Spatiotemporal Modeling of Real World Backsheets Field Survey Data: Hierarchical (Multilevel) Generalized Additive Models

Assessing photovoltaic module backsheet durability is critical to increasing module lifetime. Laboratory-based accelerating testing has recently failed to predict large scale failures of widely adopted polymeric materials. Additionally, there is a growing concern on characterizing the non-uniformity...

Full description

Saved in:

Bibliographic Details
Published in:	2022 IEEE 49th Photovoltaics Specialists Conference (PVSC) pp. 0255 - 0260
Main Authors:	Wieser, Raymond J., Li, Zelin Zack, Moffitt, Stephanie L., Zabalza, Ruben, Boucher, Evan, Ayala, Silvana, Brown, Matthew, Gu, Xiaohong, Ji, Liang, O'Brien, Colleen, Hauser, Adam W., O'Brien, Greg S., Yu, Xuanji, French, Roger H., Kempe, Micheal D., Tracy, Jared, Choudhury, Kausik R., Gambogi, William J., Bruckman, Laura S., Boyce, Kenneth P.
Format:	Conference Proceeding
Language:	English
Published:	IEEE 05-06-2022
Subjects:	Additives Analytical models Backsheet Data models Degradation Field Survey Modeling Photovoltaic systems Predictive models Spatio-temporal Stability criteria
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Assessing photovoltaic module backsheet durability is critical to increasing module lifetime. Laboratory-based accelerating testing has recently failed to predict large scale failures of widely adopted polymeric materials. Additionally, there is a growing concern on characterizing the non-uniformity of field exposure. Therefore, data from field surveys are critical to assess the performance of component lifetimes. Using a documented field survey protocol, 19 field surveys were conducted. The focus of this survey strategy is to investigate spatial continuity in degradation modes. By combining field survey data with real-time satellite weather data, stressor / response models have been trained. Generalized additive Models (GAM) model was created to predict the value of degradation based on measured predictors. Two different GAM constructions were testing using different implementations of basis splines. The model includes variables on the environmental stressors of the system and the location of each measurement in the PV mounting structure. The incorporation of hierarchical structure into the models allowed for material specific degradation rates, while maintaining the assumption of a global trend. The model performed well with an adjusted R^{2} of 0.975 for yellowness index prediction.
DOI:	10.1109/PVSC48317.2022.9938576