摘要
砂地比是表征储层侧向输导能力的重要参数,基于线性回归方法结合地震属性预测砂地比易受数据间非线性关系影响,且样本量较少,预测误差较大。以束鹿凹陷束21井区为例,研究了馆陶组砂地比平面分布趋势,通过细分层方法扩充数据样本规模及平滑因子滤波手段对地震属性进行“毛刺”平滑处理,采用皮尔逊相关性优选地震属性,结合多种地震属性和井点砂地比统计数据,使用随机森林算法预测砂地比。结果表明,随机森林法可以有效处理非线性结构数据,与线性回归、神经网络(MLP)对比,模型误差明显降低,预测精度可达80%。该方法的预测结果已成功应用于实际钻井中,且取得了显著的效果。
Sandstone‑strata thickness ratio is an important parameter to characterize lateral transport capability of reservoirs.Sandstone‑strata thickness ratio predicted based on linear regression method combined with seismic attributes is easily affected by nonlinear relationship among data,and less sample amount causes much prediction error.Taking Shu 21 well area in Shulu Sag as an example,areal distribution trend of sandstone‑strata thickness ratio of Guantao Formation is studied.Data sample size is expanded by subdivision method and smoothing factor filtering method is used to“burr”smoothing of seismic attributes.Seismic attributes are optimized by Pearson correlation,and sandstone‑strata thickness ratio is predicted by random forest method combined with various seismic attributes and well point sandstone‑strata thickness ratio statistics data.The results show that random forest method can effectively deal with nonlinear structural data.Compared with linear regression and neural network(MLP),the model error is significantly reduced,with prediction accuracy reaching 80%.Prediction results of this method are successfully applied in actual drilling,and achieves remarkable effect.
作者
付君豪
陈再贺
付宪
李云峰
胡慧婷
FU Junhao;CHEN Zaihe;FU Xian;LI Yunfeng;HU Huiting(College of Resources and Environment,Yangtze University,Wuhan 430100,China;No.5 Oil Production Company of PetroChina Huabei Oilfield Company,Xinji 052360,China;School of Earth Sciences,Northeast Petroleum University,Daqing 163318,China)
出处
《大庆石油地质与开发》
CAS
北大核心
2023年第6期34-41,共8页
Petroleum Geology & Oilfield Development in Daqing
基金
国家自然科学基金面上项目“苯基多环芳烃检测及其石油地球化学意义”(41972148)。
关键词
砂地比
随机森林算法
储层输导能力
束鹿凹陷
sandstone‑strata thickness ratio
random forest algorithm
reservoir transport capacity
Shulu Sag
