Large-scale integrations of wind power increase stochastic nature of generation schedule decision, and might make reliability index lost of load probability (LOLP) and reliability index expected energy not served (EENS) deviate each other. In this paper, unit commitment (UC) formulation is revised to optimize day-ahead generation schedule for those systems with high wind power penetration level. In the revised UC formulation, conventional deterministic constraint on spinning reserve (SR) is replaced by a chance constraint on reliability index LOLP and minimum optimization objects is extended by expected interruption costs estimated by a product of EENS and the value of lost load (VOLL). With considerations of stochastic fluctuations of load/wind power together with random outages of generators, an analytic algorithm is presented to calculate reliability indices LOLP and EENS. Genetic algorithm (GA) is utilized to solving the revised UC formulation. In addition, an intelligent mutation operator (IMO) is specifically designed to improve the performances of the GA. Simulation results on certain 10-unit case system have proven the efficiency of the methodology proposed here.