Research on Intelligent Fusion of Seismic Attributes from Spectral Decomposition Based on Automated Machine Learning for Predicting Sand Body Thickness: A Case Study of Z Oilfield in Kazakhstan

Z Oilfield has experienced more than 50 years of waterflooding development. After 2020, three appraisal wells drilled in the peripheral low-amplitude anticline area of the highly water-cut (77.1%) J-13 layer were all successful, demonstrating the potential for rolling exploration. This study aims to predict the thickness and distribution area of sand bodies in undrilled areas by extracting seismic information, providing a basis for the deployment of new appraisal wells. The study uses an algorithm based on automated machine learning to extract seismic information from different frequency bands and solve the non-linear fusion relationship of seismic attributes, so as to achieve fine quantitative prediction of sand bodies with different thicknesses and construct sand body thickness maps. The results show that this method is highly effective. It has increased the correlation coefficient between single seismic attributes and sand body thickness from 0.475 to 0.804, accurately predicting the thickness and distribution area of sand bodies and providing a reliable basis for the deployment of appraisal wells. This study innovatively applies machine automated learning with integrated multiple algorithms and automated parameter optimization to predict sand body thickness, effectively solving the problem of low correlation between seismic attributes and sand body thickness, improving the prediction accuracy, and providing key technical support for the rolling evaluation and development of the oilfield.