arXiv subjects

Shaoze Li

Publications and source records attributed to Shaoze Li.

6 recordsLinked to original sources

arXiv2026

Risk-Aware Allocation of Transmission Capacity for Large Loads

Rapid growth in data centers and other large loads is straining transmission grid interconnection processes. This paper develops a framework to quantify firm transmission grid capacity and additional risk-aware flexible capacity that can be unlocked when large loads accept a predefined level of interruption risk. We design a normalized unmet-request objective to serve large load requests. We prove that, in radial networks, every minimizer of this objective also maximizes aggregate interconnected capacity, and we show numerically that this property nearly holds on meshed networks. To allocate the firm and flexible transmission capacities among competing large loads, we use a simultaneous ascending auction (SAA) over products differentiated by {\em capacity, risk level, and bus location}. When large loads have additive, symmetric concave, unit-demand, and $K_b$-demand valuations of the firm and flexible capacity products, we show that the gross substitutes property holds, which supports SAA convergence to a competitive equilibrium. A numerical study on the IEEE 73-bus system shows that, at a 1\% risk setting, flexible capacity increases the total network capacity by 44.3\% relative to the firm capacity baseline, while the SAA reaches a competitive equilibrium.

eess.SY↗

arXiv2026

Reachable-Set Decomposition for Real-Time Aggregation of Multi-Zone HVAC Fleets

Aggregating building heating, ventilation, and air-conditioning (HVAC) fleets can provide substantial real-time flexibility to power systems. However, practical deployment requires scalable characterization of multi-zone, multi-period flexibility and guaranteed feasible disaggregation as temperature states and exogenous inputs are revealed sequentially. This paper develops an offline-online reachable-set decomposition framework that combines offline temporal decomposition with a scalable online fleet coordination policy. Offline, backward reachable sets encode remaining-horizon feasibility as per-period state constraints, while tailored polytopic inner approximations provide tractable representations of these sets for thermally coupled multi-zone buildings. Online, aggregate flexibility is computed through parallel, single-period building-level linear programs followed by the fleet-level closed-form Minkowski summation of power intervals, thereby avoiding operations on high-dimensional full-horizon sets. The resulting endpoint profiles enable a closed-form, time-causal policy that disaggregates any aggregate HVAC power command within the reported interval while preserving recursive feasibility. Case studies show that, relative to the considered fixed full-horizon baselines, the proposed framework captures greater aggregate flexibility by recomputing per-period intervals using the latest state and exogenous information, while maintaining feasible disaggregation under sequentially revealed uncertainty. Its building-separable computation further enables scalable implementation for large fleets of multi-zone buildings.

eess.SY↗

arXiv2026

Behavioral Generative Agents for Power Dispatch and Auction

This paper presents positive initial evidence that generative agents can relax the rigidity of traditional mathematical models for human decision-making in power dispatch and auction settings. We design two proof-of-concept energy experiments with generative agents powered by a large language model (LLM). First, we construct a home battery management testbed with stochastic electricity prices and blackout interventions, and benchmark LLM decisions against dynamic programming. By incorporating an in-context learning (ICL) module, we show that behavioral patterns discovered by a stronger reasoning model can be transferred to a smaller LLM via example-based prompting, leading agents to prioritize post-blackout energy reserves over short-term profit. Second, we study LLM agents in simultaneous ascending auctions (SAA) for power network access, comparing their behavior with an optimization benchmark, the straightforward bidding strategy. By designing ICL prompts with rule-based, myopic, and strategic objectives, we find that structured prompting combined with ICL enables LLM agents to both reproduce economically rational strategies and exhibit systematic behavioral deviations. Overall, these results suggest that LLM-powered agents provide a flexible and expressive testbed for modeling human decision-making in power system applications.

eess.SY↗

arXiv2026

Variable Aggregation-based Perspective Reformulation for Mixed-Integer Convex Optimization with Symmetry

This paper addresses the challenging issue of symmetry in mixed-integer convex optimization problems, which frequently arise in real-world applications such as the unit commitment problem. Although variable aggregation techniques have been employed to mitigate symmetry, their impact on tightening the corresponding continuous relaxation has not been thoroughly investigated. In this work, we propose a new formulation that integrates the perspective reformulation method into the variable aggregation framework, yielding a tighter continuous relaxation for mixed-integer convex optimization problems with symmetric structures. We prove that, in the presence of symmetry, the convex hull of the feasible region associated with each set of aggregated variables can be exactly characterized. These results demonstrate the effectiveness of the proposed reformulation and establish new theoretical foundations for achieving tightness in variable aggregation-based mixed-integer programming formulations.

math.OC↗

arXiv2025

An Efficient Solution Method for Solving Convex Separable Quadratic Optimization Problems

Convex separable quadratic optimization problems occur in many practical applications. In this paper, based on an iterative resolution scheme of the KKT system, we develop an efficient method for solving a quadratic programming problem with a convex separable objective function subject to multiple convex separable constraints. We show that the proposed approach leads to a dual coordinate ascent algorithm and provide a convergence proof. Numerical experiments support the superior performance of the proposed method to that of the Gurobi solver, especially for solving large-scale convex separate quadratic programming problems.

math.OC↗

arXiv2025

Variable aggregation-based formulations for pumped storage hydro model in the day-ahead unit commitment problem

Pumped storage hydro (PSH) plants can improve the flexibility of power systems. A well-designed formulation for a PSH model is essential when incorporating the PSH units into a day-ahead unit commitment model. In the literature, the formulation of a PSH model is generally based on the individual PSH unit. This formulation is tight if there is only one PSH unit in the reservoir. However, when there are multiple units sharing the same reservoir in a PSH plant, the existing formulation may introduce some symmetric structures which degrade the efficiency of a mixed-integer programming solver significantly. In this paper, to cope with the symmetric structure in the PSH plants that have multiple units, we propose two new formulations. The first formulation considers the case in which there are multiple identical units sharing the same reservoir. The second formulation considers a general case in which the units sharing the same reservoir have the same generating and pumping efficiency. Using the new formulations, the symmetric structures in the problem can be effectively broken. Numerical results are presented to study the computational efficiency of the new formulations.

math.OC↗