This paper sets out to develop an efficient probabilistic optimal power flow (POPF) algorithm to assess the influence of wind power on power grid. Given a set of wind data at multiple sites, the marginal distribution is fitted by a newly developed generalized Johnson system, the dependence structure of wind speeds is matched by the flexible Liouville copula. In order to lower the computational burden for solving POPF model, a Lattice sampling technique is developed to generate wind samples at multiple sites, and a Logistic mixture model is proposed to fit distributions of POPF outputs, which can quantify the effect of wind speed uncertainty on power grid. Finally, the proposed methods are illustrated by case studies on the IEEE 118-bus system.