Answers are grounded in this paper and cite their sources.
Coal quality assessment is fundamental to strategic resource utilization and the advancement of cleaner coal technologies. Among coal maturity indicators, vitrinite reflectance (VR) is widely recognized as a critical parameter for determining coal rank and reconstructing its thermal history. VR represents the percentage of incident light reflected from vitrinite macerals under microscopic examination, with higher reflectance values indicating greater thermal maturity. This parameter directly influences coal suitability for energy production and its hydrocarbon generation potential in petroleum systems. Despite its importance, accurate VR determination remains challenging due to stringent requirements for sample preparation, expert petrographic analysis, and reliable discrimination of vitrinite from other macerals, particularly in low-rank coals. Additional uncertainties arise from variations in organic composition and instrument calibration. To address these limitations, this study develops a robust predictive framework for VR estimation using a hybrid machine learning approach that integrates random forest (RF) with an enhanced gray wolf optimization (GWO) algorithm. To improve the exploration and convergence performance of GWO, five advanced population initialization strategies were incorporated: Latin hypercube sampling (LHS), Sobol sequence sampling, Boundary filling (BF), Chaos mapping (CM), and Opposition-based learning (OBL). A dataset comprising 99 petrographic blocks with quantified maceral compositions and measured VR values was used for model development and validation. Model performance was assessed using statistical indicators and non-parametric testing. The proposed BFGWO-RF model achieved superior predictive performance, with an R2 of 0.9777 and an RMSE of 0.0643. Furthermore, Shapley additive explanations (SHAP) and individual conditional expectation (ICE) analyses enhanced model interpretability, identifying liptinite content as the most influential variable governing VR prediction. The findings demonstrate the significant potential of hybrid artificial intelligence and advanced optimization strategies for reliable coal maturity prediction, contributing to improved coal quality evaluation and supporting cleaner and more efficient geoenergy resource utilization.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: