Abstract:

In order to restrain the influence of non-stationary disturbance to VCM distillation process with integral characteristics by the acetylene method, an effective adaptive MPC method was presented. Due to the integral characteristics of the high boiling tower caused by settling of high-boiling residues to the bottom of tower, liquid level of the high boiling tower was taken as integral variable. Furthermore, in order to restrain the influence of non-stationary disturbance to liquid level of the high boiling tower caused by feed temperature of VCM and other factors, firstly, RELS (recursive extended least squares) algorithm was used to estimate the disturbance; secondly, the proportion of prediction error caused by the disturbance was calculated. Lastly, twiddle factor of adaptive MPC was updated real-time to realize the control of liquid level. Industrial experiments indicated that the adaptive MPC could overcome the non-stationary disturbance effectively, and standard deviation of liquid level was 2.6616 which was reduced by 60.7% than before. The effectiveness of the adaptive MPC was proved.

Key words: model predictive control, process control, algorithm, integrating variable, rotation factor, recursive extended least squares algorithm, non-stationary disturbance

摘要：

为抑制非平稳扰动对具有积分特性的乙炔法VCM精馏过程的影响,提出了一种有效的自适应MPC方法。由于高沸塔塔釜中会不断沉积高沸物,表现出很强的积分特性,因此,将高沸塔塔釜液位作为积分变量进行控制;同时,针对VCM进料温度等因素对高沸塔塔釜液位产生的非平稳扰动,首先,利用RELS算法实时的估计影响高沸塔塔釜液位的扰动,然后,计算扰动在预测误差中所占的比例,最后,通过实时更新旋转因子的数值实现对积分过程的自适应MPC。工业试验结果表明:提出的自适应MPC方法能够有效克服非平稳扰动,高沸塔塔釜液位的标准差为2.6616,比采用自适应MPC方法之前减少了60.7%,验证了该方法的有效性。

关键词: 模型预测控制, 过程控制, 算法, 积分变量, 旋转因子, 递推增广最小二乘算法, 非平稳扰动