Target-Oriented Time-Lapse Elastic Full-Waveform Inversion Constrained by Deep Learning-Based Prior Model

Time-lapse (TL) seismic monitoring plays a vital role in reservoir characterization and management. Elastic full-waveform inversion (EFWI) has been applied to time-lapse data allow for quantitative estimation of time-varying elastic properties. However, the high-resolution can be computationally intense ill-posed. To estimate changes at reasonable cost, we utilize two key techniques inversion: 1) develop an redatuming approach retrieve virtual both base monitor target level using mainly kinematically accurate velocity, thus, reducing computational cost by focusing on zone; 2) We integrate well information target-oriented inversion, where prior model is predicted deep learning regularize inversion. A neural network (DNN) capable mappings between facies interpreted from after training process. Thus, derive mapping characterized property domain. then implement TLEFWI regularized model, redatumed jointly contributes result. The numerical examples validate that proposed enables us zone with improved resolution consistency.