透過您的圖書館登入
IP:3.88.16.192
  • 學位論文

應用數學規劃法作批次排程/熱整合/水網路之設計

A Mathematical Programming Approach for Integrated Design of Scheduling/Heat-Recovery/Water-Reuse Network in Batch Plants

指導教授 : 陳誠亮

摘要


本論文試圖發展一套整合數學模型以處理化工批次排程、熱整合及水網路之問題,整體架構可分為兩大主軸,一為考量熱整合之批次排程模組,另一為批次水網路模組,兩套模組可獨立亦可合併應用。為了兼顧模組間單獨使用之效率及協同運作之相容性,在時間軸之表達方面,皆採用連續式時間點表示法為基礎,相較於傳統離散式時間點表示法具有計算效率之優勢。此外,各模組探討之範圍均完整地包含短期及週期性操作,可適用於大多數之生產狀況。 本研究係由Castro等人所提出之資源-任務網路數學模型(Resource-Task Network, RTN)平台架構為出發點加以擴充,為了使排程之數學模性具備熱整合之能力,本文使用新增新參數與一組新的0-1決策變數之方式,將直接熱回收之計算納入,這套新排程模型相當 一般化,在與文獻上之研究案例相較後,證實其具有較佳之彈性且不失效率。 本文接著提出一套全新水網路設計模型,可考量多污染物(成份)及多儲存槽策略下之短期與週期性操作之複雜設計,並與排程模組完全對應。由於其與排程模組唯一接合之處為固定之排程參數,因此單 獨考慮水網路設計並無排程計算之負擔,在最小化原水消耗量之基本情況中並不需要任何之0-1決策變數,因此僅為較不複雜的非線性規劃(Nonlinear Program, NLP)問題,由研究案例與文獻之比較中可證實此設計優異性。 在合併前述之數學模型後,可得到一套完整之數學模型,兼具排程、熱整合及水網路之設計能力,然而不可避免須面對混合整數非線性規劃(Mixed-Integer Nonlinear Program, MINLP)求解之困難度大幅提昇之問題。一般而言,在合理計算效率下尋求可接受之解亦為重要考量,因此本研究提出符合多數情況下,排程均較水網路設計為優先考量之「兩階段計算法」,其作法為首先將水網路模組轉換為一凸包(convex hull) 之近似模型,該模型再與排程模組相結合後,可得到一容易求解之 MILP 鬆弛模型,用來求解排程之0-1決策變數;第二階段於代入第一階段之排程0-1決策變數至原模型後,可成為一較容易求解之 NLP 問題,並解得排程之外之其他變數值。為了使第一階段之鬆弛模型更貼近原模型,水網路模組中亦進行部份之延伸,在論文最後使用兩個例子說明了此計算方式的確有不錯之效果。

並列摘要


This dissertation aims to develop an integrated mathematical formulation to solve the scheduling, heat-integration and water-reuse problems. The framework is divided into two modules: one is batch scheduling incorporated with heat recovery, whereas the other one is batch water-reuse network. The later is mainly used to work in coordination with the scheduling module. However, it can also be used individually. A continuous time formulation, which is more competitive in efficiency than traditional discrete one, is adopted as the time representation in designing each subsystem. Both short-term and periodic operation modes are considered to cover most of the production scenarios. Discussions are elaborated from a novel framework of Resource-Task Network (RTN) proposed by Castro et al. To extend the capability of scheduling toward simultaneous scheduling and direct heat-recovery, a generalized mathematical formulation is firstly proposed by introducing new parameters and a set of new binary variables in managing heat-integration. In comparison with literature examples, the new model is proven competitive with its flexibility and without the loss of compromising with efficiency. Then, a new water-reuse framework is proposed with the same continuous time concept in general scheduling model. This new formulation contains water-reuse and storage facilities for multi contaminant environment. Both single and periodic operation modes are also derived in correspondence with the scheduling model. The water-reuse module is connectable with scheduling module via the decision binary variables. However, they are regarded as predefined parameters if the water-reuse module works alone. Therefore, in the two basic scenarios of water minimization, no binary variable is required; therefore, the model is formulated in NLP, which is very competitive with the result reported in literatures. An overall model is carried out thereafter by simply combining the scheduling with water-reuse module proposed previously. However, the computational issue in handling large scaled MINLP is inevitable. Instead of global optimum, an acceptable solution might be more attractive in reasonable computation time. Since the priority in scheduling is usually higher than that in water-reuse network, a hierarchical two-step solving procedure is proposed. In the first stage, the water-reuse module is converted into a 'convex hull' MILP model and then combined with the scheduling module. The resulting relaxed model is MILP and could be easily solved to obtain the decision binary variable in scheduling. In the second step, the overall model is solved with the fixed production schedule. Therefore, only a NLP, which is easier to solve than the original MINLP, must be solved. Several enhancements are carried out in the linearization of water-reuse module to narrow the difference between the original and relaxed model. Two examples are demonstrated to verify the performance of the solving procedure.

參考文獻


[59] Wu, J. Y. ; He, X. R. A new model for scheduling of batch process with mixed intermediate
[11] Chen, C. L. ; Chang, C. Y. A resource-task network approach for optimal short-
[12] Chen, C. L. ; Chang, C. Y. ; Chang, J. C. Optimal short-term scheduling of multi-purpose
[6] Brooke, A. ; Kendrick, D. ; Meeraus, A. ; Raman, R. ; Rosenthal, R. E. GMAS: a user’s guide.
[61] Yoo, C. K. ; Lee, T. Y. ; Kim, J. ; Moon, I. ; Jung, J. H. ; Han, C. ; Oh, J. M. ; Lee, I.

延伸閱讀