针对具有系统状态测量延迟和输入输出约束的酒精发酵系统,研究了其鲁棒预测控制器的设计问题。介绍了酒精连续发酵过程及其数学模型,通过在平衡工作点附近的线性化处理得到系统的线性化模型;考虑到发酵系统线性化模型的参数变化,采用多胞结构来描述发酵过程。以此为基础,借助LMI(Linear Matrix Inequality)工具将鲁棒预测控制器的在线优化问题转化为求解一组由线性矩阵不等式构成的优化问题。另外,针对系统部分状态的实时不可测性,通过二次规划求解进行状态估计,利用估算值为优化计算提供有效信息,解决了系统状态测量时延给预测控制器处理带来的困难。仿真结果证明了该方法的有效性。 Aiming at the alcohol fermentation system with system state measurement delay and input and output constraints, the design of robust predictive controller is studied. The continuous alcohol fermentation process and its mathematical model were introduced. The linearization model of the system was obtained by the linearization process near the equilibrium working point. The fermentation process was described by the multi-cell structure considering the parameter changes of the linearization model of the fermentation system. Based on this, the LMI (Linear Matrix Inequality) tool is used to transform the problem of online optimization of robust predictive controllers into solving a set of optimization problems consisting of linear matrix inequalities. In addition, aiming at the real-time untestability of the state of the system, the state estimation is carried out through quadratic programming. The estimated value is used to provide effective information for the optimization calculation. This solves the difficulties of the system state measurement delay in the processing of the predictive controller. Simulation results show the effectiveness of this method.