时滞过程中流量调节的自适应学习方法研究  

Study on Adaptive Learning Method for Flow Regulation in Time-Delay Process

作　　者：范磊 艾昌文[2] FAN Lei;AI Changwen(School of Information,Yunnan University,Kunming 650504,China;Provincial Electronic Computing Center,Yunnan University,Kunming 650223,China)

机构地区：[1]云南大学信息学院,云南昆明650504 [2]云南大学省电子计算中心,云南昆明650223

出　　处：《自动化仪表》2020年第9期64-67,共4页Process Automation Instrumentation

摘　　要：对于具有非线性、时滞等特性的复杂系统,传统的控制方法难以适用,控制品质受到严重影响。在控制过程中,关键参数的精准控制有利于提高工作效率,从而降低生产成本。针对时滞过程中流量调节的控制需求,以pH值过程控制为研究对象,对常规控制算法的不足进行了分析。在此基础上,提出了一种基于自适应增量学习的流量调节控制方法,并对流量调节经验值的学习过程进行了介绍。该学习算法通过模仿人的思维与决策过程,结合输入输出的传递关系,建立基于机器学习的面向复杂系统控制的系统架构。对时滞过程中流量调节的关键参数进行跟踪和优化,学习过程占用资源少,适用于不同的硬件平台。通过研究具有机器学习功能的多模态控制策略,为解决复杂问题的控制问题,提出一种新的思路和解决方案。试验结果表明,所提出的方法具有较好的预测精度,有利于实际生产应用。In complex system with nonlinear,time-delay and other characteristics,traditional control methods are difficult to apply,control quality is severely affected.Accurate control of key parameters in the control process is conducive to improving work efficiency,and thus reducing production costs.For the control demand of flow regulation in time-delay process,the shortcomings of conventional control algorithms were analyzed by taking pH value process control as the research object.On this basis,a flow regulation control method based on adaptive incremental learning was proposed,and the learning process of the empirical value of flow regulation was introduced.By imitating human thinking and decision-making process and combining the transfer relationship of input and output,the system architecture for complex system control based on machine learning was established.By tracking and optimizing the key parameters of flow regulation in the time-delay process,the learning process takes up less resource,and is suitable for different hardware platforms.By studying the multimodal control strategy with machine learning capability,a new idea and solution for the control of complex system was proposed.Experimental results showed that the proposed method had better prediction accuracy,and was conducive to practical production applications.

关 键 词：控制过程 关键参数 时滞系统 流量调节 pH值 自适应增量学习 机器学习 多模态 

分 类 号：TH-273[机械工程]