Monday August 26, 2019 from 10:30 to 12:10
Identification of vapour compression air conditioning system behaviour using Bayesian regularization neural network
Sholahudin Sholahudin1, Keisuke OHNO1, Seiichi YAMAGUCHI1, Kiyoshi SAITO1.
1Waseda University, Tokyo, Japan
Identification for system dynamic behaviour is necessary to develop control strategy. In this paper, the dynamic performance of air conditioning (AC) system is predicted using artificial neural network (ANN) approach. The ANN is developed to predict exergy efficiency, coefficient of performance (COP), and cooling capacity. The controllable parameters including compressor speed and evaporator and condenser fan speed are considered as the input. The datasets for prediction are generated by AC system simulator. The system was simulated by randomly varying compressor speed and evaporator and condenser fan speed with N-sample signal input. The dynamic ANN configuration with Bayesian regularization is proposed to predict one-step and multi-step ahead of system performance behaviour. The results show that the developed ANN in present study yields good prediction accuracy for all outputs. Accordingly, ANN can be further applied for predictive control application in AC system to control cooling capacity while maintaining system efficiency.