Thermal Response Estimation of De-Oiled Fresh and Marine Microalgae Based on Pyrolysis Kinetic Studies and Deep Neural Network Modeling.

As a potential alternative to fossil fuel, biofuel production from microalgae pyrolysis is a promising renewable energy resource. In this aspect, systematic investigation of thermal behavior and kinetic analysis is crucial to select suitable microalgae as a pyrolysis feedstock. The present study used model-fitted Coats Redfern (CR) and model-free distributed activation energy model (DAEM) to screen suitable de-oiled microalgae biomass as pyrolysis feedstock. Thermogravimetric data analysis of eight different CR models, based on three different reaction mechanisms, confirmed that slow pyrolysis of de-oiled microalgae biomass is governed by heat and mass diffusion mechanism. According to DAEM approach, apparent activation energy of Chlorella pyrenoidosa, Chlorella minutissima, Chlorella protothecoides, Chlorella vulgaris, and Dunaliella sp. is 55.87 ± 11.16, 56.09 ± 6.32, 46.58 ± 5.55, 55.26 ± 13.14, and 68.09 ± 10.62 kJ/mol, respectively, which is similar to CR approach. The thermodynamic parameters such as ΔH, ΔG, and ΔS of studied microalgaes are estimated in the range of 41.23–62.74 kJ/mol, 177.87–197.73 kJ/mol, and 0.19–0.22 J/mol•K, respectively. This study used a one-dimensional convolutional neural network (Conv1D) and long short-term memory (LSTM)-based Conv1D-LSTM model to predict microalgal pyrolysis data. The best deep neural network model (DNN6) showed minimum MSE (10−6) and high regression coefficient (R2 > 0.997) for 10, 20, and 30 °C/min heating rates. The proximate and ultimate results were statistically analyzed using Spearman's rank correlation and one-way analysis of variance (ANOVA). These research findings can be referenced for systematic screening of microalgae as pyrolysis feedstock and encourages artificial intelligence (AI) application in microalgae pyrolysis studies. [ABSTRACT FROM AUTHOR]