变形监测数据组合预测的串联与并联模式研究

摘要

In order to improve the accuracy and reliability of forecasting deformation rnonitoring data,two kinds of combination prediction model structure,series combination (SC)and parallel combination (PC) are proposed.In the SC time series model,grey model and neural network to predict are separately used and then the predicting results are combined.In the PC the weight is used to combine different models,and the performance of the linear and non-linear weight combination forecast are compared.Finally,a practical prediction case in building safety monitoring forecast used to explain the method.According to the experimental results,the proposed SC and PC combined method obviously can improve the prediction accuracy and therefore can be applied to deformation data analysis.%为提高变形监测数据预测的精度与可靠性,提出了变形组合预测的串联与并联两种模式:即从变形时间序列分解的角度出发,构建了串联式组合模型；在线性定权和非线性定权约束条件下,建立了并联式组合预测模型.将两种模式的组合模型应用于某高层建筑的沉降变形预测,并对其预测效果进行比较.结果表明,组合预测精度优于单一预测模型,且提高了预测结果的可靠性.