Attention-LSTM Wind Power Ultra-Short-Term Prediction Based on Kernel Principal Component Analysis

Lihao Zheng; Jiandong Duan; Pengfei Zhang; Tiancheng Mu; Yaoyun Zhang

doi:10.1007/978-981-95-7334-9_31

Attention-LSTM Wind Power Ultra-Short-Term Prediction Based on Kernel Principal Component Analysis

Lihao Zheng
, Jiandong Duan^*
, Pengfei Zhang
, Tiancheng Mu
, Yaoyun Zhang

^*Corresponding author for this work

School of Electrical Engineering and Automation, Harbin Institute of Technology

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

Abstract

To address the limitation that conventional multi-dimensional wind power prediction datasets incorporating meteorological information often fail to achieve satisfactory forecasting accuracy, this study proposes an ultra-short-term wind power forecasting framework integrating Kernel Principal Component Analysis (KPCA) with an Attention-based Long Short-Term Memory (LSTM) network. The proposed methodology employs KPCA for nonlinear dimensionality reduction of all features in the wind farm dataset, effectively extracting the most salient characteristics while eliminating redundant information. Furthermore, an attention mechanism is incorporated to enable the LSTM network to dynamically allocate higher weights to the feature elements that exert greater influence on the target prediction moment. Experimental validation using actual historical operational data from wind farms demonstrates that the proposed model significantly enhances ultra-short-term wind power forecasting performance, outperforming both traditional Back Propagation Neural Networks (BPNN) and conventional LSTM architectures in terms of prediction accuracy.

Original language	English
Title of host publication	The Proceedings of the 20th Annual Conference of China Electrotechnical Society
Editors	Qingxin Yang, Dianguo Xu, Xuerong Ye, Qiuyue Nie, Yueshi Guan
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	292-300
Number of pages	9
ISBN (Print)	9789819573332
DOIs	https://doi.org/10.1007/978-981-95-7334-9_31
State	Published - 2026
Externally published	Yes
Event	20th Annual Conference of China Electrotechnical Society, ACCES 2025 - Harbin, China Duration: 19 Sep 2025 → 21 Sep 2025

Publication series

Name	Lecture Notes in Electrical Engineering
Volume	1564 LNEE
ISSN (Print)	1876-1100
ISSN (Electronic)	1876-1119

Conference

Conference	20th Annual Conference of China Electrotechnical Society, ACCES 2025
Country/Territory	China
City	Harbin
Period	19/09/25 → 21/09/25

Keywords

Attention mechanism
Kernel principal component analysis
LSTM
Wind power prediction

Access to Document

10.1007/978-981-95-7334-9_31

Cite this

Zheng, L., Duan, J., Zhang, P., Mu, T., & Zhang, Y. (2026). Attention-LSTM Wind Power Ultra-Short-Term Prediction Based on Kernel Principal Component Analysis. In Q. Yang, D. Xu, X. Ye, Q. Nie, & Y. Guan (Eds.), The Proceedings of the 20th Annual Conference of China Electrotechnical Society (pp. 292-300). (Lecture Notes in Electrical Engineering; Vol. 1564 LNEE). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-981-95-7334-9_31

Zheng, Lihao ; Duan, Jiandong ; Zhang, Pengfei et al. / Attention-LSTM Wind Power Ultra-Short-Term Prediction Based on Kernel Principal Component Analysis. The Proceedings of the 20th Annual Conference of China Electrotechnical Society. editor / Qingxin Yang ; Dianguo Xu ; Xuerong Ye ; Qiuyue Nie ; Yueshi Guan. Springer Science and Business Media Deutschland GmbH, 2026. pp. 292-300 (Lecture Notes in Electrical Engineering).

@inproceedings{4af854aafc3e4a658d097b54fa555844,

title = "Attention-LSTM Wind Power Ultra-Short-Term Prediction Based on Kernel Principal Component Analysis",

abstract = "To address the limitation that conventional multi-dimensional wind power prediction datasets incorporating meteorological information often fail to achieve satisfactory forecasting accuracy, this study proposes an ultra-short-term wind power forecasting framework integrating Kernel Principal Component Analysis (KPCA) with an Attention-based Long Short-Term Memory (LSTM) network. The proposed methodology employs KPCA for nonlinear dimensionality reduction of all features in the wind farm dataset, effectively extracting the most salient characteristics while eliminating redundant information. Furthermore, an attention mechanism is incorporated to enable the LSTM network to dynamically allocate higher weights to the feature elements that exert greater influence on the target prediction moment. Experimental validation using actual historical operational data from wind farms demonstrates that the proposed model significantly enhances ultra-short-term wind power forecasting performance, outperforming both traditional Back Propagation Neural Networks (BPNN) and conventional LSTM architectures in terms of prediction accuracy.",

keywords = "Attention mechanism, Kernel principal component analysis, LSTM, Wind power prediction",

author = "Lihao Zheng and Jiandong Duan and Pengfei Zhang and Tiancheng Mu and Yaoyun Zhang",

note = "Publisher Copyright: {\textcopyright} Beijing Paike Culture Commu. Co., Ltd. 2026.; 20th Annual Conference of China Electrotechnical Society, ACCES 2025 ; Conference date: 19-09-2025 Through 21-09-2025",

year = "2026",

doi = "10.1007/978-981-95-7334-9\_31",

language = "英语",

isbn = "9789819573332",

series = "Lecture Notes in Electrical Engineering",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "292--300",

editor = "Qingxin Yang and Dianguo Xu and Xuerong Ye and Qiuyue Nie and Yueshi Guan",

booktitle = "The Proceedings of the 20th Annual Conference of China Electrotechnical Society",

address = "德国",

}

Zheng, L, Duan, J, Zhang, P, Mu, T & Zhang, Y 2026, Attention-LSTM Wind Power Ultra-Short-Term Prediction Based on Kernel Principal Component Analysis. in Q Yang, D Xu, X Ye, Q Nie & Y Guan (eds), The Proceedings of the 20th Annual Conference of China Electrotechnical Society. Lecture Notes in Electrical Engineering, vol. 1564 LNEE, Springer Science and Business Media Deutschland GmbH, pp. 292-300, 20th Annual Conference of China Electrotechnical Society, ACCES 2025, Harbin, China, 19/09/25. https://doi.org/10.1007/978-981-95-7334-9_31

Attention-LSTM Wind Power Ultra-Short-Term Prediction Based on Kernel Principal Component Analysis. / Zheng, Lihao; Duan, Jiandong; Zhang, Pengfei et al.
The Proceedings of the 20th Annual Conference of China Electrotechnical Society. ed. / Qingxin Yang; Dianguo Xu; Xuerong Ye; Qiuyue Nie; Yueshi Guan. Springer Science and Business Media Deutschland GmbH, 2026. p. 292-300 (Lecture Notes in Electrical Engineering; Vol. 1564 LNEE).

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

TY - GEN

T1 - Attention-LSTM Wind Power Ultra-Short-Term Prediction Based on Kernel Principal Component Analysis

AU - Zheng, Lihao

AU - Duan, Jiandong

AU - Zhang, Pengfei

AU - Mu, Tiancheng

AU - Zhang, Yaoyun

N1 - Publisher Copyright: © Beijing Paike Culture Commu. Co., Ltd. 2026.

PY - 2026

Y1 - 2026

N2 - To address the limitation that conventional multi-dimensional wind power prediction datasets incorporating meteorological information often fail to achieve satisfactory forecasting accuracy, this study proposes an ultra-short-term wind power forecasting framework integrating Kernel Principal Component Analysis (KPCA) with an Attention-based Long Short-Term Memory (LSTM) network. The proposed methodology employs KPCA for nonlinear dimensionality reduction of all features in the wind farm dataset, effectively extracting the most salient characteristics while eliminating redundant information. Furthermore, an attention mechanism is incorporated to enable the LSTM network to dynamically allocate higher weights to the feature elements that exert greater influence on the target prediction moment. Experimental validation using actual historical operational data from wind farms demonstrates that the proposed model significantly enhances ultra-short-term wind power forecasting performance, outperforming both traditional Back Propagation Neural Networks (BPNN) and conventional LSTM architectures in terms of prediction accuracy.

AB - To address the limitation that conventional multi-dimensional wind power prediction datasets incorporating meteorological information often fail to achieve satisfactory forecasting accuracy, this study proposes an ultra-short-term wind power forecasting framework integrating Kernel Principal Component Analysis (KPCA) with an Attention-based Long Short-Term Memory (LSTM) network. The proposed methodology employs KPCA for nonlinear dimensionality reduction of all features in the wind farm dataset, effectively extracting the most salient characteristics while eliminating redundant information. Furthermore, an attention mechanism is incorporated to enable the LSTM network to dynamically allocate higher weights to the feature elements that exert greater influence on the target prediction moment. Experimental validation using actual historical operational data from wind farms demonstrates that the proposed model significantly enhances ultra-short-term wind power forecasting performance, outperforming both traditional Back Propagation Neural Networks (BPNN) and conventional LSTM architectures in terms of prediction accuracy.

KW - Attention mechanism

KW - Kernel principal component analysis

KW - LSTM

KW - Wind power prediction

UR - https://www.scopus.com/pages/publications/105039335704

U2 - 10.1007/978-981-95-7334-9_31

DO - 10.1007/978-981-95-7334-9_31

M3 - 会议稿件

AN - SCOPUS:105039335704

SN - 9789819573332

T3 - Lecture Notes in Electrical Engineering

SP - 292

EP - 300

BT - The Proceedings of the 20th Annual Conference of China Electrotechnical Society

A2 - Yang, Qingxin

A2 - Xu, Dianguo

A2 - Ye, Xuerong

A2 - Nie, Qiuyue

A2 - Guan, Yueshi

PB - Springer Science and Business Media Deutschland GmbH

T2 - 20th Annual Conference of China Electrotechnical Society, ACCES 2025

Y2 - 19 September 2025 through 21 September 2025

ER -

Zheng L, Duan J, Zhang P, Mu T, Zhang Y. Attention-LSTM Wind Power Ultra-Short-Term Prediction Based on Kernel Principal Component Analysis. In Yang Q, Xu D, Ye X, Nie Q, Guan Y, editors, The Proceedings of the 20th Annual Conference of China Electrotechnical Society. Springer Science and Business Media Deutschland GmbH. 2026. p. 292-300. (Lecture Notes in Electrical Engineering). doi: 10.1007/978-981-95-7334-9_31

Attention-LSTM Wind Power Ultra-Short-Term Prediction Based on Kernel Principal Component Analysis

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this