Supporting structure of deep excavation is optimized using genetic algorithm(GA); and limitation of GA such as poor local search ability and premature is largely enhanced. Then crossover operator and mutation operator is improved while considering the adaptive method; also the combination of hill-climbing with GA enables better local search ability in post-evolution. Mathematic model of deep excavation supporting structure is discussed; according to which a GA software system is developed for deep excavation retaining system. An engineering example is studied; and the results show the validity of the amelioration.