Abstract
Accurate characterization of complex reservoirs is challenging due to limited training samples and lateral heterogeneity. To address these challenges, an intelligent elastic parameter inversion method based on non-parametric analysis and information-theoretic constraints guided by seismic waveforms is proposed. Firstly, the seismic waveforms are clustered, and then the training samples are amplified by kernel density estimation in each category to ensure that the amplified samples are consistent with logging data in both statistical distribution and geological characteristics. Furthermore, an information-theoretic constraint based on the Kullback–Leibler divergence is introduced and jointly optimized with model-driven and data-driven loss functions, thereby enforcing distributional consistency during inversion. Applications to both a thin interbedded geological model and field data demonstrate that the proposed method outperforms traditional approaches in terms of inversion accuracy and lateral continuity, while effectively capturing both global trends and local variations in elastic parameters.
Paper Information:
Chen Tengfei, Zhang Guangzhi, Yang Yahua, Hao Hongjian, Chang Dekuan, Gao Jianhu, Gao Gang. Seismic waveform guided intelligent elastic parameter inversion method based on non-parametric analysis and information-theoretic constraints[J]. Petroleum Science, 2026. Doi:https://doi.org/10.1016/j.petsci.2026.07.033.

