As every industrial process, a photovoltaic system may subject, during his operation, to various faults and abnormalities leading to a drop of its efficiency up to its unavailability. A precise diagnosis and a fault detection and isolation make it possible to reduce the maintenance costs and above all increase the yield. In this work, a focus has been made on the fault detection and isolation in the DC part of the PV system, it means of the PV array. The goal of this work is to propose, by conducting the less possible measurements to meet the economic constraints, an algorithm to detect an isolate the faults causing a drop in the array yield. To achieve this goal, the analysis of the I-V characteristic has been chosen. This analysis leads to the use of the inference method to conduct the diagnosis of the PV plant. This procedure has been validated by on site experiments, real-time simulations and non real-time simulations.
