Abstract
<title>Abstract</title> <p>The traditional extraction of humic acid (HA) from lignite is severely bottlenecked by sluggish mass transfer, excessive chemical consumption, and the inevitable structural degradation of macromolecules under harsh oxidative conditions. This study coupled with precision fluidic control and an XGBoost-SHAP machine learning framework, which achieved a 75% reduction in extraction time while optimizing the highly non-linear system. Kinetic and thermodynamic evaluations demonstrated that the dynamic fluidic regulation successfully shattered internal diffusion resistance, significantly lowering the apparent activation energy (Ea) to 46.8 kJ/mol and achieving a peak HA yield of 68.7% within just 30 min (a 75% reduction in total extraction time). The XGBoost model exhibited exceptional predictive fidelity (R2 = 0.983), while the SHAP global sensitivity analysis quantitatively revealed that while reaction temperature and pH are globally dominant, the activator drip rate is a critical operational parameter preventing the irreversible oxidative cleavage of target macromolecules. Furthermore, the high-purity HA was applied as an exogenous biostimulant for Scenedesmus obliquus to treat raw swine wastewater. At the optimal dosage of 30 mg/L, the structurally intact HA effectively buffered severe environmental stress, successfully preserving the microalgal photosystems (with chlorophyll-a content increasing to 20.8 mg/L) and boosting the COD, TN, and TP removal efficiencies to 82.5%, 80.2%, and 78.5%, respectively. However, an excessive dosage (150 mg/L) triggered a significant light-shielding effect that severely hindered photosynthesis. This study establishes a novel linkage between extraction process control, humic acid structural integrity, and microalgal remediation performance, providing a scalable strategy for lignite valorization.</p>