摘要
裂缝系统对潜山油气藏的形成至关重要,裂缝表征和评价是潜山储集层预测的关键。以成像测井及岩心分析为依据,开展了大民屯凹陷S 625区块潜山岩性、裂缝期次、产状及有效性分析,并重点分析了气测、元素、工程录井及常规测井参数对裂缝的响应,进而采用人工神经网络方法,开展常规测、录井资料建立裂缝密度计算模型的研究。在模型训练过程中,将输入参数分选为随钻测、录井数据(电阻率、自然伽马、井径、钻时、气测及其衍生数据等),元素录井数据及其他测井数据(声波时差、密度等)三类,将此三类数据逐批次输入到BP模型中进行训练,发现随着输入数据量的增加,BP模型对裂缝密度的总相关程度显著提升。现场应用结果表明,基于神经网络模型计算的裂缝密度参数与成像解释及岩心裂缝发育程度吻合度较高,能为裂缝储集层建模提供可靠的井点依据。
Fracture system is very important to the formation of buried hill reservoir,and fracture characterization and evaluation is the key to buried hill reservoir prediction.Based on imaging logging and core analysis,the lithology,fracture stages,occurrence and effectiveness of buried hill in S 625 block of Damintun Sag are analyzed.The response of gas logging,element,engineering logging and conventional log information to fracture is analyzed emphatically.Then,the artificial neural network method is used to study the fracture density computation model based on conventional well logging and mud logging data.In the process of model training,the input parameters are classified into three categories:well logging and mud logging while drilling data(resistivity,natural gamma,borehole diameter,drilling time,gas logging and its derived data),element logging data and other log data(interval transit time,density,etc.).The three categories of data are input into BP model batch by batch for training.It is found that the total correlation degree of BP model to fracture density is significantly improved with the increase of input data volume.The results show that the fracture density parameters calculated by neural network model are in good agreement with imaging interpretation and core fracture development degree,which can provide reliable well point basis for fracture reservoir modeling.
作者
徐辰浩
XU Chenhao(GWDC Mud Logging Company,CNPC,Panjin,Liaoning 124010,China)
出处
《录井工程》
2020年第3期122-129,共8页
Mud Logging Engineering
关键词
大民屯凹陷
裂缝密度
裂缝产状
裂缝有效性
神经网络
计算模型
气测
元素
Damintun Sag
fracture density
fracture occurrence
fracture effectiveness
neural network
computation model
gas logging
element
