Aimed at the complexity and time variability of forecast of dam deformation,and the shortage of traditional prediction model,combined with the overall ability of random search of genetic algorithm and the charactertics of misalignment mapping,dynamic feedback and memory function of Elman neural network, the paper built the model of genetic algorithms(GA) and Elman neural network. Compared with the Elman neural network ,the GA-Elman model has the characteristics of global convergence and can overcome the fault that Elman neural network was susceptible to fail into local minimum. The model was used to forecast some measured data of a dam deformation in a hydropower station.The result showed that the forecast precision of GA-Elman model is high and has practicability in dam deformation prediction.