A parallel algorithm for efficient large-scale dynamic optimization

Abstract:
In nowadays\' market conditions in the chemical industry, multi-product facilities, either multi-purpose batch processes or continuous processes operated in production campaigns, become more and more common. Hence, the operation of these processes does not only require optimal stationary but also optimal dynamic modes of operation. Dynamic optimization is considered a powerful tool to determine optimal control trajectories. However, the application of this method is bothered by the very high computational times in case of considering realistic models of compelte chemical plants. Therefore, besides continuing algorithmic improvement, parallel computing strategies have to be exploited. Such parallelisation strategies are presented in this contribution. Three examples of very different problem sizes are discussed. The speedup for the largest example even obtains a value close to the theoretical limit of the parallelization approach.

Please use the Online Form to request an electronic copy