In this paper, the kappa distribution and Archimedean copula are employed to construct the joint probability distribution of correlated wind speeds. A percentile matching method is proposed to parameterize the kappa distribution, such that marginal distributions of wind speeds can be well represented. A correlation coefficient matching method is adopted to determine the parameters of Archimedean copulas, and a Laplace transform based algorithm is used for sample generation. Furthermore, an efficient simplex quadrature rule is introduced to calculate the statistical moments of probabilistic power flow outputs. The case studies show that the kappa distribution fits wind speed distributions better than the Weibull distribution, the Archimedean copula can well characterize the dependence structure of wind speeds, and the simplex rule yields results as accurate as point estimate method of (2m + 1) scheme, while reducing the computational burden by half.