Temporal decoupling constrains three-product trade-offs in hydrothermal carbonization: A data-driven framework for hydrochar, aqueous-phase nitrogen, and carbon dots.
Abstract
Hydrothermal carbonization (HTC) of wet biomass yields three value-added streams: solid hydrochar, nitrogen-rich process water (a precursor for liquid fertilizer), and fluorescent carbon dots. However, single-objective optimization of individual products wastes co-products and undermines circular bioeconomy principles. Here, we establish a data-driven gradient-boosted decision tree (GBDT) framework with SHapley Additive exPlanations (SHAP), trained on 1,062 literature-derived data points across diverse feedstocks, to quantify cross-product regulation mechanisms. The models achieve test R2 of 0.94, 0.93, and 0.91 for hydrochar yield, aqueous-phase total nitrogen (TN), and carbon dots quantum yield (QY), respectively. Global SHAP analysis reveals distinct control regimes: operational parameters dominate solid/liquid products (44-45% contribution), whereas feedstock chemistry governs carbon dots optical properties (>35% contribution). Critically, while temperature (180-220 ℃) and solid-to-liquid ratio (SLR, 14-22%) permit partial cross-product coordination, SHAP dependence analysis exposes a temporal decoupling-hydrochar formation favors residence times (RT) below approximately 1.4 h, whereas carbon dots maturation generally demands RT beyond approximately 8 h-which effectively precludes single-batch three-product maximization within the explored domain. We therefore propose a conceptual two-stage configuration: Stage I (short-cycle primary HTC at approximately 217-220 ℃, RT of ca. 1-1.5 h) for co-production of hydrochar and liquid fertilizer, and Stage II (extended secondary processing at 200-220 ℃ with RT approximately 8 h) for carbon dots synthesis from residual organics, pending dedicated experimental validation with liquid-phase precursors. An open-access web application enables real-time process design for non-specialists, reducing the technical barrier to data-driven optimization. This work provides theoretical foundations and practical tools for full-stream organic waste valorization.
Abstract as supplied by the indexing source. © the authors / publisher.