基于深度学习和地震数据的海上风电场CPT预测研究  

Research on CPT Prediction for Offshore Wind Farms Based on Deep Learning and Seismic Data

作　　者：李书兆 孙国栋 申辰 李文逵 罗进华[3] 王教龙 Li Shuzhao;Sun Guodong;Shen Chen;Li Wenkui;Luo Jinhua;Wang Jiaolong(CNOOC Research Institute Ltd.,Beijing 100028,China;Institute of Geophysics&Geomatics,China University of Geosciences(Wuhan),Wuhan Hubei 430074,China;Exploration Technology Dpartment,China Oilfield Services Limited,Tianjin 300459,China)

机构地区：[1]中海石油(中国)有限公司北京研究中心,北京100028 [2]中国地质大学(武汉)地球物理与空间信息学院,湖北武汉430074 [3]中海油田服务股份有限公司勘察技术部,天津300459

出　　处：《工程地球物理学报》2025年第2期216-226,共11页Chinese Journal of Engineering Geophysics

基　　金：国家“十四五”重大科技项目(编号:KJZX-2022-12-XNY-0100)。

摘　　要：为了应对海上风电场对海底岩土参数探测的需求,以海南岛西部的莺歌海盆地东方区海上风电项目建设为例,用建立邻域的方式将地震数据和孔压静力触探试验(Cone Penetration Test,CPT)数据联系起来,构建深度学习岩土参数预测模型。通过K折交叉验证对模型进行训练,并与支持向量机、随机森林和神经网络模型进行了对比。实验结果表明,所提出的深度学习岩土参数预测模型在验证精度方面显著优于3种机器学习模型,相对误差率小85%,拟合优度大21%,均方根误差小59%,平均绝对百分比误差小71%,在测试孔位CPT-16上对比4种方法,深度学习岩土参数预测模型最拟合,能预测出CPT岩土参数的变化趋势,在预测地震剖面的岩土参数方面,深度学习岩土参数预测模型能呈现出细致且自然的地层起伏变化。总结,深度学习岩土参数预测模型具有很高的可靠性,能够对海上风电场海底岩土参数进行有效的预测。In order to meet the demand for offshore wind farms for seabed geotechnical parameter detection,this study takes the construction of the offshore wind power project in the east area of Yinggehai Basin in the west of Hainan Island as an example.The study links the seismic data and cone penetration test(CPT)drilling data by establishing a neighborhood and constructs a deep learning geotechnical parameter prediction model.The model is trained using K-cross validation and compared with support vector machines,random forests and neural network models.The experimental results indicate that the proposed deep learning model for predicting geotechnical parameters significantly outperforms three other machine learning models in terms of validation accuracy,with an 85%reduction in relative error rate,a 21%improvement in goodness of fit,a 59%decrease in root mean square error,and a 71%reduction in average absolute percentage error.Compared with the four methods on the test hole position CPT-16,the deep learning geotechnical parameter prediction model provides the best fit and can predict the trend of CPT geotechnical parameters changes.Additionally,in predicting geotechnical parameters of seismic profiles,the deep learning geotechnical parameter prediction model can show detailed and natural formation relief changes.In summary,the deep learning model for predicting geotechnical parameters demonstrates high reliability and can effectively predict seabed geotechnical parameters for offshore wind farms.

关 键 词：海上风电 岩土参数预测 地震 CPT钻孔 深度学习 

分 类 号：P631.4[天文地球—地质矿产勘探]