JZUS - Journal of Zhejiang University SCIENCE

Journal of Zhejiang University SCIENCE A

Accepted manuscript available online (unedited version)

Physics-informed deep learning for data-efficient and robust PV power forecasting

Author(s): Chang HUANG¹, Xuanbin HUANG¹, Jinmin GUO², Zhuo CAO¹, Ting HE¹, Wentao SHANG¹
Affiliation(s): ¹Energy and Electricity Research Center, Jinan University, Zhuhai 519070, China ²Beijing Tsintergy Technology Co., Ltd, Guangzhou 510630, China
Corresponding email(s): Ting HE, heting@jnu.edu.cn Wentao SHANG, wtshang@jnu.edu.cn
Key Words: Photovoltaic Power Forecasting; Physics-Informed Deep Learning; Various Forecasting Horizons; Data Efficiency; Physically-Grounded Features

Share this article to： More <<< Previous Paper \|Next Paper >>>

Chang HUANG¹, Xuanbin HUANG¹, Jinmin GUO², Zhuo CAO¹, Ting HE¹,Wentao SHANG¹. Physics-informed deep learning for data-efficient and robust PV power forecasting[J]. Journal of Zhejiang University Science A,in press.Frontiers of Information Technology & Electronic Engineering,in press.https://doi.org/10.1631/jzus.A2500582

@article{title="Physics-informed deep learning for data-efficient and robust PV power forecasting",
author="Chang HUANG¹, Xuanbin HUANG¹, Jinmin GUO², Zhuo CAO¹, Ting HE¹,Wentao SHANG¹",
journal="Journal of Zhejiang University Science A",
year="in press",
publisher="Zhejiang University Press & Springer",
doi="https://doi.org/10.1631/jzus.A2500582"
}

%0 Journal Article
%T Physics-informed deep learning for data-efficient and robust PV power forecasting
%A Chang HUANG¹
%A Xuanbin HUANG¹
%A Jinmin GUO²
%A Zhuo CAO¹
%A Ting HE¹
%A Wentao SHANG¹
%J Journal of Zhejiang University SCIENCE A
%P
%@ 1673-565X
%D in press
%I Zhejiang University Press & Springer
doi="https://doi.org/10.1631/jzus.A2500582"

TY - JOUR
T1 - Physics-informed deep learning for data-efficient and robust PV power forecasting
A1 - Chang HUANG¹
A1 - Xuanbin HUANG¹
A1 - Jinmin GUO²
A1 - Zhuo CAO¹
A1 - Ting HE¹
A1 - Wentao SHANG¹
J0 - Journal of Zhejiang University Science A
SP -
EP -
%@ 1673-565X
Y1 - in press
PB - Zhejiang University Press & Springer
ER -
doi="https://doi.org/10.1631/jzus.A2500582"

Abstract
Chinese Summary
Academic Network
Reviewer Comment

Abstract: Photovoltaic (PV) power forecasting is challenged by its inherent variability. Pure data-driven models struggle with generalization under data scarcity and complex weather conditions. In this paper, we introduce a physics-informed deep learning hybrid model (PIDL-HM) that systematically generates physically-grounded input features (e.g., plane-of-array irradiance and module temperature) for a Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM), establishing a principled integration framework beyond simple ensemble methods. Rigorously validated across multiple PV power plants in China and Australia for 15-min, 4-h, and 24-h forecasting, our approach demonstrates superior performance of up to 8.93% in root mean square error (RMSE) compared to a purely data-driven baseline. Crucially, the model shows remarkable data efficiency, maintaining high accuracy with only three months of training data, and exceptional robustness, providing a 6.87% improvement in performance under strong cross-seasonal data distribution shifts. This work has provided a reliable and data-efficient forecasting solution, establishing the PIDL-HM as a foundational element for next-generation forecasting systems.

Darkslateblue:Affiliate; Royal Blue:Author; Turquoise:Article

Reference

Open peer comments: Debate/Discuss/Question/Opinion

<1>

- Go to

Darkslateblue:Affiliate; Royal Blue:Author; Turquoise:Article

Reference