EUBCE 2025

Here I presented our study that explored how artificial neural networks (ANN) and regression models can optimize the hydrogen and carbon content in biocrude produced from food waste using thermochemical liquefaction. A mix of bread waste, apple pomace, spent tea leaves, and coffee grounds, processed with ethanol at 270 °C was used. Two predictive models were developed: a regression model with a desirability function and an ANN paired with a multi-objective genetic algorithm. The ANN performed significantly better, with higher R² values  and lower errors. Each model suggested different optimal feedstock blends; regression favored more bread waste, while the ANN emphasized coffee grounds. Both produced high-quality biocrude, though yields were relatively low, indicating a trade-off between energy content and quantity. This research highlighted ANN's strength in modeling nonlinear systems and optimizing biocrude quality. Watch out for conference proceedings here