Interval Probability Estimation of Wind Farm Cluster Ramp Events Based on Credal Network

Abstract

Large-scale wind power integration makes the impact of wind power ramp events more impossible to ignore. Compared with single station prediction, cluster prediction can reflect the impact of power mutation on the power system more intuitively, and the prediction results are more conducive to the decision-making of dispatchers. Therefore, this paper proposed an imprecise probabilistic prediction method for wind farm cluster. Data dimensionality reduction was carried out through correlation analysis and principal component analysis to avoid problems such as excessive data dimension caused by too many input variables and the influence of calculation speed. The credal network (CN) was established to express the dependent relationship between wind farm cluster ramp events and evidence variables, and the conditional dependent relationship was statistically quantified by using the imprecise Dirichlet model (IDM). Finally, combined with meteorological information, the ramp events were classified and inferred in the form of probability intervals, and the prediction performance was evaluated by using evaluation indexes. In this paper, the validity of the method was verified by using the data of a wind farm cluster in Xinjiang.