Wave-equation based velocity inversion and migration of multiples

Abstract

In recent years, oil and gas exploration is encountering the seismic imaging problems in the area with complex geological structures, thus in seismic data processing wave-equation based prestack depth migration methods with high accuracy have been widely used and intensively studied. However, there are still some problems to be solved in the methods. Firstly, these migration methods require high-resolution migration velocity model to generate correct subsurface images, but how to construct an appropriate velocity model efficiently is a difficult issue in academia and industry. We need to find a better solution to approximate the Hessian matrix, reduce the influence of amplitude information on wave-equation traveltime inversion and improve the computational efficiency of wave-equation full waveform inversion. Secondly, the conventional migration methods only take primaries into imaging process; however, the multiples also contain information of the subsurface, which can also be utilized to improve the imaging quality. Migration of multiples is significant complements for imaging the complex geological structures, and deserves further study. The dissertation focuses on wave-equation velocity inversion and migration of multiples; it involves wave-equation traveltime inversion, full waveform inversion, migration of multiples, and the elimination of migration artifacts. The main research progresses are as following:

1. This dissertation introduces wave-equation traveltime inversion. By comparing the effects of different numerical algorithms to the inversion result, limited-memory Broyden-Fletcher-Goldfarb-Shanno algorithm (L-BFGS) shows it gives the best approximation of the Hessian matrix. L-BFGS bases wave-equation traveltime inversion can improve the resolution of inversion results and computational efficiency effectively. The analysis of traveltime difference gives that traveltime difference based on the cross-correlation of waveform is strongly influenced by the amplitude and varies with the frequency of the observed data. Frequency-dependent wave-equation traveltime inversion is proposed to relieve the influence of amplitude information, reduce local minima in the inversion result and offer an initial model with high quality for subsequent full waveform inversion.
2. The second part of the dissertation introduces the time domain waveform inversion in detail. Based on multiscale early arrival waveform inversion, a pseudo Hessian matrix in time domain is proposed to precondition the gradient and accelerate the convergence rate. The real land data application shows the efficiency of the proposed method.
3. In migration of multiples, the dissertation proposes data to data migration. The approach avoids multiples prediction and wavelet estimation and it¡¯s quite effective for migration of multiples. The defect is the noises in the final image. One-way Fourier finite-difference migration and two-way reverse time migration are used to implement data to data migration. To eliminate the noises in the image generated by one-way wave-equation operator, we propose three methods: 1) using wide azimuth data and 3D data to data migration to generate high-resolution, noise-free imaging results; 2) in data domain use least squares migration to eliminate artifacts in iterative process; 3) in image domain apply Radon transform to angle domain common image gathers and eliminate the energy of noises. To improve the resolution of the image generated by two-way wave-equation operator, we generate the reverse time migration images by employing the imaging condition with wavefield separation. The real marine data application shows the efficiency of data to data migration and the approach to eliminate migration noises.
4. The dissertation combines the two research area and proposes a natural blended full waveform inversion. It use multiples in the waveform inversion process of marine data and reconstruct highly resolved velocity model with less data.