Multi Step Prediction of Landslide Displacement Time Series Based on Extended Kalman Filter and Back Propagation Trough Time

摘要

Landslide is a complex geological natural disaster that brings harm or damage to human beings and their living environment. By strengthening landslide monitoring and forecasting technology, people can avoid or reduce the impact of disasters more reasonably. At present, the single step prediction of landslide displacement time series mainly uses t time to predict the data of t+1 moment, which obviously makes it difficult for people to take appropriate measures to deal with landslide changes. In this paper, a time reverse recursive algorithm based on extended Kalman filter (EKF)and Back propagation trough time (BPTT) method, is used to predict landslide displacement in order to extend the time width of landslide prediction. The EKF is firstly used to optimize the BPTT weights, and then the network parameters are adjusted in real time to improve the reliability of the prediction. Finally, the landslide displacement data of Liangshuijing (LSJ) in the three Gorges Reservoir area is used as experimental samples to verify the feasibility and practicability of EKF-BPTT.