Research Index

Applying large language model memory architectures to enhance power grid management and renewable energy integration. The approach combines HippoRAG's associative memory techniques with non-linear control systems to create an adaptive, predictive grid management system that can handle complex multi-energy scenarios while optimizing for efficiency and stability.

A novel framework combining bio-inspired multi-agent systems with deep reinforcement learning for optimizing EV charging networks. The system uses MAS-GPT for agent coordination and implements nature-inspired optimization for both energy distribution and charging station placement, while considering real-time grid constraints and user behavior patterns.

A novel framework that applies LLM reasoning calibration techniques to optimize microgrid control decisions in real-time. By combining SEAL's thought-steering approach with power systems domain knowledge, the system can generate more efficient and reliable control strategies while reducing computational overhead.

A novel approach combining test-time reinforcement learning (TTRL) with microgrid energy management to create self-optimizing power systems. The system continuously learns from operational data without requiring explicit labels, enabling real-time adaptation to changing conditions while maintaining grid stability and optimizing energy usage patterns.