Abstract:
An inferential control (IC) strategy based on improved hybrid model combining prior knowledge with artifitial neural networks is proposed.In this paper,we presented the design approach to the inferential estimator,the PID algorithm of program variable gain and the IC system for nonadiabatic fixed led reactor,Simulation results show that the estimator possesses good steady and dynamic features,high estimation accuracy,good swiftness and generalization performance.Furthermore,it needs less training samples and is easy to be understood and interpreted.And the IC system is excellent in both setpoint traclking and disturbance resisting.