基于时间序列与GWO-ELM模型的滑坡位移预测

来源期刊:中南大学学报(自然科学版)2019年第3期

论文作者:吴益平 廖康 李麟玮 苗发盛 薛阳

文章页码:619 - 627

关键词:滑坡位移预测;时间序列;GWO-ELM模型;趋势性位移;周期性位移

Key words:landslide displacement prediction; time series; GWO-ELM model; trend displacement; periodic displacement

摘    要:针对三峡库区的阶跃型滑坡位移特征,以白水河滑坡为例,提出一种基于时间序列和灰狼优化的极限学习机(GWO-ELM)位移预测模型。首先,根据滑坡的内在演化规律和外部影响因素,建立滑坡位移的时间序列模型,将监测位移分解为趋势性位移和周期性位移,并运用稳健加权最小二乘法的三次多项式对趋势性位移进行拟合,以此得到周期性位移。其次,对位移监测数据进行分析,选取周期性位移的影响因子,分别通过GWO-ELM、极限学习机(ELM)和灰狼优化的支持向量机(GWO-SVM)模型对周期性位移进行预测。研究结果表明:GWO-ELM预测模型具有良好的泛化能力,能有效减少人为误差,在预测精度上,明显优于ELM和GWO-SVM模型。基于时间序列与GWO-ELM位移预测模型具有较高的预测精度和泛化能力,是一种有效的滑坡位移预测方法。

Abstract: Considering the landslide displacement characteristics of the Three Gorges Reservoir Area, a displacement prediction model based on time series and Extreme Learning Machine with Grey Wolves Optimization(GWO-ELM) was proposed to predict the Baishuihe Landslide. Firstly, based on the intrinsic evolution of landslides and external factors, a time series model of landslide prediction was established. The monitoring displacement was decomposed into trend displacement and periodic displacement, and the trend displacement was fitted by a cubic polynomial with a robust weighted least square method to obtain a periodic displacement. Secondly, the periodic displacement was predicted respectively by the GWO-ELM, the separate ELM and the GWO-SVM model through analyzing the influencing factors. The results show that the GWO-ELM prediction model has good generalization ability and it can reduce human error effectively. In terms of the prediction accuracy, GWO-ELM prediction model is apparently more precise than the ELM and GWO-SVM models. Based on the time series and the GWO-ELM model, the proposed model embodies a higher prediction accuracy and has generalization ability, so it is an effective landslide displacement prediction method.

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号