Verfahren zur Erzeugung Künstlicher Infrarotbilder eines Solar-Photovoltaischen Paneels zum Trainieren eines Maschinenlernalgorithmus zur Fehlererkennung

Details

Veröffentlichungs-Nr.

EP4712341

Anmeldetag

11. September 2024

Veröffentlichung

18. März 2026

Rechtsraum

EP

IPC

H02S50/00(HU XIN et al.)

Abstract

A computer implemented method for generating artificial infrared images of a solar photovoltaic panel comprising several solar cells, the artificial infrared images being intended for use as synthetic training data for a machine learning algorithm trained to detect a failure of the solar photovoltaic panel, each artificial infrared image being labeled with a type of failure, the method comprising the steps of, for each cell: Setting (202) an initial value of a photovoltaic temperature of the cell, Executing several iterations of the steps of: Determining (203) a photovoltaic operational power of the cell from an electro-thermal model of the cell function of the photovoltaic temperature of the cell, Determining (204) a new value of the photovoltaic temperature of the cell from a heat transfer model of the cell function of the photovoltaic operational power of the cell, At least a parameter of the electro-thermal model being modified from a corresponding reference parameter determined for a healthy cell to simulate a type of failure, Stopping (206) the iterations when a predefined convergence criteria is met, Providing (207,403) an artificial infrared image composed of the photovoltaic temperature of each cells and a label corresponding to the type of failure.

Anmelder

Firmen

Commissariat à l'Energie Atomique et aux Energies Alternatives
Université Savoie Mont Blanc
Centre National de la Recherche Scientifique

Land

🇫🇷 Frankreich