Optimal Design of Refinery Hydrogen Networks: Reduced Superstructure and Effective Algorithm

Abstract

In refinery hydrogen networks, the cost related to hydrogen compressors is the second largest contributor to the total annualized cost (TAC). The optimal synthesis of hydrogen networks considering compressor arrangement is a computationally difficult optimization problem because of the highly nonlinear terms caused by hydrogen compression. In this study, a reduced superstructure for optimal design of hydrogen networks is first developed based on several heuristics. This reduced superstructure includes all possible connections among hydrogen sources, hydrogen sinks, and compressors. The maximum number of compressors is reduced dramatically, and the inlet and outlet pressures of all compressors can be preassigned, which notably reduces the bilinear terms and completely eliminates the linear fractional terms in the mathematical model. A mixed-integer nonlinear programming (MINLP) model is formulated, and a tailored algorithm that incorporates successive piecewise linear approximation of concave terms and linear relaxation of bilinear terms is proposed. The case studies show that the proposed model and algorithm can give better hydrogen network designs with only 2.9 and 4.9% of the computational time in the existing publication.