H-beam is a new kind of cost-effective steel beam and is getting more extensive application. Optimizing the hot-rolling process of H-beam is of great significance to improve the mechanical properties of H-beam and to reduce the rolling energy consumption. Based on the study of the application for the ANN (Artificial Neural Network) technology in rough rolling process of H-beam, the mathematical model between the parameters in hot-rolling process and the rolling power as well as the mathematical model between the parameters and the austenite grain size were established, then the genetic algorithm was applied to optimize the rolling process. The results indicate that the capability of H-beam is improved and the energy consumption is reduced after optimizing the hot-rolling process. The research will be of vital significance to ameliorating the hot-rolling process of H-beam.