基于 ESN 的污水处理多变量自适应预测控制  

Multivariable Adaptive Predictive Control for Wastewater Treatment Process Based on ESN

作　　者：乔俊飞[1] 王莉莉[1] 韩红桂[1] 

机构地区：[1]北京工业大学电子信息与控制工程学院,北京100124

出　　处：《信息与控制》2014年第3期368-373,380,共7页Information and Control

基　　金：国家自然科学基金资助项目(61034008,61203099,61225016);北京市自然科学基金资助项目(4122006);教育部博士点新教师基金资助项目(20121103120020)

摘　　要：针对污水处理过程高度非线性、大滞后等特征,提出了一种基于回声状态网络(echo state network,ESN)模型的多变量自适应预测控制系统.首先,利用ESN建立污水处理过程的智能预测模型,该模型能够预测污水处理的输出;其次,设计污水处理过程的ESN辨识器,将辨识器输出与实际输出的差对主控制器进行误差补偿;最后,以仿真基准模型(BSM1)为平台,采用提出的多变量自适应控制方法对溶解氧浓度和硝态氮浓度进行控制,实验结果表明,该控制方法提高了系统的自适应性和抗干扰能力,能够对溶解氧浓度和硝态氮浓度实现快速、准确跟踪.Due to the highly nonlinear and time delay characteristics of wastewater treatment processes, a kind of muhivariable adaptive predictive control system based on the echo state network （ESN） model is proposed. First, an intelligent predictive model is established using ESN to predict the outputs of the wastewater treatment process. Second, the ESN identifier is established in order to compensate for error generated by the differences between the actual outputs and the identifier outputs. Finally, experiments are designed based on the BSM1. The proposed multivariable adaptive control strategy is used to control the dissolved oxygen concentration and nitrate concentration. The experimental results show that this control method improves the a- daptability and anti-interference ability, and achieves rapid and accurate tracking of dissolved oxygen concen- tration and nitrate concentration.

关 键 词：污水处理过程 ESN(echo state network)辨识器 ESN预测模型 多变量预测控制 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]