Energy Digest

A novel transformer-based architecture is used to predict generator commitment schedules over a 72-hour horizon, integrating deterministic post-processing heuristics to enforce physical constraints and minimize excess capacity. This refined prediction pipeline serves as a warm start for a downstream Mixed-integer Linear Programming (MILP) solver, achieving 100% feasibility in single-bus test system validation. The approach drastically reduces computation times while, in some cases, leading to lower overall system cost than traditional solver methods.

Power systems are vulnerable to high-impact, low-probability events due to single-dimensional resilience frameworks that fail to capture cross-dimensional effects. A new Multidimensional Resilience Index (MDRI) framework, which accounts for five interacting dimensions, shows that multi-vector attacks produce system degradation exceeding linear expectations by a factor of 5.6 and amplify it further through exogenous factors. Simultaneous dimensional failures also contribute significantly to overall system degradation.

A generic effective nodal frequency (ENF) model is developed to analyze power systems with high penetration of renewables, featuring parameters that retain only the dominant constant component governing nodal frequency dynamics. The ENI parameter is found to be most influential in determining frequency security, and a critical nodal inertia is analytically derived for ensuring system stability. A system-level frequency security index is proposed, enabling an "offline calculation and online evaluation" approach for assessing system frequency security using lookup tables and interpolation methods.

A generalized dynamic phasor framework is proposed for analyzing inverter-based resources connected to multi-machine systems under balanced and unbalanced conditions, capturing subsynchronous oscillations. The framework enables eigen decomposition analysis for root-cause identification and damping controller design. It also allows for analysis of excitation of subsynchronous oscillations in presence of data center loads.

The growth of inverter-based resources and distributed energy resources has altered modern power system voltage stability characteristics, necessitating online estimation of long-term voltage stability margin (VSM) through machine learning. An explicit analytical VSM expression is derived from offline data using a physics-informed ML-trained model and embedded into the TSO optimization problem to enable real-time enforcement of minimum VSM constraints. The proposed framework successfully enhances operational efficiency by prioritizing the most influential reactive power resources.

Bayesian Optimal Experimental Design (BOED) selects experimental designs that maximize expected information gain through the Kullback-Leibler (KL) divergence, but this approach has limitations such as support mismatch, tail underestimation, and rare-event sensitivity. An alternative IPM-based BOED framework replaces KL divergence with integral probability metrics like Wasserstein distance and Maximum Mean Discrepancy, providing stronger stability guarantees and empirical validation of concentrated credible sets. This framework is further extended to geometry-aware discrepancies using a neural optimal transport estimator.

A scalable coordination framework for managing massive distributed energy resources (DERs) has been proposed, employing a two-layer architecture that combines Eulerian modeling with convexification techniques to address non-convexity issues. The framework protects aggregator-side privacy by utilizing a local data-mixing protocol and Wasserstein-based relaxation, enabling improved economic performance. Numerical studies demonstrate the scalability, feasibility, and effectiveness of the proposed approach.

A new method, EWMA-RTTA-Based Quadratic Extrapolation, is proposed to accurately model system frequency across asynchronous simulation environments in integrated T&D co-simulation. This method improves the prediction of voltage magnitude and phase angle variations within 10 ms transmission intervals, resulting in more accurate PLL-based frequency estimation in distribution systems. The approach reduces normalized mean absolute error by a factor of 25.7 compared to existing methods.