Carbonate microfacies classification model based on dual neural network: a case study on the fourth member of the upper Ediacaran Dengying Formation in the Moxi gas field, Central Sichuan Basin

Abstract

Abstract The dual neural network (DNNW) model combines two neural networks to imitate artificial sedimentary facies division by learning the characteristics of multi-type logging curves corresponding to the sedimentary microfacies of a coring section. The model predicts many non-coring wells in the research zone through logging data. After comparing the classification performance of a fully connected neural network (FCNWW) and AutoML when dealing with three FNDA, EALS, and HRA datasets, the DNNW shows high stability and assists exploration in the Moxi gas field. Four sedimentary microfacies are identified through thin section observation, including thrombolite boundstone (MF1), laminated stromatolite boundstone (MF2), siliceous laminar boundstone (MF3), and micritic dolostone (MF4). Results suggest the fourth member of Dengying Formation in the Moxi gas field is carbonate platform facies deposition: specifically, restricted platform and platform margin.