Abstract: Distillation is a very important process in the large-scale production of ethylene and other refining and petrochemical processes.Based on the rigorous model of MESH and SRK equation of state, the optimization strategy for the operation of distillation column was studied.Since the open form equation-based model of distillation column has relatively small degrees of freedom, special structure and strong sparsity, a method based on the reduced sequential quadratic programming(SQP) algorithm and hybrid derivative method was presented to solve this kind of problems.The first order gradient information was obtained by the analytical method and preconditioned automatic differentiation.Because the objective was simple and relate to small number of variables, the analytical gradient function was obtained to get gradient information of objective.To utilize the sparsity and structure of the distillation column model, the sparse structure and compress pattern were obtained firstly by preconditioning procedures, and then gradient information of constraints was obtained through preconditioned automatic differentiation technology at each iteration.Computing results demonstrated that the proposed method was more efficient than the standard SQP algorithm, and the economic benefit could be improved significantly with optimal operation obtained under different conditions.

Key words: 序列二次规划算法, 简约空间, 自动微分, 精馏塔, 实时优化, 操作优化

摘要： 以开放式方程建立精馏塔严格机理模型，对精馏塔的操作优化进行了研究.根据精馏塔操作优化命题自由度低、模型结构稀疏且导数难以解析求解的特点，提出了基于简约空间序列二次规划算法（SQP）和混合求导方法相结合的精馏塔操作优化计算方法.在该方法中优化命题采用简约空间SQP算法求解，求解过程中需要的导数信息采用解析表达和预处理的自动微分技术求取.计算结果表明，本文方法在计算效率上大大高于基于差分求导的标准SQP算法，有利于在线实时优化的实施.另外本文优化结果也表明，精馏塔操作优化对提高产品产量、提高综合经济效益具有明显作用.

关键词: 序列二次规划算法, 简约空间, 自动微分, 精馏塔, 实时优化, 操作优化