The industrial usage of the open-source Modelica tool OpenModelica was limited so far for power plant applications, due to the performance of large fluid systems. This paper presents some efforts to improve the simulation time on benchmark fluid models proposed by Siemens Energy. The main aspects presented here to achieve a faster simulation are an efficient evaluation of the jacobian matrix by a coloring technique, that exploits the sparsity pattern of a modelica model. Therefore the techniques are scratched and applied to benchmark models provided by Siemens Energy.

In the electricity market of today, with increasing demand for electricity production on short notice, the combined cycle power plant stands high regarding fast start-ups and efficiency. In this paper, it has been shown how the dynamic start-up procedure of a combined cycle power plant can be optimized using direct collocation methods, proposing a way to minimize the start-up time while maximizing the power production during start-up. Physical models derived from first principles have been developed in Modelica specifically for optimization purposes, in that the models contain no discontinuities. Also, the models used for optimization are simpler than typical highfidelity simulation models. Two different models used for optimization in four different start-up scenarios are presented in the paper. A critically limiting factor during start-up is the stress of important components, e.g., the evaporator. In order to take this aspect into account, constraints on the stress levels of such components have been introduced in the optimization formulation. In particular, it is shown how a pressure dependent stress constraint, similar to what is used in actual operation, can be applied in optimization. Also, different assumptions about which control variables to optimize are explored. Results are encouraging and show that energy production during start-up can be significantly increased by increasing the number of control inputs available to the optimizer, while maintaining desirable lifetime of critical components by introducing constrains on acceptable stress levels.

Modeling of large fluid systems requires in-house (specialized) tools, since applicability of Modelica and existing environments is limited.Nevertheless Modelica is a very powerful and descriptive modeling language, which is best suited for physical modeling in a heterogeneous environment. Its object oriented approach, the built-in documentation and the availability of commercial and free libraries justifies the decision for Modelica as the preferred modeling language within Siemens Energy.For an appropriate analysis of transient power plant processes, there often are large fluid systems to be modeled, i.e. there can be several thousand states. For such plant models, we use our in-house tool Dynaplant (DP), which is specialized for large fluid systems. A comparison between DP and Dymola[1] reveals some deficiencies of the Modelica world concerning performance and plant model construction: Especially, successive initialization and sparse matrix solvers are important features in need.

This article describes dynamic models of the carbon dioxide (CO 2)-removal units which are coupled with conventional models to form a complete model of an IGCC power plant with CO 2 capture. Therefore some components of the Modelica_Fluid 1.0 library and packages of the Modelica.Media library from Modelica 3.0 were used. Not yet available components were developed. The results obtained with Dymola 7.1 were compared with steady state simulations calculated with other tools (ChemCAD and Aspen Plus) and a very good agreement was found.

This paper shows how different kinds of optimization related task such as offline optimization or optimal control are solved using a combination of Modelica, Optimica, JModelica.org and Python. The application examples presented in this paper are all real industrial applications in the field of Combined Cycle Power Plants. Therefore different workflows have to be combined to solve the underlying task. This paper shows that these workflows can be conveniently connected using Python.