Energy Digest

A probabilistic assessment of rare transient instability events is necessary in modern power systems due to increasing uncertainty from intermittent energy sources and variable loads. A new Kriging-based active learning framework accurately characterizes rare instability regions within input uncertainty space, estimating small instability probability with limited expensive simulations. The proposed framework outperforms existing methods in accuracy and computational efficiency.

Grid-following inverters' large-signal stability is challenged by interacting nonlinear dynamics between the phase-locked loop, DC-link voltage control, and AC terminal voltage control. An asymptotic analysis approach, bandwidth separation method, reveals that interactions degrade system stability and identify voltage instability as a root cause of transient instability. Optimal bandwidth configurations for PLL and DVC are identified to improve resilience and tolerance under various grid fault conditions.

A novel cost-guided learning approach is proposed to approximate model predictive control (MPC) policies using neural networks, providing tighter guarantees on optimality loss compared to traditional error-guided learning methods. This approach utilizes cost sensitivity information from the MPC problem to directly minimize the loss in closed-loop performance, achieving substantially better results than existing methods. The technique has been demonstrated through experiments on a continuous stirred tank reactor (CSTR) benchmark.

A new hybrid model combining a first-order Thevenin equivalent circuit with a compact neural network achieves the lowest voltage error across various conditions, including matched conditions, perturbation, temperature transfer, and drive-cycle transfer. This approach reduces mean absolute error by 48% compared to a Long Short-Term Memory (LSTM) baseline under matched conditions, while also showing lower inter-seed variability. The model's ability to correct for physical polarization mismatch improves its performance across distribution shifts.

A novel Stochastic Optimal Power Flow (SOPF)-based adaptive droop framework is proposed for hybrid AC-HVDC grids under offshore wind uncertainty, leveraging a zone-wise Beta distribution to model wind forecast uncertainty. This framework uses Polynomial Chaos Expansion (PCE) within chance-constrained SOPF to formulate the system's stochastic states analytically, extracting optimal adaptive droop gains via a Jacobian-free sensitivity analysis. The proposed approach outperforms standard fixed-coefficient approaches in minimizing active-power tracking errors during extreme wind disturbances.

Community-to-vehicle (C2V) integrates electric vehicle charging into local energy communities by allowing surplus locally generated renewable energy to be allocated and offered to external users at a community charging price, improving PV utilization and economic performance. C2V reduces EV charging costs compared to commercial alternatives and generates additional revenue streams for the community. It has the potential as a practical mechanism for integrating EV charging into local energy communities within existing regulatory frameworks.

The use of soft magnetic composites (SMCs) in stators of wound field synchronous machines for automotive applications delivers improved torque and efficiency compared to conventional designs. Integrating SMCs into full electric drive units achieves 89.7% efficiency over the WLTP drive cycle, outperforming reference permanent magnet synchronous machine-based EDUs. This novel material combination eliminates rare-earth materials, reduces cost, and offers environmental benefits through SMC utilization.

Power systems digital twins (DTs) aim to accelerate and improve decision-making across multiple time scales and geographic scopes, but current research has not delivered a practical solution. A new generation of DTs called Foundation Twins is proposed, combining generalization features of foundation models with decision-making capabilities of reinforcement learning architectures. Foundation Twins leverage recent advances in artificial intelligence (AI) and machine learning (ML) to create powerful modeling and simulation tools for power systems.