This document presents research on improving power grid management through reinforcement learning. The authors introduce a model-free approach using a masked action space that allows agents to learn optimal control strategies without requiring predefined expert knowledge. Different grid state representations are explored, including novel graph-based methods, and evaluated on their effectiveness in reducing power losses and enhancing grid stability. Experimental results using a simulated environment demonstrate the superior performance of their approach, particularly when incorporating graph observations and training against an opponent to improve robustness.
Podden och tillhörande omslagsbild pÄ den hÀr sidan tillhör
Neural Intelligence Network. InnehÄllet i podden Àr skapat av Neural Intelligence Network och inte av,
eller tillsammans med, Poddtoppen.