三峡大学学报（自然科学版）2026，Vol.48Issue(2)：8-15,8.DOI:10.13393/j.cnki.issn.1672-948X.2026.02.002
基于RWKV-TS的混凝土拱坝位移预测模型研究
Concrete Dam Deformation Prediction Model Research Based on RWKV-TS
摘要
Abstract
Accurate prediction of dam deformation is critical to ensuring the structural safety of dams.Dam deformation prediction is essentially a time series forecasting problem involving multiple influencing factors and nonlinear effects.Compared to traditional statistical models,deep learning exhibits a significant advantage in capturing nonlinear features,making it more suitable for dam deformation prediction.This study employs a time series forecasting model based on the RWKV(receptance weighted key value)model,termed the RWKV-TS model.This model integrates the strengths of both Recurrent Neural Networks and Transformers,effectively avoiding the gradient explosion issues associated with RNNs while mitigating the limitations of Transformer in memory consumption and secondary development.Through a case study on the displacement prediction of an arch dam,the RWKV-TS model demonstrates superior performance over traditional models in terms of mean absolute error(EMA),mean absolute percentage error(EMAP),and root mean square error(ERMS),thereby significantly enhancing the accuracy of dam deformation prediction with high efficiency.It offers substantial value for engineering practice and decision-making in management.关键词
大坝安全监控/变形预测/时间序列/深度学习/RWKV-TS模型Key words
dam safety monitoring/deformation prediction/time series/deep learning/RWKV-TS model分类
建筑与水利引用本文复制引用
黄之源,谷艳昌,陈波,罗诗怡,姜佩..基于RWKV-TS的混凝土拱坝位移预测模型研究[J].三峡大学学报（自然科学版）,2026,48(2):8-15,8.基金项目
国家重点研发计划项目(2024YFC3210703) （2024YFC3210703）
南京水科院基本科研业务费科研创新团队建设项目(Y722003) （Y722003）
国家自然科学基金项目(52309157) （52309157）
