Energy Digest

A novel analytical framework for power system strength has been introduced, providing a unified formulation for assessing voltage and frequency strength. Simplified analytical solutions and compact expressions are offered to translate the theoretical framework into real-world applications, while new normalized strength metrics enable device-level comparisons. The framework is implemented in a real-world study case, demonstrating its applicability as a practical tool for comprehensive strength assessments.

VSC-HVDC links offer controllable active and reactive power output, making them suitable for emergency voltage support. An analytical method is presented to adjust the active power setpoint to maximize loadability during voltage-stressed conditions, requiring local voltage measurements and an estimate of a wide-area voltage angle difference. The method has been validated on the Nordic Test System, showing that adjusting the setpoint can yield a more-than-proportional increase in loadability.

A new approach called DL-Xformer uses an attention-based Transformer classifier to detect anomalies and classify faults in power systems with inverters, achieving detection times of 0.417-1.660 ms and classification times of 2.50-50.42 ms. This method outperforms Dynamic State Estimation-Based Protection (DSE-EBP) in terms of speed, but DSE-EBP detects all anomalies, while DL-Xformer only classifies events. The two methods can be combined to create a layered protection architecture for smart grids with inverters.

CASA-KalmanNet is an online adaptation framework that integrates a neural module to monitor KalmanNet's internal features and provide indicators of reliability degradation. This allows for autonomous adaptation to changes in the system without additional state labels, enabling timely data-efficient adaptation to both abrupt and gradual changes. CASA-KalmanNet outperforms existing learning-based filters under model mismatch while approaching optimal classical methods accuracy with full domain knowledge.

Real-time N-1 contingency screening trades off between assurance against cost by skipping some credible outages, while fast linear-sensitivity screening may pass unsafe operating points. Audited Selective Verification introduces a risk-budgeted layer that balances verification and cost by proposing which outages to skip, running online audits on small random samples, and certifying violation rates with calibrated thresholds. The method reduces full power-flow studies by 29-75% per real-time operating point.

A learning-accelerated algorithm for scenario-based model predictive control (SBMPC) is presented, which leverages parallel computing and Moreau envelope learning to efficiently solve SBMPC problems. The proposed framework reformulates the SBMPC problems into consensus forms that can be decomposed via Alternating Direction Method of Multipliers (ADMM), enabling parallel updates across scenarios and time steps. This leads to substantial computational speedups while maintaining reliable closed-loop control performance in applications such as microgrid energy management.

Long-duration energy storage (LDES) investments are suppressed in incomplete risk markets due to market incompleteness. Cap-and-floor contracts can restore investment levels by reducing downside risk, but this comes at the cost of substantial expected transfers from consumers to investors. Bilaterally negotiated contracts provide weaker investment incentives than centrally administered ones, highlighting the need for policymakers to balance contract and institutional design to achieve optimal outcomes.

A probabilistic framework using deep learning models predicts battery degradation with uncertainty, enabling robust predictions under dynamic operating conditions. The framework scales to full-system data by integrating cell-level predictions with system topology and real-world operational variability, providing probabilistic estimates for entire battery energy storage systems. It demonstrates accurate SOH degradation prediction with 95% prediction intervals that align well with field system measurements.

Researchers have developed an optimization framework for assembling repurposed lithium-ion battery packs from retired electric vehicle batteries. The framework reduces the challenge of assembling heterogeneous cells into reliable packs by considering various factors such as capacity, internal resistance, and self-discharge. It achieves significant improvements over single-metric sorting heuristics in reducing mismatch between cell selection and application requirements.

The article proposes Cluster-based Sequential Feature Selection (CSFS), a novel, model-agnostic, clustering-based wrapper method for automatic feature selection in renewable energy prediction pipelines, achieving better-performing selections of features than established methods. CSFS reduces computational cost on average by 21% compared to traditional wrapper-based sequential feature selection methods. The approach is evaluated on two real-world renewable energy prediction tasks and shows comparable predictive performance.