ObjectStab - A Modelica Library for

Power System Stability Studies

Mats Larsson, Student Member
Dept. of Industrial Automation
Lund University
Sweden

November 1, 2000

Abstract
Traditionally the simulation of transient and voltage stability in power systems has been
constrained to domain-specific tools such as Simpow, PSS/E, ETMSP and EuroStag. While
being efficient and thereby able to simulate large systems, their component models are of-
ten encapsulated and difficult or impossible to examine and modify. Also, these simulators
often require substantial training and are therefore unsuitable for normal classroom use. For
academic and educational use, it is more important that the component modelling is transpar-
ent and flexible, and that students can quickly get started with their simulations. This paper
describes a freely available power system library called ObjectStab intended for voltage
and transient stability analysis and simulation written in Modelica, a general-purpose object-
oriented modelling language. All component models are transparent and can easily be modi-
fied or extended. Power system topology and parameter data are entered in one-line diagram
form using a graphical editor. The component library has been validated using comparative
simulations with EuroStag.

1 Introduction
The simulation of power systems using special-purpose tools is now a well-established area
and several commercial tools are available, e.g. Simpow, PSS/E and EuroStag. While most of
these tools are computationally very efficient and reasonably user-friendly they have a closed
architecture where it is very difficult or impossible to view or change most of the component
models. Some tools, e.g. EuroStag, provide the possibility of modelling controllers such
as governors and exciters using block diagram representation but this is often cumbersome.
Moreover, these constructs do not enable the user to modify any of the generator or network
models. The Power System Toolbox [1] is a set of Matlab program files capable of dynamical
simulation of power systems. Here the component models are accessible, but the modification
of these requires a proper understanding of the interaction of the model files in this specific
environment. Graphical editing of the models is not possible. Some of the above mentioned
tools have the capability of exporting a linearised representation of the system for further
analysis but the full nonlinear representation remains hidden to the user. Also, most of the
industrial-grade tools require substantial training before they can be used productively.
Modelica [2, 3] is an object-oriented general-purpose modelling language that is under
development in an international effort to define a unified language for modelling of physical
systems. Modelica supports object-oriented modelling using inheritance concepts taken from
computer languages such as Simula and C++. It also supports non-causal modelling, meaning
that a model’s terminals do not necessarily have to be assigned an input or output role. While
object-oriented concepts enable proper structuring of models, the capability of non-causal
modelling makes it easy to model for example power lines that are very cumbersome to model
using block-oriented languages such as Simulink.
 Causality is generally not defined in electrical systems [4]. For example, when modelling
a resistor, it is not evident ahead of time, whether an equation of the type

u=Ri (1)

will be needed, or one of the form:

u
i= (2)
R
It depends on the environment in which the resistor is embedded. Consequently the modelling
tool should relax the causality constraint that has been imposed on the modelling equations in
the past. Using causal modelling as in Simulink that does not have the capability of relaxing
such constraints is illustrated in [5] where a model for transient stability simulation is de-
scribed. With this approach, the network has to be modelled using the traditional admittance
matrix method and the power system topology can not be visualized in the graphical editor.
In Modelica, models can be entered as ordinary or differential-algebraic equations, state-
machines and block diagrams etc. Modelica also supports discrete-event constructs for hybrid
modelling. Object-oriented modelling of power systems using general purpose simulation
languages has previously been described in [6], where a block library was developed using
the modelling language Omola. However, the Omola project is now discontinued and the
experiences from this have been brought into the Modelica project.
At the time of publication, the simulation tool Dymola [7] was the only tool supporting
the Modelica language although some other producers of general-purpose simulators have
announced that future versions of their programs will support Modelica. There is already a
number of Modelica libraries for various application domains such as; 3-dimensional multi-
body systems, rotational and translational mechanics, hydraulics, magnetics, thermodynami-
cal systems and chemical processes.
This paper describes a component library written in Modelica for the simulation of tran-
sient and voltage stability in power systems. It is designed for use by undergraduate and
graduate students in the teaching of power system dynamic stability and for rapid testing of
research ideas. All component models are transparent and can easily be modified or extended.
Emphasis has been given to keeping the component models transparent and simple. Power
system topology and parameter data are entered in one-line diagram form using a graphical
editor. Using the Dymola simulator, the full hybrid nonlinear or linearities equation system
can be exported for use as a block in Simulink simulations and the full range of MATLAB
tools for visualization and control design can then be used. The component library has been
validated using comparative simulations with EuroStag [8].

2 The Component Library

The ObjectStab library presently contains the following component models :
• Generators with constant frequency and voltage as slack or PV nodes, or using 3rd or
6th order dq-models with detailed models of excitation and governor control systems,
or as quasi-steady state equivalents according to [9] .
• Transmission lines in pi-link or series impedance representation.
• Reactive power compensation devices; shunt reactors, shunt capacitances and series
capacitances according to [10].
• Fixed ratio transformers.
• On-load tap changing transformers (OLTC) modelled as detailed discrete models or
using their corresponding continuous approximations according to [11].
• Static and dynamic loads, including generic exponential recovery loads [12].
• Buses.
• Faulted lines and buses with fault impedance.
Standard assumptions for multi-machine power system stability simulations are made,
i.e., generator stator and network time constants are neglected and voltages and currents are
assumed to be sinusoidal and symmetrical. Except where other references are given, the com-
ponents are modelled according to the guidelines given in [10]. Furthermore each generator’s
set of equations are referred to an individual dq-frame and the stator equations are related to
the system (common) reference frame through the so-called Kron’s transformation [10].
All models can be modified and extended through inheritance. New models or extensions
of old ones can be entered as differential and algebraic equations, block diagrams or state-
graphs. Voltages and currents are described by their phasor representation:

I = ia + jib (3)
V = 1 + va + jvb (4)

Using this representation a connection point for power system components can be defined by
the following Modelica connector definition
connector Pin
Real va;
Real vb;
flow Real ia;
flow Real ib;
end Pin;
By definition all Pins use a p.u. system based on the system reference of 100 MVA,
although the generator models support parameter entry on their own individual base.

2.1 Modelling example: Impedance line model

The inheritance hierarchy for the impedance transmission line model is given below:
partial model TwoPin
Pin T1;
Pin T2;
end TwoPin;
The definition implies that a TwoPin is a model with two pins named T1 and T2 that
will be used as external connectors for the line. The Impedance model is defined using
the TwoPin as a base class and thereby inherit its attributes. Furthermore equations defining
the real and imaginary voltage drop on the line have to be given:
model Impedance
extends Base.TwoPin;
parameter Base.Resistance R=0.0 "Resistance";
parameter Base.Reactance X=0.1 "Reactance";
equation
[T1.va-T2.va; T1.vb-T2.vb] = [R, -X;
X, R]*[T1.ia; T1.ib];
[T1.ia; T1.ib] + [T2.ia; T2.ib] = [0; 0];
end Impedance;

2.2 Modelling example: 3rd order Generator Model

The dynamic generator models inherit from the OnePin class and its declarations are split
into three classes in order to simplify the implementation of new generator models. The class
DetGen defines the swing equation and the coordinate transformations as follows:
partial model Partials.DetGen
extends Base.OnePin;
parameter Base.Voltage V0=1;
parameter Base.ActivePower Pg0=1;
parameter Base.VoltageAngle theta0=0;
parameter Boolean isSlack=true;
parameter Base.ApparentPower Sbase=100;
parameter Base.InertiaConstant H=6;
parameter Base.DampingCoefficient D=0;
Base.AngularVelocity w(start=1);
 Base.VoltageAngle delta;
Base.MechanicalPower Pm;
Base.VoltageAmplitude Efd;
Base.Current id,iq;
Base.Voltage vd,vq;
Base.Current Iarm=sqrt(idˆ2 + iqˆ2);
Base.ActivePower Pe;
equation
// swing equations
der(w) = 1/(2*H)*(Pm-Pe-D/Base.ws*(w-Base.wref));
der(delta) = Base.ws*(w - Base.wref);
// Kron’s transformation
-[T.ia; T.ib] = [-sin(delta), cos(delta);
cos(delta), sin(delta)]*[id; iq];
// Load-flow initialization equation
if initial() then
if isSlack then
1 + T.va = V0*cos(theta0);
T.vb = V0*sin(theta0);
else
V = V0;
Pg = Pg0;
end if;
else
[1 + T.va; T.vb] = [-sin(delta), cos(delta);
cos(delta), sin(delta)]*[vd; vq];
end if;
end DetGen;
During the initialization which is equivalent to a load-flow calculation, generators are
represented as either PV-nodes or Slack nodes [10] depending on the attribute isSlack. It
is declared as a partial class, i.e. non-instantiable, since it lacks equations for the stator and
EMF equations that are dependent on the type of generator model used. For the 3rd order
generator model with a single transient EMF in the q-axis the definition is as follows:
model ObjectStab.Generators.Partials.DetGen3
extends DetGen;
parameter Base.Resistance ra=0;
parameter Base.Reactance xd=0.8948;
parameter Base.Reactance xq=0.84;
parameter Base.Reactance xdp=0.30;
parameter Base.Time Td0p=7;
Base.Voltage Eqp(start=1);
equation
// Transient EMF equation
Td0p*der(Eqp) = Efd - Eqp + id*(xd - xdp);
// stator equations
vd = -ra*id - xq*iq;
vq = Eqp + xdp*id - ra*iq;
// electrical power
Pe = Eqp*iq + (xdp - xq)*id*iq;
end DetGen3;
Also DetGen3 is declared as a partial class since it does not contain equations for the
governor and excitation system. These are provided in the instantiable class GovExc3rdGen
which contains models of the governor and excitation system.
model GovExc3rdGen
extends Partials.DetGen3;
replaceable Controllers.ConstPm Gov;
replaceable Controllers.ConstEfd Exc;
equation
// connection equations for governor and exciter
Gov.u = w;
Pm = Gov.y;
Exc.u1 = sqrt(vdˆ2 + vqˆ2);
Exc.u2 = w - Base.wref;
Efd = Exc.y;
// initialization
when initial() then
reinit(w, Base.wref);
reinit(delta, atan2(T.vb-ra*T.ib-xq*T.ia,
1.0+T.va-ra*T.ia+xq*T.ib));
reinit(id, -(-sin(delta)*T.ia+cos(delta)*T.ib));
reinit(iq, -(cos(delta)*T.ia+sin(delta)*T.ib));
reinit(Eqp,(Pg+ra*(idˆ2+iqˆ2)-id*iq*xdp+id*iq*xq)/iq);
reinit(Gov.Pm0,Pg + ra*(idˆ2 + iqˆ2));
reinit(Exc.Ef0,(Pg+ra*(idˆ2+iqˆ2)+id*iq*xq-id*xd*iq)/iq);
reinit(Exc.Vref, V + Exc.Ef0/Exc.K);
end when;
end GovExc3rdGen;
This model also contains equations for the initialization of the generator dynamic states
from the values given by the Pin variables. The values of the Pin variables are automat-
ically calculated in the load-flow that precedes every simulation. Since the Governor and
Exciter models have been declared using the modifier replaceable, they can be replaced
for models taken from a controller library or for user-defined models in instantiations of the
GovExc3rdGen class.

3 Extending the Library

The goal has been to provide a framework with the basic models necessary for power system
stability studies. However, such a library can never be complete and users can create their own
models inheriting from the (partial) base classes and providing the equations given below:
• Generic current injector models OnePin and TwoPin - equations for each of the real
and imaginary currents of each connector (Pin).
• Loads - equations for active and reactive load.
• Generators - equations for active and reactive production.
• Governor, Exciter - Generator control systems, equations for the mechanical shaft power
or the field voltage must be given.
The equations in user-defined model can be defined by block-diagrams, state-graphs or dif-
ferential or algebraic equations. Internal dynamic states in user-defined models should be
initialized from their terminal variables, similarly to the generator model in Sec. 2.2.

3.1 Extension Example: Exciter IEEE ST1A [13]

New exciter models can be constructed by creating a new class that inherits from the partial
class Exciter. This partial class contains the declarations necessary for communication
with the generator models and calculation of the initial values of the field voltage and voltage
reference. The block diagram can then be drawn using components from the block libraries
as shown in Figure 1, or alternatively, the corresponding differential or algebraic equations
can be entered in the equation window of the new model. The design of new governor system
is analogous. Note that also equations for dynamic blocks which at the start of the simulation
have non-zero output must be given as shown in Figure 1.
Figure 1: Diagram and equation windows for the IEEE ST1A Excitation system model [13].

Figure 2: The four-generator test system from [13].

 Figure 3: Textual representation of the four-generator test system.

4 Case Study: Four Generator Test System

The following steps must be carried out to simulate a power system using the ObjectStab
library and the Dymola simulator:
1. Create a new model window.
2. Draw the power system in the model window as shown in Figure 2 using the graphical
model editor. New components are created by dragging them from a library window
into the model window. By double-clicking on a component in the model window, a
dialog box appears and the parameters can be entered.
3. Compile the model by choosing ’Translate’ in the ’File’ menu of the model editor.
4. Simulate the system by choosing ’Simulate’ from the ’Simulation’ menu of the simu-
lator window. If desired, parameter values can be adjusted using a dialog box prior to
starting the simulation.
As an alternative to steps 1-2, the model can be entered in the textual format illustrated in
Figure 3. If the system is built using the graphical editor, parameter values are entered using
dialog boxes and the corresponding textual representation is automatically generated. Figure 2
shows the graphical display of the four generator test system [13] used for validation of the
library. Figure 4 shows simulation results obtained using the ObjectStab library and results
from simulation results obtained using EuroStag [8]. At simulation time 2 s, a ground fault
occurs close to bus 7. The faulted line is disconnected at both ends at 2.07 s, and later at 3.5 s
the fault has cleared and the line is reconnected. From the figure we can se that there is a
near-perfect agreement between the results of the two programs.

5 Further Work
The current version (4.1) of the Dymola simulator does not exploit the sparse structure of
the differential-algebraic equation system resulting from a large power system model. The
simulation of large systems is therefore very slow with Dymola compared to domain-specific
 ObjectStab EuroStag

Bus 8 Voltage (p.u.)

1 1

0.8 0.8

0.6 0.6

0.4 0.4
0 5 10 0 5 10
Time (s)
Bus 8 Angle (deg)

50 50

0 0

−50 −50
0 5 10 0 5 10

6 6
G1 Efd (p.u.)

4 4
2 2
0 0
−2 −2
0 5 10 0 5 10

Figure 4: Simulation results obtained using the ObjectStab library and EuroStag.

tools such as PSS/E or EuroStag. As an example, the a simulation of a ground fault with
the four node system takes about 10 s with EuroStag and 15 s with Dymola on a 150 MHz
SUN UltraSparc which is comparable. The corresponding times for the same simulation with
the Nordic 32 test system [14] which contains 41 buses and 23 generators, are 20 s with
EuroStag and 1330 s with Dymola - a factor 66 slower. Integration algorithms using sparse
matrix techniques may be included in a later version of Dymola and will hopefully make the
simulation times comparable with industrial-grade tools even for large systems. Currently, the
generator and transformer models do not contain saturation modelling. This may be included
in a later version of the library.

6 How To Get ObjectStab Running

The ObjectStab library was constructed for use with the simulation tool Dymola, which
runs on Microsoft Windows as well as SUN Solaris and Linux platforms. In addition to the
components models, the library includes models of the four generator test system used in this
paper and the Nordic 32 test system from [14]. A demo version of Dymola for Windows can
be downloaded from the Dynasim website1 . The library itself can be downloaded from the
ObjectStab website2 , and is free for educational use.

7 Conclusions
A component library called ObjectStab for power system transient and voltage stability sim-
ulations has been presented. Using the library, students can quickly enter their power system
in one-line diagram form using a graphical editor, without having to type cryptic data files
in text format or entering parameter data in countless dialog boxes. The library has an open
structure and all models can be modified or extended using various Modelica constructs, such
as state machines, block diagrams or plain algebraic or differential equations. The models
also have reasonable default parameter values defined such that students can play around with
the models even before they have full understanding of the meaning of all system parameters.
It is primarily intended as an educational tool, which can be used to experiment with and
observe the effect of the different levels of modelling or trying out new types of controllers.
1 http://www.dynasim.se/
2 http://www.iea.lth.se/˜ielmatsl/ObjectStab/
Because of some drawbacks described in Sec. 5 it can not yet compete with special-purpose
tools such as EuroStag in terms of computational efficiency, and is therefore most suitable for
analysis of small systems.

8 Acknowledgement
Parts of the library are based on models developed in a master’s thesis [15]. A special thank
goes to Inés Romero Navarro for her dedicated work in the masters project. Also the contri-
butions from Professor Gustaf Olsson and the financial support from the Sydkraft Research
Foundation is gratefully acknowledged. Thanks also to the development crew at Dynasim
AB, especially Hans Olsson and Sven-Erik Mattsson for modelling help and quick bug fixes.

References
[1] J H Chow and K W Cheung, “A toolbox for power system dynamics and control engi-
neering education and research”, IEEE Transactions on Power Systems, vol. 7, no. 4,
pp. 1559–64, November 1992. 1
[2] Modelica Design Group, Modelica - A Unified Object-Oriented Language for Physical
Systems Modeling, Language Specification, 1999, http://www.modelica.org/
documents.shtml. 1
[3] Modelica Design Group, Modelica - A Unified Object-Oriented Language for Physical
Systems Modeling, Tutorial and Rationale, 1999, http://www.modelica.org/
current/modelicarational12rev.pdf. 1
[4] K J Åström, H Elmqvist, and S E Mattson, “Evolution of continuous-time modeling and
simulation”, in The 12th European Simulation Multiconference, ESM’98, June 16-19,
1998, Manchester, UK, 1998. 2
[5] T Hiyama, Y Fujimoto, and J Hayashi, “Matlab/Simulink based transient stability sim-
ulation of electric power systems”, in Power Engineering Society Winter Meeting. 1999,
vol. 1, pp. 249–53, IEEE. 2
[6] S-E Mattsson, “Modelling of power systems in omola for transient stability studies”, in
IEEE Symposium on Computer-Aided Control System Design, 1992. 2
[7] H Elmqvist, D Brück, and M Otter, Dymola, Users Manual, Dynasim AB, Sweden,
2000, WWW Document: http://www.dynasim.se. 2
[8] J Deuse and M Stubbe, “Dynamic simulation of voltage collapses”, IEEE Transactions
on Power Systems, vol. 8, no. 3, pp. 894–904, August 1993. 2, 7
[9] T Van Cutsem and C Vournas, Voltage Stability of Electric Power Systems, Power
Electronics and Power Systems Series. Kluwer Academic Publishers, 1998. 2
[10] J Machowski, J W Bialek, and J R Bumby, Power System Dynamics and Stability, John
Wiley, New York, 1997. 2, 3, 4
[11] P W Sauer and M A Pai, “A comparison of discrete vs. continuous dynamic models of
tap-changing-under-load transformers”, in Proceedings of NSF/ECC Workshop on Bulk
power System Voltage Phenomena - III : Voltage Stability, Security and Control, 1994.
2
[12] D Karlsson and D J Hill, “Modelling and identification of nonlinear dynamic loads
in power systems”, IEEE Transactions on Power Systems, vol. 9, no. 1, pp. 157–163,
February 1994. 2
[13] P Kundur, Power System Stability and Control, Power System Engineering Series.
McGraw-Hill, New York, 1994. 5, 6, 7
[14] “Long term dynamics phase II”, Tech. Rep., CIGRE Task Force 38.02.08, 1995. 8
[15] I Romero Navarro, “Object-oriented modelling and simulating of power systems using
dymola”, Master’s thesis, Department of Industrial Electrical Engineering and Automa-
tion, Lund University, June 1999. 9