Seismic inversion transforms seismic reflection data into quantitative rock-property descriptions of a reservoir, e.g., P-impedance, S-impedance, and density. Regardless of the inversion method used, the accuracy of and confidence in the inversion results rely highly on the quality of gathers or stacks obtained from seismic imaging.
Seismic data bandwidth is limited by signal-to-noise ratio (S/N), absorption, source wavelet, and shot and receiver ghosts. As a result, conventional seismic data lack low frequencies below 7 Hz. A typical deterministic seismic inversion workflow uses the low frequencies of existing well logs by extrapolating or interpolating along stratigraphic layers. The interpolation result is often biased on the well locations and quality of the well logs and can be affected by the interpolation method.
We propose a 3-stage method to minimize the dependency of seismic inversion on a well-log based initial model and improve confidence in the final result. The method includes 1) pre-migration deghosting to remove ghosts in the seismic data, subsequently extending seismic signal to lower frequencies; 2) high-resolution velocity model building with full waveform inversion (FWI) and fault-constrained tomography (FCT) to improve velocity resolution, extending the spectrum to higher frequencies; and 3) simultaneous seismic inversion using the FWI-derived model as the initial model to invert for P-impedance and Vp/Vs.
Presentation Date: Monday, September 25, 2017
Start Time: 3:55 PM
Presentation Type: ORAL