酵母培养过程的数学模型及参数和状态的实时估计

由于缺乏实时测量生化过程某些参数的传感器,本文提出了一种应用气体分析和估计原理相结合的方法,去建立酵母培养过程的数学模型。首先在带计算机检测系统的ISF_200型发酵装置上作了充分的实验。在积累了大量实验数据基础上,根据氧消耗和能量产生之间的关系建立了酵母培养过程的观测方程。又根据物料衡算建立了酵母培养过程的状态方程。然后用单纯形—阿达姆斯寻优方法确定系统模型参数,用卡尔曼滤波方法实时估计发酵液中的一些关键状态(菌体浓度、比生长速率及基质浓度),并进一步用扩展卡尔曼滤波方法进行模型参数和状态同时的实时估计。从采集数据到得出估计结果所经历的时间不超过3分钟。与实验结果对比,说明所建立的数学模型能较精确地描述酵母培养过程。由于实现了过程状态和模型参数的实时估计,为进一步实现过程的最优控制奠定了基础。

THE MATHEMATICAL MODEL OF YEAST FERMENTATION PROCESS AND REAL-TIME ESTIMATION OF PARAMETER AND STATE

Because of the absence of sensors for measurement of biochemical process data, in this paper a new method is presented, which applies gas analysis and estimation theory to real-time estimating some important state varibles in culture medium, such as biomass concentration, specific cell growth rate and substrate concentration. In response to relatation of oxygen consumption and energy production, observation model of yeast culture process is obtained. In response to material balance, the state model of yeast culture process is built. The parameters of model are estimaed by Simplex-Adams method. Real-time estimates of states by Kalman fiiter achive satisfactory results. The conbined real-time estimation of model parameters and states are obtained by the extead Kalman filter. The efficacy of this method in engineering is illustrated by experiment.