This study is focused on computation optimization of the geometry for the flow channels in a micro-reformer used for methanol steam reforming. Three-dimensional mass and momentum transport phenomena with a pure fluid simulation in a micro-reformer are predicted using a commercial computational fluid dynamics code. Meanwhile, a simplified conjugate-gradient method is adopted to seek the optimal manifold shape and channel width of the micro-reformer iteratively using a Python interface. In the present study, the geometrical optimization tasks involve the designs of the inlet manifold and outlet manifold shapes as well as channel width distribution, and the design purpose is to obtain a uniform flow distribution throughout the entire micro-reformer so as to increase the hydrogen gas production rate. Cubic-spline interpolation is used in shape design to fit the points on the manifold shape more smoothly. The results show that the velocity standard deviation decreased from 0.14 to 0.048 and 0.051 after searching the optimal manifold shapes and channel widths, respectively. The manifold shapes of the inlet and outlet as well as the channel widths can efficiently lead to significant uniformity in the flow fields using a simplified conjugate-gradient method. (C) 2014 Elsevier Ltd. All rights reserved.