Fuel Cell Performance Prediction Using an AutoRegressive Moving-Average ARMA Model

In economy, finance, transport, and many other fields, a set of sequentially ordered observations can be recorded, called chronological or time series. The analysis of these time series make it possible to understand their dependencies and their time evolution for various purposes. In order to carry out the good corrective actions to improve the studied system, the times series prediction is a good solution and one of its main purposes. Fuel cells integrated in a hybrid electrical vehicles are subject to degradations that affect its performance. In order to limit these degradations, the fuel cell performance must be estimated during its lifetime through the prediction of its degradation. This study focuses on the analysis of times series representing observations on a fuel cell stack voltage, which is used as a degradation indicator for prognosis purpose. An Auto Regressive Moving Average model is used to model and predict the degradation of the fuel cell stack voltage. To obtain an accurate prediction, a suitable model must be chosen. In this paper, the concept of time series analysis is introduced as well as the information criteria used to choose the appropriate Auto-Regressive Moving Average model. The chosen model has been adjusted on 30%, 40% and 50% of the time series in order to check its reliability. The prediction results give an accurate estimation of the fuel cell voltage in the long term forecasting.