A data-driven approach for multi-objective unit commitment under hybrid uncertainties. (1st December 2018)

• Record Type: Journal Article
• Authors: Zhou, Min; Wang, Bo; Li, Tiantian; Watada, Junzo
• Is Part Of: Energy. Volume 164(2018)
• Page Start: 722
• Page End: 733
• Publication Date: 2018-12-01
• Subjects: Multi-objective unit commitment -- Non-parameter kernel density method -- Hybrid uncertainties -- Reinforcement learning-based particle swarm optimization algorithm; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2018.09.008
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: Recent years, renewable energy has taken growing penetration in power systems due to the energy shortage and environmental concerns. As a result, system operators encounter increasing difficulties in solving unit commitment optimization. In this paper, a data-driven unit commitment model is proposed to handle the hybrid uncertainties of wind power and future load. First, a non-parameter kernel density method is utilized to represent the above hybrid uncertainties, and a novel bandwidth selection strategy for the above method is then proposed to capture the inherent correlation between uncertainty representation and unit commitment. Second, a Monte Carlo simulation is developed to integrate the hybrid uncertainties into Value-at-Risk to get a comprehensive system reliability measurement. Third, considering that system operators might be interested in the inherent conflict between reliability and economy, minimizing operation costs and maximizing system reliability are taken as two objectives in the model. To get more practical schedules, the transmission line constraint is considered as well when building the mathematical model. Additionally, by integrating the reinforcement learning mechanism, a novel multi-objective particle swarm optimization algorithm is proposed to solve the complicated nonlinear model. Finally, several experiments were performed to demonstrate the effectiveness of this research. Highlights: Non-parameter kernel density method is used to represent hybrid uncertainties. A novel bandwidth selection strategy is proposed to adjust uncertainty distribution. A data-driven multi-objective unit commitment model is proposed. A reinforcement learning-based particle swarm optimization algorithm is developed. The model and algorithm are effective to solve unit commitment under uncertainties.