Energy Digest

Nuclear fusion offers a scalable, low-carbon power source suitable for rapidly growing AI-driven data center demand, with high-capacity-factor, firm baseload generation that can meet continuous power requirements. Preliminary techno-economic analysis suggests that several Nth-of-a-kind fusion concepts may become cost-competitive with firmed renewable systems and advanced fission. Nuclear fusion's favorable safety profile and reduced waste burden improve its long-term social and political viability relative to traditional nuclear power options.

Voltage certificates can be designed to retain local disturbance bounds, droop slopes, inverter limits, and feeder dependent sensitivities while recovering the worst-case certificate as a special case. A new droop architecture with virtual coordination and affine feedforward compensation overcomes deterioration of voltage sensitivity bounds caused by network topology. The approach ensures forward invariance and satisfaction of reactive power reserve constraints.

A distributed unscented Kalman filter for nonlinear networks without a fusion center is developed, allowing local nodes to update at their own sampling instants and exchange estimates within one-hop neighborhoods. The filter's performance is demonstrated through testing on a stochastic nonlinear benchmark and a three-inertia network with non-observable individual nodes. The method outperforms covariance-averaging consensus in terms of accuracy and stability.

Small modular reactors can generate low-carbon power for hyperscale data centers at a price comparable to or even lower than traditional grid supply, with costs ranging from 49-62% more than grid supply in 2023. However, the cost of absorption cooling system dispatch and installation can add significant expenses to the overall project, ranging from $9.2 million per year to $570 kW$_\mathrm{c}^{-1}$ on a legacy-efficiency campus. The viability of reactor cogeneration depends on factors such as carbon prices, market conditions, and financing costs.

Battery safety can be diagnosed with high accuracy from sparse voltage measurements using a cross-modal diagnostic framework called DeFault. The framework mathematically unfolds one-dimensional voltage sequences into multi-dimensional topologies to decode compounded fault modes, achieving an average accuracy of 0.96 and an F1 score of 0.84 on a field dataset of 16.4 million data records. This method can diagnose battery faults without requiring new sensors or hardware upgrades.

The proposed reinforcement learning (RL) control framework for hybrid wind-wave energy systems achieves substantial Pareto improvements over conventional control strategies. The RL controller can generate up to 75% higher wave energy capture at the same platform motion level or nearly 50% lower motion at the same energy capture level, extending the performance boundary of such systems. This approach offers a promising solution for reducing costs in offshore renewable energy.

AIDC microgrids are increasingly vulnerable to power coordinated attacks due to the variability in renewable energy generation and fluctuations in AI data center workloads, posing stability risks to low-carbon AIDCs. The integration of renewable energy sources introduces cross-domain challenges that affect both supply and demand sides, creating interconnected stability risks. Computing-power coordinated attacks can induce sustained inverter frequency excursions exceeding 20% of the nominal value and reach instability conditions unattainable by single attacks.