Energy Digest

Oscillations in renewable energy integrated power systems pose a threat to grid secure operation due to the large-scale integration of renewables. A novel analytical framework explains the mechanism of stimulated oscillations, highlighting the importance of pole positioning and relationships between poles, zeros, and other characteristics on the complex plane. The research finds that its theory surpasses classical stability-based theories rather than negating them.

Machine learning has emerged as a faster alternative to traditional model-based methods for addressing the challenges in modern power systems, offering a scalable framework for incorporating topological dependencies. The integration of graph machine learning (GML) methods has shown promise for various applications such as forecasting, state estimation, optimization, and control in power systems. Despite its potential, GML faces open challenges including limited real-world deployment and a need for interpretable models in safety-critical settings.

A universal ontology framework is proposed to unify heterogeneous data silos in modern power systems and smart grids through a unified knowledge graph, integrating cyber-physical simulators via IEC standards. The framework achieves sub-linear scaling in knowledge graph size and construction time, with millisecond-level query performance for real-time decision support. It provides a robust, scalable information corpus for autonomous smart grid operations.

A reinforcement learning framework for distributed energy resource allocation learns optimal policies directly from operational data, adapting to stochastic demand variations through data-driven updates. The framework models DERA dynamics as a deterministic linear system and exogenous net load as a feature-based linear Markov process, capturing short-range temporal dependencies without explicit forecasting. It achieves high tracking accuracy and stable regulation across heterogeneous DER aggregators without requiring any demand prediction.

A two-layer model predictive control (MPC) framework is proposed for real-time control of sustainable data centers, integrating on-site renewable energy, battery storage, waste heat recovery, and district heating. The upper layer employs stochastic optimization to optimize intraday market participation, workload scheduling, and energy management under uncertainty, while the lower layer uses an adaptive MPC strategy to track dispatch references despite short-term disturbances. The framework reduces real-time dispatch deviations and imbalance costs compared to single-layer control strategies, offering a sustainable and grid-supportive operation approach for data centers.