基于小波递归神经网络的间歇过程迭代学习优化控制

针对间歇过程提出了基于小波神经网络的迭代学习优化控制算法，实现产品终点质量指标的控制。小波递归神经网络用于建立提供长期预测的间歇过程模型。由于模型误差以及未知干扰的影响，基于预测模型得到的控制变量在实际应用中得不到期望的终点质量指标。利用间歇过程的重复特性，采用迭代学习优化控制改善批次间的产品质量，根据以前批次的模型预测误差均值来修正神经网络模型预测输出，继而计算出下一个批次的控制输入。随着批次的进行，模型误差逐渐消失，控制输入达到最优控制。仿真实验结果验证了该算法的有效性。

Batch Process Iterative Learning Control Based on Wavelet Recurrent Neural Network

An iterative learning control (ILC) algorithm based on wavelet recurrent neural network(WRNN) was proposed to control product final quality in batch process. Wavelet recurrent neural network was used to establish the model of long range batch process. Due to model errors and unmeasured disturbances, the calculated control policy based on WRNN model may not be optimal when applied to the actual process. By utilizing the repetitive characteristic of batch process, ILC was used to improve product final quality from batch to batch. According to the mean value of previous prediction errors, the prediction output of neural network model was modified and the control input was computed for the next batch. And then, the model errors were gradually reduced from batch to batch, and the control inputs attained the optimal control policy. The effectiveness is verified via a simulated batch process.