摘要
大坝位移数据受各种因素的影响,具有非平稳非线性的特征,针对数据预测精度低的问题,提出了一种CEEMDAN-PCCs-TCN-XGBoost组合预测模型.以某重力坝监测数据为例,首先,引入CEEMDAN算法捕捉非平稳数据的趋势信息和波动信息,联合PCCs算法确定影响数据波动的主要因素;其次,为了提高预测精度,趋势信息使用传统HST模型进行预测,波动信息利用主要因素作为输入变量进行预测,再分别应用TCN模型和XGBoost模型对数据进行预测最后累加得到最终预测结果,并将预测结果与EEMD-ARIMA、EEMD-LSTM-MLR等模型对比,结果表明,CEEMDAN-PCCs-TCN-XGBoost组合预测模型对波动频繁的大坝位移数据的预测更精确.
The dam displacement data is influenced by various factors and has non-stationary nonlinear characteristics.To address the problem of low data prediction accuracy,a CEEMDAN-PCCs-TCN-XGBoost combined prediction model is proposed.Taking the monitoring data of a certain gravity dam as an example,firstly,the CEEMDAN algorithm is introduced to capture the trend and fluctuation information of non-stationary data,and the PCCs algorithm is combined to determine the main factors affecting data fluctuations.Secondly,in order to improve prediction accuracy,the traditional HST model is used for trend information prediction,and the main factors are used as input variables for fluctuation information prediction.Finally,the TCN model and XGBoost model are respectively applied to predict the data,and the prediction result is accumulated.The prediction results are compared with models such as EEMD-ARIMA and EEMD-LSTM-MLR.The results show that the CEEMDAN-PCCs-TCN-XGBoost combined prediction model is more accurate in predicting dam displacement data with frequent fluctuations.
作者
傅露莹
齐慧君
李同春
姜鹏辉
杜效鹄
FU Luying;QI Huijun;LI Tongchun;JIANG Penghui;DU Xiaohu(College of Water Conservancy and Hydropower Engineering,Hohai University,Nanjing 210098,China;General Institute of Hydropower and Water Resources Planning and Design,Beijing 100120,China)
出处
《三峡大学学报（自然科学版）》
北大核心
2024年第1期1-6,共6页
Journal of China Three Gorges University：Natural Sciences
基金
国家重点研发计划(2022YFC3005403)
中国电建集团科技项目(DJ-ZDXM-2021-10)。
作者简介
通信作者:傅露莹(1999-),女,硕士研究生,主要从事水工结构安全监测研究.E-mail:402333475@qq.com。
