大坝的变形受到多种因素的影响，对于一些没有长期、连续、可靠监测数据的工程，采用传统的多变量灰色模型MGM(1,n)进行大坝变形预测时，往往随着预测时间的推移，预测精度降低。本文采用自适应MGM(1,n)模型，根据有限的监测资料，综合考虑各个变量之间的相互影响，通过置入最新信息取代最老的信息，来反映坝体变形过程中的随机因素或扰动对系统的影响。以此为基础，利用马尔科夫链确定位移时序的状态转移概率矩阵，通过对实测值、拟合值以及所处状态的分析，对大坝变形进行更高精度的预测。实例表明，和传统多变量灰色模型MGM(1,n)以及自适应MGM(1,n)模型相比，自适应MGM(1,n)-马尔科夫链模型（MGM-MC模型）具有更高的精度。

Dam deformation is influenced by a lot of factors. For those projects without long-term, continuous, and reliable monitoring data, the prediction precision of dam deformation based on the traditional MGM(1,n) model decreases with time. In this paper, the self-adaptive MGM(1,n) model was applied. The proposed model characterizes the interaction between each variable, and replaces the oldest information with new information, which can reflect the effects of random factors or perturbation on dam deformation. On the basis, the state transition probability matrix of the time series was determined by Markov chain, and the monitoring data and forecast data were analyzed to predict the dam deformation with a higher precision. Compared with the traditional MGM(1,n) model and self-adaptive MGM(1,n) model, the MGM-MC model has higher precision.