A Python-based Software Tool for Power System Analysis
Dr. Federico Milano
E-mail: Federico.Milano@ucd.ie
School of Electrical, Electronic and Communications Engineering
University College Dublin
Ireland
Dublin, October 11, 2013 Introduction - 1
Why Python?
Python is a modern scripting language which merges together the flexibility of
class-oriented programming (such as Java) and neat functional programming schemes
(such as Haskell).
It includes a huge number of third-party modules for mathematical applications, 2D
and 3D plotting, html and xml parsing, etc.
However, Python is an interpreted language, hence it is slow . . .
Dublin, October 11, 2013 Introduction - 2
Python is just a Glue!
Another property of Python is that it can be easily linked to C, C++ and Fortran libraries.
So, Python can be used as a glue to link together efficient (compiled) mathematical
libraries.
In conclusion Dome can be quite fast!
For example, to solve the power flow analysis of the UCTE test system (1254 buses
and 1944 lines) takes about 0.05 seconds (neat CPU time of the NR routine) on a
standard Dell workstation.
Dublin, October 11, 2013 Introduction - 3
Dome: Objectives
A Python-based Tool for Power System Analysis.
Dome has been designed with the following objectives in mind:
Modularity
Fast prototypying of models, algorithms and libraries
Parallel computing
Lazyness
Dublin, October 11, 2013 Introduction - 4
Dome: Structure
Data User
Format Filters
Filters
Output Formats
Device
Models
Dome Core
User
Custom Output Models
Routines User Routines
Dublin, October 11, 2013 Introduction - 5
Dome: Methodological Rules
Dome has been designed based on systematic methodological (even philosophical)
rules:
Bottom-up approach (or divide et impera)
No code duplicates (heavily based on Python classes)
Everything can be customized (Hegels approach!)
Separation between models and solvers
Minimizing the amount of essential code sections
Never complain (use defaults and warn the user!)
Code and documentation live (and update) together
Layered structure
The world is described by differential algebraic equations and/or if-then rules.
Dublin, October 11, 2013 Introduction - 6
Layered Structure
User Interface Surface usage
Toplevel device functions
Expert Lowlevel device functions
developers
Solver algorithms
Software core Deep usage
Dublin, October 11, 2013 Introduction - 7
Mathematical Model: DAE
Dome requires a set of nonlinear differential algebraic equations (DAE) with discrete
variables, as follows:
x = f (x, y, , u, t) (1)
0 = g(x, y, , u, t)
where x (x Rnx ) indicates the vector state variables, y (y Rny ) are the
algebraic variables, ( Rn ) are the controllable parameters, u (u Rnu ) are
discrete variables, f ( : Rnx Rny Rn Rnu R+ 7 Rnx ) are the
differential equations, and g ( : Rnx Rny Rn Rnu R+ 7 Rny ).
Discrete variables u can be often translated into if-then rules.
Dublin, October 11, 2013 Introduction - 8
Mathematical Model: SDAE
Dome can take into account stochastic processes leading to stochastic differential
algebraic equations (SDAE):
x = f (x, y, , u, t) + B(x, y, , u, t) (2)
0 = g(x, y, , u, t)
where are white noises, i.e., the time derivative of Wieners processes:
dW
=
dt
and B is the diffusion tensor.
Dublin, October 11, 2013 Introduction - 9
Stochastic Time Domain Analysis
Dome allows solving time domain simulations in parallel.
Solving 1000 simulations with a fixed time step of 0.05 s and a standard dishonest NR
solver for a 200-variable system takes about 8.5 seconds on a 12-CPU workstation.
Dublin, October 11, 2013 Introduction - 10
Mathematical Model: DDAE
Dome can take into account delays leading to delayed differential algebraic equations
(DDAE):
x = f (x, y, xd , y d , , u, t) (3)
0 = g(x, y, xd , y d , , u, t)
where xd and y d are delayed variables:
xd = x(t )
yd = y(t )
and are the delays (non-necessarily time-independent).
Dublin, October 11, 2013 Introduction - 11
Small-Signal Stability Analysis of Delayed Power Systems
Dome allows finding approximated solutions of the characteristic equation of a DDAE.
It turns to be the solution of a huge eigenvalue problem (thousands of eigenvalues!).
Dublin, October 11, 2013 Introduction - 12
Some Statistics
Dome currently includes:
45 data format (e.g., PSS/E, GE PSLF, DigSilent, etc.)
357 device models (e.g., synchronous machines and their primary controllers,
FACTS, wind turbines, DERs, energy storage systems, etc.)
10 power flow algorithms ranging from the standard NR to a GPU-based BFS.
13 mathematical libraries (e.g., KLU, UMFPACK, SLUDIST, PETSC, MAGMA, etc.)
Several static and dynamic analysis tools (eigenvalue analysis, CPF, OPF, time
domain analysis, short-circuit analysis, equivalencing techniques, polynomial
recasting, electric vehicles management, etc.).
Dublin, October 11, 2013 Introduction - 13
Supported Models
Dome currently supports:
Standard quasi-static phasors representing single-phase equivalents of
three-phase balanced and symmetrical devices
Park vectors
EMT models
DC, AC and power electronics devices
Three-phase unbalanced systems
Physical micro-controller devices (beta!)
A mix of all the above (with proper interfaces!)
Dublin, October 11, 2013 Introduction - 14
Research
Dome is an efficient research tool.
It is currently used by a small group of researchers under the supervision of Federico
Milano to study the following topics:
Modeling power systems through stochastic differential equations.
Stability of power system controllers including functional delays.
Parallellization of power flow analysis and time domain integration.
Detailed modeling of DERs and energy storage devices.
Dublin, October 11, 2013 Introduction - 15
Power System Education
Dome proved to be a challenging educational tool.
It has been used at the University of Castilla-La Mancha for the lab activities of the
course Power System Control.
It is currently used for teaching basic power system control and stability concepts at
Stage 4 of the Electrical Engineering program at UCD.
Dublin, October 11, 2013 Introduction - 16
Benchmarking
Dome can be used to test mathematical libraries.
For example, one can test the performance of libraries for sparse matrix factorization.
The table below refers to the power flow solution of the 1254-bus 1944-line network
that models that UCTE 2002 Winter Off-peak.
Library Total CPU 1st fact. Next fact.
time [s] time [s] time [s]
KLU 0.0933 0.0044 0.0026
CXSPARSE 0.0936 0.0043 0.0027
UMFPACK 0.1750 0.0126 0.0095
SUPERLU 0.1927 0.0247 0.0082
LUSOL 0.3112 0.0360 0.0195
Dublin, October 11, 2013 Introduction - 17
Consulting
Dome can be easily extended, can run on any platform supported by Python and the
resulting code is 100% open source.
These features makes Dome ideal for consulting activities.
It has been successfully used to develop an optimal load management algorithm, data
format filtering, stability analysis, etc., for a variety of European and American
companies.
A spin-out based on Dome is under evaluation at UCD.
Dublin, October 11, 2013 Introduction - 18
Current Challenges
Extensions that are currently under study and/or development are:
Real-time simulation capability
Hardware-in-the-loop
Capability of being a OPC (OLE for Process Control) server
Include diverse energy system models (e.g., gas and water)
Parallel computing based on heterogeneous architectures (CPUS and GPUs)
Dublin, October 11, 2013 Introduction - 19
Hardware-in-the-loop with Arduino
A very first attempt to include hardware-in-the-loop: Arduino micro-controllers.
Dublin, October 11, 2013 Introduction - 20
To-Do List
Graphical user interface (originally omitted by purpose!)
Full support on Windows operating systems
Improve documentation (e.g., automatic generation of device equations)
Support for discrete-event-simulation (DES) models (see for example the tools
developed at Oak Ridge National Laboratory, US)
...
Dublin, October 11, 2013 Introduction - 21
Try Dome!
Further details and a demo version of Dome are available at:
http://faraday1.ucd.ie/dome.html
If you are interested in giving Dome a try, just let me know and I will open an account
for you on my server.
Dublin, October 11, 2013 Introduction - 22
UCD-ERC Software Tools
A new page that collects information on software tools developed at ERC has been
included in the ERC web-page:
http://erc.ucd.ie/outputs/software
The page can be accessed from the ERC main page:
Menu Options Sotware Tools
Dublin, October 11, 2013 Introduction - 23
Thanks for your attention!
Dublin, October 11, 2013 Introduction - 24
