Application of machine learning to map Photovoltaic-Solid Oxide Fuel Cell (PV-SOFC) systems

Growing demand for energy paired with the environmentally fuelled emphasis on the transition to more sustainable methods of energy production has led to increased research, development and investment in renewable energy systems. This study aims to examine the feasibility of applying machine learning algorithms to map Photovoltaic-Solid Oxide Fuel Cell (PV-SOFC) systems. This study applies different machine learning models to a PV-SOFC system to accurately predict the hydrogen flow rate, levelised cost of product, system efficiency and CO2 emissions index system parameters. The prediction models considered in this study are Linear Regression, Random Forest, Seasonal AutoRegressive Integrated Moving Average with eXogenous regressors (SARIMAX) and Gradient Boosting models. The key evaluation metrics used to measure model accuracy and performance were R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> and RMSE. The results of the study determine the Random Forest prediction model to be the most accurate predictor for the majority of system parameters. In this study, various train/test split ratios were examined for the Random Forest and the Gradient Boosting models. The results demonstrate that the proposed prediction models successfully predicted the system parameters of a PV-SOFC and can be used to optimise these parameters.