Physics-based computational modeling is an essential aspect of research and development of new photovoltaic (PV) technologies. For example, PV modeling allows us to improve our understanding of device physics, evaluate the potential of new device architectures prior to experimentation, and predict the energy production of PV systems. This dissertation describes the development of improvements to PV modeling at both the device and system levels, as well as the application of these models to obtain new insights into device physics and PV system performance.The first half of this thesis focuses on solar cell modeling using the drift-diffusion approach. Investigations into numerical instabilities of a drift-diffusion model for bilayer organic solar cells are described, and techniques for improving convergence behavior are presented. The drift-diffusion approach is then adapted for planar perovskite solar cells, validated against literature results, and utilized to understand a new experimental result involving the impact of ultrathin fullerene layers on device performance. The second half of this thesis expands our modeling to PV systems. Here, we focus on the effects of temperature and soiling (accumulation of particulate matter on PV module surfaces) on PV system energy production. A general semi-physical model for predicting annual PV soiling losses is developed and integrated with open-source PV performance models and commercial algorithms for PV system cleaning schedule optimization. Additionally, the potentials of machine learning approaches to soiling modeling are discussed and a proof-of-concept is demonstrated. In order to consider the impact of temperature on PV module performance, a coupled thermal-electrical modeling approach is developed by combining temperature-dependent equivalent circuit models with a commercial heat transfer solver. This approach also allows for predicting energy production of PV modules or films installed on irregular surfaces, such as vehicles. Application of the coupled thermal-electrical approach to simulation of residential rooftop and vehicle-integrated PV systems is demonstrated. Overall, this work has resulted in an improved understanding of the numerical methods necessary to ensure stability of drift-diffusion codes, insight into the role of fullerenes in perovskite solar cells, and the development of modeling approaches than can aid in PV system engineering and improve accuracy in PV system energy production forecasting.
Read
