Conference Agenda

Past studies have demonstrated that optimizing the control of the integrated urban drainage system can enhance system performance instead of incurring the high costs of updating expensive facilities. This study focuses on the drainage system management of the Wutong Mountain River Basin in Longgang District, Shenzhen. For the drainage network, wastewater treatment plant, and river water systems in the study area, SWMM, ASM, and Delft3D models were constructed. By coupling these models, control objectives were set with two sub-objectives: environmental and economic. The environmental sub-objective includes considerations for node pressure, wastewater plant water quality, and river water quality, while the economic sub-objective includes factors such as reagent dosage, aeration blower operation time, pump operation time, and operational frequency. The study aims to optimize the operation of the integrated urban drainage system. To address the low operational efficiency of mechanistic models, a deep reinforcement learning model was developed to improve computational speed, enabling real-time control. The expected results, compared to existing strategies, suggest a 5% reduction in economic costs while achieving a 10% reduction in river pollution. Moreover, the computational time, compared to mechanistic models, is reduced from 30 minutes to under 10 seconds.