Remaining useful life prediction for lithium-ion batteries using particle filter and artificial neural network
| Date | 09 October 2019 |
| Pages | 312-328 |
| Published date | 09 October 2019 |
| DOI | https://doi.org/10.1108/IMDS-03-2019-0195 |
| Author | Wei Qin,Huichun Lv,Chengliang Liu,Datta Nirmalya,Peyman Jahanshahi |
| Subject Matter | Information & knowledge management |
Remaining useful life prediction
for lithium-ion batteries
using particle filter and
artificial neural network
Wei Qin, Huichun Lv, Chengliang Liu and Datta Nirmalya
Shanghai Jiao Tong University, Shanghai, China, and
Peyman Jahanshahi
Nobleo Technology, Eindhoven, The Netherlands
Abstract
Purpose –With the promotion of lithium-ion battery, it is more and more important to ensure the safety usage of
the battery. The purpose of this paper is to analyze the battery operation data and estimatet heremaining life of the
battery, and provide effective information to the user to avoid the risk of battery accidents.
Design/methodology/approach –The particle filter (PF) algorithm is taken as the core, and the
double-exponential model is used as the state equation and the artificial neural network is used as the observation
equation. After the importance resampling process, the battery degradation curve is obtained after getting the
posterior parameter, and then the system could estimate remaining useful life (RUL).
Findings –Experiments were carried out by using the public data set. The results show that the Bayesian-based
posterior estimation model has a good predictive effect and fits the degradation curve of the battery well, and the
prediction accuracy will increase gradually as the cycle increases.
Originality/value –This paper combines the advantages of the data-driven method and PF algorithm. The
proposed method has good prediction accuracy and has an uncertain expression on the RUL of the battery.
Besides, the method proposed is relatively easy to implement in the battery management system, which has
high practical value and can effectively avoid battery using risk for driver safety.
Keywords Artificial neural network, Neural network, Particle filter, Lithium battery,
Remaining useful life, State of health, Double-exponential model
Paper type Research paper
1. Introduction
For different engineering application areas, various uncertainties are inevitable during the
implementationof each project, which leads to somerisks and accidents, so risk management
gets more attention. Project risk management uses various risk response measures,
management method to effectively control the risks of the project, and it ensures the overall
goal of the projectwith the least cost. Cooper et al.,(2005) proposed that all engineeringproject
risk managementincludes the following processes: establishing risk factorsets to accomplish
risk identification, risk analysis, risk assessment and risk response.
For new energy vehicles, due to the characteristics of high energydensity and long service
life, lithium-ion batteryplays an irreplaceable rolein the field of consumerelectronics and new
energy vehicles(Etacheri et al., 2011). At thesame time, the safety and reliability of lithium-ion
batteries have also gotten more and more attention, and the failure of the battery can easily
lead to other accidents and losses, which is mainly reflected in (Fotouhi et al., 2016):
(1) most of the electrolytes are organic combustible, and it may lead to the increase of
battery temperature, fire or even explosion; and
(2) overcharge and overdischarge will lead to changes in the material properties of the
battery, resulting in irreversiblecapacity loss and remaining useful life (RUL) decline,
and once the battery exceeds the life threshold, there will be great risks for users.
Industrial Management & Data
Systems
Vol. 120 No. 2, 2020
pp. 312-328
© Emerald PublishingLimited
0263-5577
DOI 10.1108/IMDS-03-2019-0195
Received 30 March 2019
Revised 21 May 2019
22 July 2019
Accepted 28 August 2019
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/0263-5577.htm
312
IMDS
120,2
For the electric vehicle battery, the reduction of cost and the improvement of safety
performance are the determinants for the large-scale application of electric vehicles. The life
test specification and life prediction of power batteries are important parts of power battery
technology. The RUL means how long the battery can be used in the current vehicle state,
which is a very important issue in the safety of the battery. As a microsystem, lithium
battery also needs a management system to monitor the risk of battery usage, so the battery
health management system (BMS) has been proposed by some scholars, which is the link
between the battery and the user and can greatly improve the practicability, economy and
safety of the lithium battery (Lu et al., 2013).
Generally, there are two important functions of BMS: estimation of state of charge (SOC) and
state of health (SOH). The SOC mainly predicts the remaining usable charge in the current cycle
of the battery (Andre et al., 2013). The SOH predicts the health status of the battery, which is
closely related to the RUL of the battery (He et al., 2011). According to the SOH prediction of the
battery, the BMS could get the RUL and the user or the battery manufacturer can get the
notification of replacing the battery in time. There are two main ways to define SOH. One is
definedbythebatterycapacity,andtheotherisdefined by the internal resistance of the battery
(Sepasi et al., 2015). This paper takes the definition of battery capacity, SOH is equal to the ratio
of the battery discharge capacity to the new battery nominal capacity under certain conditions
(Berecibar et al., 2016). The failure threshold for lithium-ion batteries is defined as the time
(or cycle) when the battery capacity falls below 80 percent of its initial rated value (Wang et al.,
2016). The RUL prediction of a lithium-ion battery corresponds to the prediction of the remaining
time or cycle before its capacity reaches the failure threshold (Xing et al., 2013). However, the
aging rate of the battery cannot be measured, and the battery is a non-linear system. In addition,
the use of the battery is greatly affected by the working environment. The main factors are
temperature, charge and discharge cutoff voltage, charge and discharge current and so on, all of
them have a great influence on the RUL prediction of the battery.
Lithium battery RUL prediction methods can be mainly divided into model-based,
data-driven and fusion model. Most of the model-based prediction methods use the equivalent
circuit model (Kim et al., 2014; Andre et al., 2013), which builds the circuit model composed of
electrical components based on the operating principle of the system, and makes the RUL
prediction by approximation of the dynamic characteristics of the system. However, in the
approximation process, the implicit relationship between some parameters that have a decisive
effect on the system characteristics may be neglected. It is difficult to consider all the complicated
external conditions, which results in the comprehensive ability of the equivalent circuit model to
describe the dynamic and static characteristics of the battery weekly (Liu et al., 2015). Another
model-based prediction method is the empirical degradation model. These methods generally use
the random filtering algorithm to track the battery degradation information, obtain the optimal
parameters of the battery empirical degradation model and establish the empirical degradation
model to predict the RUL. These filtering algorithms include particle filtering algorithms (Qiang
et al., 2013; Miao et al., 2013), unscented particle filter (PF) algorithm (Qiang et al., 2013), etc.
However, under the condition of many parameters, the initialization of empirical degradation
model takes much time, which affects the real-time performance of the algorithm (Xian et al.,
2014). The data-driven approach mainly uses the data collected during current and previous data
in the charge and discharge cycles to predict the battery RUL. These data are charge and
discharge current, voltage, temperature, internal resistance, time and battery capacity, etc. The
mainstream data-driven RUL prediction methods mainly include AR/ARMA model (Saha et al.,
2009), artificial neural network (ANN) (Zhang et al., 2018), support vector machine (SVM) (Patil et
al., 2015), Gaussian process regression (Liu et al., 2013) and so on. The data-driven prediction
methods have good performance in both real-time and prediction accuracy, but they lack of
uncertainty expression. The fusion prediction method uses the above two kinds of methods to
improve the accuracy and stability of the prediction results through weighting or other fusion
313
Life prediction
for lithium-ion
batteries
To continue reading
Request your trial