Aggregation and Disaggregation of Power Flexibility to Support Reinforcement Learning in Large-Scale Grid Management
Sharaf Alsharif
ACM SIGEnergy Energy Informatics Review, September Issue 3 2025
Efficient coordination of power flexibility is vital for the secure operationof modern power systems, especially in Active Distribution Grids (ADGs). As the number of Flexibility-Providing Units (FPUs) grows, managing flex-ibility becomes increasingly complex, particularly under time-dependentconstraints like battery state of charge. Reinforcement Learning (RL) offersa model-free solution for real-time grid control, but faces challenges such asactuator conflicts and performance degradation due to changing marginaldistributions. This work proposes a data-driven approach for aggregatingflexibility into non-conflicting interfaces suitable for RL decision-making. Se-lected set points are then disaggregated to minimize their long-term impact,enhancing the stability and effectiveness of RL-based control in large-scaleADGs.
09 / 2025
inproceedings
the 14th DACH+ Conference on Energy Informatics 2025
OFFIS Autoren
