In modern power systems, the increasing penetration of renewable energy provides promising ways to operate the system at low cost and low pollution. Unfortunately, the noticeable uncertainties caused by the forecast errors of renewable generation as well as power load restrict the utilization of renewable energy (e.g. cause wind curtailment), and moreover, bring unprecedented challenges to maintain system reliability (e.g. cause load shedding which generally leads to high compensation). One prevailing solution is to prearrange sufficient spinning reserve of thermal units, which however, deviates from the ambition of cost saving. This paper analyzes the relationship between operation cost and wind curtailment amount using multi-objective unit commitment. The main contents include: First, systems with mixed generation sources including thermal units, wind farms and battery-based energy storage are investigated, whereafter Value-at-Risk-based worst-case estimations on load shedding and wind curtailment are established under hybrid uncertainties. Second, operation cost with generation, emission and load shedding concerns is established, whereafter multi-objective optimization is performed to investigate the inherent nature between operation cost and wind curtailment. Third, a multi-objective particle swarm optimization algorithm with reinforcement learning is developed to solve the complicated model. Finally, experiments on two case studies were performed to demonstrate the research effectiveness. (C) 2019 Elsevier Ltd. All rights reserved.