ABSTRACTDue to the sustainable nature and stimulus plans from government, renewable energy (such as wind and solar) has been increasingly used in power systems. However, the intermittency of renewable energy creates challenges for power system operators to keep the systems reliable and cost-effective. In addition, information about renewable energy is usually incomplete. Instead of knowing the true probability distribution of the renewable energy course, only a set of historical data samples can be collected from the true (while ambiguous) distribution. In this article, we study two risk-averse stochastic unit commitment models with incomplete information: the first model being a chance-constrained unit commitment model and the second one a two-stage stochastic unit commitment model with recourse. Based on historical data on renewable energy, we construct a confidence set for the probability distribution of the renewable energy and propose data-driven stochastic unit commitment models to hedge against the incomplete nature of the information. Our models also ensure that, with a high probability, a large portion of renewable energy is utilized. Furthermore, we develop solution approaches to solve the models based on deriving strong valid inequalities and Benders’ decomposition algorithms. We show that the risk-averse behavior of both models decreases as more data samples are collected and eventually vanishes as the sample size goes to infinity. Finally, our case studies verify the effectiveness of our proposed models and solution approaches.