Experiments found that under identical weather patterns, zero-shot LLM agents can achieve management outcomes comparable to reinforcement learning agents.