Zongbin Liu ¹ Jianmin Zhu ¹ Bo Tian ¹ Rui Zhang ¹ Yongheng Fu ¹ Yuan Liu ¹ Lixin Wang ^2*

• ¹ CNOOC China Limited, Tianjin Branch, Tianjin, China, Tianjin, China
• ² School of Geosciences, Yangtze University, Wuhan, China

As the demands for exploration continue to rise, the identification of thin sand bodies becomes significantly important for subsequent petroleum exploration and development efforts. To enhance the clarity and detail of seismic readings, seismic inversion plays an important technique, especially in predicting inter-wells sand bodies as hydrocarbon reservoirs. However, traditional inversion techniques struggle with complex subsurface structures. More sophisticated approaches demand considerable computational efforts. In this study, we adopted a novel approach combining spectral decomposition with convolutional neural networks (CNNs) within a genetic algorithm (GA) framework for inversion. The CNNs are adept at recognizing and interpreting the spatial configurations in the data, thereby establishing a dynamic linkage between seismic attributes and sand body distributions. This fusion of CNNs and GA facilitates an efficient and high-fidelity interpretation of sand formations, closely mirroring geological conditions, and does so with rapid processing times. The outcomes reveal that the model's sand thickness predictions closely match the actual measurements at well sites, with a new horizontal well's alignment with the predicted outcomes reaching an impressive accuracy of 85.1%.