[en] This paper presents a model based supervisory control strategy for online control of building central cooling water systems to enhance their energy efficiency. The supervisory control strategy seeks the minimum energy input to provide adequate cooling energy for buildings, taking into account the characteristics and interactions of central cooling water systems as well as the requirements and constraints of practical application. Simplified semi-physical chiller and cooling tower models are used to predict the system energy performance and environment quality as well as the system response to changes of control settings. A hybrid optimization technique, namely the PMES (performance map and exhaustive search) based method, is developed and utilized to seek optimal solutions to the optimization problem. The control performance and energy performance of this model based supervisory control strategy are evaluated on the central cooling water system of a high rise commercial office building by comparing with that of the model based supervisory control strategy using a genetic algorithm (GA) as the optimization tool, and the performance map based near optimal control strategy as well as other conventional control strategies for cooling water systems in terms of energy efficiency, control accuracy, computational cost etc. The results showed that this strategy is more energy efficient and computational cost effective than other methods for online practical application