LAUSR

Macroalgae‐derived hydrogen production is a promising strategy for developing a sustainable energy era. The literature statistical analysis showed that dark fermentation (DF) fed with brown macroalgae species is the most studied due to its operational flexibility, low energy requirement, and high efficiency. Some physics‐based models have been developed to predict this process, but they are limited to a few specific operating conditions. Therefore, this study applied machine learning (ML) techniques based on the Scopus database to predict and optimize the brown macroalgae‐fed DF performance, which depends on substrate properties (e.g., contents of carbohydrates, proteins, lipids, etc.) and operating conditions (e.g., temperature, pH, inoculum amount, etc.). Among 22 proposed models, the artificial neural network (ANN) model was identified as the best model, with a high coefficient of determination value of 0.977 and a low root mean square error (RMSE) value of 5.1. Shapley additive explanations (SHAPs), partial dependence plots (PDPs), and individual conditional expectation (ICE) analyses revealed that the pretreatment application, substrate concentration, and operating time are the most influential factors. Finally, the optimal conditions for the brown macroalgae‐fed DF process were detected by implementing particle swarm optimization (PSO). The optimization results indicated that the maximum hydrogen yield could reach 213.6 mL/g‐VS when Saccharina japonica was used as the substrate, representing an increase of ~20% compared to the highest reported value. However, the optimal conditions vary depending on the performance target (i.e., hydrogen yield or production rate) and feedstock type (i.e., raw or pretreated), suggesting that energy efficiency and economics models need further investigation.