中文摘要：

化工过程中,由于外部环境变化等因素引起的模型失配及不可测扰动会对预测控制效果带来一定的影响。对化工过程施加预测控制时,预测控制中通常采用反馈校正的方式来解决此类问题,即直接将预测输出与实际输出的差值作为校正量补偿到预测控制算法中。其不足之处是当模型失配程度增大时,无法实现良好的控制。鉴于状态估计能够利用实际测量值来估计系统状态的优点,本文提出一种卡尔曼滤波校正的预测控制方法,通过分析模型内在机制来补偿上述问题对控制效果造成的影响。该方法首先将底层回路和被控对象当作广义对象模型,预测控制器直接控制广义对象模型;然后将模型参数的变化等效成扰动,通过分析得到扰动的统计特性后,用卡尔曼滤波方法估计模型参数发生变化后系统的真实状态;最后将估计状态代入预测控制算法的优化求解中,实现系统的优化控制。仿真实例验证了该方法的有效性。

英文摘要：

In chemical processes, when there is model mismatch and immeasurable disturbance, the prediction model is no longer accurate, resulting poor performance of the model predictive control（MPC） strategy. In traditional MPC, feedback correction is usually adopted to compensate the difference between the predictive output and the actual plant output. However, When the model mismatch is large, the compensation strategy does not work. In this paper, a kalman filter correction method is proposed, in which the system is compensated by analyzing the intrinsic mechanism of the model. In the method, the PID loop and the plant are firstly regarded as a unified model. And then the change of the parameters of the plant model is modelled as disturbance. After analyzing the statistical characteristics of the disturbance, the kalman filter is used to estimate the real state of the system. Finally, the estimated state are incorporated into the MPC optimization problem. A simulation example shows that the proposed method is effective.