This paper first presents five empirically based models for steady-state modeling of cen-trifugal chillers, i.e. Simple linear regression model (SL), Bi-quadratic regression model (BQ), Multivariate polynomial regression model (MP), Gordon-Ng model and Quadratic homogeneous polynomial regression model (QHP). They are applied to predict the coefficient of performance using 4523 chiller data sets from chiller manufacturers. The data sets comprise four broad classifications, including (1) constant speed, con-stant condenser and chilled water flow, (2) variable speed, constant condenser and chilled water flow, and (3) variable speed, constant condenser and variable chilled water flow, and (4) variable speed, variable con-denser and chilled water flow. The regression parameters for each performance model are obtained using least squares method. The comparison demonstrates that QHP model shows the best prediction accuracy for all kinds of data sets (>94%) and MP model has less regression parameters while shows less prediction ac-curacy than QHP model (±2%).
centrifugal chillersempirically based modelsleast square methodregression
NETL
NSTL
万方数据
维普
