Energy Digest

Small-signal stability analysis of power systems dominated by power electronics converters faces scalability challenges due to system size increase. A new scalable stability assessment method, based on frequency-domain quadratic constraints, provides quantitative stability indices for multi-converter systems and is shown to clarify requirements for existing conditions. The proposed method reduces conservatism and enables efficient device-level diagnosis with hierarchical screening-diagnosis procedure.

A novel neural network framework called Hard-Constrained Probabilistic Factor Graph Neural Network (HCP-PINN) is proposed for distribution system state estimation in the presence of non-Gaussian uncertainty. The approach combines physical principles with probabilistic constraints to produce physically feasible state estimates, offering improved accuracy and robustness compared to existing methods. HCP-PINN delivers more accurate and consistent results than deep learning benchmarks and numerical estimators, while exhibiting greater computational efficiency under realistic measurement scenarios.

The achievable inertia constant of inverter-based resources is bounded by energy reserve, converter power rating with voltage ride-through obligations, and inertia-emulation control scheme activation delay. A unified envelope function combines these bounds into a single value that quantifies the plant's ability to contribute inertia at given loading, grid voltage, and time. The analysis shows that de-loading yields a linear, quantifiable inertia gain and collocated energy storage removes source-coupling constraints, making grid-following plants with storage equivalent post-activation.

The rate of change of frequency (RoCoF) is measured over tens to hundreds of milliseconds, and the inferred inertia depends on the window size and response delivered during that time. The system inertia is estimated using a closed-form prediction that applies to linear response models, which can inflate the estimate with fast frequency responses. This allows operators to compute their measurement reports and understand the impact of different system configurations on inertia estimates.

This paper proposes an adaptive model reduction approach based on nonlinear participation factors to accelerate time-domain simulation in power systems. The method enables a dynamic transition between full-order and reduced models, using modal energies to rank system modes under contingencies. It achieves significant simulation speedup while preserving accuracy, particularly when non-linear behaviors are present or near-resonance conditions occur.

A general minimum-order functional observer framework has been developed that encompasses both Darouach and Luenberger constructions as special cases. The required functional dimension is determined by a set of observability indices from the triple (A,C,L), and necessary-and-sufficient spectral feasibility conditions ensure this dimension is achievable. The framework enables minimum-order functional observers of intermediate order, even when neither classical observer is available.

Grid-forming (GFM) Enhanced STATCOMs (E-STATCOMs) are proposed as a solution for stable integration of large-scale AI data centers into weak or low-inertia networks, providing fast, well-damped responses to abrupt load changes. E-STATCOMs mitigate low-frequency oscillations and interactions with network resonances through tunable dynamic behavior, enabling precise shaping of system response and improved regulation at the point of connection. The proposed approach is scalable and flexible, supported by a design-oriented framework that facilitates parameter selection and performance assessment without requiring extensive EMT studies.

Doubly-Spectral Stochastic (DSS) expansion models uncertain node embeddings as random graph signals to capture structural variation and latent stochastic variation. DSS-GNN achieves the lowest Brier score without post-hoc correction on 14 node classification benchmarks and performs well in deployment modes, including standalone and hybrid modes. The model's robustness is established empirically across multiple tasks and evaluation settings.

An integrated modeling framework using nonlinear model predictive control optimizes thermal management, workload scheduling, and battery operation in a wave-powered subsea data center. The framework maintains thermal safety while adapting to changes in workloads, battery states, and renewable energy availability. Simulation results show that increasing battery capacity and wave generation capacity can improve flexibility, but excessive renewable energy may lead to reduced efficiency.