Abstract: Fermentation process is characterized by high non-linearity,time-delay and uncertainty,which directly influence the control performance of the fermentation system.In this paper,a fed-batch fermentation process predictive control method based on non-linear quadratic Gaussian(NLQG)was proposed.The method was composed of an extended Kalman filter(EKF)and non-linear quadratic regulator(NLQR)connected in series.In order to realize the dynamic predictive control of fed-batch fermentation process,the optimal state estimation and the real-time state feedback of the controlled variable were realized by EKF and NLQR respectively.The NLQG controller was designed in the LabVIEW by the strategy using ActiveX to invoke the COM components which were compiled and generated in MATLAB.The controller was applied to the predictive control of the dissolved oxygen concentration in penicillin fed-batch fermentation process.The experiments showed that the proposed fed-batch fermentation process predictive control method had good tracking for the change of set value,high control accuracy and good robustness in different noisy environments.

Key words: fed-batch fermentation, predictive control, non-linear quadratic Gaussian controller, extended Kalman filter

摘要： 发酵过程具有高度非线性、时滞性和不确定性等特征,直接影响着发酵过程的有效调控。提出了一种基于非线性二次高斯(NLQG)的分批补料发酵过程预测控制方法。该方法由扩展Kalman滤波器(EKF)和NLQR串联构成,EKF给出被控变量的最优状态估计,NLQR获得被控变量的实时状态反馈,以实现分批补料发酵过程的动态预测控制。在LabVIEW软件平台中,利用ActiveX控件调用MATLAB环境下编译生成的COM组件设计了NLQG控制器,并用于青霉素分批补料发酵过程溶氧浓度的预测控制。实验结果表明,所提出的分批补料发酵过程预测控制方法对于被控变量的设定值变化有良好的跟踪效果,在不同的噪声环境下均能获得较高的控制精度,具有较强的鲁棒性。

关键词: 分批补料发酵过程, 预测控制, 非线性二次高斯控制器, 扩展Kalman滤波器