Energy Digest

The Generalized Fiedler Vector (GFV) metric evaluates dynamic connectivity in power systems, capturing interplay between topology and dynamic properties. It leverages network connectivity and nodal inertia distribution to provide insights for optimal siting of stochastic generation to mitigate frequency variability. The approach is demonstrated on the IEEE 68-bus test system through Monte Carlo simulations.

The increasing penetration of renewable energy sources relies on grid-forming inverters for voltage and frequency regulation, but standard control schemes have limitations that affect their transient stability. A novel adaptive function is proposed to enhance the transient stability of droop-controlled GFM inverters by autonomously adjusting active power reference and droop gain based on terminal voltage. The method maximizes critical clearing time (CCT) while preventing acceleration of phase angle.

The article proposes an AC-Informed DC Optimal Transmission Switching (AIDC-OTS) scheme that uses admittance sensitivities-based constraints to guide the decision-making process towards AC-feasible topologies, resulting in a computationally tractable model. This approach enhances the original DC Optimal Transmission Switching (DC-OTS) model and validates its effectiveness through extensive simulations using PGlib test cases. The AIDC-OTS scheme outperforms other OTS formulations such as LPAC-OTS and QC-OTS.

A hybrid physics-based residual learning approach is proposed for accurate modeling of battery discharge and state-of-charge evolution in electric vehicles, combining a vehicle dynamics model with a neural network residual learner to correct complex discrepancies. This framework balances interpretable physics-based models with data-driven methods while reducing errors by approximately 0.8%. The approach outperforms traditional physics-only models while maintaining computational efficiency and physical interpretability.

A new tool based on Unbalanced Optimal Power Flow (UBOPF) is introduced to assess cost-effective local inverter control strategies for mitigating overvoltages in PV-rich distribution networks. The approach effectively maintains voltage levels within regulatory limits, with one method leading to lower PV curtailment rates than the other. The analysis highlights the need to consider these control actions as ancillary services to be properly compensated by the grid.

A digital twin-based cooling system optimization study was conducted on a liquid cooling infrastructure at the Frontier exascale supercomputer, achieving significant energy savings through various optimization strategies. The validated surrogate model enabled the evaluation of three constrained optimization frameworks, resulting in energy savings ranging from 20.4% to 27.8%. Co-optimizing supply temperature with flow rate nearly doubles the savings achievable by flow reduction alone.