Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


18ms-thesis-fada.pdf (27.18 MB)

ETD Abstract Container

Abstract Header

Modeling Degradation of Photovoltaic Modules using Machine Learning of Electroluminescent Images

Fada, Justin S, Fada
http://orcid.org/0000-0002-0029-5051
http://rave.ohiolink.edu/etdc/view?acc_num=case1522838106560133

Abstract Details

2018, Master of Sciences (Engineering), Case Western Reserve University, EMC - Mechanical Engineering.
Understanding lifetime performance of field deployed photovoltaic (PV) systems is paramount to the long-term success of the PV industry. Of the many characterization methods available for study of modules undergoing indoor accelerated tests as well as outdoor real-world system monitoring, imaging techniques provide a great volume of high resolution information regarding the health of a module. Electroluminescent (EL) imaging captures the spatial variance of bulk recombination resolved at the wafer front-side surface. Intensity distributions reveal features, such as cracking and corrosion, providing insights into the extent of degradation the module has experienced. Using step-wise imagery, the evolution of degradation has been studied and quantified for five brands of damp-heat exposed commercial modules. Using supervised machine learning, three degradation features were classified to greater than 95% accuracy by training a scalable classification pipeline which can now be used by PV researchers for unbiased, repeatable, and time-efficient feature identification in EL images. For busbar darkening specifically, a lifetime performance prediction model was developed for determining degradation of field deployed modules using only images.This method removes the issue of accelerated indoor to real-world time scaling for power performance prediction. Additionally, current-voltage (I-V) module data has been correlated with the numeric intensity distribution from EL images showing strong correlation between series resistance, maximum power, and wafer darkening. A normalized busbar corrosion width was calculated, which was demonstrated to correlate well with I-V electrical properties, while also providing greater insight to the mechanistic behavior governing I-V characteristics. A composite Feature Ratio metric for busbar darkening was developed and applied to 3 brands of modules. The slope of Pmp vs. the Feature Ratio for brands A, B, and D were -6.8, -42.0, and -3.7, respectively, which trended similarly to the degradation rates of Pmp after 3000 hours of damp-heat exposure with values of -6.0%, -30.4% and -2.7%, respectively, clearly identifying brand B as a poor performing module under damp-heat conditions.
Alexis Abramson (Committee Chair)
Roger French (Advisor)
Jenný Brynjarsdóttir (Committee Member)
Jennifer Braid (Committee Member)
137 p.
Mechanical Engineering

Recommended Citations

Citations

  • Fada, Fada, J. S. (2018). Modeling Degradation of Photovoltaic Modules using Machine Learning of Electroluminescent Images [Master's thesis, Case Western Reserve University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=case1522838106560133

    APA Style (7th edition)

  • Fada, Fada, Justin. Modeling Degradation of Photovoltaic Modules using Machine Learning of Electroluminescent Images. 2018. Case Western Reserve University, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=case1522838106560133.

    MLA Style (8th edition)

  • Fada, Fada, Justin. "Modeling Degradation of Photovoltaic Modules using Machine Learning of Electroluminescent Images." Master's thesis, Case Western Reserve University, 2018. http://rave.ohiolink.edu/etdc/view?acc_num=case1522838106560133

    Chicago Manual of Style (17th edition)

case1522838106560133
1,382
© 2018, all rights reserved.
This open access ETD is published by Case Western Reserve University School of Graduate Studies and OhioLINK.