IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO.

3, AUGUST 2005 1199

An Open Source Power System Analysis Toolbox

Federico Milano, Member, IEEE

Abstract—This paper describes the Power System Analysis TABLE I

Toolbox (PSAT), an open source Matlab and GNU/Octave-based MATLAB-BASED PACKAGES FOR POWER SYSTEM ANALYSIS
software package for analysis and design of small to medium size
electric power systems. PSAT includes power flow, continuation
power flow, optimal power flow, small-signal stability analysis,
and time-domain simulation, as well as several static and dy-
namic models, including nonconventional loads, synchronous and
asynchronous machines, regulators, and FACTS. PSAT is also
provided with a complete set of user-friendly graphical interfaces
and a Simulink-based editor of one-line network diagrams. Basic
features, algorithms, and a variety of case studies are presented in
this paper to illustrate the capabilities of the presented tool and its
suitability for educational and research purposes. control scheme design. For these reasons, several Matlab-based
Index Terms—Continuation power flow, GNU/Octave, Matlab, commercial, research and educational power system tools have
optimal power flow, power flow, smaill-signal stability analysis, been proposed, such as Power System Toolbox (PST) [3],
time-domain simulation. MatPower [4], Toolbox (VST) [5], MatEMTP [6], SimPow-
erSystems (SPS) [7], Power Analysis Toolbox (PAT) [8], and
I. INTRODUCTION the Educational Simulation Tool (EST) [9]. Among these, only
MatPower and VST are open source and freely downloadable.

S OFTWARE packages for power system analysis can be

basically divided into two classes of tools: commercial
softwares and educational/research-aimed softwares. Commer-
This paper describes a new Matlab-based power system anal-
ysis tool (PSAT) which is freely distributed on line [10]. PSAT
includes power flow, continuation power flow, optimal power
cial software packages available on the market (e.g. PSS/E, flow, small-signal stability analysis, and time-domain simula-
EuroStag, Simpow, and CYME) follows an “all-in-one” phi- tion. The toolbox is also provided with a complete graphical
losophy and are typically well-tested and computationally interface and a Simulink-based one-line network editor. Table I
efficient. Despite their completeness, these softwares can result depicts a rough comparison of the currently available Matlab-
cumbersome for educational and research purposes. Even based tools for power system analysis and PSAT. The features
more important, commercial softwares are “closed”, i.e., do illustrated in the table are the power flow (PF), the continua-
not allow changing the source code or adding new algorithms. tion power flow and/or voltage stability analysis (CPF-VS), the
For research purposes, the flexibility and the ability of easy optimal power flow (OPF), the small-signal stability analysis
prototyping are often more crucial aspects than computational (SSA), and the time-domain simulation (TD), along with “aes-
efficiency. On the other hand, there is a variety of open source thetic” features such as the graphical user interface (GUI) and
research tools, which are typically aimed to a specific aspect of the graphical network editor (GNE).
power system analysis. An example is UWPFLOW [1], which An important but often missed issue is that the Matlab en-
provides an extremely robust algorithm for continuation power vironment is a commercial and “closed” product, thus Matlab
flow analysis. However, extending and/or modifying this kind kernel and libraries cannot be modified nor freely distributed.
of scientific tools also requires keen programming skills, in To allow exchanging ideas and effectively improving scientific
addition to a good knowledge of a low level language (C in the research, both the toolbox and the platform on which the toolbox
case of UWPFLOW) and of the structure of the program. runs should be free [11]. At this aim, PSAT can run on GNU/Oc-
In the last decade, several high-level scientific languages, tave [12], which is a free Matlab clone.
such as Matlab, Mathematica, and Modelica, have become more The paper is organized as follows. Section II illustrates the
and more popular for both research and educational purposes. main PSAT features while Section III describes the models and
Any of these languages can lead to good results in the field of the algorithms for power system analysis implemented in PSAT.
power system analysis (see, for example, [2]); however, Matlab Section IV presents a variety of case studies based on the IEEE
proved to be the best user choice. Key features of Matlab are the 14-bus test system. Finally, Section V presents conclusions and
matrix-oriented programming, excellent plotting capabilities future work directions.
and a graphical environment (Simulink) which highly simplifies
II. PSAT FEATURES
Manuscript received November 10, 2004; revised March 2, 2005. Paper no.
TPWRS-00596-2004.
A. Outlines
The author is with the Department of Electrical Engineering, Uni- PSAT has been thought to be portable and open source. At this
versity of Castilla-La Mancha, Ciudad Real 13071, Spain (e-mail: fmi-
[email protected]). aim, PSAT has been developed using Matlab, which runs on the
Digital Object Identifier 10.1109/TPWRS.2005.851911 commonest operating systems, such as Unix, Linux, Windows,
0885-8950/$20.00 © 2005 IEEE
1200 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 3, AUGUST 2005

TABLE II
FUNCTIONS AVAILABLE ON MATLAB AND GNU/OCTAVE PLATFORMS

— Market Data: Power supply bids and limits, generator

power reserves, and power demand bids and limits.
— Switches: Transmission line faults and breakers.
— Measurements: Bus frequency measurements.
— Loads: Voltage dependent loads, frequency dependent
loads, ZIP (polynomial) loads, thermostatically con-
trolled loads, and exponential recovery loads [14].
— Machines: Synchronous machines (dynamic order from 2
to 8) and induction motors (dynamic order from 1 to 5).
— Controls: Turbine Governors, AVRs, PSSs, Over-excita-
tion limiters, and secondary voltage regulation.
— Regulating Transformers: Under load tap changers and
phase shifting transformers.
Fig. 1. Synoptic scheme of PSAT.
— FACTS: SVCs, TCSCs, SSSCs, UPFCs.
— Wind Turbines: Wind models, constant speed wind turbine
and Mac OS X. Nevertheless, PSAT would not be completely with squirrel cage induction motor, variable speed wind
open source if it run only on Matlab, which is a proprietary turbine with doubly fed induction generator, and variable
software. At this aim PSAT can run also on the latest GNU/Oc- speed wind turbine with direct drive synchronous gener-
tave releases [12], which is basically a free Matlab clone. In the ator.
knowledge of the author, PSAT is actually the first free software — Other Models: Synchronous machine dynamic shaft, sub-
project in the field of power system analysis. PSAT is also the synchronous resonance model, solid oxide fuel cell, and
first power system software which runs on GNU/Octave plat- subtransmission area equivalents.
forms. Besides mathematical algorithms and models, PSAT includes
The synoptic scheme of PSAT is depicted in Fig. 1. Observe a variety of additional tools, as follows.
that PSAT kernel is the power flow algorithm, which also takes 1) User-friendly graphical user interfaces.
care of the state variable initialization. Once the power flow has 2) Simulink library for one-line network diagrams.
been solved, the user can perform further static and/or dynamic 3) Data file conversion to and from other formats.
analyses. These are as follows. 4) User defined model editor and installer.
1) Continuation Power Flow (CPF). 5) Command line usage.
2) Optimal Power Flow (OPF). The following subsections will briefly describe these tools.
3) Small-signal stability analysis. Observe that, due to GNU/Octave limitations, not all algo-
4) Time-domain simulations. rithms/tools are available on this platform (see Table II).
PSAT deeply exploits Matlab vectorized computations and
sparse matrix functions in order to optimize performances. Fur- B. Getting Started and Main Graphical User Interface
thermore, PSAT is provided with the most complete set of algo- PSAT is launched by typing at the Matlab prompt
rithms for static and dynamic analyses among currently avail-
able Matlab-based power system softwares (see Table I). PSAT
also contains interfaces to UWPFLOW [1] and GAMS [13],
which will create all structures required by the toolbox and open
which highly extend PSAT ability to solve CPF and OPF prob-
the main GUI (see Fig. 2). All procedures implemented in PSAT
lems, respectively. These interfaces are not discussed here, as
can be launched from this window by means of menus, buttons,
they are beyond the main purpose of this paper.
and/or short cuts.
In order to perform accurate and complete power system The main settings, such as the system base or the maximum
analyses, PSAT supports a variety of static and dynamic models, number of iteration of Newton-Raphson (NR) methods, are
as follows. shown in the main window. Other system parameters and
— Power Flow Data: Bus bars, transmission lines and specific algorithm settings have dedicated GUIs (see Figs. 8
transformers, slack buses, PV generators, constant power and 11). Observe that PSAT does not rely on GUIs and makes
loads, and shunt admittances. use of global variables to store both setting parameters and
MILANO: AN OPEN SOURCE PSAT 1201

Fig. 4. GUI for data format conversion.

Fig. 2. Main graphical user interface of PSAT.

Fig. 3. PSAT simulink library.

data. This approach allows using PSAT from the command line Fig. 5. GUI for user defined models.
as needed in many applications (see Section II-E).
include: IEEE, EPRI, PTI, PSAP, PSS/E, CYME, MatPower and
C. Simulink Library PST formats. On Matlab platforms, an easy-to-use GUI (see in
PSAT allows drawing electrical schemes by means of picto- Fig. 4) handles the DFC.
rial blocks. Fig. 3 depicts the complete PSAT-Simulink library The UDM tools allow extending the capabilities of PSAT and
(see also Fig. 7, which illustrates the IEEE 14-bus test system). help end-users to quickly set up their own models. UDMs can be
The PSAT computational engine is purely Matlab-based and created by means of the GUI depicted in Fig. 5. Once the user
the Simulink environment is used only as graphical tool. As a has introduced the variables and defined the DAE of the new
matter of fact, Simulink models are read by PSAT to exploit model in the UDM GUI, PSAT automatically compiles equa-
network topology and extract component data. A byproduct of tions, computes symbolic expression of Jacobians matrices (by
this approach is that PSAT can run on GNU/Octave, which is means of the Symbolic Toolbox) and writes a Matlab function
currently not providing a Simulink clone. of the new component. Then the user can save the model defi-
Observe that some Simulink-based tools, such as PAT [8] and nition and/or install the model in PSAT. If the component is not
EST [9], use Simulink to simplify the design of new control needed any longer it can be uninstalled using the UDM installer
schemes. This is not possible in PSAT. However, PAT and EST as well.
do not allow representing the network topology, thus resulting
in a lower readability of the whole system. E. Command Line Usage
GUIs are useful for education purposes but can in some cases
D. Data Conversion and User Defined Models limit the development or the usage of a software. For this reason
To ensure portability and promote contributions, PSAT is pro- PSAT is provided with a command line version. This feature
vided with a variety of tools, such as a set of Data Format Con- allows using PSAT in the following conditions.
version (DFC) functions and the capability of defining User De- 1) If it is not possible or very slow to visualize the graphical
fined Models (UDMs). environment (e.g., Matlab is running on a remote server).
The set of DFC functions allows converting data files to and 2) If one wants to write scripting of computations or include
from formats commonly in use in power system analysis. These calls to PSAT functions within user-defined programs.
1202 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 3, AUGUST 2005

3) If PSAT runs on the GNU/Octave platform, which cur- TABLE III

rently neither provides GUI tools nor a Simulink-like en- EXPONENTIAL RECOVERY LOAD DATA FORMAT (Erload.con)
vironment.

III. MODELS AND ALGORITHMS

A. Power System Model
The standard power system model is basically a set of non-
linear differential algebraic equations, as follows:

(1)
where most parameters are defined in Table III and , and
where are the state variables ; are the algebraic are initial powers and voltages, respectively, as given by the
variables ; are the independent variables ; power flow solution. Observe that a constant PQ load must be
are the differential equations ; and connected at the same bus as the ERL to determine the values
are the algebraic equations . of , , and .
PSAT uses (1) in all algorithms, namely power flow, CPF, Exponential recovery loads are defined in the structure Er-
OPF, smaill-signal stability analysis, and time-domain simula- load, whose fields are as follows:
tion, as discussed in the following Sections III-B to III-F. The 1) con: exponential recovery load data.
algebraic equations are obtained as the sum of all active and 2) bus: Indexes of buses to which the ERLs are connected.
reactive power injections at buses 3) dat: Initial powers and voltages ( , and ).
4) n: Total number of ERLs.
(2) 5) xp: Indexes of the state variable .
6) xq: Indexes of the state variable .
where and are the power flows in transmission lines as
B. Power Flow
commonly defined in the literature [15], is the set of network
buses, and are the set and the power injections of PSAT included the standard NR method [15], the fast decou-
components connected at bus , respectively. pled power flow (XB and BX variations [16]), and a power flow
PSAT is component-oriented, i.e., any component is defined with a distributed slack bus model [17]. The latter is a novelty
independently of the rest of the program as a set of nonlinear among Matlab-based power system softwares. The power flow
differential-algebraic equations, as follows: problem is formulated as (1) with zero first time derivatives

(5)
(3) Differential equations are included in (5) although some dy-
namic components are initialized after power flow analysis. This
where are the component state variables, the algebraic
is needed if the known input data of the component are not the
variables (i.e., and at the buses to which the component is
input parameters of its dynamic model. For example, the user
connected) and are independent variables. Then differential
does not generally know field voltages and mechanical torques
equations in (1) are built concatenating of all components.
of synchronous machines. However the user does know desired
Equations (3) along with Jacobians matrices are defined in a
voltages and active powers injected into the network by gener-
function which is used for both static and dynamic analyzes. In
ators. Thus, one can solve the power flow first, using PV buses
addition to this function, a component is defined by means of a
and then initialize synchronous machine state variables using
structure, which contains data, parameters and the interconnec-
the power flow solution. Nevertheless, other components can be
tion to the grid.
included in the power flow as one typically knows the input pa-
For the sake of clarity, let us consider the following example,
rameters of the dynamic model. For example, in the case of load
namely the exponential recovery load (ERL) [14]. The set of
tap changers, it is likely the user knows the regulator reference
differential-algebraic equations are as follows:
voltage rather than the transformer tap ratio.
The distributed slack bus model is based on a generalized
power center concept and consists in distributing losses among
all generators [17]. This is obtained by rewriting active powers
of slack and PV generators as
(6)
where are the desired generator active powers, is a
scalar variable which distributes power losses among all gen-
(4)
erators and are the participation factors of the generators to
MILANO: AN OPEN SOURCE PSAT 1203

the losses. Observe that is an unknown insofar as losses are clearing model with inclusion of voltage stability constraints,
unknown. Assuming that (6) has been written for all generators, based on what was proposed in [21] and [22], as follows:
is balanced by the phase reference equation.

C. Continuation Power Flow (CPF)

The CPF function included in PSAT is a novelty among avail-
able Matlab-based packages for power system analysis. The
CPF algorithm consists in a predictor step which computes a
normalized tangent vector and a corrector step that can be ob-
tained either by means of a local parametrization or a perpen-
dicular intersection [18]. The CPF problem is defined based on (11)
(1), as follows:
In (11), a second set of power flow variables and equa-
tions , together with the constraints
, are introduced to represent the solution as-
(7)
sociated with a loading parameter , where represents an in-
where is the loading parameter, which is used to vary crease in generator and load powers, as follows:
base case generator and load powers, , and re-
spectively, as follows:
(12)
where and are total generator and load powers for the
(8) current market condition.
Two objective functions are available: the maximization of
the distance to the maximum loading condition
D. Optimal Power Flow (OPF)
The OPF is defined as a nonlinear constrained optimization (13)
problem. The Interior Point Method (IPM) with a Mehrotra’s
and a multi-objective objective function
predictor-corrector method is used to solve the OPF problem
[19]. Notice that PSAT is the only Matlab-based software
which provides an IPM algorithm to solve the OPF-based
market clearing problem. A variety of objective functions are
(14)
included in PSAT, as follows.
where is a factor which allows weighting the influ-
1) Market Clearing Procedure: The “standard” OPF-based
ence of the system security on the market clearing procedure.
market model is represented in PSAT as follows:
E. Small-Signal Stability Analysis
PSAT allows computing and plotting the eigenvalues and the
participation factors of the system, once the power flow has been
solved. The eigenvalues can be computed for the state matrix of
(9) the dynamic system, and for the power flow Jacobian matrix
( sensitivity analysis) [23]. Unlike other softwares, such as
where and are defined as in (1), the control variables are the PST and Simulink-based tools, eigenvalues are computed using
power demand and supply bids and , while analytical Jacobian matrices, thus ensuring high-precision re-
and are the objective function and the inequality sults.
constraints, respectively. 1) Dynamic Analysis: The Jacobian matrix of a dy-
The goal is to maximize the social benefit; thus, the objective namic system is defined by linearizing (5), as follows:
function is defined as
(15)
(10)
where , , , and
where and are quadratic functions of supply and demand . Then the state matrix is obtained by eliminating ,
bids in $/MWh, respectively. and thus implicitly assuming that is nonsingular (i.e., no
The physical and security limits included in PSAT are sim- singularity-induced bifurcations)
ilar to what is used in [20], and take into account transmission (16)
line thermal limits, transmission line power flow limits, gener-
ator reactive power limits, and voltage “security” limits. The computation of all eigenvalues can be a lengthy process
2) VSC-OPF Market Clearing Model: The following op- if the dynamic order of the system is high. At this aim, PSAT
timization problem is used for representing an OPF market allows computing a reduced number of eigenvalues based on
1204 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 3, AUGUST 2005

sparse matrix properties and eigenvalue relative values (e.g. TABLE IV

largest or smallest magnitude, etc.). PSAT also computes par- PERFORMANCE OF PSAT SOLVERS FOR THE IEEE 14-BUS TEST SYSTEM
ticipation factors using right and left eigenvector matrices [15].
2) Sensitivity Analysis: The sensitivity analysis is
computed on a reduced matrix, as it was proposed in [23]. Let
us assume that the power flow Jacobian matrix is divided
into four submatrices

(17)

Then the reduced matrix used for QV sensitivity analysis is de-

fined as follows:

(18)

where it is assumed that is nonsingular [23]. Observe that

the power flow Jacobian matrix used in PSAT takes into account
all static and dynamic components, e.g. tap changers models,
etc.

F. Time-Domain Simulation
1) Integration Methods: Two integration methods are avail-
Fig. 6. GUI for power flow reports. The results refer to the IEEE 14-bus test
able, i.e., backward Euler and trapezoidal rule, which are im- system.
plicit -stable algorithms and solve (1) together [simultaneous-
implicit method (SI)]. This method is numerically more stable
than the partitioned-explicit method, which solves differential
and algebraic equations separately [15]. Observe that PSAT is
currently the only Matlab-based tool, which implements a SI
method for the numerical integration of (1).
2) Handling Disturbances: The commonest perturbations
for transient stability analysis, i.e., faults and breaker op-
erations, are handled by means of embedded functions. Step
perturbations can be obtained by changing parameter or variable
values after completing the power flow. All other disturbances
can be defined through custom “perturbation” functions, which
can include and modify any global structure of the system.

IV. CASE STUDIES

This section illustrates some PSAT features for static and
dynamic stability analysis by means of the IEEE 14-bus test
system (authors interested in reproducing the outputs could
retrieve the data from the PSAT web site [10]). All results
have been obtained on Matlab 7 running on a Intel Pentium
IV 2.66 GHz. Table IV depicts simulation times for the 14-bus
test system. Results were double-checked by means of other
software packages, namely PST [3], UWPFLOW [1], and
GAMS [13].
Fig. 7 depicts the model of the IEEE 14-bus network built
using the PSAT Simulink library. Once defined in the Simulink
Fig. 7. PSAT-Simulink model of the IEEE 14-bus test system.
model, one can load the network in PSAT and solve the power
flow. Power flow results can be displayed in a GUI (see Fig. 6)
and exported to a file in several formats including Excel and by Matlab, so that computation times increase slowly as the net-
LaTeX. PSAT also allows displaying bus voltages and power work size increase. Table V illustrates net power flow computa-
flows within the Simulink model of the currently loaded system tion times for a variety of tests network, with different solvers,
(e.g. see the bus voltage report in Fig. 7). Notice that PSAT uses namely NR method and fast decoupled power flows (both XB
vectorized computations and sparse matrix functions provided and BX variations). Results were obtained using the command
MILANO: AN OPEN SOURCE PSAT 1205

TABLE V
PERFORMANCE OF PSAT POWER FLOW SOLVERS

Fig. 10. Nose curves at bus 14 for different contingencies for the IEEE 14-bus
test system.

Fig. 8. GUI for continuation power flow settings.

Fig. 11. GUI for OPF settings. Observe that the weighting factor is set to 1 in
order to obtain the objective function (13).

TABLE VI
MAXIMUM LOADING CONDITION OPF FOR THE IEEE 14-BUS NETWORK

Loading Condition , and the Avail-

Fig. 9. GUI for plotting CPF results. The plots illustrate voltages at buses 12, able Loading Capability for the base case
13, and 14 for the IEEE 14-bus test system with no contingency. and the lines 2–3 and 2–4 outages. The OPF problem used to
compute the MLC is (11) and (13). Notice that, because of the
definitions of generator and load powers and given in (8)
line version of PSAT (times are about 0.5 s slower if using
and (12), one has .
GUIs).
The test case presented in [24] is reproduced here to illus-
CPF analysis is handled by a dedicated GUI, as illustrated
trate smaill-signal stability analysis and time-domain simula-
in Fig. 8. Nose curves can be plotted using the GUI for plotting
tion available in PSAT. Firstly it has been used the IEEE 14-bus
simulation results, which is depicted in Fig. 9. Fig. 10 illustrates
system with a 40% load increase with respect to the base case
the nose curves obtained using the CPF algorithm imple-
loading, and no PSS at bus 1. As illustrated by the time-domain
mented in PSAT. The curves refers to mere static equations, i.e.,
simulation depicted in Fig. 12, a Hopf bifurcation occurs for the
the differential equations of synchronous machines and controls
line 2–4 outage resulting in undamped oscillations of generator
are ignored during the CPF analysis. Fig. 10 depicts three dif-
angles. A similar analysis can be carried on the same system
ferent nose curves considering the base case network and line
with a 40% load increase but considering the PSS of the gener-
2–4 and line 2–3 outages, respectively. Notice that contingen-
ator connected at bus 1. Fig. 13 depicts the GUI for eigenvalue
cies are simulated by setting the status of breakers as “open” in
analysis and shows that the system is stable.
the Simulink model.
The GUI depicted in Fig. 11 allows adjusting parameters and
preferences for OPF analysis. For the sake of comparison with V. CONCLUSIONS
the CPF analysis, Table VI depicts the maximum loading param- This paper has presented a new open-source PSAT which
eter , the base case power , the Maximum runs on Matlab and GNU/Octave. PSAT comes with a variety
1206 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 3, AUGUST 2005

[5] A. H. L. Chen, C. O. Nwankpa, H. G. Kwatny, and X. Yu, “Voltage

stability toolbox: an introduction and implementation,” in Proc. 28th
North American Power Symp., Nov. 1996.
[6] J. Mahseredjian and F. Alvarado, “Creating an electromagnetic transient
program in MATLAB: MatEMTP,” IEEE Trans. Power Delivery, vol.
12, no. 1, pp. 380–388, Jan. 1997.
[7] G. Sybille. (2004, Oct.) SimPowerSystems User’s Guide, Version 4.
published under sublicense from Hydro-Québec, and The MathWorks,
Inc.. [Online] Available at: http://www.mathworks.com.
[8] K. Schoder, A. Hasanović, A. Feliachi, and A. Hasanović, “PAT: a power
analysis toolbox for MATLAB/Simulink,” IEEE Trans. Power Syst., vol.
18, no. 1, pp. 42–47, Feb. 2003.
[9] C. D. Vournas, E. G. Potamianakis, C. Moors, and T. Van Cutsem, “An
educational simulation tool for power system control and stability,”
Fig. 12. Generator speed oscillations for the IEEE 14-bus test system due to IEEE Trans. Power Syst., vol. 19, no. 1, pp. 48–55, Feb. 2004.
Hopf bifurcation triggered by line outage at 40% overload. [10] F. Milano. (2002) PSAT, Matlab-Based Power System Analysis
Toolbox. [Online] Available at: http://thunderbox.uwaterloo.ca/~fmi-
lano
[11] R. M. Stallman, Free Software, Free Society: Selected Essays of Richard
M. Stallman. Boston, MA: Free Software Foundation, 2002.
[12] J. W. Eaton, GNU Octave Manual. Bristol, U.K.: Network Theory
Ltd., 1997.
[13] A. Brooke, D. Kendrick, A. Meeraus, R. Raman, and R. E. Rosenthal.
(1998, Dec.) GAMS, a User’s Guide. GAMS Development Corporation.
[Online] Available at: http://www.gams.com/
[14] D. Karlsson and D. J. Hill, “Modeling and identification of nonlinear
dynamic loads in power systems,” IEEE Trans. Power Syst., vol. 9, no.
1, pp. 157–166, Feb. 1994.
[15] P. W. Sauer and M. A. Pai, Power System Dynamics and Sta-
bility. Upper Saddle River, NJ: Prentice-Hall, 1998.
[16] R. A. M. van Amerongen, “A general-purpose version of the fast decou-
pled loadflow,” IEEE Trans. Power Syst., vol. 4, no. 2, pp. 760–770, May
1989.
[17] W. R. Barcelo and W. W. Lemmon, “Standardized sensitivity coefficients
for power system networks,” IEEE Trans. Power Syst., vol. 3, no. 4, pp.
1591–1599, Nov. 1988.
[18] C. A. Cañizares, Ed., “Voltage Stability Assessment: Concepts, Prac-
Fig. 13. GUI for eigenvalue analysis. The plot illustrates eigenvalues for the tices and Tools,” IEEE/FES Power System Stability Subcommittee,
IEEE 14-bus test system with PSS, for a line 2–4 outage at 40% overload. Final Document, Tech. Rep., Aug. 2002.
[19] G. L. Torres and V. H. Quintana, “Introduction to interior-point
of procedures for static and dynamic analysis, several models methods,” in Proc. IEEE PICA, Santa Clara, CA, May 1999.
[20] K. Xie, Y.-H. Song, J. Stonham, E. Yu, and G. Liu, “Decomposition
of standard and unconventional devices, a complete GUI, and model and interior point methods for optimal spot pricing of electricity
a Simulink-based network editor. These features make PSAT in deregulation environments,” IEEE Trans. Power Syst., vol. 15, no. 1,
suited for both educational and research purposes. As a matter pp. 39–50, Feb. 2000.
[21] W. D. Rosehart, C. A. Cañizares, and V. H. Quintana, “Multi-objective
of fact, PSAT is currently used by several undergraduates, optimal power flows to evaluate voltage security costs in power net-
Ph.D. students, and researchers, and has an active mailing works,” IEEE Trans. Power Syst., vol. 18, no. 2, pp. 578–587, May 2003.
list (http://groups.yahoo.com/groups/psatforum) currently [22] F. Milano, C. A. Cañizares, and M. Invernizzi, “Multi-objective opti-
mization for pricing system security in electricity markets,” IEEE Trans.
counting over 450 members. Among future projects, there are Power Syst., vol. 18, no. 2, May 2003.
extending the CPF algorithm to dynamic bifurcation analysis [23] G. K. Morison, B. Gao, and P. Kundur, “Voltage stability analysis using
and including new control schemes and renewable energy static and dynamic approaches,” IEEE Trans. Power Syst., vol. 8, no. 3,
pp. 1159–1171, Aug. 1993.
generator models. Any suggestion and/or bug report are very [24] S. K. Mena Kodsi and C. A. Cañizares. (2003) Modeling and Simula-
welcome. tion of IEEE 14 Bus System With FACTS Controllers. Elect. Comput.
Eng. Dept., Univ. Waterloo, Waterloo, ON, Canada. [Online] Available:
http://www.power.uwaterloo.ca
REFERENCES
[1] UWPFLOW, Continuation and Direct Methods to Locate Fold Bifurca-
tions in AC/DC/FACTS Power Systems, C. A. Cañizares and F. L. Al-
varado. (1999). http://www.power.uwaterloo.ca [Online]
[2] M. Larsson, “ObjectStab—an educational tool for power system sta-
bility studies,” IEEE Trans. Power Syst., vol. 19, no. 1, pp. 56–63, Feb. Federico Milano (M’03) received the Elect. Eng. degree and the Ph.D. degree in
2004. electrical engineering in 1999 and 2003, respectively, both from the University
[3] J. H. Chow and K. W. Cheung, “A toolbox for power system dynamics of Genoa, Genoa, Italy.
and control engineering education and research,” IEEE Trans. Power From September 2001 to December 2002, he was with the Electrical and
Syst., vol. 7, no. 4, pp. 1559–1564, Nov. 1992. Computer Engineering Department, University of Waterloo, Waterloo, ON,
[4] R. D. Zimmerman, C. E. Murrillo-Sánchez, and D. Gan. (2005) Canada, as a Visiting Scholar. He is currently an Assistant Professor of
Matpower, Version 3.0.0, User’s Manual. Power System Engineering Electrical Engineering at the University of Castilla-La Mancha, Ciudad Real,
Research Center, Cornell Univ., Ithaca, NY. [Online] Available: Spain. His research interests are voltage stability, electricity markets, and
http://www.pserc.cornell.edu/matpower/matpower.html computer-based power system analysis and control.