A Real-time Optimization Strategy for Petroleum Processes with Successive Adaptive Model Refinement

To enhance models for refinery processes it seems inevitable to increase the level of compositional detail to a much higher level than today. In previous work, we have developed a simulation technique for steady-state simulation, which pays respect to the particularities of this type of detailed composition models. With this new technique, it seems to be possible to significantly lower the computation time for obtaining high-resolution results. This work deals with the extension of our previous work to optimization problems. Using an adaptive composition representation in combination with a multigrid optimization strategy, the time span for the optimization of petroleum processes can significantly be lowered. The large potential of this approach for the implementation in real-time optimization schemes is illustrated by investigating a simple flash problem. It can be shown that the reduced dead time in which the process runs off the optimum operating points, yield a large economical potential.