Premium
Modeling and optimization of refinery hydrogen network – a new strategy to linearize power equation of new compressor
Author(s) -
Mahmoud Ahmed,
Adam Abdel Samed M.,
Sunarso Jaka,
Liu Shaomin
Publication year - 2017
Publication title -
asia‐pacific journal of chemical engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.348
H-Index - 35
eISSN - 1932-2143
pISSN - 1932-2135
DOI - 10.1002/apj.2131
Subject(s) - linearization , gas compressor , refinery , solver , mathematical optimization , power (physics) , nonlinear system , nonlinear programming , computer science , control theory (sociology) , mathematics , engineering , mechanical engineering , thermodynamics , physics , control (management) , quantum mechanics , artificial intelligence , waste management
Refinery hydrogen network problem is highly nonlinear due to the equation that describes the power of new compressor. Most of the previous attempts to linearize this equation have been made by assuming constant suction and discharge pressure while taking the inlet flow rate as a variable. Such assumption may not be practical in real condition because the calculated power requirement for new compressor may not be compatible with the pressure ratio of the selected compressor. This work proposed a new linearization method for the power of new compressor that provides additional degree of freedom by allowing the solver to choose the optimum new compressor(s) that satisfied the pressure requirement of process sinks. Using our proposed model, mixed‐integer nonlinear programming (MINLP) formulation can be converted into mixed‐integer linear programming. The applicability of our model was validated using two different refinery case studies. Mixed‐integer linear programming results obtained using our model require substantially lower computational cost than their MINLP counterparts where at least 60% savings in terms of iteration number and computational processing time were achieved. The approach demonstrated here can be potentially used to approach more complex refinery hydrogen network cases where the initial guess can be obtained from the linearized MINLP problem. © 2017 Curtin University and John Wiley & Sons, Ltd.