Energy Digest

Synthetic power-grid scenarios can be improved by incorporating AC feasibility through a learning framework that integrates topological, branch electrical, and time-varying load profile constraints. The proposed framework enables generator-controlled diffusion sampling to produce operationally feasible grid scenarios without requiring post-generation validation or optimization. This approach significantly improves operational feasibility and contingency robustness while maintaining statistical fidelity.

A new power system stability criterion allows distinct subsets of converters to satisfy different local requirements at the same frequency, enabling a more efficient use of converter capabilities. The criterion partitions between gain and phase bounds based on network-dependent quadratic constraints, providing a technology-aware stability certificate that balances conservatism and practicality. This approach is demonstrated to outperform standard decentralized conditions in heterogeneous systems.

Local Flexibility Markets require DSOs to determine both quantity of flexibility and willingness to pay, which is often neglected in existing approaches. A new methodology proposes embedding DSO's willingness to pay into market-clearing objective by monetizing network aging, losses, and congestion, enabling more efficient clearing. This framework demonstrates higher market liquidity, efficient procurement, and improved network operations while preserving transparency and non-discrimination principles.

A new automated decision-making framework is proposed to enhance grid resilience against wildfires by accounting for decision-dependent uncertainty (DDU) through preventive and corrective measures. The framework incorporates a multistage optimization model that considers DDU, influencing probabilities and parameters of future wildfire scenarios. Simulation results demonstrate the effectiveness of the approach in providing more realistic and operationally resilient solutions.

A novel automated decision-support framework proposes a stochastic multi-stage programming approach to enhance power system resilience and operational resilience against evolving wildfires, considering preventive and corrective actions, and optimizing response strategies based on potential scenarios. The framework aims to minimize wildfire risk and operational costs while reducing load curtailment. A proposed algorithm and stochastic dual dynamic programming approach enable adaptive decisions and achieve a global optimum for the framework, demonstrating improved results over traditional single-stage optimization strategies.

A new hybrid quantum formulation is developed to address the NP-hard problem of controlled islanding in power systems, capturing essential decisions with a physics-informed compact encoding and combining quantum optimization with classical refinement. The approach reduces phase-separator and per-layer gate complexity from quadratic to linear scaling on sparse graphs. This method produces feasible, high-quality solutions under practical circuit and sampling budgets for eight IEEE systems across multiple quantum backends.

ContinualSkillBench evaluates the ability of large language models to evolve their capabilities through in-context continual skill learning, finding that sequential execution generally improves performance but with varying gains across models and domains. The framework suggests that adaptation to prior context and feedback is a key factor in improvement, while explicit skills provide selective benefits for tasks requiring reusable procedures or precise outputs. Less capable models tend to accumulate larger, more fragmented collections of task-specific skills.

POEM, a phase-aware forecasting framework, uses $\mathrm{SO}(2)$ feature rotation to address periodicity drift in time series forecasting by learning a phase-correction coordinate and applying an invertible rotation to paired latent features. POEM achieves competitive performance through the use of Directional Phase Increment Attention (DPIA) for extrapolating phase corrections from historical phase increments. The framework results in more regular latent trajectories.

Artificial intelligence (AI) data centers drive rapid electricity load growth across US regions, raising system costs and carbon exposure. A new game framework models a regulator's control over carbon penalties and subsidies, ISO capacity market clearing, and technology investors' decisions under incomplete information to capture the impact of AI on energy demand. The study investigates how second-life battery storage competes for capacity-market revenue against new/first-life storage in a framework that also considers the effects of a carbon tax, renewable subsidy, and SLB subsidy.

A novel coordination strategy for wind turbine generators is proposed to address mismatches between grid-forming control and primary frequency regulation, enabling consistent participation in frequency regulation while maintaining appropriate power points. The strategy involves well-designed relationships among key variables, preserving conventional GFM control structure and dynamic performance. Comparative case studies demonstrate the effectiveness of the proposed method under varying operating scenarios.