Passive cooling of electronic components by natural convection heat transfer is the least expensive, quietest and most reliable method of heat rejection. However, some characteristics intrinsic to the phenomenon, such as a non-linear interaction between heat sources difficult its practical application. The objective of this work is to study the natural convection heat transfer optimization through genetic algorithms. Four cases were studied: one heat source, two heat sources with same dissipation rate, and two heat sources with different dissipation rates on a vertical wall. The results showed good agreement with literature and confirmed the methodology as computationally feasible for the optimal location of heat sources in a vertical wall. (c) 2006 Elsevier Ltd. All rights reserved.