A Simplified LSTM Neural Networks for One Day-Ahead Solar Power Forecasting
In recent years, exploration and exploitation of renewable energies are turning a new chapter toward the development of energy policy, technology and business ecosystem in all the countries. Distributed energy resources (DERs) are being largely interconnected to electrical power grids. This disperse...
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2021-01-01
|
Series: | IEEE Access |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/9333638/ |
id |
doaj-477f4f05fa06457ea6067bdd99ace3f7 |
---|---|
record_format |
Article |
spelling |
doaj-477f4f05fa06457ea6067bdd99ace3f72021-03-30T15:16:03ZengIEEEIEEE Access2169-35362021-01-019171741719510.1109/ACCESS.2021.30536389333638A Simplified LSTM Neural Networks for One Day-Ahead Solar Power ForecastingChun-Hung Liu0Jyh-Cherng Gu1Ming-Ta Yang2https://orcid.org/0000-0002-4874-3589Department of Electrical Engineering, National Taiwan University Science and Technology, Taipei, TaiwanDepartment of Electrical Engineering, National Taiwan University Science and Technology, Taipei, TaiwanDepartment of Electrical Engineering, National Penghu University of Science and Technology, Magong, TaiwanIn recent years, exploration and exploitation of renewable energies are turning a new chapter toward the development of energy policy, technology and business ecosystem in all the countries. Distributed energy resources (DERs) are being largely interconnected to electrical power grids. This dispersed and intermittent generational mixes bring technical and economic challenges to the power systems in terms of operations, stability, reliability, interoperability and the policy making. In additional, DERs cause the significant impacts to the operation of traditional centralized generation power plants and the dispatch control centers. Under such circumstances, the accuracy of DERs power forecasting is one of the critical problems for TSO and DSO such as unit commitment, smooth fluctuations, peak load shifting, demand response, etc. In this paper, a simplified LSTM algorithm built over the architecture of Machine Learning methodology to forecast one day-ahead solar power generation is introduced. Through the machine learning processes of data processing, model fitting, cross validation, metrics evaluation and hyperparameters tuning, the result shows that the proposed simplified LSTM model outperform the MLP model. Moreover, the forecast of LSTM model can successfully capture intra-hour ramping on different weather scenarios. The average RMSE is 0.512 which is quite promising to inspire that the proposed methodology and architecture can best fit the short-term solar power forecasting applications.https://ieeexplore.ieee.org/document/9333638/Artificial neural networksDERLSTMmachine learningsolar power forecasting |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Chun-Hung Liu Jyh-Cherng Gu Ming-Ta Yang |
spellingShingle |
Chun-Hung Liu Jyh-Cherng Gu Ming-Ta Yang A Simplified LSTM Neural Networks for One Day-Ahead Solar Power Forecasting IEEE Access Artificial neural networks DER LSTM machine learning solar power forecasting |
author_facet |
Chun-Hung Liu Jyh-Cherng Gu Ming-Ta Yang |
author_sort |
Chun-Hung Liu |
title |
A Simplified LSTM Neural Networks for One Day-Ahead Solar Power Forecasting |
title_short |
A Simplified LSTM Neural Networks for One Day-Ahead Solar Power Forecasting |
title_full |
A Simplified LSTM Neural Networks for One Day-Ahead Solar Power Forecasting |
title_fullStr |
A Simplified LSTM Neural Networks for One Day-Ahead Solar Power Forecasting |
title_full_unstemmed |
A Simplified LSTM Neural Networks for One Day-Ahead Solar Power Forecasting |
title_sort |
simplified lstm neural networks for one day-ahead solar power forecasting |
publisher |
IEEE |
series |
IEEE Access |
issn |
2169-3536 |
publishDate |
2021-01-01 |
description |
In recent years, exploration and exploitation of renewable energies are turning a new chapter toward the development of energy policy, technology and business ecosystem in all the countries. Distributed energy resources (DERs) are being largely interconnected to electrical power grids. This dispersed and intermittent generational mixes bring technical and economic challenges to the power systems in terms of operations, stability, reliability, interoperability and the policy making. In additional, DERs cause the significant impacts to the operation of traditional centralized generation power plants and the dispatch control centers. Under such circumstances, the accuracy of DERs power forecasting is one of the critical problems for TSO and DSO such as unit commitment, smooth fluctuations, peak load shifting, demand response, etc. In this paper, a simplified LSTM algorithm built over the architecture of Machine Learning methodology to forecast one day-ahead solar power generation is introduced. Through the machine learning processes of data processing, model fitting, cross validation, metrics evaluation and hyperparameters tuning, the result shows that the proposed simplified LSTM model outperform the MLP model. Moreover, the forecast of LSTM model can successfully capture intra-hour ramping on different weather scenarios. The average RMSE is 0.512 which is quite promising to inspire that the proposed methodology and architecture can best fit the short-term solar power forecasting applications. |
topic |
Artificial neural networks DER LSTM machine learning solar power forecasting |
url |
https://ieeexplore.ieee.org/document/9333638/ |
work_keys_str_mv |
AT chunhungliu asimplifiedlstmneuralnetworksforonedayaheadsolarpowerforecasting AT jyhchernggu asimplifiedlstmneuralnetworksforonedayaheadsolarpowerforecasting AT mingtayang asimplifiedlstmneuralnetworksforonedayaheadsolarpowerforecasting AT chunhungliu simplifiedlstmneuralnetworksforonedayaheadsolarpowerforecasting AT jyhchernggu simplifiedlstmneuralnetworksforonedayaheadsolarpowerforecasting AT mingtayang simplifiedlstmneuralnetworksforonedayaheadsolarpowerforecasting |
_version_ |
1724179741188030464 |