Qualitative Analysis of Dusting and Soiling in solar PV System using Learning Algorithms

Nowadays the demand for renewable energy is being increased, solar plays a major role in supplying the energy demand. However, there are several environmental parameters like humidity, rainfall, high temperature, dust, air pollution, wind speed, snow, aging, and so on block the demonstration and efficiency of solar PV schemes. But particularly the dust buildup on the PV modules has an impact that is negative in various ways. Because it reduces the spectral transmittance which leads to reduced power output and efficiency. Sometimes, it damages the solar panels and the lifespan gets minimized. But there are solutions to prevent dust deposition by periodic cleaning which allows the panels to regain their function and perform to produce better outputs. This paper aims to classify panels as clean and light dusted, medium dusted, and heavy dusted. Using image processing the percentage of the dust that is deposited can be found by differentiating the dusted area in the panel and finding the pattern of dust accumulation related to locality is also analyzed. Convolution Neural Networks and Binarization can determine the accuracy of dust deposited in the panels. Though this paper has used certain algorithms, Otsu’s thresholding concepts resulted in more accuracy of about 89%. It automatically finds the optimal threshold intensity by classifying the image into two classes using a class variance. Thus, the system preserves the economic value of the asset.