Lithology identification using electrical imaging logging image: A case study in Jiyang Depression, China

Abstract

Lithology identification plays a significant role in stratigraphic evaluation and geological analysis. Traditional lithology identification method is by modeling the relationship between well logging and lithology. However, well logging are not always sufficient to identify lithology since sometimes the curves are similar for different lithologies. Recently, electrical imaging logging image (EILI) with high resolution plays an increasingly important role in logging interpretation since EILI can intuitively reflect the characteristics of lithology. Unlike traditional lithology identification method by using well logging, in this paper, we propose a novel multi-dimensional automatic lithology identification method by applying deep learning to EILI. First, Filtersim algorithm is employed to fill the blank strip of the EILI. Then, an integrated convolutional neural networks (CNNs) model is designed to extract the resistivity feature, texture feature, and holistic feature of the EILI, respectively. Specifically, the integrated CNNs model can realize automatic recognition for different geological structures (massive, bedded, lamellar) and lithology (mudstone, sand-mudstone, lime-mudstone). Finally, lithology identification can be achieved by combining with multi-dimensional features. The efficacy of proposed integrated model is validated experimentally on the EILI of shale oil reservoir in the Jiyang Depression of China. Experimental results show the effectiveness and superiority of the integrated CNNs method for lithology identification.