Probability Interval Prediction of Wind Power Based on KDE Method With Rough Sets and Weighted Markov Chain

Abstract

Research on the uncertainty of wind power has a significant influence on power system planning and decision-making. This paper proposes a novel method for wind power interval forecasting based on rough sets theory, weighted Markov chain, and kernel density estimation (KDE) method. Since the wind power prediction is significantly correlated to its historical record, this method first applies the Markov chain method to predict the power at different steps based on historical power data, and then the overall power is calculated via rough set weighted summation. Finally, the obtained forecasting power is fed into the KDE forecasting model to obtain both upper and lower bounds of the probability interval of the wind power at a certain confidence level. The predicted interval coverage probability and average bandwidth are two of the criterions used to evaluate the proposed method. Moreover, the simulation results obtained via the Markov chain-KDE method and the weighted Markov chain-KDE method are compared against the results of the proposed method. These comparisons show that the proposed method based on rough set theory and weighted Markov chain KDE method offers unique advantages over the other methods for probability interval prediction of wind power, which are higher coverage, narrower average bandwidth, and a more accurate result.