Integrated artificial intelligence approach for well-log fluid identification in dual-medium tight sandstone gas reservoirs

Frontiers in Earth Science, Год журнала: 2025, Номер 13

Опубликована: Апрель 28, 2025

Introduction With the development of complex tight sandstone oil and gas reservoirs, accurately cost-effectively characterizing these reservoirs have become a critical yet challenging task. To address limitations conventional machine learning algorithms, which low accuracy due to data inhomogeneity weak fluid logging responses, this study introduces novel method for evaluation in dual-medium reservoirs. Methods The integrates core, thin section, scanning electron microscope observations, taking into account effect fractures. Results Reservoirs are divided three types: fractured (FR), porous (PR), microfracture-pore composite (MPCR), highlighting distinct responses each type. Reservoir classification based on geological genetic mechanism significantly reduces noise prediction ambiguity, thereby improving efficiency model training. Discussion final is constructed by an ensemble that multiple sub-models, including fuzzy C-means clustering (FCM), gradient boosting decision tree (GBDT), backpropagation neural network (BPNN), random forests (RF), light machines (LightGBM). Applied West Sichuan Depression Basin, validation reached 91.96%. In summary, reliable log prediction, improved its robustness compared with single models traditional methods, providing comprehensive perspective across geophysical disciplines

Язык: Английский

Discovering Optimal Markov Blanket for High-Dimensional Streaming Features

Information Sciences, Год журнала: 2025, Номер 716, С. 122240 - 122240

Опубликована: Апрель 29, 2025

Язык: Английский