|dc.description.abstract||It is desired to make the Economic Plantwide Control design procedure proposed by Skogestad (2000, 2004, 2012) available for engineers without deep knowledge of process control or optimization. The integration of the use of commercial process simulators to obtain the process model could be a useful tool for the automation of the economic plantwide design procedure. However, process simulators are set up in "design mode" and often work poorly in "operation mode". In this thesis, the use of commercial process simulators to generate process models suitable for an automated economic plantwide control procedure is explored. The analyzed process is methanol production, as it consists of: a reactor, a separator, and a recycle stream with purge. Simulations were made in UniSim R400 Design Suite.
The optimization for nominal conditions and disturbed process was done using a gradient-free algorithm, implemented in Python. As the active constraint area was scanned using different tolerances, more than 1000 optimization procedures and 60000 function evaluations (simulations) were performed. Four active constraint regions were found and a self-optimizing control structure was designed for one of them. The results of the resulting control structure were satisfying, with a consistently small loss.
It was shown that the tolerance for the optimizer is an important parameter in terms of finding consistent solutions. As a matter of comparison, the optimization at nominal conditions was also performed using Matlab gradient-based NLP fmincon algorithm. It was demonstrated that the gradient-free solver required less function evaluations than the gradient-based algorithm. This has a positive effect on reducing the time for evaluation and for performing step 2 in the plantwide design procedure.||