The field of seismic full-wave inversion (FWI) and high-resolution reservoir simulation are in a transition period where methods based on simplified wave propagation and flow physics phenomena are successively replaced by fully numerical approaches of dense earth model representations running on high-performance computers (HPC) that allow exploration geophysicists and reservoir engineers to unlock, leverage tomorrow's reserves and mitigate subsurface uncertainties in a cost-effective way. The objective is on the one hand to exploit the complete seismic recordings for the benefit of improved vertical resolution of subsurface elastic anisotropic heterogeneous Earth models; and on the other hand, to understand physics at small scale to improve microscopic recovery and improve reservoir modeling predictability towards optimized field development with proprietary in-house developed technology run on HPC. The promise of elastic FWI for seismic imaging and interpretation is to employ waveforms (raw observed seismograms recorded with long/broad range and densely sampled offset/azimuths and full frequency bandwidth) to account for refractions, reflections and high-order scattering, and make NO physical assumptions in the simulation of any observed amplitudes. However, FWI is an ill-posed problem with non-unique solutions, i.e., many combinations of earth elastic parameters can fit the data equally well. The non-uniqueness of solution triggers the uncertainty in the earth model parameters which equally affect both seismic imaging and reservoir modeling workflows. As a result, there is a need to create multiple sets of models in an attempt to optimally explain the data (seismograms). Therefore, the solution of a seismic inverse problem has very high computational complexity that can only be efficiently handled using high-performance computers (HPC). Such computers contain large numbers of nodes interconnected via the high throughput networks; each node combines conventional CPU cores and GPU (graphic processing units) accelerators. Through a combination of theory, methodology and case studies, we demonstrate the recent progress and value added with cost-effective fit-for-purpose Total's proprietary technology efficiently deployed on heterogeneous HPC to yield new acreages in frontier domains, better characterize/predict subsurface reservoirs, and mitigate the geosciences and drilling uncertainties.
展开▼
