Zhang et al. (2025) Deriving reservoir operating rules of spillway gates based on deep reinforcement learning

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-07-15
• Authors: Aonan Zhang, Pan Liu, Cheng Qian, Lei Cheng, Weibo Liu, Yalian Zheng, Huan Xu, Wei Zhang, Dongyang Han, Hao Ye
• DOI: 10.1016/j.jhydrol.2025.133858

Research Groups

• State Key Laboratory of Water Resources Engineering and Management, Wuhan University
• Hubei Provincial Key Lab of Water System Science for Sponge City Construction, Wuhan University
• Research Institute for Water Security (RIWS), Wuhan University
• Bureau of Hydrology, Changjiang Water Resources Commission
• Science and Technology Research Institute, China Three Gorges Corporation

Short Summary

The study proposes a deep reinforcement learning approach to derive operating rules for spillway gates rather than general reservoir outflow, aiming to optimize flood control and reduce operational frequency.

Objective

• To derive spillway gate operating rules that simultaneously minimize peak outflow and the number of gate operations while adhering to practical constraints such as symmetrical openings and specific opening sequences.

Study Configuration

• Spatial Scale: Three Gorges Reservoir, China.
• Temporal Scale: Real-time operation during flood seasons.

Methodology and Data

• Models used: Deep Reinforcement Learning (DRL), specifically the Proximal Policy Optimization (PPO) algorithm.
• Data sources: Hydrological data from the Three Gorges Reservoir.

Main Results

• Spillway gate operation schemes can be effectively described by three variables: current inflow, current reservoir water level, and previous decisions.
• Compared to conventional operating rules, the proposed DRL-based method reduced peak outflow by an average of 3.54%.
• The number of spillway gate operations was reduced by an average of 62.34% compared to conventional rules.

Contributions

• Shifts the optimization focus from aggregate reservoir outflow to the actual status of spillway gates (fully open or closed), incorporating real-world operational constraints.
• Provides a more practical and directly applicable framework for real-time reservoir operation compared to traditional outflow-based rules.

Funding

• Not specified in the provided text.

Citation

@article{Zhang2025Deriving,
  author = {Zhang, Aonan and Liu, Pan and Qian, Cheng and Cheng, Lei and Liu, Weibo and Zheng, Yalian and Xu, Huan and Zhang, Wei and Han, Dongyang and Ye, Hao},
  title = {Deriving reservoir operating rules of spillway gates based on deep reinforcement learning},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.133858},
  url = {https://doi.org/10.1016/j.jhydrol.2025.133858}
}