将灰色系统（GM(1,1)）、BP神经网络、灰色神经网络（GNNM(1,1)）3种智能预测模型分别应用于深大基坑锚碇基础的基底变形预测过程中，以润扬大桥北锚碇基础基底土压力的监测资料为例进行动态预测分析，并与实测值进行了比较。结果表明：3种模型土压力预测值的相对误差分别为1.11%，0.77%和0.43%。GNNM（1,1)模型的预测结果更接近于实测值，与GM(1,1)和BP神经网络相比，GNNM(1,1)更适宜对波动较大的线性数据和非线性数据进行拟合，可以在工程中推广应用。

Intelligent models including Grey Model (GM(1,1)), BP neural network, and the combination of the two models Grey Neural Network Model (GNNM(1,1)) were employed in the prediction of anchorage foundation deformation. Monitored soil pressure of the north anchorage foundation of Runyang Bridge was taken to dynamically predict the deformation by these three models. The predictions were further compared with the measured soil pressures. The comparison showed that there is a relative error of 1.11% ，0.77%and 0.43% respectively of each model’s prediction result. Compared with the other two models, the prediction of GNNM(1,1) was closer to the measured soil pressure, and it can be applied in actual prediction process as it is more appropriate for curve fitting nonlinear data and largefluctuation data.