Short-term wind power probabilistic forecasting using a new neural computing approach: GMC-DeepNN-PF

Abstract

With the increasing penetration of renewable energy in power generation, the increasing uncertainty of renewable energy has significant influence on the stability of energy system. In order to handle the system uncertainty as well as satisfying the energy demand and supply, a short-term wind power probabilistic forecasting under uncertainty using the Gaussian mixed clustering-Deep neural network probabilistic forecasting (GMC-DeepNN-PF) is proposed. To illustrate the applicability of the proposed method, a case study of a wind power data set from Kaggle is presented and comparisons among four different data situations and four different forecasting models are shown in this paper. The raw wind power data is firstly pre-processed based on Gaussian Mixture Model reducing data defects. After that, the cleaned data is extended for more information which can improve the output accuracy and time as a property is added to forecasting model. According to DeepNN-PF, deep conventional neural network concatenated with T distribution is then utilized for forecasting using extended data. The simulations and comparisons demonstrate the advancement of the proposed method.