Energy Digest

A unified co-optimization framework integrates Network Topology Optimization (NTO), Dynamic Line Rating (DLR), and Variable Impedance Devices (VIDs) to alleviate congestion and minimize operational costs. The framework uses a McCormick relaxation approach to linearize bilinear terms, achieving improved tractability and solution accuracy. It is validated on the IEEE 118-bus system and additional benchmark systems, demonstrating its effectiveness in improving solution quality and computational efficiency.

Thermal-Mass Coupled (TMC) modeling is proposed to address operational safety constraints in aluminum smelting demand response participation, capturing complex physical mechanisms of electrolytic cells. The model achieves net economic improvements 122.4% and 325.1% higher than the Equivalent Thermal Parameter (ETP) model in illustrative case studies. The proposed framework supports deeper demand response participation while explicitly assessing operational safety.

Noise Steering, a model-free reinforcement learning strategy, jointly controls turbine yaw angles and power setpoints to maximize wind farm power while satisfying a local noise constraint, achieving an average power deficit of 2% and up to 7.7% relative to unimpeded control. The strategy outperforms industry-standard noise-reduced operation, with an average power loss of 17%, and retains performance in higher-fidelity simulators. It successfully satisfies the 40 dB noise constraint throughout the full range of turbulent winds.

The proposed TANFL neural network uses voltage and current phasors, feeder connection relationships, and line parameters to locate faults in radial distribution feeders. It can handle missing measurements by introducing explicit missing data indicators and random bus data block masking. TANFL outperforms a conventional multilayer perceptron and a formula-based localization method in fault location accuracy, especially in scenarios with missing measurements.

A new model predictive control framework is proposed to mitigate periodic power fluctuations caused by AI training workloads. By coordinating the relative phases of independent training jobs, the method reduces power fluctuations by 74.39% and 46.6% in mean-square power fluctuations and power standard deviation, respectively. The framework updates launch decisions in real-time to minimize delays and maintain scheduling feasibility.

A proposed solution for coordinating EV charging with energy supply involves a greedy heuristic and a deep reinforcement learning controller that balances energy cost and carbon emissions. The system assigns EVs based on minimizing service delays and controls battery operation to optimize energy dispatch. The proposed model achieves the lowest accumulated service delay, energy cost, and carbon emissions in simulations.