Ye et al. (2026) Agent-based intelligent real-time control for pluvial flood mitigation at urban scale

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-03-29
• Authors: Chenlei Ye, Zongxue Xu, Kai Liu, Ming Wang, Xinyi Shu, Lei Yu
• DOI: 10.1016/j.jhydrol.2026.135412

Research Groups

• School of National Safety and Emergency Management, Beijing Normal University, Zhuhai, China
• School of National Safety and Emergency Management, Beijing Normal University, Beijing, China
• Joint International Research Laboratory of Catastrophe Simulation and Systemic Risk Governance, Beijing Normal University, Zhuhai, China
• College of Water Sciences, Beijing Normal University, Beijing, China

Short Summary

This study develops an Urban Flooding Control Model (UFCM) by integrating hydrological-hydrodynamic models with deep reinforcement learning (DRL) for real-time pluvial flood mitigation at the urban scale. Applied to Jinan, China, UFCM significantly reduced inundated areas compared to traditional methods, demonstrating efficient and accurate real-time decision-making for complex urban flood systems.

Objective

• To develop and validate an intelligent, real-time control model (UFCM) for urban pluvial flood mitigation at the urban scale, integrating physically-based hydrological-hydrodynamic models with deep reinforcement learning.
• To propose a FUSE scheme with a voting mechanism to enhance the performance and stability of DRL agents in flood control.

Study Configuration

• Spatial Scale: Urban scale (meso-scale), applied to Jinan, China.
• Temporal Scale: Real-time control, validated through multiple storm flood events, including 2-year and 50-year rainfall scenarios.

Methodology and Data

• Models used:
  • Hydrological-hydrodynamic coupled models (Physically-Based Model, UFCM-PBM)
  • Deep Reinforcement Learning (DRL) (Intelligent Control Model, UFCM-ICM)
  • Agent-based DRL
  • Urban Flooding Control Model (UFCM)
  • FUSE scheme with a voting mechanism (combining multiple DRL agents)
  • Heuristic algorithms (for model calibration)
• Data sources:
  • Real-time flooding state
  • Uncertain and unknown rainfall scenarios
  • Multi-functional subareas (for calibration)
  • Multiple storm flood events (for validation)

Main Results

• Heuristic algorithms effectively calibrated the pluvial flooding model across multi-functional subareas, yielding excellent simulation results for multiple flooding events.
• The UFCM, coupling physical mechanisms with DRL, provided efficient and accurate decision-making solutions for complex urban flood system scheduling and control.
• UFCM demonstrated significant improvements over traditional water level-based real-time control (RTC) strategies.
• Under the regulation of the FUSE agent, the inundated area was reduced by 56.31% (to 43.69% of the original scenario) under 2-year rainfall, compared to scenarios without water engineering.
• Under the regulation of the FUSE agent, the inundated area was reduced by 17.55% (to 82.45% of the original scenario) under 50-year rainfall, compared to scenarios without water engineering.

Contributions

• Novel integration of physically-based hydrological-hydrodynamic models with deep reinforcement learning for urban-scale pluvial flood control.
• Introduction of the Urban Flooding Control Model (UFCM), comprising a physically-based simulation module and an intelligent DRL-based control module.
• Development of the FUSE scheme with a voting mechanism, enhancing the stability and reliability of DRL-generated control strategies.
• Demonstration of superior performance in real-time flood mitigation and decision-making compared to traditional water level-based control strategies.
• Provides valuable insights for intelligent urban pluvial flood modeling and real-time control at the meso-scale.

Funding

Not specified in the provided text.

Citation

@article{Ye2026Agentbased,
  author = {Ye, Chenlei and Xu, Zongxue and Liu, Kai and Wang, Ming and Shu, Xinyi and Yu, Lei},
  title = {Agent-based intelligent real-time control for pluvial flood mitigation at urban scale},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.135412},
  url = {https://doi.org/10.1016/j.jhydrol.2026.135412}
}