Energy Digest

Large language models (LLMs) and agentic AI systems are being applied to smart grids for tasks like forecasting, optimization, and control, but current approaches lack a unified design principle. A new approach presents a solver-grounded design, where trusted tools compute results that are verified by an explicit verification process. This framework has been successfully applied in four case studies, achieving significant improvements over LLM-only baselines.

The increasing penetration of inverter-based resources has led to a reduction in system inertia, making frequency deviations more pronounced in low-inertia power systems. A new load model incorporating dynamic behavior improves the representation of load-frequency control (LFC) models, increasing the accuracy of frequency response studies. Static load modeling can lead to inaccurate results and potentially misleading conclusions.

A new framework is proposed for assessing extreme-weather impacts on distribution system resilience, integrating probabilistic event-track modeling, branch-level fault screening, and downstream impact propagation analysis to better understand weather-driven disruptions. The framework can distinguish between geographic exposure and topology-dependent operational impacts, providing a unified workflow for assessment under forecast uncertainty. It is demonstrated through case studies using the IEEE 33-bus distribution feeder and visualized on the CURENT Large-scale Testbed platform.

A phased development framework enables islanded operation of sustainable AI data centers with onsite grid-following and grid-forming energy architectures, addressing scalability challenges through modular construction and integrated energy systems. The approach aims to meet aggressive time-to-market objectives while navigating interconnection approvals and component equipment procurement lead times. Electromagnetic transient simulations demonstrate the reliability of on-site natural gas generation and grid-forming energy storage in supporting data center operations during early deployment phases.

A novel energy management system using proximal policy optimization (PPO) is proposed for hydrogen-enabled community microgrids, which can improve renewable energy utilization and local resilience. The system achieved a net annual operating revenue of A$195,690.67 under normal conditions, with high load satisfaction and low carbon intensity. Under increased grid-outage probability, the system retained significant revenue while increasing battery discharge and diesel generation to meet demand.

Multisine electrochemical impedance spectroscopy is a non-invasive technique that allows for high-fidelity measurements of Li-ion batteries over a wide frequency range. The method uses a sum of sines to reduce experiment time while providing comparable results to single-sine excitation. Multisine EIS measurements offer new insights into battery dynamics, material properties, charge transfer processes, and thermal performance under various operating conditions.

A digital twin-based method assesses local collective tariffs in distribution-level energy systems by integrating agent-based modeling of household consumption and generation with virtual aggregation and explicit representation of tariff logic. This approach demonstrates cost reductions through efficient allocation of tariff components, but also highlights sensitivity to temporal alignment of consumption and generation. The method supports systematic assessment of tariff mechanisms by capturing interactions between infrastructure, user behavior, and regulatory design.