Preview

PDF 142988 - Decomp...pdf | Download (4MB) | Preview

Abstract

Wind energy penetration has increased significantly and is playing a crucial role in the conversion of power systems to renewable energy. Remote and isolated power systems are increasing wind generation due to high cost of diesel fuel and transportation. To address the concerns of system frequency and scheduling from high penetration of stochastic wind generation, accurate short-term wind power forecasting is required. The research Investigates temporal resolution of wind energy data to improve neural network based forecast models. High resolution wind power data is used to simulate different temporal resolution, for both 10 minute and 1 hour forecast horizons. Three decomposition methods are compared wavelet, empirical mode, and variable mode decomposition. They each decomposed the sampled data into different modes, firstly a long-term component of lower frequencies, then more modes with detailed higher frequency components. To evaluate the temporal resolution and decomposition methods Back propagation neural network (BP), long short-term memory neural network (LSTM) and a convolutional neural network (CNN) are evaluated using wind power data from the King Island power system.

Item Type:	Conference Publication
Authors/Creators:	Aitken, W and Negnevitsky, M and Semshchikov, E
Keywords:	temporal resolution, renewable energy
Journal or Publication Title:	Proceedings of the 2020 Australasian Universities Power Engineering Conference (AUPEC)
Publisher:	IEEE-Inst Electrical Electronics Engineers Inc
Copyright Information:	Copyright 2020 University of Tasmania
Related URLs:	Google Scholar: Negnevitsky, M UTAS Profile: Negnevitsky, M
Item Statistics:	View statistics for this item

Actions (login required)

Item Control Page