Wind Power Interval Prediction Based on Robust Kernel Density Estimation

Abstract Wind power output has a high degree of randomness, so it is difficult to describe accurately with some typical probability distribution. The extreme values influence the sample’s general non-parametric kernel density estimation method, and estimated results are relatively conservative. A wind interval prediction method based on robust proposed improve compactness accuracy prediction. In estimation, this will assign small weight sample data reduce its better robustness than method. addition, bandwidth dynamic parameter, which can be adjusted avoid over-conservative