顾及诱发因素响应的灰色极限学习机滑坡位移预测

Prediction of landslide displacement based on GM(1,1)-ELM model by considering characteristics of inducing factors

基于滑坡变形演化特征与诱发因素分析,提出了一种顾及诱发因素响应的灰色极限学习机滑坡位移预测模型。该模型将滑坡位移量分解成具有确定性的趋势项和具有不确定性的随机项,采用动态GM(1,1)模型提取滑坡位移趋势项,引入降雨量、库水位等滑坡诱发因素,利用新型智能算法极限学习机ELM(extreme learning machine)对随机项进行逼近,最后将各分项位移叠加,实现滑坡位移的预测。选取新滩滑坡B3监测点89期数据及三峡库区某滑坡GPS2-2监测点29期数据进行预测研究,结果表明,该模型的预测结果与实测数据吻合度较高。

A new landslide displacement prediction model coupling GM(1,1)and extreme learning machine was proposed based on characteristics of landslide evolution and inducing factors analysis.This model firstly decomposes landslide deformation displacement into trend term and stochastic term and utilizes GM(1,l)model to extract trend term of deformation.Then,stochastic term of ELM neural network forecast displacement based on rainfall,reservoir water level and inducing factors is established.Finally,the accumulative displacement was obtained by adding up all items displacement.A total number of 89 monitoring data of B3 point of Xintan Landslide and a total number of 29 monitoring data of GPS2-2 point in the area of Three Gorges Reservoir are taken to test the predication performance of the proposed model.The example indicates that the proposed model has high prediction accuracy.