A review on biomass thermal-oxidative decomposition data and machine learning prediction of thermal analysis

Thermochemical conversion is the most economical approach to recovering energy and alternative fuels from biomass feedstock. This work first reviews the literature data on thermal-oxidative decomposition for common biomass types and forms a database of 18 parameters, including element, proximate, and thermogravimetric analysis (TGA). Then, an Artificial Neural Network (ANN) model is developed for the prediction of TGA data. Pearson correlation coefficient analysis reveals that the influence of environment heating rate on biomass thermal decomposition is larger than that of fuel properties. By inputting biomass elemental/proximate analysis and heating rate, the ANN model successfully predicts 8 key TGA parameters, namely, pyrolysis-onset temperature, peak pyrolysis temperature, oxidation-dominant temperature, peak oxidation temperature, oxidation-end temperature, peak pyrolysis rate, oxidation-dominant rate, and peak oxidation rate, with R² values greater than 0.98. A better performance can be achieved when all ten input features are considered. Final, an open-access online software, Intelligent Fuel Thermal Analysis (IFTA), is developed to predict thermal-oxidative decomposition across a wide range of heating rates and biomass types. This work provides a better understanding of biomass thermal-oxidative decomposition dynamics and a shortcut to obtain key parameters of biomass degradation without TGA tests.