摘要
准确描述并预测裂缝空间分布是裂缝型潜山油藏储层开发的前提条件,裂缝型潜山油藏构造复杂、断层发育、地震反射多表现为空白或杂乱反射,传统的依赖模型的反演方法难以表征复杂的潜山断裂系统。为此,提出了以测井裂缝敏感参数FIC c为训练目标的深度信念网络反演方法。首先基于碳酸盐岩裂缝在测井曲线上的响应特征,综合伽马、井径等曲线,构建敏感性裂缝识别参数FIC c;然后以基于优选井计算的FIC c曲线作为训练目标、以井旁地震数据作为训练特征构成训练样本,通过训练深度信念网络,建立井旁地震数据与FIC c曲线之间的非线性映射关系;最后将训练成熟的网络模型应用到整个地震数据体,反演得到裂缝识别参数FIC c,进而预测裂缝空间分布。S区潜山裂缝预测的应用结果表明,测井裂缝识别参数FIC c识别结果与成像测井裂缝识别结果基本吻合,FIC c作为训练目标在S区裂缝预测中具有较好的可靠性;应用深度信念网络反演的解释结果表明S区潜山主要发育北东向裂缝,呈带状沿断层大面积发育,与熵属性刻画的裂缝发育带一致性较好,钻井吻合率达71%。
Accurate description and prediction of fractures is key to the effective development of fractured buried hill reservoirs.The structure of these reservoirs is complex:as they feature developed faults,the seismic reflections are mostly blank or random.The traditional model-dependent inversion method can hardly characterize the fracture system of complex buried hills.Therefore,a deep belief network method was proposed which uses petrophysical fractured sensitive characteristic curves as the training target.Firstly,based on the logging response of carbonate fractures,the sensitive fracture identification parameter FICc was constructed by integrating gamma and caliper curves.Then,a nonlinear relationship between borehole-side seismic data and FICc was established by training the deep belief network,with the optimized well FICc curve as the training target and the borehole-side seismic data as the training features.Finally,the trained network model was applied to the entire seismic data to predict the spatial distribution of fractures.The results of the application to a buried hill in the S area showed that the results obtained by logging the fracture identification parameter FICc were consistent with those using the imaging logging.Moreover,using FICc as learning samples provided a better applicability in fracture prediction.The inversion results using deep belief network in the S area showed a well-developed north-east trending fracture along the fault.These results were consistent with those obtained through the entropy attribute with a well drilling anastomosis rate of 71%.
作者
丁燕
杜启振
Qamar Yasin
张强
刘力辉
DING Yan;DU Qizhen;YASIN Qamar;ZHANG Qiang;LIU Lihui(Key Laboratory of Deep Oil and Gas,China University of Petroleum(East China),Qingdao 266580,China;Laboratory for Marine Mineral Resources,Qingdao National Laboratory for Marine Science and Technology,Qingdao 266237,China;Key Laboratory of Geophysical Prospecting,CNPC,China University of Petroleum(East China),Qingdao 266580,China;Research Institute of Exploration and Development,Shengli Oilfield Branch,SINOPEC,Dongying 257015,China;Beijing Rockstar Petroleum Technology Co.,LTD,Beijing 100000,China)
出处
《石油物探》
EI
CSCD
北大核心
2020年第2期267-275,共9页
Geophysical Prospecting For Petroleum
基金
中国科学院战略性先导专项(XDA14010303)
高等学校学科创新引智计划(“111计划”)“致密油气地质与勘探创新引智基地、深层超深层油气地球物理勘探创新引智基地”联合资助。
关键词
裂缝预测
潜山
碳酸盐岩
测井
裂缝识别参数
深度学习
非线性反演
熵属性
fracture prediction
buried hill
carbonate
logging
fracture identification parameter
deep learning
nonlinear inversion
entropy attribute
