Quantitative monitoring for enhanced geothermal systems using double-difference waveform inversion with spatially-variant total-variation regularization

Double-difference waveform inversion is a promising tool for quantitative monitoring for enhanced geothermal systems (EGS). The method uses time-lapse seismic data to jointly invert for reservoir changes. Due to the ill-posedness of waveform inversion, it is a great challenge to obtain reservoir changes accurately and efficiently, particularly when using time-lapse seismic reflection data. To improve reconstruction, we develop a spatially-variant total-variation regularization scheme into double-difference waveform inversion to improve the inversion accuracy and robustness. The new regularization scheme employs different regularization parameters in different regions of the model to obtain an optimal regularization in each area. Utilizing a spatially-variant regularization scheme, the target monitoring regions are well reconstructed and the image noise is significantly reduced outside the monitoring regions, and the results are significantly better than those obtained obtained using a constant regularization parameter. Our numerical examples demonstrate that the spatially-variant total-variation regularization scheme provides the flexibility to regularize local regions based on the a priori spatial information without increasing computational costs and the computer memory requirement.