研究了遗传算法在系统辨识中的应用 ,根据系统辨识自身的特点 ,设计了自适应选取遗传算法适值函数的方法 ,该方法的计算量要比排序选择操作的计算量小得多 ,在种群进化中采用自适应交叉、变异概率 ,以此对一差分方程描述的系统进行参数在线辨识 .仿真结果说明了该策略能够有效地避免算法非成熟收敛 . The application of genetic algorithm in system identification is studied. According to the characteristics of system identification, a method to adaptively select the fitness function of genetic algorithm is designed. The computational cost of this method is much less than the computational cost of the sequence selection operation. In this paper, an adaptive crossover and mutation probability is adopted to identify the parameters of a system described by a difference equation online.The simulation results show that this strategy can effectively avoid the immature convergence of the algorithm.