[3S05Aa] 21. Open Topics in Biotechnology - Part 1

3:00 PM - 3:15 PM JST(6:00 AM - 6:15 AM UTC)

[3Open-06] Dark fermentation of Clostridium sp. supplemented with Chlorella vulgaris FSP-E microalgae hydrolysate for biohydrogen production: Experimental and AI prediction modelling studies

○Adityas Agung Ramandani¹, John Chi-Wei Lan², Kuan Shiong Khoo¹ (1. Department of Chemical Engineering and Materials Science, Yuan Ze University, Taoyuan, Taiwan (Taiwan), 2. Biorefinery and Bioprocess Engineering Laboratory, Department of Chemical Engineering and Materials Science, Yuan Ze University, Taoyuan, Taiwan (Taiwan))

Keywords:

Chlorella vulgaris FSP-E,Biohydrogen production,Dark fermentation,Artificial Intelligence,Machine Learning

Conversion of microalgae biomass into hydrolysate for Clostridium sp. for dark fermentation biohydrogen production has emerged as a promising approach for sustainable bioenergy production and effective food waste valorization. In our previous study, Chlorella vulgaris FSP-E microalgae was cultivated using food waste as a primary nutrient source for its biomass production. The microalgae biomass achieved a final concentration of 4.6 g/L, was subjected to hydrolysis process to produce fermentable hydrolysate enriched in organic compounds. The microalgae hydrolysate was then used for dark fermentation experiments that were conducted using Clostridium sp. as the hydrogen-producing bacterium under anaerobic conditions at a controlled temperature of 37°C to ensure optimal microbial activity. To enhance the analytical and predictive capability of hydrogen production performance, our preliminary research integrates Artificial Intelligence (AI)-based modeling techniques using Python programming. Unlike traditional kinetic models, AI-based modeling approaches offer improved flexibility in capturing complex, non-linear relationships between process variables and system responses. In this work, five machine learning algorithms, such as Random Forest (RF), Extreme Gradient Boosting (XGBoost), Artificial Neural Networks (ANN), k-Nearest Neighbors (kNN), and Support Vector Machine (SVM), were systematically developed and applied to model cumulative hydrogen production and related fermentation dynamics based on numerical experimental datasets. The input features for the models included key fermentation parameters such as substrate concentration, fermentation time, pH, temperature stability, and reducing sugar, while the outputs focused on H₂ yield, production rate, and CO₂ production. The models were trained, validated, and tested using appropriate data splitting strategies to ensure robustness and generalization. Performance evaluation was conducted using widely accepted statistical indicators, including the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE), allowing objective comparison among the applied algorithms. The results demonstrate that AI-based models successfully captured the sigmoidal trends and dynamic behavior of hydrogen production observed during dark fermentation. Among the tested models, ensemble learning approaches such as Random Forest and XGBoost exhibited superior predictive performance due to their ability to handle nonlinear interactions and reduce overfitting, while ANN showed strong capability in approximating complex biological systems.