An automatic data acquisition system for on-line training of artificial neural network-based air handling unit modeling

Air-conditioning systems are highly non-linear. A conventional air-conditioning control cannot achieve thermal comfort environments all the times. Artificial neural networks have been proven to be successfully applicable in this area. However, they required on-line field data to re-tune themselves. In this paper, a fully automatic data acquisition system was developed for this purpose. The system consisted of a real air-handling unit, thermal sensors, data acquisition boards and a computer with an acquisition program. Substantial experiments were carried out to capture the operating characteristics of the real air-conditioning plant. With the help of the acquisition system, a neural network-based model was built which could accurately predict the behaviors of the plant. A practical optimal control is thus possible. The methodology in this paper, the proposed acquisition system with the neural network under on-line training, can be applied in other areas for better control performances.