利用Kohonen神经网络划分二维地震相——以柴达木盆地E区风险勘探为例

Application of Kohonen neural network to 2D seismic facies division in E area of Qaidam Basin

为研究柴达木盆地E区大型背斜构造沉积相及砂体展布规律,在二维地震数据闭合差校正、邻区井标定引入及精细层位解释基础上,利用改进算法的Kohonen神经网络技术开展二维地震相划分研究,识别出三角洲前缘水下分流河道、分流间湾及滩坝等微相。本文研究认为,研究区古流向为南东—北西向,储集砂体较发育,主要富集于研究区中部,现今构造东高点位于有利沉积相带。改进算法的Kohonen神经网络二维地震相划分技术补充了沉积相研究成果,适合于西部二维地震资料覆盖的风险探区,具较强的推广价值。

The situation of E area is hard to proceed deep research and risk assessment by the absent of prospecting well data.Based on seismic mis-tie calibration,adjacent well calibration and fine horizon interpretation,Kohonen neural network method is applied to carry out two-dimensional seismic facies classification of target zone.Microfacies of delta front such as distributary channel,interdistributary bay and sand bar are recognized.The paleo-current direction is suspected from southeast to northwest.Reservoir sand bodies developed well in the middle of the study area,preliminary prospecting well is located in favorable sedimentary facies.Sedimentary facies division are supplied and refined by the result of seismic facies,which can supply significant foundation for geometric arrangement of risk wells and regional breakthrough.