View in EDS

Investigation on solar irradiance prediction model based on physics-informed neural network.

Saved in:
Bibliographic Details
Title: Investigation on solar irradiance prediction model based on physics-informed neural network.
Authors: Sun, Haibo1 (AUTHOR) sun_haibo@tju.edu.cn, Zhao, Xueqing1 (AUTHOR), Liu, Zongkuan2 (AUTHOR) liuzongkuan@tju.edu.cn, Cui, Yue1 (AUTHOR), Li, Bowen2 (AUTHOR), Ma, Zhuang1 (AUTHOR)
Source: International Journal of Green Energy. 2026, Vol. 23 Issue 9, p2009-2023. 15p.
Subject Terms: *Photovoltaic power generation, *Global radiation, *Clean energy, *Renewable energy sources, *Weather & climate change, Artificial neural networks, Meteorological databases
Abstract: Amid global efforts to address climate change and ensure energy security, renewable energy's large-scale development is key to energy transition. However, grid-connected photovoltaic power generation has volatile output due to dynamic weather, endangering grid stability. Thus, improving solar radiation prediction accuracy and generalization is a research focus. This paper proposes a physics-informed neural network (PINN) for solar irradiance prediction, embedding a physical loss function into traditional neural networks to boost performance. Taking Tianjin Port as the study area, it calculates solar azimuth and altitude via geometric principles, combines them with measured meteorological data as PINN inputs, and uses the physical coupling between solar irradiance and surface temperature as constraints to build the model. The results show that the PINN outperforms DNN and BPNN in terms of R2 on both the training set (0.974) and the test set (0.972) (its performance on the validation set is slightly weaker but still acceptable). On the test set, it achieves a MAPE of 4.01%, an MAE of 28.41, and an RMSE of 45.79, with a higher proportion of relative errors falling within the [−10%, 10%] range. Moreover, it maintains the highest R2 in cross-year seasonal tests, demonstrating its enhanced generalization capability. [ABSTRACT FROM AUTHOR]
Copyright of International Journal of Green Energy is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: GreenFILE
Description
Abstract:Amid global efforts to address climate change and ensure energy security, renewable energy's large-scale development is key to energy transition. However, grid-connected photovoltaic power generation has volatile output due to dynamic weather, endangering grid stability. Thus, improving solar radiation prediction accuracy and generalization is a research focus. This paper proposes a physics-informed neural network (PINN) for solar irradiance prediction, embedding a physical loss function into traditional neural networks to boost performance. Taking Tianjin Port as the study area, it calculates solar azimuth and altitude via geometric principles, combines them with measured meteorological data as PINN inputs, and uses the physical coupling between solar irradiance and surface temperature as constraints to build the model. The results show that the PINN outperforms DNN and BPNN in terms of R2 on both the training set (0.974) and the test set (0.972) (its performance on the validation set is slightly weaker but still acceptable). On the test set, it achieves a MAPE of 4.01%, an MAE of 28.41, and an RMSE of 45.79, with a higher proportion of relative errors falling within the [−10%, 10%] range. Moreover, it maintains the highest R2 in cross-year seasonal tests, demonstrating its enhanced generalization capability. [ABSTRACT FROM AUTHOR]
ISSN:15435075
DOI:10.1080/15435075.2026.2643393