You are on page 1of 6

Robust Switched Interval Observer for the Energy

Management System in Hybrid Vehicles


Fatah Bejaoui Chokri Mechmeche Ali Sghaier Tlili
Laboratory of Advanced Systems, Laboratory of Advanced Systems, Laboratory of Advanced Systems,
Polytechnic School of Tunisia (EPT) Polytechnic School of Tunisia (EPT) Polytechnic School of Tunisia (EPT).
BP. 743, 2078, La Marsa, Tunis BP. 743, 2078, La Marsa, Tunis BP. 743, 2078, La Marsa, Tunis
National High School of Engineering of National High School of Engineering of High Institute of Computer Science (ISI),
Tunis (ENSIT) Tunis (ENSIT) University of Tunis El Manar
Bejaouifatah@gmail.com chokri.mechmeche@esstt.rnu.tn ali.tlili@ept.rnu.tn, ali.tlili@isi.utm.tn

Abstract— In this paper, the energy management system model - the plug-in hybrid: To increase the autonomy of the
of a hybrid vehicle (EMSHV) is elaborated. Based on this model, batteries, they can also be recharged by plugs connected to the
a switched interval observer with unknown input is proposed to domestic current.
estimate the output voltage of the DC bus. Sufficient conditions Among the problems of the hybrid vehicle architecture, is
of the stability of such an observer are expressed in terms of a particularly critical issue which is the energy management in
LMI. A finite state machine is elaborated to control the EMSHV.
between the different organisms [9]. To control the system, a
To demonstrate the efficacy of the designed observation strategy,
a large study, conducted by numerical simulations using hybrid interval observer is designed by using a sensorless
Matlab/Simulink, is performed. algorithm to estimate the unmeasurable magnitude, to ensure
the robustness of the controller and to increase the reliability
Keywords— Interval observer, Energy management system, of the system.
Hybrid model, DC-DC converters, LMI approach.

I. INTRODUCTION
Unlike traditional cars running exclusively on gasoline or
diesel, the hybrid car operates with a heat engine coupled with
an electric motor powered by batteries as a main energy
source. The start-up operations can be done through a
secondary source of energy [1]. The system is formed by two
DC-DC converters and an ultra-capacitor (Fig. 1). In fact,
coupling the ultra-capacitor and the batteries seems to be more
promising in the short term [2], [3], [4]. In this case, the ultra-
capacitor is presented as an auxiliary source to the batteries
where each device is made to play a complementary role [5],
[6]. This is the case of the hybrid vehicle where traction would
be ensured by the batteries, while start-up and acceleration
would benefit from the power supply of the ultra-capacitor
that could be recharged during braking [7], [8]. According to Fig. 1: Hybrid vehicle Scheme
the strategy adopted, there are several families of hybrid
vehicles: Our system often suffers from quite significant disruption
- The micro-hybrid: this is the simplest stage of that is why we use an interval observer. In fact, this observer
hybridization: an alternator-starter allows the momentary is founded on positive dynamical systems theory which gives
shutdown of the engine and its automatic restart. the possibility to estimate the state of the system even in the
- The Serial Hybrid: the electric motor ensures the case where significant disturbances are present [10], [11],
propulsion of the vehicle but the engine work as a source of [12].
energy that recharges the battery if necessary. The aim of this paper is to estimate state variables of a
- The parallel hybrid: the engine provides propulsion as in switched system represented by the EMSHV by using a
a conventional vehicle, but an electric motor is placed between switched interval observer, even in the presence of exogenous
it and the gearbox to optimize the fuel the consumption. disturbances considered as unknown inputs. The synthesis of
- The serial/parallel hybrid: this solution combines the such an observer is addressed via the LMI technique to derive
advantages of the two previous modes. sufficient conditions ensuring the state observer stability.
The paper is composed of four main sections. A hybrid
model for the EMSHV in an electric traction is presented in
section two. The third section presents a design of a switched  diL V V 
interval observer for the proposed system. The fourth section   Bat  OUT  1 VOUT
BOOST

illustrates the performances of the designed method through  dt L1 L1 L1


numerical simulations.  diL V R V
II. HYBRID MODEL OF THE ENERGY MANAGEMENT SYSTEM   UC  UC iL  OUT 3
BUCK
(2)
 dt L2 L2 L2
BUCK

A fuel cell or a battery can provide energy management in


an electric vehicle as a main power source and an ultra-
 dV i i V
capacitor module as an auxiliary one.
 OUT  L (1  1 )  L 3  OUT
BOOST BUCK

Two DC-DC converters control and stabilize the DC-bus


 dt CF CF RL CF
voltage. In fact, a boost DC-DC converter forms the system,
which raises the battery voltage. The other reversible DC-DC We can group the two models into one by selecting a variable
converter functions on a buck mode in case the electrical ρ4 given by
power is converted from the load to the ultra-capacitor (when
it brakes) and functions on a boost mode otherwise (when 4  (1  2 )U  3 (1  U )
energy demands) [13], [14].
The global scheme of the system is illustrated by figure 2. where U=1 in the boost mode and U=0 in the buck mode.
By combining (1) with (2), we get the global system model. It
yields:

 diL V V 
  Bat  OUT  1 VOUT
BOOST

 dt L1 L1 L1
 diL V R V
  UC  UC iL  OUT  4
BUCK
(3)
 dt L2 L2 L22


 dVOUT  iL (1   )  iL   VOUT
BOOST BUCK

 dt CF
1
CF
4
RL CF

 dVUC iL
 dt   C
BUCK

 UC

Additionally, system (3) provides a hybrid system given by


Fig. 2 : Global EMSHV scheme.  x(t )  Aq ( t ) x (t )  Bu (t )  d (t )
It is noted that three IGBT are controlled by ρ1, ρ2 and ρ3  (4)
signals.  y (t )  Cx (t )
where x(t )  , u (t )  , y(t )  reflect respectively
n m p
To stabilize Vout, we propose the two following cases:
the state vector, the output vector and the control input vector.
Case 1: The second converter functions as a boost. Thereby,
Aq(t)   , q(t)=1...4, are the matrices given by different
4x 4
the state-space representation of the system is given by
states of ρ1 and ρ4. The unknown input d(t) is bounded by two
 diL V V  known functions such that
  Bat  OUT  1 VOUT
BOOST

t  0 d  (t )  d (t )  d  (t )
 dt L1 L1 L1
 diL V R V The four matrices A1, A2, A3 and A4 are given by the following
  UC  UC iL  OUT (1   2 )
BUCK
(1) expressions:
 dt L2 L2 L2
BUCK

 dV i i V
 OUT  L (1  1 )  L (1   2 )  OUT
BOOST BUCK

 dt CF CF RL CF
Case 2: The second converter functions as a buck. It follows
the system state-space such as
 0 
1 
0 0 0 0   0 0

 RUC 1  
L1
 where
0 - 0   R 1 

A1  
L2 L2 
A 
 0  UC
L2
0
L2 


T12  T43  x  R , x1  iL1max
4

1
 x  R , x 
0 0 - 0 
2
 
 RL CF  C
1
0 
1
0  T21  T34 4
1
 iL 1min
   F
RL C F

0 1
0  T14  T23  x  R , x  iL2 max 
4
- 0  1 
 
2
CUC  0  0 0 
   x  R , x 
CUC
T41  T32 4
 iL2 min
 0 
2
1
0  0 0 
  0 0 0

L1
  R 1 1  Mode q(t)= 1 T’14
0    Mode q(t)=4
 1 
UC
RUC 1
 0    L2  ρ1=1 ρ4=1 ρ1=1 ρ4=0
L 
L L2
2
T’41
A  
L2 L2
A3   Boost1=On Boost1=On
2
1 1
 1 1 1 
4
0  0  Buck2=On
   Buck2=Off
C 0  C F
RL C F
 F
CF RL C F
  
  0  1 0 0 

1
  T’21 T’12 T’34 T’43
 0 0 0  C
 
UC
CUC

The state, output and input vector, together with the input and Mode q(t)=2 T’23 Mode q(t)=3
observation matrix are expressed as follows: ρ1=0 ρ4=1 ρ1=0 ρ4=0
Boost1=Off T’32 Boost1=Off
1 Buck2=On Buck2=Off
i  L  1 0 0 0  iLBOOST 
LBOOST

     
1

x (t )  
i
LBUCK ,  
B   0  , C  0 1 0 0 , y (t )   iL Fig. 4: The finite state machine for transitions in buck mode.
V    
0
BUCK

 OUT
 0 0 0 1  V  with
   UC 
 V   0 
 
UC

T '12  T '43  x  R , x1  iL1max


4

and u (t )  VBat .
T '21  T '34  x  R , x  iL 1min
4
1
It is worth noting that both DC-DC converters are controlled
T '14  T '23  x  R , x 
 i ' L2 max
4
by the current. In order to move from one mode to another, 2
reference currents iL1max, iL1min, iL2max, iL2min, i’L2max and i’L2min
T '41  T '32  x  R , x 
 i 'L2 min
4
must be fixed for each converter (Fig. 3 for boost mode and 2
Fig. 4 for buck mode). Then, the following two conditions will
be satisfied. III. SWITCHED INTERVAL OBSERVER DESIGN FOR THE ENERGY
MANAGEMENT SYSTEM
iL  iL  iL and iL  iL  iL
1 min 1 1 max 2 min 2 2 max
In this section, we present a prominent technique of
switching interval observer for system (4) in the following
form [15]:
Mode q(t)= 1 T14 Mode q(t)=4  x  (t )  ( Aq ( t )  Lq ( t ) C ) x  (t )  Bu (t )  Lq ( t ) y (t )  d  (t )
ρ1=1 ρ4=1 ρ1=1 ρ4=0  
T41  
Boost1=On Boost1=On  x (t )  ( Aq ( t )  Lq ( t ) C ) x (t )  Bu (t )  Lq ( t ) y (t )  d (t) (5)
Boost2=On Boost2=Off   
 x (t 0 )  x(t)  x (t 0 )
 
T21 T12 T34 T43 where x (t)  x(t)  x (t) t  0
The interval error is given by e(t) = x (t) − x−(t), which is the+

difference between the upper and lower estimated states. The


Mode q(t)=2 T23 Mode q(t)=3
global dynamic error e(t) may be given by
ρ1=0 ρ4=1 ρ1=0 ρ4=0
T32
Boost1=Off Boost1=Off    
e (t )  ( Aq ( t )  Lq ( t ) C ) e (t)   (t )
Boost2=On Boost2=Off    
(6)
e (t )  ( Aq ( t )  Lq ( t ) C ) e (t)   (t )

Fig. 3: The finite state machine for transitions in boost mode.
with By putting M q ( t )  PLq ( t ) and    , then we obtain the
2


 
(t )  d (t )  d (t ) LMI constraint presented in (8).


 
(t )  d (t )  d (t )  AqT( t ) P  C T M qT( t )  PAq ( t )  M q ( t )C   q ( t ) P P 
  0
 I n 
and the total dynamic error is expressed by  P

Moreover, the inequality (9) given by PAq (t )  M q (t )C   P  0


e(t )  ( Aq ( t )  Lq ( t ) C )e(t )   (t ) (7)
is obtained by considering the matrix ( Aq (t )  Lq (t )C) as a
where  (t )   (t )   (t ) .
 
Metzler matrix so that ( Aq (t )  Lq (t )C)  I n  0 . If multiplying
Theorem: If there exist a positive definite diagonal matrix P, the left side of this inequality by P then it yields (9). This ends
real positive scalar ξ and matrices M q (t ) which satisfy the the proof.
following LMI: IV. NUMERICAL SIMULATIONS

 AqT( t ) P  C T M qT( t )  PAq ( t )  M q ( t ) C   q ( t ) P P  The numerical simulations were performed by using Matlab
  0
 I n 
(8) YALMIP toolbox for the LMI synthesis. The used numerical
 P parameter values are given by L1=0.330H, L2=0.330H,
PAq (t )  M q(t )C   P  0 Cb=0.0008F, Rsc=0.001Ω, Csc=70F and RBus=100Ω.
As a matter of fact, three cases of the DC-DC converter are
(9)
tested by simulation as follows:
with αq(t), ξ and σ are positive real constants, and
Case 1: The reversible DC-DC converter works in boost mode
1
Lq ( t )  P M q ( t ) to elevate the Vout voltage (30V in our case), the control mode
signal q(t) is generated from the finite state machine as shown
Then, the switched interval observer (5) leads to a positive in Fig. 5, q(t) changes every time when the current exceeds the
estimation error (6) such that the global estimation error is gaps defined in the chart diagram for transitions modes
stable asymptotically and bounded uniformly. illustrated by Fig. 3.
Proof: To enhance the performance of the switching interval
observer, we define  q (t ) a variable such as for a common

Lyapunov function V (t )  e(t ) Pe(t ) with P  PT  0. The


T

time derivate of V (t ) satisfies


V (t )   V (t ) q (t ) (10)
for each mode q(t) of the observation, we have
Fig. 5: The control signal in boost mode.
V  e (( Aq ( t )  Lq ( t ) C ) P  P ( Aq ( t )  Lq ( t ) C ))e   Pe  e P
T T T T
(11)
From (10) and (11), we can deduce that Fig. 6 shows the current IL1 in the Boost mode, this DC-DC
eT (( Aq (t )  Lq (t )C )T P  P( Aq (t )  Lq (t )C ))e   T Pe  converter is current command to maintain the value in certain
limit. IL1+ and IL1- present the upper and lower state estimation
eT P   q (t ) eT Pe (12) of the current IL1.
The total error provided in (7) is stable if the following
inequality is satisfied for each q(t) mode
e (( Aq ( t )  Lq ( t ) C ) P  P ( Aq ( t )  Lq ( t ) C ))e   Pe 
T T T

e P   q ( t ) e Pe  0
T T
(13)

According to the H  criteria, we define    satisfying


n

e 2   2
(14)
which is equivalent to Fig. 6: The IL1 current and its upper and lower estimated states
e e    0
T 2 T
(15) in boost mode.
Owing to constraints formulated in (14) and (15) and the S- Fig. 7 shows the current in the buck/boost DC-DC converter, in
procedure formula [16], we obtain the following inequalities: this case IL2 is postive because the buck/boost converter
eT (( Aq (t )  Lq (t )C )T P  P( Aq (t )  Lq (t )C ))e   T Pe  eT P  (16)
fucntions as a boost. The ultra-capacitor provides the IL2
current to support the baterie in start-up mode for the hybrid
 q (t ) eT Pe   2 T   0 vehicule. IL2+ and IL2- reflect the upper and lower state
estimation of the current IL2. Fig. 8 and Fig. 9 depict
respectively the Vout and Vsc voltage and their upper and lower Fig. 12 highlights the current in the second DC-DC converter.
state estimation in boost mode. In this case IL2 is negative because the second converter works
as a buck, the the ultra-capacitor recovers the IL2 current in
braking mode for the hybrid vehicule. IL2+ and IL2-
demonstrate the upper and lower state estimation of the
current IL2. Fig. 11 illustrates the current in the first DC-DC.
Fig. 13 and Fig. 14 show respectively the Vout and Vsc voltage
and their upper and lower state estimation in buck mode.

Fig. 7: The IL2 current and its upper and lower state estimation
in boost mode.

Fig. 12: The IL2 current and its upper and lower state
estimations in buck mode.

Fig. 8: The Vout voltage and its upper and lower estimated
states in boost mode.

Fig. 13: The Vout voltage and its upper and lower state
estimations in buck mode.

Fig. 9: The Vuc voltage and its upper and lower estimated states
in Boost mode.
Case 2: The reversible DC-DC converter works in buck mode
which causes the charge of the ultra-capacitor and the
decrease of the Vout voltage (25V in our case).

Fig. 14: The Vuc voltage and its upper and lower state
estimations in buck mode.

Case 3: The reversible DC-DC converter works in buck/boost


mode, which causes the charge and discharge of the ultra-
capacitor. The Vout voltage increase when the second
Fig. 10: The control signal in buck mode. converter works as a boost and decrease when it works as a
buck.

Fig. 11: The IL1 current and its upper and lower state
estimations in buck mode. Fig. 15: The control signal in buck/boost mode.
carried out numerical simulations have demonstrated the
exceptional performances of the observer's design technique.
Such an observer has given prominent, precise and robust state
estimation despite the inherent unknown input considered as
an external disturbance affecting the behavior of the
considered hybrid process.
REFERENCES
Fig. 16: The IL1 current and its upper and lower estimated [1] S. Ifqir, N. Ait Oufroukh, D. Ichalal, S. Mammar, "Switched interval
states in buck/boost mode. observer for uncertain continuous-time systems", in Proceedings of the
20th IFAC WC, 2017.
[2] R. Mura, V. Utkin and S. Onori, "Energy management design in hybrid
Fig. 17 shows that the IL2 current changes from positive value electric vehicles: A novel optimality and stability framework", IEEE
(in start-up of the vehicle) to negative (in braking mode). Fig. Transactions on Control Systems Technology, vol. 23, no. 4, pp. 1307–
16 highlights the current in the first DC-DC. Fig. 18 and Fig. 1322, 2015.
19 depict respectively the Vout and Vsc voltage and their upper [3] A. R. Mayyas, S. Kumar, P. Pisu, et al., "Model-based design validation
and lower state estimation in the buck/boost mode. for advanced energy management strategies for electrified hybrid power
trains using innovative vehicle hardware in the loop (VHIL) approach",
Applied Energy, vol. 204, pp. 287–302, 2017.
[4] M. Castañeda, A. Cano, F. Jurado, H. Sanchez, "Sizing optimization
dynamic modeling and energy management strategies of a stand-alone
pv/hydrogen/battery-based hybrid system", International Journal of
Hydrogen Energy, vol. 38, no. 10, pp. 3830-3845, 2013.
[5] D. Xu, J. Liu, X. Yan, et al., "Novel Adaptive Neural Network
Constrained Control for a Multi-Area Interconnected Power System
with Hybrid Energy Storage", IEEE Transactions on Industrial
Electronics, vol. 65, no. 8, pp. 6625–6634, 2018.
[6] W. Zhenhua, L. Cheng-Chew, S. Limb Yi, "Interval observer design for
Fig. 17: The IL2 current and its upper and lower state uncertain discrete-time linear systems", Systems & Control Letters, vol.
estimations in buck/boost mode. 116, pp 41-46, 2018.
[7] D. Jezuz, J.A. Moren, "Interval observer design for nonlinear systems:
Stability radii approach", IEEE Access, vol 6, pp 52801-52813, 2018.
[8] D. Efimov, T. Raïssi, "Design of interval observers for uncertain
dynamical systems", Automation and Remote Control, vol 77, no. 2, pp
191-225, 2016.
[9] N. Ellero, D. Gucik-Derigny and D. Henry, "Interval observer for Linear
Time Invariant (LTI) uncertain systems with state and unknown input
estimations", Journal of Physics: Conference Series, 12th European
Workshop on Advanced Control and Diagnosis, 2015.
[10] D. Efimov, A. Polyakov, and J.-P. Richard, “Interval observer design for
Fig. 18: The Vout voltage and its upper and lower state estimation and control of time-delay descriptor systems", European
estimations in buck/Boost mode. Journal of Control, vol. 23, no. 5, pp. 26–35, 2015.
[11] F. Bejaoui, C. Mechemeche, A. S. Tlili, "Faults Detection and Isolation
for the Energy Management System based on a Hybrid Observer", 19th
international conference on Sciences and Techniques of
Automaticcontrol & computer engineering (STA), Sousse, Tunisia,
March 24-26, 2019.
[12] F. Cacace, A. Germani, and C. Manes, "A new approach to design
interval observers for linear systems", IEEE Trans. Automatic Control,
vol. 60, no. 6, 2015.
[13] L. Meyer, D. Ichalal, V. Vigneron, "Interval observer for LPV systems
with unknown inputs", IET Control Theory & Applications, vol 12, no.
Fig. 19: The Vuc voltage and its upper and lower state 5, pp 649-660, 2018.
estimations in buck/Boost mode. [14] F. Mazenc, T.N. Dinh, "Continuous-DiscreteInterval Observers for
Systems with Discrete Measurements", 52th IEEE Conference on
Decision and Control, Florence, Italy, pp. 1584-1589, 2013.
V. CONCLUSION [15] S. Chebotarev, D. Efimov, T. Raïssi, and A. Zolghadri, "Interval
In this paper, the EMSHV, formed by the two buck and buck- observers for continuous-time LPV systems with L1=L2 performance",
Automatica, vol. 58, no. 8, pp. 82–89, 2015.
boost DC-DC converters was modeled. In addition, a robust
[16] A. Martínez-Sibaja, C.M. Astorga-Zaragoza, A. Alvarado-Lassman et
switched interval observer with an unknown input was al, "Simplified Interval Observer Scheme: A New Approach for Fault
designed to rebuild the unmeasurable state variables. The Diagnosis in Instruments", Sensors, vol 11, pp 612-622, 2001.

You might also like