Thermochemical behavior and physics-informed neural network kinetic inversion of microalgal components for blast furnace slag waste heat valorization

Blast furnace slag (BFS) carries considerable high-grade sensible heat and reactive mineral phases, offering opportunities for thermochemical waste heat recovery coupled with biomass conversion. In this work, glycine (Gly) and medium-chain triglycerides (MCT) were selected as model compounds for the protein and lipid fractions of microalgae to investigate the thermal conversion behavior in BFS-assisted pyrolysis. Thermogravimetric analysis coupled with mass spectrometry was performed at different BFS blending ratios and heating rates. To bridge the gap between complex experimental data and underlying reaction mechanisms, an Enhanced Physics-Informed Neural Network ( E -PINN) was developed to reconstruct activation energy distributions with improved stability and accuracy. The results showed that BFS exerted distinct structure-dependent effects on the two substrates. For Gly, the slag lowered the average activation energy monotonically from 129.95 to 113.57 kJ⋅mol^-1 and shifted the energy distribution toward lower values, while the gas product distribution remained largely unchanged. In contrast, MCT exhibited a non-monotonic response. At 30 wt% loading, the average activation energy reached a maximum of 171.33 kJ⋅mol^-1 with a compressed reaction interval, accompanied by a marked shift in gas distribution from CO₂-dominated to CO-enriched. These converging observations suggest that BFS does not simply promote decomposition but instead redirects the reaction toward a higher-barrier pathway. The findings demonstrate that BFS can serve as both a heat carrier and a catalytic medium, with its role depending strongly on the molecular structure of the feedstock. This study provides a reliable tool for resolving complex thermal conversion kinetics and offers guidance for process design in thermochemical recovery of metallurgical waste heat.