TY - JOUR
T1 - Multi-step ahead prediction of vapor compression air conditioning system behaviour using neural networks
AU - Sholahudin, S.
AU - Ohno, K.
AU - Yamaguchi, S.
AU - Saito, K.
N1 - Publisher Copyright:
© Published under licence by IOP Publishing Ltd.
PY - 2019
Y1 - 2019
N2 - Cooling capacity and super heat temperature control for air conditioning (AC) system operation is necessary to ensure that the system operates efficiently. In this paper, multi-step-ahead prediction of AC system behaviour is presented using backpropagation neural network model as the first effort to develop the effective control strategy. Several step-ahead cooling capacity and superheat temperature performance are predicted under modulation of compressor speed and expansion valve opening. The prediction is proposed to capture the dynamic behaviour of system that can be applied in predictive control purpose. The configuration of ANN model is developed based on nonlinear autoregressive network with exogenous input (NARX) structure. Input and output data for training and validation of ANN model are generated by AC simulator. The ANN model is optimized by investigating the effect of number of neuron and time delay input on prediction accuracy. The results show that the ANN model developed in present study has good accuracy in predicting several step-ahead of cooling capacity and superheat temperature. Accordingly, this ANN model is applicable for predictive control in future study.
AB - Cooling capacity and super heat temperature control for air conditioning (AC) system operation is necessary to ensure that the system operates efficiently. In this paper, multi-step-ahead prediction of AC system behaviour is presented using backpropagation neural network model as the first effort to develop the effective control strategy. Several step-ahead cooling capacity and superheat temperature performance are predicted under modulation of compressor speed and expansion valve opening. The prediction is proposed to capture the dynamic behaviour of system that can be applied in predictive control purpose. The configuration of ANN model is developed based on nonlinear autoregressive network with exogenous input (NARX) structure. Input and output data for training and validation of ANN model are generated by AC simulator. The ANN model is optimized by investigating the effect of number of neuron and time delay input on prediction accuracy. The results show that the ANN model developed in present study has good accuracy in predicting several step-ahead of cooling capacity and superheat temperature. Accordingly, this ANN model is applicable for predictive control in future study.
UR - http://www.scopus.com/inward/record.url?scp=85068994445&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85068994445&partnerID=8YFLogxK
U2 - 10.1088/1757-899X/539/1/012003
DO - 10.1088/1757-899X/539/1/012003
M3 - Conference article
AN - SCOPUS:85068994445
SN - 1757-8981
VL - 539
JO - IOP Conference Series: Materials Science and Engineering
JF - IOP Conference Series: Materials Science and Engineering
IS - 1
M1 - 012003
T2 - 1st International Conference on Design, Energy, Materials and Manufacture, ICDEMM 2018
Y2 - 24 October 2018 through 25 October 2018
ER -