【作者】 张亚中；

【导师】 张贵宾；

【作者基本信息】 中国地质大学（北京） ， 地球探测与信息技术， 2010， 博士

【摘要】 丘陵油田位于吐哈盆地台北凹陷,构造被南北向断裂切割成东西两块,含油层系为中侏罗统七克台组、三间房组和西山窑组。2005年吐哈油田实施了第一块针对已开发油田的二次高精度地震资料采集、处理、解释一体化项目。本次研究基于高精度三维地震资料,结合井、测井和开发动态资料,针对影响丘陵油田开发的小断层和储层进行了精细解释,取得了较为丰富的研究成果。油田开发面临难题多种多样,实质上是油藏地质特征认识不够深入,即对小断层识别和储层预测的精度不足以满足油田开发的需要。为此,本次研究将小断层识别和储层预测作为重点。本次研究将模型正演、谱分解技术应用于小断层的识别和解释,识别了断距为10m以下的小断层,构建了一套适合丘陵油田及类似地震地质条件下的高精度地震资料小断层解释技术框架。原有构造研究成果表明本区断层主要有两组,本次研究应用构造族系和地质力学的理论,将区内断裂系统划分为三个构造族系,使得断层组合、切割关系分析、断层平面展布特征和小断层的识别更加符合丘陵构造带地质应力环境和地质规律。鉴于小波分析具有良好的时频局部性质,而神经网络具有良好的容错能力和自学习功能,因此本次研究采用小波神经网络进行地震约束反演。在实施约束反演时,首先整理探区内钻井、测井、三维地震数据等基础资料;在此基础上优选反演约束条件,给小波神经网络各参数赋予随机初始值,并不断调整网络参数,直至完成网络训练。将未作为约束条件的测井曲线叠加到地震反演结果上进行分析发现,以电阻率为约束条件的地震反演结果层次清楚,剖面高阻抗和自然电位测井曲线异常具有较好的对应关系。本次研究提出了模糊支持向量机方法,并将其用于了储层预测。首先仔细分析丘陵油田的地层、沉积与岩石地球物理等特征,确定探区内主要目的层段;然后采用单道时窗提取与多道时窗提取等手段提取目的层段属性,以构成学习样本。在此基础上训练FSVM,并用训练结果对目的层段实施预测。根据本次研究可知,丘陵油田具有二次开发潜力,区内下侏罗统、三叠系和二叠系构造落实可靠、储盖组合发育,油源丰富,具有良好的勘探前景。更多还原

【Abstract】 Qiuling oilfield is located in Taibei depression, Tulufan-Hami basin, where structure is cut into two blocks by north-south faults and where Qiketai group, Sanjianfang group and Xishanyao group of the Middle Jurassic are the source formations. In 2005, the integrated high-precision 3D seismic project including data acquisition, processing and interpretation was implemented in Tuha oilfield, which was to resolve the problems found in oil-field development. Based on high-precision 3D seismic data, combing with borehole data, logging data and development dynamic information, we subtly interpreted the small faults and predicted reservoirs.Materially, the problems facing to oil-filed development can be taken that the ability to recognize small faults and the precision of reservoir prediction are not meet the requirement in oil-field development. In the paper, small faults which displacement is less than 10m can be identified by forward modeling and spectral decomposition. Based on high-precision 3D seismic data, technique framework of seismic data interpretation which is suitable for geological conditions like Qiuling oil-filed was constructed.The old research results show there are two sets of faults. According to structure family theory and geologic mechanics theory, the new fault system is classified as three structure families in the research area, which makes fault combination, cutting relationship and distribution in plane clearly, and is according with geologic stress circumstance and geologic rules of Qingling structure belt.Wavelet analysis has good local time-frequency character and neural network has allowing-error ability and auto-learning ability, so wavelet neural network method is selected to perform seismic constrained inversion. Before seismic constrained inversion, it is necessary that well data, logging data and 3D seismic data in the research area are neatened. After selecting optimum constrained condition, we give the random initial value of each parameter of wavelet neural network, and continuously adjust network parameters until the completion of network training. Comparing the inversion results and the log curves which are not used as constrained condition, it can be found that the inversion results of resistivity curve used as constrained condition is clearly, and there is a good corresponding between high impedance in inversion result and abnormity in spontaneous potential curve when. In our research, FSVM is used to reservoir prediction.After analyzing the characteristics of lithology, sedimentary and petrophysics of the targets in Qiuling oil-field, the attributes of targets are extracted by the means of single or multi-channel time-window. Learning samples is composed of these attributes. After training FSVM, the results is used to predict the reservoir. Random comparing prediction results with drilling data, prediction errors are within the allowable range. The research result shows that there are potential of secondary development in Lower Jurassic, Triassic and Permian the structures are reliable,and the source is abundant.更多还原