Reliability Assessment of Grid Connected Solar Inverters in 1.4 MW PV Plant from Anomalous Classified Real Field Data

In this work, a top-down analysis is carried out to investigate the impacts of environmental factors on the health, and hence on the reliability, of solar inverters (SI). Five years of real field data from 46 string inverters in a 1.4 MW Photovoltaic (PV) plant located at Florida International University (FIU) are used for the analysis. Collected data is classified and examined based on inverter faults, failures, and stress conditions using the classification and regression tree (CART) algorithm. Results have shown that inverter performance is highly correlated to ambient conditions, i.e. sunrise and sunset timing, relative humidity, and irradiance profile, and therefore adequate specific ventilation management can be a useful tool to mitigate some major inverter health issues. Triggered by this study, a prognostic analysis from the information in service tickets and machine learning (ML) outcomes will be carried out as future work.