基于EMD-SVR的高挡墙护坡位移预测方法

Displacement Prediction Method of High Retaining Wall Slope Protection Based on EMD-SVR

针对传统基于单一模型的护坡位移预测方法预测精度较低的问题,提出了一种基于经验模态分解(EMD)和支持向量回归(SVR)的预测模型(EMD-SVR)。首先对护坡位移时间序列进行EMD分解,将原始复杂序列分解为一系列结构相对简单的本征模函数(IMF);然后分别对每个IMF建立SVR模型进行变形趋势预测,同时提出改进的粒子群算法对SVR核参数进行全局寻优,提升预测性能;最后综合叠加各IMF预测值获得最终护坡位移变化趋势预测结果。实验结果表明,该方法比SVR、BP神经网络等单一模型的预测精度更高,最大预测误差等4项指标表现更优。

Aiming at the low prediction accuracy of traditional slope protection displacement prediction methods based on a single model,we proposed a prediction model based on empirical mode decomposition(EMD)and support vector regression(SVR).Firstly,we performed EMD decomposition on the time series of slope protection displacement,decomposing the original complex sequence into a series of relatively simple intrinsic mode functions(IMF).Then,we established SVR model for each IMF to predict deformation trend,and proposed an improved particle swarm algorithm to globally optimize the SVR kernel parameters and improve prediction performance.Finally,we obtained the final prediction trend of slope protection displacement by comprehensively stacking various IMF prediction values.The test results show that the proposed method has higher prediction accuracy than single models such as SVR and BP neural network,and performs better in four indicators such as maximum prediction error.