Prediction of Beach-Bar Sand Reservoir Based on Machine Learning
P. An, S. Yang, Fang-Yuan Xu, J. Zhang, Z. Liu, Fei Liu, P. Li, Ganghua Huang · 83rd EAGE Annual Conference & Exhibition · 2022
Summary Beach-bar sand reservoir is a special type of thin interbedded reservoir, which has become a new growth point of lithologic oil and gas reservoir. Due to the characteristics of deep burial, thin single-layer sand body, rapid lateral change and low resolution of seismic data, it is difficult to realize the fine prediction of beach-bar sand reservoir by using conventional seismic technology. Aiming at the difficulty of beach bar sand reservoir prediction, this paper uses the “spectrum inversion” technology based on matching tracking algorithm to complete the frequency extension processing of seismic data; Further, the multi-attribute information is used as the input of radial basis function neural network to predict the internal variation characteristics of beach bar sand reservoir; Finally, the prediction results of fluid activity attributes are fused with paleogeomorphic information to clarify the favorable reservoir forming area of beach bar sand body. The prediction results show that this technology can effectively identify the lateral variation details of beach bar sand reservoir and effectively guide the exploration and development of beach bar sand reservoir.