Development of Data Driven Model for Proton Exchange Membrane Fuel Cell Using Machine Learning Approaches

The world is moving towards a sustainable source of transportation and energy. Proton exchange membrane fuel cells (PEMFCs) are the major key for that. A multi-input data-based predictive model of PEM fuel cell is built using machine learning approach, such as, linear regression approach, support vector machine (SVM), decision tree (DT) method, random forest (RF) approach etc. Since fuel cell deterioration is very sensitive to operating circumstances, the transfer learning approach is used to accurately forecast fuel cell deterioration for varied activities, particularly in the early stages. The results illustrate the effectiveness and repeatability of the intended strategy. The created voltage prediction digital model is cross-validated using fuel cell real-time voltage collected during operation. This article's goal is to create an early-stage fuel cell voltage prediction digital model that can predict deterioration in real-time and is tolerant to various prediction patterns. The developed models are able to predict the voltage, and RF has the more accurate data model among all.