Multi-objective optimization is performed for methane steam reforming in a porous ceramic membrane reactor using nitrogen and steam as sweep gases. The non-dominated sorting genetic algorithm (NSGA) is applied in solving the optimization problems. The Pareto optimal solutions have been obtained for the simultaneous maximization of the hydrogen production rate and the recovery yield, and for the simultaneous maximization of the hydrogen production rate and minimization of the sweep gas flow rate or the membrane area. The comparisons of the membrane reactor performances for nitrogen and steam as sweep gases at optimal conditions illustrate that the use of steam as a sweep gas can produce more hydrogen at higher recovery yield, or produce more hydrogen using less sweep gas or membrane area. (c) 2007 Elsevier Ltd. All rights reserved.