An Evolutionary Approach to Derive Adaptive Optimal Control Policy for Chemical Processes

Abstract

To obtain a near optimal control policy for real world chemical processes, in this paper, we focused our attention on a new meta-heuristic method termed differential evolution (DE). Compared with conventional approaches characterized by variational logic as well as by the inconveniences of simultaneous optimization methods for differential-algebraic equation, we have shown DE’s ability to derive a near optimal solution adaptive to various requirements in practice. Since such a simulation-based search does not need any differential information and additional relations like adjoint equations and disregards complexities, the algorithm is straightforward and flexible to manage various conditions that other conventional approaches could not cope with effectively. These properties present great advantages when we need to cope with high dimensionality and outstanding non-linearity peculiar to chemical process in real world applications. In numerical experiments, we provided three popular reaction processes, applied the proposed method under various meaningful conditions, and validated its adaptability in comparison with other methods.