Hafdaoui, HichemHichemHafdaouiKleebauer, MaximilianMaximilianKleebauerBouzekri, AbdelhafidAbdelhafidBouzekriBelhaouas, NasreddineNasreddineBelhaouasCharki, AbdérafiAbdérafiCharkiBouchakour, SalimSalimBouchakour2026-01-272026-01-272026-01-272026-01https://publica.fraunhofer.de/handle/publica/50489510.1016/j.nexres.2026.101385The rapid expansion of photovoltaic (PV) power plants requires automated, accurate, and scalable monitoring methods to support energy planning, environmental assessment, and grid management. Satellite remote sensing combined with artificial intelligence provides an effective solution for large-scale PV infrastructure detection. This study proposes a hybrid detection framework that integrates Convolutional Neural Networks (CNNs) and Support Vector Machines (SVMs) for identifying PV power plants in high-resolution multispectral Sentinel-2 satellite imagery. CNNs are employed to automatically extract discriminative spatial and spectral features, while SVMs are used as a robust classifier to enhance class separability and reduce false alarms caused by complex backgrounds and spectral similarity with other man-made surfaces. The proposed approach is evaluated through a case study of a 24 MWp utility-scale PV power plant located in El Abyed Sidi Chikh, El Bayadh, Algeria, with additional validation on rooftop and large-scale installations. Experimental results demonstrate that the hybrid CNN–SVM model significantly outperforms standalone CNN and SVM classifiers and providing reliable delineation of PV installations across different spatial scales.enCNNSVMPV Power PlantsImage ClassificationRemote Sensingjournal article