Energy Digest

The integration of first-principles models with learning-based components using linear fractional representations (LFRs) offers higher accuracy and faster convergence compared to black-box approaches. A new formulation addresses three key limitations, including ensuring well-posedness and stability guarantees through constraint-free parametrization, handling non-smooth cost functions, and facilitating automatic model order selection. These contributions are demonstrated on various simulation and benchmark identification examples.

A new stability assessment method for grey-box doubly fed induction generator (DFIG) systems uses a frequency sweeping technique to acquire a MIMO model, then applies a determinant-based criterion to analyze stability without requiring detailed system models. The proposed method also includes critical-pole estimation with trajectory information to assess the dominant mode of the system. This approach has been demonstrated effective through simulation and hardware-in-loop experiments.

The study explores the controllability of lithium-ion batteries, revealing a connection between the condition number of a battery's controllability matrix and the amount of control effort required. Poorly mathematically conditioned cell dynamics result in increased control effort, while aging affects all parameters equally. The analysis helps determine the best conditioned assemblies for cells with varying levels of new and second-life parameters.

The Thermodynamic Liquid Manifold Network introduces a new approach to solar irradiance forecasting that projects meteorological and geometric variables into a Koopman-linearized Riemannian manifold, achieving accurate results with zero-magnitude nocturnal error and a sub-30-minute phase response during high-frequency transients. The model is validated against five years of testing in a severe semi-arid climate, achieving an RMSE of 18.31 Wh/m2 and a Pearson correlation of 0.988. This ultra-lightweight design enables robust, thermodynamically consistent solar forecasting for edge-deployable microgrid controllers.