A multi-objective optimization framework for sustainable rural wastewater treatment and agricultural reuse

Rural wastewater management stands at the nexus of public health, environmental stewardship, and climate resilience. Addressing the growing demand for sustainable sanitation in decentralized contexts, this study presents a multi-objective optimization framework that integrates classical process modeling with machine learning to enhance the performance of bio-ecological treatment systems. By coupling the Activated Sludge Model with a fully connected neural network, our integrated approach provides near real-time decision support for meeting stringent discharge standards and agricultural reuse requirements. The framework demonstrates strong predictive capability, achieving coefficients of determination (R²) of 0.860, 0.854, 0.879, and 0.871 for COD_Cr, NH₄⁺-N, TN, and TP, respectively, on an independent validation dataset. The Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) is incorporated to guide adaptive aeration strategies. Results indicate an energy reduction of 1.19 kWh/day under high-standard discharge mode and a nutrient retention benefit of 22.5 g/day of ammonia nitrogen under irrigation reuse mode, thereby supporting circular nutrient flows. This work offers a scalable and resilient pathway toward low-carbon, resource-recovering sanitation in rural areas by combining data-driven control with ecological design, contributing to sustainable development and climate adaptation goals.