In photovoltaic (PV) power plants, quickly finding faults is crucial for identifying what is causing them and fixing major problems to maintain good efficiency. Many studies have used drones to inspect PV plants, but these drone-based methods usually struggle to address dangerous issues. To solve this latters, this work introduces a smart system for finding damaged PV panels using a drone equipped with two cameras (one thermal, one digital). The drone does regular checks and uses both cameras to take pictures. The digital camera finds problems caused by outside factors, while the thermal camera shows problems inside the panels. A Convolution-Neural-Network method is used at the processing unit to analyses pictures and finds the problems. This study looks at four different states (normal, shadowed, dust, & thermal-faults). The developed method is very accurate (over 0.8), especially when using Euler angle (0.98), showing it can effectively find and locate problems in PV plants.