A Deep Learning Structure Correction Method Under the Influence of Special Lithology Body

Abstract In the process of precise exploration and development of oil and gas reservoirs, the accurate implementation of the target layer structure plays a vital role in the increase of oil reserves and the success of drilling. However, with the gradual deepening of the research, it is found that in the overlying strata of oil and gas reservoirs, special lithological bodies with strong heterogeneity and abnormal velocity are usually locally developed, which directly affects the construction of seismic data migration velocity, resulting in structural distortion at the local position of the reservoir in the process of data interpretation. Focusing on the structural change caused by the abnormal velocity, the deep learning algorithm is used to accurately predict the velocity of the special lithologic body and realize the accurate correction of the structure. Firstly, based on the combination of seismic and geological understanding, the geological model of the target layer and the overlying special lithologic body is constructed, and the distribution law of the special lithologic body is identified by the forward simulation method. Then use the well-side seismic trace and P-wave velocity curve to train the velocity data through the deep learning method and apply the training results to the post-stack seismic data to obtain the accurate P-wave velocity body. Finally, the real formation velocity over the target layer is restored to achieve reservoir structure correction. Based on this method, the research experiment was carried out in the M block of the Tarim Basin, and the spatial distribution range of the special lithology body was carved. The data error between the migration velocity body and the abnormal high-velocity body was calculated, and the target layer structure in the study area was corrected.