摘要
An analytical framework for identifying key factors of the degradation of photovoltaic efficiency over time is presented. We demonstrate that, in many photovoltaic experimental settings, reliability data sets are easily cast in a multi- or N-way format. We adopt a statistical technique, N-way partial least squares, that generates a multi-linear model using all of the data simultaneously. With this approach, we are able to model variables of interest such as cell efficiency while representing the data in a lower-dimensional space in which salient features are more easily identified. We illustrate our approach with reliability data for CdS/CdTe heterojunction solar cell devices. Even with the inclusion of a noisy parameter such as the net acceptor density, and with a relatively small number of devices, we automatically identify key factors that are highly related to performance degradation. In particular, the conductance at the back contact is related to short stress-time degradation (0-300h), whereas the net acceptor density near the junction (at +0.08VDC bias) is correlated with more gradual, long stress-time degradation (300-1000h). These notable degradation modes are explained with respect to our processing conditions and Cu-diffusion in the cells.
- 出版日期2015-1