Energy Digest

A new framework is established to capture interaction sensitivity in power systems under perturbations, providing analytic expressions and physically interpretable mechanisms to explain stability relationships. This framework addresses the limitations of traditional participation factor (PF) analyses, which are structurally unable to analyze event-dependent mode-state interactions under perturbations. The framework offers a general analytical basis for understanding perturbation-driven interaction reconfigurations in complex power systems.

Dynamic operating envelopes (DOEs) for low-voltage distribution networks aim to reduce unnecessary curtailment of generation and load by computing flexible power exchange limits that trade off optimality with flexibility. The proposed approach sets both upper and lower limits on active and reactive power exchange, reducing curtailment and operational costs while maintaining voltage magnitudes within desired limits. Flexible DOEs outperform non-flexible approaches in this context, demonstrating superior performance in the Australian low-voltage distribution network validation study.

The article analyzes the transient synchronization stability (TSS) of a parallel system consisting of synchronous generators (SG) and doubly fed induction generators (SG-DFIG), focusing on the effects of complete low-voltage ride-through processes. A unified generalized swing equation is derived to describe the system's behavior, allowing for easy evaluation of TSS under various conditions. The analysis is supported by experiments and simulations, providing a clearer understanding of the TSS mechanism in hybrid systems.

VeraGrid-Agent is a tool-augmented LLM that autonomously solves complex power flow problems by executing an open-source solver and providing accurate answers. The model achieves an accuracy of 97.3% to 100.0%, significantly outperforming results without simulator access, which range from 42.7% to 49.3%. The few remaining errors are attributed to incorrect interpretations during multi-step reasoning, rather than simulator execution failures.

An event-driven neuromorphic framework for energy-efficient open-circuit fault diagnosis in three-phase inverters has been proposed, achieving a significant reduction in inference energy while maintaining 100% diagnostic accuracy. The method achieves this by exploiting the sparse structure of trajectory matrices to activate computation only in informative regions. A prototype implementation using Loihi-based neuromorphic energy estimation reduces inference energy by 382 times compared to a GPU-based CNN, from 11 microjoules to just 290 nanojoules per diagnosis.

Current-reference generation based on lookup tables (LUTs) is proposed for star-connected symmetrical six-phase nonsalient PMSMs, enabling unbalanced nonsinusoidal currents. A lexicographic optimization minimizes torque ripple and then copper loss, subject to various constraints. Simulations show a 77% reduction in peak-to-peak torque ripple with cogging-torque compensation.

Multimodal large language models (LLMs) can provide accurate answers for grid diagnosis, but there is a lack of evidence that they use task-appropriate information. A proposed framework evaluates self-reported reliance, intervention-derived behavioral reliance, and engineering importance to detect discrepancies in task-conditional faithfulness, which can lead to improved grounding without compromising performance. The framework's results validate its ability to detect, diagnose, and correct faithfulness failures across three LLMs on different scaled scenarios.

An integrated electric-hydrogen-transport system (EHTS) is modeled using a hierarchical optimization framework that combines vehicle scheduling and energy dispatch to minimize operational costs. A greedy heuristic algorithm is used for real-time vehicle charging and refueling, while a deep reinforcement learning approach optimizes battery operation, hydrogen tank operation, and PV generation allocation. The proposed framework demonstrates effectiveness in various transport-demand scenarios without retraining.