• 沒有找到結果。

類神經網路於非線性程序系統控制之應用 劉權輝、涂瑞澤

N/A
N/A
Protected

Academic year: 2022

Share "類神經網路於非線性程序系統控制之應用 劉權輝、涂瑞澤"

Copied!
2
0
0

加載中.... (立即查看全文)

全文

(1)

類神經網路於非線性程序系統控制之應用 劉權輝、涂瑞澤

E-mail: 8919315@mail.dyu.edu.tw

摘 要

模式預測控制(model predictive control, MPC)已成為最常用的程序控制架構之一,其原理係在控制器內建立一組程序模式,

由模式來預測程序的響應。而藉由此模式來決定未來長期的操作值,使得未來響應與設定值間的差達到最小,也因此發揮 控制效果。以往MPC中常使用的是一個線性化模式,對真實程序必然存在一定的誤差。然而,目前很多複雜的程序,多以 神經網路(artificial neural network, ANN)來建立MPC中的預測模式,除了因為神經網路能做為良好的非線性模式外,主要更 由於其擁有學習的能力,可由線上資料來決定網路中的權值,進而獲得正確的程序動態。以往,這類的神經預測控 制(neural predictive control, NPC),大都將ANN學習成一個ARX (AutoRegressive model with eXogenous)或ARMAX模式,因 此,當需要長時間預測時,預測的結果需要以ANN迭代產生出來,故而誤差較大,且需要較長的運算時間。為改善上述缺 點,本研究根據DMC(dynamic matrix control)的控制原理,以ANN學習摺合 (convolution) 模式。我們稱此種控制為神經網路 預測控制(neural network predictive control, NNPC)。而基於ANN的學習方式與用途,研究中設計了一種以PRBS(pseudo random binary sequence)信號識別程序動態的方法,由這種識別的結果產生ANN學習所需的樣本。而在決定未來的操作時

,則以Levenberg-Marquardt法依經過學習的ANN模式使得模式輸出與設定點的誤差最小化。 針對這新設計的控制器,我 們以一個CSTR反應器進行測試。對溫度及濃度二個環路分別控制,結果顯示,無論是設定點或程序的干擾,NNPC都有 很好的效果。當進一步,應用在一個動態緩慢的生化反應器程序時,由於ANN模式內節點的增加,造成網路學習時的困難

。這時,我們減少預測模式中,操作與響應的次數。在這種簡化的模式預測控制器(simplified neural network predictive control, SNNPC)下,使得ANN更容易學習程序的動態。在和傳統PI及DMC控制器比較後,顯示這種控制器仍然維持著相 當不錯的控制品質,但偏移(off-set)卻明顯地增加。

關鍵詞 : 神經網路 ; 模式預測控制 ; 簡化神經網路模式預測控制 目錄

封面內頁 頁次 簽名頁 授權書1 iii 授權書2 iv 中文摘要 vi ABSTRACT viii 誌謝 x 目錄 xi 圖目錄 xv 表目錄 xviii 第一章 緒論 1 第二章 研究背景 5 2.1 前言 5 2.2 神經網路 (ANN) 6 2.2.1 節點(神經元)的組成 7 2.2.2 神經網路的拓蹼 8 2.3 神經網路的學 習訓練 10 2.4 模式預測控制 (MPC) 12 2.4.1 動態矩陣控制 (DMC) 14 2.4.2 DMC之簡化 19 符號說明 22 第三章 神經網路模式 預測控制 24 3.1 前言 24 3.2 NNPC控制架構 25 3.2.1 程序動態 26 3.2.1.1預測模式(ANN)的建立 27 3.2.1.2 訓練樣本的準備 28 3.2.1.3 訓練網路 29 3.2.1.4 ANN模式 33 3.2.2 模式預測控制 34 3.2.3 最適化運算 37 3.3 CSTR反應槽 39 3.3.1 溫度環路 42 3.3.2 濃度環路 42 3.4 結果與討論 43 3.4.1 程序的開環響應 43 3.4.1.1 冷卻水流率對程序之影響 43 3.4.1.2 進料流率對程序之 影響 45 3.4.2 溫度環路之控制 48 3.4.2.1 不同λ值之控制效果 50 3.4.2.2 NNPC與DMC控制結果之比較 52 3.4.3 濃度環路之 控制 54 3.4.3.1 不同λ值之控制效果 56 3.4.3.2 NNPC與DMC控制結果之比較 58 3.5 結論 60 符號說明 62 第四章 簡化神經 網路模式預測控制 65 4.1 前言 65 4.2 SNNPC控制架構 67 4.2.1 模式預測控制 67 4.2.2 最適化運算 69 4.3 SNNPC與NNPC之 控制差異 70 4.3.1 簡化神經網路模式預測控制結果 70 4.3.2 SNNPC與NNPC之控制結果比較 74 4.3.2.1 溫度環路 74 4.3.2.2 濃度環路 76 4.4 CASE STUDY--CSTR發酵槽 78 4.4.1 程序之模式建立 79 4.4.2 產生學習樣本 80 4.4.3 訓練樣本取決方式 81 4.4.4 以神經網路為模式 82 4.5 結果與討論 83 4.5.1 連續式發酵程序開環響應 83 4.5.2 控制結果及最適控制條件 84 4.5.3 與 其他控制器的比較 86 4.6 結論 88 符號說明 90 第五章 結論與展望 92 5.1 結論 92 5.2 展望 92 參 考 文 獻 94 附錄 99 附錄A 99 附錄B 101

參考文獻

王進德、蕭大全,類神經網路與模糊控制理論入門,全華科技圖書,台北市,1994。 黃明潔,簡化的模式預測控制,國立台灣科技大 學,碩士論文,台北,1990。 靳蕃、範俊波、譚永東,神經網路與神經計算機原理, 儒林, 台北,1992。 葉怡成,類神經網路模式應用 與實作,儒林,台北,1993。 蔡宏明、涂瑞澤,線上學習神經控制器結合最佳動態學習速率-應用於非線性系統的程序控制,中華民國 八十六年電腦程序控制研討會論文集,國立台灣大學,台灣,196-205,1997。 Aoyama, A., Internal model control framework using neural networks for the modeling and control of a bioreactor, Engng. Applic. Artif. Intell., 8(6), 689-701, 1995. Baheri, H. R., G. A. Hill, and W. J.

Roesler, Optimum design for production of tryptophan synthetase subunit from recombinant Escherichia coli in a two-stage continuous flow stirred tank reactor, The Chemical Engineering Journal, 62, 143-148, 1996. Chang, W. F., and M. W. Mak, A conjugate gradient learning algorithm for

(2)

recurrent neural networks, Neurocompution, 24, 173-189, 1999. Chen, J., Systematic derivations of model predictive control based on artificial neural network, Chem. Eng. Comm., 164, 35-59, 1998. Chen, J., D. S. H. Wang, and S. L. Jang, Product and process development using artificial neural-network model and information analysis, AIChE Journal, 44(4), 876-887, 1998. Chien, I. L., and P. S. Fruehanf, Consider IMC tuning to improve controller performance, Chemical Engineering Progress, 86, 33-38, 1990. Cutler, C. R., and B. L. Ramaker, Dynamic matrix control: A computer control algorithm. Proc. American Control Conf., WP5-B, San Francisco, CA, 1980. Cutler, C. R., and R. B. Hawkins, Constrained multivariable control of a hydrocracker reactor, Proc. American Control Conf., Minneapolis, Minnesota, 1014-1020, 1987. Dutta, P., and R. R.

Rhinehart, Application of neural network control to distillation and an experimental comparison with other advanced controllers, ISA Transactions, 38, 251-278, 1999. Henson, M. A., and D. E. Sebrog, Nonlinear control strategies for continuous Fermentors, Proc. American Control Conf., San Diego, CA, 2723-2728, 1990. Hsu, Y. Y., and C. C. Yu, A self-learning fault diagnosis system based on reinforcement learning, Ind. Eng. Chem. Res., 31, 1937-1947, 1992. Hunt, K. J., and D. Sbarbaro, Neural networks for non-linear internal model control. Proc. IEE Pt. D, 138, 431-438, 1991. Hunt, K. J., D. Sbarbaro, R. Zbikowski, and P. J. Gawthrop, Neural networks for control systems-A survey, Automatica, 28, 1083-1112, 1992. Kambhampati, C., J. D. Mason, and K. Warwick. A stable one-step- ahead predictive control of non-linear system, Automatica.

36, 485-495, 2000. Kreyszig, E., Advanced Engineering Mathematics, 8th ed., Wiley, New York, 1999. Kurtanjek, Z., Principal component ANN for modeling and control of baker’s yeast production, Journal of Biotechnology, 65, 23-25, 1998. Lebourgois, F., IDCOM applications and experiences on a PVC production plant, JointAutomatic Control Conference, San Fransisco, FA9-C, 1980. Lippmann, R. P., An introduction to computing with neural nets, IEEE ASSP Mag, 68-76, 1987. Luyben, M. L., Essentials of Process Control, McGraw-hill, New York, 1999. Masters, T., Advanced Algorithms for Neural Networks: A C++ sourcebook, Wiley, New York, 1995. Masters, T., Practical neural network recipes in C++, Academic Press, New York, 1993. Maurath, P. R., D. A. Mellichamp, and D. E. Seborg, Predictive controller design for single-input/ single-output (SISO) system, Ind. Eng. Chem. Research, 27,956-966, 1988. Mayne, D. Q., and H. Michalska, Receding horizon control of nonlinear systems.

Trans. IEEE on Aut. Control, 35, 814-824, 1990. Mehra, R. K., and R. Rouhani, Theoretical considerations on model algorithmic control for mon-minimum phase systems. Proc. American Control Conf., San Francisco, CA, TA8-B, 1980. Miller, W. T., R. S. Sutton, and P. J. Werbos, Neural networks for control. MIT Press, Cambridge, MA, 1990. Narendra, K. S., and A. M. Parthasarathy, Identification and control for dynamic systems using neural networks, IEEE Trans. on Neural Networks, 1, 4-27, 1990. Press, W. H., S. A. Teukolsky, W.T. Vetterling, and B. P.

Flannery, Numerical recipes in C. Second edition, Cambridge university press, New York, 1992 Robitaille, B., B. Marcos, M. Veillette, and G.

Payre, Modified quasi-newton methods for training neural networks, Computers Chem. Engng, 20, 1133-1140, 1996. Rohani, S., M. Haeri, and H. C. Wood, Modeling and control of a continuous crystallization process Part 2. Model predictive control, Computer and Chemical Engineering, 23, 279-286, 1999. Russell, N. T., H. H. C. Bakker, and R. I. Chaplin, Modular neural network modeling for long-range prediction of an evaporator, Control Engineering Practice, 8, 49-59, 2000. Saha, P. K., M. Shoib, and J. Kamruzzaman, Development of a neural network based integrated control system of 120 ton/h capacity boiler, Computers & Electrical Engineering, 24, 423-440, 1998. Salomon, R., and J. L. Hemmen, Accelerating Backpropagation through dynamic self-adaptation, Neural Networks, 9, 589-601, 1996. Smith, C. A., and A. B. Corripio, Principles and Practice of Automatic Process Control, 2nd ed., 471-475, Wiley, New York, 1997. Sorensen, P. H., M. Norgaard, O. Ravn, and N. K. Poulsen, Implementation of neural network based non-linear predictive control, Neurocomputing, 28, 37-51, 1999. Su, H. T., N. Bhat, P. A. Minderman, and T. J. McAvoy, Integrating neural networks with first principles models for dynamic modeling, In 3rd IFAC Symposium on Dynamics and Control of Chemical Reactors, Distillation Columns and Batch Processes, USA, 77-82, 1992. Syu, M. J., and C. L. Hou, Neural network dynamics identification of a fermentation system. in the proceedings of the 7th Congress of the Asian Pacific Confederation of Chemical Engineerings, TICC, Taipei, 2:625-630, 1996. Taso, J. H., and W. T. Wu, Global control of a continuous stirred tank bioreactor, The chemical Engineering Journal, 56, B69-B74, 1994. Temeng, K. O., P. D. Schnelle, and T. J. McAvoy. Model predictive control of an industrial packed bed reactor using neural networks, J. Proc. Cont., 5, 19-17, 1995. Trajanoski, Z., W. Regittnig, and P. Wach, Neural predictive controller for closed-loop control of glucose using the subcutaneous route: A simulation study, Control Eng. Practice, 5, 1727-1730, 1997. Trajanoski, Z., W. Regittnig, and P. Wach, Simulation studies on neural predictive control of glucose using the subcutaneous route, Computer Methods and Programs in Biomedicine, 56, 133-139, 1998. Wang, H., Y. Oh, and E. S. Yoon, Strategies for modeling and control of nonlinear chemical processes using neural networks, Computers Chem. Engng, 22, s823-s826, 1998. Werbos, P. J., Backpropagation through time: what it dose and how to do it?, Proc. of IEEE, 78, 1550-1560, 1990. Yu, X. H., and G. A. Chen. Efficient backpropagation learning using optimal learning rate and momentum, Neural Networks, 10(3), 517-527, 1997. Yu, X. H., G. A. Chen, and S. X. Cheng. Dynamic learning rate optimization of the backpropagation algorithm, IEEE Tran.

on Neural Networks, 6(3), 669-677, 1995. Zamarreno, J. M., and P. Vega, Neural predictive control. Application to a highly non-linear system, Engng. Applic. Artif. Intell., 12, 149-158, 1999。

參考文獻

相關文件

Infusing higher-order thinking and learning to learn into content instruction: A case study of secondary computing studies in Scotland. Critical thinking: What it is and why

Briefing Seminar for School Leaders of WiFi 900 Schools... Coherent Development of IT in Education Strategies and Curriculum Reform for Fostering Students’ Life-long Learning

Is end-to-end congestion control sufficient for fair and efficient network usage. If not, what should we do

{ Title: Using neural networks to forecast the systematic risk..

Since it is so, what do we cultivate for?People are looking for the ways to improve the mental state, and the courage or wisdom to face the hard moments.. But the ways of improving

In the work of Qian and Sejnowski a window of 13 secondary structure predictions is used as input to a fully connected structure-structure network with 40 hidden units.. Thus,

OpenGL 4.2 Reference card: Blue means!. deprecated

Text messaging (SMS) allows users to send and receive short text messages on a phone or other mobile device or computer Picture messaging allows users to send pictures and