动态博弈框架下的分布式动态优化

Distributed dynamic optimization in the framework of dynamic games

为了实时在线求解复杂的大规模动态优化问题,本文基于动态博弈理论提出了一种分布式动态优化方案,滚动合作博弈优化(RCGO).首先基于滚动时域优化框架,该方案将原本复杂的大规模动态优化问题分解为若干简单的小规模局部优化子问题,使得计算复杂度降低从而保证优化求解的实时性.之后本文基于动态博弈提出了分解迭代法求解各局部动态优化子问题,并对RCGO优化方案下系统稳定性进行分析.最后本文选择一个化工过程网络作为仿真案例,基于RCGO方案得到了极大化经济效益下该网络的最优操作.优化结果表明在求解复杂大规模动态优化问题时,RCGO方案较传统的集中式优化方案在由系统经济效益、闭环控制性能及优化求解实时性等组成的综合指标上有较大优势.

In order to solve complex large-scale dynamic optimization problems online and in real time,this paper proposes a distributed dynamic optimization scheme called receding cooperative game optimization(RCGO)based on dynamic games.Firstly,based on the receding horizon optimization framework,the scheme decomposes original complex large-scale dynamic optimization problems into several simple small-scale local dynamic optimization subproblems,which reduces the computational complexity and ensures the real-time solutions of the dynamic optimization problems.Then,based on dynamic games,the iterative decomposition method is proposed to solve local dynamic optimization subproblems,and the system stability under the RCGO scheme is analyzed.Finally,this paper uses a chemical process network as a simulation case.Based on the RCGO scheme,the optimal operations of the network with the maximum economic benefit are obtained.The optimization results show that when solving complex large-scale dynamic optimization problems online,the RCGO has more advantages than the traditional centralized dynamic optimization scheme in terms of the system economic benefit,closed-loop control performance and real-time optimization solution.