Probabilistic optimal power flow analysis incorporating correlated wind sources

Abstract

In this article, the Latin hypercube sampling (LHS) based probabilistic optimal power flow (P-OPF) technique is employed to assess the performance of power systems under large wind power penetration. In the case where only wind speed samples are available, the kernel tuned polynomial expansion series are developed to recover marginal distributions of wind speeds. Because the dependence among wind speeds has a considerable impact on P-OPF solutions, 13 Archimedean copulae are derived to represent the dependence structure of correlated wind speeds at multiple sites. The rejection sampling method and Metropolis-Hastings algorithm are deployed to generate samples of copula models, and a midpoint LHS technique is proposed to produce correlated low discrepancy sequences for P-OPF computation. Finally, a case study is performed on an IEEE 118-bus system to illustrate the proposed methods.