Real-time control of variable air volume system using a SPSA based neural controller

A neural control scheme for variable air volume (VAV) air-conditioning system is proposed. The neural network is trained online by the simultaneous perturbation stochastic approximation (SPSA) method instead of the standard back-propagation algorithm. The closed-loop stability of the proposed control scheme is guaranteed based on the conic sector theory. The new control scheme provides the desired functionality as well as the adaptation of the VAV control system for a wide range of disturbances and parameter changes. To demonstrate the applicability of the proposed method, real-time experiments were carried out on a pilot VAV air-conditioning system and good testing results are obtained.