自组织人工神经网络算法在储层预测中的应用

Application of Self-organizing Artificial Neural Network in Reservoir Prediction

Su10井区位于苏里格气田,该气田是一个有效储离散分布的大规模低效气田。通过对测井曲线和取心岩样数据建立测井解释标准模型,寻找有效砂体并确定其岩性物性参数范围。从地震资料中提取的地震属性资料,运用自主神经网络算法学习标准储层参数与这些属性之间的权值关系,通过学习的权值关系来分析目标储层的地震属性特征,反推其岩性物性参数得到有效砂体的分部范围和厚度预测结果,为进一步的储层预测给出参考意见。

Su10 well field is located in Sulige gas field, which is a large-scale and inefficient gas field with discontinuous effective storage. Firstly, the standard model of log interpretation was established by logging curve and core sample data. Effective sand body was searched, and the range of lithologic and physical parameters was determined. Seismic attribute data were extracted from seismic data, and the weight relationship between the standard reservoir parameters and these attributes was studied by using the autonomous neural network algorithm, the lithologic and physical parameters of the target reservoir were calculated. Finally, the qualitative analysis results of reservoir obtained by wave impedance inversion and the definite analysis results obtained by seismic attribute analysis were synthetically analyzed to provide reference opinions for further reservoir prediction.