Regenerative Braking Modeling, Control, and Simulation of A Hybrid Energy Storage System For An Electric Vehicle in Extreme Conditions
Regenerative Braking Modeling, Control, and Simulation of A Hybrid Energy Storage System For An Electric Vehicle in Extreme Conditions
Regenerative Braking Modeling, Control, and Simulation of A Hybrid Energy Storage System For An Electric Vehicle in Extreme Conditions
4, DECEMBER 2016
465
I. I NTRODUCTION
A. General
Manuscript received April 19, 2016; revised July 23, 2016; accepted
August 29, 2016. Date of publication September 12, 2016; date of current
version December 1, 2016.
K. Itani is with the Electrical Department, Institut Suprieur des Sciences Appliques et EconomiquesCnam Liban, 113-6175 Hamra Beirut,
Lebanon (e-mail: [email protected]).
A. De Bernardinis and Z. Khatir are with Systmes et Applications des
Technologies de lInformation et de lEnergie/Institut Franais des Sciences et Technologies des Transports, de lAmnagement et des Rseaux,
78000 Versailles Cedex, France (e-mail: [email protected];
[email protected]).
A. Jammal is with the Ministry of Higher Education, Beirut, Lebanon
(e-mail: [email protected]).
M. Oueidat is with Institut Universitaire de TechnologieSaida Lebanese
University, Saida, Lebanon (e-mail: [email protected]).
Digital Object Identifier 10.1109/TTE.2016.2608763
2332-7782 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
466
Fig. 1.
Fig. 2.
467
TABLE I
TABLE II
TABLE III
P0
Ic0max
(1)
(2)
2 3
C0 2
2
3
Uc0max Uc0min
+ kc Uc0max Uc0min
.
2
3
(3)
V2
.
150
(4)
V EHICLE S IZING
(Vbrakem/s v w )2
2
(5)
where Table II presents the different parameters.
The inertia effect of all the rotating elements is neglected.
The total power to brake is therefore 316 kW.
The power has to be shared between front and rear wheels.
The ratio between the friction front force and the total friction
force respecting the ECE R13 regulations [27] while assuring
that a maximal front/rear braking ratio is expressed by the
coefficient
2 0.07L b h g + L b + 0.07h g
.
(6)
hb-max =
0.85L
Table III presents the vehicle dimensions.
Eighty-one percent of the total power should be recuperated
by the front wheels and dispatched between electrical and
mechanical brakes. The electrical flow power should not
exceed the maximal power of 60 kW, which constitutes, in
this particular case, 23.4% of the front power braking. It is
468
TABLE IV
PI G AIN C ONTROLLERS AND E LECTRICAL S PECIFICATIONS
FOR THE T HREE -L EVEL DC/DC C ONVERTER
Fig. 3.
Fig. 4.
Fig. 5.
1
2
Mtotal Vbrake
.
m/s
2
(7)
VBUSmax
.
16 i c0max f sw
(8)
2
Vmin
.
Pmax
(9)
(10)
(11)
Fig. 6.
469
Fig. 7. Bode magnitude and phase plot set for the UC current open-loop
transfer function.
kiic0
1 + ic0 s
= k pic0
(17)
s
ic0 s
where kpic0 and ki ic0 are, respectively, the proportional and
integral gain of the current controller.
The transfer functions of sampling, algorithm calculus, and
the zero-order hold element will be merged as a one-order
transfer function with a global time constant Tsi = 2 Tsw .
The open-loop transfer function in this case will be
G ic0 (s) = k pic0 +
k pic0
1
.
(18)
ic0 s R L0 (1 + Tsi s)
The optimal modulus design criterion is used in order to
calculate the gain of the PI current controller, where the
damping factor is = 0.707. Based on this criterion, the
OLTFic0 =
Fig. 8.
k pic0 = ic0
R L0
2Tsi
(20)
k pic0
.
(21)
ic0
3) UC Current Closed-Loop Stability: The stability is measured by the gain margin and phase margin of the open-loop
transfer function. Using the designed controller synthesis, and
for a d varying from 0.01 to 0.1, the Bode diagram is plotted
in Fig. 7.
For different values of d, the gain margin of 15.2 dB at
a cutoff frequency of 22 263 rad/s and the phase margin of
63 at 0dB cutoff frequency of 5988 rad/s remain the same.
The major perturbation is the current signal perturbation i B .
The simulation results in Section VI show the robustness of
the UC current controller.
kiic0 =
470
Fig. 9.
C 0 +2kc Uc0
Rc0 [C 0 +2kc [Uc0 Rc0 Ic0 ]]2
2kc Ic0
+ [C +2k [U
2
c c0 Rc0 Ic0 ]]
0
s+
s
= Rc0
(24)
(25)
i c0max
Uc0max
Fig. 11.
CLTFuc0 (s)
s + z
s + p
(23)
G BUS (s) =
Fig. 10.
(26)
+ ( p + Rc0 z k puc0 )2 = 0.
(28)
( p 2z z k puc0 Rc0 )
Rc0
( p 2z z k puc0 Rc0 )2 ( p + z k puc0 Rc0 )2
+
.
Rc0
(29)
(30)
471
(32)
Fig. 12. Root locus of the dc bus voltage closed-loop characteristic equation.
1 + k puc0 Rc0
sCBUS VBUS
1 + sCBUS RESR
= C V (1+k R ) .
(33)
puc0 c0
s BUS BUS
k puc0 Uc0
(36)
(37)
(39)
(40)
(41)
with K i = K p /Ti .
The characteristic equation of a second-order equation can
be written as
1 2 (r1 + r2 )
s +
s+1=0
(42)
r1 r2
r1 r2
CLTFvbus(s) = X
p1 =
(35)
(r1 + r2 )
.
r1 r2
(45)
and
1
Td s
.
(38)
+
C(s) = k p 1 +
Ti s
1 + Td s
Kp =
(44)
(34)
1
r1 r2
p2 =
a1 ( p1 b1 )
p2 b1 ( p1 b1 )
(46)
k pvbusmax
.
( p1 b1 )
(47)
and
kivbusmax =
Uc0 k puc0
.
VBUS (1 + k puc0 Rc0 )
(49)
(48)
472
Fig. 15.
Fig. 16.
Fig. 13. Bode magnitude and phase plot set for the dc bus voltage open-loop
transfer function.
Fig. 14.
6) System Robustness: The major perturbation is the regenerative power coming from the inverters. This perturbation
should negatively affect the dc bus voltage value if the dc bus
voltage controller is poorly designed. The effect of parameter
value uncertainties and model approximation assumptions will
not be treated in this paper.
The simulation results in Section VI show the robustness of
the system due to several tests.
D. PI Antiwindup Controllers
PI antiwindup controllers have been used for the UC voltage
controller and the dc bus voltage controller. When a controller
saturates, it therefore operates in a nonlinear region. The action
consisting of increasing the control signal will have no effect
on the system output, known as wind-up effect. This leads to a
characteristic step response with a large overshoot and a very
high settling time [34], [35]. The windup effect will postpone
the intervention of the controllers, which will delay the UC
recovering and create a high transient and hardly controlled
signal on the UC voltage and dc bus voltage sides.
In order to solve the problem, it has been decided to use an
internal tracking feedback loop to discharge the PI controller
integrators for the UC voltage and dc bus voltage controllers,
known as backcalculation antiwindup. In Fig. 14, the backcalculation antiwindup structure of the UC voltage PI controller
is represented.
(50)
(51)
473
(53)
Y = x (e + y) + x y (e + f )
(54)
and
Fig. 17.
S1 = ab
c + a b d
(52)
474
Fig. 18.
Fig. 19.
Fig. 20.
Fig. 21.
Fig. 22.
Fig. 23. DC bus voltage regulation for medium friction road type at Uc0min .
Fig. 26.
Fig. 27.
Fig. 28.
475
Fig. 24. Battery measurement for medium friction toad type test at Uc0min .
Fig. 25. Storage elements recovered energy for medium friction toad type
test at Uc0min .
Fig. 29. DC bus voltage regulation for 5-cm shallow snow road type at
Uc0 = 290 V.
The UC voltage and current follow their references according to Fig. 21. The maximal current UC reached is 210 A.
At the end of the stopping time of 6.97 s, the UC voltage
and SOC (estimated by MATLAB/Simulink) have attained
229 V and 66%, respectively.
The maximum power recovering of the UC is 39.2 kW
(Fig. 22). The 3-kW difference between the UC power and
the HESS maximal power input of 42.3 kW largely comes
from the joule losses performed on the 70-m resistor of
the inductor. Fig. 23 shows that the dc bus voltage regulation
476
Fig. 30.
290 V.
Fig. 31. Storage elements recovered energy for 5-cm shallow snow road
type at Uc0 = 290 V.
Fig. 32.
Fig. 33.
Fig. 34.
Fig. 35. Recovered energy and dissipated energy by the battery and controlled
resistor, respectively.
Fig. 36.
Simulation test comparison on an asphalt dry road type for the first simulation.
Fig. 37.
Simulation test comparison on a shallow snow road type for the second simulation.
477
478
479