State Selection and Cost Estimation for Deep Reinforcement Learning-Based Real-Time Control of Urban Drainage System

In recent years, a real-time control method based on deep reinforcement learning (DRL) has been developed for urban combined sewer overflow (CSO) and flooding mitigation is more advantageous than traditional methods in the context of drainage systems (UDSs). Since current studies mainly focus analyzing feasibility DRL comparing them with methods, there still need to optimize design cost methods. this study, state selection estimation are employed analyze influence different states performance provide relevant suggestions practical applications. A real-world UDS used as an example develop models states. Their effect data monitoring costs then compared. According results, training process difficult when using fewer nodes information or water level input state. Using both upstream downstream improves DRL. Also, effective nodes; flow likely have better level, while cannot significantly further improve effect. Because higher monitoring, number use flow/water be balanced cost-effectiveness.