Prediction of Macrolithotypes Distribution for Deep Coalbed Methane Exploration in Eastern Ordos Basin

Deep coal seams (>2000 m) are increasingly targeted for coalbed methane (CBM) development due to their abundant resources in China. Macrolithotypes, as key indicators of coal heterogeneity, significantly influence porosity and permeability, especially under deep burial conditions where diagenetic effects further obscure lithological boundaries. Traditional identification methods based on core observations are impractical in deep formations due to high cost, limited recovery, and subjective interpretation. To address this, this study applies a backpropagation (BP) neural network model to predict macrolithotypes distribution in deep coal seams, using well-logging data (GR, DEN, AC, RLLD) as input features. The model successfully captures nonlinear relationships between logging parameters and bright component content, enabling accurate vertical discrimination of macrolithotypes at depth. Validation results show a high prediction accuracy of 97.1%, with most deviations below 1.0, confirming the model’s robustness even in deep coal intervals. Prediction errors are mainly limited to lithotype transition zones where overlapping log responses reduce resolution. Based on the BP model framework, a single-well brightness model is further developed by defining a brightness coefficient (Bc) to quantify coal brightness in deep strata. Results indicate that the no. 8 coal seam is vertically dominated by bright coal and semibright coal. Regionally, a clear north–south brightness gradient is intensified, with brightness increasing from the northeast to southwest. This spatial variation is interpreted as a result of paleo-depositional controls: deeper and more stable water tables in the southwestern peatlands promoted continuous reductive conditions and the development of bright coal, whereas the northeastern areas experienced more dynamic hydrological regimes and clastic input, leading to higher inertinite content and fragmented lithotypes. These findings provide a reliable basis for deep CBM reservoir evaluation and target optimization.