An adaptative approach for estimating the remaining useful life of a heavy-duty fuel cell vehicle
Nicolas Morizet, Perceval Desforges, Christophe Geissler, Elodie Pahon, Samir Jemeï, Daniel Hissel
Abstract
The aim of this paper is to predict the remaining useful lifetime without prior knowledge of the load profile of a heavy duty vehicle. To achieve this, the data is indexed by cumulative current load instead of time. Furthermore, the stress load profile, that represents the different operation modes of a fuel cell on a heavy-duty fuel cell vehicle, is split into several ranges of current. Then, the remaining useful life in each cluster is predicted with echo state networks. These predictions are afterwards combined to compute the global useful remaining life. The approach could be generalized to any type of fuel cell regardless of the load profile with a high accuracy highlighted by three different performance metrics.