Prediction of photovoltaic resources based on deep learning

Under the background of national "double carbon", new energy represented by photovoltaic will usher in spurt development. At present, there are some problems such as low efficiency, low accuracy and high cost in surveying the potential data of assemblable photovoltaic roofs by manpower. To solve this problem, a photovoltaic resource prediction method based on deep learning was proposed. Firstly, meteorological data and image data including roof area were collected by UAV equipped with various sensors, and then the roof image data was identified by deep learning semantic segmentation algorithm, and the roof where photovoltaic panels can be installed was extracted. Finally, combined with meteorological and image data, deep learning Knime was used for analysis. The accuracy rate is 97.222%, and the kappa coefficient is 0.85.The results show that this method can predict the photovoltaic resources on the roof.