大坝变形是一个复杂的动态过程,传统的单变量预测模型未能考虑各监测点之间的相关性,不足以体现出大坝整体变形规律。因此采用多变量MGM(1,m)模型预测,根据MGM(1,m)的建模原理以及误差来源,重新构建了MGM(1,m)的背景值,并在此基础上对残差修正进行综合优化。实验结果表明,基于综合优化的MGM(1,m)模型拟合预测精度最高。 Dam deformation is a complex dynamic process. The traditional univariate prediction model fails to consider the correlation between monitoring points and does not reflect the overall deformation law of the dam. Therefore, the multivariable MGM (1, m) model is used to predict the background value of MGM (1, m) based on the modeling principle of MGM (1, m) and the source of error. Based on this, For comprehensive optimization. Experimental results show that the prediction accuracy of the MGM (1, m) model based on synthetic optimization is the highest.