This paper presents modelling and analyzing method of centrifugal chiller which has a rated capacity of 200 RT(703 kW) through on-site performance test. Field performance data of a chiller installed in a research building of KIST have been collected. Simple models were developed for predicting the heat exchanger and system performances by regression of chiller operation data during 5 days in August. The models proposed here account for the effect of variations of cooling capacity, temperatures and flew rates of secondary fluids. The proposed models can predict the actual performance data from June to September within $pm$ 5% error. The COP of centrifugal chiller are estimated under the standard rating conditions and reduced mass flow rates of chilled and cooling water.