Innovative Approach for Yield Evaluation of PV Systems Utilizing Machine Learning Methods

PV plant owners and O&M providers heavily rely on the flawless and reliable operation of their PV systems. Unscheduled maintenance interventions due to incorrect defect or error messages as well as unpredicted shading due to plants or soiling lead to high yield losses and unnecessary additional operation costs, which adversely affect the profitability of the PV system. This paper presents innovative methods from machine learning to analyze monitoring data. Various approaches such as artificial neural networks and clustering of multi-dimensional data will be introduced exemplarily and it will be shown how they can be used for detection and identification of defects and degradation effects. Assumptions, databases, working methods and results will be presented in this work to show the effective utilization of machine learning methods in big-data evaluation of PV systems.