Data-Driven Model Development for Prediction and Optimization of Biomass Yield of Microalgae-Based Wastewater Treatment

Microalgae-based nutrient recovery has the potential to efficiently recover nutrients while simultaneously treating wastewater. However, the absence of an optimization model for this technology hinders its full potential. This study has developed a model to optimize the biomass yield in micro algae-based wastewater treatment system. Seven machine learning models, including Decision Trees (DT), Random Forest (RF), K-Nearest Neighbours (KNN), Gradient Boosting Regressor (GBR), Multi-Layer Perceptron Regression (MLPR), Support Vector Regression (SVR), and Artificial Neural Networks (ANN), were compared. Among other algorithms, ANN performed superiorly, achieving an R2 value of 0.98 with the lowest error. Furthermore, the optimal biomass yield of 948 mg/L (31.7% higher) was obtained when the COD, phosphate, nitrate, nitrite, pH, and retention times were maintained at 350 mg/L, 50 mg/L, 60 mg/L, 140 mg/L, 7.1, 9 days respectively. The pH and Retention time were found to be critical factors for the prediction of biomass yield. 20% of variation in the train test split ratio caused a 21% increase in the error value and the 75:25 ratio was found to be optimal for better performance of the model. This study serves as a valuable reference point for efficient AI - integrated algae-based wastewater treatment.