Generative Adversarial Network based Data Augmentation for PV Module Defect Pattern Analysis

In order to solve the problem of insufficient defective images of photovoltaic (PV) modules, more advanced image augmentation technique is required to augment the available dataset. This paper proposed a mixed model consisting of Wasserstein generative adversarial network (WGAN) and a method judging convergence, which based on the PV module defective images obtained by unmanned aerial vehicles (UAVs). The performance of the adopted WGAN is validated in comparison with two existing models through extensive experiments. In addition, the generated images are evaluated by the convolutional neural network (CNN) classifier as a supplementary dataset for the model training process. The numerical result demonstrates the feasibility of the generated images in data augmentation.

[1]  Kilian Q. Weinberger,et al.  An empirical study on evaluation metrics of generative adversarial networks , 2018, ArXiv.

[2]  Mario Luca Bernardi,et al.  A UAV infrared measurement approach for defect detection in photovoltaic plants , 2017, 2017 IEEE International Workshop on Metrology for AeroSpace (MetroAeroSpace).

[3]  Qiang Yang,et al.  Deep Learning Based Module Defect Analysis for Large-Scale Photovoltaic Farms , 2019, IEEE Transactions on Energy Conversion.

[4]  Yoshua Bengio,et al.  Generative Adversarial Nets , 2014, NIPS.

[5]  Mohammadreza Aghaei,et al.  Innovative Automated Control System for PV Fields Inspection and Remote Control , 2015, IEEE Transactions on Industrial Electronics.

[6]  Andrew Zisserman,et al.  Very Deep Convolutional Networks for Large-Scale Image Recognition , 2014, ICLR.

[7]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[8]  Koray Kavukcuoglu,et al.  Pixel Recurrent Neural Networks , 2016, ICML.

[9]  Max Welling,et al.  Auto-Encoding Variational Bayes , 2013, ICLR.

[10]  Léon Bottou,et al.  Towards Principled Methods for Training Generative Adversarial Networks , 2017, ICLR.

[11]  Mario Luca Bernardi,et al.  UAV system for photovoltaic plant inspection , 2018, IEEE Aerospace and Electronic Systems Magazine.

[12]  Soumith Chintala,et al.  Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks , 2015, ICLR.

[13]  Xuanqin Mou,et al.  Low-Dose CT Image Denoising Using a Generative Adversarial Network With Wasserstein Distance and Perceptual Loss , 2017, IEEE Transactions on Medical Imaging.

[14]  Tsuyoshi Murata,et al.  {m , 1934, ACML.

[15]  Qiang Yang,et al.  IET Renewable Power Generation Special Issue: Performance Assessment and Condition Monitoring of Photovoltaic Systems for Improved Energy Yield Visible defects detection based on UAV-based inspection in large-scale photovoltaic systems , 2020 .

[16]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[17]  Lin Chen,et al.  An end-to-end generative adversarial network for crowd counting under complicated scenes , 2017, 2017 IEEE International Symposium on Broadband Multimedia Systems and Broadcasting (BMSB).

[18]  Yoshua Bengio,et al.  Mode Regularized Generative Adversarial Networks , 2016, ICLR.

[19]  Ya Tu,et al.  Digital Signal Modulation Classification With Data Augmentation Using Generative Adversarial Nets in Cognitive Radio Networks , 2018, IEEE Access.

[20]  Mohammadreza Aghaei,et al.  Light Unmanned Aerial Vehicles (UAVs) for Cooperative Inspection of PV Plants , 2014, IEEE Journal of Photovoltaics.