基于自适应时序分解筛选的大坝变形预测模型

Dam deformation prediction model based on self-adaptive temporal decomposition screening

高精度的大坝变形分析和预测是掌握大坝工作性态、诊断大坝异常的重要手段。针对现有模型信息特征挖掘不充分、泛化能力弱、难以实现精准预测等问题,采用灰狼算法优化自适应噪声完备经验模态分解解决多维参数标定问题,使用阈值评价指标保留变形时序数据的有效信息特征;引入交叉验证的递归特征选择法通过多个学习器综合筛选出最优因子集,移除冗余特征、提取有效信息并增强模型可解释性;考虑时序数据特性优化双向长短期记忆神经网络时间窗步数,结合大坝变形数据降噪、最优特征因子输入等多种方法,构建大坝变形预测模型。以实际工程为例,结合多种预测模型进行对比分析,结果表明该模型具备挖掘非线性信息能力,预测性能得到改善,可为大坝安全监测提供参考。

High precision analysis and prediction of dam deformation is an important means to master dam working behavior and diagnose dam anomalies.Aiming at the problems such as insufficient information feature mining,weak generalization ability and difficulty in accurate prediction of existing models,grey Wolf algorithm was used to optimize the complete ensemble empirical mode decomposition with adaptive noise to solve the multidimensional parameter calibration problem,and threshold evaluation indexes were used to retain the effective information features of deformation time series data.The cross-validation method is combined with recursive feature selection method,and the optimal factor subset is selected by multiple learners to remove redundant features,extract effective information and enhance the interpretability of the model.Considering the characteristics of time series data,the number of steps in the time window of the bidirectional long short term memory neural network is optimized,and in order to construct dam deformation analysis and prediction model,several methods such as noise reduction of dam deformation data and input of optimal feature factors are used.The results show that the model has the ability of accurately mining nonlinear information,and the prediction performance has been significantly improved,which can provide reference for dam safety monitoring.