A Prestack Elastic Parameter Seismic Inversion Method Based on xLSTM-Unet

Prestack seismic data retain the amplitude variation with offset (AVO) characteristics, providing more geophysical information reflecting lateral reservoir variations, thus facilitating the identification of oil and gas reservoirs. However, due to the band-limited nature of seismic data, the precision of forward modeling, and the accuracy of algorithms, traditional prestack approaches suffer from ambiguity and uncertainty. With the development of deep learning and big data, an increasing number of deep learning methods have been proposed. We integrate the extended long short-term memory (xLSTM) modules with the Unet framework, and design a novel neural network architecture, that is, xLSTM-Unet, for elastic parameter inversion (<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$V_{\text {P}}$ </tex-math></inline-formula>, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$V_{\text {S}} $ </tex-math></inline-formula>, and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\rho $ </tex-math></inline-formula>) from prestack seismic gathers. Through testing on synthetic seismic records and field data, the proposed xLSTM-Unet outperforms both the traditional Unet and LSTM-Unet models in predicting elastic parameters from prestack seismic data. The xLSTM-Unet proposed in this article provides a stable and effective approach for predicting prestack elastic parameters, offering new insights for the intelligent development of seismic exploration.