Short-term probability forecasting of wind power based on D-Vine quantile regression

Abstract

Wind power is stochastic and volatile, and largescale wind power grid connection affects system frequency and peak regulation, which requires accurate short-term time-series prediction of wind power. Based on the D-vine quantile point regression model, a short-term probabilistic prediction method for wind power that takes into account the time-series correlation of power output is proposed. Firstly, the D-vine structure is used to model the time-series correlation of wind power to obtain a joint probabilistic model of wind power output. Secondly, a conditional quantile regression algorithm is used to determine the number of variables for time-series prediction, and the probability of wind power output at the next moment is projected based on the successive joint probability distribution based on the wind power output in the previous sequence. The proposed probabilistic prediction method is tested using data from an offshore wind farm in Jiangsu, and the results show that the proposed algorithm can meet the requirements of offshore wind power output prediction.