Internal Model Control Based On Recurrent Networks And Its Application In Thermal Plant

When the temperature of main steam is taken as the object of control, it will be a typical object of large time-delay process. Satisfactory control effects can hardly be obtained with conventional PID cascade control method that could not adapt itself to the character of dead-time varying and parameter–varying. A kind of Elman dynamic recurrent network model is introduced and its BP study algorithm is discussed in this paper. This type of network is used as internal model and inverse main controller in base of identification of main steam temperature object in thermal plants. This is proved having a dynamic performance superior to conventional PID cascade control. And it is featured by stronger robustness and self-adaptability to accommodate itself to variations of the object's parameters that could be seen in the simulation results in a wide range of load units.