Waste activated sludge (WAS) is a significant byproduct of municipal wastewater treatment plants (MWTPs), but its disposal poses a threat to the environment due to secondary pollution. However, a promising technology that converts WAS into volatile fatty acids (VFAs) could provide sustainable solutions. VFAs can be used in the synthesis of bio-degradable plastics and the removal of nutrients in MWTPs. While VFA fermentation is a complex process, researchers have found a potential solution using machine learning (ML) to optimize VFA production.
A study conducted by a team of researchers from Hangzhou Dianzi University explored the application of ML in predicting and optimizing VFA production from riboflavin-mediated sludge fermentation. The researchers tested three ML algorithms – Artificial Neural Network (ANN), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost). The team utilized input variables such as pH, temperature, fermentation time, soluble protein, total carbohydrates, reducing sugar, NH-N, and riboflavin dosage to predict VFA production.
The results showed that XGBoost exhibited the best prediction performance and excellent generalization ability, with a high testing coefficient of determination (R) of 0.93 and a low root mean square error (RMSE) of 0.070. The study also analyzed the importance and interaction of input features using the Shepley Additive Explanations (SHAP) method. pH and soluble protein were identified as the top two input features in the modeling process.
Furthermore, the optimization algorithms of genetic algorithm (GA) and particle swarm optimization (PSO) were employed to predict the maximum VFA output. The researchers found that the optimal solution for VFA output was a predicted maximum of 650 mg COD/g VSS. This data-driven approach not only improved the production efficiency of VFAs but also paved the way for sustainable waste activated sludge management.
By combining chemical treatment and machine learning, this research highlights the potential for efficient VFA production and opens up new possibilities for the management of waste activated sludge. The findings contribute to environmental protection and resource recovery, with significant implications for urban sewage treatment.
Overall, this study demonstrates the value of machine learning in optimizing the complex fermentation process of VFA production. The results provide insights into the interactions between various input variables and showcase the potential for using ML models in predicting and maximizing VFA output. As waste management and sustainability continue to be pressing global concerns, this research offers a promising strategy for unlocking sustainable solutions in the field of wastewater treatment.
