Application des Techniques d’Intelligence Artificielle dans les Systèmes Photovoltaïques en vue de Diagnostic des Défauts

Abstract

This thesis deals with application of artificial intelligence techniques for the diagnosis, detection and classification of defects in photovoltaic systems. These latter like all electrical and electronic systems, can break down and degrade during the operating period. This requires a diagnostic whose main objective is to provide an automatic tool that can early detect defects to protect the persons and installations, and in addition can classify these defects. At the end of 2016, 303 GW of photovoltaic energy was installed around the world. About 75 GW installed only in the year of 2016, and this comes from the fact that new solutions have encouraged government to rely more and more on this kind of energy. For the development of fault classification algorithms in photovoltaic systems, at the beginning, a database is collected using real time emulator. Then, classifiers based on artificial intelligence were built, such as the fuzzy classifier, neuronal and the neuro-fuzzy classifier. Finally, the diagnostic task was sophisticated with the introduction of a new classifier "multi-class neurofuzzy classifier (MC-NFC)". This latter has been implemented on a DSPACE platform "DS1104" to demonstrate its ability to detect and classify faults in real time