基于神经网络的双进双出磨煤机混合模型研究
|
摘要
火电厂中最重要的部分是由制粉系统组成的,而磨煤机作为火电厂的制粉系统其运行的安全性、可靠性和经济性直接影响到锅炉以及整个火电厂机组的安全性、可靠性和经济性。随着发电厂对节能减排的要求不断提高,双进双出磨煤机具有能耗低、生产效率高、研磨煤种范围广等优点,所以称为火电厂普遍采用的制粉系统。双进双出磨煤机设备的安全、经济运行的前提是对其进行有效的控制,而对双进双出磨煤机有效控制必须要有高精度的模型做依据,因此,建立一个高精度的双进双出磨煤机模型是提高发电厂磨煤机优化控制有待解决的问题。
双进双出磨煤机是强耦合、多输入多输出、大滞后的非线性系统,它的工作特性会随着运行工况的变化而变化,到目前为止,传统的建模方法存在测量精度不高,抗干扰性差缺点,本文针对难以建立双进双出磨煤机高精度模型这一问题展开研究。
首先介绍了双进双出磨煤机的基本结构、工作原理及工作特性曲线,根据特性曲线对其运行参数及影响因素进行分析。
然后本文提出了一种混合建模方法,该方法是在建立双进双出磨煤机机理模型的基础上建立了神经网络模型,基于两种模型的加权融合来求得混合模型。在神经网络模型中利用免疫聚类确定最佳的RBF神经网络隐含层中心位置,利用粒子群算法优化神经网络权值从而可以提高模型的精确度。混合建模方法既反应了双进双出磨煤机实际系统的主要规律又能体现外界条件引起的不确定因素对实际系统的影响,改善了双进双出磨煤机难以建立精确模型难题。
最后应用上述双进双出磨煤机混合智能建模方法对阜新电厂机组BBD-4360型双进双出磨煤机进行建模研究,仿真实验表明,混合模型能正确地反映出双进双出磨煤机制粉过程的输入输出参数之间的关系,具有较高的精度,可以很好地为其优化控制算法提供很好的依据。
The milling system is the most important part of the whole coal. fired power plant. Ball mill as the milling system of power plant, whether the operation of the ball mill is good will not only directly impact on the safety, economic and stability of the operation of the boiler, but also directly relate to the safety, economic and stability of the operation for the entire unit of the plant. With the Power of the increasing demands of energy saving, a BBD ball mill has low energy consumption, high efficiency, wide range of grinding coal, etc. The BBD ball mills is commonly used in thermal power plant milling system BBD ball mills equipment, safety, economic operation on the premise that their effective control, while the BBD ball mill must have a valid basis to do high. Precision models, the establishment of the precise parameter model of BBD ball mill system is to be used to solve the problem.
In this paper, high accuracy is difficult to establish precise parameter BBD ball mill, first introduced the basic structure of the mill, working principle and operating characteristics, operating parameters under the curve and influence of its factors were analyzed.
BBD ball mill is a nonlinear system of strong coupling, large time delay and multi input output. Its features of the work will change with the change of operating conditions. In this paper, a hybrid modeling method is presented to establish mechanism and Mill network models and to obtain the hybrid model through the weighted fusion of these two models. Under neural network model, the location of optimal Radial Basis Function (RBF) neural network hidden layer centre is determined using Immune Clustering and the accuracy of the model is improved by implementing Particle Swarm Optimization (PSO) to optimize weight of the neural network. The proposed method achieves excellent effect of global modeling and takes System's operating characteristics into account at the same time.
Finally, using the above hybrid model based on neural networks and Mechanism models study on the historical data at Fuxin BBD.4360 ball mill. Simulation results show that the hybrid model precisely reflects the relationship between input and output parameters of this Double Ball Mill milling process, and has high accuracy. Can be available basis for well optimizing control algorithms.
双进双出磨煤机是强耦合、多输入多输出、大滞后的非线性系统,它的工作特性会随着运行工况的变化而变化,到目前为止,传统的建模方法存在测量精度不高,抗干扰性差缺点,本文针对难以建立双进双出磨煤机高精度模型这一问题展开研究。
首先介绍了双进双出磨煤机的基本结构、工作原理及工作特性曲线,根据特性曲线对其运行参数及影响因素进行分析。
然后本文提出了一种混合建模方法,该方法是在建立双进双出磨煤机机理模型的基础上建立了神经网络模型,基于两种模型的加权融合来求得混合模型。在神经网络模型中利用免疫聚类确定最佳的RBF神经网络隐含层中心位置,利用粒子群算法优化神经网络权值从而可以提高模型的精确度。混合建模方法既反应了双进双出磨煤机实际系统的主要规律又能体现外界条件引起的不确定因素对实际系统的影响,改善了双进双出磨煤机难以建立精确模型难题。
最后应用上述双进双出磨煤机混合智能建模方法对阜新电厂机组BBD-4360型双进双出磨煤机进行建模研究,仿真实验表明,混合模型能正确地反映出双进双出磨煤机制粉过程的输入输出参数之间的关系,具有较高的精度,可以很好地为其优化控制算法提供很好的依据。
The milling system is the most important part of the whole coal. fired power plant. Ball mill as the milling system of power plant, whether the operation of the ball mill is good will not only directly impact on the safety, economic and stability of the operation of the boiler, but also directly relate to the safety, economic and stability of the operation for the entire unit of the plant. With the Power of the increasing demands of energy saving, a BBD ball mill has low energy consumption, high efficiency, wide range of grinding coal, etc. The BBD ball mills is commonly used in thermal power plant milling system BBD ball mills equipment, safety, economic operation on the premise that their effective control, while the BBD ball mill must have a valid basis to do high. Precision models, the establishment of the precise parameter model of BBD ball mill system is to be used to solve the problem.
In this paper, high accuracy is difficult to establish precise parameter BBD ball mill, first introduced the basic structure of the mill, working principle and operating characteristics, operating parameters under the curve and influence of its factors were analyzed.
BBD ball mill is a nonlinear system of strong coupling, large time delay and multi input output. Its features of the work will change with the change of operating conditions. In this paper, a hybrid modeling method is presented to establish mechanism and Mill network models and to obtain the hybrid model through the weighted fusion of these two models. Under neural network model, the location of optimal Radial Basis Function (RBF) neural network hidden layer centre is determined using Immune Clustering and the accuracy of the model is improved by implementing Particle Swarm Optimization (PSO) to optimize weight of the neural network. The proposed method achieves excellent effect of global modeling and takes System's operating characteristics into account at the same time.
Finally, using the above hybrid model based on neural networks and Mechanism models study on the historical data at Fuxin BBD.4360 ball mill. Simulation results show that the hybrid model precisely reflects the relationship between input and output parameters of this Double Ball Mill milling process, and has high accuracy. Can be available basis for well optimizing control algorithms.
引文
[1]袁钢.双进双出钢球磨煤机控制系统研究：(硕士学位论文).长沙理工大学,2009.
[2]张珂.神经网络在直吹式制粉系统中速磨煤机建模中的应用研究：(硕士学位论文).东南大学,2006.
[3]崔宝侠,曲星宇,段勇.基于多信息融合的双进双出磨煤机料位检测.沈阳工业大学大学报,2010,32(2)：182-186.
[4]章臣樾.锅炉的动态特性及数学模型.水利电力出版社,1987.
[5]谢铮桂,钟少丹,韦玉科.基于免疫聚类的神经网络集成的研究.微计算机信息(测控自动化),2010,26(21)：154-180.
[6]童健华,谭洪舟,郭雷勇.一种改进的人工免疫文本聚类算法.系统仿真学报,2010,22(1)：29-32.
[7]范哗.基于免疫机制的神经网络及其应用研究：(硕士学位论文).上海交通大学,2008.
[8]郝勇生,张柯,李军.基于新型RBF神经网络的中速磨煤机建模.热力发电,2007,(12)：14-18.
[9]李林,李建兵,牛鹏超.基于粒子群算法的RBF神经网络优化方法.山东电力高等专科学校学报,2010,13(1)：51-53.
[10]王学厚,韩璞,李岩等.基于PSO的RBF神经网络在热工系统辨识中的应用.华北电力大学学报,2008,35(1)：52-56.
[11]崔海青,刘希玉.基于粒子群算法的RBF网络参数优化算法.计算机技术与发展,2009,19(12)：117-119.
[12]张君,岳恒,柴天佑.钢球磨煤机制粉过程的混合智能建模.控制工程,2009,16(3)：282-285.
[13]王东风,韩璞,纂守荣.中储式钢球磨制粉系统的建模与优化控制研究：(博士学位论文).华北电力大学,2008.
[14]中国动力工程学会.火力发电设备技术手册(第四卷.火电站系统与辅机).机械工业出版社,2000.
[15]韦巍.智能控制技术.北京：机械工业出版社,2004：61-65.
[16]潘吴,侯清兰.基于粒子群优化算法的BP网络学习研究.计算机工程与应用,2006,17(10)：41-43.
[17]熊伟丽,徐保国,孙俊.改进PSO算法在BP网络训练中的应用.系统仿真学报,2005,17(09)：2078-2080.
[18]F-van den Bergh, A.P.Engelbrecht.A study of particle swarm optimization particle trajectories. Information Sciences,2006,176(08):937-941.
[19]Powell M J D. Radial basis functions for multivariable interpolations:a review. In IMA Conference on Algorithms for Approximation of Functions and Data. RMCS, Shriven ham UK,1985, 143-167.
[20]Sutanto E L,et al. Mean. Tracking clustering algorithm for radial basis function centre selection.Int.J.Control,1997,67 (6):961-977.
[21]Chen S, Wu Y, Luck B L. Combined genetic algorithm optimization and regularized orthogonal least squares learning for radial basis function. IEEE Transactions on Neural Networks,1999,10 (5):1239-1213.
[22]Karakinos N B. Gradient descent learning of radial basis neural networks. Proceedings of the IEEE Conference on Artificial Neural Networks,1997, vol.3.
[23]Li Y, Sunderararajan N, Saratchandran P-Analysis of minimal radial basis function network algorithm for real-time identification of nonlinear dynamic systems. IEEE Proc-Control Theory Application,2000,147 (4):476-484.
[24]Castro L N D, Ruben F J V. An evolutionary immune network for data clustering. In Proceedings of the IEEE Suborn (Brazilian Symposium On Artificial Neural Networks), Rio De Janeiro,2000,82-89.
[25]CUI Bao-xia, LI Rui, DUAN Yong et al. Study of BBD Mill Measure Method Based on Rough Set and NN Information Fusion.2008 IEEE Pacific. Asia Workshop on Computational Intelligence and Industrial Application,2008:585-587.
[26]Bao-xia CUI, Xing-yu QU, DUAN Yong. Study of BBD Ball Mill Material Measure Based on Rough Sets and RBF Neural Network Data Fusion. International Conference on Intelligent Human. Machine Systems and Cybernetics,2009:237-240.
[27]Kolacz.Measurement System of the Mill Charge in Grinding Ball Mill Circuits. Minerals Engineering,1997:1329-1338.
[28]陈凯.基于智能算法的双进双出钢球磨煤机建模与优化控制：(硕士学位论文).沈阳：沈阳工业大学,2008.
[29]崔宝侠,苏桂华.采用智能方法对双进双出钢球磨煤机进行优化控制：(硕十学位论文).沈阳：沈阳工业大学,2006.
[30]李瑞.双进双出磨煤机料位检测与控制：(硕士学位论文).沈阳：沈阳工业大学,2008.
[31]田传波.专家控制技术在双进双出磨煤机中的应用研究：(硕士学位论文).长春：东北电力大学,2008.
[32]明学星,吕震中,于向军.对于球磨机制粉系统控制及优化策略的研讨.测控技术,2007,26(1)：38-40.
[33]华烟旻.基于神经网络的中储式球磨机控制的研究：(硕士学位论文).南京：东南大学,2006.
[34]张小桃,倪维斗,等.基于现场数据的中速磨煤机动态建模研究。热能动力工程,2004,6(11)：614-616.
[35]刘志远.径向基函数神经网络及在电厂热工工程建模和控制中的应用研究：(博十学位论文).南京：东南大学,2003.
[36]李月英,申东日,陈义俊.基于RBF神经网络的非线性系统的预测.计算机测量与控制,2006,14(3)：319-321。
[37]谭代伦,刘益,张世禄.运用机理分析法巧解数学模型题.洛阳师范学院学报,2007(2)：29-31.
[38]胡雄辉,刘武林,李劲柏.直吹式双进双出钢球磨煤机仿真料位自动的研究.湖南电力,2009,29(1)：14-17.
[39]常敬涛,马平.基于机理模型的锅炉过热器的动态特性仿真及研究.电力科学与工程,2009,25(10)：28-30.
[40]岳俊红.复杂L业过程多模型预测控制策略及其应用研究：(博士学位论文).北京：华北电力大学,2007.
[41]王艳芹.非线性系统的径向基函数神经网络建模方法研究：(硕士学位论文).大庆：大庆石油学院,2005.
[42]肖本贤,王晓伟,刘一福.MPSO-RBF优化策略在锅炉过热系统辨识中的仿真研究.系统仿真学报,2007,19(6)：1382-1387.
[43]张继东,唐国华,姚树建等。基于神经网络算法计算双进双出磨煤机工作参数的研究.热力发电,2007,12(5)：23-26.
[44]常绿,杨涛,姚树建等.基于神经网络和遗传算法的磨煤机结构和工作参数的优化.热能动力工程,2007,22(1)：69-72.
[2]张珂.神经网络在直吹式制粉系统中速磨煤机建模中的应用研究：(硕士学位论文).东南大学,2006.
[3]崔宝侠,曲星宇,段勇.基于多信息融合的双进双出磨煤机料位检测.沈阳工业大学大学报,2010,32(2)：182-186.
[4]章臣樾.锅炉的动态特性及数学模型.水利电力出版社,1987.
[5]谢铮桂,钟少丹,韦玉科.基于免疫聚类的神经网络集成的研究.微计算机信息(测控自动化),2010,26(21)：154-180.
[6]童健华,谭洪舟,郭雷勇.一种改进的人工免疫文本聚类算法.系统仿真学报,2010,22(1)：29-32.
[7]范哗.基于免疫机制的神经网络及其应用研究：(硕士学位论文).上海交通大学,2008.
[8]郝勇生,张柯,李军.基于新型RBF神经网络的中速磨煤机建模.热力发电,2007,(12)：14-18.
[9]李林,李建兵,牛鹏超.基于粒子群算法的RBF神经网络优化方法.山东电力高等专科学校学报,2010,13(1)：51-53.
[10]王学厚,韩璞,李岩等.基于PSO的RBF神经网络在热工系统辨识中的应用.华北电力大学学报,2008,35(1)：52-56.
[11]崔海青,刘希玉.基于粒子群算法的RBF网络参数优化算法.计算机技术与发展,2009,19(12)：117-119.
[12]张君,岳恒,柴天佑.钢球磨煤机制粉过程的混合智能建模.控制工程,2009,16(3)：282-285.
[13]王东风,韩璞,纂守荣.中储式钢球磨制粉系统的建模与优化控制研究：(博士学位论文).华北电力大学,2008.
[14]中国动力工程学会.火力发电设备技术手册(第四卷.火电站系统与辅机).机械工业出版社,2000.
[15]韦巍.智能控制技术.北京：机械工业出版社,2004：61-65.
[16]潘吴,侯清兰.基于粒子群优化算法的BP网络学习研究.计算机工程与应用,2006,17(10)：41-43.
[17]熊伟丽,徐保国,孙俊.改进PSO算法在BP网络训练中的应用.系统仿真学报,2005,17(09)：2078-2080.
[18]F-van den Bergh, A.P.Engelbrecht.A study of particle swarm optimization particle trajectories. Information Sciences,2006,176(08):937-941.
[19]Powell M J D. Radial basis functions for multivariable interpolations:a review. In IMA Conference on Algorithms for Approximation of Functions and Data. RMCS, Shriven ham UK,1985, 143-167.
[20]Sutanto E L,et al. Mean. Tracking clustering algorithm for radial basis function centre selection.Int.J.Control,1997,67 (6):961-977.
[21]Chen S, Wu Y, Luck B L. Combined genetic algorithm optimization and regularized orthogonal least squares learning for radial basis function. IEEE Transactions on Neural Networks,1999,10 (5):1239-1213.
[22]Karakinos N B. Gradient descent learning of radial basis neural networks. Proceedings of the IEEE Conference on Artificial Neural Networks,1997, vol.3.
[23]Li Y, Sunderararajan N, Saratchandran P-Analysis of minimal radial basis function network algorithm for real-time identification of nonlinear dynamic systems. IEEE Proc-Control Theory Application,2000,147 (4):476-484.
[24]Castro L N D, Ruben F J V. An evolutionary immune network for data clustering. In Proceedings of the IEEE Suborn (Brazilian Symposium On Artificial Neural Networks), Rio De Janeiro,2000,82-89.
[25]CUI Bao-xia, LI Rui, DUAN Yong et al. Study of BBD Mill Measure Method Based on Rough Set and NN Information Fusion.2008 IEEE Pacific. Asia Workshop on Computational Intelligence and Industrial Application,2008:585-587.
[26]Bao-xia CUI, Xing-yu QU, DUAN Yong. Study of BBD Ball Mill Material Measure Based on Rough Sets and RBF Neural Network Data Fusion. International Conference on Intelligent Human. Machine Systems and Cybernetics,2009:237-240.
[27]Kolacz.Measurement System of the Mill Charge in Grinding Ball Mill Circuits. Minerals Engineering,1997:1329-1338.
[28]陈凯.基于智能算法的双进双出钢球磨煤机建模与优化控制：(硕士学位论文).沈阳：沈阳工业大学,2008.
[29]崔宝侠,苏桂华.采用智能方法对双进双出钢球磨煤机进行优化控制：(硕十学位论文).沈阳：沈阳工业大学,2006.
[30]李瑞.双进双出磨煤机料位检测与控制：(硕士学位论文).沈阳：沈阳工业大学,2008.
[31]田传波.专家控制技术在双进双出磨煤机中的应用研究：(硕士学位论文).长春：东北电力大学,2008.
[32]明学星,吕震中,于向军.对于球磨机制粉系统控制及优化策略的研讨.测控技术,2007,26(1)：38-40.
[33]华烟旻.基于神经网络的中储式球磨机控制的研究：(硕士学位论文).南京：东南大学,2006.
[34]张小桃,倪维斗,等.基于现场数据的中速磨煤机动态建模研究。热能动力工程,2004,6(11)：614-616.
[35]刘志远.径向基函数神经网络及在电厂热工工程建模和控制中的应用研究：(博十学位论文).南京：东南大学,2003.
[36]李月英,申东日,陈义俊.基于RBF神经网络的非线性系统的预测.计算机测量与控制,2006,14(3)：319-321。
[37]谭代伦,刘益,张世禄.运用机理分析法巧解数学模型题.洛阳师范学院学报,2007(2)：29-31.
[38]胡雄辉,刘武林,李劲柏.直吹式双进双出钢球磨煤机仿真料位自动的研究.湖南电力,2009,29(1)：14-17.
[39]常敬涛,马平.基于机理模型的锅炉过热器的动态特性仿真及研究.电力科学与工程,2009,25(10)：28-30.
[40]岳俊红.复杂L业过程多模型预测控制策略及其应用研究：(博士学位论文).北京：华北电力大学,2007.
[41]王艳芹.非线性系统的径向基函数神经网络建模方法研究：(硕士学位论文).大庆：大庆石油学院,2005.
[42]肖本贤,王晓伟,刘一福.MPSO-RBF优化策略在锅炉过热系统辨识中的仿真研究.系统仿真学报,2007,19(6)：1382-1387.
[43]张继东,唐国华,姚树建等。基于神经网络算法计算双进双出磨煤机工作参数的研究.热力发电,2007,12(5)：23-26.
[44]常绿,杨涛,姚树建等.基于神经网络和遗传算法的磨煤机结构和工作参数的优化.热能动力工程,2007,22(1)：69-72.