This paper presents a systematic approach for the synthesis of the heat recovery network in total site using a Mixed Integer Linear Programming model. This model returns a near-to-optimal network configuration with minimum utility cost while allows to select geographically closest matches. The Heat Load Distribution is the subproblem of the network design and has been reported to be quite expensive to solve for large-scale problems. The computational complexity of HLD resides in the number of streams and the feasible networks. An additional challenge, raising particularly in industrial problems, has been the intermediate heat transfer network which aggravates the combinatorial complexity. The presented methodology deals with those difficulties by priority consideration based on the location of process units. It helps significantly reducing the computational time and also comes with a realistic network sketch with respect to the plant layout. Several examples are discussed along with a real industrial case study. (C) 2014 Elsevier Ltd. All rights reserved.