An artificial intelligence modeling framework based on microbial community structure prediction enhances the pollutant removal efficiency of the algae-bacteria granular sludge system

Algae-bacteria granular sludge (ABGS) technology is a new energy-saving and low-carbon water treatment technology based on the algae-bacteria symbiotic system. However, due to its complex internal microbial system, the regulation mechanism of ABGS is unclear. To address this issue, the present study constructed a two-stage optimal control model for the ABGS system, which includes prediction of microbial community structure and planning of pollutant removal efficiency. This model enabled intelligent optimization of the system's pollutant removal efficiency through the regulation of operational parameters. In the first stage, seven machine learning (ML) algorithms were compared to predict the succession process of microbial community structure under the different conditions (R² > 0.94). In the second stage, six ML algorithms were compared to predict the pollutant removal efficiency of the ABGS system, combining regulatory indicators and microbial community structure (R² > 0.94). Finally, the non-dominated sorting genetic algorithm was used to integrate the prediction models of the two stages, and the microbial community structure was selectively shaped to enhance the removal efficiency of any two of the carbon, nitrogen, and phosphorus pollutants in the ABGS system (removal rate >90 %). The results of this study provided a universally applicable and quantitative intelligent guidance model for the performance optimization of ABGS technology and other biological systems.