Industrial utility systems, traditionally the "poor forgotten cousin" of the processing industry, are now receiving the attention they deserve. While the thermodynamics and process models of utility systems are well established, the optimization of utility systems typically involves solving large-scale optimization problems including integer constraints, and hence are less commonly tackled. The integer formulation arises naturally in circumstances for example where process flows drop to zero, or processing equipment is brought into, and out of service. Both these cases are far more common when modelling utility systems as opposed to the now routine modelling of the parent chemical processes. This paper investigates various alternative modelling and subsequent optimisation strategies for a large industrial steam utility system. The results show that by applying suitable models, good modelling choices and appropriate software, one can readily solve large problems to obtain significant economic savings in solution times that are even amenable for design and/or real time optimization.