臺科大化工系建置之乙酸異丙酯反應蒸餾試驗場於產能擾動測試時，系統在PID控制下產品濃度呈下降趨勢，導致許多次級品產生，因此本研究欲建立一套模型預測控制技術提供試驗場溫度操作建議值，使系統於擾動測試時能維持產品濃度且快速回到穩定狀態，達到即時預測與製程導航。本研究根據試驗場收集之實驗數據，將數據進行彙整與分析後，使用Aspen Plus V11.0建立符合試驗場製程之第一原理模型，再用Aspen Plus Dynamics V11.0建置與試驗場規格相同和控制架構之動態模型。使用動態模型模擬試驗場產能擾動之數據，作為類神經網路模型訓練之數據。
本研究使用Python建立類神經網路模型，並選用GRU模型作為類神經網路模型架構，模型的輸入變數分別為:異丙醇進料流量、第2板與第12板溫度控制器之設定點、第2板與第12板溫度當前值、異丙醇與醋酸進料之莫爾分率，輸出變數為側流產品之乙酸異丙酯濃度，本研究亦探討在不同輸入變數組合下，模型訓練與預測結果。GRU模型架構中，模型隱藏層數、神經元數、Time step等超參數之調諧則透過Bayesian optimization 來進行。
本研究使用Python進行模型預測控制，將GRU模型作為模型預測控制之預測模型，與動態模型進行連接對產品進行導航與排除干擾，並與傳統PID控制進行比較。模型預測控制中控制變數為側流產品之乙酸異丙酯濃度，由GRU模型提供，操作變數為第2板與第12板溫度控制器之設定點，模型預測控制主要調諧參數有:預測水平、控制水平以及權重矩陣，本研究將根據控制響應之結果調諧參數，並進行三種策略產能擾動測試。
結果顯示，類神經網路之測試結果MPAE可達0.002 %以下。模型預測控制於±10 %產能擾動測試中，CASE B與CASE C能有效控制產品濃度，產品濃度與濃度設定值積分誤差IAE約為3×10-4，產品濃度變化幅度降低至PID控制一半，達到穩定時間相較於PID控制縮短4小時以上，因CASE C模型輸入變數並無使用兩進料組成份，可使未來模型上線時減少兩進料量測點，可節省測量進料組成時間與成本，因此對於產能擾動測試而言，CASE C為最佳策略。

In this study, the isopropyl acetate reactive distillation pilot-scale plant was run, and it is found that lots of sub-product would appear due to the offset from PID controllers. To less the sub-products and make the system recovery faster, this study would develop model predictive control that can provide predicted operating guidance for the pilot plant. Aspen Plus V11.0 was used to establish a first-principle model to simulate the pilot plant. Then, dynamic model was built with the same specifications and control constructure as the real plant by Aspen Plus Dynamics V11.0. The data calculated by the dynamic model after throughput disturbance would be applied as the training data for the artificial neural network model.
This study builds ANN model with Python, and GRU model is selected as the ANN model architecture. The input variables of the GRU model are：isopropanol feed flow rate, setpoints of the temperature controllers, the process value of temperature controllers, and feed composition of IPA and acetate acid.On the other side, the output variable is IPAc mole faction of the product. Moreover, different combinations of input variables were discussed in this study. In the GRU model architecture, the tuning of hyperparameters, such as the number of hidden layers, hidden size, and time step, is performed through Bayesian optimization.
Python is used for building MPC. GRU model as the input of MPC connected with the dynamic model, and the results would be compared with that of traditional PID control. The control variable in MPC is the concentration of IPAc product of the fourth plate, which is provided by the GRU model. The manipulated variables are the setpoints of the temperature controllers. The main parameters of MPC are prediction horizon, control horizon and weight matrix.
The results show that the mean absolute percentage error of the testing set of GRU model can be less than 0.002 %. After ±10 % throughput disturbances, CASE B and CASE C can effectively control the product purity with MPC. The integral absolute error of product concentration and its setpoint is 3×10-4. The differences of product purity after disturbance with MPC are half of those with PID controllers, and the time span to make system steady with MPC decreases at least 4 hours. In terms of throughput disturbance, CASE C is the best strategy due to the lack of feed composition as input variables, which saves the cost and the time of measurement.

參考文獻
英文
1. Agreda, V. H. and Partin, L. R. (1984). Reactive Distillation Process for the Production of Methyl Acetate. U.S. Patent 4435595.
2. Beckhaus, A. A. (1921). Apparatus for the Manufacture of Esters. U.S. Patent 1400849A.
3. Bo, C., Li, J., Yang, L., Yi, H., Tang, J., & Qiao, X. (2017). MPC of distillation column with side reactors for methyl acetate. Chinese journal of chemical engineering, 25(12), 1798-1804.
4. Cho, K., Van Merriënboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv preprint arXiv:1406.1078.
5. Gadewar, S. B., Malone, M. F., Doherty, M. F. (2002). Feasible region for a countercurrent cascade of vapor‐liquid CSTRS. AIChE Journal, 48 (4), 800-814.
6. Gao, S., Huang, Y., Zhang, S., Han, J., Wang, G., Zhang, M., & Lin, Q. (2020). Short-term runoff prediction with GRU and LSTM networks without requiring time step optimization during sample generation. Journal of Hydrology, 589, 125188.
7. Garcia, C. E., Prett, D. M., & Morari, M. (1989). Model predictive control: Theory and practice—A survey. Automatica, 25(3), 335-348.
8. Harvianto, G. R., Ahmad, F., & Lee, M. (2017). A thermally coupled reactive distillation and pervaporation hybrid process for n-butyl acetate production with enhanced energy efficiency. Chemical Engineering Research and Design, 124, 98-113.
9. He, R., Dong, Y., Zou, Y., Zhao, J., Yaseen, M., Mu, C., & Tong, Z. (2020). Simulation and optimization of reactive distillation for the production of ethyl acetate using [BMIM] HSO4 as catalyst. Chemical Engineering Research and Design, 161, 218-231.
10. Henson, M. A. (1998). Nonlinear model predictive control: current status and future directions. Computers & Chemical Engineering, 23(2), 187-202.
11. Henson, M. A., & Seborg, D. E. (1997). Nonlinear process control (pp. 5-8). Upper Saddle River, New Jersey: Prentice Hall PTR.
12. Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735-1780.
13. Hopfield, J. J. (1982). Neural networks and physical systems with emergent collective computational abilities. Proceedings of the national academy of sciences, 79(8), 2554-2558.
14. Kalman, R. E. (1960a). Contributions to the theory of optimal control. Bulletin de la Societe Mathematique de Mexicana, 5, 102-119.
15. Kalman, R. E. (1960b). A new approach to linear filtering and prediction problems. Transactions of ASME, J. Basic Eng., 87, 35-45.
16. Kumar, A. S., & Ahmad, Z. (2012). Model predictive control (MPC) and its current issues in chemical engineering. Chemical Engineering Communications, 199(4), 472-511.
17. Lai, I. K., Hung, S. B., Hung, W. J., Yu, C. C., Lee, M. J., Huang, H. P. (2007). Design and control of reactive distillation for ethyl and isopropyl acetates production with azeotropic feeds. Chem. Eng. Sci, 62, 878-898.
18. Lai, I. K., Liu, Y. C., Yu, C. C., Lee, M. J., & Huang, H. P. (2008). Production of high-purity ethyl acetate using reactive distillation: experimental and start-up procedure. Chemical Engineering and Processing: Process Intensification, 47(9-10), 1831-1843.
19. Lee, H. Y., Li, S. Y., Chen, C. L. (2016). Evolutional Design and Control of the Equilibrium-Limited Ethyl Acetate Process via Reactive Distillation-Pervaporation Hybrid Configuration. Ind. Eng. Chem. Res., 55, 8802-8817.
20. Li, C., Duan, C., Fang, J., & Li, H. (2019). Process intensification and energy saving of reactive distillation for production of ester compounds. Chinese Journal of Chemical Engineering, 27(6), 1307-1323.
21. Liu, L., Jiang, H., He, P., Chen, W., Liu, X., Gao, J., & Han, J. (2019). On the variance of the adaptive learning rate and beyond. arXiv preprint arXiv:1908.03265.
22. Ma, J., & Yarats, D. (2018). Quasi-hyperbolic momentum and adam for deep learning. arXiv preprint arXiv:1810.06801.
23. Malone, M. F., & Doherty, M. F. (2000). Reactive distillation. Industrial & Engineering Chemistry Research, 39(11), 3953-3957.
24. Muske, K. R., & Rawlings, J. B. (1993). Model predictive control with linear models. AIChE Journal, 39(2), 262-287.
25. Osuolale, F. N., & Zhang, J. (2016). Energy efficiency optimisation for distillation column using artificial neural network models. Energy, 106, 562-578.
26. Oyelere, S. S. (2014). The application of model predictive control (MPC) to fast systems such as autonomous ground vehicles (AGV). IOSR J. Comput. Eng.(IOSR-JCE), 16(3), 27-37.
27. Pearson, R. K., & Ogunnaike, B. A. (1997). Nonlinear process identification. Nonlinear process control, 11-109.
28. Powell, M. J. (1994). A direct search optimization method that models the objective and constraint functions by linear interpolation. In Advances in optimization and numerical analysis (pp. 51-67). Springer, Dordrecht.
29. Prikryl, J., Novotný, J., & Šmídl, V. (2015, September). On distributed traffic signal control. In 2015 IEEE 18th International Conference on Intelligent Transportation Systems (pp. 894-899). IEEE.
30. Psichogios, D. C., Ungar, L. H. (1992). A Hybrid Neural Network-First Principles Approach to Process Modeling. AIChE J., 38, 1499-1511.
31. Raghavan, S. V., Radhakrishnan, T. K., & Srinivasan, K. (2011). Soft sensor based composition estimation and controller design for an ideal reactive distillation column. ISA transactions, 50(1), 61-70.
32. Rastegari, H., Ghaziaskar, H. S., Yalpani, M., & Shafiei, A. (2017). Development of a continuous system based on azeotropic reactive distillation to enhance triacetin selectivity in glycerol esterification with acetic acid. Energy & Fuels, 31(8), 8256-8262.
33. Ravi Chandra, P. V. S. and Venkateswarlu Ch. (2007). Multistep model predictive control of ethyl acetate reactive distillation column. Indian Journal of Chemical Technology, 14, 333-340.
34. Richalet, J., Rault, A., Testud, J. L., & Papon, J. (1976). Algorithmic control of industrial processes. 4th IFAC symposium on identification and system parameter estimation, 1119-1167.
35. Rumelhart, D. E., McClelland, J. L., & PDP Research Group. (1988). Parallel distributed processing (Vol. 1, pp. 354-362). Massachusetts: IEEE.
36. Sakhre, V. (2020). A Review on AI Control of Reactive Distillation for Various Applications. In Wastewater Treatment. IntechOpen.
37. Seborg, D. E., Edgar, T. F., Mellichamp, D. A., Doyle, F. J. (2017). Process Dynamics and Control. Wiley.
38. Sharma, N., & Singh, K. (2012). Model predictive control and neural network predictive control of TAME reactive distillation column. Chemical Engineering and Processing: Process Intensification, 59, 9-21.
39. Sundmacher, K., & Kienle, A. (Eds.). (2006). Reactive distillation: status and future directions. John Wiley & Sons.
40. Tang, Y. T., Chen, Y. W., Huang, H. P., Yu, C. C., Hung, S. B., & Lee, M. J. (2005). Design of reactive distillations for acetic acid esterification. AIChE Journal, 51(6), 1683-1699.
41. Victoria, A. H., & Maragatham, G. (2021). Automatic tuning of hyperparameters using Bayesian optimization. Evolving Systems, 12, 217-223.
42. Wang, H., & Mo, R. (2020). Neural network algorithm and its application in reactive distillation. arXiv preprint arXiv:2011.09969.
43. Yamanee-Nolin, M., Andersson, N., Nilsson, B., Max-Hansen, M., & Pajalic, O. (2020). Trajectory optimization of an oscillating industrial two-stage evaporator utilizing a Python-Aspen Plus Dynamics toolchain. Chemical Engineering Research and Design, 155, 12-17.
44. Yang, A., Lv, L., Shen, W., Dong, L., Li, J., & Xiao, X. (2017). Optimal design and effective control of the tert-amyl methyl ether production process using an integrated reactive dividing wall and pressure swing columns. Industrial & Engineering Chemistry Research, 56(49), 14565-14581.
45. Yu, T., & Zhu, H. (2020). Hyper-parameter optimization: A review of algorithms and applications. arXiv preprint arXiv:2003.05689.
中文
1. 林政綱，「試驗場級乙酸異丙酯反應蒸餾程序啟動策略與擾動排除指引之研究」，國立臺灣科技大學化學工程研究所，碩士論文，2020
2. 陳建穎，「使用類神經網路及模式預測控制於乙酸異丙酯反應蒸餾製程之控制研究」，國立臺灣科技大學化學工程研究所，碩士論文，2020
3. 薛宇玲，「使用非均相觸媒Amberlyst39之乙酸異丙酯合成反應動力行為研究」，國立臺灣科技大學化學工程研究所，碩士論文，2019