Energy Digest

A digitally controlled adaptive power conversion framework dynamically co-optimizes inductor configuration and switching frequency for high efficiency across a wide range of load conditions. The proposed reconfigurable inductor architecture adjusts effective inductance and switching frequency at runtime, resulting in improved average efficiency (4.5 percentage-point improvement) and reduced power loss (up to 32%) compared to fixed configurations. A hysteresis-based control strategy ensures robust operation under fast load transients.

Time-series foundation models (TSFMs) can recover a system's input-response relationship when provided with sufficient independent control excitation, but fine-tuning pipelines and feature smoothing reduce this need. TSFMs require informative control variation in the inference context to accurately capture system responses to alternative actions considered by controllers. Initial closed-loop results show promise for shorter context windows.

Simulations of a rural low-voltage feeder with 97 buses found that decentralized cost minimization of electric vehicles and heat pumps can stress the transformer beyond twice its rated capacity at moderate penetration levels. At high penetration, peak loads exceed 478% of rated capacity due to device-mix dependent effects. A central schedule capping the transformer can mitigate these demand-driven violations, while a regulated tariff provides only marginal relief for certain devices.

Establishing a separation principle allows the observer and controller to be designed independently for nonlinear multi-agent systems under denial-of-service attacks, using parametric Lyapunov equations and time-varying feedback to decouple coupled terms. The separation principle can be achieved regardless of the presence of event-triggered control and/or DoS attacks, provided certain linear-growth-type conditions are met. A simple design approach involves selecting only three parameters for each agent, including one for convergence time and two others for controller and hybrid observer gains.

RTKNNC, a graph-based method, detects cluster structure at multiple scales without pre-defining the number of clusters. The approach treats both directions of neighbor relations and refines its results through recursive forward-and-reverse traversal and optional refinement steps. It performs competitively against existing clustering methods while preserving a label-free construction process.

An electrochemically constrained residual-bias compensation dual extended Kalman filter (RBC-DEKF) reduces mean state-of-charge RMSE for lithium iron phosphate batteries with uncertainty in initial SOC. The RBC-DEKF improves SOC estimation compared to Single-EKF, reducing error by 89% and 92% for additive bias and multiplicative gain faults, respectively. The algorithm's ability to handle uncertainties and electrochemical constraints enables stable state recovery after fault introduction.

A curated collection of 22 public field-data releases from 16 research groups has been created, containing over 223 GB of data from at least 131,096 lithium-ion cells across various applications. The dataset includes 293 labeled faults, providing insights into real-world battery performance and safety issues. This collection aims to improve battery diagnostics, fault detection, and health estimation for researchers and manufacturers alike.