Data-driven pre-stack AVO inversion method based on fast orthogonal dictionary

Abstract

Abstract Regularization techniques are commonly used as prior constraints to address or mitigate the ill-posed problems for pre-stack AVO inversions. Because of the difficulty of applying conventional regularization methods for complex reservoirs, in recent years, a data-driven method has been proposed by scholars to intelligently obtain prior information for formations. This method uses the dictionary learning algorithm to log data to learn a dictionary that can characterize the structural characteristics of elastic parameters. The unambiguity of the inversed elastic parameters can be reduced by applying information of dictionary sparse representation as a prior constraint. This method has been proved to greatly improve the inversion accuracy. However, applications of the popular KSVD algorithm suffers from its slow operation efficiency and many operational parameters. This paper aims to address this problem by proposing a novel data-driven intelligent pre-stack inversion method based on the orthogonal dictionary (ORTD) learning method. Because the atoms are orthogonal, the efficiency of dictionary operation is nearly 100 times higher than that of KSVD redundant dictionary. Instead of regulating the sparsity, atomic size and atomic length for the KSVD method, the proposed method depends on threshold parameters and dictionary atomic length. Experiments indicate that the performance of the proposed pre-stack AVO inversion method is mainly affected by the threshold size and the sparse reconstruction accuracy is strongly influenced by the hard threshold value. Because the size of the hard threshold coefficient weakly influences the reconstruction efficiency, the proposed method can have comparable dictionary learning and sparse representation abilities with KSVD dictionary by regulating the threshold value in applications. Moreover, the adjustment of parameters has little impacts on the operation efficiency, which greatly improves the applicability of the data-driven inversion method. We use model tests and field data applications to verify the proposed sparse representation regularization pre-stack AVO inversion that is based on the fast orthogonal dictionary method. Compared with the KSVD inversion algorithm, the proposed inversion can obtain comparable results and has advantages in tuning parameters and inversion efficiency.