Remaining Useful Life Estimation for Proton
Exchange Membrane Fuel Cell Based on Extreme
Learning Machine
Xiaoling Xue, Yanyan Hu, and Shuai Qi
School of Automation and Electrical Engineering, University of Science and Science and Technology Beijing
Beijing, 100083, China
xuexiaoling@xs.ustb.edu.cn; huyanyan@ustb.edu.cn; qishuai@xs.ustb.edu.cn
Abstract—Remaining useful life estimation (RUL), as an
essential part in prognostics and health management (PHM), has
becoming the hot issue and one of the challenging problem with
the high requirement on the reliability and safety of the
equipment. Extreme learning machine (ELM) is a Single-hidden
Layer Feed-forward Neural Networks (SLFNs) learning algorithm
which is easy to use. As the new generation of fuel cell, proton
exchange membrane fuel cell (PEMFC) is promising in electronic
system. In this paper, we study the RUL of the PEMFC using the
PEMFC dataset in IEEE PHM 2014 Data Challenge. We analyze
the PEMFC degradation trend, at the same time construct the
corresponding degradation model utilizing the ELM and realize
RUL estimation. Finally, the feasibility and effectiveness of the
proposed method are illustrated by a numerical simulation.
Keywords—extreme learning machine; proton exchange
membrane fuel cell; remaining useful life estimation
I. INTRODUCTION
There are higher requirements on reliability and safety in
equipment with the development of the modern industrial
society. From the view of the cost, the electronic system is
changing from periodic maintenance to condition-based
maintenance. As a result, prognostics and health management
(PHM) becomes a hot issue, in which remaining useful life
(RUL) is highly discussed. RUL supplies the most important
information for plant maintenance. It will help us manage the
plant, and improve the sustainability to avoid the heavy loss
caused by the abrupt equipment performance degradation [1].
As one of the central component of the system, cell is used
for energy supplying. Meanwhile, power-supply-mode based on
cell has being seen in many fields varies from cellphone to
aerospace and regarded as the core unit. For this reason, it may
cause catastrophic fault when cell degradation happens.
Therefore by constructing degradation model, utilizing RUL,
having health management, we will highly improve the
feasibility and reliability of the system. In mechanical
equipment, we call RUL as the length of time from the operating
time (t) to the broken time (T), which means the RUL is a
random variable, and is dependent on the current operating state
of the plant, the operating environment and the state information
observed at present (seen from Fig. 1).
Usually, RUL has two classes: physically-based method and
data-driven-based method [2]. Physically-based method needs
fully constructed mathematical model to describe the physical
property and failure mode of the plant [3][5]. While data-
driven-based method needs little priori knowledge
(mathematical model and expertise) of the system. Dependent
on the condition-based inspecting data (history data), it uses
various data analysis methods to find the implicit information for
RUL [6]. On account of the complicated physical model of the
things to study, as well as the external disturbance such as
temperature and press, it is hard to construct an equation of state
based on the established physical model, thus the data-driven-
based method has more advantage.
Extreme Learning Machine (ELM) is a new learning
algorithm aiming at Single-hidden Layer Feed-forward Neural
Networks (SLFNs). Proposed by Huang et al. in 2004, it has the
advantage of strong learning ability and fast training rate [7].
Thus the ELM is wildly used in prediction field. Singh et al. use
ELM in time series analysis to predict the trend of the time series
[8]; a wavelet decomposition and wind power prediction model
based on multi-model ELM is devised by Huang et al. in [9]
making use of the ELM in wind power detection; Moreover, the
ELM has been improved and used in RUL estimation by Javed
[10].
Proton exchange membrane fuel cell (PEMFC) is based on
the contradictory equipment of the water electro-analysis. With
the merit of the low operating temperature, high power
generating efficiency, little noise pollution, flexible use and easy
maintenance, the PEMFC has attracted the developed country
and famous companies, including general motors, Toyota and
Daimler Chrysler, which are now developing electro-mobile to
realize commercialization development [11]. Fig. 2 is the
PEMFC working schematic diagram [12].
PEMFC is now regarded as the most promising alternative
energy sources. The general life of PEMFC is 2000 hours, while
many applications require a longer life. In consequence, the
URL estimation for PEMFC can fully get the degradation
phenomena and have a positive effect on the development of the
PEMFC. In addition, the URL estimation can effectively avoid
massive loss.
R. K. Jaworski constructs parametric statistical models for
predicting the cell life [13]; electrochemical impedance
This work was supported in part by the National Natural Science Foundation
of China under Grant 61304105, and in part by the Fundamental Research Funds
for the Central Universities with Grant FRF-TP-15-060A2.
VW<RXWK$FDGHPLF$QQXDO&RQIHUHQFHRI&KLQHVH$VVRFLDWLRQRI$XWRPDWLRQ
:XKDQ&KLQD1RYHPEHU
,(((