Energy Digest

A planner-initiated siting framework has been introduced to mitigate stress on power grids due to explosive growth in energy-intensive AI data centers. The framework assesses potential interconnection sites using reliability-gated screening, system-wide market-impact assessment, and entropy-weighted multi-criteria scoring, expanding the siting frontier by 9-21% compared to firm operation. This technology-agnostic framework enables faster deployment of large flexible loads while preserving grid reliability and market stability.

Model-based optimization methods are computationally costly and insufficient for addressing terminal constraints in demand response scheduling. A new approach integrates Goal-Space Planning with Deep Deterministic Policy Gradient to improve sample efficiency and mitigate credit-assignment challenges. The proposed method demonstrates improved performance in satisfying terminal storage constraints while reducing myopic control behavior.

A distributionally robust model predictive control framework is proposed for optimal Virtual Power Plant operation under electricity price uncertainty, incorporating data-driven forecasting with quantile-based uncertainty quantification to construct time-varying Wasserstein ambiguity sets. The approach improves economic performance relative to standard forecast-based MPC when the ambiguity radius is chosen appropriately, resulting in consistent gains of up to 0.8%. Proper radius selection is crucial to avoid overly conservative results that reduce revenue.