基于多影响因素联合的某抽水蓄能电站主厂房洞室围岩变形预测  

Prediction of Surrounding Rock Deformation of Cavern in Main Powerhouse of Pump Storage Power Station Based on Combination of Multiple Influencing Factors

作　　者：张翌娜[1] 江琦 张建伟 李香瑞 ZHANG Yi-na;JIANG Qi;ZHANG Jian-wei;LI Xiang-rui(School of Civil and Transportation Engineering,Yellow River Conservancy Technical Institute,Kaifeng 475004,China;School of Water Resources,North China University of Water Resources and Electric Power,Zhengzhou 450046,China)

机构地区：[1]黄河水利职业技术学院土木与交通工程学院,河南开封475004 [2]华北水利水电大学水利学院,河南郑州450046

出　　处：《水电能源科学》2024年第3期161-165,共5页Water Resources and Power

基　　金：教育部重点实验室开放基金资助项目(RMHSE1902);广州市科技计划项目(202002030467)。

摘　　要：为保证地下洞室围岩环境的安全状态,提出了一种以变分模态分解(VMD)方法分解原始数据和粒子群优化算法(PSO)提高预测精度为基础,基于多影响因素联合核极限学习机(KELM)方法的洞室围岩变形预测方法。该方法首先采用VMD方法将监测位移分解为受趋势性因素影响的趋势项位移和受周期性因素影响的周期项位移,去除影响因素的干扰项,其次将演化状态及影响因素作为PSO-KELM的输入数据,预测各影响因素所对应的趋势项或周期项位移,最后叠加两种分项位移,并将多影响因素结合的KELM方法与其他预测方法进行精度比较。对某抽水蓄能电站工程实测数据的验证结果表明,预测结果与原始位移的RRMSE相差仅0.76%,且二者的R为0.986,所提预测方法具有较高的预测精度,可为同类工程的围岩变形预测提供参考。In order to ensure the safe state of the underground cavern surrounding rock environment,a cavern surrounding rock deformation prediction method based on the joint kernel-extreme learning machine method(KELM)with multiple influencing factorswas proposed on the basis of the Variable Modal Decomposition(VMD)method to decompose the original data and the Particle Swarm Optimization(PSO)algorithm to improve the prediction accuracy.The method firstly used the VMD method to decompose the monitored displacements into trend term displacements affected by trending factors and cycle term displacement affected by the cyclical factors,and removed the interference terms of the influencing factors.Secondly,the evolutionary state and influencing factors were taken as the input data of the PSO-KELM to predict the trend term or cycle term displacements corresponding to the influencing factors.Finally,the two types of displacementswere superimposed,and the accuracy of the KELM method combining the multi-influencing factors to other prediction methods was compared.The validation results of the measured data of a pumped storage power station project show that the difference between the prediction results and RRMSE of the original displacement is only 0.76%,and the R of the two is 0.986.The above prediction method has higher prediction accuracy,and it can be used as a reference for the prediction of the surrounding rock deformation of the same kind of projects.

关 键 词：洞室围岩 变形 变分模态分解 粒子群优化 核极限学习机 影响因素 

分 类 号：TV731.6[水利工程—水利水电工程]