Asian Modelica and FMI Conference 2026

Baodi Zhang


Session

09-21
12:05
25min
Research on Simulation Efficiency of Supercritical Carbon Dioxide Brayton Power Generation System Based on Modelica
Siqiang Yi, Baodi Zhang

Existing Modelica libraries have demonstrated effective modeling of sCO2 systems. However, the influence of momentum-equation formulation on solver behavior has not been systematically quantified under an identical system model and consistent numerical settings. This study develops a one-dimensional dynamic Modelica model of an sCO2 Brayton power system and systematically compares four momentum-equation formulations within the same system under identical numerical settings, quantitatively examining how state selection, flow inertia, and pressure feedback affect solver steps, Jacobian evaluations, and error-dominant variables. The conventional steady-state algebraic momentum formulation is compared with three differential formulations incorporating flow inertia, and their computational performance is assessed under both steadystate and transient conditions. For the investigated system model and operating cases, the differential formulations containing a mass-flow derivative term completed the transient simulations, whereas the conventional steadystate algebraic formulation did not converge under the same numerical settings. Furthermore, by introducing a first-order inertia term for the average pipe flow rate together with inlet/outlet pressure feedback, the steadystate computation time is reduced from 750.5 s to 73.2 s, the number of transient solver steps decreases from 7878 to 3967, the number of Jacobian evaluations is reduced by 52.8%, and the occurrence frequencies of the reported error-dominant variables decrease by approximately 62% to 77% relative to Fluid Equation 2. These results provide a quantitative, model-specific characterization of how state selection, flow inertia, and pressure feedback affect solver behavior under the tested conditions.

Fluid Simulation Performance
Modelica Technology & AI (R1001)