A hybrid machine learning–based multi-objective supervisory control strategy of a full-scale wastewater treatment for cost-effective and sustainable operation under varying influent conditions

摘要: Abstract This study develops a multi-objective supervisory control (MOSC) strategy for wastewater treatment based on hybrid machine-learning algorithms that search optimal setpoints of multiple controllers under varying influent conditions. A plant (WWTP) operation was modeled by Benchmark Simulation Model No. 2 (BSM2), and conditions were generated in consideration H-WWTP South Korea. Two proportional-integral (PI) dissolved oxygen biogas, one cascade-PI controller nitrate used as local loops. The MOSC identified five scenarios using fuzzy c-means nitrogen-to-carbon ratios. Then, the performance according to changes gauged employing deep-learning-based approximation model, determined non-dominated sorting genetic algorithm. results demonstrate an intelligent can identify improve WWTP outperform reference across range possible ratios total Kjeldahl nitrogen chemical demand (TKN/COD) disturbances. also able accommodate extreme conditions, reduce operational costs 8%, maintain effluent quality, produce biogas sustainable operation.