Kinetic model derived from machine learning for accurate prediction of microalgal hydrogen production via conversion from low thermally pre-treated palm kernel expeller waste.

The depletion of fossil fuel sources and increase in energy demands have increased the need for a sustainable alternative energy source. The ability to produce hydrogen from microalgae is generating a lot of attention in both academia and industry. Due to complex production procedures, the commercial production of microalgal biohydrogen is not yet practical. Developing the most optimum microalgal hydrogen production process is also very laborious and expensive as proven from the experimental measurement. Therefore, this research project intended to analyse the random time series dataset collected during microalgal hydrogen productions while using various low thermally pre-treated palm kernel expeller (PKE) waste via machine learning (ML) approach. The analysis of collected dataset allowed the derivation of an enhanced kinetic model based on the Gompertz model amidst the dark fermentative hydrogen production that integrated thermal pre-treatment duration as a function within the model. The optimum microalgal hydrogen production attained with the enhanced kinetic model was 387.1 mL/g microalgae after 6 days with 1 h thermally pre-treated PKE waste at 90 °C. The enhanced model also had better accuracy (R 2 = 0.9556) and net energy ratio (NER) value (0.71) than previous studies. Finally, the NER could be further improved to 0.91 when the microalgal culture was reused, heralding the potential application of ML in optimizing the microalgal hydrogen production process. [Display omitted] • Thermal pre-treated palm kernel expeller as a substrate for dark fermentation. • Machine learning-based kinetic model in predicting microalgal hydrogen production. • Integrated Gompertz model with thermal pre-treated function for optimization study. • Energy feasibility to produce green hydrogen from thermal pre-treated substrate. [ABSTRACT FROM AUTHOR]