Exponential smoothing method based on wavelet transform for slope displacement prediction

Abstract

It has important significance in engineering to analyze rock slope's evolution rule and forecast its development trend based on the safety monitoring displacement data. The actual slope monitoring sequence is non-stationary time series containing a number of errors, therefore, firstly discrete stationary wavelet transform (DSWT) are used to denoising for monitoring data, then the reconstruction series are transformed into a stationary sequence by first-order difference, finally exponential smoothing method is used to prediction for the stationary differential sequence. The combination forecasting model is applied to high slope displacement prediction on the left bank of Jinping I Hydropower Station, the calculation results show that the combined model have higher forecast accuracy compared with other prediction methods, most of the relative errors of the prediction results are less than 5%, meeting engineering prediction requirements.