Nonlinear Amplitude Inversion Using Deep Extreme Learning Machine Optimized by Improved Sparrow Search Algorithm
Y. Chen, Mengxuan Song, D.M. Shen, W.K. Wu, Xiaping Fu · 2024
Summary Amplitude variation with offset (AVO) inversion estimates elastic properties of subsurface and thus is an essential tool in oil and gas exploration. Nonlinear inversion based on the exact Zoeppritz equation provides an opportunity to obtain accurate but time-consuming results. In this abstract, we propose a Deep Extreme Learning Machine method optimized by improved Sparrow Search Algorithm (ISSA-DELM) for nonlinear AVO inversion using the exact Zoeppritz equation. Compared with the traditional Deep Extreme Learning Machine method optimized by sparrow search algorithm (SSA-DELM), ISSA-DELM includes a Cauchy-Gaussian mutation strategy to enhance global search capabilities. Numerical tests show that compared with SSA-DELM, ISSA-DELM requires few searches to converge to the global optimal solution, and its inversion results have higher accuracy, efficiency, and robustness. The field data application verifies the good consistency between the inversion parameters and the actual logging results.