模块环境下的filter-SQP用于过程优化  

Filter-SQP in modular simulator environment for process optimization

作　　者：岳金彩[1] 杨霞[2] 郑世清[2] 韩方煜[2] 

机构地区：[1]华南理工大学化工学院,广东广州510640 [2]青岛科技大学计算机与化工研究所,山东青岛266042

出　　处：《化工学报》2006年第3期614-619,共6页CIESC Journal

基　　金：国家重点基础研究发展规划项目(2000026308)~~

摘　　要：Sequential Quadratic Programming(SQP) is the most efficient algorithm for nonlinear optimization.But a penalty function is usually used for linear search,causing some problems.Filter-SQP developed by Roger Fletcher and Sven Leyffer avoids using penalty function.In the view of filter-SQP,NLP problem has two objectives,one is minimizing objective function,the other is satisfying the constraints.The concept of filter is proposed on the basis of these two objectives.In this paper flowsheet optimization using filter-SQP in modular simulator environment was studied.Infeasible path strategy was used and the constraint function was composed of tear stream equation,specific design and unsatisfied inequality constraint.When filter could not find a step as the starting point of the next iteration,in order to avoid algorithm failure three strategies were used.They were restarting strategy,converging recycle strategy and feasible path strategy.A successive scaling strategy was proposed for filter-SQP to improve the efficiency of optimization.A case study of process optimization with filter-SQP was very encouraging.Sequential Quadratic Programming （SQP） is the most efficient algorithm for nonlinear optimization. But a penalty function is usually used for linear search, causing some problems. Filter-SQP developed by Roger Fletcher and Sven Leyffer avoids using penalty function. In the view of filter-SQP, NLP problem has two objectives, one is minimizing objective function, the other is satisfying the constraints. The concept of filter is proposed on the basis of these two objectives. In this paper flowsheet optimization using filter-SQP in modular simulator environment was studied. Infeasible path strategy was used and the constraint function was composed of tear stream equation, specific design and unsatisfied inequality constraint. When filter could not find a step as the starting point of the next iteration, in order to avoid algorithm failure three strategies were used. They were restarting strategy, converging recycle strategy and feasible path strategy. A successive scaling strategy was proposed for filter SQP to improve the efficiency of optimization. A case study of process optimization with filter-SQP was very encouraging.

关 键 词：模块环境 过程优化 filter-SQP 规格化 

分 类 号：TQ015.9[化学工程]