HV Battery SOC Estimation Algorithm Based on PSO
Qiang Sun
1
, Haiying Lv
1
, Shasha Wang
2
, Shixiang Jing
1
, Jijun Chen
1
and Rui Fang
1
1
College of Engineering and Technology, Tianjin Agricultural University, Jinjing Road, Tianjin, China
2
Beijing Electric Vehicle Co., Ltd., East Ring Road, Beijing, China
sunqiang@tjau.edu.cn, wangshasha@bjev.com.cn
Keywords: HV battery, SOC, PSO, individual optimum, global optimum.
Abstract: In view of the traditional algorithm to estimate the electric vehicle battery state of charge (SOC), lots of
problems were caused by the inaccurate battery model parameter estimation error. Particle swarm
optimization (PSO) algorithm is a global optimization algorithm based on swarm intelligence, it has
intrinsic parallel search mechanism, and is suitable for the complex optimization field. PSO is simple in
concept with few using of parameters, and easily in implementation. It is proved to be an efficient method to
solve optimization problems, and has successfully been applied in area of function optimization. This article
adopts the PSO optimization method to solve the problem of battery parameters. The paper make a new
attempt on SOC estimation method, applying PSO to SOC parameters optimization. It can build
optimization model to get a more accurate estimate SOC in the case of less parameters with faster
convergence speed.
1 INTRODUCTION
Lithium battery has many advantages such as high
energy density, high power density, long cycle life
and so on. As the energy storage element of electric
vehicles, it is used widely. In order to improve the
safety and reliability of battery pack, fully play its
efficiency, and prolong its life, the battery pack must
be managed effectively in using. Battery state of
charge (SOC) estimation is the important contents
and basis of battery management system (BMS), but
the battery is a kind of nonlinear uncertain system,
so it is very difficult to estimate battery SOC
accurately (Ramadass, 2003).
The SOC of the battery is refers to the amount of
residual capacity of charged battery, SOC estimation
is the basis of the battery thermal management,
balance management, and safety and reliability
management; But due to the complexity of lithium
ion battery structure, the charged state by the
working current of battery, battery internal
resistance and the surrounding environment
temperature, self-discharge, and the influence of
factors such as aging, all of that makes the SOC
estimation difficult (Takeno, 2004). Particle swarm
optimization (PSO) algorithm is a global
optimization algorithm based on swarm intelligence,
has intrinsic parallel search mechanism, and is
suitable for the complex optimization field. The
paper make a new attempt on SOC estimation
method, applying PSO to SOC parameters
optimization, It can build optimization model to get
a more accurate estimate SOC in the case of less
parameters, faster convergence speed. (Kang, 2014)
2 PSO ESTIMATION METHOD
2.1 Particle swarm optimization theory
PSO is evolutionary computation technique
developed by Dr. Eberhart and Dr. Kennedy in 1995,
inspired by social behavior of bird flocking or fish
schooling. Recently, PSO algorithm has been
gradually attracted more attention over another
intelligent algorithm. (Shi, 2008) It was proved to be
an efficient method to solve optimization problems,
and has successfully been applied in the area of
function optimization, neural network training and
fuzzy control systems, etc. (Plett, 2004/2006)
2.2 The main model of Particle swarm
algorithm
PSO is colony intelligence which simulates the
swarm flock in search of food, each individual is to
Sun, Q., Lv, H., Wang, S., Jing, S., Chen, J. and Fang, R.
HV Battery SOC Estimation Algorithm Based on PSO.
In 3rd International Conference on Electromechanical Control Technology and Transportation (ICECTT 2018), pages 667-670
ISBN: 978-989-758-312-4
Copyright © 2018 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
667
be a particle of the flock, representative with space
position and speed of two-dimensional vector. In the
process of looking for food (problem solving), and
each particle according to the current location, itself
has experienced the best location (individual
optimum) and the entire population of optimal
location optimal (global optimum) to determine the
flight direction of next step.
The basic PSO algorithm model is the foundation
of other improved model, its core idea is the group
of individuals experience learning by itself and
social experience for reference, and group of other
individuals in the information exchange and sharing,
dynamic change itself, the speed of the optimized
group.
Figure 1: PSO Model algorithm process
3 LI-ION BATTERY SOC
ESTIMATION METHOD
At present, There are many different kinds of battery
model, Thevenin model of battery has the
characteristics of simple operation, and can well
reflect the dynamic and static characteristics of the
battery, so in this paper, this model is adopted to
establish the equation of the battery.
3.1 Modelling of key parameters for
SOC estimation
The type of new energy car battery system, the
system is composition of cells with P parallel and S
series, single series battery module in four
dimensional space vector.
Table 1: Key parameters of SOC estimation
Symbol
Description
Values
V Voltage 2.75V~4.2V
C Capacity --
R Resistance --
T Temperature -20℃~55℃
3.2 Model algorithm process
The specific calculation steps are as follows.
i) The initial time,
0
tt
. There are many
battery blocks as particles. Assuming that the
quantity of block is S, and
,,,
iiiii
s
oc C V T R
1, 2,iS
. The initial particle swarm,
123
,,,
iS
SOC soc t soc t soc t soc t
,
,,,
iiiii
s
oc t C t V t T t R t
. Initialize
individual particle experienced the optimal position
(individual optimal), record for
t
p
best t
, initialize
group particle experienced the optimal position
(global optimal), record for
t
g
best t
.
ii).
+1tt
, update the voltage, the temperature,
the resistance, the capacity of particle, get particle
swarm on the next moment, through the following
formula 1, formula 2, formula 3 and formula 4.
1
2
1
ii
V t V t rand pbest t p t
rand pbest t p t
(1)
3
4
1
ii
T t T t rand pbest t p t
rand pbest t p t
(2)
5
6
1
ii
Rt Rt rand pbestt pt
rand pbest t p t
(3)
11 21
31
111
1
ii
Ct Ct Vt Tt
Rt
(4)
Typically, the first for the original voltage of
battery module, reflects the battery state of a
moment, namely under the voltage influence of the
last time. The second reflects their thinking, which is
ICECTT 2018 - 3rd International Conference on Electromechanical Control Technology and Transportation
668
the current state of the battery module from their own
experience. The third part reflects the perception, the
group cooperation and information sharing among
the battery module. The battery module according to
the needs of the three parts for better search of the
optimal residual capacity
1
,
2
, …
6
.. For
learning factor, the main regulating speed with their
choice of optimal partner selection and optimal
location affect weight, rand () is a random number
between [0, 1].
iii) If t > Maximum iterations or match the end
conditions, then it terminated after the output
calculation, or cycle through the second step.
4 THE EXPERIMENT AND
SIMULATION ANALYSIS
In this paper, the PSO algorithm’s superiority in
battery state estimation through the battery charge
and discharge at the set condition experiment is
verified. The subjects for the lithium ion power
battery in a company.
Battery figure 2 (a) and (b) as shown in the
condition of current charge and discharge, the
experimental data of the sampling frequency of
1Hz.Figure 2 (a) for a battery charge and discharge
current, charge and discharge cycle figure 2 (b) as
the working condition of the battery working current.
200
-50
100
150
0 2000
4000
6000
8000
10000
-100
-150
50
0
Current (A)
Time
(
s
)
12000
(a) Single driving cycle
200
-50
100
150
0
2000
4000
6000
8000 10000
-100
-150
50
0
Current (A)
Time
(
s
)
1200
0
(b) Driving cycle of entire
Figure 2: Driving cycle condition
50
90
100
020
40
60
80
100
30
20
80
70
SOC (%)
Number of Cell
60
40
Measurement
Estimation
(a) SOC of cells estimation curves
-0.02
0.02
0.03
020
40
60
80
100
-0.04
-0.05
0.01
0
Error (%)
Number of Cell
-0.01
-0.03
0.05
0.04
30% SOC
50% SOC
70% SOC
90% SOC
(b) Estimation error curves
Figure 3: SOC of cells on individual optimum
70
20
50
60
02000
4000
6000
8000
10000
10
0
40
30
SOC (%)
Time
(
s
)
12000
80
90
100
Estimation
Measurement
(a) SOC of pack estimation curves
0.02
-0.03
0
0.01
0 2000
4000
6000
8000
10000
-0.04
-0.01
-0.02
Error (%)
Time
(
s
)
12000
0.03
0.04
(b) Estimation error curves
Figure 4: SOC of battery pack on global optimum
HV Battery SOC Estimation Algorithm Based on PSO
669
According to the estimation performance, the
parameters for estimating SOC was optimized by
using PSO algorithm, so it can improve the
estimation precision obviously.
5 CONCLUSIONS
HV battery is a kind of strongly nonlinear uncertain
system, so it is very difficult to establish battery
model accurately. In this paper, based on Thevenin
model of battery, the battery model is set up
adopting PSO to optimize the key parameters for the
battery SOC estimation. The estimated and
experimental results show the PSO method is robust
and the estimation precision can be improved base
on more accurate battery model.
ACKNOWLEDGEMENTS
This work was financially supported by the Tianjin
Science and Technology Development Strategy
Research Program (Grant No.: 17ZLZXZF00960),
the Undergraduate Teaching Quality and Teaching
Reform Project of Colleges and Universities in
Tianjin (Grant No.: 171006107C), and the National
Training Programs of Innovation and
Entrepreneurship for Undergraduates (Grant No.:
201710061022 and 201710061257).
REFERENCES
Ramadass, P., Haran, B., White, R., Popov, B. N., 2003.
Mathematical modeling of the capacity fade of Li-ion
cells. Journal of Power Sources.
Takeno, K., Ichimura, M., Takano, K., Yamaki, J., Okada,
S., 2004. Quick testing of batteries in lithium-ion
battery packs with impedance-measuring technology.
Journal of Power Sources.
Kang, L., Zhao, X., and Ma, J., 2014. A new neural
network model for the state-of-charge estimation in
the battery degradation process. Applied Energy.
Shi, Q., Zhang, C., and Cui, N., 2008. Estimation of
battery state-of-charge using ν-support vector
regression algorithm. International Journal of
Automotive Technology.
Plett, G. L., 2004. Extended kalman filtering for battery
management systems of LiPB-based HEV battery
packs part 3 state and parameter estimation. Journal of
Power Sources.
Plett, G. L., 2006. Sigma-point Kalman filtering for
battery management systems of LiPB-based HEV
battery packs Part 1: Introduction and state estimation.
Journal of Power Sources.
ICECTT 2018 - 3rd International Conference on Electromechanical Control Technology and Transportation
670
