Energy Digest

Slashers is a general system for modulating the power of Azure datacenters in response to various scenarios such as infrastructure failures or grid services, aiming to minimize negative impact on hosted workloads while meeting power targets. The system coordinates datacenter resources to handle power draw reduction from individual racks to regional multi-datacenter events. Slasher coordinates datacenter resources with a goal of meeting power targets and minimize negative impact on workloads.

A unified graph-based AC-OPF framework called UNION is proposed to ensure secure real-time grid operation in heterogeneous systems with topology-changing conditions. UNION achieves a 1.23% mean objective gap while satisfying operational limits on 99.56% of test instances, even under zero-shot N-1 contingencies and time-varying topologies. The model supports real-time AC-OPF across diverse systems, achieving inference times of 55-114 ms per instance.

Dynamic droop specifications for Grid-Forming Inverter-Based Resources propose a simple data-enabled model to capture small-signal dynamics of inverter-based resources, defining bounds on gain and phase of dynamic droop coefficients to ensure grid-forming capabilities. The results provide insight into certifying an inverter as grid-forming and clarify the notion of frequency control ancillary services. Dynamic droop specifications can screen IBR dynamics for potential adverse interactions with common controls.

Industrial production modeling provides operational constraints for industrial users participating in demand response programs. A new linear model of the production process balances computational complexity with modeling accuracy, offering significant improvements over existing methods. The proposed model's numerical results verify its accuracy and efficiency in evaluating demand response applications.

TrustFormer, a task-specific multi-dimensional trust evaluation framework, synchronizes heterogeneous trust-related data across historical collaborations using task identifiers and device-generated timestamps, then employs cross-temporal and cross-dimensional attention mechanisms to jointly model temporal dynamics and inter-dimensional correlations. The framework evaluates potential collaborators' multi-dimensional resource trust according to the multi-dimensional resource requirements of tasks, enabling optimal collaborator selection. TrustFormer outperforms existing methods by yielding a 40.8% improvement in trust evaluation accuracy.

A proposed bidirectional Mamba-enabled model (BM) evaluates device behavior based on long-term collaborations considering both forward and backward temporal dependencies to address challenges in assessing device reliability. The model aggregates short-term representations across all time intervals to produce a stable and reliable long-term behavior evaluation for each device. Experimental results show that BM achieves higher evaluation accuracy than baseline methods, enabling the selection of trustworthy collaborators.

System operators may allow virtual power plants (VPPs) to submit their feasible region for market clearing and dispatch, which requires VPPs to determine a feasible region based on individual operation models of internal distributed energy resources. A data-driven approach is proposed to approximate the energy-regulation feasible region of VPPs using a virtual battery model and inverse optimization. This method improves adaptability and accuracy compared to existing analytical approaches.

Virtual power plants (VPPs) aggregate distributed energy resources for ancillary services, but overlooking response requirements can reduce earnings or lead to disqualification. A new operational framework integrates fast-response capability into VPP bidding models using historical control commands and chance constraints to ensure a specified probability of meeting ancillary service requirements. This approach enhances VPP operation through case studies verifying the benefits of considering fast-response capabilities.

Accurate PV output power forecasting is crucial for optimizing grid-connected microgrids' operation, with proposed LSTM-based model showing 6% reduced RMSE compared to persistence model. The forecast also increases PV self-consumption ratio by 6.4% and reduces grid injections by 82%. Higher battery throughput can lead to accelerated aging.