Title:	Deep Learning Temperature Estimation Model for PV Modules
Author(s):	A. Kladas, B. Herteleer, J. Cappelle
Keywords:	PV System, Pyranometer, PV Module, Photovoltaic, Modelling
Topic:	Photovoltaic Modules and BoS Components
Subtopic:	PV Module Performance – Modelling, Testing, Standards
Event:	8th World Conference on Photovoltaic Energy Conversion
Session:	3BO.14.4
Pages:	517 - 519
ISBN:	3-936338-86-8
Paper DOI:	10.4229/WCPEC-82022-3BO.14.4
Price:	0,00 EUR
Document(s):	paper, presentation

Abstract/Summary:

---

Simulations and modelling are very important in order to estimate the produced energy of photovoltaic systems, which helps its management and optimization. This work presents the development of a reliable deep learning PV temperature prediction model based on experimental data. The model performance is evaluated from the data of two different sites with different climates and compared with state-of-the-art models. The results show that the developed model provides very accurate and stable estimations for 1-minute data (MAE 1.03°C, RMSE 1.34°C) on the first site, but it shows weaker performance when tested in a different climate and different mounting conditions, especially during weather patterns that have not been included in its training. The comparative analysis shows that the developed model outperforms the other examined models on PV temperature estimations, yet the coefficients translate poorly for other climates: the developed model must be calibrated for each site for optimal results. Data-driven models developed for universal usage must contain a wide range of possible meteorological cases.