Research Area:  Machine Learning
Conventional wind power forecasting (WPF) methods adopt deterministic forecasting methods to produce a definite value of wind power output at a future time instant. However, any forecasting involves inherent uncertainty, and the uncertainty in WPF cannot be described by deterministic forecasting methods. Because WPF has the properties of time series data and long short-term memory (LSTM) is a time recursive neural network, the latter has significant advantages in forecasting the time series events. Therefore, in this study, a short-term WPF method based on the improved LSTM model is proposed, and the output power of a wind farm is calculated. The results show that the 4-h, 24-h, and 72-h forecasting accuracies of LSTM are higher than those of the back propagation (BP) neural network, the Particle swarm optimization and back propagation neural network (PSO-BP) hybrid model, and the wavelet neural network (WNN) at different time scales and seasons. The uncertainties in WPF performed by different forecasting models at different time scales are qualitatively described by the expectation, entropy, and hyper-entropy of cloud model. The uncertainties in WPF are quantitatively calculated by the confidence intervals based on the non-parametric kernel density estimation (NPKDE). The calculated results show that the proposed method can accurately predict the uncertainties in WPF at different confidence levels. The optimal operation results of reserve capacity based on the uncertainty in WPF and the optimal operation of the distribution network containing wind power and electric vehicles show that the proposed method can further improve the economic benefits of wind farm and distribution network.
Keywords:  
wind power forecasting
Long Short-Term Memory
Cloud Model
Non-Parametric Kernel Density Estimation
Deep Learning
Machine Learning
Author(s) Name:  Bo Gu,Tianren Zhang,Hang Meng,Jinhua Zhang
Journal name:  Renewable Energy
Conferrence name:  
Publisher name:  ELSEVIER
DOI:  10.1016/j.renene.2020.09.087
Volume Information:  Volume 164, February 2021, Pages 687-708
Paper Link:   https://www.sciencedirect.com/science/article/abs/pii/S0960148120315123