RESIDUAL LIFE PREDICTION OF HIGH POWER PROTON EXCHANGE MEMBRANE FUEL CELL FOR VEHICLE USING.

The service life of proton exchange membrane fuel cell (PEMFC) is an important breakthrough direction in the field of vehicle fuel cell, but the current life prediction method has the problem of low accuracy and efficiency. To solve this problem, this paper proposes an improved CNN network model based residual life prediction method for vehicle high-power proton exchange membrane fuel cells. In this method, the improved dropout and shortcut units are used to improve the convolutional neural network (CNN) network to solve the over fitting problem of prediction network training and improve the overall performance of PEMFC life prediction network model. At the same time, in order to realize the realtime adaptability of the prediction network, Adam optimization algorithm is used to minimize the loss function to support the fast tracking performance of PEMFC life prediction network. Finally, the prediction experiment is carried out based on the general data set. The experimental results show that the prediction error of the proposed method is guaranteed to be 1.5 h, and the maximum relative error is 0.473%. [ABSTRACT FROM AUTHOR]