Pyrolysis characteristics of microalgae and machine learning modelling for activation energy

Xiang Zhang, Sitao Zhu, Yijie Wang, Chen Hong,
Pyrolysis characteristics of microalgae and machine learning modelling for activation energy,
Journal of Environmental Chemical Engineering,
Volume 13, Issue 6,
2025,
119876,
ISSN 2213-3437,
https://doi.org/10.1016/j.jece.2025.119876.
(https://www.sciencedirect.com/science/article/pii/S2213343725045737)
Abstract: Microalgae biomass, as a representative third-generation bioenergy source, holds significant potential for sustainable fuel production. This study systematically investigated the pyrolysis characteristics and kinetic parameters of Chlorella vulgaris (CV) using thermogravimetric analysis coupled with Fourier-transform infrared spectroscopy (TG-FTIR), model compound decoupling, kinetic modeling, and machine learning (ML). Pyrolysis exhibited three distinct phases: dehydration, devolatilization, and residual decomposition. Increasing heating rates accelerated mass loss rates but did not affect final residue yields, confirming intrinsic compositional control. Model compounds (starch, glutamic acid, soybean oil) revealed synchronous decomposition mechanisms, with overlapping temperature ranges explaining DTG peak broadening. Kinetic analysis via Friedman, Flynn-Wall-Ozawa (FWO), and Kissinger-Akahira-Sunose (KAS) methods yielded an average activation energy of 198.5 kJ/mol for the main devolatilization phase. The reaction followed an order reaction model (O3), described by: dα/dt = 2.38(1-α)3exp(-2.39 ×104/T). The most probable mechanism function is the order model O3. To identify key factors influencing activation energy, an ML model was developed using 1003 datapoints from 51 microalgal species. Decision Tree (DT) outperformed BPNN (Backpropagation Neural Network) and SVR (Support Vector Regression) models, achieving exceptional accuracy (R2=0.998, RMSE=4.224). Feature importance analysis quantified C and H content as primary controllers of activation energy, underscoring organic composition’s critical role in reducing pyrolysis energy barriers. This work provides novel insights into microalgae pyrolysis mechanisms and establishes a robust ML framework for activation energy prediction.
Keywords: Microalgae; Model compound; Pyrolysis kinetics; Activation energy; Machine learning