采用遗传算法求解大型空间结构的模型降阶问题。给出了模型降阶的数学表示和所用遗传算法的有效形式，即编码方式、适值方式、交叉方式和变异方式等。仿真实例说明，遗传算法优于标准内平衡法和Ｈｏｐｆｉｅｌｄ神经网络方法 Using GA to Solve Model Reduction Problem of Large Space Structure. The mathematical representation of the model reduction and the effective forms of the genetic algorithms used are given, ie, coding, fitness, crossover and mutation. Simulation examples show that genetic algorithm is superior to the standard internal balance method and Hopfield neural network method