Energy Digest

The article introduces a formal robustness certification framework for neural networks against adversarial attacks, particularly data poisoning attacks. The framework models gradient-based training as a discrete-time dynamical system and formulates poisoning robustness as a formal safety verification problem. This approach provides sufficient conditions to certify a robust radius ensuring the terminal model remains safe under worst-case ℓp-norm based poisoning.

The article presents a multimodal framework for assessing pedestrian well-being in urban environments, which integrates physiological sensing, geospatial tracking, and self-reports. This framework aims to capture the dynamic and subjective aspects of everyday walking experiences, providing a more comprehensive understanding of how urban environments influence well-being. The proposed approach offers a scalable and transferable method for advancing human-centered urban analytics.

This article proposes the use of high voltage circuit breaker operated braking resistors at data center transmission substations to enhance grid resilience during large load loss scenarios. The approach has been theoretically demonstrated and tested on a gigawatt scale data center load cluster connected to a 345 kV network, showing promising results in reducing frequency rate of change and providing time for governor response and capacitor switching. The method integrates with existing protection systems and can be scaled up or down according to the growth of the data center facility load.

The article presents a decision-focused learning (DFL) framework for optimizing day-ahead scheduling in underground pumped hydro energy storage (UPHES) systems. The proposed methodology addresses the computational complexity of nonlinear turbine performance characteristics and discrete operational modes by employing neural networks to predict penalty weights. This enables efficient solution quality and computation time, with significant improvements in profitability.

The article presents a Lyapunov-based small-gain theorem for establishing fixed-time input-to-state stability (FxT-ISS) guarantees in interconnected nonlinear dynamical systems. The proposed framework provides robustness with respect to external inputs and complements existing Lyapunov-based small-gain theorems for asymptotic and finite-time stability. It enables a systematic analysis of interconnection structures exhibiting fixed-time stability.

Here is a summary and key takeaways based on the provided article: The article discusses the Systematization of Knowledge (SoK) framework for understanding resilience and fault tolerance in Cyber-Physical Systems (CPS). The SoK taxonomy provides a structured lens for analyzing how faults arise, propagate, and amplify across CPS systems. By mapping representative systems onto the taxonomy, researchers identified four structural gaps that span multiple CPS domains.

The article discusses an approximation method for designing energy-gain control policies for linear time-varying systems with finite receding-horizon preview of model parameters. The proposed synthesis leverages the strict contraction of lifted Riccati operators to establish a sufficient number of preview steps for performance loss below a set tolerance. This is achieved while avoiding the dependence on all forward model data, which is necessary in standard formulations.

The article presents a bio-inspired 3D flocking algorithm for autonomous drone swarms, which uses minimal local interactions to achieve sharp transitions between swarming and schooling phases. By tuning interaction gains, the swarm exhibits enhanced responsiveness to external disturbances, rapid collective turns, and reliable recovery of alignment in the presence of intruders. The results demonstrate the effectiveness of simple gain modulation in adjusting stability, flexibility, and resilience.

The article proposes a new power and energy systems framework, TSA-LLM, for transient stability analysis that leverages large language models to achieve universality across diverse operating conditions and unseen faults. The framework overcomes the limitations of existing dynamics prediction frameworks by introducing innovative data processing pipelines, parameter-efficient architecture modifications, and a two-stage fine-tuning scheme. This universal framework demonstrates zero-shot generalization on multiple systems with minimal fine-tuning.

A proposed feedback control system is designed to balance water levels in irrigation channels with uncertain supply-demand mismatch. The decentralized controller adjusts gate flows between pools to regulate neighboring water-level errors to zero, ensuring steady-state balancing. The sequential SISO loop-shaping procedure for designing local flow controllers is developed and verified through numerical simulations and field trial results.