Dynamic load sharing property optimization of a 2K-H planetary gear reducer is carried out based on genetic algorithm and the bending torsional coupling model of the planetary gear train. Some design parameters which have significant influence on the load sharing property of the reducer are chosen by using the method of numerical calculation about the dynamic response. According to principle that parameter is easy to modify in actually engineering project, clearance of the external and internal gear pairs are selected as design variables from the above chosen parameters. According to the clearance region of a gear pair in actually engineering project, the scope of design parameters is fixed. A global optimum clearance of the external and internal gear pairs are gotten by using genetic algorithm, and numerical test proved that the load sharing property of the planetary gear reducer improve significantly after optimization.