.png){kind=icon}
.png){kind=icon}
Return
AI-Assisted hotspot detection and Optimal SuDoKu reconfiguration for photovoltaic arrays under partial shading conditions
.png){kind=icon}
PRE
AI
K
A
DOI：10.1063/5.0322474.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Partial shading of photovoltaic (PV) arrays leads to mismatch losses, multiple P-V peaks, and hotspot formation, causing reduced energy yield, long-term degradation, and potential safety hazards. This study proposes an integrated AI-assisted hotspot detection and optimal PV reconfiguration framework to address both thermal and electrical performance issues in real-time. The hotspot detection module employs an OpenCV-based preprocessing pipeline with median filtering, Contrast Limited Adaptive Histogram Equalization enhancement, and K-means segmentation, followed by YOLOv8 deep learning inference, achieving a detection accuracy of 94% on thermal datasets. Upon hotspot identification, an irradiance-aware Optimal SuDoKu Total Cross-Tied (TCT) reconfiguration was implemented in MATLAB/Simulink, dispersing shading across the array while minimizing wiring losses. The simulation results demonstrate up to 28% reduction in resistive losses, 15%-25% improvement in global Maximum Power Point (MPP) power output, and similar to 30% mismatch loss mitigation compared to conventional TCT under varied shading scenarios. The proposed hybrid AI-electrical approach enables scalable autonomous PV monitoring and optimization by combining predictive maintenance capabilities with enhanced energy harvesting under partial shading conditions.
Journal
IF:
1.9
Papers:
373
Citations:
4.4K
Organization
Cited Papers