Home – Home – Jornals – Russian Geology and Geophysics 2018 number Неопубликованное

Seismic inversion is a widely used method of quantitative interpretation in seismic exploration aimed at recovering the elastic properties of the subsurface from seismic data. Subjectivity in selecting computational parameters, including regularization weights and the wavelet, remains a significant problem in practical inversion. Existing approaches to their determination, such as L-curve analysis and related methods, generally involve repeatedly solving the inverse problem for parameter values on a regular grid, followed by an analysis of the stability of the resulting solutions. This strategy effectively reduces to a search through the parameter space and is characterized by low computational efficiency. In addition, the L-curve-based methods considered here are primarily focused on minimizing the misfit between modeled and observed seismic data, whereas the consistency of the inverted model with well-log data is not used as a separate criterion. In this study, we propose a parameter selection approach based on multi-objective optimization. A vector-valued objective function is constructed, consisting of criteria that quantify consistency with both well-log data and seismic data. The optimal solution is selected from the set of Pareto-optimal models as a compromise between these criteria. The method was tested on data from a field in the East Siberian petroleum province. Quantitative evaluation of the inversion results showed an improvement in the selected metrics compared with an L-curve-based scheme for selecting the regularization parameter in the experimental setting considered.