An improved wind power uncertainty model for day-ahead robust scheduling considering spatio-temporal correlations of multiple wind farms

Abstract

Robust optimization (RO) is an important method to deal with the uncertainty of wind power. The main challenge is to reduce the redundancy of the uncertainty model, thereby improving the economy of the scheduling plan while ensuring its robustness. In this paper, aiming at the day-ahead robust scheduling problem, the uncertainty model of wind power is improved by fitting 3 typical characteristics. First, the kernel density estimation (KDE) model and the non-parametric Copula model are combined to fit the nonlinear correlation between forecast power and forecast error, and the forecast error boundary constraints are established. Second, by combining the Mallet algorithm, the autoregressive integrated moving average (ARIMA) model, and the t distribution model, the temporal dimension constraints are established to describe the time-series characteristics of wind power. Third, the spatial dimension constraints are established based on the high-dimensional non-parametric regular vine (R-vine) Copula model, and the spatial correlation of multiple wind farms is reflected. Based on the above wind power uncertainty model, a 2-stage day-ahead robust scheduling model is established. The case study shows that the proposed wind power uncertainty model helps to achieve the balance between economics and robustness of the scheduling plan.