Advanced Detection and Prediction of Defects in Photovoltaic Power Plants Using Artificial Intelligence and Smart Sensors: A Comprehensive Review

Abstract

The rapid expansion of photovoltaic (PV) power installations worldwide has made reliable operation and maintenance of these systems critically important. While PV systems offer significant environmental and economic advantages, their long-term performance is continuously threatened by a wide range of defects and faults that can silently reduce power output and shorten system lifetime. This paper presents a comprehensive review of defect detection and fault diagnosis methods for PV power plants. A structured taxonomy of defects is first presented, covering physical defects such as cracks, delamination, and corrosion; environmental defects including partial shading, soiling, and hotspots; and electrical faults on both the DC and AC sides of the system. The main fault detection approaches are then reviewed, encompassing electrical-based methods, visual and thermal inspection techniques, Artificial Intelligence (AI) and machine learning algorithms, and IoT-enabled remote monitoring platforms. The growing role of AI in enabling automated and scalable fault diagnosis is highlighted as a central theme. Failure prognosis strategies and the key challenges facing the field are also discussed. The findings provide a consolidated reference for researchers and practitioners working toward more intelligent and cost-effective PV monitoring solutions. Unlike prior reviews that address imaging, electrical, or deep-learning methods in isolation, this work uniquely integrates all detection families electrical, visual-thermal, AI-based, and IoT-enabled - alongside failure prognosis and a structured defect taxonomy, providing a unified and actionable framework for the field.