Design of a GMDH-PID controller using a database

Abstract

PID control schemes have been widely used in most industrial control systems. However, it is difficult to determine a suitable set of PID parameters because most industrial systems have nonlinearity. In order to overcome such a problem, a data-driven PID (DD-PID) control scheme based on utilizing a database has been proposed and its effectiveness has been investigated. However, the DD-PID controller has two problems. One is that training of the database in an on-line manner takes a long time. The other is that a database requires large amount of memory and high computational cost for some micro-controllers. In order to train a database in an off-line manner, the DD-FRIT scheme which is a combination of a database and the fictitious reference iterative tuning (FRIT) scheme has been proposed in a previous research. According to the DD-FRIT scheme, a DD-PID controller can be trained in an off-line manner by using a set of operating data. In this paper, to address the problem of required memory and computational cost, a method that expresses a DD-PID controller as a simple nonlinear function by using the group method of data handling (GMDH) is proposed. According to the proposed method, a DD-PID controller which is trained in advance by using a set of operating data is replaced by a network constructed in N-Adalines (units expressed by a simple nonlinear function). The proposed method is first explained and the effectiveness of the proposed method is numerically evaluated by a simulation example.