基于多尺度小波分析的大坝变形自回归预测模型

针对大坝变形监测数据序列一般具有较明显的多尺度特征、年周期性和非平稳趋势性等特点,为克服自回归模型在大坝非平稳性变形预测方面存在的不足,利用多尺度小波分析理论在处理非平稳信号上的优势对变形监测数据序列进行分解和重构,然后对重构的不同尺度下的数据子序列分别建立自回归预测模型(AR模型),最后叠加各尺度下的预测结果,从而将多尺度小波分析与自回归模型有机地结合起来,为大坝变形预测提供了一种新的预测模型,并给出了一个工程实例.研究表明:与传统的单一自回归预测模型相比,该大坝变形预测模型能较大地提高预测精度.

Forecast model of dam deformation based on multi-scale wavelet analysis and autoregressive method

According to characteristics of the deformation monitoring data sequence of general obvious multi-scale features,annual periodicity and non-stationary trend,in order to overcome the deficiency of autoregressive model in the non-stationary dam deformation forecast,this paper uses the advantage of multiscale wavelet analysis theory in dealing with the non-stationary signal to decompose and reconstruct the deformation monitoring data sequences.Then different autoregressive models are built to forecast the reconstruction data sequence mining of different scales;the forecasting results of each scale are accumulated finally.Thus multi-scale wavelet analysis and autoregressive model are combined organically so as to offer a new model for dam deformation forecast.Finally,an engineering example is given.Research shows that: compared with the traditional single autoregressive forecasting model,the proposed dam deformation forecasting model in this paper can be used to improve the forecasting accuracy considerably.