Energy Digest

Busbar protection is crucial for industrial power system reliability but faces practical challenges due to CT placement, dynamic zone selection, and evolving fault scenarios. Different busbar protection schemes, including leakage-to-frame, high-impedance differential, interlocking overcurrent, electronic, and numerical, have varying operating principles and implementation requirements. A structured evaluation of these schemes highlights trade-offs and operational limitations in maintaining system reliability.

Grid-forming voltage source converters are proposed to address reduced conventional inertia in electric power systems due to the substitution of traditional synchronous generators with converter-interfaced renewable energy sources, which can impact system stability and resilience during transients. The dynamic response and operational limits of the primary energy source should be explicitly modeled and accounted for in DC-link control design to ensure reliable grid operation. A new methodology using a Genetic Algorithm is proposed to tune control parameters and enhance system resilience.

Transient-stability assessment in power systems relies on critical clearing time (CCT) estimation, which is challenging due to complex fault-clearing dynamics requiring repeated simulations. An event-structured physics-informed neural network (ES-PINN) has been proposed to estimate CCT by aligning its representation with pre-fault, fault-on, and post-clearing swing dynamics, enabling accurate boundary extraction and local sensitivity analysis. Experiments on various power systems demonstrate the effectiveness of ES-PINN in improving trajectory and stability-boundary accuracy over neural-surrogate baselines.

Cyber attacks on digital substations can manipulate circuit breaker operations, causing severe system impacts, by abusing IEC 61850 communication. A new method called Substation Cyber Attack Strategy Phasing (SubCASP) uses a Hidden Markov Model to fuse IDS data logs and infer the current attack phase and next attack phase. This approach is trained on a reproducible dataset and demonstrates robustness for various IDS observability levels and missing IDS data log scenarios.

This paper develops explicit dynamic models of a hyperscale data center and integrates them into a stability framework to investigate the impact of demand response on power system inter-area oscillations, showing that UPS-based demand response can enhance damping. The HVAC subsystem is found to be incapable of providing effective oscillation damping due to its limited thermal response bandwidth. A gradient-based optimization algorithm tunes the UPS controller gain to maximize damping ratio in critical modes.

HARGO, a heterogeneity-aware reward-guided optimization method, is proposed to address the challenges of optimizing large language models (LLMs) on high-performance computing (HPC) tasks. The method uses confidence-modulated advantage to compute per-response weights, achieving state-of-the-art performance across four HPC tasks and nine methods. HARGO outperforms uniform-weight reinforcement learning methods by up to 58x, demonstrating its effectiveness in handling heterogeneous HPC tasks.

A study investigated an adaptive demand-driven control strategy for a district heating system integrated with phase change materials (PCMs) to improve operational flexibility and techno-economic performance. The results showed that the adaptive demand-driven approach effectively smoothed heat demand profiles, achieved up to 5.3% peak-load reduction, and maintained thermal comfort. However, the payback period was 25.3 years, indicating a need for further cost reductions and supportive market incentives.