Dynamic Optimization using Evolutionary Algorithm with Progressive Step Reduction
Dynamic optimization is the process of determining the optimal control strategy for a dynamic system over a finite time period to maximize or minimize some performance index. The system is normally described by a set of differential equations that are solved numerically. Each of the control variables to be optimized is described by a vector, each element of which represents a value in time (for a batch process) or space (for a continuous process). Such problems can be solved by iterative dynamic programming or by some evolutionary algorithm. In this paper, an acceleration technique termed Progressive Step Reduction (PSR) is proposed. At the beginning of the optimization, the time (or space) history of the control variables between the start and end of the process is first divided into a very small number of steps to allow a quick search for the overall features. As optimization progress, the number of time/space steps is repeatedly increased to allow the control profile to be gradually refined. The PSR technique is used in an evolutionary algorithm to optimize test problems in chemical reaction engineering and a factorial analysis was conducted to determine the effects of various factors and to optimize the algorithm itself. PSR was found to be effective in increasing the speed and reliability of the optimization.