Energy Digest

The rapid growth of artificial intelligence (AI) data centers poses challenges to power system operation, as their variable and increasing demand requires effective characterization of demand flexibility. A proposed framework assesses the grid-compatible demand flexibility of AI data centers through batch workload temporal shifting, which preserves execution continuity while reducing server power demand. Numerical results show that workload temporal shifting provides quantifiable and grid-compatible demand flexibility with limited disruption to computing workloads.

A simulation-based framework for decentralized energy networks, called EnergyNet, is developed to model and evaluate the performance of interconnected power electronic routers. The framework represents various components and links within EnergyNet and uses receding-horizon optimization to calculate energy allocation and exchange, as well as support disturbance and reliability analysis. A case study on a Swedish EnergyNet system shows that it can maintain full service even in winter conditions with one grid interface or link lost, but reliability screening indicates that losing two links could partition the network.

PowerMarketJax introduces a benchmark suite for multi-agent reinforcement learning (MARL) in power markets, featuring five different environments with unique clearing, pricing, and settlement rules. The learning behavior of agents depends strongly on the market design, highlighting the need for tailored approaches to each market setting. PowerMarketJax offers a GPU-optimized framework for training MARL models, achieving up to 33X speedup over CPU-based baselines.

Decentralized power dispatch for integrated heat and power systems involves coordinating electric power systems and district heating networks while preserving privacy, using a new framework that enables variable-flow and variable-temperature operations. The method uses Bernstein-Galerkin equivalent projection to formulate the dynamic model without disclosing internal system topology or states. This approach preserves thermal state consistency and exploits flexibility, improving economic and computational performance.

A Quantum optimization Unit Commitment framework is developed, benefiting from Pauli Correlation Encoding and Sample-based Quantum Diagonalisation techniques to address scalability issues. The framework transforms a binary problem into a QUBO problem, mapped onto an Ising representation suitable for adiabatic quantum evolution within the Qibo framework. Using PCE and SQD techniques can reduce effective computational cost and improve solution reliability.

Existing analyses of grid-forming converter networks under current limiting primarily focus on single-converter systems, with network-level stability analysis remaining unresolved. A new energy-based framework is proposed to address this challenge, incorporating current limiting while retaining a provable dissipation property for stable analysis. The framework has been applied successfully to two-grid forming converter networks with reactive circular current limiter and impedance insertion.

Wide-area monitoring, protection, and control systems in smart grids are vulnerable to false data injection (FDI) attacks due to reliance on synchronized PMU measurements. An attack-resiliency analytics framework is presented that models the dynamics of these systems and formulates an optimal stealthy FDI attack as a mixed-integer linear program. This approach can reduce the attack objective by 15.5-55% and prevent over-frequency relay trips, while still allowing for feasible stealthy attacks.

A fixed-point framework reuses a single machine learning model trained on basecase topology data to predict post-contingency AC power flow states under non-critical single-line outages, offering a faster and more efficient approach than existing methods. The proposed method converges accurately across different topologies and demonstrates convergence rates in terms of network parameters. Numerical tests show promising results with prediction errors comparable to basecase predictions on a synthetic Texas system.

Rapid growth in renewable generation is changing power-system operation and increasing the importance of ancillary services for security and reliability. Market designs for ancillary services vary across seven countries, including Great Britain, Germany, Denmark, Italy, California, Texas, and Australia, with emerging trends such as faster frequency-response products, broader participation by various resources, and greater use of technology-neutral procurement. Despite some challenges, the comparison finds no single market design and supports differentiated but coordinated ancillary-service frameworks reflecting each service's characteristics.

Massachusetts recalibrated its Clean Peak Standard (CPS) in 2026, lowering minimum standards and expanding qualifying storage. This change reduced new battery additions by 53.1% and system cost by 3.7%, but lowered modeled emissions. Fossil generation during CPS hours increased by 11.7%.