A Parallel Rolling Horizon Scheme for Large Scale Security Constrained Unit Commitment Problems with Wind Power Generation

The Unit Commitment Problem (UCP) is an important category of power planning problems. The purpose of UCP is to determine when to start up and shut down the generator units and how to dispatch the committed units to meet the electricity demands, ancillary services requirements and security constraints. In this paper, we improve the traditional Lagrange Relaxation (LR) approach and analyze the effectiveness of using parallel computing in solving large unit commitment problems with wind penetration and investigate the potential of combining parallel computing with a rolling horizon scheme to improve the solution quality when a large amount of wind power is present. In particular, we first formulate a security constrained unit commitment problem by taking into account power generation costs, ancillary costs, wind power and a variety of security constraints employed in real New York State dayahead power market. We then propose a parallelized version of the LR method to solve the problem in a single step, analyze the scalability issue of parallel computing, and investigate the impact of increased wind energy penetration. Finally, when a large amount of wind power is present, we further propose an approach that combines parallel computing with a rolling horizon technique to solve the UCP online.