A new approach for the optimisation of steam levels of total site utility systems satisfying varying utility demands is presented, accounting for interactions between total site utility systems and chemical processes. The optimisation problem involves the selection of the steam levels with respect to their temperatures. In this paper, by exploiting total site analysis techniques, a new transhipment network is developed to represent the heat flows of a total site. Base on the transhipment network representation of the total site, a general multi-period mixed-integer linear programming (MILP) model is presented for identifying the optimal steam levels of the total site utility system. By using engineering and thermodynamic knowledge, a boiler hardware model (BHM) is developed to describe the performance of boilers, and the turbine hardware model (THM) is applied for the shaft-work targeting of steam turbines. Both models are capable of predicting the real efficiency trends of the units. The application of the proposed optimisation approach is illustrated through two case studies including single operation scenario and multiple operation scenarios. (C) 2004 Elsevier Ltd. All rights reserved.