A remotely located plant transmits sensor measurements to an operator over a network that may be under attack, and the goal is to detect attacks without prior knowledge of the system's model or structure. Two types of attacks are considered: model-free replay attacks and model-based stealthy attacks, for which closed-form expressions are derived for optimal attack policies against a χ^2 detector. A model-structure-free detector based on TimesFM achieves comparable or superior attack detection performance, demonstrated numerically on the IEEE 14-bus power system.

A new power flow model has been developed to address voltage unbalance (VU) in distribution grids, which is exacerbated by increasing single-phase loads and distributed generation. The model benchmarks two approaches: strict VU limit enforcement and penalization, with a proposed "Improved Hybrid Limits" formulation that balances compliance and optimization efficiency. The IHL approach converges faster and more reliably than penalization methods, making it a practical solution for mitigating VU in market-based operation of unbalanced distribution systems.

The article proposes a novel framework called GraphOPF that accelerates the solution of alternating current optimal power flow (AC-OPF) problems with topology changes, achieving speeds up to 200 times faster in neural network training and 66 times faster in solving large-scale AC-OPF problems. The framework incorporates toplogy-adaptability, scalability, self-supervision, and feasibility considerations. It achieves over 99% feasibility in solving large-scale power systems, including the real Korean power system.

A learning-assisted day-ahead energy scheduling framework is proposed to enhance power system inertial response and frequency stability in grids with grid-forming battery energy storage systems. The framework uses a surrogate model to represent the frequency support dynamics of the batteries, enabling accurate frequency metrics capture while reducing computational complexity. This approach outperforms analytical frequency-constrained day-ahead energy scheduling methods in terms of accuracy and efficiency.

A hybrid energy storage system and differentiable predictive control framework is proposed to mitigate AI datacenter power fluctuations, reducing residual power disturbance injections into the grid by up to 80%. The system allocates load deviation between a battery energy storage system and supercapacitor based on frequency characteristics. Simulations show that the system improves SC state-of-charge sustainability and reduces generator peak-to-peak frequency deviations.

Conformal Risk Sharing is a mechanism that provides trustworthy caps on participants' future obligations in cost allocation arrangements, ensuring no participant is made worse off. The framework pairs an interpretable sharing policy with split conformal calibration to produce distribution-free guarantees under exchangeability. Experiments confirm the method can reduce extreme obligations for high-risk agents while controlling harm to others.

A bi-level computation-electricity coordination framework is proposed to capture the bidirectional interactions between data centers (DCs) and power grids, enabling geo-distributed DCs to share computational resources and reduce energy consumption through peer-to-peer cloud service markets. The framework achieves a win-win coordination outcome, increasing total DC operational profit by 22.8% while alleviating grid congestion and reducing energy consumption by 3.2%. A proposed algorithm ensures rigorous convergence of the framework.

Smart Grids integrate digital communication infrastructure with distributed energy resources and other technologies to facilitate efficient energy operation and management. Emerging use cases such as smart distributed voltage control, real-time fault detection, and predictive maintenance have been identified with quantified service performance requirements. 5G and 6G network capabilities, including AI-driven optimization and edge computing, are enabling the development of next-generation Smart Grid systems.