Dam Operation by Deep Reinforcement Learning Based on Operational Rules for Special Disaster Prevention Operations and Regular Operation

Key Points

• This research aims to enhance dam operation efficiency during floods by employing AI models based on operational rules.
• Developed a deep reinforcement learning model for dam operation during special disaster events.
• The model determines gate discharge based on historical and predicted dam metrics and downstream river levels.
• Incorporated regression equations to forecast the impact of discharge on downstream river levels.
• Applied the model to Haji Dam and Okawa Dam for comparison with existing operational rules and performance.
• The AI model demonstrated improved decision-making for gate discharges during special disaster scenarios.
• Predicted downstream river levels were effectively integrated into the operational process.
• The model's application presented a viable approach for modernizing dam operations in response to floods.