利用Aspen Plus ActiveX自動化伺服器進行塔序列、整合與整廠最適化

Abstract

本篇利用模擬最適化，藉由在客戶端MATLAB使用模擬退火演算法與在伺服端Aspen Plus建立單一蒸餾塔，來完成近似最適化的塔序列與熱整合程序設計。塔序列中的每座單塔都能獨立完成，且每個塔序列的總年度成本可以被加總計算。此法優點不僅在程序模擬器中建立嚴謹模型與序列程序靈活運用，還能在程式語言中建構非微分導向最適化方法。使用此法，單塔可以在六分鐘以內完成最適化。再者，在適當的堆疊整合流程與可行性溫度範圍的蒸氣營運條件下，近似最適化的熱整合序列塔可以自動化完成。本篇用兩個塔序列程序來演示結果。最後本篇提出兩個結合序列塔與熱整合的整廠程序。反應段的出料決定分離段的進料，規格定義用來控制此兩段的進料與出料。結果顯示在整廠程序眾多的設計變數下 (13與11個變數)，最適化是可行的。

Simulation-optimization is applied to determine nearly optimal process designs in column sequencing and stacking problems using a simulated annealing algorithm (SAA) in the client (MATLAB) and a single distillation column in the server (Aspen Plus). Each column in every sequence is optimized separately and the total cost of each sequence is calculated. This method benefits both from the rigor of column modeling in a process simulator and the flexibility of implementing sequential procedures and derivative-free optimization in a procedural programming language. Using this method, a single column can be optimized within six minutes of computer time. Furthermore, when stacking rules and available utilities are specified, a nearly optimal heat-integrated distillation column sequence can be designed automatically. Results are illustrated using two case studies of column sequencing problems. Finally, two plantwide processes combined with column sequencing and stacking were proposed. The outputs of reaction section will determine the inputs of separation section. Some specifications are defined to govern the inputs and outputs between two sections. The results demonstrate that optimization from numerous design variables (13 and 11 variables) in plantwide process is feasible.