Research on surface defect detection method for photovoltaic panels based on MSP2-YOLOv11

Abstract

Solar photovoltaic technology is experiencing a period of rapid growth, and the share of photovoltaic systems in the overall energy structure is steadily increasing. However, due to their widespread distribution and the inherent difficulties associated with their maintenance, surface defects on PV panels—such as cracks, dust accumulation, and foreign object occlusion—are often difficult to detect in a timely manner, thereby severely compromising power generation efficiency and system safety. To address the aforementioned challenges, this study designs a photovoltaic panel defect identification algorithm based on an optimized YOLOv11 architecture. By incorporating a Multi-Scale Dilated Attention (MSDA) mechanism, optimizing the loss function, and adding a detection layer specifically for small objects, the MSP2- YOLOv11 model is constructed. Experimental data fully validated the reliability of the model in complex environments, and its detection sensitivity for small target defects reached a high level. The relevant technical solutions help to achieve automated inspection and fault early warning of photovoltaic power plants, thereby improving operation and maintenance efficiency.