Design of a Data-Driven PID Controller Using Operating Data

Abstract PID controllers have been widely used in a lot of process systems because a control structure is simple and physical meanings of PID parameters are clear. Among the methods for determining PID parameters, the VRFT (Virtual Reference Feedback Tuning) and the FRIT (Fictitious Reference Iterative Tuning) have gotten much attention as methods which determine control parameters directly by using a set of operating data in recent years. However, these methods target only linear systems, thus applicable scopes of these methods are narrow because most process systems have nonlinearity. On the other hand, a Data-Driven PID (DD-PID) controller has been proposed as an e ective method for nonlinear systems. In this method, PID parameters are recursively tuned based on a large database. Moreover, a database can store new operating data in an on-line manner, thus PID parameters are tuned flexibly even if a system has nonlinearity. However, the method needs a lot of experimental trials to obtain an initial set of PID parameters in a database, and it has been pointed out as a weak point of this method. In this research, an o -line updating method of a DD-PID controller by using the FRIT is proposed. According to this method, a DD-PID controller is able to update an initial set of PID parameters in a database by only using a set of operating data. Moreover, this method can expand applicable scopes of the FRIT to nonlinear systems. In this research, the e ectiveness of the proposed method is numerically evaluated by some simulations.