Hourly solar radiation forecasting using LS-based volterra filters

L. Ma; K. Khorasani; Y. Xiao; N. Yorino; F. I. Wahyudi

doi:10.1109/ICPRE.2017.8390670

Hourly solar radiation forecasting using LS-based volterra filters

L. Ma^*, K. Khorasani, Y. Xiao, N. Yorino, F. I. Wahyudi

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

1 Citation (Scopus)

Abstract

In this paper, a new solar radiation (insolation) prediction method is proposed that uses the Volterra filter which is trained based on the least squares (LS) criterion. Only historical insolation data are used as input information to train the Volterra filter that consists of the first- and second-order kernels. The proposed method is applied to real datasets of hourly insolation data of four years (2012-2015), which were downloaded from the website of Japan Meteorological Agency. Extensive simulations demonstrate the forecasting superiority of the proposed method over the naive persistence model (NPM), the autoregressive (AR) model, and the feedforward neural network (FFNN) schemes. Furthermore, the proposed method requires much lower training cost as compared to the FFNNs.

Original language	English
Title of host publication	2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	930-936
Number of pages	7
ISBN (Electronic)	9781538621561
DOIs	https://doi.org/10.1109/ICPRE.2017.8390670
Publication status	Published - 19 Jun 2018
Externally published	Yes
Event	2nd International Conference on Power and Renewable Energy, ICPRE 2017 - Chengdu, China Duration: 20 Sept 2017 → 23 Sept 2017

Publication series

Name	2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017

Conference

Conference	2nd International Conference on Power and Renewable Energy, ICPRE 2017
Country/Territory	China
City	Chengdu
Period	20/09/17 → 23/09/17

Keywords

Volterra filter
feedforward neural network
insolation prediction
least squares criterion
solar radiation forecasting

Access to Document

10.1109/ICPRE.2017.8390670

Cite this

Ma, L., Khorasani, K., Xiao, Y., Yorino, N., & Wahyudi, F. I. (2018). Hourly solar radiation forecasting using LS-based volterra filters. In 2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017 (pp. 930-936). (2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICPRE.2017.8390670

@inproceedings{aa710b94dc484b6da5bd715ebdf749d1,

title = "Hourly solar radiation forecasting using LS-based volterra filters",

abstract = "In this paper, a new solar radiation (insolation) prediction method is proposed that uses the Volterra filter which is trained based on the least squares (LS) criterion. Only historical insolation data are used as input information to train the Volterra filter that consists of the first- and second-order kernels. The proposed method is applied to real datasets of hourly insolation data of four years (2012-2015), which were downloaded from the website of Japan Meteorological Agency. Extensive simulations demonstrate the forecasting superiority of the proposed method over the naive persistence model (NPM), the autoregressive (AR) model, and the feedforward neural network (FFNN) schemes. Furthermore, the proposed method requires much lower training cost as compared to the FFNNs.",

keywords = "Volterra filter, feedforward neural network, insolation prediction, least squares criterion, solar radiation forecasting",

author = "L. Ma and K. Khorasani and Y. Xiao and N. Yorino and Wahyudi, {F. I.}",

note = "Publisher Copyright: {\textcopyright} 2017 IEEE.; 2nd International Conference on Power and Renewable Energy, ICPRE 2017 ; Conference date: 20-09-2017 Through 23-09-2017",

year = "2018",

month = jun,

day = "19",

doi = "10.1109/ICPRE.2017.8390670",

language = "English",

series = "2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "930--936",

booktitle = "2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017",

address = "United States",

}

Ma, L, Khorasani, K, Xiao, Y, Yorino, N & Wahyudi, FI 2018, Hourly solar radiation forecasting using LS-based volterra filters. in 2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017. 2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017, Institute of Electrical and Electronics Engineers Inc., pp. 930-936, 2nd International Conference on Power and Renewable Energy, ICPRE 2017, Chengdu, China, 20/09/17. https://doi.org/10.1109/ICPRE.2017.8390670

Hourly solar radiation forecasting using LS-based volterra filters. / Ma, L.; Khorasani, K.; Xiao, Y. et al.
2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017. Institute of Electrical and Electronics Engineers Inc., 2018. p. 930-936 (2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Hourly solar radiation forecasting using LS-based volterra filters

AU - Ma, L.

AU - Khorasani, K.

AU - Xiao, Y.

AU - Yorino, N.

AU - Wahyudi, F. I.

PY - 2018/6/19

Y1 - 2018/6/19

N2 - In this paper, a new solar radiation (insolation) prediction method is proposed that uses the Volterra filter which is trained based on the least squares (LS) criterion. Only historical insolation data are used as input information to train the Volterra filter that consists of the first- and second-order kernels. The proposed method is applied to real datasets of hourly insolation data of four years (2012-2015), which were downloaded from the website of Japan Meteorological Agency. Extensive simulations demonstrate the forecasting superiority of the proposed method over the naive persistence model (NPM), the autoregressive (AR) model, and the feedforward neural network (FFNN) schemes. Furthermore, the proposed method requires much lower training cost as compared to the FFNNs.

AB - In this paper, a new solar radiation (insolation) prediction method is proposed that uses the Volterra filter which is trained based on the least squares (LS) criterion. Only historical insolation data are used as input information to train the Volterra filter that consists of the first- and second-order kernels. The proposed method is applied to real datasets of hourly insolation data of four years (2012-2015), which were downloaded from the website of Japan Meteorological Agency. Extensive simulations demonstrate the forecasting superiority of the proposed method over the naive persistence model (NPM), the autoregressive (AR) model, and the feedforward neural network (FFNN) schemes. Furthermore, the proposed method requires much lower training cost as compared to the FFNNs.

KW - Volterra filter

KW - feedforward neural network

KW - insolation prediction

KW - least squares criterion

KW - solar radiation forecasting

UR - http://www.scopus.com/inward/record.url?scp=85050367793&partnerID=8YFLogxK

U2 - 10.1109/ICPRE.2017.8390670

DO - 10.1109/ICPRE.2017.8390670

M3 - Conference contribution

AN - SCOPUS:85050367793

T3 - 2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017

SP - 930

EP - 936

BT - 2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2nd International Conference on Power and Renewable Energy, ICPRE 2017

Y2 - 20 September 2017 through 23 September 2017

ER -

Ma L, Khorasani K, Xiao Y, Yorino N, Wahyudi FI. Hourly solar radiation forecasting using LS-based volterra filters. In 2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017. Institute of Electrical and Electronics Engineers Inc. 2018. p. 930-936. (2017 2nd International Conference on Power and Renewable Energy, ICPRE 2017). doi: 10.1109/ICPRE.2017.8390670

Hourly solar radiation forecasting using LS-based volterra filters

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this