基于变分模态分解和深度置信神经网络模型的滑坡位移预测

三峡库区共有滑坡1 000余处,频繁发生的滑坡灾害极大威胁着人民生命财产安全,因此开展合理有效的滑坡位移预测对减少财产损失和拯救人民的生命具有重要的研究意义。以三峡库区白家包滑坡为例,针对当前滑坡位移预测中常用分解方法的局限,在位移时间序列的分解中引入可以控制分解模态数目的变分模态分解方法,选取不同模态参数进行对比,以提高分解模型的精度和有效性;并基于滑坡触发因子建立深度置信网络模型对位移子序列进行预测,重构所有子序列预测结果得到总的位移预测值。总位移预测均值绝对误差3.657 mm,平均绝对百分比误差为0.010%,总体预测精度高,该方法误差小,具有良好的应用指导意义。

Landslide Displacement Prediction Based on Variational ModeDecomposition and Deep Confidence Neural Network Model

Frequent landslide disasters in the Three Gorges Reservoir area threaten the safety of people’s lives and property. Predicting landslide displacement rationally and effectively is of crucial significance for reducing property losses and protecting people’s lives. In view of the limits of conventional decomposition methods, the variational decomposition method which could control the number of decomposition modes was introduced into the decomposition of displacement time series. The Baijiabao landslide in the Three Gorges reservoir area was taken as a case study. The parameters are compared to improve the accuracy and effectiveness of the decomposition model. Moreover, a deep confidence network model involving landslide triggers was established to predict the displacement subsequences, and the results of all subsequences are reconstructed to obtain the total displacement prediction value. The mean average error of predicted total displacement is 3.657 mm, and the mean average percentage error is 0.010%, indicating a high accuracy.