Prosumer-based decentralized unit commitment for future electricity grids

Power systems;Unit commitment;Optimization

Costley, Mitcham Hudson ; Grijalva, Santiago Egerstedt, Magnus Ahmed, Shabbir Meliopoulos, Sakis Ferri, Bonnie Saeedifard, Maryam Electrical and Computer Engineering ; Grijalva, Santiago

University:Georgia Institute of Technology

Department:Electrical and Computer Engineering

关键词: Power systems;    Unit commitment;    Optimization;   

Others  :  https://smartech.gatech.edu/bitstream/1853/54890/1/COSTLEY-DISSERTATION-2015.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

The contributions of this research are a scalable formulation and solution method for decentralized unit commitment, experimental results comparing decentralized unit commitment solution times to conventional unit commitment methods, a demonstration of the benefits of faster unit commitment computation time, and extensions of decentralized unit commitment to handle system network security constraints. We begin with a discussion motivating the shift from centralized power system control architectures to decentralized architectures and describe the characteristics of such an architecture. We then develop a formulation and solution method to solve decentralized unit commitment by adapting an existing approach for separable convex optimization problems to the nonconvex domain of unit commitment. The potential computational speed benefits of the novel decentralized unit commitment approach are then further investigated through a rolling-horizon framework that represents how system operators make decisions and adjustments online as new information is revealed. Finally, the decentralized unit commitment approach is extended to include network contingency constraints, a crucial function for the maintenance of system security. The results indicate decentralized unit commitment holds promise as a way of coordinating system operations in a future decentralized grid and also may provide a way to leverage parallel computing resources to solve large-scale unit commitment problems with greater speed and model fidelity than is possible with conventional methods.