The traditional deterministic unit commitment cannot adequately address the safe and economic operation of power systems with large-scale volatile and uncontrollable wind power integration. By combining an uncertainty analysis of wind power based on confidence interval and cost-benefit analysis in economics, an improved unit commitment model considering the uncertainty risk of wind power predictions is proposed to appropriately apply wind power predictions into unit commitment. Cost-benefit analysis is utilized in the proposed model to obtain a single objective function consisting of the generation cost of units and loss-of-load risk of power systems. The proposed model remedies the defects of the existing models where the selection of confidence interval is not given, and realizes a scheduling decision compromising the economic efficiency and the risk of wind power. Using mixed integer linear programming method, simulation studies on the IEEE 26-generator reliability test system connected to a wind farm are presented to verify the effectiveness and advantage of the proposed model.