Solving large-scale Modelica models: new approaches and experimental results using OpenModelica

Abstract

Modelica-based modeling and simulation is becoming increasingly important for the development of high quality engineering products. Therefore, the system size of interest in a Modelica-based simulation is continuously increasing and the traditional way of generating simulation code, e.g. involving symbolic transformations like matching, sorting, and tearing, must be adapted to this situation. This paper describes recently implemented sparse solver techniques in OpenModelica in order to efficiently compile and simulate large-scale Modelica models. A proof of concept is given by evaluating the performance of selected benchmark problems.

Modeling and simulation of large-scale systems: A systematic comparison of modeling paradigms

“…We discuss the concepts and techniques the experts consider having high potential (IM ≥ 3.5 ) in more detail below. [29,30].…”

“…The inefficiency of using dense solvers for sparse systems has been discussed by Casella [31] and Braun et al [30] among others. Sparse ODE and DAE solvers are available.…”

Coalesced Gas Turbine and Power System Modeling and Simulation using Modelica

“…Some additional advantages of Modelica are the open distribution of several libraries meant to represent physical systems, and the fact that models are independent from IDEs and solvers (Gómez et al, 2015). In addition, Modelica tools are now supporting the required numerical techniques to simulate large power grids (Braun, Casella, & Bachmann, 2017;Casella, Leva, & Bartolini, 2017;Dassault Systemes, 2018).…”

DAE Solvers for Large-Scale Hybrid Models

“…To meet those demands, special model structures are typically analyzed and exploited (Casella, 2015). Examples of strategies that have been successfully realized in Modelica tools, such as Dymola and the 3DEXPE-RIENCE platform, involve: multirate simulation (Thiele et al, 2014), mixed-mode simulation (Schiela and Olsson, 2000;Thiele et al, 2014), model decoupling and parallel execution (Elmqvist et al, 2014), and sparse solvers (Braun et al, 2017).…”

“…For certain models, such DAE mode simulations can be orders of magnitude faster, among other things, due to more efficient treatment of algebraic loops. Significant speed-ups have, for example, been observed when simulating national-and continental-sized models of electrical power systems (Braun et al, 2017). Rosenbrock DAE integrators were used by (Olsson et al, 2017) to achieve fast and predictable run times for model-based embedded control.…”