Enhancing energy circularity and achieving carbon neutrality in water resource recovery facilities through an...

This study addresses a significant gap in applying artificial intelligence to optimize biogas production in real-world settings, emphasizing the necessity for minimal daily changes to sensitive processes for effective implementation. The comprehensive approach integrated multiple AI models, specifically Long Short-Term Memory (LSTM) networks and Extreme Gradient Boosting (XGBoost), alongside Particle Swarm Optimization (PSO) algorithm. This study also evaluates the impact of optimization on energy circularity, thereby advancing energy sustainability within wastewater treatment processes. The methodology began with data preprocessing, utilizing online sensor data and experimental measurements that are often available after extensive analysis. To overcome this limitation, LSTM was employed as a soft sensor to predict key experimental values daily, facilitating real-time control over time-intensive parameters in anaerobic digestion. Once LSTM predictions were combined with online data, an enriched dataset was created for optimization. The PSO-XGBoost model was then developed to identify optimal settings for operational parameters including temperature, pH, alkalinity, and sludge flowrate. Post-optimization smoothing process was applied for optimized parameters to maintain consistency and prevent drastic daily fluctuations in real-world operations. Results demonstrated a significant increase in biogas yield, along with a 156 % reduction in net CO₂ emissions. The optimized process improved energy circularity index from 0.71 to 1.15, making it energy self-sufficient. This study demonstrates how AI-driven optimization can transform wastewater treatment into sustainable, energy-efficient facilities.