摘要
生物技术作为21世纪经济发展的关键技术之一,在化工、医药卫生、农林牧渔、轻工食品、能源和环境等行业和领域都发挥着越来越重要的作用。随着生物技术的迅猛发展,生物工程的规模不断扩大,生物工程对自动控制技术的要求越来越迫切。生物反应过程依赖于多个环境因子,具有非线性、多变量、强耦合等特征,内部机理非常复杂,特别是一些直接反映生物反应品质的关键生化参量无法在线实时检测,成为制约生物反应过程优化控制的瓶颈问题,严重影响了生物反应过程的自动化生产。生物反应过程的先进检测和控制技术已成为生物工程领域的重要研究内容,其研究对于提高装备自动化水平,促进生物工程技术革新,提高生产效率和经济效益,培育生物新品种等都具有重大意义。
论文在国家高技术研究发展计划“基于模糊神经逆的生物反应过程软测量方法及优化控制研究(2007AA04Z179)”、高等学校博士学科点专项科研基金“基于模糊神经逆的生物反应过程软测量方法及优化控制(20070299010)”等项目资助下开展研究,基于数据驱动方法,建立了生物反应过程菌体浓度、基质浓度、产物浓度等关键生化参量的混沌粒子群最小二乘支持向量机(CPSO-LSSVM)软测量模型；提出了基于广义预测控制(GPC)与粒子群(PSO)滚动优化的生物反应过程闭环控制方法；基于.NET开发平台和C#语言,开发了具有关键生化参量软测量、在线优化控制等功能的工程化软件,并以氨基酸类典型菌种赖氨酸为例,开展了相关的实验研究。主要研究工作如下：
1.针对生物反应过程关键生化参量无法在线检测的难题,提出生物反应过程不可直接测量参量的数据驱动软测量建模方法。将最小二乘支持向量机(LS-SVM)的非线性函数逼近能力、混沌运动的随机性和遍历性、带极值扰动粒子群(tPSO)的全局优化能力等优势有机结合,建立了典型生物反应过程菌体浓度、基质浓度和产物浓度等关键参量的CPSO-LSSVM软测量模型,并与常规LS-SVM、BP神经网络等软测量方法进行了对比研究,结果表明,CPSO-LSSVM软测量方法具有建模精度高、泛化能力强、收敛速度快等一系列优点。
2.针对生物反应过程多变量、非线性、机理模型难以获得等难题,提出基于数据驱动的生物反应过程精确化模型建立方法。基于反应过程数据,以径向基函数(RBF)为核函数,建立典型反应过程的LS-SVM (?)(?)线性控制模型,并基于采样点线性化方法,建立了典型反应过程的分段线性精确化模型,模型具有结构简单、精度高等优点。
3.以分段线性精确化模型为参考模型,提出了基于广义预测控制的生物反应过程连续流加补料控制方法,实现典型生物反应过程的在线优化控制；提出基于PSO优化的预测控制律设计方法,解决基于Diophantine方程的传统预测控制律设计方法存在的难以获得最优解、计算复杂度高等难题,有效提高了生物反应过程优化控制的实时性与控制精度。
4.基于.NET平台和C#语言,给出了CPSO-LSSVM软测量、广义预测控制等核心算法的实现策略,研发了典型生物反应过程的智能化、模块化测控软件系统。系统具有温度、压力、溶氧等物理参量测量、菌体浓度、基质浓度和产物浓度等生化参量软测量、多变量优化补料控制等功能。性能稳定,可靠性高。
Biology technology gives scope to more and more function in chemical industry, medicine hygienism, agriculture, forestry, husbandry, fishery industry, foods light industry, energy and environment as one of the critical technology in the economy development in 21 century. Biology project puts forward higher and higher demand to automatic control technology as the biotechnology boosts sharply, the scale of bioengineering enlarges. The process of biological fermentation depends on multiple environmental factors featured as dynamic, nonlinear, multivariable and strong coupling. What's more, biochemical reaction mechanism in biological fermentation is so complicated that the key biological parameters which directly reflect on the characters in biological fermentation can't be measured real-time on line. This issue has become the bottleneck for restricting optimal control in biological fermentation, which seriously influences the automatic production in the process of biological fermentation. The important research in the field of biological project has focused on the advanced measurement and control technology in the process of biological fermentation process. The research is so significant to raise the automatic level for the equipment, to promote innovation for biological engineering technology, to enhance production efficiency and economic effectiveness and cultivate new variety in biology.
The dissertation was supported by the National High Technology Research and Development Program ("863 Program"), one project of which is the Research for the Soft Sensor Method and Optimal Control based on FNN Inverse in the Process of Biological Fermentation under Grant 2007 AA04Z179. It is also supported by Specialized Research Fund for the Doctoral Program of Higher Education of China, one project of which is the Research for the Soft Sensor Method and Optimal Control based on FNN Inverse in the Process of Biological Fermentation under Grant 20070299010. The dissertation focuses on building the CPSO- LSSVM soft sensor model which is vital to key biological parameter in cell concentration, substrate concentration, production concentrate in the process of biological fermentation based on the method of data driven. It also puts forward the closed loop control method based on generalized predictive control (GPC) and particle swarm optimization (PSO) rolling optimization in the process of biological fermentation. The dissertation also develops the engineering soft functioning in key biological parameter and optimal control on line based on NET technology platform and C# language. It also demonstrates amino acids and typical strain lysine, conducts the relevant experimental research. The main research is as the following:
1. The data-driven softsensor model in bio-chemical fermentation process has been put forward as key biological parameters checked straight and online are not available. The CPSO-LSSVM soft sensor model, including key parameters in typical bio-chemical reaction like cell, substrate, and product concentration, is established based on the combination of approximate ability of nonlinear function of LS-SVM. randomness and ergodicity of chaos motion and overall optimization of tPSO. and owns a series of advantages such as high precision of modeling, capacity of generalization and high speed of convergence with contrast to regular soft-sense measurement like LS-SVM, BP nerve net, and so on.
2. Ways on how to set precise models have been offered as there are many variables in the bio-chemical fermentation and it is difficult to obtain nonlinear and mechanism model. Based on data in reaction and RBF as core function, LS-SVM nonlinear control model is set in typical fermentation process, as well as the piecewise linear precision model, with the adoption of sampling point and linearization, which possesses some advantages like simple structure and high accuracy.
3. On the basis of the generalized predictive control theory, how to control the continuous feeding in the process and how to optimize the control online has been raised, as well as design methods on predictive control based on PSO optimization. The real-time optimal control and precision gets improved effectively for some though problems like optimum unavailable and complicated calculation get resolved as there are deficiencies in the traditional predictive control design based on Diophantine formula.
4. Based on NET platform and C language, strategies of core calculation on CPSO-LSSVM soft-sensor and generalized predictive control have been explored. as well as the intelligent and modularized sense-and-control software system, stable and reliable, which can measure physical parameters like temperature, pressure and dissolved oxygen and sense bio-chemical parameters like cell, substrate, and product concentration, and can optimize feeding control with variables.
引文
[1]葛爱冬.软测量技术在青霉素发酵过程中的应用[D].山东大学,2004.
[2]黄丽.基于神经网络软测量的ε-聚赖氨酸发酵过程控制的研究[D].江苏大学,2006.
[3]童豪.生物发酵过程中的软测量技术应用研究[D].浙江大学,2004.
[4]王鲜芳.生化过程动态建模及优化控制研究[D].江南大学,2009.
[5]王博.基于模糊神经逆的生物反应过程关键参数软测量的研究[Z].20081.
[6]刘瑞兰.软测量技术若干问题的研究及工业应用[D].浙江大学,2004.
[7]Beluhan D, Gosak D, Pavlovic N, et al. Biomass estimation and optimal control of the baker's yeast fermentation process[J]. Computers\& chemical engineering,1995,19:387-392.
[8]Zhao Y. Studies on modeling and control of continuous biotechnical processes[D]. Norway: Norwegian University of Science and Technology,1996.
[9]Tatiraju S, Soroush M, Mutharasan R. Multi-rate nonlinear state and parameter estimation in a bioreactor[J]. Biotechnology and bioengineering,1999,63(1):22-32.
[10]Chae H J. Delisa M P. Cha H J, et al. Framework for online optimization of recombinant protein expression in high-cell-density Escherichia coli cultures using GFP-fusion monitoring[J]. Biotechnology and bioengineering,2000,69(3):275-285.
[11]Lubenova V. On-line estimation of biomass concentration and non stationary parameters for aerobic bioprocesses[J]. Journal of biotechnology,1996,46(3):197-207.
[12]Yuan Z, Bogaert H, Devisscher M, et al. On-line estimation of the maximum specific growth rate of nitrifiers in activated sludge systems[J]. Biotechnology and bioengineering,1999,65(3):265-273.
[13]Takiguchi N, Shimizu H, Shioya S. An on-line physiological state recognition system for the lysine fermentation process based on a metabolic reaction model[J]. Biotechnology and bioengineering, 1997,55(1):170-181.
[14]Feng M, Glassey J. Physiological state-specific models in estimation of recombinant Escherichia coli fermentation performance[J]. Biotechnology and bioengineering,2000,69(5):495-503.
[15]Warnes M R, Glassey J, Montague G A, et al. On data-based modelling techniques for fermentation processes[J]. Process Biochemistry,1996,31(2):147-155.
[16]张泉灵,金晓明.基于主元分析和模糊模型的链霉素发酵过程建模[J].无锡轻工业学院学报,2000，19(005)：446-450.
[17]Mcgovern A C, Broadhurst D, Taylor J, et al. Monitoring of complex industrial bioprocesses for metabolite concentrations using modern spectroscopies and machine learning:application to gibberellic acid production[J]. Biotechnology and bioengineering,2002,78(5):527-538.
[18]Montague G A, Morris A J, Tham M T. Enhancing bioprocess operability with generic software sensors[J]. Journal of biotechnology,1992,25(1-2):183-201.
[19]Glassey J, Ignova M, Ward A C, et al. Bioprocess supervision:neural networks and knowledge based systems[J]. Journal of biotechnology,1997,52(3):201-205.
[20]Bhowmik U K, Saha G, Barua A, et al. On-Line Detection of Contamination in a Bioprocess Using Artificial Neural Networks[J]. Chemical engineering\& technology,2000,23(6):543-549.
[21]Cimander C, Carlsson M, Mandenius C F. Sensor fusion for on-line monitoring of yoghurt fermentation[J]. Journal of biotechnology,2002,99(3):237-248.
[22]Valdez-Castro L, Baruch I, Barrera-Cortes J. Neural networks applied to the prediction of fed-batch fermentation kinetics of Bacillus thuringiensis[J]. Bioprocess and biosystems engineering,2003,25(4): 229-233.
[23]Dacosta P, Kordich C, Williams D, et al. Estimation of inaccessible fermentation states with variable inoculum sizes[J]. Artificial intelligence in engineering,1997,11(4):383-392.
[24]高学金,王普,孙崇正,等.微生物发酵过程建模与优化控制[J].控制工程,2006,13(002)：152-153.
[25]乔晓艳,贾莲凤.RBF神经网络在菌体细胞浓度软测量中的应用[J].计算机工程与设计,2003,24(3)：55-57.
[26]黄明志,杭海峰.人工神经网络在红霉素发酵过程状态预估中的应用[J].华东理工大学学报：自然科学版,2000,26(002)：162-164.
[27]李运锋,袁景淇.头孢菌素C发酵过程状态变量及效益函数预报方法[J].化工学报,2005,56(007)：1281-1283.
[28]Ronen M, Shabtai Y, Guterman H. Hybrid model building methodology using unsupervised fuzzy clustering and supervised neural networks[J]. Biotechnology and bioengineering,2002,77(4): 420-429.
[29]殷铭,张兴华,戴先中.基于模糊神经网络的发酵过程溶解氧预估控制[J].控制与决策，2000,15(5)：523-526.
[30]方千山,王永初.基于神经网络的模糊预测控制及其应用[J].自动化仪表,2002,23(008)：11-13.
[31]高学金,王普,孙崇正,等.一种建立发酵过程模型的新方法[J].北京工业大学学报,2006,32(005)：405-409.
[32]马勇,黄德先,金以慧.基于支持向量机的软测量建模方法[J].信息与控制,2004,33(004)：417-421.
[33]孙宗海,孙优贤.关联向量机在微生物发酵传感器故障诊断中的应用[J].高校化学工程学报,2004,18(004)：483-487.
[34]许光,俞欢军,陈德钊.基于支持向量机的柠檬酸发酵过程统计建模[J].化学反应工程与工艺，2004，20(001)：59-63.
[35]常玉清,王福利,王小刚,等.基于支持向量机的软测量方法及其在生化过群中的应用[J].仪器仪表学报，2006，27(003)：241-244.
[36]桑海峰,王福利,何大阔,等.基于多支持向量机的诺西肽发酵中菌体浓度软测量[J].系统仿真学报，2006，18(007)：1983-1986.
[37]Wang J, Yu T, Jin C. On-line Estimation of Biomass in Fermentation Process Using Support Vector Machinel[J]. Chinese Journal of Chemical Engineering,2006,14(3):383-388.
[38]徐进荣,潘丰.基于粒子群和支持向量机为青霉素发酵建模[J].计算机与应用化学,2007,24(011)：1458-1460.
[39]俞佩菲,吴燕玲,卢建刚,等.基于PCA和LS—SVM的软测量建模与应用[J].江南大学学报：自然科学版,2006,5(002)：182-186.
[40]桑海峰,干福利,何大阔,等.基于最小二乘支持向量机的发酵过程混合建模[J].仪器仪表学报,2006，27(6)：629-633.
[41]桑海峰,苑玮琦,王福利,等.基于机理知识与最小二乘支持向量机的诺西肽发酵过程混合建模方法[J].系统仿真学报,2008,20(002)：468-472.
[42]许光,俞欢军,陶少辉,等.与机理杂交的支持向量机为发酵过程建模[J].化工学报，2005，56(4)：659-664.
[43]隋青美,王正欧.发酵过程混合神经网络模型及其仿真[J].系统仿真学报,2002,14(004)：415-417.
[44]Dai X. Wang W, Ding Y, et al. "Assumed inherent sensor" inversion based ANN dynamic soft-sensing method and its application in erythromycin fermentation process[J]. Computers& Chemical Engineering,2006,30(8):1203-1225.
[45]Takiguchi N, Shimizu H, Shioya S. An on-line physiological state recognition system for the lysine fermentation process based on a metabolic reaction model[J]. Biotechnology and bioengineering, 1997.55(1):170-181.
[46]Pinchuk R J, Brown W A, Hughes S M. et al. Modeling of biological processes using self-cycling fermentation and genetic algorithms[J]. Biotechnology and bioengineering,2000,67(1):19-24.
[47]Woo S H, Park J M, Rittmann B E. Evaluation of the interaction between biodegradation and sorption of phenanthrene in soil-slurry systems[J]. Biotechnology and bioengineering,2001,73(1):12-24.
[48]Nguang X D, Others. Robust [infinity] control of a class of continuous fermentation processes[J]. Applied mathematics letters,1999,12(6):61-69.
[49]Renard F, Wouwer A V, Valentinotti S, et al. A practical robust control scheme for yeast fed-batch cultures-an experimental validation[J]. Journal of Process Control,2006,16(8):855-864.
[50]Stoyanov S, Simeonov I. Robust compensator control of continuous fermentation processes[J]. Bioprocess and Biosystems Engineering,1996,15(6):295-300.
[51]孙玉坤,王博,嵇小辅,等.基于模糊神经网络α阶逆系统的发酵过程多变量解耦控制(英文)[J].控制理论与应用，2010(2)：188-192.
[52]孙晓天,靳其兵,张瑶,等.基于逆系统的赖氨酸发酵多变量解耦内模控制[J].电子技术应用，2010(2)：91-94.
[53]Ben Youssef C,Guillou V, Olmos-Dichara A. Modelling and adaptive control strategy in a lactic fermentation process[J]. Control Engineering Practice,2000,8(11):1297-1307.
[54]Soons Z, Voogt J A, Van Straten G, et al. Constant specific growth rate in fed-batch cultivation of Bordetella pertussis using adaptive control[J]. Journal of biotechnology,2006,125(2):252-268.
[55]孙西,金以慧.双线性系统的自适应控制及其在谷氨酸pH值控制中的应用[J].自动化学报,1995,21(002)：209-213.
[56]Horiuchi J I, Kishimoto M. Application of fuzzy control to industrial bioprocesses in Japan[J]. Fuzzy sets and Systems,2002,128(1):117-124.
[57]Karakuzu C,T U Rker M, O Zt U Rk S I I. Modelling, on-line state estimation and fuzzy control of production scale fed-batch baker's yeast fermentation[J]. Control Engineering Practice,2006,14(8): 959-974.
[58]Lennox B, Kipling K, Glassey J, et al. Automated production support for the bioprocess industry[J]. Biotechnology progress,2002,18(2):269-275.
[59]Hrni I K P, N A HI I K J, Vovs I K J. The BIOGENES system for knowledge-based bioprocess control[J]. Expert Systems with Applications,2002,23(2):145-153.
[60]王斌,王孙安.生物发酵过程的温度控制模型研究[J].西安交通大学学报,2004,38(7)：737-740.
[61]隋青美,王正欧.基于神经网络的多变量发酵过程自适应控制[J].信息与控制,2002,31(4)：371-374.
[62]Hisbullah H M A, Ramachandran K B. Design of a fuzzy logic controller for regulating substrate feed to fed-batch fermentation[J]. Food Biopro Proc,2003,81(C2):138-146.
[63]王斌,王孙安.生物发酵过程的温度控制模型研究[J].西安交通大学学报,2004(7)：737-740.
[64]Rodrigues J, Maciel Filho R. Production optimization with operating constraints for a fed-batch reactor with DMC predictive control[J]. Chemical engineering science,1999,54(13-14):2745-2751.
[65]Ashoori A, Moshiri B, Khaki-Sedigh A, et al. Optimal control of a nonlinear fed-batch fermentation process using model predictive approach[J]. Journal of Process Control,2009,19(7):1162-1173.
[66]Ramaswamy S. Cutright T J. Qammar H K. Control of a continuous bioreactor using model predictive control[J]. Process Biochemistry.2005.40(8):2763-2770.
[67]Nagy Z K. Model based control of a yeast fermentation bioreactor using optimally designed artificial neural networks[J]. Chemical Engineering Journal.2007,127(1-3):95-109.
[68]习毅.发酵过程预测控制研究[D].江南大学,2008.
[69]冯絮影.发酵过程混合建模及带动态补偿的非线性预测控制方法研究[D].北京化工大学，2010.
[70]谢磊,张泉灵,王树青,等.基于多模型的自适应预测函数控制[J].浙江大学学报(工学版),2003,37(2)：64-67.
[71]Taitel Y, Dukler A E. Flow pattern transition in gas—liquid systems[J]. Multiphase Science and Technology,1986,2.
[72]Hetsroni G. Handbook of multiphase systems[J].1982.
[73]Matsui G. Identification of flow regimes in vertical gas-liquid two-phase flow using differential pressure fluctuations[J]. International journal of multiphase flow,1984,10(6):711-719.
[74]李修亮.软测量建模方法研究与应用[D].浙江大学，2009.
[75]方瑞明著.支持向量机理论及其应用分析[M].北京：中国电力出版社，2007.
[76]李超峰.基于支持向量机的非线性系统广义预测控制研究与应用[D].浙江大学,2010.
[77]陈明忠.基于支持向量机的微生物发酵过程软测量方法研究[D].江苏大学,2008.
[78]李涛.基于SVM和PSO的非线性模型预测控制及应用研究[D].上海交通大学,2008.
[79]杨琳.基于支持向量机的微生物发酵过程生物量软测量方法研究[D].北京化工大学,2005.
[80]潘红华.预测控制技术及其应用研究[D].浙江大学,2000.
[81]诸静等.智能预测控制及其应用[M].杭州市：浙江大学出版社,2002.
[82]王伟.广义预测控制理论及其应用[M].北京市：科学出版社,1998.
[83]丁宝苍.预测控制的理论与方法[M].北京市：机械工业出版社,2008.
[84]李国勇.智能预测控制及其MATLAB实现[M].北京市：电子工业出版社,2010.
[85]施冬梅.基于SVM的船舶航向广义预测控制[D].苏州大学,2008.
[86]Kennedy J, Eberhart R. Particle swarm optimization[C].1995.
[87]Eberhart R, Kennedy J. A new optimizer using particle swarm theory[C].1995.
[88]胡旺.图像融合中的关键技术研究[D].四川大学,2006.
[89]袁君,陈贝,朱光灿.采用混沌粒子群优化算法的水质模型参数辨识[J].东南大学学报(自然科学版),2009(5)：1018-1022.
[90]仇国庆,唐贤伦,庄陵,等.基于混沌粒子群优化的神经网络在旋转机械故障诊断中的应用[J].中国机械工程,2008(21)：2642-2645.
[91]吕晓明,黄考利,连光耀.基于混沌粒子群优化的系统级故障诊断策略优化[J].系统工程与电子技术,2010(1)：217-220.
[92]唐贤伦,庄陵,李学勤,等.基于混沌粒子群支持向量回归的高炉铁水硅含量预测[J].四川大学学报(工程科学版),2009(4)：141-145.
[93]王波,徐泽柱,高松波.混沌粒子群优化算法在短期负荷预测中的应用[J].水电能源科学.2009(2)：208-211.
[94]徐生林,史燕,杨成忠.应用混沌粒子群优化训练的BP神经网络预报高炉铁水含硅量[J].冶金分析,2010(5)：9-13.
[95]李勇,邵诚.一种新的灰关联分析算法在软测量中的应用[J].自动化学报,2006(2)：311-317.
[96]郭健,朱瑞军.一类非线性系统的自适应预测控制[J].控制理论与应用,2002,19(001)：68-72.