Record data of the training process of multi-agent reinforcement learning algorithm for a large-scale ground photovoltaic station

(1) Data 2-4: Reinforcement Learning Training Process Data (MAPPO/SAC/MADDPG) This section contains the training process records of the MAPPO algorithm proposed in this article and the comparative algorithms SAC and MADDPG in the same experimental environment, for the purpose of reproducing algorithm performance and conducting fair comparisons. Field Explanation: Step: Index of Time Steps within a Round Value: Single step instant reward value (2) Data 5: Loss Function Data (MAPPO Loss) This data contains round by round records of the loss of the policy network (Actor) during the training process of the MAPPO algorithm, reflecting the dynamic process of model parameter optimization and policy updates. Field Explanation: Step: Index of Time Steps within a Round Value: Policy gradient loss of policy network