A Multi-Time-Scale RO-BD Coordinated Fine-Grained Scheduling Method for Power Systems Considering Source–Load Temporal Coupling

Abstract

Abstract With the high level of uncertainty introduced by large-scale renewable energy integration and the inability of traditional hourly scheduling to capture minute-level power fluctuations, this paper puts forward a fine-grained coordinated scheduling method for source, grid, load, and storage that combines Rolling Optimization and Benders Decomposition. First, a unified modeling framework with a 5-minute resolution is established. Based on supply-side and demand-side characteristics, a Mixed Integer Linear Programming model is developed, incorporating temporal coupling constraints related to thermal power ramping, renewable energy accommodation, and energy storage and demand response. Second, a spatiotemporal coordinated decoupling strategy is designed. In the time domain, the rolling mechanism of Rolling Optimization dynamically corrects long-horizon forecast deviations. In the decision space, the Benders Decomposition algorithm decouples unit commitment decisions represented by integer variables from economic dispatch strategies represented by continuous variables. Dynamic cutting planes are constructed using dual information fed back from subproblems, which fundamentally overcomes the computational bottleneck of high-resolution fine-grained simulation. Case studies based on a typical regional power system, using a modified IEEE Reliability Test System, show that the proposed model effectively increases the average utilization rate of renewable energy to 95.7% (0.60% higher than the second-best method) and reduces the load peak-to-valley difference by 30.9% to 284.7 MW compared with the traditional hourly-resolution model. Curtailed power during peak generation periods decreases by 69.35%, from 43.4 MWh to 13.3 MWh. In particular, for high-dimensional temporal optimization problems, the proposed method reduces computation time by 57.2% compared with the Lagrangian relaxation method under standard test cases. For ultra-large-scale cases where existing methods fail, the proposed method demonstrates unique solvability. By combining the parallel computing potential of Benders Decomposition, this study provides a technical path for real-time fine-grained operation of new-generation power systems.

Keywords

Identifiers

DOI

10.1088/2631-8695/ae49e9

Journal

Engineering Research Express

Year

2026