ABSTRACT
The quality of power interval forecast is mainly determined by three aspects: the accuracy of wind power point forecast, the accuracy of error distribution estimation, and the length of the error distribution confidence interval. In this paper, a method of achieving optimal interval forecast based on the Kernel Density Estimation (KDE) and Deep Belief Network (DBN) is proposed. Firstly, DBN is used to forecast wind power and it can improve the accuracy of wind power point forecast; secondly, KDE is used to obtain the error distribution and it can improve the accuracy of error distribution estimation; finally, the trust region method is used to obtain the minimum confidence interval. Simulation results show that this method not only can effectively improve the wind power interval forecast quality, but also has general applicability.
