对大坝变形进行预测时，观测序列中的粗差不可避免地影响预测精度，同时，监测序列较短也会影响预测的精度，针对这些情况，将未确知滤波法和灰色模型应用到大坝变形预测中。首先利用未确知滤波法对大坝变形监测序列进行粗差识别，并修正粗差，再采用灰色模型预测大坝的变形。实例分析表明，通过 2 种方法有机地结合，可以显著地提高大坝变形预测的精度，具有较高的工程实用价值。

Unascertained filtering and grey model are applied to predict dam deformation as the prediction accuracy is inevitably influenced by the gross error of observation sequence on the one hand and the short observation sequence itself on the other. The gross error of dam deformation observation sequence is detected and corrected by unascertained filtering method, and subsequently the grey model is employed to predict the dam deformation. The integration of the two processes is proved by case study to be of high practical value with remarkably improved prediction accuracy.