Optimization of an Artificial Neural Network for Forecasting the Operating Conditions of CO2 Fixation Rate By Chlorella Genus Algae

Le résumé fourni par la source

Algae-based CO2 capture technology, specifically from the Chlorella genus algae, has emerged as a highly effective and feasible natural system for CO₂ capture with minimal environmental consequences. However, optimizing the algae’s cultivation conditions, such as temperature, initial pH, and inlet CO₂ concentration required for an ideal CO2 fixation rate, remains challenging due to the parameters’ complex and nonlinear interdependencies. Artificial neural networks (ANNs), combined with optimization tools such as the genetic algorithm (GA) and the Bayesian optimization algorithm (BOA), are widely used to simulate biological processes. In this study, an ANN model combined with optimization tools (GA or BOA) is employed to predict algal CO₂ fixation rates using temperature, initial CO₂ concentration, and initial pH as inputs. A total dataset of 46 experimental observations from peer-reviewed studies (2010–2025) was used to train the model. The models generated from this study, ANN standalone, ANN optimized with GA (ANN-GA), and ANN optimized with BOA (ANN-BOA), were compared in terms of the fitness performance (root mean squared error (RMSE), Mean Absolute Error (MAE), and correlation coefficient (R²)). It is found that the ANN optimized with GA has the best fitness performance across the overall datasets (R2 = 0.8495, RMSE = 0.1384, MAE = 0.1038). The ANN-GA model also identified the best operating conditions for a high CO2 fixation rate (pH = 7.0-8.5, temperature = 30-35°C, CO2 inlet concentration = 5-10 %). The findings from this study demonstrate moderate predictive performance under controlled conditions, providing proof-of-concept support for data-driven optimization and a foundation for future work before scaling up.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.