Distributionally Robust Economic Dispatch Considering the Uncertainty and Correlation of Wind Farm Outputs

Abstract

With the proportion of renewable energy in the power system increases, modeling the renewable energy generation with uncertainty accurately is significant in the economic dispatch of power system. Distributionally robust optimization (DRO) is widely used in power system dispatching considering the uncertainty of renewable energy generation, where the variation of uncertain parameters is defined using ambiguity sets (ASs). However, existing moment-based AS does not consider the correlation of uncertain parameters properly, which makes the DRO solution overly conservative. This paper proposes an AS construction method using a principal component analysis (PCA) algorithm, which considers the uncertainty and correlation of wind farm outputs. The economic dispatch problem is formulated using two-stage DRO with the AS constructed by PCA (DRED-PCA for short). The DRED-PCA model is reformulated into a semidefinite programming, and is solved by the delayed constraint generation algorithm. The performance of DRED-PCA is compared with the two-stage distributionally robust economic dispatch with the AS constructed without using the PCA algorithm. Comparative simulations show that the proposed method can improve the economic efficiency of the solution, and solve the economic dispatch problem more efficiently.