GM(1,1)模型的改进及其在变形预测中的应用

为了提高GM(1,1)模型预测精度,采用积分优化、二次拟合优化以及残差改化方法,分步对GM(1,1)模型进行改进,建立灰色多重修正模型。具体改进步骤为:首先,利用积分优化方法对背景值进行纠正,减小模型误差并提高预测精度; 接着,对模型参数(发展系数和灰作用量)进行二次拟合优化,使参数更加接近理论真值; 然后,根据预测结果进行适当的残差改化,提高模型整体的预测精度; 最后,建立根据GM(1,1)模型改进的灰色多重修正模型。以重庆南川地区甄子岩崩塌为例,建立灰色多重修正模型对危岩裂缝累计位移值进行模拟和预测,并与GM(1,1)模型进行对比。精度检验结果表明:灰色多重修正模型后验差比值(0.082 39)明显好于GM(1,1)模型(0.192 67),平均相对残差比(0.073 9)更远好于GM(1,1)模型(0.259 6),表明灰色多重修正模型在预测精度上有较大提高,可靠性更好。

Improvement of GM(1,1)Model and Its Application on Deformation Prediction

In order to improve the precision of prediction, GM(1,1)Model was modified by the means of integral optimization, quadratic fitting optimization and residual error modification methods step by step, and grey multiple correction model(GMCM)was built. The steps of modification included that background value was corrected by the means of integral optimization in order to reduce the error and improve the precision of prediction firstly; secondly, the parameters including developing coefficient and grey action were optimized by the means of quadratic fitting in order to make the parameters closer to the theoretical value; thirdly, residual error modification was properly done according to the prediction results in order to improve the whole precision of prediction; finally, GMCM, which was modified based on GM(1,1)Model, was built. Collapse of Zengzi Rock in Nanchuan of Chongqing was taken as an example, GMCM was built to simulate and predict the accumulated displacement of crag crack and was compared with GM(1,1)Model. The results of accuracy test showed that posterior error ratio of GMCM(0.082 39)was better than that of GM(1,1)Model(0.192 67), and average relative residual error ratio of GMCM(0.073 9)was better than that of GM(1,1)Model(0.259 6), so that the precision of prediction and reliability of GMCM were improved significantly.