Research And Application On PID Auto-tuning Control Method Based On Step Identification

PID controller is the most widely used controller in process industrialprocess control for its simple theory, good robustness. However，simplePID tuning algorithm cannot be accustomed to the complex system. At thesame time, manual PID tuning method should have strong experience andtime-consuming. Therefore in order to strength the automation andintelligence of system, more and more controllers use auto-tuningtechnology. This paper will research PID auto-tuning control method basedon step identification. The mainly researches are focused on stepidentification, controller design and PID auto-tuning software design.This paper researches the model identification methods based on stepresponse and tuning method of PID controller based on internal modelcontrol (IMC) for stable dead-time processes and unstable dead-timeprocesses. By combining the above identification method with controlalgorithms, the PID quantitative auto-tuning strategy is developed. At last,the PID quantitative auto-tuning strategy will be used in PID auto-tuningsoftware. The main contribution lies in:(1) For the second-order plus dead-time model, a simple identificationapproach is proposed based on step response. Firstly the relationshipbetween step response and the transfer function is researched. Then anidentification method based on the above relationship is proposed. Thismethod almost has ideal identification performance while the noise issmall. At the same time, for the environment under mounts of noise, aleast-squares method based on genetic algorithm is presented. This methodsolves the issue that the choice of initial point and end point of data is hardto chosen in the tradition least-squares method in the presence of mountsof noise. (2) For the first-order differential dead-time model which is typicallyused in stream turbine control system, a step identification based onleast-squares method is proposed. This method uses the simple step signaland identifies the entire model’s parameter in the single step test.(3) For the unstable first-order plus dead-time model, a stepidentification method based on relay feedback is proposed. Thisidentification has the advantage of both relay feedback and step response.For that this method is executed in closed loop will lead to capability ofdisturbance rejection and step identification will result in excellentidentification accuracy.(4) For the unstable first-order plus dead-time model, which is difficultto control, a PID controller tuning method is proposed based on internalmodel control (IMC). Meanwhile, the system adds a setpoint filter toeliminate the overshoot. It is a dominant merit that the controller is easy tobe turned. Besides, the unstable processes can be successfully stabilizedand effective controller without overshoot.(5) A new PID auto-tuning software is designed using the abovecontribution. This software can both identify and control stable dead-timesystem. This software identifies the model by step identification and tunesthe parameter of PID controller effectively. This package supports lots ofindustrial communication protocol and can be used in both embeddeddevice and large PLC or DCS system.