Data-Driven Photovoltaic Module Performance Analysis with FAIR Data

Due to rapid growth of the photovoltaic (PV) market, a huge amount of data is generated everyday. However, networking and data exchange remains challenging due to the scarcity of interoperability and due to the fact that most datasets are stored locally. The "Findable, Accessible, Interoperable, Reusable" (FAIR) Data Principles have been developed to provide guidelines to improve the management of digital assets, so that metadata and data are both human-readable and machinereadable. This work showcases an example where data collected from current and voltage (I - V) measurements of over 1,500 photovoltaic (PV) modules of varying types (e.g., monocrystalline, multicrystalline; aluminium back surface field (Al-BSF), passivated emitter rear cell (PERC), heterojunction technology (SHJ) and interdigitated back contact module (IBC) and histories of exposure (e.g., as manufactured, installed in the field and exposed to accelerated aging) can be FAIRified and stored as linked data. This is an important step towards unified documentation of data with multiple registries. Subsequently, the FAIR data is used to build a data-driven PV module performance analysis model. The I - V characteristics of a PV module carry a huge amount of information. Traditional practice of I- V analysis involves curve fitting with predefined mathematical equations. The simplified equation however could create obstacles in understanding the data and the actual fault and degradation happening within the PV module could be overlooked. Data-driven model on the other hand will be able to extract information that is carried by the real data. This work adapts the previous developed data-driven I- V analysis model and implements additional features to collectively analyse dark I - V curves, light I - V curves and pseudo I - V curves. The integration of data FAIRification and use of datadriven models play an important role in long-term scalability and maintainability in PV research.