Influence of Feed Rate on Temperature Controllers in Multicomponent Distillation Columns (in cooperation with Statoil/Gassco at Kårstø)
MetadataShow full item record
In this thesis two simulation programs were compared, Unisim and D-spice. This was done by creating a model of three columns in series in Unisim and comparing the dynamic behavior of these models with the models in D-spice. The latter is a used by Statoil as a simulation tool. The biggest distinction between the programs is calculation procedure. To make the calculations faster and more robust numerically, D-spice takes advantages of thermodynamical tables, where thermodynamic properties are given for various temperature and pressure, but linearized with respect to composition. Originally, D-spice was doubtless better fitted to the real plant. However, some of the dynamics were found to be more appropriate in Unisim rather than in D-spice. Even though simplicity of column environment and complications in convergence of the solver with more complicated distillation structures was found in Unisim. Later in this study, control structure was elaborated at lower feed rate, with temperature controllers in priority. It was found that a right operation, that was also found to be the most profitable operation, will not result in critically unstable behavior. Even though the same can not be said about nonprofitable operation, gain scheduling was found to be unnecessary tool if process is operated at profitable operation with controllers tuned by SIMC tuning rules. Finally, optimization, with respect to the cost function of the most energy consumption column at full rate, was executed. The optimization was carried out with regard to impurity fraction of light component in the bottom product, whereas heavy component in the top product was held at two constant values, setpoint for supervisory control layer (Septic) and the specification of stream. The result was somewhat lower impurity fraction of light component bottom than the specification. In fact, J function reduction of 1.6% from specification to optimal operation was found. Consequently, a real time optimization, that would consequently supply control layer below with updated setpoints, was recommended.