Title :
Adaptive PID control with neural network based predictor
Author :
Tan, Yonghong ; De Keyser, Robin
Author_Institution :
Ghent Univ., Belgium
Abstract :
In this paper, a novel approach to adaptive PID control with neural network based d-step ahead predictor is presented to deal with the control problem for a nonlinear process with time-delay. The paper is organized as follows:In section 2, both recursive and non-recursive d-step ahead predictors are presented to compensate for the influence of the time-delay on nonlinear control systems. In section 3, an adaptive PID control strategy with neural network predictor is proposed. Then, a RLS type learning algorithm is described in section 4 to speed up the learning of the feedforward neural network, thus allowing for on-line adaptive control strategies to become feasible. Finally, simulation results as well as the real-time control of a heat exchanger are illustrated.
Keywords :
adaptive control; delays; feedforward neural nets; filtering and prediction theory; heat exchangers; learning (artificial intelligence); nonlinear control systems; predictive control; three-term control; RLS type learning algorithm; adaptive PID control; feedforward neural network; heat exchanger; neural network based d-step ahead predictor; nonlinear control systems; nonlinear process; online adaptive control strategies; real-time control; time-delay;
Conference_Titel :
Control, 1994. Control '94. International Conference on
Conference_Location :
Coventry, UK
Print_ISBN :
0-85296-610-5
DOI :
10.1049/cp:19940357