Researchers have developed a new AI system for controlling the frequency of microgrids. Microgrids are small-scale power systems that can provide resilience and renewable energy integration. The proposed method, called Prompt-DT, uses few-shot learning to adapt to different microgrid configurations without requiring explicit system parameters. It also incorporates self-supervised contrastive learning to improve environment recognition and prompt utilization efficiency. A physics-informed prompt design technique is used to filter high-quality physical guidance during online execution. The system can generalize well in unseen environments with limited data, thanks to a lightweight finetuning approach. --- Why it matters: This matters because microgrids are becoming increasingly important for next-generation power systems, and reliable frequency control is crucial for their stability. This work provides a promising solution for addressing the challenges of generalization and adaptability in microgrid frequency control. Source: https://arxiv.org/abs/2608.21858