Defects Inspection in Polycrystalline Solar Cells Electroluminescence Images Using Deep Learning

Abstract

Solar cells defects inspection plays an important role to ensure the efficiency and lifespan of photovoltaic modules. However, it is still an arduous task because of the diverse attributes of electroluminescence images, such as indiscriminative complex background with extremely unbalanced defects and various types of defects. In order to deal with these problems, this paper proposes a new precise and accurate defect inspection method for photovoltaic electroluminescence (EL) images. The proposed algorithm leverages the advantage of multi attention network to efficiently extract the most important features and neglect the nonessential features during training. Firstly, we designed a channel attention to exploit contextual representations and spatial attention to effectively suppress background noise. Secondly, we incorporate both attention networks into modified U-net architecture and named it multi attention U-net (MAU-net) to extract effective multiscale features for defects inspection. Finally, we propose a hybrid loss which combines focal loss and dice loss aiming to solve two problems: a) overcome the class imbalance problem, and b) allowing the network to train with irregular image labels for some complex defects. The proposed multi attention U-net is evaluated on real photovoltaic EL images datasets using 5-fold cross validation technique. Experimental results demonstrate that the proposed network can segment and detect various complex defects correctly. The proposed method achieved the mean intersection over-union (m-IOU) of 0.699 and F-measure of 0.799 which outperforms the previous methods.

Investigating the Impact of Cracks on Solar Cells Performance: Analysis Based on Nonuniform and Uniform Crack Distributions

“…1(a), represents a valuable means to detect the location of cracks over solar cells. It is nondestructive and moderately fast, with computation times varying from 1 ms to a few seconds [9]. The EL system comprises a black environment to minimize the lights absorption whilst taking the EL images.…”

“…For =1000 W/m 2 , the power loss amounts to 2.55 and 2.02 W for the nonuniformly-and uniformly-cracked cells, respectively. The output power loss of both solar cells is calculated using (8) and (9).…”

“…They also outlined that the percentage of the output power loss ranges from 0.2% to 12%, depending on the crack size. A similar deep learning model was also proposed by Rahman & Chen [9]; they have developed a multi-attention U-net (MAU-net) algorithm for solar cell cracks identification with a calibration time of 75 ms, excluding the EL imaging time.…”

Deep learning‐based automatic detection of multitype defects in photovoltaic modules and application in real production line

“…The purpose of automatic detection is to replace the manual inspection in production line, and it has two requirements: (1) different types of defects should be concerned, and (2) every single defect should be localized and classified, which is essentially an object detection task. But current researches only studied crack, break, and finger interruption 4–16 and cannot handle localization problem well for multitype defects, which is what we aimed to achieve in this paper. We sum up our main contributions as follows: (1) we gathered 5983 EL images of defective modules and labeled all of them, with 19 categories of defects found and introduced.…”

“…Hence, many researches aiming to achieve automatic detection of defects in EL images have been done in the past decade. These studies can be divided into two groups according to their approaches: using conventional signal processing algorithms 4–8 and using artificial intelligence (AI) techniques 9–16 …”