Abstract: In view of the disadvantages of simple genetic algorithm: lowconvergence rates, inferior ability in local optimization and so on, a decimal encoding scheme, improved arithmetic crossover and non- uniform mutation operation were adopted to improve IGA, then an IGA- BP model was analyzed and built. The experiment in which the IGA- BP model was applied to forecast damhorizontal displacement indicated the IGA- BP model was much better than traditional models in convergence speed and prediction precision.

摘要： 针对基本遗传算法( SGA) 收敛速度慢、局部寻优能力差等缺陷, 采用十进制编码, 引入改进的算术交叉、非均匀变异操作等算法, 分析和建立了改进的遗传神经网络( IGA- BP) 模型, 并将该模型应用于大坝水平位移的预测。结果表明, 该模型在收敛速度、预报精度等方面比传统模型有较大的改善。