ANOVA method applied to proton exchange membrane fuel cell ageing forecasting using an echo state network

摘要

This paper presents the results of a sensibility analysis applied to a Reservoir Computing method. The method ANOVA has been chosen in order to analyse the influence of the different parameters of a new kind of Neural Network paradigm: the Echo State Network. The three sets of parameters used correspond to the more influential parameters in the case of the fuel cell voltage ageing forecasting. The simulations are then compared with experimental data obtained after a 1700 h long duration test. The results obtained, with a Mean Average Percentage Error of less than 5%, prove that Echo State Network is an interesting and promising tool.