Energy Digest

Stability issues in converter-dominated grids are influenced by static, small-signal, and large-signal phenomena, requiring a rethinking of conventional modeling and analysis tools. Data-driven approaches can provide insight, but raise questions about data completeness, model generalization, physical interpretability, and uncertainty awareness. Model-based and data-driven methodologies for stability analysis are discussed, examining their effectiveness in capturing the complex nature of instability phenomena.

Event-triggered data selection is proposed for remote frequency-domain identification in high-performance systems, reducing communication demands by transmitting only relevant measurements. The approach combines uncertainty information with model mismatch evidence to select measurements locally, preserving accuracy and responsiveness. This method can accommodate increasing sensing rates and subsystems, making it suitable for modern machines.

Optimal energy conversion in port-Hamiltonian systems is studied through analysis of state- and control-constrained finite-horizon optimal control problems that maximize energy conversion efficiencies. Asymptotic optimal efficiencies exist and coincide with optimal cyclic efficiencies under suitable reachability conditions, and a generalized Carnot-type efficiency bound is established. Intrinsic limits on achievable energy conversion are also derived for a class of port-Hamiltonian systems.

NOCE-Net, a deep nested sequence model, learns control-enabling representations directly from raw battery operational data, achieving 13.7% MAPE in voltage response prediction and outperforming strong baselines by over 43% with 40 times fewer parameters. The model's representations encode degradation dynamics transferable to downstream tasks, providing evidence for its generalisability. NOCE-Net operates on diverse operational regimes using a multi-patch transformer encoder and GRU backbone.

The generation interconnection queue introduces uncertainty to power grid congestion price estimates, and Monte Carlo sampling is a more effective alternative to conventional approaches for estimating congestion distribution. This method yields good estimates of congestion distribution even with a small number of samples, and includes correlation in the distribution model without introducing significant changes. The approach is applicable to both small-load and large-load interconnection scenarios.