【摘要】 针对可编程逻辑控制器(PLC)和虚拟PLC的PID难以优化整定的难题,将建模、控制、优化和深度学习与强化学习相结合,提出无模型PID在线自优化整定算法.将工业云及边缘计算、软件定义实时及可靠保障机制的双通道通信架构与所提出的PID整定算法相结合,提出云端协同的软件定义智能控制系统.云为基于云服务器的智能控制软件开发平台;端为基于工业服务器的智能控制软件.智能控制软件包括虚拟PLC PID、PID预优化整定和控制过程数字孪生以及在线自优化整定、自适应切换机制.采用研制的软件定义智能控制系统研究实验平台,进行所提出的控制系统与国外先进PLC和工业PC的无模型整定软件PID控制系统的仿真与物理对比实验.实验结果表明本文的软件定义智能控制系统可进行控制器参数自优化整定,控制性能显著优于国外无模型整定软件的PID控制系统.更多还原

【Abstract】 To address the challenge of achieving optimal PID tuning in both physical programmable logic controller(PLC) and virtual PLC, a model-free PID online self-optimizing tuning algorithm is proposed by integrating modeling, control, optimization with deep learning and reinforcement learning. A cloud-edge collaborative software-defined intelligent control system is developed by combining the industrial cloud and edge computing as well as the proposed PID tuning algorithm and a software-defined dual-channel communication architecture based on real-time and reliability assurance mechanisms. In this system, the cloud serves as an intelligent control software development platform based on cloud servers, while the edge comprises intelligent control software deployed on industrial servers. The intelligent control software includes virtual PLC PID, pre-optimization PID tuning, digital twin of the control process, online self-optimizing tuning and an adaptive switching mechanism. Simulation and physical comparative experiments on the developed software-defined intelligent control system research experimental platform are conducted between the proposed control system and model-free PID tuning control systems on advanced foreign PLCs and industrial PCs. The experimental results indicate that the proposed software-defined intelligent control system is capable of self-optimizing controller parameter, and its control performance significantly outperforms that of advanced foreign model-free tuning PID control systems.更多还原