Cheng Wang | Searcharxiv

arXiv2026

Regional Frequency Constrained Dispatch Method Considering Spatial-joint Stochastic Disturbances and Contingencies

The increasing penetration of renewable energy challenges frequency stability due to high variability and declining inertia. Traditional frequency security constrained dispatch methods fail to capture regional frequency heterogeneity and spatially correlated stochastic disturbances, resulting in inaccurate frequency security enforcement. To address this, a regional frequency constrained dispatch method is proposed, considering the spatially joint stochastic disturbances and contingencies. Firstly, a multi-regional frequency response model is constructed, incorporating the Vine-Copula based characterization of regional stochastic disturbances and regional frequency support. Then, a progressive latent-bottleneck physics-informed neural network is applied to characterize differential frequency nadir terms and regional frequency support integral terms via an encoder-decoder architecture, which enables optimization compatible expressions of system frequency dynamics. Finally, a day ahead dispatch model is developed in which regional stochastic frequency constraints are embedded using a CVaR-based formulation. A case study on the IEEE 118-bus system shows that the proposed method outperforms the unified center of inertia embedded approach in mitigating regional frequency violations, with the regional frequency nadir and RoCoF improved by 32.76% and 2.06%, respectively.

eess.SY↗

arXiv2026

DRIP-R: A Benchmark for Decision-Making and Reasoning Under Real-World Policy Ambiguity in the Retail Domain

LLM-based agents are increasingly deployed for routine but consequential tasks in real-world domains, where their behavior is governed by inherently ambiguous domain policies that admit multiple valid interpretations. Despite the prevalence of such ambiguities in practice, existing agent benchmarks largely assume unambiguous, well-specified policies, leaving a critical evaluation gap. We introduce DRIP-R, a benchmark that systematically exploits real-world retail policy ambiguities to construct scenarios in which no single correct resolution exists. DRIP-R comprises a curated set of policy-ambiguous return scenarios paired with a realistic customer personas, a full-duplex conversational simulation with tool-calling capabilities and a multi-judge evaluation framework covering policy adherence, dialogue quality, behavioral alignment, and resolution quality. Our experiments show that frontier models fundamentally disagree on identical policy-ambiguous scenarios, confirming that ambiguity poses a genuine and systematic challenge to LLM decision-making.

cs.CL↗