Abstract: This paper proposes simulated annealing genetic algorithm (SAGA), considering the fact that the conventional standard genetic algorithm (SGA) suffers from shortcomings such as slow convergent speed and low stability when applied to when applied to beam pattern synthesis of an array.The improved algorithm makes some improvements in several factors of SGA, which include fitness function, crossover operator and mutation operator.In addition, annealing algorithm (SA) is also introduced into this algorithm.The convergent speed and stability are both effectively promoted separately, when SAGA is applied to beam pattern synthesis.The results of simulation show that the proposed method achieves a much lower sidelobe level (SLL) than conventional method.