Learning-driven multi-timescale operation simulation and hierarchical boundary optimization for renewable-dominated...

Area of Science:

• Power Systems Engineering
• Artificial Intelligence in Energy
• Optimization Theory

Background:

• Conventional power system scheduling struggles with high renewable energy penetration and uncertainty.
• Decoupled frameworks lack operational consistency and adaptability across multiple time scales.

Purpose of the Study:

• To develop a unified framework for multi-timescale power system operation under uncertainty.
• To enhance coordination between long-term planning and real-time control.

Main Methods:

• Integration of deep sequence forecasting, reinforcement learning, and rolling model predictive control.
• A hierarchical boundary optimization framework for dynamic boundary construction and updates.
• Simulation on a renewable-dominated IEEE 118-bus system.

Main Results:

• Reduced total operating cost by ~9.6%.
• Increased system reliability by ~15.8%.
• Lowered renewable curtailment from 9.5% to 5.7%, with boundary deviations below 2%.

Conclusions:

• The proposed framework significantly improves economic efficiency and operational robustness.
• Establishes a new paradigm for integrating learning and optimization in multi-timescale energy management.
• Demonstrates effective adaptive boundary learning for enhanced system performance.