.png){kind=icon}
Return
Feasibility-Informed Upper-Bound Improvement for Accelerating Large-Scale Unit Commitment
.png){kind=icon}
PRE
AI
DOI：10.1109/tii.2026.3728086.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In electricity market operations and large-scale power system dispatch, unit commitment (UC) must deliver high-quality schedules within strict clearing and operational time limits. As modern power systems integrate more controllable generators and energy storage devices, mixed-integer linear programming (MILP)-based UC becomes increasingly difficult to solve efficiently. However, existing MILP solver acceleration methods often lack explicit feasibility guidance for producing high-quality feasible schedules and are rarely tailored to UC structure. To address these issues, this article proposes two feasibility-informed upper bound improvement methods for UC based on analytic center theory: arc search rounding (ASR) and distance cost neighborhood search (DCNS). The proposed methods are customized to UC by incorporating the priority relationships and decision characteristics of UC variables into search, rounding, and variable-fixing. ASR uses feasibility information from the analytic center to generate an initial feasible solution at the root node by rounding points along an arc between the analytic center and the linear programming (LP) relaxation solution, with feasibility further enhanced through domain propagation and conflict analysis. During branching, DCNS improves the incumbent solution by combining hard fixing based on distance cost indicators with soft fixing based on Hamming-distance constraints to construct local neighborhoods. The methods are integrated into an open-source solver and evaluated on randomly generated instances from the French 1888-bus and Polish 2383-bus systems. Results show improvements in computational efficiency and solution quality for large-scale UC problems.
Keywords:
Arc search rounding (ASR)
distance cost neighborhood search (DCNS)
mixed-integer linear programming (MILP)
unit commitment (UC)
upper bound (UB) improvement
Journal
IF:
9.9
Papers:
8.6K
Citations:
6.0W
Organization
Cited Papers
No cited papers available