自适应变步长迭代动态规划方法及其在间歇过程优化中的应用

迭代动态规划(IDP) 作为一种求解非线性问题的离散算法, 其寻优精度和收敛速度受到时间段划分的影响. 通常, 时间段划分依赖主观经验, 缺乏科学有效的指导. 针对终端时刻固定的动态优化问题, 提出一种自适应变步长IDP 算法, 综合考虑控制变量与目标函数值的变化, 对时间段数量、长度和切换点进行优化. 将该方法应用于间歇过程优化, 结果表明其能够智能分配时间段数量与长度, 可有效提升寻优精度.

Self-adaptive variable-step approach for iterative dynamic programming with applications in batch process optimization

As a discrete algorithm to solve nonlinear optimization problems, iterative dynamic programming(IDP) algorithm is rather vulnerable to the stage of time in several aspects such as accuracy as well as the convergence rate. Traditionally, the time division associated with IDP algorithm relies on human’s subjective experiences, lacking effective guidance. Motivated by this observation and targeted at fixed terimal time optimizaton problem, a self-adaptive variable-step IDP algorithm is introduced in this paper, which can adjust the number, length and switching point of the time stages taking account of the performance and control variables, in order to improve the performance of IDP. The approach is applied to batch process optimization simulations. The results show that the time stages can be self-adjusted and the optimization performance can be improved.