Konferensartikel

Performance Benchmark of Modelica Time-Domain Power System Automated Simulations using Python

Sergio A. Dorado-Rojas
Department of Electrical, Systems and Computer Engineering, Rensselaer Polytechnic Institute, Troy, NY, USA

Manuel Navarro Catalán
Department of Electrical, Systems and Computer Engineering, Rensselaer Polytechnic Institute, Troy, NY, USA

Marcelo de Castro Fernandes
Department of Electrical, Systems and Computer Engineering, Rensselaer Polytechnic Institute, Troy, NY, USA

Luigi Vanfretti
Department of Electrical, Systems and Computer Engineering, Rensselaer Polytechnic Institute, Troy, NY, USA

Ladda ner artikelhttps://doi.org/10.3384/ecp2016928

Ingår i: Proceedings of the American Modelica Conference 2020, Boulder, Colorado, USA, March 23-25, 2020

Linköping Electronic Conference Proceedings 169:3, s. 28-34

Visa mer +

Publicerad: 2020-11-03

ISBN: 978-91-7929-900-2

ISSN: 1650-3686 (tryckt), 1650-3740 (online)

Abstract

In this paper, a benchmark between solvers and Modelica tools for time-domain simulations of a power system model is presented. A Python-based approach is employed to automate Modelica simulations and compute performance metrics. This routine is employed to compare the performance of a commercial (Dymola) against an open-source (OpenModelica) simulation tool with different solver settings. Python scripts are developed to execute a dynamic simulation of a common model for power system studies with 49 states and 420 variables in three different scenarios. This degree of automation makes it easier to change solver settings and tools during execution. The performance of each of the tools is assessed through metrics such as execution time and CPU utilization. The quantitative comparison results provide a clear reference to the performance of the tools and solvers for the execution of time-domain simulations with a significant degree of complexity. The commercial tool offers better performance for variable-step solver, but the performance of the open-source software shows significantly faster results for fixed-step solvers.

Nyckelord

Modelica, Python-Dymola Interface, Python-OpenModelica Interface, CPU performance.

Referenser

Inga referenser tillgängliga

Citeringar i Crossref