude and phase angle. Point p
the block diagram of the generator subsystem as implemented is an internal non-physical node. The ratio tsr is real for in-phase
in OpenModelica. transformers (θs =θp ) and non-real (complex) for phase shifting
transformers. The mathematical description of an in-phase
E. Transmission line transformer is given in (21), where ysr =1/zsr .
A transmission line is represented by an equivalent pi-
I a2 y –ay V
circuit with a series impedance Zs =Rs +jωLs and shunt  s  =  sr y sr   s  (21)
admittance Ysh =Gsh +jωBsh at the sending and receiving ends as Ir –aysr sr Vr
shown in Fig. 5. The total shunt admittance is equal to 2Ysh and
G. Load model
The existing load models in OpenModelica are static loads,
which express the characteristic of the load at a given instant as
a function of the bus voltage and frequency.
The general representation of the voltage dependency of the
load characteristic is given by the following exponential model:
P = P0 (V/V0 )a (22)
Q = Q0 (V/V0 )b (23)
where V0 is the initial voltage at the load bus obtained from
power flow solutions. The variables P and Q are the active and
reactive power components of the load at voltage V, and P0 , Q0
are the corresponding values at the initial operating conditions.
The active power exponent a, and the reactive power exponent
b are the parameters of the model. For a = b = 0, 1 or 2, the
model represents constant power, constant current or constant
Fig. 4. The new generator subsystem graphical model in OpenModelica
impedance characteristics, respectively.

47
 The polynomial load model is another model used in later time t = 35s at bus 4 and cleared after 50ms. Fig. 8 shows
OpenModelica, which consists of a combination of constant the response of the generator-rotational speed to the applied
impedance, constant current and constant power components. transients. The dotted lines represent the response of the
The load power is a quadratic expression of the bus voltage network with OldGen generators, while the solid lines represent
given by the response of the network with NewGen generators. It is
observed that both systems regain stability after the faults are
P = P0  pz (V/V0 )2 + pc (V/V0 ) + pp  (24) cleared. However, the old generator system without frequency
regulation experiences a change in operating speed and has a
Q = Q0 qz (V/V0 )2 + qc (V/V0 ) + qp  (25) longer settling time. A further deviation from the initial
operating speed is observed when an additional fault is applied
The active power coefficients (pz , pc , pp ) and reactive power at a later time.
coefficients (qz , qc , qp ) are the parameters of the model and A detailed view of the NewGen behavior is shown in Fig. 9.
define the composition proportion of each component [17], It is observed that the introduced NewGen subsystem returns to
[21]. the initial operating speed and has a fast settling time. This is
The representation of the frequency dependency is obtained due to the operation of the included new governor system that
by multiplying a factor to the voltage dependent models. adjusts the mechanical power of the synchronous machine in
Further details of the frequency dependent load model can be response to speed deviations.
found in [17] and [21].
B. Comparison with DIgSILENT PowerFactory
III. SIMULATION AND RESULTS The IEEE 9-bus network is also simulated in DIgSILENT
PowerFactory with the same generator and controller
A. Fault simulations parameters as used in NewGen in OpenModelica, in order to
verify the results of the new generator system. Fig. 10 shows
The developed generator subsystem (3) – (19), cf. Fig. 4 in the response of the generator system rotational speed to short
Sec. II D, is compared with the existing OpenModelica circuit faults in DIgSILENT PowerFactory.
generator subsystem to assess the differences in generator
response and accuracy of the model. The old generator Fig. 9 and Fig. 10 show that both system return to the initial
subsystem (OldGen) is modeled with constant mechanical operating speed after the faults are cleared. However, the new
power input while the new generator subsystem (NewGen) is generator system introduced in OpenModelica responds with
modeled with a governor-regulated input mechanical power. less overshoot and faster settling time than the one implemented
in DIgSILENT PowerFactory.
The IEEE 9-bus test feeder [1] (with 3 generators, 3 loads
and 9 buses) is selected for the tests. Different transient cases
are simulated to analyze the system response to disturbances.
Fig. 7 shows the 9-bus network model with a fault event at
bus nine. The parameters of the model are taken from an
existing IEEE 9-bus example network in OpenModelica.
Similar governor parameters are used for the three generators
in NewGen with the only difference being the reference power
(Pref ) and the speed droop (R) as shown in Fig. 7.
A three-phase short circuit fault is applied at bus nine at time
t = 10s and cleared after 50ms. A second fault is applied at a

Fig. 8. Generator responses to short circuit faults in OpenModelica with

OldGen-subsystem and NewGen-subsystem (3) – (19)

Fig. 7. IEEE 9-Bus test feeder with the new generator subsystem Fig. 9. NewGen-subsystem response to short circuit faults in OpenModelica

48
 x Abrupt increase in load ΔL1 = (20 + j0.5)MVA at
t = 20 for a duration of 5s.
x Abrupt decrease in load ΔL2 = – (20 + j0.5)MVA at
t = 30 for a duration of 5s.
The simulated responses of the generator rotational speed
and the turbine mechanical power output are shown in Fig. 12
and Fig. 13.
The observed changes in the rotational speed are Δ1 =
0.00154 pu and Δ2 = 0.00154 pu for an increase and
decrease in load, respectively. The two turbines vary their
power output as follows:
x Gen1: ΔP1 = 0.15462 and ΔP1 = – 0.154624
Fig. 10. Generator rotational speed response in DIgSILENT PowerFactory
x Gen2: ΔP2  = 0.03090 and ΔP2 = – 0.03090
C. Simulation of Load changes When the load is increased by ΔL1 , the units slow down and
In order to further test the operation of the frequency regulator, this cause the governors to increase the output power of the
the new generator model is tested in a real network. Fig. 11 corresponding units until they attain a new common operating
shows the validated model of KIT-Campus North (KIT-CN) 20 speed given by ω1 = ω0 – Δω1 . The units respond in a similar
kV network in OpenModelica. The system consists of 2 way to an increase in generator speed caused by a decrease in
generators, 176 buses, and 35 loads. The KIT-CN model is part load ΔL2 . In this case, the governors decrease the output power
of the grid simulation framework within the EnergyLab-2.0 until a new common operating speed  =  + Δ is
project [22]. reached. The change in output power of each unit is equivalent
to the expected amount of load picked up by each unit computed
The simulated transient in this case is system load variation using the speed droop characteristic
at a specific time. In order to test the operation of the generator
subsystem, the following load changes are applied: 1 (26)
ΔP = – Δω
R
It is observed that the governors respond to the speed
deviation caused by the load change and adjust the output
power of the corresponding units in order to maintain
synchronism and minimize speed deviation. However, there is
a significant deviation from the nominal operating speed which
is attributed to the large load change which cannot be matched
by the generator set power. In order to restore the operation
speed to the nominal value, it is the role of the automatic
generation control (AGC) to adjust the reference power of
selected units. Modeling of the AGC will be considered in the
next steps.

Fig. 12. Generator rotational speed response

Fig. 11. KIT-CN 20kV Network in OpenModelica

49
 [6] K. K. Kaberere, K. A. Folly and A. I. Petroianu, "Assessment of
commercially available software tools for transient stability: Experience
gained in an academic environment," IEEE AFRICON, pp. 711 - 716,
2004.
[7] F. Milano, Power System Modelling and Scripting, London: Springer,
2010.
[8] F. Milano and L. Vanfretti, "State of the Art and Future of OSS for
Power," in Power & Energy Society General Meeting, Calgary, AB,
Canada, Oct 2009.
[9] P. Fritzson, Principles of Object-Oriented Modeling and Simulation
with Modelica 3.3: A Cyber-Physical Approach, Wiley-IEEE Press,
2015.
[10] P. Fritzson, P. Aronsson, H. Lundvall, K. Nyström, A. Pop, L. Saldamli
and D. Broman, "The OpenModelica Modeling, Simulation, and
Software Development Environment," In Simulation News Europe,
December 2005.
[11] N. Masoud and N. Ramine, "Modeling and simulation of differential
equations in Scicos," in Modelica Conference, September, 2006.
Fig. 13. Turbine mechanical power response [12] L. Vanfretti, W. Li, T. Bogodorova and P.Panciatici, "Unambiguous
Power System Modeling and Simulation using Modelica Tools," in
IEEE PES General Meeting, 2013.
IV. CONCLUSION
[13] M. A. A. Murad, F. J. Gomez and L. Vanfretti, "Equation-Based
Simulation and analysis of a wide range of dynamics in Modeling of FACTS using Modelica," in IEEE PowerTech, 2015.
power systems can be undertaken by increasing the details of [14] "ThermoPower, Open library for thermal power plant simulation,"
the system elements considered during modeling. The present [Online]. Available: https://casella.github.io/ThermoPower/.
paper shows an extension of the generator subsystem, which is [15] L. Vanfretti, T. Bogodorova and M. Baudette, "A Modelica Power
the main source of system dynamics. A turbine-governor System Component Library for Model Validation and Parameter
system is added to an existing generator model in an open Identification," in 10th International Modelica Conference, Lund,
source simulation tool. This is executed using special libraries Sweden, March 10 -12, 2014.
and Modelica standard components. The new generator [16] L. Vanfretti, T. Rabuzin, M. Baudette and M. Murad, "iTesla Power
subsystem with primary frequency regulation responds to Systems Library (iPSL): A Modelica library for phasor time-domain
simulations," SoftwareX, vol. 5, pp. 84 - 88, 2016.
network frequency changes by adjusting the generated output
power in order to maintain system stability. The introduced [17] P. Kundur, Power System Stability and Control, Newyork: McGraw-
Hill, Inc, 1994.
generator subsystem shows similar behavior compared to
models in DIgSILENT PowerFactory. [18] J. Machowski, J. W. Bialek and J. R. Bumby, Power System Dynamics,
Stability and control, John Wiley & Sons, Ltd, 2008.
The next step in this work will take renewable energy [19] IEEE Std 421.5-2005, "IEEE recommended practice for excitation
sources into consideration in order to assess the impact of system models for power system stability studies," IEEE, Newyork,
integration of large amounts of renewable energy sources and 2006.
the replacement of conventional sources on power system [20] A. Leva, C.Maffezzoni, "Modelling of power plants," in Thermo power
dynamics. Models of renewable energy sources like wind plant simulation and control, London, United Kingdom, The Institute
energy and solar photo voltaic will be included in the studied of Engineering and Technology, 2003, pp. 17 - 57.
networks. [21] F. Milano, "Power System Analysis Toolbox Documentation for
PSAT," February 14, 2008.
[22] Hagenmeyer et al, "Information and Communication Technology in
ACKNOWLEDGMENT Energy Lab 2.0: Smart Energies System Simulation and Control Center
This work is part of the “Energy System 2050” initiative of with an Open-Street-Map-Based Power Flow Simulation Example,"
Energy Technology, vol. 4, no. 1, pp. 145 - 162, 2016.
the Helmholtz Association [23].
[23] "Energy System 2050 – A Contribution of the Research Field Energy,"
Helmholtz Association, [Online]. Available:
https://www.helmholtz.de/en/research/energy/energy_system_2050/.
REFERENCES
[1] P. M. Anderson and A.-A. A. Fouad, Power System Control and
Stability, Piscataway, NJ : IEEE Press ; Hoboken, NJ ; Wiley-
Interscience , 2003.
[2] Hydro-Québec and T. Technologies, "SimPowerSystems, for use with
Simulink," TransÉnergie Technologies Inc, 2003.
[3] "PSS/E User Manual," Siemens Power Transmission & Distribution,
Inc, Dec 2007.
[4] "DIgSILENT PowerFactory User manual Version 2016," DIgSILENT
GmbH, Gomaringen, Germany, March 2016.
[5] BCP, "Neplan-Power System Analysis Software,"
http://www.neplan.ch, 2007.

50