Real-time neural optimal controller for a direct expansion (DX) air conditioning (A/C) system

A discrete-time neural inverse optimal control scheme for the simultaneous control of indoor air temperature and humidity of a DX A/C system is reported in this paper. The plant model is identified using a recurrent high order neural network (RHONN), and a discrete-time inverse optimal control law is derived with this model. Kalman filtering is used to perform on-line the neural network learning. This novel proposed control scheme is tested via implementation in real time. The obtained results for trajectory tracking illustrate the effectiveness of the proposed approach.