Abstract
With the increasing integration of photovoltaic (PV) systems, challenges
in condition monitoring, fault diagnosis, and modeling of PV plants have
intensified. To address the cluster division aspect of these challenges,
a novel and efficient method using an improved spectral clustering (SC)
approach is proposed. The approach starts with preprocessing the PV data
to normalize variations in magnitude. The clustering process is then
executed using the modified SC algorithm. To validate its effectiveness,
comparative analyses are conducted using various clustering indices
across different test cases. To comprehensively evaluate the performance
of the approach, a comprehensive index is proposed. The results confirm
that the method significantly enhances the efficiency and speed of
cluster division in large-scale PV plants, making it a promising tool
for PV system management.