ADAPTIVE INTELLIGENT PID CONTROL WITH PREDICTIVE PERFORMANCE

摘要

A new simple solution of deriving adaptive PID control based on artificial neural network (ANN) policy is presented. The design problem consists of training of both, a recurrent and a feedforward neural network model. The recurrent and the feedforward neural network are used as a predictor and adaptive neural feedback controller, respectively. They are trained using the conjugate gradient and stochastic approximation algorithms, respectively. The multi-layer feedforward ANN is trained so as to achieve the control objective. The controller, which consists of a multi-layer feedforward ANN and a discrete-time PID algorithm, operates in predictive mode. The network serves for adaptive tuning of the PID controller. To demonstrate the feasibility and the performance of this control scheme, the linear second-order system as well as a continuous-flow stirred biochemical reactor model have been chosen as simulation case studies. Simulation results demonstrate the usefulness and the robustness of the control system proposed.