Energy Digest

A high-fidelity digital twin dataset has been generated for inverter-based microgrids to address the limitations of public power-system datasets. The dataset includes synchronized three-phase PCC voltages and currents, per-DG active power, reactive power, frequency, and scenario labels from eleven operating and disturbance scenarios. The resulting dataset provides a consistent, labeled EMT benchmark for surrogate modeling, disturbance classification, robustness testing, and cyber-physical resilience analysis in inverter-dominated microgrids.

A data-driven successive linearization approach for voltage control under nonlinear power flow constraints is proposed, which leverages the fact that the deviation between the nonlinear solution and its linearization is bounded by the distance from the operating point. This method converges to a neighborhood of KKT points, establishing fast convergence and adaptability to changes in net load. The approach achieves better results than traditional fixed linear approximations under heavy power injection from distributed energy resources.

V2G adoption is hindered by uncertainties about its effects on battery lifetime and vehicle usability, but leveraging real-world Californian BEV usage data reveals that participation is most feasible for drivers who charge their vehicles daily. The impact of V2G on battery capacity loss depends on the calendar aging sensitivity of the batteries, with low-sensitivity batteries experiencing increased capacity loss across all drivers, while high-sensitivity batteries show negligible changes or even improved capacity retention. A user-centric V2G strategy can help better assess its potential and viability for California BEV drivers.

Battery degradation prognosis can be predicted by forecasting the state-of-health (SOH) trajectory over future cycles using a world model approach that encodes raw data into a latent state and propagates it forward in time. A Single Particle Model constraint improves prediction at the degradation knee, reducing forecast error compared to direct regression methods. This approach halves the trajectory forecast error compared to baseline direct regression methods.