Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

2011 IEEE International Conference on

Automation Science and Engineering


Trieste, Italy - August 24-27,2011

SaA1.1

Hybrid Modeling and Control of Switching


DC-DC Converters via MLD Systems
Mohammad Hejri* and Alessandro Giuat
* School of Electrical Engineering
KTH-Royal Institute of Technology, Stockholm, Sweden
Email: moh arnmad.hejri@ee.kth.se
tDepartment of Electrical and Electronic Engineering
University of Cagliari, Italy, Email: giua@diee.unica.it

Abstract-Modeling and control of DC-DC converters is a non


trivial problem in power electronics. Such converters exhibit
both autonomous and non-autonomous switching phenomena
which are commonly modeled by general classes of hybrid
systems. This paper addresses the modeling and control of a
DC-DC buck converter via Mixed Logical Dynamical (MLD)
systems and hybrid predictive control based on Mixed Integer
Quadratic Programming. The main contribution of the paper is
the novel concept of Forward MLD as opposed to the classical
modeling approach that we call Backward MLD. The advantage
of the proposed method is to introduce a prediction of the
continuous behavior at a the switching instant. This results in
a more accurate model and allowing to reduce by one step the
prediction horizon necessary to compute the optimal control.
The proposed technique reduces the complexity of the resulting
MIQP program. The transient and steady state performances of
the closed-loop control system over a wide range of operating
points show satisfactory operation of the proposed modeling
and control scheme.

I. INTRODUCTION
Hybrid modeling and control of DC-DC converters has
attracted interest in the fields of power electronics and auto
matic control theory. However, developing a general model
and control technique which considers all possible dynamics
of a converter is still a challenging issue.
Several approaches have been reported in literature for
modeling and control of DC-DC converters [1]-[11]. In
[12] and [13], the principles of operation of these methods
are reviewed and the shortcomings of such methods are
mentioned. The main drawback of all the existing methods is
that their design is based on one operating mode of operation,
i.e. either continuous current mode (CCM) or discontinuous
current mode (DCM). In CCM the current in the inductor
fluctuates during a cycle but never goes down to zero. On
the contrary, in DCM the current in the inductor fluctuates
during the cycle, going down to zero at or before the end
of each cycle. Therefore, the performance and stability of
the closed-loop system is guaranteed only for one operation
mode. Recently in [14], several hybrid control techniques
from different research groups were introduced for DC-DC
This work was done when the first author was a Ph.D student in
Department of Electrical and Electronic Engineering,University of Cagliari,
Italy,Email: mohammad.hejri@diee.unica.it

978-14577-1732-1/11/$26.00 2011 IEEE

buck and boost converters. However, all of these works deal


with only the CCM of operation.
In some applications like battery powered systems [15], to
improve the overall efficiency, it is desirable the converter
operates in CCM under full-load conditions and in DCM
under light-load conditions. In such cases, the operation of
the converter may move from CCM to DCM or vice versa
during load and line disturbances.
MLD modeling and predictive control of a DC-DC buck
converter in both CCM and DCM operations are presented
in [16]. However, the approach is based on averaging and
approximated modeling techniques. Hybrid modeling and
control of buck and boost converters in MLD framework,
considering all possible dynamics in CCM and DCM opera
tion was already introduced in [12], [13]. However, in these
works the reset dynamics in transition from continuous to
discontinuous current modes were not considered.
In our work, the reset dynamic in transition from CCM
operation to DCM one is also considered. The reset dy
namic arises from the discrete-time nature of MLD modeling
framework and the inability of exact detecting uncontrolled
switching surfaces because of technical limitation in selecting
small descretization sample times. Considering the reset
dynamic, results in a most exact discrete-time model which
can be obtained for the converter in MLD framework.
This paper introduces the novel concept of Forward
Mixed Logical Dynamical (FMLD) systems and compares
it with the existing one [17] which we call as a Backward
Mixed Logical Dynamical (BMLD) systems. In BMLD,
when combining continuous affine dynamics into a single
MLD framework, each dynamic is multiplied by a binary
variable which represents the previous and current values
of the hybrid automaton discrete state. In this modeling
approach, an unnecessary one step sampling delay appears
among the switching of different continuous dynamics of
the hybrid automaton. To solve this problem, a new MLD
modeling approach, i.e. FMLD is introduced in this paper. In
FMLD, different continuous dynamics of hybrid automaton
are merged together by mUltiplying each continuous dynamic
by an integer variable which represents the current and next
values of the discrete state of the hybrid automaton. With this
approach, the continuous dynamic of the system changes as

714

soon as the discrete state changes and any unnecessary delay


in the BMLD disappears.
The FMLD and BMLD models presented in this work are
improved as compared to the ones in [12], [13] in terms of
model accuracy and less optimization complexity.
The one step delay in BMLD systems does not allow the
continuous state of the system to react synchronously to the
change of discrete states. This implies that to achieve the
same performance objectives in BMLD and FMLD systems
for a DHA containing both controlled and uncontrolled
switching phenomena, the prediction horizon in the BMLD
model is set at least one step more than that of the FMLD
one. This is a possible problem for BMLD models in terms of
complexity of the resulting MIQPIMILP problem. This is due
to the fact that all auxiliary integer and continuous variables
appear one time more in the associated optimization problem
of the BMLD model as compared to the FMLD one [18].
The FMLD approach, that we describe in this paper for the
case of the DC-DC buck converter may also be applied to
the modeling and control of other complex power electronics
converters. The paper is organized as follows. Principles of
mixed logical dynamical modeling are discussed in section
II. The FMLD and BMLD modeling of a buck converter
in three modes of operation is described in section III. The
hybrid predictive control and objective function selection
is explained in section IV. Simulation results are presented
in section V. We discuss the computational burden of the
optimization procedure in section VI. Finally, concluding
remarks are made in section VII.
II. MIXED LOGICAL DYNAMICAL(MLD) SYSTEMS

yields continuous part of hybrid dynamic. The interaction


between these two parts are made with two interconnecting
elements, i.e., Mode Selector (MS) and Event Generator
(EG). The overall MLD model equations in (1) is extracted
Event
Generator
(EG)

oe(k)

FinHe State
Machine

(FSM)

/0

0-\:0

Switched
Affine
System (SAS)
Mode

xb(k) Selector
:-+I(MS)
.,..
u-..,.
b(k)
oe(k)
=Cx>-

l..----

i(k)

1-------'

=D-

Fig. I.

discrete hybrid automaton and its components [19]

by the modeling of different blocks in Fig. 1 separately and


then combining the models [17], [19].
III. MLD MODELING

OF

A BUCK CONVERTER

A conventional DC-DC buck converter with parasitic el


ements is shown in Fig. 2. Let the continuous states of the

Discrete-time hybrid systems are modeled as Mixed Log


ical Dynamical (MLD) systems in [17]. The main idea in
MLD is transforming logic and constraints of a complex
system into mixed-integer inequalities with propositional
calculus. An MLD system can be described by the following
relations [17]:

x(k + 1) = Ax(k) + B1U(k) + B26(k) + B3Z(k)


y(k) = Cx(k) + D1U(k) + D2 6(k) + D3Z(k)
E2 6(k) + E3Z(k) :::; E1u(k) + E4X(k) + Es
where

Rnc

kEN

{O, l}nl denotes the states,

{O, l}ml are the inputs and

y =

U =

[ ; ] E

[ ; ] E

is the discrete-time instant,

[ ;; ]E

RPc

(1)

Rmc

{a, l}PI are

the outputs, with both real and bmary components.


Furthermore, 6E {O,1 yl and zE RTc represent binary
and auxiliary continuous variables, while A, B/s, C, D i'S
and Ei'S are suitable matrices.
Fig. 1 shows a schematic diagram of a discrete hybrid
automaton (DHA). Such a model evolves in discrete time
and is a decomposed model of hybrid systems. A DHA is
obtained from the interconnection of two main parts: a Finite
State Machine (FSM) which provides discrete dynamic of
hybrid system and a Switched Affine System (SAS) which

Fig. 2.

Buck Converter

system be defined as xc(t) = [idt), vo(t) V which are the


inductor current and converter output voltage respectively.
There are three discrete modes of operation for a buck
converter. In mode 1, the switch is ON and the diode is
OFF. In this mode, energy is supplied from the supplier to
magnetic field of inductor. The state equation in this mode,
is:
(2)
where:

715

_'!:.J..
L
r
-.l
c
Ro+rc (l-Crc LL)
v,
L

v: Ro TC
SRo+TC L

1
-r

_-.l_
_
l
(l+Crcfi)
c Ro+rc
L

(3)

In Mode 2, the switch is OFF and the diode is ON. The


stored energy in magnetic field of inductor is delivered to
the load and capacitor. The state equation is:
(4)
where:

A2

AI,h2

[ ]

The Discrete-Time Hybrid Automaton (DHA) correspond


ing to the buck converter with different modes of operation
is shown in Fig. 3. In this figure, N in idk +N) is set to 2
for BMLD model and 1 for FMLD one. A reset dynamic as

(5)

In Mode 3, both the switch and diode are OFF. The required
energy for the load is supplied from stored energy in electric
field of capacitor. The state space equation in this mode is:
(6)

xe(k + 1) = Ae,xC<k) + he,

where:

ud(k)=O

(7)
We define Ud(t)E {O, I} as a discrete input which shows
the state of the switch. Ud(t) = 1 when the switch is ON
and Ud(t) = 0 when it is OFF. The basic converter equation
in continuous time is described as:
if
if
if

Ud(t) = 1
Ud(t) = 0/\ idt) > 0
Ud(t) = 0/\ idt) :::; 0

(8)

To use the MLD framework and MPC algorithm, It IS


required to change all continuous-time state space equations
to discrete-time domain with suitable sampling switching
period Ts.
Therefore, by defining xc(k) = [idk), vo(k) V as the
discrete time and normalized continuous state vector, the
discrete time system in (8) is described as:

Ac,xc(k) + hc, ifud(k) = 1


AC2XC(k) + hC2 if ud(k) = 0/\ idk) > 0
AC3XC(k) + hC3 if ud(k) = 0/\ idk) :::; 0
(9)
where ACi = eAi Ts, hCi = J Ts eAi dT hi, iE {I, 2,3}.
o

xc(k+1) =

Because of the discrete time nature of (9), there are some


instants at which in transition from mode 2 to mode 3, the
inductor current takes negative values that is unrealistic result
in ideal analysis of electric circuits. To solve this problem,
it is necessary to use the predictive enabling [18], [19]. It
means that when the converter is in mode 2, the enabling
of mode 3 is predicted one step ahead. However, since
switching conditions in (9) depend on the discrete input,
such prediction is impossible. This is due to the fact that at
instant k, the value ud(k + 1) is not predictable. To use the
advantage of predictive enabling, it is necessary to describe
the converter's dynamic as a DHA. With this approach,
the transition from mode 2 to mode 3 is represented as a
autonomous switching with a reset dynamic. Therefore, one
can force the inductor current to zero via one step predictive
enabling of autonomous edge and FMLD idea. Also, since the
continuous dynamic in mode 2 is independent of the discrete
input, one can force the inductor current to zero via two step
predictive enabling of uncontrolled edge in BMLD approach.

Fig. 3.

Xd, =

[]

DHA of a buck converter

in (10) is considered to the edge between discrete modes 2


and 3 which are shown by binary vectors [1 O]T and [0 oV
respectively.

xc(k + 1) =

[ ] (AC3XC(k) + hC3)

AC4XC(k) + hC4

(10)
We define xd(k) = [xd,(k),Xd2(k) V as a discrete state
vector in the hybrid automaton of the buck converter in Fig.
3. This figure has three discrete states, i.e. [1 O]T,[O I V
and [0 O]T that are corresponding to the modes 1,2 and 3,
respectively.
First, the condition idk +N) :::;
a binary variable 51 (k) such that:

0 can be associated

with

(11)
where, iL(k + N) = [1 O](Axc(k) + L:Ol A2hc2) ' The
evolution of the logic state vector Xd(k) can be written as
the following equation:

where:

5j, = 0/\512 = 0
5j, = 0/\512 = 1
--+ 5j, = 0/\512 = 1
--+ 5j, = 1/\512 = 0
1 /\
1 /\ (j f2
0 /\ <h
0 /\
0 --+ (j It
Xd, = 1/\ Xd2 = 0/\51 = 1 --+ 5j, = 0/\512 = 0
Xd1 = 1/\ Xd2 = 0/\ Ud = 1 --+ 5j, = 0/\512 = 1

Xd, = 0/\ Xd2 = 0/\ Ud = 0


Xd, = 0/\ Xd2 = 0/\ Ud = 1
xd, = 0/\ xd2 = 1/\ Ud = 1
Xd, = 0/\ Xd2 = 1/\ Ud = 0
Xd,

Xd2

Ud

--+

--+

(13)
In BMLD approach, we define a new discrete state bit Xd3(k)
that denotes the value of Xd,(k) during the previous step, i.e.,

xd3(k) = xd,(k - 1).

716

The new binary variable

5;;3 is now introduced as:

Now, the overall continuous dynamic of the converter can be


written as:

defined in the following form to reduce the output voltage


error and switching frequency.

xc(k + 1)= (AC1 XC(k) + hcJXd2(k) + (AC2XC(k)+


BC2)Xd1 (k) + (AC3XC(k) + hc,)(l - Xd2(k) - Xd1 (k)
-83(k)) + (AC4XC(k) + BC4)83(k)= AC3XC(k) + hC3+
+[(ACl - AC3)Xc(k) + hCl - hC3lxd2(k) + [(AC2 - Ac,)
xc(k) + hC2 - hC3lxdl (k) + [(AC4 - AC3)Xc(k) + hC4 -hc3l83(k)
New auxiliary variables
defined as:

Zl (k)
z!2(k)
z!3(k)

Zll (k), z2(k)

(15)
and z3(k) are

[(AC2 - Ac,)xc(k) + hC2 - hC3lxd1 (k)


[(ACl - AC3)Xc(k) + hCl - hC3lxd2(k)
[(AC4 - Ac,)xc(k) + hC4 - hC3l83(k)
(16)

Using the definition in (16), (15) can be expressed as:

xc(k + 1)=
AC3XC(k) + hC3 + Zl (k) + z2(k) + z3(k)

(17)

For the FMLD formulation, we define the integer variable

8!'3(k) as:

min

{u(O),... ,u(Tu-1),
,,(Olt),. ,,,(Tp-llt),
z(Olt),. ,z(Tp-llt)}

Er==lIIQ(y(klt) - r(t))11 2+

+ Er:::olIIR( Ud(k) - ud(k - 1)11 2


s.t.

(22)

x(Olt)= x(t)
FMLD or BMLD model
x(llt), ,x(Tp)

Xmin

Xmax

Where Tp is prediction horizon, Tu is control horizon, y(klt)


and x(klt) are the predicted output and state at instant k +t,
r(t) is output reference signal, Q is weighting coefficient to
penalize output signal error and R is a weighting coefficient
to penalize big changes in input signal. The main controller
in (22) is slightly modified to force the output to follow the
reference signal with zero steady state error in the presence of
unmeasured disturbances and plant-model mismatches. This
is the same approach which is used in linear model predictive
control to achieve zero steady state error tracking [20]. All
is necessary is to measure the difference between the latest
plant output and the latest model output which is an estimate
for the unknown disturbance d(k) at instant k, i.e.:

d(klk)= y(k) - y(klk - 1)

where the hat sign indicates the predicted value of the cor
responding parameter. The disturbance is assumed constant
during the prediction horizon, thus:

Now the overall continuous dynamic can be written as:

Xc(k + 1)= (AC1 XC(k) + hc1 )8h(k) + (AC2XC(k)+


BC2)8h (k) + (AC3XC(k) + hc,)(l - 812(k) - 8h (k)
-8!,3(k)) + (AC4XC(k) + hc4)8!'3(k)= AC3XC(k) + hC3+
+[(ACl - AC3)Xc(k) + hd1 - hd3l8h(k) + [(AC2 - AC3)
xc(k) + hC2 - hc3l8h (k) + [(AC4 - Ac,)xc(k) + hC4 -hc3l8!,3(k)

w;: define the new auxiliary variables Z 1 (k),


ZX23(k) are defined as:
Z;ll (k)
Z;22(k)
Z;23(k)

(19)
Zi,2(k) and

(20)
By substituting of (20) in (19), the overall continuous dy
namic in the FMLD framework is written as:

IV.

y(k + ilk)

Cx(k + ilk) + DI U(k + i) + D28(k + i) +


(24)
+D3Z(k + i) + d(k + ilk)
where d(k + ilk)= d(klk).
This equation is now used in place of y(k) in (1) which
offers more realistic prediction of the plant output under
disturbances and plant-model mismatches.
Problem in (22) is a mixed integer quadratic programm
that is solved via a free matlab function miqp.m [21].
V.

[(AC2 - AC3)Xc(k) + hC2 - hc3l8h (k)


[(ACl - AC3)Xc(k) + hCl - hc3l8h(k)
[(AC4 - AC3)Xc(k) + hC4 - hC3l83(k)

xc(k + 1)=
AC3XC(k) + hC3 + Z;ll (k) + Zi,2(k) + Zi,3(k)

(23)

(21)

HYBRID PREDICT IV E CONT ROL

In general, the control objective in DC-DC converters is to


regulate the average output voltage as fast as possible with
the lowest overshoot despite changes in the input voltage
Vs or changes in the load resistance Ro considering to the
constraints on the inductor current. The objective function is

SIMUL AT ION RESULTS

In this section, simulation results of the proposed control


strategy are presented. In this paper, all the states and
elements of the converter are expressed in per unit (p.u.)
system. The value of the various parameters in p.u. system
lies in a narrow range, though the actual values vary widely.
This provides computational tractability by the p.u. values
as compared to the real ones, and therefore, p.u. values are
ideal in computer simulations. The real values can be easily
computed from the p.u. values by selecting suitable base
values. For example, if the frequency base in a converter
is ib = 10k.H z, then the time base is Tb = "* = 100p,sec,
and therefore, each time step O.lp.u. in the simulations of
this paper is equal to: 0.1 x 100= lOp,sec.
Circuit parameters of the plant in the simulation are given
in Table I [22]. Sampling switching period Ts is taken
equal to 0.1 p.u .. In all cases, the penalty matrix Q, R
and control horizon Tu in (22) are selected as 1, 0.00003

71 7

TABLE I
CONVERTER PARAMETERS USED FOR SIMULATIONS

c
TC

Parameters of the converter (in p.u.)


p.u.
L
ilmax
2'" p.u.
0.005 p.u.
TL
0.05 p.u.
Ro
2

(a)

3 p.u.
1 p.u.

FMLD.
... .. BMLD.

__

10
nme(p.u.)
(b)

FMLD.
BMLD.
. . . BMLD.

.. . .

___

10
nme(p.u.)

15

10

T =1
p
T =2
p

15

__

(a)

20

30

50
40
Time(p.u.)
(b)

60

70

80

60

70

80

-FMLD
BMLD.I
I........
20
10

T =1
p
T =2
p
T =1
p

20

30

40
50
Time(p.u.)

Fig. 5.
System performance for large output resistance variation, (a)
inductor current,(b) output voltage

(a)

....... BMLDI
I-FMLD

20

Fig. 4.
Converter start by the M PC for FMLD and BMLD models (a)
inductor current,(b) output voltage

10

and 1, respectively. First, the prediction horizon is taken


as Tp = 1. Because of one step delay in BMLD model,
the converter is not started by the MPC for this modeling
framework, and the simulation results backs this fact, Fig.
4 (dashed line). However, as it can be seen from Fig. 4
(solid line), using the FMLD model the converter starts well.
To achieve the same results with the FMLD model, the
prediction horizon in BMLD model is increased from Tp = 1
to Tp = 2, Fig. 4 (dotted-line). As a result, the complexity
of the resulting optimization is increased considerably as
compared to the FMLD one. This is due to the fact that
all auxiliary integer and continues variables are appeared
one time more in the resulting optimization problem. Figs.
Sea) and (b) show the inductor current and output voltage
of the converter associated with the FMLD and BMLD
models when a very large load change from Ro = 1 p.u.
to Ro = 10 p.u. at step time k = 20 and then from
Ro
10 p.u. to a very small value Ro
0.05 p.u. at
step time k = 50 occur. This small value for the output
resistance is a simulation for a short circuit on the output of
the converter. Converter nominal load is recovered at k = 60.
As it can be seen, because of the light load at k = 20, the
converter shifts from continuous mode to deep discontinuous
mode of operation. Also, when a short circuit occurs at the
output of the converter at k = 50, it is clear from Figs. Sea)
and (b) that the controller holds the current limit and forces
the output voltage to drop to a level that the maximum current
limit of the inductor is held. Current constraint satisfaction
during the controller design is a valuable protective action in
practical applications. Figs. 6(a) and (b) show the inductor
=

20

30

40
50
Time(p.u.)
(b)

60

70

80

40
50
Time(p.u.)

60

70

80

....... BMLDI
I-FMLD

:;
Co
:;
o

10

20

30

Fig. 6.
System performance when reference signal variation occurs,(a)
current waveform,(b) voltage waveform

current and output voltage of the converter during output


voltage reference voltage variation. In this simulation, the
output voltage reference VOTe! changes from 1p.u. to a lower
value of 0.2p.u. at step time k = 20. It then returns to its
initial value 1p.u. at step time k = 60. As it can be seen, the
converter enters DCM operation for lower reference signal
values. The closed-loop performance over a wide range of
operation points is desirable.
VI.

COM PLEX ITY OF COM PUTAT IONS

It is well known that computational complexity of the


MILP and MIQP programs is strongly dependent on the num
ber of optimization variables in which the integer variables
have a very important role [23]. MIQP problem in (22) has
Tu(mc + ml) + Tp(rl + rc) optimization variable of which
Tu(mt} + Tp(rl) are integer. Any MIQP problem with nd
binary variables can be associated to the 2(nd+1) -1 quadratic

718

programmes in the tree structure of the branch and bound


algorithm [23]. Since in general, the prediction horizon in
FMLD approach is one step less than the corresponding
BMLD one, we can conclude that the size of explored
tree in optimal control of FMLD model is less than the
BMLD one. By considering Tp and Tp + 1 as the prediction
horizons for the optimal control of FMLD and BMLD models
respectively, and Tu as a common control horizon, one can
quantify the reduction in the size of the exploration spaces
2Tu(=I)+Tp(rl) -1

as RFB = 2Tu(=I)+(Tp+l)(rl)_1' where RFB IS the ratlO of


the number of nodes in the related search trees of MIQP
problems in FMLD and BMLD. Assuming the power terms
in RFB are much greater than 1, this quantity is simplified
as RFB = 2;1 . This means that the size of exploration tree
in MIQP problem of BMLD model is larger than FMLD
one from the order 2Tl. Therefore, in problems with a
large number of auxiliary binary variables, the computational
advantage of FMLD over BMLD is more highlighted.
VII.

CONCLUSION

In this paper, the novel concept of Forward Mixed Logical


Dynamical systems is introduced as compared to the existing
one, i.e. Backward Mixed Logical Dynamical systems. In the
FMLDs continuous dynamics at different discrete states of
discrete hybrid automaton are multiplied by the next value
of discrete states. While in the BMLDs, they are multiplied
by the existing value of discrete states. One of the advantages
of FMLDs is the elimination of unnecessary one step delay
in BMLDs that results in the ability of choosing shorter
prediction horizon in FMLDs and complexity reduction in
the optimal control of such systems. A new application
of the proposed modeling approaches to global modeling
and control of a DC-DC buck converter operating in both
continuous and discontinuous current modes is presented and
compared. Simulation results show that the proposed control
scheme is quite successful in both CCM and DCM, and
the transient response is desirable with respect to line and
load disturbances. Hybrid predictive control scheme enables
us to take into account physical constraints of the system
during controller design especially the maximum inductor
current. With this ability, a short circuit at the converter
output is handled by dropping the output voltage such that
the inductor current limit is held. Online computation of
the optimal control problem in FMLD model requires less
computation times as compared with BMLD one. But it is
still above the sampling times of the high frequency con
verters. Thus, for real time high frequency applications, the
state feedback controller can be calculated as a look-up table
by off-line solving the optimization problem in the overall
state space with multi-parametric programming techniques.
Hopefully, with the adventure of high-speed processors the
online computations can be done in the future.
REFERENCES
[1] F. H. F. Leung, P. K. S. Tam, and C. K. Li, "The control of
switching dc-dc converters- a general lqr problem," IEEE Transactions
on Industrial Electronics" vol. 38,no. 1,pp. 65-71,February 1991.

[2] H. SiraRamirez and P. LischinshyArenas,"The differential algebraic


approach in nonlinear dynamical compensator design for dcdc power
converters," International journal control, vol. 54,no. 1,pp. 111-134,
1991.
[3] H. Sira-Ramirez, R. A. Perez-Moreno, R. Ortega, and M. Garcia
Esteban, "Passivity-based controllers for stabilization of dc-dc power
converters," automatica, vol. 39,no. 4,pp. 499-513,1997.
[4] S. R. Sanders and G. C. Verghese, "Lyapanov based control for
switched power converters," IEEE Transaction on power electronics,
vol. 7,no. 1,jan. 1992.
[5] M. Lazar and R. D. Keyser,"Non-linear predictive control of a dc
dc converter,
" Symposium on Power Electronics, Electrical Drives,
Automation and Motion (SPEEDAM),Capri, Italy, 2004.
[6] H. Sira-Ramirez,"On the generalized pi sliding mode control of dc-dc
power converters:a tutorial;' International journal of control, vol. 76,
no. 9-10,pp. 1019-1033,2003.
[7] P. Gupta and A. Patra,"Hybrid mode-switched control of dc-dc boost
converter circuits," IEEE Transactions on Circuits and Systems, vol. 52,
no. 11,November 2005.
[8] 1. Buisson,P.-Y. Richard,,and H. Cormerais,"On the stabilisation of
switching electrical power converters," Hybrid Systems: Computation
and Control, Springer Verlag, vol. 3414,pp. 184--197,2005.
[9] A. Nobuhiro, Z. Tadanao, and 1. Muneaki, "Optimal control of dc
dc converter using mixed logical dynamical system model," IEEJ
Transactions on Industry Applications, vol. 127, no. 3, pp. 339-346,
2007.
[10] S. Mathew,E. Gabriel,and T. J. Koo, "Hybrid modeling and control
of power electronics," Hybrid Systems: Computations and Control,
Lecture Notes in Computer Science,Springer, pp. 450-465,2003.
[11] G. Escobar, A. 1. van der schaft, and R. Ortega, "A hmiltonian
viewpoint in the modeling of switching power converters," Automatica,
vol. 35,no. 3,pp. 445-452,march 1999.
[12] M. Hejri and H. Mokhtari, "Global hybrid modeling and control of
a buck converter: A novel concept," International Journal of Circuit
T heory and Applications, John Wiley and Sons, Inc., vol. 37, no. 9,
pp. 968 - 986,November 2009.
[13]
,"Hybrid predictive control of a DC-DC boost converter in both
continuous and discontinous current modes of operations," Optimal
Control, Applications and Methods, John Wiley and Sons, Inc., vol. 32,
no. 3,pp. 270-284,May/June 2011.
[14] S. Marilthoz,S. Almer,M. Baja, A. Beccuti,D. Patino, A. Wemrud,
1. Buisson,H. Cormerais,H. Fujioka,T. Geyer,U. Jonsson,C.-Y. Kao,
M. Morari,G. Papafotiou, A. Rantzer,and P. Riedinger,"Comparison
of Hybrid Control Techniques for Buck and Boost DC-DC Converters,"
IEEE Transactions on Control Systems Technology, vol. 18,no. 5,pp.
1126 - 1145, August 2010.
[15] S. Zhou and G. A. Rincn-Mora, "A high efficiency, soft switching
dcdc converter with adaptive current-ripple control for portable appli
cations," IEEE Transactions on Circuits and SystemsII. Express Briefs,
vol. 53,no. 4,pp. 319-323, April 2006.
[16] A. G. Beccuti, G. Papafotiou, and M. Morari, "Optimal control
of the buck dc-dc converter operating in both the continuous and
discontinuous conduction regimes," in IEEE Conference on Decision
and Control,USA, December 2006,pp. 6205-6210.
[17] A. Bemporad and M. Morari, "Control of systems integrating logic,
dynamic and constraints," Automatica, vol. 35, no. 3, pp. 407-427,
March 1999.
[18] M. Hejri, "Hybrid modeling and control of power electrinic convert
ers;' Ph.D. dissertation,University of CagJiari Cotutorship with Sharif
University of Technology,May 2010.
[19] F. D. Torrisi and A. Bemporad, "Hysdel-a tool for generating com
putational hybrid models for analysis and synthesis problems," IEEE
Transactions on Control Systems Technology, vol. 12,no. 2,p. 235249,
March 2004.
[20] K. R. Muske and J. B. Rawlings,"Model predictive control with linear
models," A.I.Ch.E. Journal, vol. 39,no. 2,pp. 262-287,1993.
[21] A. Bemporad and D. Mignone,"miqp.m: a matlab function for solving
mixed integer quadratic programs," Technical Report, 2000.
[22] T. Geyer, G. Papafotiou, and M. Morari, "On the optimal control of
switch-mode dc-dc converters," In Hybrid Systems:Computation and
control,Springer-Verlag, vol. 2993,pp. 342-356,March 2004.
[23] C. A. Floudas, Nonlinear and mixed-integer optimization.
Oxford:
University Press,1995.

719

--

You might also like