为实现大坝安全性态的准确预报,以西南某坝的多年扬压力极值为例,以扬压力实测序列建立GM(1,1)预测模型,基于预测模型的残差序列建立马尔科夫残差模型,对比分析GM(1,1)残差预测模型和马尔科夫残差模型;综合GM(1,1)模型和马尔科夫残差模型建立灰色马尔科夫模型,并利用灰色马尔科夫模型预测扬压力极大值。结果表明,灰色马尔科夫模型提高了预报的精度并恰当地反映了过程的摆动性。 In order to realize accurate prediction of dam safety status, taking the extreme value of multi-year thrust pressure of a dam in southwest China as an example, a GM (1,1) forecasting model is established based on actual measured pressure of uplift and Markov is established based on the residual sequence of prediction model. The residual GMM (1,1) model and the Markov residual model were comparatively analyzed. The gray Markov model was established by combining GM (1,1) model and Markov residual model. The Cove model predicts the maximum pressure. The results show that the gray Markov model improves the accuracy of forecasting and properly reflects the volatility of the process.