Please use this identifier to cite or link to this item: https://hdl.handle.net/10356/76046
Title: Deep learning for power system time series forecasting
Authors: Sun, Weijia
Keywords: DRNTU::Engineering::Electrical and electronic engineering
Issue Date: 2018
Abstract: Power system time series forecasting is an essential part of smart electric grid. It enhances the reliability and reliability and efficiency of the power system. However, the traditional forecasting methods are unable to satisfy the much higher demand of precision in forecasting. In this dissertation, two kinds of power system datasets are tried, which are electricity load and wind power. Long Short Term Memory (LSTM) network is e↵ective for these sequential data based tasks and some signal preprocessing methods could improve prediction performance. Since wind power generation rely on wind speed, which is stochastic and intermittent, it is more difficult to forecast in short term compared with electricity load forecasting. After implementing di↵erent forecasting methods, a novel approach, which combines LSTM network and Empirical Mode Decomposition (EMD), is proposed. Original data is decomposed into constitutive series through EMD. The decomposition is expressed as a function of a combination of several components. LSTM networks are used to fit the components with di↵erent complexity for prediction. In the proposed model, the network is simplified and computational efficiency is improved. Keywords—Long Short Term Memory, Empirical Mode Decomposition, Wavelet Transform, Short-Term Load Forecasting, Wind Power Forecasting
URI: http://hdl.handle.net/10356/76046
Schools: School of Electrical and Electronic Engineering 
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:EEE Theses

Files in This Item:
File Description SizeFormat 
SunWeiJia_2018.pdf
  Restricted Access
Main article2.16 MBAdobe PDFView/Open

Page view(s)

294
Updated on Jul 13, 2024

Download(s)

10
Updated on Jul 13, 2024

Google ScholarTM

Check

Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.