无线网络环境下数据驱动混合选别浓密过程双率控制方法  

Data-driven Dual-rate Control for Mixed Separation Thickening Process in a Wireless Network Environment

作　　者：吴倩 范家璐 姜艺 柴天佑 WU Qian;FAN Jia-Lu;JIANG Yi;CHAI Tian-You(State Key Laboratory of Synthetical Automation for Process Industries, Northeastern University, Shenyang 110819)

机构地区：[1]东北大学流程工业综合自动化国家重点实验室

出　　处：《自动化学报》2019年第6期1122-1135,共14页Acta Automatica Sinica

基　　金：国家自然科学基金(61333012,61533015,61304028);中央高校基本科研专项资金(N160804001)资助~~

摘　　要：无线网络环境下赤铁矿混合选别浓密过程控制问题是以底流矿浆泵频率为内环输入,以底流矿浆流量为内环输出外环输入,以底流矿浆浓度为外环输出的非线性串级工业过程控制问题.其外环反馈回路存在丢包,且模型参数难以辨识,故本文利用工业运行过程的在线数据,设计不依赖模型参数的跟踪控制器.首先,利用浓密过程运行在工作点附近的特点进行线性化,对流量过程设计Q-学习控制器,保证流量过程能够跟踪给定的流量设定值;然后采用提升技术,得到统一时间尺度的以底流矿浆流量设定值为输入,以底流矿浆浓度为输出的被控对象;最后,考虑到在无线网络环境下浓度过程存在反馈丢包,当前的状态可能无法获得,故采用史密斯预估器的思想,利用历史的数据估计系统当前的状态,设计丢包Q-学习设定值控制器为流量过程提供最优设定值.通过仿真实验验证所提算法的有效性.The mixed separation thickening process(MSTP) of hematite beneficiation in a wireless network environment is a nonlinear cascade process with the frequency of underflow slurry pump as the inner loop input, the slurry flow-rate as the inner loop output and the concentration as the outer loop output. The dropout occurs in the outer feedback loop,making it difficult to identify the parameters of the model, so the tracking controller only using the data generated by operational processes and independent of the knowledge of model parameters is designed in this paper. First, linearize the thickening system near the steady states, then design a controller based on Q-learning algorithm to make the inner process trace the set-point of the slurry flow-rate. Second, use the lifting technology to obtain the uniform time scale controlled object with the set-point of the slurry flow-rate as the input and the concentration as the output. Finally, considering that the networked-induced feedback dropout exists in the feedback process, meaning the current state information may be lost, a novel Smith predictor is developed to predict the current state from historical measured data, and a dropout Q-learning method is designed to provide the optimal set-point of lower loop. A simulation experiment on MSTP is given to show the effectiveness of the proposed method.

关 键 词：混合选别浓密过程 Q-学习 丢包 史密斯预估器 

分 类 号：TN92[电子电信—通信与信息系统] TD951[电子电信—信息与通信工程]