基于变分模态分解和灰狼优化极限学习机的隧道口边坡位移预测

Displacement prediction of tunnel entrance slope based on variational modal decomposition and grey wolf optimized extreme learning machine

针对高铁隧道口边坡位移监测数据非平稳、非线性的特点,以及极限学习机(ELM)模型起始参数随机生成导致预测性能不佳等问题,建立了基于变分模态分解(VMD)和灰狼优化算法(GWO)的ELM位移预测模型VMD-GWO-ELM。首先,通过经验模态分解的自适应分解层数确定VMD的最佳分解数k,得到周期项、趋势项和波动项位移。然后,利用灰狼算法优化ELM的输入权值和隐含神经元阈值。最后,对各子序列进行预测和叠加。实例验证结果表明:本文模型的均方根误差为0.3822 mm,平均绝对百分比误差为1.0047%,拟合优度为0.9837,表明该模型具有更高的预测精度及适用性。

In view of the non-stationary and nonlinear characteristics of the slope displacement monitoring data of high-speed railway tunnel entrance and the poor prediction performance caused by random generation of initial parameters of extreme learning machine(ELM)model,an ELM displacement prediction model based on variational mode decomposition(VMD)and grey wolf optimizer(GWO)was established.The optimal decomposition number of VMD was determined by the adaptive decomposition layers of Empirical Mode Decomposition,and the displacement of periodic term,trend term and wave term were obtained by VMD.The GWO was used to search for the optimal weight matrix connecting the input and hidden layers and the threshold of the hidden layer neurons of ELM.Each subsequence was predicted and the cumulative displacement was obtained by combining the results.Example verification shows that the root-mean-square error,mean absolute percentage error and goodness of fit of VMD-GWO-ELM model are 0.3822 mm,1.0047%and 0.9837,respectively.The VMD-GWO-ELM model has higher prediction accuracy and applicability.