利用程序模擬自動化伺服器設計熱整合變壓蒸餾系統

Abstract

本篇利用MATLAB程式與常用化工程序模擬軟體Aspen Plus伺服器端進行自動化處理，並使用模擬退火法來進行變壓蒸餾系統的最適化設計，以尋找變壓蒸餾系統的最小年度總成本(TAC)，並與其他分離方法比較，此方法可以降低尋找最適化設計的時間與人力，也能確保找尋到全域最佳解。變壓蒸餾系統總有九個變數(design variable)，包含總板數(N1,N2)、進料板(N1FS,N2FS)、回流板(N1RS)、塔壓(P1,P2)以及與共沸物組成之距離(δ1,δ2)，自動化過程可以同時考慮各變數對年度總成本之影響，並找出系統全域最佳解。為了計算年度總成本，還需考慮塔序列(sequence)、真空系統(vacuum system)以及兩塔間的熱整合(heat integration)，針對此自動化處理程序，本篇會探討三種二成分共沸物系統－(1) 正己烷(N-hexane) /乙酸乙酯(ethyl acetate) (2) 丙酮(acetone)/氯仿(chloroform) (3) 苯(benzene)/ethanol(乙醇) ，利用模擬退火法找尋最小年度總成本，並與之常用的分離方法比較。三成分共沸物系統產生的共沸物可能不只一個，因此在最適化的過程會變得更加複雜，為了簡化問題，會針對一個共沸物的系統先進行探討。

Pressure-swing distillation (PSD) is an attractive alternative for separation of azeotropic mixtures when the composition of the azeotrope changes significantly with pressure. Two columns are operated at significantly different pressures and can produce nearly pure products. Since there are a large number of design variables, it would be quite tedious and time-consuming to determine the optimal value of each variable manually, the process is modeled in Aspen Plus and the simulated annealing algorithm is implemented in MATLAB to search optimum values. MATLAB acts as a client calling Aspen Plus to run simulations and simulation results would be transferred back to MATLAB to calculate total annual cost (TAC). Combined with heat integration and vacuum system, three pressure-swing distillation cases are studied. Different initial input variables could get almost the same output variables and the results are reproducible. The optimization process executes about two hours to get the minimum total annual cost.