模糊神经网络在大滞后非线性系统中的应用
|
摘要
随着科学技术的发展，现代工业生产过程的一个共同特征是控制系统的复杂性和不确定性日趋明显，即各个子系统之间或其内部会有较强的关联性，参数的高维性、时变性和随机性，且系统和环境具有许多未知的和不确定的因素，这些因素还会随环境、工况和时间等发生不可预料的变化。因此己不可能利用那些基于定量数学模型的传统控制方法对其实现有效的控制，必须寻求新的控制策略。
模糊控制是一种不依赖于被控过程数学模型的仿人思维的控制技术。它可以利用领域专家的操作经验或知识建立被控系统的模糊规则，有较好的知识表达能力。但在工程实际应用中却缺乏自学习或自调整的能力。尽管神经网络是一类黑箱式的非线性映射，但它具有良好的自学习能力。将二者有机结合起来，取长补短，用以提高整个系统的学习能力和表达能力。目前这个方向的研究正方兴未艾。
本文首先对模糊控制、神经网络及模糊神经网络的发展、背景和原理等进行了综述。对于大滞后非线性系统控制品质要就比较高的场合，模型辨识是至关重要的，针对大滞后非线性系统难以用经典方法辨识的原因，模拟人脑记忆事物的过程，本文提出了一种新型联想记忆神经网络结构，作为建模工具，以再现被控对象的特征，为控制做好铺垫。针对温度控制系统的非线性特点，本文利用模糊神经网络控制器，应用多层前馈网络构造模糊变量隶属函数和模糊椎理控制模型，使神经网络不再表现为黑箱式映射，其所有节点和参数都具有模糊系统等价意义。该算法可根据在线调整确定初始隶属函数和发现规则的存在，并可优化调整隶属函数，获得理想输出。本文利用该控制策略进行了仿真研究。结果表明，该控制策略可以使温度参数很好的达到要求。
温度控制系统是典型的非线性大滞后控制系统，根据本文提出的联想记忆神经网络辨识器，与模糊神经网络控制器相结合，建立模型参考自适应控制方案，成功的实现了对单腔电阻加热炉的控制，模型参考自适应控制方案可以向其它大滞后非线性特性的过程控制参量(如流量、压力、液位等)推广。
Along with the development of technology there are complexity and uncertainty of control systems in industry manufacture process. Many uncertain factors, such as correlation, randomicity, will change incidentally when environment and time change. So traditional control technique based on mathematical model is unuseful. New control strategy has being seeking.
Fuzzy control is a kind of human imitating technique which is independent on the controlled plant's mathematical model. It utilizes the knowledge and experience of experts to carry out rationalization. As a result, it has good robustness. But it is lack of the ability of self-learning or self-tuning after the fuzzy control rules have been set off-line. Neural network has the ability of self-learning in spite of its nonlinear mapping similar with Black-Box. The abilities of self-learning and expression of the whole system will be improved when they hand together. The research of this combination is in the ascendant.
First the backgrounds, improvements and principles of fuzzy control, neural networks and fuzzy neural networks are introduced. For the high controlled condition for larger nonlinear system, the identification which the classed method is difficult is important. The paper purposed a new associative memory neural network which can replaced system and ready for the request of control. This paper on the basis of fuzzy neural network technique and optimization it's parameter can get ideal output. The results of simulation show that this strategy is feasible in industrial, so it has good perspective of industrial application.
In industrial the temperature system is classed large nonlinear system. With associative memory neural network as identification and fuzzy neural network as control established model reference adaptive control system, succeeded realized single dealing control. Model reference adaptive control system supplied new method for other lager nonlinear system such as flux, stress and fluid system.
模糊控制是一种不依赖于被控过程数学模型的仿人思维的控制技术。它可以利用领域专家的操作经验或知识建立被控系统的模糊规则，有较好的知识表达能力。但在工程实际应用中却缺乏自学习或自调整的能力。尽管神经网络是一类黑箱式的非线性映射，但它具有良好的自学习能力。将二者有机结合起来，取长补短，用以提高整个系统的学习能力和表达能力。目前这个方向的研究正方兴未艾。
本文首先对模糊控制、神经网络及模糊神经网络的发展、背景和原理等进行了综述。对于大滞后非线性系统控制品质要就比较高的场合，模型辨识是至关重要的，针对大滞后非线性系统难以用经典方法辨识的原因，模拟人脑记忆事物的过程，本文提出了一种新型联想记忆神经网络结构，作为建模工具，以再现被控对象的特征，为控制做好铺垫。针对温度控制系统的非线性特点，本文利用模糊神经网络控制器，应用多层前馈网络构造模糊变量隶属函数和模糊椎理控制模型，使神经网络不再表现为黑箱式映射，其所有节点和参数都具有模糊系统等价意义。该算法可根据在线调整确定初始隶属函数和发现规则的存在，并可优化调整隶属函数，获得理想输出。本文利用该控制策略进行了仿真研究。结果表明，该控制策略可以使温度参数很好的达到要求。
温度控制系统是典型的非线性大滞后控制系统，根据本文提出的联想记忆神经网络辨识器，与模糊神经网络控制器相结合，建立模型参考自适应控制方案，成功的实现了对单腔电阻加热炉的控制，模型参考自适应控制方案可以向其它大滞后非线性特性的过程控制参量(如流量、压力、液位等)推广。
Along with the development of technology there are complexity and uncertainty of control systems in industry manufacture process. Many uncertain factors, such as correlation, randomicity, will change incidentally when environment and time change. So traditional control technique based on mathematical model is unuseful. New control strategy has being seeking.
Fuzzy control is a kind of human imitating technique which is independent on the controlled plant's mathematical model. It utilizes the knowledge and experience of experts to carry out rationalization. As a result, it has good robustness. But it is lack of the ability of self-learning or self-tuning after the fuzzy control rules have been set off-line. Neural network has the ability of self-learning in spite of its nonlinear mapping similar with Black-Box. The abilities of self-learning and expression of the whole system will be improved when they hand together. The research of this combination is in the ascendant.
First the backgrounds, improvements and principles of fuzzy control, neural networks and fuzzy neural networks are introduced. For the high controlled condition for larger nonlinear system, the identification which the classed method is difficult is important. The paper purposed a new associative memory neural network which can replaced system and ready for the request of control. This paper on the basis of fuzzy neural network technique and optimization it's parameter can get ideal output. The results of simulation show that this strategy is feasible in industrial, so it has good perspective of industrial application.
In industrial the temperature system is classed large nonlinear system. With associative memory neural network as identification and fuzzy neural network as control established model reference adaptive control system, succeeded realized single dealing control. Model reference adaptive control system supplied new method for other lager nonlinear system such as flux, stress and fluid system.
引文
[1]阎平凡，张长水．人工神经网络与模拟进化计算．清华大学出版社,2002．123～132
[2]戴汝为．人工智能．化学工业出版社,2002．63～72
[3]申东日，冯少辉．BP网络改进方法概述．化工自动化及仪表，2000．27(1)：30～32
[4]高雪鹏，丛爽．BP网络改进算法的性能对比研究．控制与决策，2001．16(2)：167～171
[5] Psaltis D, A.Sideris and A.Yamamura. A multilayered neural network controller. IEEE Control Systems Magazine, 1988.8(2):17～21
[6] 徐湘元，毛宗源．过程控制的发展方向智能控制．化工自动化及仪表，1998．25(2)：1～5
[7] Beartrice Lazzerini. Leonardo M.Marcello Chiaberge. A neural-fuzzy approach to hybrid intelligent control[J]. IEEE Trans on Industry Applications, 1999.35(2):413～425
[8]李金厚．神绎网络研穷进展与展望．华东冶金学院学报,2000．17(4)：275～281
[9] X.Dong, Q.xu.Identification and control for chain boiler combustion system based on neural networks.Instrumentation and Control Technology,2003.5253(6): 620～623
[10] Xu Lina.Neural networks control.Harbin Institute of Technology,1999
[11]陈在平，杜太行等．控制系统计算机仿真与CAD．天津大学出版社，2001
[12]孙增圻．智能控制理论与技术．清华大学出版社广西科学技术出版社,2002．126～135
[13] Yang H T, Chang W Y, Huang C L.Power system Distributed On-Line Fault Sectin Estimation Using Decision Tree based Neural Nets Approach. IEEE Transactions on Power Delivery, 1995,10(1):540～546
[14] Sun Y, Jiang H,Wang D.Fault Synthetic Recognition for an EHV Transmission Line Using a Group of Neural Networks with a Time-Space Property. IEE Proceedings—Generation, Transmission and Distribution, 1998,145(3):265-270
[15]肖田元，张燕云，陈加栋．系统仿真导论．清华大学出版社，2002
[16]杨宪勇．人工过程自动控制．清华大学出版社，2000．149～164
[17]方崇智，萧德云．过程辨识．清华大学出版社，2001．98～104
[18]高隽，胡勇，胡良梅．关于AM学习样本选择的实验研究．模式识别与人工智能，2002
[19]杨斌，田永青，朱仲英．智能建模方法中的数据预处理．信息与控制，2002
[20] Sittithumwat A, Tomsovic K, Soudi E Optimizing Maintenance Resources in Distribution System with limited Information. In:ISAP'99. Rio de Janeiro(Brazil):1999
[21] Wen Fushuan, Hart Zhenxiang. A Refined Genetic Algorithm for Fault Section Estination In Power Systems Using the Time Sequence Information of Circuit Breakers. Electric Machines and Power Systems, 1996,24:801～815
[22] Wen Fushuan, Chang C S. A Probabilistic Approach to Alarm Processing in Power systems Using a Refined Genetic Algorithm. In: Proceedings of 1996 International Conference on Intelligent Systems
Applications to Power Systems(ISAP'96).Orlando(USA):1996.14～19
[23] When Fushuan, Han Zhenxiang. Fault Section Estimation in Power Systems Using a Genetic Algorithm. Electric Power Systems Research,1995,34:165～172
[24]王耀南，姚志红．神经网络自适应模糊控制其设计与应用．湖南大学学报,1996
[25]徐英，徐用懋，杨尔辅．时变大纯滞后系统的单神经元预测控制．清华大学学报．2002
[26]王桂增，王诗宓．高等过程控制．清华大学出版社，2002．65～70
[27] MICHAEL J, PIOVOSO and KARLENE A.KOSANOVICH. Applications of multivariate statistical methods to process monitoring and controller design. INT.J.CONTRLO, 1994,59(3):743～765
[28] LeeSC,LeeET. Fuzzy Neural Networks. Mathematical Biosciences, 1975,23,151-177
[29] Fukuda Toshio,Ghibata Takanori. Theory and Applications of Neural Networks for Industrial Control Systems. IEEE Transactions on Industrial Electronics, December 1992,39(6):472～489
[30]何坚勇．运筹学基础．清华大学出版社，2000
[31] Narendra K,K. Parthasarathy, Identification and control of dynamical systems using neural networks, IEEETrans. On Neural Networks, 1990, 1(1): 4～27
[32] Narendra K. S, and S. Mukhopadhyay, Adaptive control of nonlinear multivariable systems using neural networks, 1994,7(5): 737～752
[33]丁国峰，王孙安，林廷析等．基于径向基函数网络的一步超前预测控制研究．控制与决策．1996， 11(4)：485～489
[34]林茂琼，陈增强，袁著祉．基于阻尼最小二乘法的神经网络预测偏差补偿自校正控制器．信息与控制，2000．29(1)：27～33
[35]程涛，左静等．基于单个神经元的自适应模糊控制器及其应用．控制理论与应用，1999．16(4)：621～624
[36]姚尹武，熊金涛，毛宗源．一种神经网络自组织模糊控制．控制理论与应用 1996．13(6)：728～743
[37]陈小红，候春海，钱积新 逆系统方法的径向基网络实现．信息与控制，1998，13(2)：146～150
[38] Behera L., M. Gopal and S. Chaudhury. Inversion of RBF networks and applications to adaptive control of nonlinear systems. IEEE Proc. Control Theory Appl. 1995.142(6): 617～624
[39] Garcia C.E. and M.Morari. Internal modelcontrol-l: A uniying review and some new results. Ind.Eng.Chem.Process Des.Dev. 1982.21(2) 308～323
[40]田明，戴汝为．基子动态BP神经网络的系统辨识方法．自动化学报，1993．19(4)：450～453
[2]戴汝为．人工智能．化学工业出版社,2002．63～72
[3]申东日，冯少辉．BP网络改进方法概述．化工自动化及仪表，2000．27(1)：30～32
[4]高雪鹏，丛爽．BP网络改进算法的性能对比研究．控制与决策，2001．16(2)：167～171
[5] Psaltis D, A.Sideris and A.Yamamura. A multilayered neural network controller. IEEE Control Systems Magazine, 1988.8(2):17～21
[6] 徐湘元，毛宗源．过程控制的发展方向智能控制．化工自动化及仪表，1998．25(2)：1～5
[7] Beartrice Lazzerini. Leonardo M.Marcello Chiaberge. A neural-fuzzy approach to hybrid intelligent control[J]. IEEE Trans on Industry Applications, 1999.35(2):413～425
[8]李金厚．神绎网络研穷进展与展望．华东冶金学院学报,2000．17(4)：275～281
[9] X.Dong, Q.xu.Identification and control for chain boiler combustion system based on neural networks.Instrumentation and Control Technology,2003.5253(6): 620～623
[10] Xu Lina.Neural networks control.Harbin Institute of Technology,1999
[11]陈在平，杜太行等．控制系统计算机仿真与CAD．天津大学出版社，2001
[12]孙增圻．智能控制理论与技术．清华大学出版社广西科学技术出版社,2002．126～135
[13] Yang H T, Chang W Y, Huang C L.Power system Distributed On-Line Fault Sectin Estimation Using Decision Tree based Neural Nets Approach. IEEE Transactions on Power Delivery, 1995,10(1):540～546
[14] Sun Y, Jiang H,Wang D.Fault Synthetic Recognition for an EHV Transmission Line Using a Group of Neural Networks with a Time-Space Property. IEE Proceedings—Generation, Transmission and Distribution, 1998,145(3):265-270
[15]肖田元，张燕云，陈加栋．系统仿真导论．清华大学出版社，2002
[16]杨宪勇．人工过程自动控制．清华大学出版社，2000．149～164
[17]方崇智，萧德云．过程辨识．清华大学出版社，2001．98～104
[18]高隽，胡勇，胡良梅．关于AM学习样本选择的实验研究．模式识别与人工智能，2002
[19]杨斌，田永青，朱仲英．智能建模方法中的数据预处理．信息与控制，2002
[20] Sittithumwat A, Tomsovic K, Soudi E Optimizing Maintenance Resources in Distribution System with limited Information. In:ISAP'99. Rio de Janeiro(Brazil):1999
[21] Wen Fushuan, Hart Zhenxiang. A Refined Genetic Algorithm for Fault Section Estination In Power Systems Using the Time Sequence Information of Circuit Breakers. Electric Machines and Power Systems, 1996,24:801～815
[22] Wen Fushuan, Chang C S. A Probabilistic Approach to Alarm Processing in Power systems Using a Refined Genetic Algorithm. In: Proceedings of 1996 International Conference on Intelligent Systems
Applications to Power Systems(ISAP'96).Orlando(USA):1996.14～19
[23] When Fushuan, Han Zhenxiang. Fault Section Estimation in Power Systems Using a Genetic Algorithm. Electric Power Systems Research,1995,34:165～172
[24]王耀南，姚志红．神经网络自适应模糊控制其设计与应用．湖南大学学报,1996
[25]徐英，徐用懋，杨尔辅．时变大纯滞后系统的单神经元预测控制．清华大学学报．2002
[26]王桂增，王诗宓．高等过程控制．清华大学出版社，2002．65～70
[27] MICHAEL J, PIOVOSO and KARLENE A.KOSANOVICH. Applications of multivariate statistical methods to process monitoring and controller design. INT.J.CONTRLO, 1994,59(3):743～765
[28] LeeSC,LeeET. Fuzzy Neural Networks. Mathematical Biosciences, 1975,23,151-177
[29] Fukuda Toshio,Ghibata Takanori. Theory and Applications of Neural Networks for Industrial Control Systems. IEEE Transactions on Industrial Electronics, December 1992,39(6):472～489
[30]何坚勇．运筹学基础．清华大学出版社，2000
[31] Narendra K,K. Parthasarathy, Identification and control of dynamical systems using neural networks, IEEETrans. On Neural Networks, 1990, 1(1): 4～27
[32] Narendra K. S, and S. Mukhopadhyay, Adaptive control of nonlinear multivariable systems using neural networks, 1994,7(5): 737～752
[33]丁国峰，王孙安，林廷析等．基于径向基函数网络的一步超前预测控制研究．控制与决策．1996， 11(4)：485～489
[34]林茂琼，陈增强，袁著祉．基于阻尼最小二乘法的神经网络预测偏差补偿自校正控制器．信息与控制，2000．29(1)：27～33
[35]程涛，左静等．基于单个神经元的自适应模糊控制器及其应用．控制理论与应用，1999．16(4)：621～624
[36]姚尹武，熊金涛，毛宗源．一种神经网络自组织模糊控制．控制理论与应用 1996．13(6)：728～743
[37]陈小红，候春海，钱积新 逆系统方法的径向基网络实现．信息与控制，1998，13(2)：146～150
[38] Behera L., M. Gopal and S. Chaudhury. Inversion of RBF networks and applications to adaptive control of nonlinear systems. IEEE Proc. Control Theory Appl. 1995.142(6): 617～624
[39] Garcia C.E. and M.Morari. Internal modelcontrol-l: A uniying review and some new results. Ind.Eng.Chem.Process Des.Dev. 1982.21(2) 308～323
[40]田明，戴汝为．基子动态BP神经网络的系统辨识方法．自动化学报，1993．19(4)：450～453