Abstract: This paper proposes a novel deep reinforcement learning approach, named GARL (Generative Adversarial Reinforcement Learning), for intra-day optimization dispatch in high-proportion new ...