河南水利与南水北调2025，Vol.54Issue(4)：86-88,3.
基于3σ+LSTM耦合模型的地下水位预测研究
Research on the Prediction of Groundwater Level Based on 3σ+LSTM Coupling Model
摘要
Abstract
To address the problems of complex groundwater level data series and low prediction accuracy,a groundwater level prediction algorithm based on 3σ+LSTM coupling model was established.The 3σ method was used to correct and normalize outliers.A LSTM network structure with fully connected layers was constructed,and the prediction error was gradually reduced through iterative learning.Finally,the groundwater level of a certain city was predicted in practice,and compared with other groundwater level prediction models.The results show that the average absolute error of the groundwater level prediction results of the 3 σ+LSTM coupling model is 0.07,the mean square error is 0.008,and the root mean square error is 0.099 respectively,which are all lower than those of the BP neural network and ARIMA prediction models.The prediction accuracy of groundwater level data series is improved.关键词
地下水位预测/耦合模型/3σ修正/LSTM神经网络/误差分析Key words
prediction of groundwater level/coupling model/3σ correction/LSTM neural network/error analysis分类
地质学引用本文复制引用
张静..基于3σ+LSTM耦合模型的地下水位预测研究[J].河南水利与南水北调,2025,54(4):86-88,3.基金项目
2025年度河南省高等学校重点科研项目指导性计划(25B510010) （25B510010）
水利部黄河下游河道与河口治理重点实验室开放课题基金项目(2025007). （2025007）
