Geology-Aware Multiscale Fractal Coupling Graphs for Cross-Well Heterogeneity Characterization of Coal-Measure Well Logs in the Eastern Ordos Basin

Closely spaced lithologic contacts create scale-dependent variations in coal-measure well logs, while incomplete logging suites complicate cross-well comparison. A geology-aware multiscale stability-regularized fractal coupling graph (GeoMSR-FCG) links local analysis support to stratigraphic boundaries and represents cross-log coupling at explicit physical scales. A multilog gradient field regulates Higuchi fractal dimension (HFD) support; detrended cross-correlation analysis (DCCA) defines graph edges at 3, 6, and 12 m. Separate roof and floor contrasts capture stratigraphic asymmetry, and projection onto measured target edges enables transfer between logging suites. Evaluation combines 960 held-out synthetic realizations with a five-well proof of concept comprising 35 coal intervals in the eastern Ordos Basin. For a 25-sample embedded layer, boundary regulation reduces HFD root-mean-square error from 0.1931 to 0.1868; overall error is 0.1828, compared with 0.1823 for regularization without the boundary term. In the field wells, median effective weighted support decreases from 5.42–5.77 m to 5.28–5.49 m. At 3 m, well-balanced coal–host coupling-magnitude contrasts are −0.652 for gamma ray–deep resistivity and −0.160 for acoustic slowness–deep resistivity, while density–neutron contrasts are positive in the three wells with both logs. Transfer to the reduced Jin2 and Gao2 suites yields ROC-AUC values of 0.762 and 0.781. GeoMSR-FCG characterizes coal-measure heterogeneity through boundary-adaptive fractal support and scale-specific coupling among the measured logs.