化工过程优化中的自动微分法  

AUTOMATIC DIFFERENTIATION IN CHEMICAL ENGINEERING PROCESS OPTIMIZATION

作　　者：李翔[1] 仲卫涛[1] 钱积新[1] 

机构地区：[1]浙江大学系统工程研究所,杭州310027

出　　处：《石油学报（石油加工）》2001年第6期79-83,共5页Acta Petrolei Sinica(Petroleum Processing Section)

摘　　要：针对化工过程优化中普遍存在的大规模求导问题 ,引入自动微分法生成子程序 ,形式函数的导数表达式 ,并用该表达式计算Jacobian矩阵 ,使命题固有的稀疏结构得以利用 ,并克服了传统差商法不准确的缺陷。实例计算表明 ,该方法较差商法更加快速、有效。Aiming at large scale differentiation in chemical engineering process optimization problems, automatic differentiation is employed to generate the derivative expressions of the original functions, which are usually coded in subroutines. The obtained expressions are used for the evaluation of the Jacobian matrix. Compared with traditional finite difference approximation, the derivatives are exact and the intrinsic sparse structure of the original functions could be exploited in this approach. Numerical experiments demonstrate that the new approach is faster and more efficient than finite difference approximation.

关 键 词：化工过程 优化 自动微分法 导数表达式 Jacobian矩阵 

分 类 号：TQ021.8[化学工程]