Nonlinear Model Predictive Control (NMPC) of Separation Trains

Abstract

An evaluation of Modelica Fluid Library 1.0 has been performed and the new library has been found to be more efficient both during modelling and during computations than the beta version of the library. The separator model in CyberneticaLib has been upgraded to the new Modelica 3.1 standards, and has been equipped with some new features for more robust simulations. Changes has also been made to some other components in CyberneticaLib. A work flow for building an NMPC application with Dymola, ModleFit and CENIT has been described. The work flow is mostly based on personal experience and input from people with expertice at Cybernetica. The total work flow consist of many steps, and several of these should be possible to automate for a more efficient work process. A process model of the separator train at Troll C has been developed. The process model has been developed according to the work flow described earlier. Several assumptions and siplifications has been done during the development of the model, and they are described and argumented in this report. Four different cases of disturbances to the separator train was simulated. Performance of the control system with NMPC manipulationg the setpoints was compared to performance of the control system without NMPC. The simulations showed a significant improvment both to robustness and performance of the process plant. However, some weaknesses were pointed out and possible changes to the model were suggested. All new Modelica equipment models and the Troll C model can be found in appendix [AppendixCD]. Also C-files for the model component project, MATLAB-files and simulation results can be found there.