Displacement Prediction of Landslide Based on GA-ELM and Optimization of Inducing Factors

摘要

The surface of the landslide in the Three Gorges Reservoir area has a "step-like" feature. Landslide displacement prediction method based on displacement response component model is one of the main methods for the prediction of landslide displacement. In order to solve problem about displacement prediction of Landslide fluctuation term in reservoirs, the reorganization and optimization of the main priming factors have not been considered yet. And a method for predicting landslide displacement HP-CEEMDAN-GA-ELM based on the reorganization and optimization of time-series CEEMDAN inducing factors has been proposed. Taking the surface displacement data of Baishuihe landslide from January 2008 to December 2012 as an example, the surface displacement time series is decomposed into the trend term displacement and the fluctuation term displacement by using HP filter. This trend item is predicted by GA-ELM. It decomposes predisposing factors through CEEMDAN, and uses gray correlation analysis to determine the optimal recombination factors for inducing factors. Based on Inducer component of reorganization, a GA-ELM model is established to predict the fluctuation term. Compared with multiple prediction models, the experimental results show that the prediction error of the model is small, which further prove the validity of the method.