摘要
天然气水合物与天然气储层识别一直是海洋能源勘探开发阶段的重点任务。然而,由于测井数据与储层之间的复杂非线性关系以及测井数据的不均衡性,导致传统储层识别方法往往精度不高,严重限制了研究区域的勘探进展。为解决上述问题,提出了一种用于储层识别的混合方法,即采用改进的SMOTE算法增加少数类储层样本数量,并进行去噪处理,可有效地解决数据不均衡的问题,再利用XGBoost算法对储层进行识别。结果表明:相比于传统的机器学习方法,RLSMOTE-XGB方法在储层识别方面具有更高的有效性和准确性,该方法解决了传统机器学习方法在样本类别不均衡时的局限性,储层识别精度从66.7%提高至86.4%,算法的性能得到显著提升。该研究可有效提高天然气水合物与天然气储层识别效果,对实现智能化识别储层有重要意义。
Natural gas hydrates identification and characterization are the key tasks throughout the exploration and development phase of marine energy resources.However,due to the complex nonlinear relationship between logging data and reservoirs,as well as the imbalance of logging data,traditional reservoir identification methods often show low accuracy,which severely limited the progress of energy exploration in the study area.To address the above issues,a composite method for reservoir identification is proposed.The improved SMOTE algorithm is used to increase the number of minority class reservoir samples and denoise the data,which effectively solves the issues of data imbalance.The XGBoost algorithm is then used to identify reservoirs.The results show that compared with traditional machine learning method,the RLSMOTE-XGB method has higher effectiveness and accuracy in reservoir identification.This method addresses the limitations of traditional machine learning methods in the case of imbalanced sample classes,increasing the reservoir identification accuracy from 66.7%to 86.4%and significantly improving the algorithm′s performance.This study can effectively improve the identification effect of natural gas hydrates and natural gas reservoirs,which is of great significance for achieving intelligent reservoir identification.
作者
杜睿山
黄玉朋
付晓飞
孟令东
张轶楠
靳明洋
蔡洪波
Du Ruishan;Huang Yupeng;Fu Xiaofei;Meng Lingdong;Zhang Yi′nan;Jin Mingyang;Cai Hongbo(Northeast Petroleum University,Daqing,Heilongjiang 163318,China;Key Laboratory for Evaluation of Oil and Gas Reservoir and Underground Storage Integrity in Heilongjiang Province,Daqing,Heilongjiang 163318,China;PetroChina Liaohe Oilfield Company,Panjin,Liaoning 124010,China)
出处
《特种油气藏》
CAS
CSCD
北大核心
2024年第5期11-19,共9页
Special Oil & Gas Reservoirs
基金
国家重点研发计划“区域二氧化碳捕集与封存关键技术研发与示范”(2022YFE0206800)
黑龙江省自然科学基金“基于多源深度强化学习的复杂场景视频事件检测”(LH2021F004)。
