You are on page 1of 10

Particle swarm optimization of the fuzzy logic controller for a

hybrid energy storage system in an electric car

Marek Michalczuk, Bartlomiej Ufnalski and Lech M. Grzesiak


WARSAW UNIVERSITY OF TECHNOLOGY
Institute of Control and Industrial Electronics
75 Koszykowa St., Warsaw 00-662, Poland
Phone: +48 22 234-6138
Fax: +48 22 234-6023
Email: {marek.michalczuk, bartlomiej.ufnalski, lech.grzesiak}@ee.pw.edu.pl
URL: http://www.ee.pw.edu.pl

Acknowledgments
The research was partially supported by the statutory fund of Electrical Drive Division within the
Institute of Control and Industrial Electronics, Faculty of Electrical Engineering, Warsaw University of
Technology.

Keywords
hybrid energy storage system, fuzzy logic controller, particle swarm optimizing, electric vehicle, power
management

Abstract
This paper describes the particle swarm optimization of the fuzzy logic controller for a hybrid energy
storage system in an electric vehicle. The Sugeno-type fuzzy inference system has been applied to divide
power between the battery and ultracapacitor energy storage systems, as well as to manage the amount of
energy stored in ultracapacitors. The two output signals of the described fuzzy logic controller represent
power for each energy storage system, and are the weighted sums of all inference rule outputs. The
particle swarm optimization has been proposed to determine the weights of rules. A simulation and
experimental validation of the power management algorithm is also introduced in this paper.

1 Introduction
An interest in electric vehicles (EVs) is constantly growing and it is an incontrovertible fact that the
world is already undergoing an automotive revolution. EVs, because of its unique selling points, are
currently gaining in popularity, especially as vehicles for urban and sub-urban areas [1–3]. Even though
cars powered by electrochemical batteries have a relatively small range, in this regard, they are able to
meet expectations for typical urban vehicles [4, 5]. A battery pack, in a constructional manner, is a
cumbersome component of EV, because it has a significant impact on the mass of the vehicle and takes
up a lot of space. The urban vehicle for which the range is not crucial, can afford a relatively small energy
storage system. The reduced size of the battery pack, compared with the one for a long distance EV, not
just facilitates the construction of small urban vehicles, but it also influences their price. Consequently,
it has a positive effect on the maintenance cost, as it is necessary to replace the vehicle’s batteries after
several years of use. However, the limited size of energy storage system reduces not only the range of the
vehicle, which is secondary here, but also the power ability. The performance of electrochemical battery
decreases, including the most popular lithium battery solutions for EV, if the battery is discharged with
high currents [6, 7]. This effect is particularly exacerbated when the battery is used at temperatures below
the nominal one. The poor performance is defined, among others, as a deterioration in the efficiency and
an accelerated ageing process of the battery.
Negative effects caused by high currents might be reduced through energy storage system hybridization
[8]. An example of such approach for EV are hybrid energy storage systems (HESS) consisting of high-
energy Li-ion batteries and high-power ultracapacitors (UCs) [9–11]. Therefore, the power management
algorithm that shares out power demand between the individual energy storage systems (ESSs) is required.
Power management principles for such a system are presented in [8, 12–14]
This research paper describes the fuzzy logic controller (FLC) for the power management purpose. The
expert knowledge is stored in the system by intuitive inference rules, which mimic the human way of
reasoning. This aspect allows an expert to conveniently formulate the rules of inference. The FLC has
been created based on the controller described in [15] and represents its development. The FLC controller
uses information from a road infrastructure and other vehicles to better adjust an instantaneous power
of batteries and ultracapacitors, depending on circumstances on the road. The merits of the use of such
signals are presented in [15]. The controller tuning involving a determination of rule weights, has a
significant impact on the controller performance. This process is complex and not evident, especially if
the inference system has a lot of rules. In this study the particle swarm optimization (PSO) is proposed
to determine the rule weights. This multidimensional optimization technique allows us to find the
appropriate balance between the battery and ultracapacitor power. This paper introduces a complete
design of FLC, including a method to optimize the inference rule weights.

2 Hybrid energy storage system for electric vehicle


In this study the fully active HESS with parallel topology is considered (Fig. 1). Parameters of the
urban vehicle and its powertrain are listed in Tab. I and Tab. II. The significant element is that the
energy stored in ultracapacitors is relatively small. In fact, it is comparable to the kinetic energy of the
vehicle at its maximum speed, and this energy is negligible in terms of the vehicle’s range. However,
ultracapacitors can serve as an auxiliary ESS and are able to temporarily provide a high power impulse
during dynamic states of the vehicle, namely accelerating and braking. Due to the resistant forces, such
a small ultracapacitor energy storage system is not sufficient to cover the full demand for power during
acceleration from zero to maximum speed. Thus, the power demand must be distributed between the
two ESSs. It follows that the power management algorithm should supervise the energy stored in UCs in
order to avoid accelerating the vehicle with the fully discharged UCs, as well as braking with the fully
charged UCs.
The model of hybrid energy storage system used for designing and optimization is described in detail in
[16] and is available on [17, 18]. The dynamic models of battery and ultracapacitor allow the simulation
of their terminal voltage and power losses at different working conditions. The parameters of battery and
ultracapacitor models have been estimated from experimental data.
Table II: Parameters of drive system and vehicle
Parameter Vehicle
Table I: Parameters of hybrid energy storage system Total mass 1300 kg
Maximum speed 90 km/h
Parameter Battery pack Ultracapacitor
stack Maximum 4 km/h/s
acceleration 1.1 m/s
Nominal voltage 375 V 486 V
Maximum 60 kW
Voltage operation 316 V – 421 V 243 V – 486 V powertrain power
range
Powertrain average
Capacity of a cell 40 Ah 310 F efficiency (drive 85 %
Number of cells 117 360 converter and motor)
Stored energy 15 kWh 112 Wh Aerodynamic
coefficient – Cx 0.37
Rolling friction 0.012
coefficient

3 Power management
3.1 The general scheme of power management
The power required by the drive system is provided by the power management algorithm in such a way
that it does not affect the vehicle dynamic performance, and at the same time strives to minimize the
instantaneous battery current. A schematic representation of the control strategy is shown in Fig. 1. There
are two inner current control loops associated with the two DC/DC converters. Maintaining desirable
output voltage of the hybrid source, along with the voltage of UCs, and the appropriate power partitioning
between the two energy sources, are all achieved by a proper reference current value determination by
means of the FLC.
An instantaneous power required by the drive system is significantly affected by the road grade. A larger
energy reserve should be maintained in the UCs in the course of driving uphill, due to higher power
X
ibat
Controller -
ibat + ·
ubat
v
eq. (1)

DC/DC 1
Battery

2
vveh

pbatref įest
Fuzzy vexp

system
Drive
Logic pload +
Ultracapacitor

Controller
DC/DC 2
pucapref

udc udc_ref + uerr


-

Controller eucap
+ · eq. (2)
iucap -
iucap uucap
X

Fig. 1: The control scheme for the hybrid energy storage system

pulses during acceleration and lower energy recovery. In turn, if the vehicle is driving downhill, the
power consumption is lower and the energy recovery increases. This suggests that the performance of
HESS might be improved, if the information on slope of terrain is taken into consideration by the FLC.
Such information can be provided by a GPS navigation system compatible with digital elevation model
(DEM) data. More potentially useful signals can be provided by other vehicles on the road, as a part of
vehicle-to-vehicle (V2V) communication. An example of such information is the speed of other vehicles
in the vicinity, which provides the basis to estimate the expected speed of the vehicle. Additionally,
a maximum permissible speed for a given road can be provided as a part of infrastructure-to-vehicle
communication (I2V).

3.2 The structure of FLC


The fuzzy logic controller has six inputs and two outputs. The output signals directly correspond with
the reference power for individual ESSs, rather than one signal describing a ratio of selected ESS power
with respect to load power. Those two signals allow to determine a direct flow of energy between storage
systems and control the amount of energy stored in UCs, when the load power is zero. The power
flow management system has been implemented by computationally efficient Sugeno type fuzzy logic
controller. The input variables (Fig. 2) are: output voltage of the hybrid source error (uerr ), power load
(pload ), expected speed of the vehicle (vexp ), square of speed normalized to the maximum value expressed
as
vveh
2
= (v/vmax ) 2 , (1)
the amount of usable energy stored in ultracapacitor as presented in the equation
2
uucap − (0.5uucapmax ) 2
eucap = ( 2
), (2)
0.75uucapmax

and the slope of terrain (δest ). An operating voltage range for ultracapacitor ESS has been set between
half of nominal voltage to the nominal voltage. The minimum voltage needs to be limited because if not,
it results in the need for extremely high currents at low UC voltage. Consequently, the usable energy is
75 % of the total energy stored in ultracapacitor cells. The following criteria have been undertaken to
normalize the signal eucap : eucap equal to zero means the fully discharged UCs, and eucap equal to one
means fully charged UCs. The value of δest factor is calculated as:
(h (200) − h (0))
δest = 100% (3)
200
where h(n) is the elevation within n meters from the current location. The output variables are: a
reference battery power (pbatref ) and a reference ultracapacitors power (pucapref ). The output membership
functions of the first-order Sugeno fuzzy inference system are either a linear or a constant function as
neg zero pos low med high veryH low high
1 1 1






0 0 0
-20 -12 -4 4 12 20 0 12 24 36 48 60 0 18 36 54 72 90

under veryL low mid high veryH zero low high max down normal up
1 over 1 1




 

0 0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 -10 -6 -2 2 6 10

Fig. 2: The membership functions of input variables

follows: disch = 20, dischL = −0.22vexp , zero = 0, chargeL = 0.22vexp , charge = 20, line = −3uerr ,
line2 = −pload .
The 159 rules are employed in the inference process. Each rule has a crisp output (zi ) resulting from a
rule’s degree of applicability and the output function defined for the particular rule. The overall output is
obtained via a weighted sum, computed as

159
pucapref, pbatref = wi z i , (4)
i=1

where wi is the weight of an i-th rule. For the discussed controller, most of weights are calculated in the
optimization process. However, for rules from 1 to 7:
1. If (uerr is pos) then (pbat is line) and (pubat is zer o)
2. If (uerr is neg) and (eucap is over ) then (pbat is line) and (pubat is zer o)
3. If (uerr is neg) and (eucap is not over ) and (δslope is down) then (pbat is line) and (pubat is zer o)
4. If (uerr is zer o) then (pbat is zer o) and (pubat is zer o)
5. If (uerr is neg) and (eucap is not over ) and (δslope is not down) then (pbat is zer o) and (pubat is line2)
6. If (uerr is neg) and (eucap is not over ) and (δslope is down) then (pbat is zer o) and (pubat is line2)
7. If (eucap is not under ) then (pbat is zer o) and (pubat is line)

weights are already defined, and they are: w1 =w2 =w4 =w5 = w7 = 1, w3 = 0.4, w6 = 0.6. These rules are
responsible for keeping output voltage at a desired level. The HESS should always ensure the appropriate
output voltage, so the weights of the rules no. 1–7 are fixed based on expertise and not included into
optimization. Rules from 8 to 39 correspond to dividing the positive load power between ESSs. Rules
no. 8–11 are formulated as:
8. If (pload is ver yH ) and (vexp is high) and (eucap is not under ) and (vveh
2 is max) and (δ
slope is not up) then (pbat is zer o) and
(pubat is char geL)
9. If (pload is high) and (vexp is high) and (eucap is not under ) and (vveh
2 is max) and (δ
slope is not up) then (pbat is zer o) and
(pubat is char geL)
10. If (pload is med) and (vexp is high) and (eucap is not under ) and (vveh
2 is max) and (δ
slope is not up) then (pbat is zer o) and
(pubat is char geL)
11. If (pload is low) and (vexp is high) and (eucap is not under ) and (vveh
2 is max) and (δ
slope is not up) then (pbat is zer o) and
(pubat is char geL)

Subsequently, rules no. 12–23 are defined when vveh 2 is either high, low, zero in subsequent rules.

Similarly, rules no. 24–39 are defined when δest is up. The succeeding rules are responsible for the
exchange of energy between batteries and ultracapacitors and are sensitive only to eucap , vveh
2 and δ
est
signals. These rules do not affect the resultant power provided to the drive system, but enable changing
the energy stored in the UCs. Analogous to rules no. 40–43:
40. If (eucap is ver yL) and (vveh
2 is zer o) and (δslope is nor mal) then (pbat is dischL) and (pubat is char geL)
41. If (eucap is ver yL) and (vveh
2 is low) and (δslope is nor mal) then (pbat is dischL) and (pubat is char geL)
42. If (eucap is ver yL) and (vveh
2 is high) and (δslope is nor mal) then (pbat is disch) and (pubat is char ge)
43. If (eucap is ver yL) and (vveh
2 is max) and (δslope is nor mal) then (pbat is disch) and (pubat is char ge)

rules no. 44–59 are formulated when eucap is in one of the states low, mid, high, vh. Similarly rules
no. 60–79 and 80–99 are defined as δest is up and down, respectively. The rules 40–99 correspond to
the energy flow from batteries to ultracapacitors. Rules 100–159 are complementary to rules no. 40–99,
i.e. they have the same antecedents, but the consequents are always as follows: ’pbat is charge’ and
’pubat is disch’, what corresponds to reversed energy flow.
3.3 Particle swarm optimization
The core of power distribution between batteries and UCs is defined in rules no. 8–169 and the
performance of the management controller varies with the change in rule weights. The controller
should present satisfying performance for a variety of speed profiles. Because the speed profile of the
vehicle is not known a priori, the selection of weights must take into account numerous dynamic states
of the vehicle. Due to high number of decision variables, selection of weights by a trial and error method
based on an expertise, is very problematic. In this study the weights are determined in the particle swarm
optimization [19]. An illustrative PSO of the FLC for the HESS is presented in [20]. The objective is to
minimize the battery current for defined drive cycle, so initially the cost function has been defined as
 tstop
J0 = i 2bat dt . (5)
0

Due to the fact that there exist some constraints on the rule weights, the number of parameters in the
search-space of the PSO is smaller than the number of the sought weights. Firstly, as a result of rules
no. 8–39, it is assumed that the relationship between pucapref and pload is monotonic and the pucapref is
negative. This means that the higher pload , the higher discharge power of the UCs, thus weights of these
rules are calculated as follows:
w8+4i = min(1, pi+1 ) , w9+4i = min(0.75, pi+1 ) ,
(6)
w10+4i = min(0.5, pi+1 ) , w11+4i = min(0.25, pi+1 ) ,

for i = 0, 1 ... 7, where pn is the n-th parameter in the search-space of PSO. Secondly, sets of rules no.
40–99 and 100–159 are complementary, so it is reasonable to activate only one of a pair of complementary
rules. Thus, weights from 40 to 159 are calculated according to the formulae:
 
p for p9+i ≥ 0 0 f or p9+i ≥ 0
w40+i = 9+i , w100+i = , (7)
0 f or p9+i < 0 −p9+i for p9+i < 0

for i = 0, 1 ... 59. That clearly specifies that, the swarm is looking for an optimal solution in 68-
dimensional space.
Other constraints for rules no. 40–159 are introduced into the PSO by a penalty method. The cost
function (5) is extended by the penalty function

J = J0 + gi (w) . (8)
i

It is required for the relations between power to be exchanged internally in HESS and the amount of
energy stored in the UCs and kinetic energy of the vehicle are to be monotonic. Consequently, for each
eucap and δest , the higher vveh
2 , the lower p
ucapref and higher pbatref . This is defined in the penalty function
as

2
gi+1 = max(0, w9+n+4i+1 − w9+n+4i ) , (9)
n=0

2 and δ , the higher e


for i = 0, ... , 14. Additionally for each vveh est ucap , the higher pucapref and lower pbatref .
This is defined as

2 
3+5m
gi+16 = ( max(0, w9+4n+i+4 − w9+4n+i )) , (10)
m=0 n=0+5m

for i = 0, ... , 3. It is assumed that the share of battery power in coverage of total power demand is lower
if speed of the vehicle is lower. Thus, the penalty function for this constraint is calculated as:


2
gi+20 = max(0, wn+4i+1 − w8+n+4i+2 ) , (11)
n=0

for i = 0 or 1.
There are 24 particles implemented in the optimization. The PSO applies the reflecting boundary
conditions and stops after 60 iterations. In each iteration, each particle represents a potential solution
Table IV: Optimized weights of fuzzy logic inference rules
c
pr c
1 2 3 4
1 0.75 0.70 0.53 0.38
2 0.82 0.78 0.57 0.45
Table III: Parameters of the test diving cycle 3 0.91 0.83 0.71 0.00
4 0.86 0.79 0.49 −0.33

δslope [%] δest [%] ΔV eucap (t = 0) 5 0.75 0.30 0.26 −0.49


[km/h] 6 0.62 0.22 −0.28 −0.60
7 −0.12 −0.43 −0.41 −0.70
10, 20,
8 0.94 0.94 0.74 0.24
30, 40, 0, 0.7, 0.17, r
−4, 0, 4 0, δslope 50, 60, 0.3, 0.45, 9 0.85 0.80 0.71 0.19
70, 80, 0.62, 0.8, 1 10 0.82 0.60 0.42 −0.17
90 11 0.73 0.50 −0.15 −0.38
12 −0.11 −0.21 −0.54 −0.59
13 0.82 0.73 0.52 0.32
14 0.69 0.49 0.24 −0.12
15 0.47 0.37 0.01 −0.80
16 0.41 −0.04 −0.16 −0.89
17 −0.13 −0.52 −0.85 −0.99

and is evaluated during a test drive cycle with the objective function. The performance of optimized
controller strongly depends not only on the objective function, but also on the defined test drive cycle.
The whole test diving cycle is a set of many acceleration and braking phases, occurring at different
conditions. Fig. 3 illustrates the speed profiles for a single acceleration and deceleration phase. The test
driving cycle includes all possible combinations of acceleration and braking tests for parameters given
in Tab. III, where δslope is a slope of the route and ΔV is an absolute value of difference between Vf and
Vi (Fig. 3). A positive acceleration is 4 km/h/s or 3 km/h/s if δslope is 4 % and deceleration is 6 km/h/s.
All this tests are executed assuming that vexp is 90 km/h and Vf is 90 km/h for accelerating and zero for
braking. The rule base indicates that the lower vexp , the higher the share of UCs in the load power. This
relation has been rigidly defined by chargL and dischL output function in certain rules. Thus, a change
of vexp is not included in the test driving cycle.
The optimization process for a model driving cycle with initial condition eucap = 0.2 is illustrated in
Fig. 4. The final values of the parameters pr c are summarized in Tab. IV. Sample characteristics obtained
in the optimization process are shown in Fig. 5. These graphs show a component of ultracapacitor power
that describes the energy exchange between EESs. The positive value of pucapref means that the current
eucap should be increased for a given speed of the vehicle, and conversely, the negative pucapref means that
eucap is too high.

100
120
Vf 80
   

80
 

60
  
40 40
Vi
20
0
0
0 10 t1 t 1+ 0 20 40 60 0 20 40 60
Vi  
120 120

80 80
 

 

 
Vf
40 40

0 0
0 10 t1 t 1+
  0 20 40 60 0 20 40 60
 
  

Fig. 3: Acceleration (a) and
braking (b) test speed profiles Fig. 4: Speed profile (a) and battery current response after 3 (b), 15 (c), 60 (d)
iterations.
20 20 20





0 0 0

-20 0 -20 0 -20


0 0 0
0.5 0.5 0.5
0.5 0.5 0.5
1 1 
 1 1 
 1 1 

  

Fig. 5: Relationships between eucap , vveh


2 and p
ucapref , when δest is f lat (a), δest is uphill (b), δest is downhill (c) at
uerr = 0, pload = 0, vexp = 90

4 Simulation
A performance of the designed control scheme has been analysed in simulation studies for various driving
cycles. The power distribution between storage systems in the IM240 driving cycle [21] is shown in Fig. 6.
The maximum power of the battery in relation to load power is significantly reduced and less efficient
battery charging process is almost eliminated. The nonlinear control algorithm for the energy flow
management uses information about the estimated slope of the terrain and the expected speed. The effect
of using data on the expected speed of a vehicle for the normalized Artemis driving cycle is presented
Fig. 7 [21]. In this case it was assumed that the vexp is 60 km/h. Using the same speed profile, Fig. 8
illustrates the results for a vehicle being driven in heavy traffic or in the centre of a city, where the
permissible speed is limited. In these conditions, if information on expected speed is accessible then a
battery power profile is smoother, causing more efficient operation.
100
 

75
 50  30
25
 

15
0
100 0 40
0 80
0 120
2 160
6 200
0 240
4 0
 

75
 -15
50
25 -30    
0
0 40 80 120 160 200 240 0 40 80 120 160 200 240
 
  


Fig. 6: Speed (a) and corresponding power distribution (b) and ultracapacitor state of charge (c) for IM240 driving
cycle

75 20
 

10
 

50 
 0
25
0 150 300 450 600
0 20
0 150 300 450 600
 

30
10
  

15 

0
0

-15
0 150 300 450 600
0 150 300 450 600 
 
 


Fig. 8: Battery power in the Artemis driving without


Fig. 7: Speed (a) corresponding load power (b) for the the use of signal vexp (a) and when signal vexp available
Artemis High Urban (ArtHiURB) driving cycle
(b)

Fig. 9 demonstrates the speed profile for a dynamic real driving—recorded using GPS receiver—in
a mountainous suburban area. The power drawn from the battery is compared for two cases: when
information on the slope is not used (Fig. 10a) and when it is available (Fig. 10b). Both positive and
negative impulses of battery power are limited when the topography of the route is known. It is particularly
distinctive e.g. in 350-400 and 620-650 seconds, as indicated in Fig.10. This example highlights the fact
that the additional information from navigation systems can be beneficial and should be included in the
algorithm.

100 60
 

75
 50 35
25

 
0  10
0 150 300 450 600 750 900
    

60 -15
45
30
15 -40
 0
-15
-30 0 150 300 450 600 750 900
-45 60
-60
0 150 300 450 600 750 900 35
 


 
 10
    

400
 -15
200
0 -40
0 5 10 15 20

   0 150 300 450 600 750 900

 

Fig. 9: Vehicle speed (a) in real driving cycle recorded Fig. 10: Battery power in real driving cycle when slope
near Sanok City, corresponding load power (b) and information is not used (a) and is used (b)
relative elevation (c)

5 Laboratory experiment
Experimental validation of the proposed power management algorithm has been performed through tests
using a laboratory setup. This non-mobile setup is equipped with AC/DC converter fed by the grid instead
of battery pack. Additionally, two DC/DC converters and ultracapacitor stack are included to complete
the HESS, as well as two machine sets are incorporated to imitate a drive system of the vehicle. Each
machine set consists of two PMSMs (Permanent Magnet Synchronous Motors). One PMSM operates as
a drive, the other operates as a load. The control algorithm for HESS including the fuzzy system has
been implemented on TMS320F28335 microcontroller. Compared with the vehicle, non-mobile systems
have been designed for reduced power (scale of 1 : 20). The value 0 of eucap corresponds to the minimal
voltage level. This level has been set so that the energy stored in UCs matches to the state of affairs in the
full-scale vehicle.

Fig. 11: Laboratory setup: a) power electronic equipment, b) four-motor set. The setup built within the ECO-
Mobility Project (POIG01.03.01-14-154/09-04)
Speed profile and corresponding demand for power from HESS by the drive system is illustrated in
Fig. 12. Power distribution for that speed profile is presented in Fig. 13. Power sharing between emulated
battery pack and ultracapacitor storage system takes place in accordance with the expectations. While
the vehicle is accelerating and high power is needed, a majority of the output power is provided by the
ultracapacitors. Thereby, the power drawn from the main energy storage during the acceleration phase
is close to the power during constant speed phase. A modest share of ultracapacitors in output power is
exclusively for low speed, but due to low power demand at that moment, it is not at the expense of a high
battery current. It allows to maintain the energy reserve in the auxiliary energy storage system, so that
it is possible to successfully reduce the battery power at the later acceleration phase when here is a high
power demand.

  
 


 

   


Fig. 12: Speed of the vehicle (a) and corresponding power output power (b) of laboratory HESS

2.25

1.5

0.75
 

 0

-0.75

-1.5

-2.25 
  
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170

100
 

75
 50
25
0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170
 

Fig. 13: Power distribution (a) and ultracapacitor state of charge (b) in laboratory HESS

6 Conclusions
The fuzzy logic controller for the hybrid energy storage system has been optimized using PSO and
evaluated in simulation. The aim of the power management system is the reduction of an instantaneous
battery current, without deterioration of the vehicle performance. The effectiveness of optimized FLC
has been examined in various driving scenarios. The usefulness of an additional information related
to the topography of the road and the expected vehicle speed, in the context of reducing battery peak
currents, has been investigated. The laboratory test with the physical emulator of the electric vehicle
powertrain shows the ability to implement the proposed control structure in real application.
References
[1] Gis, W. and Waśkiewicz, J.: Future of electric cars in the eu, Journal of KONES, vol. 21, 2014.
[2] Reininger, C. and Salmon, J.: Systems feasibility study for implementing electric vehicles into
urban environments, 2015 9th Annual IEEE International Systems Conference (SysCon), 2015,
pp. 734–739.
[3] Akhavan-Rezai, E., Shaaban, M., El-Saadany, E., and Karray, F.: Demand responce through in-
teractive incorporation of plug-in electric vehicles, Power Energy Society General Meeting, 2015
IEEE, 2015, pp. 1–5.
[4] Pearre, N. S., Kempton, W., Guensler, R. L., and Elango, V. V.: Electric vehicles: how much range
is required for a day’s driving?, Transportation Research Part C: Emerging Technologies, vol. 19,
no. 6, 2011, pp. 1171 –1184.
[5] Tamor, M. A., Gearhart, C., and Soto, C.: A statistical approach to estimating acceptance of electric
vehicles and electrification of personal transportation, Transportation Research Part C: Emerging
Technologies, vol. 26, 2013, pp. 125 –134.
[6] Omar, N., Monem, M. A., Firouz, Y., Salminen, J., Smekens, J., Hegazy, O., Gaulous, H., Mulder,
G., Bossche, P. V. den, Coosemans, T., and Mierlo, J. V.: Lithium iron phosphate based battery –
assessment of the aging parameters and development of cycle life model, Applied Energy, vol. 113,
2014, pp. 1575 –1585.
[7] Michalczuk, M., Ufnalski, B., Grzesiak, L. M., and Rumniak, P.: Power converter-based electro-
chemical battery emulator, Przeglad ˛ Elektrotechniczny, vol. 90, no. 7, 2014, pp. 18–22.
[8] Michalczuk, M., Grzesiak, L., and Ufnalski, B.: Hybridization of the lithium energy storage for
an urban electric vehicle, Bulletin of the Polish Academy of Sciences: Technical Sciences, vol. 61,
no. 2, 2013, pp. 325–333.
[9] Cain, S. R., Anderson, A., Tasillo, E., Infantolino, W., and Wolfgramm, P.: Empirical evaluation of
the improvement of battery output when coupled with a capacitor bank, Journal of Power Sources,
vol. 268, 2014, pp. 640 –644.
[10] Carter, R., Cruden, A., and Hall, P.: Optimizing for efficiency or battery life in a battery/su-
percapacitor electric vehicle, IEEE Transactions on Vehicular Technology, vol. 61, no. 4, 2012,
pp. 1526–1533.
[11] Biernat, P., Rumniak, P., Michalczuk, M., Galecki, A., Grzesiak, L., Ufnalski, B., and Kaszewski,
A.: Powertrain system with the ultracapacitor-based auxiliary energy storage for an urban battery
electric vehicle, The Archives of Transport, no. vol. 27-28, no 3-4, 2013, pp. 45–64.
[12] Song, Z., Hofmann, H., Li, J., Han, X., and Ouyang, M.: Optimization for a hybrid energy storage
system in electric vehicles using dynamic programing approach, Applied Energy, vol. 139, 2015,
pp. 151 –162.
[13] Wang, T., Deng, W., Wu, J., and Zhang, Q.: Power optimization for hybrid energy storage system
of electric vehicle, Transportation Electrification Asia-Pacific (ITEC Asia-Pacific), 2014 IEEE
Conference and Expo, 2014, pp. 1–6.
[14] Wang, B., Xu, J., Cao, B., Li, Q., and Yang, Q.: Compound-type hybrid energy storage system and
its mode control strategy for electric vehicles, Journal of Power Electronics, vol. 3, no. 3, 2015.
[15] Michalczuk, M., Ufnalski, B., and Grzesiak, L. M.: Fuzzy logic based power management strat-
egy using topographic data for an electric vehicle with a battery-ultracapacitor energy storage,
COMPEL - The International Journal for Computation and Mathematics in Electrical and Elec-
tronic Engineering, vol. 34, no. 1, 2015, pp. 173–188.
[16] Michalczuk, M., Grzesiak, L., and Ufnalski, B.: Experimental parameter identification of battery-
ultracapacitor energy storage system, 2015 IEEE 24th International Symposium on Industrial
Electronics (ISIE), 2015, pp. 1260–1265.
[17] Michalczuk, M.: Li-ion battery model, url: http://www.mathworks.com/matlabcentral/f
ileexchange/48234.
[18] Michalczuk, M.: Ultracapacitor (Supercapacitor) model, url: http://www.mathworks.com/m
atlabcentral/fileexchange/51243.
[19] Clerc, M.: Particle swarm optimization, vol. 93, John Wiley & Sons, 2010.
[20] Michalczuk, M.: Particle Swarm Optimization of Fuzzy Logic Controller, url: https://www.ma
thworks.com/matlabcentral/fileexchange/57653.
[21] Barlow, T.: A Reference Book of Driving Cycles for Use in the Measurement of Road Vehicle
Emissions, Published project report, IHS, 2009.

You might also like