Invited Paper IPF+MS-TuM1
Mesoscale Evolution & Temporal Analytics of Photovoltaic Energy Materials: A Degradation Science Approach
Tuesday, October 20, 2015, 8:00 am, Room 210F
Degradation science1 combines physical and statistical approaches to examine degradation mechanisms and pathways of a material or system in order to improve materials and reduce system failures by incorporating modeling, monitoring, and prediction of lifetime performance. Degradation of PV modules evolves over long time-frames and length scales, which is a characteristic of mesoscale science. Degradation arises due to the distinct, complex, and interactive phenomena which lead to failure. Real-world studies under diverse environmental conditions must be combined and cross-correlated with accelerated in-lab studies, using data science and analytics methodologies, so as to span the time and length scales that control the system’s behavior over lifetime. Semi-supervised generalized structural equation (semi-gSEM) modeling can be used to relate physical mechanistic submodels with data-driven statistical submodels as networks of mechanisms and modes with statistically significant pathway relationships. The relationships and coupling strengths (ij) amongst variables can be rank-ordered in their contributions to the system’s degradation. Temporal evolution, damage accumulation and change points among mechanisms/modes (variables) are accounted for in the semi-gSEM models. Towards these goals, a statistical methodology has been developed and applied to investigate the response of full sized PV modules to accelerated stress conditions. The results of this initial study indicate that a correlation exists between system level power loss and the buildup of acetic acid resulting from the hydrolytic degradation of EVA polymer encapsulant. To further explore this proposed mechanistic pathway, studies are underway to characterize the degradation of minimodule samples under a broader range of similar multifactor accelerated stress conditions. Sample types feature frontside silver gridlines of two different widths and exposure conditions vary in irradiance level and temperature. Samples are measured non-destructively at many points along their lifespan, using confocal Raman microscopy to capture chemical signals and various techniques to gather electrical performance information, with the goal of observing the co-evolution of EVA degradation and gridline corrosion. This represents an important first step towards exploring the often misunderstood role of EVA degradation in PV module performance loss, and building a more integrated picture of PV module degradation as a whole. Initial data analytics of six months’ real-world performance data of 60 c-Si PV modules on the SDLE SunFarm shows deviation of performance ratio among modules at the same geometric location. Grouping of samples with similar performance patterns was performed with hierarchical clustering, K-means clustering was used to confirm the optimum number of clusters. A brand dependent module performance model was developed based on a subgroup of 21 modules from 7 manufactures. Over 1.5 million I-V curves measured every 5 minutes for 500 days on 10 modules with/without mirror on dual-axis trackers were analyzed use an automated analytic functions we developed. Maximum power point, open circuit voltage, short circuit current, slope of the curve near open circuit voltage, and slope of the curve near short circuit current are either directly extracted or estimated from measured I-V curves. An algorithm based on moving local regression model was developed to detect the change points on I-V curves, which caused by bypass diode turning on when I-V curve was measured under non-uniform irradiance. These examples of the use of degradation science, with its physical and statistical foundation and data analytics approach, will hopefully enable the community to address the long-term reliability uncertainty of photovoltaics as they become a major component in the world’s energy systems.
1. French, Roger H., Rudolf Podgornik, Timothy J. Peshek, Laura S. Bruckman, Yifan Xu, Nicholas R. Wheeler, Abdulkerim Gok, et al., 2015, “Degradation Science: Mesoscopic Evolution and Temporal Analytics of Photovoltaic Energy Materials,” Current Opinion in Solid State and Materials Science, Doi: 10.1016/j.cossms.2014.12.008