An Efficient Scalogram Generator for Partial Shading Analysis of Photovoltaic Array

Photovoltaic (PV) systems are negatively affected by many troubles, such as conductor faults, broken glass, cell aging, etc. Among them, partial shading conditions (PSC) not only reduce the output power of the PV systems but also cause hot spots on the surface of PV panels. This paper presents a Convolutional Neural Network-based method for PSC detection and classification. The highlights of this proposal are a scalogram extracted from the I-V characteristics of the PV array and a preprocessing phase implemented by a periodic function-based filter and the wavelet transform. The best accuracy of the proposed model is 98.41% through 819 test images after 50 training epochs, which proves its effectiveness. The future improvement of this work is also discussed in the conclusions.