The paper proposes a new approach for the optimisation of utility networks. The approach makes use of the targeting and conceptual developments presented in Part I, and the available knowledge is exploited to accomplish a decomposition scheme that accommodates for both simple and complex units. The decomposition establishes a common modelling basis, and provides for variable efficiencies in the operation of the steam turbines. Its principles are exploited to set up a network superstructure that is comprised by a managable number of generic elements. The units are developed to take into account operational variations against which the optimisation objectives determine the component sets that need be used. As a result of the insights thrown into the formulation and the targeting models used, the optimisation effort is simply facilitated by a Mixed Integer Linear Programming formulation whose application is illustrated with several examples and an industrial case study.