Stability-enhanced prestack seismic inversion using hybrid orthogonal learning particle swarm optimization

Prestack seismic inversion quantitatively translates seismic data into multiple elastic properties, supporting the exploration of subsurface reservoirs in detail. Whereas, nonlinearity existing in the physical model causes the ill-posed inverse problem. Although global optimization algorithms own the advantages of achieving better results for highly nonlinear problems compared to linear/local optimization algorithms, such algorithms may suffer from stochastic instabilities or premature convergence when applied to prestack seismic inversion involving multiparameter and multimodal results, especially under complex geological conditions. In this paper, the Metropolis-based probabilistic acceptance criterion is incorporated into orthogonal learning particle swarm optimization (OLPSO), which leads to the proposed hybrid OLPSO (HOLPSO) algorithm. The HOLPSO is introduced to solve prestack seismic inversion problem with special intention of mitigating the instability of the results and the premature convergence of the algorithm. In specific, the stability of multiple results is enhanced by the orthogonal learning (dimensional combination) during model update/perturbation. In addition, the searching capability of the algorithm is improved by selecting guidance particle based on probabilistic acceptance mechanism. Synthetic tests demonstrate the stability and accuracy of the proposed method. Field application shows the method is capable of obtaining fine description of subsurface properties for better identifying potential reservoirs under complex geological conditions.