论文部分内容阅读
为了克服模型与过程间的偏差,提出了一个基于时变偏扰模型的间歇过程迭代学习控制方法。利用主元回归(principal component regression,PCR)和部分最小二乘(partial least squares,PLS)方法,可以得到过程对象在正常运行轨迹附近线性化的模型。前一批次的模型预报误差被用来修正当前批次的模型预报值。每完成一个批次就利用新得到的数据对模型进行更新,该更新的模型也是在前一批次的控制轨迹基础上进行线性化得到的。主元回归和部分最小二乘方法能克服批次内不同阶段的控制量存在的相关关系从而得到更准确的模型。仿真结果表明:基于PCR和PLS模型的控制效果要好于基于多元线性回归(MLR)模型的控制效果。
In order to overcome the deviation between the model and the process, an iterative learning control method based on the time-varying disturbance model is proposed. Using principal component regression (PCR) and partial least squares (PLS) methods, a model of the process object can be linearized near the normal operation trajectory. The previous batch of model prediction error is used to correct the current batch model prediction value. The model is updated with new data every time a batch is completed. The updated model is also linearized on the basis of the previous batch of control trajectories. The principal component regression and the partial least squares method can overcome the correlation between the control quantities at different stages in a batch and thus obtain a more accurate model. The simulation results show that the control effect based on PCR and PLS model is better than the control effect based on multiple linear regression (MLR) model.