摘要
对单一地震属性采用神经网络技术进行分类,在低信噪比地区很难准确进行地震相分析和砂体预测。根据地震波形理论,基于Seisfacies的多属性体分类、PCA主成分分析、混合聚比法和自组织人工神经网络技术对层段内多种地震属性体按照相似性原则进行统计聚类分析,并在区域内进行地震相自动划分,得到地震相三维数据体。结合钻井资料,借助三维可视化技术在三维空间中预测储层砂体的空间分布,大大减少了地震相划分的多解性,提高了储层预测的精度和工作效率。对华北油田万庄地区扇三角洲砂体进行预测,在三维空间中精细刻画了沙三中段扇三角洲砂体的边界和空间展布,预测了有利储层的发育区,发现并落实了3个有利的岩性圈闭,部署钻探的T12X、T47等井均获工业油流,效果很好。
Conventional neural network technology for seismic waveform classification by using one single seismic attribute is very difficult to be used to predict seismic facies and sand body distribution in low signal and noisy ratio areas.Seisfacies multi-attribute volume classification technology is based on seismic wave theory,principal component analysis (PCA),hybrid classification method and self-organizing neural network technology,by using the principle of similarity to cluster analysis seismic attribute and also seismic facies are automatically analyzed.Thus a seismic facies classification volume is obtained.Integrated with well data,the seismic facies volume is analyzed in 3D visualization.It is better to predict reservoir sand body distribution in three-dimensional space and greatly reduce uncertainty caused by single attribute seismic facies analysis.This technology is used in Wanzhuang area,Huabei Oilfield.Fan delta sand body distribution is correctly described,three prospective lithologic reservoirs are predicted clearly.Based on these results,well T12X and T47 were drilled and encountered thicker oil pays,which proved the multi-attribute volume classification prediction results.
出处
《西南石油大学学报(自然科学版)》
CAS
CSCD
北大核心
2010年第1期57-62,共6页
Journal of Southwest Petroleum University(Science & Technology Edition)
关键词
三维可视化
地震属性
砂体预测
岩性油气藏
神经网络
地震相
three-dimensional visualization
seismic attribute
sand body prediction
lithologic reservoir
neural network
seismic facies