Design and application of a data-driven PID controller

Abstract

Research on database utilization has received much attention in recent years. This paper presents a data-driven proportional-integral-derivative (DD-PID) controller design method that uses a large database. In this method, a large data-set comprising input/output (I/O) data and control parameters is stored in a database in advance, and current control parameters are tuned by using the database in an online manner. The DD-PID controller design method utilizes the fictitious reference iterative tuning (FRIT) approach to learn control parameters in the database in an offline manner by utilizing initial experimental I/O data. This method is called the DD-FRIT method. In the study, the DD-FRIT method was applied to a nonlinear tank system to evaluate its effectiveness. This paper is structured as follows. First, the DD-PID controller design method is presented. Next, the offline learning algorithm that combines the design method with FRIT is explained. Finally, the results of a simulation and experiment are presented to demonstrate the effectiveness of the proposed method. These results showed that control parameters in the database were learned appropriately by the proposed method, and good control results were obtained.