Landslide Groundwater Level Time Series Prediction Based on Phase Space Reconstruction and Wavelet Analysis-Support Vector Machine Optimized by PSO Algorithm
Huang Faming
Yin Kunlong
Zhang Guirong
Zhou Chunmei
Zhang Jun
Abstract：It is of great significance to predict the dynamic evolution process of landslide underground water level for landslide stability analysis.For the problem that the evolution process of groundwater level in reservoir landslide is a highly non-linear and non-stationary time series affected by many factors,to predict landslide groundwater level time series,a coupling model based on phase space reconstruction and wavelet analysis-support vector machine (WA-PSVM)optimized by particle swarm optimization is proposed.Firstly,the groundwater level time series was decomposed into several different frequency compo-nents to transform the non-stationary groundwater level time series into stationary time series.Secondly,the PSVM model was established for each component prediction based on the phase-space reconstruction.At last,the final prediction result was ob-tained by adding the predicted values of all frequency components.Taking daily average groundwater level time series of STK-1 hydrology hole on Sanzhouxi Landslide in the Three Gorges Reservoir Area for example,the influencing factors of landslide groundwater level fluctuation were analyzed and WA-PSVM model was used to predict the STK-1 groundwater level values. Meanwhile,the single PSVM model and wavelet analysis-back propagation neural network (WA-BP)model were also used for groundwater level prediction.The results show that reservoir water level fluctuation and rainfall are the main factors of ground-water level fluctuation in the reservoir landslide leading edge.We also find that the root-mean-square error (RMSE)of the pro-posed model for groundwater level time series prediction in STK-1 hydrology holes is 0.073 m,the goodness of fit is 0.966,re-spectively.The prediction accuracy of WA-PSVM model is higher than the single PSVM model and WA-BP model.What is more,WA-PSVM model solves the non-linear and non-stationary problem.WA-PSVM model also has a high operating effi-ciency and strong applicability without considering the impacts of reservoir water level fluctuation and seasonal rainfall.
Keywords：reservoir landslidegroundwater level time seriesphase-space reconstructionwavelet analysisparticle swarm optimizationsupport vector machinegroundwatergeological hazard
Publication Date：2015-01-01
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：12( 1254-1265 )