Convolutional Neural Network for Dust and Hotspot Classification in PV Modules

Abstract

This paper proposes an innovative approach to classify the losses related to photovoltaic (PV) systems, through the use of thermographic non-destructive tests (TNDTs) supported by artificial intelligence techniques. Low electricity production in PV systems can be caused by an efficiency decrease in PV modules due to abnormal operating conditions such as failures or malfunctions. The most common performance decreases are due to the presence of dirt on the surface of the module, the impact of which depends on many parameters and conditions, and can be identified through the use of the TNDTs. The proposed approach allows one to automatically classify the thermographic images from the convolutional neural network (CNN) of the system, achieving an accuracy of 98% in tests that last a couple of minutes. This approach, compared to approaches in literature, offers numerous advantages, including speed of execution, speed of diagnosis, reduced costs, reduction in electricity production losses.

Characterization of Soiling Bands on the Bottom Edges of PV Modules

“…Temperature differences of up to 23 • C have been observed between non-covered cells and cells covered by soiling bands (Lorenzo et al, 2013). Unlike localized overheating caused by semi-homogeneous deposits, hot bands, or bird droppings (Cipriani et al, 2020), the entire area covered by soiling bands exhibits a higher temperature than other areas, which was also observed in our field investigation.…”

Photovoltaic Panels Classification Using Isolated and Transfer Learned Deep Neural Models Using Infrared Thermographic Images

“…In the literature, linear and nonlinear, feature-based, deep network-based classifiers, etc. are extensively used for PV system classification [1,[11][12][13][26][27][28][29]. Linear and feature-based classifiers such as SVM, naïve Bayes, etc., are used to differentiate PV panels based on their health or defects into two and three categories.…”

“…The power loss in a PV system arises due to different factors, including manufacturing defects, transportation, installation, short circuit, open circuit, partial shading, shading, line to line fault, arc fault, bird drops, dust accommodation, module mismatch, environmental degradation, ageing, cell cracking due to mechanical stresses, etc., in real-time operations [ 1 , 8 , 11 , 12 , 13 , 14 , 15 , 16 , 17 ]. Details on PV system defects are provided in Basnet et al [ 18 ].…”

“…Akram et al [ 12 ] developed an isolated light convolution isolated neural network trained on PV system EL thermographs, and they used a transfer learning approach on PV system IR thermography for defect detection with 99.23% accuracy. Similarly, the isolated convolution neural networks to classify IR images based on dust and hotspots with 98% accuracy were proposed by Cipriani et al [ 13 ]. Kurukuru et al [ 28 ] used IR thermographs of eight different panels (one healthy and seven faulty), extracted texture features, and used a scaled conjugate gradient back propagation algorithm to adjust the weight of the neural network classifier and achieved 91.7% testing accuracy.…”

Automatic Inspection of Photovoltaic Power Plants Using Aerial Infrared Thermography: A Review

“…A common cause of hot spots in PV power plants is soiling and shadow over the modules, which hinders the evaluation of results since they are not considered real defects of the PV modules [6]. Cipriani et al [112] approached this issue by using a CNN to differentiate hot spots caused by faults from soiling, obtaining an accuracy of up to 98%.…”