Chinese steel industry faces increasing pressure to reduce water consumption and wastewater impact. To perform the optimization of steel production process from the perspective of water use, this work describes a new superstructure-based water network optimization model using typical steel park as our study use. The water systems at different scales (e.g. unit-scale, plant-scale and park-scale) in the water network and their interactions are modeled for optimization. To construct the optimization model: 1) All types of unit-scale typical water systems in the park are identified, and short-cut unit models are developed as the basic elements for network superstructure; 2) Intra- and inter-plant superstructures are further established to describe potential configurations of the water network at plant-scale and park scale respectively; 3) Built up on the element models and superstructures, a Mixed Integer Non-Linear Programming model using total annual cost as objective is set up to investigate potentials for water network optimization in a steel park. To illustrate its applicability and effectiveness, an industrial case study is carried out, in which two schemes are illustrated, one scheme (termed as scheme A) considering only indirect integration between plants, the other (termed as scheme B) considering both direct and indirect integration strategy (mixed integration). In scheme A, due to the higher water efficiency and optimal configuration in the case study, the freshwater consumption and total annual cost are reduced by 22% and 21% respectively. Scheme B shows 23% savings in total annual cost, while the reduction of freshwater consumption keeps unchanged. Besides, the information about the integration strategies of all scales (e.g. unit-, plant- and park-scale) can all be obtained, which is valuable for decision-making. (C) 2017 Elsevier Ltd. All rights reserved.