基于EMD-LSTM的边坡位移预测模型应用  

Application of slope displacement prediction model based on EMD-LSTM

作　　者：白继元 肖航 武志高 BAI Jiyuan;XIAO Hang;WU Zhigao(CHN Eenergy Xinjiang Mining Hongshaquan Second Mine Co.,Ltd.,Changji 831100,China)

机构地区：[1]国能新疆矿业红沙泉二矿有限公司,新疆昌吉831100

出　　处：《露天采矿技术》2025年第2期36-40,共5页Opencast Mining Technology

摘　　要：为提高边坡位移预测的准确性,解决现有方法在准确率和滞后性等方面的问题;提出了基于经验模态趋势分解和长短期记忆神经网络的结合模型。预测模型将位移变化视为多个简单分量信号的叠加,通过经验模态分解将位移关系分解为多个周期项与趋势项,再利用长短期记忆神经网络分别预测这些分量,最终实现非线性关系的预测。露天矿排土场边坡实例应用表明:模型的预测准确性超过90%,显著优于传统的BP神经网络方法,能够满足工程实际需求。In order to improve the accuracy of slope displacement prediction and solve the problems of accuracy and lag in existing methods,a combined model based on empirical mode trend decomposition and long short-term memory neural network is proposed.The predicted model considers displacement changes as the superposition of multiple simple component signals,decomposes displacement relationships into multiple periodic and trend terms through empirical mode decomposition,and then uses long short-term memory neural networks to predict these components separately,ultimately achieving the prediction of nonlinear relationships.The application of example of the slope of an open-pit mine dump show that the prediction accuracy of the model exceeds 90%,which is significantly better than the traditional BP neural network method and can meet the practical needs of engineering.

关 键 词：位移预测 趋势分解 神经网络 非线性 结合模型 

分 类 号：TD824.7[矿业工程—煤矿开采]