Adaptive neural network tuner of pid-controller for heating furnaces control

The existing neural network tuner of PI-controller is improved in such a way that to be able to adjust parameters of a PID-controller. The new structure of a neural network for the tuner is defined for this purpose, its rule base is updated, and the stability criterion is proposed for the system with the tuner and the PID-controller. The new version of the tuner is applied to control a typical heating furnace during numerical and full-scale experiments in order to maintain the required quality of transients under the condition of the furnace parameters non-stationarity. It made it possible to reduce the furnace power consumption by 8,4% through the full-scale experiments in comparison to common PID-controller.