Abstract：In this work multi-stream heat exchanger network (MSHEN) synthesis approach is applied to the crude oil pre-heating network reformation of atmospheric and vacuum distillation unit of a petrochemical company. Based on stage-wised superstructure of MSHEN, a more general and rigorous mathematical model eliminating the unreasonable hypothesis of isothermal mixing of stream splits by previous literatures is set up for MSHEN synthesis problem. A novel approach is put forward to convert the mathematical model formulated as a mixed integer nonlinear programming program (MINLP) to a nonlinear programming problem (NLP) through appropriately designing a valid solution generation method for MSHEN. The approach brings forth the advantage of shunning difficulty to solve the original MINLP problem. A genetic/simulated annealing hybrid algorithm is adopted to successively solve this problem. The comparison of the synthesis results obtained by the present mathematical model and optimization algorithm with those of commercial software HEXTRAN and the existing heat exchangers network indicates that the methodology proposed in this paper is quite effective, and MSHEN can be superior in industrial engineering as to economy and technical advantages.