Fracture network characterization with deep generative model based stochastic inversion

The characterization of fracture networks is challenging for enhanced geothermal systems, yet crucial the understanding thermal distributions, and behaviors flow field solute transport. A novel inverse modeling framework proposed estimation networks. hierarchical parameterization method adopted in this work. For a small number large fractures, each characterized by length, azimuth coordination center. dense density fractal dimension are utilized to characterize Moreover, we adopt variational auto-encoder generative adversarial network (VAE-GAN) fuse GAN objective with prior constraint information capture distribution parameters complex satisfy knowledge fields, thereby mapping high-dimensional parameter into low-dimensional continuous field. Afterwards, relying on Bayesian framework, ensemble smoother based collected data from hydraulic tomography reduce uncertainty distribution. Two numerical cases different complexity used test performance framework. results show that algorithm can estimate effectively fields.