A Probability Distribution Estimation Based Method for Dynamic Optimization

摘要

Engineering optimization of a highly nonlinear complex system is always a challenge methodologically and computationally.This is especially true when multistage dynamic optimization is involved.While significant progress has been made in rigorous deterministic algorithms for dynamic optimization,meta-heuristic-based optimization may offer an attractive alternative.This paper introduces a general mathematical framework,called the Population-based Probability Distribution Estimation(PPDE)method,for tackling constrained multistage complex process dynamic optimization problems.Solution identification is accomplished through probability distribution estimation based search in a continuous space,where special solution migration and penalty assignment techniques are integrated.Besides an optimal parameter estimation problem for a reactor system,an automotive coating curing optimization problem is also investigated,where the PPDE successfully minimizes oven energy consumption under various process/product constraints.Optimization results demonstrate superiorities of the method over the Ant Colony System(ACS)based dynamic optimization method.