铁合金冶炼过程能耗监测与分析研究
|
摘要
铁合金企业是高耗能企业,企业的竞争力主要取决于产品的成本。能耗是成本中可控的主要部分,所以对能耗的研究有助于控制成本中的能耗水平,降低成本,从而增强企业竞争力。要解决铁合金企业的高能耗问题,目前迫切需要解决的两个问题是：一是提高铁合金企业能源信息化管理水平,使企业的计量手段和方法更趋于合理化；二是掌握科学、有效的能耗分析手段。本论文就是针对这两个问题进行研究的。通过对铁合金企业能耗的监测与分析,研究影响能耗的因素,从而找出其中规律,明确企业节能降耗的方向,为做出正确的节能决策提供可靠依据。
本文首先建立铁合金企业能耗监测系统。结合中钢集团吉林铁合金股份有限公司铁合金冶炼过程的生产工艺,开发了基于现场总线、组态技术及以太网络通讯技术等的能耗监测系统,阐述了该系统的设计方法及各模块功能,本系统实现了能源计量的集中化管理、实时能源监测、历史查询、能源统计、数据图形和报表等功能。能耗监测系统提高了冶金企业能源信息化管理水平,提高了企业工作效率,并保证了数据的可靠性,为下一步能耗分析打下良好基础。
接着,在能耗监测系统提供的可靠数据基础上,利用智能算法对铁合金企业能耗进行科学分析和能耗影响因素Pareto构序。首先,构造了铁合金企业能耗分析模型。建立能耗分析模型时,分别采用了三种方法对模型进行训练。第一种方法是标准BP算法；第二种方法是改进的BP算法,采用附加动量法、自适应学习率法及LM算法等来改进BP算法;第三种方法是遗传算法与改进BP网络结合的算法,称为GA-BP算法,此法充分利用了遗传算法全局搜索能力强的特点与人工神经网络模型学习能力强的特点,用遗传算法来优化神经网络初始权重；同时,为了有效提高BP网络收敛速度,采用LM算法作为GA优化后的BP网络的后续训练。有效克服了标准BP算法网络收敛速度慢和容易陷入局部极小值的缺陷。采用上述三种算法,通过Matlab软件在计算机上编程模拟对模型进行训练、仿真,证实了GA-BP算法,在训练速度、精度及泛化能力上均优于其它两种算法,所以在建立能耗分析模型的时,我们选择的是GA-BP算法。能耗分析模型建立好了之后,结合中钢集团吉林铁合金股份有限公司八分厂的实际生产数据,利用训练后的神经网络具有联想记忆和推测的特性,找出各主要因素变化对铁合金生产能耗的影响程度,把各因素对能耗的影响程度进行量化表示,对能耗影响因素进行了定量分析,并对能耗的影响因素进行了Pareto构序,为企业的能源决策提供了理论依据,促进了节能降耗。
Ferroalloy enterprises are high energy consuming enterprises.The competence of company lies on the cost of product.The level of cost reflects the benefit of company.Energy consumption is a main repressible part. So the research of energy consumption helps to control the level of energy consuming of cost,reduce the cost and raise the competence of corporation.At present, to solve the problem of high energy consumption of Ferroalloy enterprises,we urgently need to address two issues:First, we can improve ferroalloy enterprise energy information management level and rationalize enterprise's means and methods of measurement. the second is to master scientific and effective means of analysis of energy consumption.This paper aims to study these investigate some factors influential to energy consumption system these two issues,by monitoring and analysis the Situation of energy consumption in ferroalloy enterprises.It helps to find out the regularity, clear the direction of energy saving companies,and provide the base of right decision of energy saving.
In the first part,Combined with physical circumstance of Sinosteel Jilin Ferroalloy Co., Ltd.'s ferroalloy smelting production process,a energy consumption monitoring system has been designed base on the profibus,configuration software and ethernet network communication technology in this paper. The author expains the constitutions of the system, discusses the problems in realization of the system, and introduces the function of every module in details.This system realizes many functions,such as energy neasurement centralization management,real-time data monitor,historical inquiry, Trend curve, analysis of energy consumption,print statiscal report and so on. This work not only improves managing level of metallurgical enterprise energy information and improves labor productivity effectively but also establish. a good foundation for the following to the analysis of energy consumption and ensure the reliability of the data.
In the second part, The main factors influencing the energy consumption in the smelting process of ferroalloy were quantitatively analyzed and were structured the Pareto order based on reliable data of the energy consumption monitoring.The scientific planning method was obtained. First, we established a energy consumption model of ferroalloy enterprise.Training of model energy consumption, three methods are applied, the first measure is the standard BP algorithm. the second is improved BP algorithm, the use of additional momentum method,adaptive learning rate method and LM algorithm Etc. to improve the BP algorithm.The third way is a synthetic arithmetic based on genetic algorithm (GA) and improved nerve network of BP algorithm (BP)which is called GA-BP algorithm. This measure makes use of excellent global searching ability of GA and fine learning ability of ANN.I use GA to optimize initial weights of neural network to design GA-BP algorithm. Meanwhile, in order to effectively improve the convergence rate of BP network, the use of LM algorithm as GA optimized BP network follow-up training.In a sense, local optimizing problemsand slow convergence,-which is widely existed in BP neural network model training, can be overcome.Testifying the three kind of algorithm above, applying MATLAB software to simulation and comparing these method. confirmed the GA-BP algorithm, the training speed, accuracy and generalization performance are superior to two other algorithms, so the process of establishing model of energy consumption analysis, we Select the synthetic arithmetic based on GA and the improved BP algorithm. Second,After establishing model of energy consumption analysis, the energy intensity of Jilin Ferroalloy Co., Ltd. Eight branch factories was analyzed by the method of intelligent analysis with statistical data. I made use of the characteristics of associative memory and speculation of trained neural network to identify and quantify the extent of which change of the main factors impact on the energy consumption in ferroalloy production. The main factors influencing the energy consumption in the smelting process of ferroalloy were quantitatively analyzed and were structured the Pareto order.The scientific planning method was obtained.The intelligent algorithm will provide essential foundation for seeking method of energy saving, drawing plan of energy saving and making decision of energy saving.
本文首先建立铁合金企业能耗监测系统。结合中钢集团吉林铁合金股份有限公司铁合金冶炼过程的生产工艺,开发了基于现场总线、组态技术及以太网络通讯技术等的能耗监测系统,阐述了该系统的设计方法及各模块功能,本系统实现了能源计量的集中化管理、实时能源监测、历史查询、能源统计、数据图形和报表等功能。能耗监测系统提高了冶金企业能源信息化管理水平,提高了企业工作效率,并保证了数据的可靠性,为下一步能耗分析打下良好基础。
接着,在能耗监测系统提供的可靠数据基础上,利用智能算法对铁合金企业能耗进行科学分析和能耗影响因素Pareto构序。首先,构造了铁合金企业能耗分析模型。建立能耗分析模型时,分别采用了三种方法对模型进行训练。第一种方法是标准BP算法；第二种方法是改进的BP算法,采用附加动量法、自适应学习率法及LM算法等来改进BP算法;第三种方法是遗传算法与改进BP网络结合的算法,称为GA-BP算法,此法充分利用了遗传算法全局搜索能力强的特点与人工神经网络模型学习能力强的特点,用遗传算法来优化神经网络初始权重；同时,为了有效提高BP网络收敛速度,采用LM算法作为GA优化后的BP网络的后续训练。有效克服了标准BP算法网络收敛速度慢和容易陷入局部极小值的缺陷。采用上述三种算法,通过Matlab软件在计算机上编程模拟对模型进行训练、仿真,证实了GA-BP算法,在训练速度、精度及泛化能力上均优于其它两种算法,所以在建立能耗分析模型的时,我们选择的是GA-BP算法。能耗分析模型建立好了之后,结合中钢集团吉林铁合金股份有限公司八分厂的实际生产数据,利用训练后的神经网络具有联想记忆和推测的特性,找出各主要因素变化对铁合金生产能耗的影响程度,把各因素对能耗的影响程度进行量化表示,对能耗影响因素进行了定量分析,并对能耗的影响因素进行了Pareto构序,为企业的能源决策提供了理论依据,促进了节能降耗。
Ferroalloy enterprises are high energy consuming enterprises.The competence of company lies on the cost of product.The level of cost reflects the benefit of company.Energy consumption is a main repressible part. So the research of energy consumption helps to control the level of energy consuming of cost,reduce the cost and raise the competence of corporation.At present, to solve the problem of high energy consumption of Ferroalloy enterprises,we urgently need to address two issues:First, we can improve ferroalloy enterprise energy information management level and rationalize enterprise's means and methods of measurement. the second is to master scientific and effective means of analysis of energy consumption.This paper aims to study these investigate some factors influential to energy consumption system these two issues,by monitoring and analysis the Situation of energy consumption in ferroalloy enterprises.It helps to find out the regularity, clear the direction of energy saving companies,and provide the base of right decision of energy saving.
In the first part,Combined with physical circumstance of Sinosteel Jilin Ferroalloy Co., Ltd.'s ferroalloy smelting production process,a energy consumption monitoring system has been designed base on the profibus,configuration software and ethernet network communication technology in this paper. The author expains the constitutions of the system, discusses the problems in realization of the system, and introduces the function of every module in details.This system realizes many functions,such as energy neasurement centralization management,real-time data monitor,historical inquiry, Trend curve, analysis of energy consumption,print statiscal report and so on. This work not only improves managing level of metallurgical enterprise energy information and improves labor productivity effectively but also establish. a good foundation for the following to the analysis of energy consumption and ensure the reliability of the data.
In the second part, The main factors influencing the energy consumption in the smelting process of ferroalloy were quantitatively analyzed and were structured the Pareto order based on reliable data of the energy consumption monitoring.The scientific planning method was obtained. First, we established a energy consumption model of ferroalloy enterprise.Training of model energy consumption, three methods are applied, the first measure is the standard BP algorithm. the second is improved BP algorithm, the use of additional momentum method,adaptive learning rate method and LM algorithm Etc. to improve the BP algorithm.The third way is a synthetic arithmetic based on genetic algorithm (GA) and improved nerve network of BP algorithm (BP)which is called GA-BP algorithm. This measure makes use of excellent global searching ability of GA and fine learning ability of ANN.I use GA to optimize initial weights of neural network to design GA-BP algorithm. Meanwhile, in order to effectively improve the convergence rate of BP network, the use of LM algorithm as GA optimized BP network follow-up training.In a sense, local optimizing problemsand slow convergence,-which is widely existed in BP neural network model training, can be overcome.Testifying the three kind of algorithm above, applying MATLAB software to simulation and comparing these method. confirmed the GA-BP algorithm, the training speed, accuracy and generalization performance are superior to two other algorithms, so the process of establishing model of energy consumption analysis, we Select the synthetic arithmetic based on GA and the improved BP algorithm. Second,After establishing model of energy consumption analysis, the energy intensity of Jilin Ferroalloy Co., Ltd. Eight branch factories was analyzed by the method of intelligent analysis with statistical data. I made use of the characteristics of associative memory and speculation of trained neural network to identify and quantify the extent of which change of the main factors impact on the energy consumption in ferroalloy production. The main factors influencing the energy consumption in the smelting process of ferroalloy were quantitatively analyzed and were structured the Pareto order.The scientific planning method was obtained.The intelligent algorithm will provide essential foundation for seeking method of energy saving, drawing plan of energy saving and making decision of energy saving.
引文
[1]陈和平.中国能源可持续发展战略刍议[J].山西能源与节能.2001,22(3)：1-2.
[2]杨宏韬,张德江,李秀兰等.遗传神经网络能耗预测模型在钢铁企业中的应用[J].长春工业大学学报(自然科学版).2007,28(B07)：186-189.
[3]凌振华.钢铁企业能源管理信息系统的研究[D].长沙：中南大学,2005.
[4]胡敏艺,马荣骏.神经网络在冶金工业中的应用[J].湖南有色金属,2000,16(5)：16-19.
[5]刘祥官,刘显著,蒋美华等,智能控制专家系统在莱钢高炉上的应用[J].冶金工业自动化,2002,(6)：15-19.
[6]肖树成,杨清敏,徐业鹏.钢铁厂系统节能及其经济效益[J].北京科技大学学报,1994,16(4)：309-314.
[7]朱里红.铁合金冶炼过程能耗预测与决策优化的研究[D].长春：长春工业大学,2009.
[8]栾心汉,唐琳,李小明等.铁合金生产节能及精炼技术.西安：西北工业大学出版社,2006.
[9]唐喜文,李祥飞,柏森.热装法生产中低碳铬铁[J].铁合金,2002,(5)：7-9.
[10]高峰,铁合金节能技术[J].中国冶金,2001,(6)：23-25.
[11]张加云.铁合金生产过程能耗系统预测及优化的研究[D].长春：长春工业大学,2009.
[12]杨世明.我国铁合金工业能耗概况及节能措施[J].铁合金,1999,(4)：38-42.
[13]俞金寿.过程自动化及仪表[M].北京：化学工业出版社,2003.
[14]李正军.现场总线与工业以太网及其应用系统设计[M].北京：人民邮电出版社,2006.
[15]廖学勤.现场总线技术的最新进展[J].自动化仪表,2002,7：11-13.
[16]Graham Loose.Fieldbus and ints impact on the automation user.COMPUTING& CONTROL ENGINEERING JOURNAL,1995:259-262.
[17]AV Scott,WJ Buchanan.Truly Distributed Control Systems using Fieldbus Technology. Engineering of Computer Based Systems. April,2000:165-173.
[18]罗成方.漫谈现场总线[J].四川轻化工学院学报,1996,(3)：59-62.
[19]李昌禧.智能仪表原理与设计[M].北京：化学工业出版社,2005,(2)：155-158.
[20]Wayne Tomasi著.数据通信与联网技术[M].张宝生,孙岩译.北京：清华大学出版社,2006,(2)：533.
[21]杨金奇,刘学军.工业以太网技术及其应用现状与发展[J].四川工业学院学报,2002,21(3)：34-36.
[22]紫金桥监控组态软件用户手册[G].大庆：大庆紫金桥软件技术有限公司.2007.
[23]方彦军,王运涛,李世红.冶金企业能源计量与管理系统的实现探讨[J].微计算机信息,2004,9(20)：39-42.
[24]杨行俊,郑君里.人工神经网络[M].北京：高等教育出版社,1992.
[25]邱忠文.玻璃器皿模具号的计算机视觉识别[D].广西：广西师范大学,2000.
[26]袁曾任.人工神经网络及其应用[M].北京：清华大学出版社,1999.
[27]Paul J Werbos.Backpropagation Through Time:What It Does and How It.Proceedings of The IEEE,1990, (10):1550-1560.
[28]王科俊,王克成.神经网络建模、预报与控制[M].哈尔滨：哈尔滨工程大学出书版社,1996.
[29]刘金琨.智能控制[M].北京：电子工业出版社,2005.5.
[30]田景文,高美娟.人工神经网络算法研究及应用[M].北京：北京理工大学出版社,2006.
[31]杨忠林,张静,川培根.改进的BP网络及其在电路故障诊断中的应用[J].舰船电子工程,2006,(6)：103-106.
[32]楼顺天,施阳.基于MATLAB的系统分析与设计--神经网络[M].西安：西安电子科技大学出版社,1998.
[33]飞思科技产品研发中心.神经网络理论与MATLAB7实现[M].北京：电子工业出版社,2005.3.
[34]J.Holland.Adaptation in natural and artificial system [J].Ann, Arbor:The University of Miehigan Press.1975.
[35]Holland J H.AdaPtation in Nature and Artifieial Systems [M].Ann Arbor, MI: University of Michigan Press,1975.
[36]周明,孙树栋.遗传算法原理及应用[M].北京：国防工业出版社.1999.
[37]Bornholdt A. General Asymmetric Neural Networks and Structrue Design by Generic Algorithms[J].Neural Networks.1992,5(2):67-71.
[38]焦李成.神经网络系统理论[M].西安：西安电子科技大学出版社,1990.
[39]王耀南.智能控制系统[M].长沙：湖南大学出版社,1996.
[40]李敏强,徐博艺,寇纪淞.遗传算法与神经网络结合[J].系统工程理论与实践,1999.
[41]刘洁,魏连雨,杨春风.基于遗传-神经网络的交通量预测[J].长安大学学报,2003,23(1)：68～70.
[42]李伟超,宋大猛,陈斌.基于遗传算法的人工神经网络[J].计算机工程与设计.2006,27(2)：316～318.
[43]陈郁.遗传神经网络在模拟电路故障诊断中的应用[D].长春：长春理工大学.2009.
[44]谢小山.基于遗传算法和BP神经网络的铁路客运量预测研究[D].成都：西南交通大学.2010.
[45]庄镇泉,王熙法,王东生.神经网络与神经计算机.北京：科学出版社,1992.
[46]周继成.人工神经网络[M].北京：科学普及出版社,1993.
[47]谢安国.钢铁工业和企业能源消耗系统分析方法的研究[D].沈阳：东北大学,1997.
[48]蔡红鑫,谢安国,苑安民.人工神经网络在炼铁工序能耗分析中的应用[J].上海应用技术学院报,2006,(2)：75-80.
[49]SHEN W X,CHAN C C, LO E W C, et al.A new battery available capacity indicator for electric vehicles using neural network[J].Journal of Power Sources,2002,43 (6): 817-826.
[50]Zhong Luo, Liu Lisheng, Zou Chengming, et al. The application of neural network in life time prediction of concrete[J].Journal of Wuhan University of Technology,2002, 17(1):79-81.
[51]徐丽娜.神经网络控制[M].北京：电子工业出版社,2003.
[52]刘威,李小平,毛慧欧,柴天佑.基于实数编码遗传算法的神经网络成本预测模型及其应用[J].控制理论与应用,2004,21(3)：423-431.
[53]Tan K C, Tang H J.New Dynamical optimal learning for Linear Multilayer FNN[J].IEEE Transactions on Neural Network,2004, (99):1-7.
[54]杨行峻,郑君里.人工神经网络[M].高等教育出版社,1992.
[55]张鸿宾.神经网络的基本结构和功能.计算机科学[J],1989,(6)：7-13.
[56]刘洁,魏连雨,杨春风.基于遗传-神经网络的交通量预测[J].长安大学学报,2003,23(1)：68-70.
[57]阎平凡,张长水.人工神经网络与模拟进化计算[M].北京：清华大学出版社,2000.
[58]谢安国,陆钟武.人工神经网络及在钢铁企业能耗分析中应用方法的研究[J].钢铁,1997,32(7)：60-63,56.
[2]杨宏韬,张德江,李秀兰等.遗传神经网络能耗预测模型在钢铁企业中的应用[J].长春工业大学学报(自然科学版).2007,28(B07)：186-189.
[3]凌振华.钢铁企业能源管理信息系统的研究[D].长沙：中南大学,2005.
[4]胡敏艺,马荣骏.神经网络在冶金工业中的应用[J].湖南有色金属,2000,16(5)：16-19.
[5]刘祥官,刘显著,蒋美华等,智能控制专家系统在莱钢高炉上的应用[J].冶金工业自动化,2002,(6)：15-19.
[6]肖树成,杨清敏,徐业鹏.钢铁厂系统节能及其经济效益[J].北京科技大学学报,1994,16(4)：309-314.
[7]朱里红.铁合金冶炼过程能耗预测与决策优化的研究[D].长春：长春工业大学,2009.
[8]栾心汉,唐琳,李小明等.铁合金生产节能及精炼技术.西安：西北工业大学出版社,2006.
[9]唐喜文,李祥飞,柏森.热装法生产中低碳铬铁[J].铁合金,2002,(5)：7-9.
[10]高峰,铁合金节能技术[J].中国冶金,2001,(6)：23-25.
[11]张加云.铁合金生产过程能耗系统预测及优化的研究[D].长春：长春工业大学,2009.
[12]杨世明.我国铁合金工业能耗概况及节能措施[J].铁合金,1999,(4)：38-42.
[13]俞金寿.过程自动化及仪表[M].北京：化学工业出版社,2003.
[14]李正军.现场总线与工业以太网及其应用系统设计[M].北京：人民邮电出版社,2006.
[15]廖学勤.现场总线技术的最新进展[J].自动化仪表,2002,7：11-13.
[16]Graham Loose.Fieldbus and ints impact on the automation user.COMPUTING& CONTROL ENGINEERING JOURNAL,1995:259-262.
[17]AV Scott,WJ Buchanan.Truly Distributed Control Systems using Fieldbus Technology. Engineering of Computer Based Systems. April,2000:165-173.
[18]罗成方.漫谈现场总线[J].四川轻化工学院学报,1996,(3)：59-62.
[19]李昌禧.智能仪表原理与设计[M].北京：化学工业出版社,2005,(2)：155-158.
[20]Wayne Tomasi著.数据通信与联网技术[M].张宝生,孙岩译.北京：清华大学出版社,2006,(2)：533.
[21]杨金奇,刘学军.工业以太网技术及其应用现状与发展[J].四川工业学院学报,2002,21(3)：34-36.
[22]紫金桥监控组态软件用户手册[G].大庆：大庆紫金桥软件技术有限公司.2007.
[23]方彦军,王运涛,李世红.冶金企业能源计量与管理系统的实现探讨[J].微计算机信息,2004,9(20)：39-42.
[24]杨行俊,郑君里.人工神经网络[M].北京：高等教育出版社,1992.
[25]邱忠文.玻璃器皿模具号的计算机视觉识别[D].广西：广西师范大学,2000.
[26]袁曾任.人工神经网络及其应用[M].北京：清华大学出版社,1999.
[27]Paul J Werbos.Backpropagation Through Time:What It Does and How It.Proceedings of The IEEE,1990, (10):1550-1560.
[28]王科俊,王克成.神经网络建模、预报与控制[M].哈尔滨：哈尔滨工程大学出书版社,1996.
[29]刘金琨.智能控制[M].北京：电子工业出版社,2005.5.
[30]田景文,高美娟.人工神经网络算法研究及应用[M].北京：北京理工大学出版社,2006.
[31]杨忠林,张静,川培根.改进的BP网络及其在电路故障诊断中的应用[J].舰船电子工程,2006,(6)：103-106.
[32]楼顺天,施阳.基于MATLAB的系统分析与设计--神经网络[M].西安：西安电子科技大学出版社,1998.
[33]飞思科技产品研发中心.神经网络理论与MATLAB7实现[M].北京：电子工业出版社,2005.3.
[34]J.Holland.Adaptation in natural and artificial system [J].Ann, Arbor:The University of Miehigan Press.1975.
[35]Holland J H.AdaPtation in Nature and Artifieial Systems [M].Ann Arbor, MI: University of Michigan Press,1975.
[36]周明,孙树栋.遗传算法原理及应用[M].北京：国防工业出版社.1999.
[37]Bornholdt A. General Asymmetric Neural Networks and Structrue Design by Generic Algorithms[J].Neural Networks.1992,5(2):67-71.
[38]焦李成.神经网络系统理论[M].西安：西安电子科技大学出版社,1990.
[39]王耀南.智能控制系统[M].长沙：湖南大学出版社,1996.
[40]李敏强,徐博艺,寇纪淞.遗传算法与神经网络结合[J].系统工程理论与实践,1999.
[41]刘洁,魏连雨,杨春风.基于遗传-神经网络的交通量预测[J].长安大学学报,2003,23(1)：68～70.
[42]李伟超,宋大猛,陈斌.基于遗传算法的人工神经网络[J].计算机工程与设计.2006,27(2)：316～318.
[43]陈郁.遗传神经网络在模拟电路故障诊断中的应用[D].长春：长春理工大学.2009.
[44]谢小山.基于遗传算法和BP神经网络的铁路客运量预测研究[D].成都：西南交通大学.2010.
[45]庄镇泉,王熙法,王东生.神经网络与神经计算机.北京：科学出版社,1992.
[46]周继成.人工神经网络[M].北京：科学普及出版社,1993.
[47]谢安国.钢铁工业和企业能源消耗系统分析方法的研究[D].沈阳：东北大学,1997.
[48]蔡红鑫,谢安国,苑安民.人工神经网络在炼铁工序能耗分析中的应用[J].上海应用技术学院报,2006,(2)：75-80.
[49]SHEN W X,CHAN C C, LO E W C, et al.A new battery available capacity indicator for electric vehicles using neural network[J].Journal of Power Sources,2002,43 (6): 817-826.
[50]Zhong Luo, Liu Lisheng, Zou Chengming, et al. The application of neural network in life time prediction of concrete[J].Journal of Wuhan University of Technology,2002, 17(1):79-81.
[51]徐丽娜.神经网络控制[M].北京：电子工业出版社,2003.
[52]刘威,李小平,毛慧欧,柴天佑.基于实数编码遗传算法的神经网络成本预测模型及其应用[J].控制理论与应用,2004,21(3)：423-431.
[53]Tan K C, Tang H J.New Dynamical optimal learning for Linear Multilayer FNN[J].IEEE Transactions on Neural Network,2004, (99):1-7.
[54]杨行峻,郑君里.人工神经网络[M].高等教育出版社,1992.
[55]张鸿宾.神经网络的基本结构和功能.计算机科学[J],1989,(6)：7-13.
[56]刘洁,魏连雨,杨春风.基于遗传-神经网络的交通量预测[J].长安大学学报,2003,23(1)：68-70.
[57]阎平凡,张长水.人工神经网络与模拟进化计算[M].北京：清华大学出版社,2000.
[58]谢安国,陆钟武.人工神经网络及在钢铁企业能耗分析中应用方法的研究[J].钢铁,1997,32(7)：60-63,56.