Energy Digest

A single neural-network based model generates simultaneous point and interval forecasts by treating point and interval forecasting as a multi-objective optimization problem, utilizing multi-gradient descent to adaptively select optimal weights. The proposed framework ensures non-crossing prediction intervals through a new loss function that guarantees coverage while maximizing sharpness, and outperforms existing methods in intra-day solar irradiance forecasting applications. It achieves target coverage with the narrowest point interval widths, remaining highly competitive compared to LSTM encoder-decoder and Transformer architectures.

Power system expansion is constrained by grid-supporting equipment (GSE) supply chains, which require critical materials such as copper to be manufactured and delivered on time. By 2030, shortages of GSE could reach 28.5-28.6% in the US under high-growth conditions, with trade disruption exacerbating the issue. Grid-enhancing technologies offer limited relief from these constraints.

The article proposes a novel approach for stabilizing fast power disturbances in grid-following inverters used at data centers, which are affected by query-driven power transients on millisecond timescales. The method is based on singular perturbation-based modeling and control, which yields explicit gain bounds linking inverter parameters to disturbance rejection performance. Numerical simulations validate the theoretical predictions under stochastic AI transients.

A scenario-based stochastic model predictive controller (SMPC) has been developed to optimize energy management in microgrids with electric vehicle fleets under persistent grid outages. The SMPC achieves performance within 1% of a perfect-forecast benchmark, outperforming naive MPCs that assume continuous grid availability and offering economic and sustainability advantages. Incorporating a deterministic buffer against EV consumption uncertainty eliminates over 90% of state-of-charge violations with negligible impact on operating costs.

The proposed decentralized stability-constrained Optimal Power Flow framework for inverter-based power systems incorporates algebraic decentralized small-signal stability criteria that can be used in optimization formulations, enabling tractable representations of stability conditions. The framework defines a Nodal Stability Shadow Price (NSSP) for each inverter, allowing for clear theoretical and practical interpretation of the stability contribution from each inverter. Considering reactive power costs in the optimization process yields strictly positive shadow prices and meaningful economic impacts on stability constraints.

Financial Transmission Rights (FTR) payouts can exceed the available merchandising surplus if the used network models are misaligned, leading to structural underfunding. This misalignment occurs independently of bidding behavior and can be quantified using a geometric framework that assigns weights to transmission element-contingency constraints. Misaligned FTR modeling choices result in hedging inefficiencies, even when DAM shadow prices remain constant over time.

Q-value iteration (Q-VI) identifies optimal action classes in finite time, but reaches its final limit $Q^*$ only in the general case. The distance from an iterate to the set of practically optimal solutions $\mathcal X^*$ decays exponentially at a rate governed by the joint spectral radius. This exponential decay can be faster than the standard convergence rate if the restricted spectral radius is strictly smaller.

Forecasting Ionospheric Irregularities uses dynamic graphs with ephemeris conditioning to predict ionosphere anomalies up to 2 hours ahead, achieving BSS of 0.49 and PR-AUC of 0.75 compared to persistence, with larger gains at longer lead times. The model outperforms previous methods by 35% in BSS and 52% in PR-AUC, especially for satellites that rise during the forecast horizon. The results suggest a viable alternative to grid-based representations for ionospheric irregularity forecasting using dynamic graph forecasting on evolving lines of sight.