Life prediction study of proton exchange membrance fuel cell for vehicle 2019-01-0385
In order to know whether the vehicle’s proton exchange membrane fuel cell (PEMFC) meets the future requirements of vehicle conditions, this paper proposes a model for predicting the life of the vehicle PEMFC.
Firstly, the voltage outputs of PEMFC stack are be taken as the PEMFC life evaluation index; two methods are be used to establish the life decline evaluation criteria of the PEMFC stack: 1) Least Squares Fitting (LSF) method establishes the standard for stack voltage degradation behavior. 2) The Back Propagation (BP) neural network learns the stack’s voltage decay characteristics and establishes the standard for stack’s voltage degradation behavior.
Secondly, the Autoregressive Moving Average (ARMA) time series model learns part of the known stack’s voltage decay behavior to predict unknown stack’s lifetimes.
Finally, in order to verify the accuracy of ARMA prediction, combined with the operational data of the vehicle PEMFC, the prediction results are compared with the established two voltage decay standards and the real voltage output. It can be seen that the ARMA prediction accuracy meets the operational requirements of the vehicle PEMFC. After obtaining the real PEMFC’s voltage decay data within the ARMA prediction range, the PEMFC life prediction for the next period can be performed by using ARMA.