Forecasting OPV outdoor performance, degradation rates and diurnal performances via machine learning

Predicting the potential diurnal performance and degradation of organic photovoltaics (OPV) in outdoor conditions is of key interests for users and industrialists. Therefore, machine learning methods are herein employed in order to model and predict the diurnal variation in performance parameters. Subsequently, this allows the expected power output of the modules to be determined. Accurate modelling of the diurnal performance is achieved via a multilayer perceptron algorithm, trained using only the climatic conditions. Furthermore, the degradation rate of the OPV modules is predicted using a separate multivariate regression model. This allows for the main factors that influence the degradation to be found, which in rank order are the 1) Irradiance, 2) Module Temperature 3) Dew point, 4) UV dose, 5) humidity, 6) time, 7) wind speed, in rank order. Using the regression model for degradation, improved understanding of the sources of outdoor degradation is possible.