Вы находитесь на странице: 1из 8

SICE Journal of Control, Measurement, and System Integration, Vol. 3, No. 1, pp.

035042, January 2010

A New Model Predictive Control Approach to DC-DC Converter


Based on Combinatory Optimization

Abu Zaharin AHMAD and Kang-Zhi LIU

Abstract : DC-DC converters by nature present hybrid behavior, which is described by a set of discrete modes associated
with continuous dynamics. The control objective is accomplished by switching among the discrete modes. This paper
presents a new model predictive control approach to optimize the performance of DC-DC converters. The proposed
control algorithm is tailored to DC-DC converters, making full use of the fact of finite number of modes to transform the
performance optimization problem into a combinatory optimization task. This approach has made the on-line computa-
tion very simple and eective for implementation.

Key Words : model predictive control, DC-DC converter, hybrid control.

1. Introduction tions. Motivated by these facts, a variety of control strategies


DC-DC converters are now widely employed in a variety have been proposed for the improvement of DC-DC convert-
of applications, ranging from computer to medical electronic ers [5][11]. But, the proposed design methods are still based
systems, adapters of consumer electronic devices, spacecraft on the averaged model of the DC-DC converters.
power systems, and telecommunication equipments [1],[2]. Recently, there have been several papers, which apply
The wide range of applications has made this an active research the hybrid control to DC-DC converters. Particularly, in
field where various control strategies have been employed and [3],[4],[12][15] the discrete time model predictive control
investigated. (MPC) is introduced, based on a discrete time piecewise ane
The switched circuit topologies entail dierent modes of op- (PWA) modeling of the converters in which the duty ratio is re-
eration, i.e., on and o modes of switching, which comprises garded as a constrained input. This discrete time PWA model
two semiconductor switches (usually one switching transis- can capture the hybrid dynamics of systems at the sampling in-
tor and one switching diode) that are turned on and o at a stants. Then, MPC schemes have been developed by minimiz-
very high frequency [1]. Each has its own associated linear ing the 1-norm of a certain output voltage error and the change
continuous-time dynamics. These switching operations lead to of duty ratio. This shows that hybrid control methods can be
its hybrid nature, which makes it dicult to analyze and control applied to DC-DC converters. But the implementation has to
DC-DC converters [3],[4]. rely on look-up tables due to the computation complexity.
Traditionally, power electronics switching circuits have In this paper, a new and more direct model predictive control
been controlled in industry using the pulse-width-modulation approach is proposed based on the sampled data from the orig-
(PWM) combined with linear controllers or its modification to inal continuous-time hybrid model of converters. The key idea
regulate the voltage of the load. Much of the typical analysis is to make full use of the property that the number of modes
of switching circuits relies on averaging or discretization tech- is finite and to transform the optimization of the squared area
niques [1],[5]. State-space averaging is perhaps the most pop- of output voltage error over a finite time interval into a search
ular averaging technique that models the converters as a single problem for the optimal switching sequence. This combina-
equation so as to describe the behavior of converter approx- tory optimization problem can be computed eectively within
imately over a switching cycle, by taking the switching duty the stringent switching period. It is shown that the computa-
ratio as an input in the range of [0, 1]. The switching frequency tion of the control index can be greatly simplified via analytic
does not appear in this model and the system trajectories have methods so that the on-line computation reduces to extremely
continuous first derivatives. Another approach is to use a small- simple algebraic calculations. Comparison with traditional
signal linearization of the hybrid model [2],[6]. In these control PID-PWM method shows that the proposed approach is quite
techniques, it is clear that the simplified models are rather lim- promising.
ited in their ability to describe the hybrid system dynamics, thus This paper is organized as follows. In section 2, the hybrid
limiting the resultant control performance and accuracy. model of DC-DC converters is discussed. In section 3, the de-
The control problem is further complicated by the operating tailed MPC algorithm and the treatment of load variations are
point changes due to input voltage change and the load varia- presented. In section 4, the performance of the proposed al-
gorithm is demonstrated through simulations and a comparison
with PID-PWM illustrated.
Department of Electrical and Electronics Engineering, Chiba
University, 133, Yayoi-cho, Inage-ku, Chiba 2638522, Japan
E-mail: zaharin@graduate.chiba-u.ac.jp
(Received April 17, 2009)
(Revised July 9, 2009)

JCMSI 0001/10/03010035 
c 2009 SICE
36 SICE JCMSI, Vol. 3, No. 1, January 2010

index is formulated as a mixed-integer quadratic programming


problem. This problem consists of solving quadratic program-
ming (QP) problem while doing a tree search. Since the QP
problem can only be solved via recursive computation, the
computation time is in general long and not suitable for the
on-line implementation to switching circuits where the com-
putation must be finished within several to tens of micro sec-
Fig. 1 Step-down (buck) DC-DC converter. onds. Method of [12] developed for step-down DC-DC con-
verter along this line bears the same feature and the on-line
2. Dynamic Model of DC-DC Converter computation time is rather long.
A DC-DC converter is a switching circuit, which transforms As an alternative approach that can be implemented on-line
the voltage of a DC source into other desired voltage in the using a look-up table,[3],[4],[13][15] developed an MPC for
load side (the step-down converter in Fig. 1 is used to lower the boost (step-up) converter where the hybrid dynamics is approx-
source voltage). This is achieved through suitable switching of imated as a PWA system, then the minimization of output and
the circuit. So, the objective of the switching control in DC-DC duty ratio errors (1 norm) is solved o-line by using multi-
converter is to realize high power transfer eciency and good parametric programming. The optimal input (duty ratio) is ob-
tracking of output voltage. tained as a PWA function of the state in accordance with a poly-
The inherent switching operation mechanism of DC-DC con- hedral partition of the state space. Meanwhile, in [17] a similar
verter results in the feature that the state transits among a finite MPC method for discrete systems with input constrained in a fi-
number of circuit topologies (modes). As an example, let us nite set is proposed. The main idea is to solve the unconstrained
look at a step-down (buck) converter. The circuit topology of optimization problem first, then using the nearest neighbor vec-
the step-down DC-DC converter with equivalent series resistor tor quantizer to search for the optimal input in the finite set. The
of capacitor and inductor [1] is shown in Fig. 1. algorithm is implemented through a polyhedral partition of the
The state is taken as x = [il vo ]T where il is the inductance state space determined o-line. In these two methods, the de-
current and vo is the output voltage. In practice, the inductance termination of the input boils down to checking on-line which
current il and output voltage vo can be directly measured. Both polyhedron the sampled state is located at.
L and C have equivalent series resistances rl and rc . The con- Our method diers from those of [12],[16] and [13],[17] in
verter features two operation of switching modes (on and o) that on-line implementation is possible without using any table.
and the state equations of each switching mode are described Particularly, compared with [13],[17], in our method the search
by the following ane continuous-time state-space equations: for the optimal input is carried out via on-line computation of
the values of control index along the search tree and comparison
x(t) = Aon x(t) + Bon v s (1) of their values. As will be verified through simulations, for DC-
DC converter a long prediction horizon is not needed. So this
x(t) = Ao x(t) + Bo v s (2) approach can be implemented on-line. Another dierence is
that we can deal with the continuous-time subsystems directly,
where
instead of discretizing the subsystems.

rl 1 The key observation here is that all possible state trajectories
Aon = Ao =  L rr  L  can be predicted on-line since the number of modes is finite,
R+rc C L
R 1 c l R 1 rc
R+rc RC L the dynamics is ane and autonomous (i.e. free of external sig-
nals). So we are able to find the best sequence of switching

, Bo = 0 .
1
Bon = L
Rrc
mode such that the tracking error of output voltage over cer-
L(R+rc ) 0 tain finite time interval (horizon) is minimized. This process
can be reduced to a simple combinatory optimization task. The
Further, the output voltage across the load R is given by main issue to be resolved is how to simplify the computation of
y = vo = Cx, C = [0, 1]. (3) tracking error prediction.

This means that the state will evolve between these two 3.1 Computation of Error Index
modes according to the switching operation. The tracking of For simplicity, let index = 1 denote the on mode and = 2
reference output voltage is achieved via suitable switching con- denote the o mode. Then the dynamics (1)-(3) of converter is
trol. The step-up (boost) converter has a similar feature. conveniently described as
As shown in this example, each mode of DC-DC converters x(t) = A x(t) + B u. (4)
has an ane continuous-time dynamics, in the sense that the
y(t) = vo (t) = Cx(t)
input signal is a constant. So, the switching control of such
ane systems will be discussed in detail. in which = {1, 2}. It is worth noting that u = v s is a constant
and all A are stable. So, the solution x(t) in t (t0 , t) with
3. MPC Algorithm for Ane Continuous-Time initial state x(t0 ) is computed as
Switching Systems x(t) = eA (tt0 ) x(t0 ) (I eA (tt0 ) )A1
B u
For hybrid systems, there have been many MPC methods
= eA (tt0 ) (x(t0 ) + A1 1
Bu) A B u. (5)
proposed. For example, in [16] a mixed-integer predictive con-
trol method is proposed in which the optimization of control Therefore, the output is calculated as
SICE JCMSI, Vol. 3, No. 1, January 2010 37

y(t) = CeA (tt0 ) (x(t0 ) + A1 1


Bu) CA B u. (6)

To enforce the tracking of the reference output voltage vr , we


consider the minimization of the squared area of the tracking
error e(t) over a finite prediction horizon:

e(t) = y(t) vr (7)


t1
t1
J(t1 , t0 ) = e2 (t)dt = (y(t) vr )2 dt. (8) Fig. 2 Illustration of search tree for prediction horizon l = 2.
t0 t0

Here, only the tracking error of voltage is penalized because 3.2 Model Predictive Control Algorithm
the sole purpose of a converter is to track the reference voltage. In our approach, the fundamental switching period1 is set as
Due to this simple index, the computation burden is reduced the sampling period T . The control problem is to calculate a
substantially in implementation. This is dierent from the con- sequence of switching modes so that the output voltage vr tracks
trol indexes of [3],[4],[12][15] where the errors of current and the reference as good as possible, via the minimization of the
duty ratio are also penalized. index function (8) over a finite prediction horizon l. We denote
As the first step, we must simplify the computation of this the corresponding switching sequence by = [(1), . . . , (l)]
index function. This is done as follows. First, assuming that where (k) {1, 2}
the state is in mode , we define a vector and a scalar as The optimizing sequence must be recalculated at each con-
follows: secutive sampling instant because the influence of disturbance
cannot be predicted. This influence of disturbance will change
= x(t0 ) + A1
Bu, = CA1
B u vr . (9) the consecutive optimal switching mode.
The optimizing algorithm is summarized as follows:
Expanding e2 (t) yields
i Compute Q (T ), CA1
(e
A T
I) o-line for all P. Then
T
e (t) =
2
2 + T eA (tt0 )C T CeA (tt0 ) + 2Ce A (tt0 )
. (10) repeat the following procedure.

Then it is easy to see that ii Sample the state at t = t0 and calculate and of (9) for

all P.
t1
T
J (t1 , t0 ) = (t1 t0 )2 + T eA (tt0 )C T CeA (tt0 ) dt iii Compute the predicted state at all consecutive switching
t0

t1
instants from t0 + T up to t0 + lT for all switching sequence
+ 2C eA (tt0 ) dt (11) .
t0
iv Compute the index J(k) (t0 +kT, t0 +(k1)T ) for all switch-
Here, a subscript is added to the index function in order to in- ing sequence and k {1, . . . , l}.
dicate that the state is in mode . This equation can further
be simplified by using Lyapunov equation to compute the ob- v Search for the optimal switching sequence such that
servability grammian. Since A is stable, there exists a matrix
t0 +lT
X > 0 satisfying
I (t0 + lT, t0 ) = e2 (t)dt
t0
AT X + X A + C T C = 0 (12)
= J(1) (t0 + T, t0 ) + . . . + J(l) (t0 + lT, t0 + (l 1)T ) (15)
Then it holds that

is minimal. That is,
t1
AT (tt0 )
Q (t1 t0 ) := e C Ce T A (tt0 )
dt = arg min I (t0 + lT, t0 ) (16)
t0

t1 t0
=
T
d(eA t X eA t ) As an example, the search tree for the case of prediction hori-
0 zon l = 2 is illustrated in Fig. 2 where the value of index for
= X eA (t1 t0 ) X eA (t1 t0 ) .
T
(13) each branch is:
a. J1 (t0 + T, t0 ) + J1 (t0 + 2T, t0 + T )
Substituting equation (13) back into (11), the index function is
rewritten as: b. J1 (t0 + T, t0 ) + J2 (t0 + 2T, t0 + T )

J (t1 , t0 ) = (t1 t0 )2 + T Q (t1 t0 ) c. J2 (t0 + T, t0 ) + J1 (t0 + 2T, t0 + T )


+ 2CA1
(e
A (t1 t0 )
I) . (14) d. J2 (t0 + T, t0 ) + J2 (t0 + 2T, t0 + T )

What is nice with this formula is that all data except can be Comparison of these values leads to the optimal switching
computed o-line, and can be computed once the initial state sequence.
is sampled. This significantly simplifies the on-line implemen- 1
The fundamental switching period is the shortest time interval
tation. between any two consecutive switching.
38 SICE JCMSI, Vol. 3, No. 1, January 2010

Table 1 Step-down DC-DC converter parameters.

Parameters Value
vs 24[V]
vr 12[V]
L 15.91[mH]
Fig. 3 Robustification of MPC with external PID control loop.
C 94[F]
R 4[]
3.3 Load Variations
rc 0.001[]
It is obvious that the MPC algorithm depends on the param-
rl 0.5179[]
eter of load. Through simulations, it is observed that when the
load changes the output voltage converges to a dierent con- Table 2 Parameters of PID-PWM controller.
stant value. To robustify the proposed MPC algorithm, we have
Parameters Value
tried adding an external loop with PI or PID compensators (as
shown in Fig. 3). It is discovered that the PID compensator Carrier frequency, fc 100[kHz]
yields better transient response. So the configuration of Fig. 3 KP 115
is adopted. The PID controller in the external loop of the system KI 0.9
adjusts the output voltage reference of MPC to so as to provide kD 0.01
the output tracking when the load varies. In principle, the inte-
grator I compensates the steady-state tracking error, while the Table 3 Parameters of PID compensator.
PD part adjusts the transient response. Parameters Value

4. Simulation Results KP 40

In this section, the eectiveness of the proposed control KI 700


algorithm will be investigated through computer simulation. KD 0.009
The results of a step-down (buck) converter will be presented
Table 4 Computational time for prediction horizon l.
and comparison will be made with well tuned traditional PID-
PWM controller. The nominal DC-DC converter parameters Prediction horizon l Computational time
are shown in Table 1. The parameters of PID-PWM controller l=4 1.63[s]
are shown in Table 2, while Table 3 provides the parameters l=5 3.39[s]
of PID compensator used for the robustification of MPC (refer
l=6 6.91[s]
to Fig. 3). The simulation is carried out on the SimPowerSys-
l=7 13.95[s]
tems toolbox of MATLAB and four cases are investigated. In
l=8 28.03[s]
practical, the DC-DC converters usually operate over 100[kHz]
of switching frequency. For comparison the carrier frequency
of PWM is set as the fundamental switching frequency and the SH7206 in which each multiplication takes 5 109[s], the
proposed MPC operating with a sampling period T = 10[s]. increasing of computational time is shown in Table 4 (based on
Numerous prediction horizons l have been tested in simu- SH7206 microcontroller).
lations. The results showed that better output responses were
4.1 Case 1
obtained for l 4 (some of them are illustrated in Fig. 4). But
the results also indicated that the performance for l > 4 were The output response and the inductance current il during the
roughly the same as l = 4, no significant improvement was start-up are illustrated in Figs. 4(ad) and 5(a) for the proposed
observed. Therefore, in order to reduce the computation time MPC, and in Figs. 4(e) and 5(b) for the PID-PWM, respectively.
l = 4 is selected. In [13] a prediction horizon l = 2 is used Compared with the PID-PWM, the output voltage settles to its
and the corresponding polyhedral partition has 121 regions af- stationary value with virtually no overshoot for the proposed
ter simplification. So the determination of the state location in MPC.
these regions is still quite time-consuming. Meanwhile, in our The number of switchings in 0 0.02[s] is 526 for the pro-
approach the number of combinations is only 16 even for l = 4, posed MPC, while the PID-PWM controller needs 1514 of
hence the computation burden is relatively lighter. switchings, as depicted in Figs. 6(a, c) and 6(b, d) respectively.
In computer implementation, the computation time is mainly Since the number of switching obtained for proposed MPC is
determined by the number of multiplications. The number of less than PID-PWM controller, the power loss associated with
combinations (branches of the search tree) is 2l for a prediction switching control is reduced which enhances the power e-
horizon l. In each branch of l = 1, the number of multiplications ciency of DC-DC converters.
is 9. Meanwhile, for each branch behind l = 1 the number of
4.2 Case 2
multiplications is 11. So, the total number N of multiplications
becomes The second case is devoted to the investigation of the step
variation of reference output vr from 12[V] to 8[V]. The output
N = 18, l=1 response and inductance current il are depicted in Figs. 7 and
N = 18 + (4 + 8 + . . . + 2 ) 11,
l
l > 1. 8 for the proposed MPC and PID-PWM, respectively. It is ob-
viously shown that the proposed MPC has no overshoot during
When the algorithm is implemented on a microcontroller, e.g. the start-up for vr = 8[V], compared with the PID-PWM.
SICE JCMSI, Vol. 3, No. 1, January 2010 39

Fig. 5 Inductance current il .

Fig. 4 Start-up behaviour of vo .


Fig. 6 Switching sequence. (continue on next page)
40 SICE JCMSI, Vol. 3, No. 1, January 2010

Fig. 6 (continued)

Fig. 9 Eect of increase of input voltage v s (24[V] to 32[V] at t =


0.03[s]).

Fig. 7 Start-up behaviour of vo (vr = 8[V]).

Fig. 10 Eect of decrease of input voltage v s (24[V] to 15[V] at t =


0.03[s]).

voltage v s in which v s is increased to 32[V] from its nominal


value at 0.03[s] during the steady-state operation is illustrated in
Figs. 9(a) and 9(b) for the proposed MPC and PID-PWM con-
troller, respectively. The magnitude of output error increases
due to the increase of input voltage. The result shows that the
proposed MPC has roughly a similar level of output error com-
pared with the PID-PWM. The average value and variance of
Fig. 8 Inductance current il (vr = 8[V]). output voltage when the input voltage increase are 12.083[V]
and 0.00396[V] for proposed MPC, while the PID-PWM yields
4.3 Case 3 12.092[V] and 0.00567[V].
The third case examines the output response for the varia- The input voltage is then reduced to 15[V] from the nominal
tion of input voltage v s . The eect of change in the input value at 0.03[s] during the steady-state operation as shown in
SICE JCMSI, Vol. 3, No. 1, January 2010 41

Fig. 11 Load variation (4 to 6 at t = 0.03[s], vo ). Fig. 13 Load variation (4 to 2 at t = 0.03[s], vo ).

Fig. 12 Load variation (4 to 6 at t = 0.03[s], il ). Fig. 14 Load variation (4 to 2 at t = 0.03[s], il ).

Figs. 10(a) and 10(b) for the proposed MPC and the PID-PWM, current il are illustrated in Figs. 12(a) and 12(b) for proposed
respectively. It is obvious that the proposed MPC is not aected MPC and PID-PWM, respectively.
significantly and comparable with PID-PWM. Figures 13(a) and 13(b) illustrate the drop of load resistance
R from 4 to 2 (50% decrease from nominal value), for the
4.4 Case 4 proposed MPC and the PID-PWM controller, respectively. The
The fourth case investigates the step variation of the load re- result of proposed MPC is roughly similar to the PID-PWM
sistance R during the steady-state operation. First, load varia- controller. However, the propose MPC has a better transient
tion is tested with the load resistance R raised from 4 to 6 response for inductance current il compared to PID-PWM con-
(50% increase from nominal value), as depicted in Figs. 11(a) troller, as depicted in Figs. 14(a) and 14(b).
and 11(b) for the proposed MPC and PID-PWM controller, re-
spectively. The result shows that the proposed MPC controller 5. Conclusion
has less undershoot and the steady-state performance is com- In this paper, a new model predictive control approach to DC-
parable to the PID-PWM controller. The results of inductance DC converters has been proposed. By making full use of the hy-
42 SICE JCMSI, Vol. 3, No. 1, January 2010

brid feature of DC-DC converters, the MPC has been reduced Riedinger, T. Geyer, A. Beccuti, G. Papafotiou, M. Morari,
to a combinatory optimization procedure, which can be imple- A. Wernrud, and A. Rantzer: Hybrid control techniques for
mented on-line eectively. The control index has been simpli- switched-mode DC-DC converters Part I: The step-down topol-
fied via analytic methods. The continuous-time hybrid dynam- ogy, Proc. ACC07, pp. 54505457, 2007.
[15] A. G. Beccuti, G. Papafotiou, M. Morari, S. Almer, H. Fujioka,
ics of converters is not approximated in any sense, contrary to
A. Wernrud, Rantzer, M. Baja, H. Cormerais, and J. Buisson:
many existing MPC methods. Generally, the simulation results Hybrid control techniques for switched-mode DC-DC convert-
show that the proposed MPC controller has good performance ers Part II: The step-up topology, Proc. ACC07, pp. 5464
when operated at high frequency. 5471, 2007.
Further, to improve the steady-state performance it is impor- [16] A. Bemporad and M. Morari: Control of systems integrating
tant to introduce the notion of duty ratio into the proposed logic, dynamics and constraints, Automatica, Vol. 35, No. 3,
MPC method. This will be addressed in a forthcoming paper. pp. 407427, 1999.
[17] D. E. Quevedo, G. C. Goodwin, and J. A. De Dona: Finite
Finally, although only the example of step-down (buck)
constraint set receding horizon quadratic control, International
converter is shown in this paper due to page limitation, the Journal of Robust and Nonlinear Control, Vol. 14, No. 4,
proposed MPC algorithm can be applied to step-up (boost) pp. 355377, 2004.
converters directly.

Acknowledgements Abu Zaharin AHMAD


He received the B.S. and M.S. degrees in electrical
The authors would like to thank the anonymous reviewers
engineering from University Technnology MARA, Shah
for their comments and suggestions which helped improving Alam, Malaysia in 2001 and 2004, respectively. He then
the quality of this paper. worked as a maintainance engineer at PFC Engineer-
ing before joining University Malaysia Pahang. He is
References
presently pursuing his PhD degree at Chiba University,
[1] A. Simon and O. Alejandro: Power Switching Converters, 2nd Japan.
Edition, Taylor & francis, CRC, 2005.
[2] H. Sira-Ramirez and R. Silva-Ortigoza: Control Design Tech- Kang-Zhi LIU (Member)
niques in Power Electronics Devices, Springer, 2006. He received the Ph.D. degree from Chiba University in
[3] T. Geyer, G. Papafotiou, and M. Morari: Model predictive 1991 and has been with this university since then. He
control in power electronics: A hybrid systems approach, is currently an Associate Professor at the Department of
Proc. CDC-ECC05, pp. 56065611, 2005. Electrical and Electronics Engineering. His interest is in
[4] G. Papafotiou, T. Geyer, and M. Morari: Hybrid modeling and power electronics control, switching control and robust
optimal control of switch-mode DC-DC converters, Proc. IEEE control. He has authored four books and numerous jour-
Workshop Computers in Power Electronics, pp. 148155, 2004. nal papers. He has also received four awards from SICE,
[5] A. J. Forsyth and S. V. Mollov: Modeling and control of DC- Japan.
DC converters, Power Engineering Journal, Vol. 12, No. 5,
pp. 229236, 1998.
[6] A. Nogueiras-Melendez, J. D. Gandoy, C. Martinez-Penalver,
and A. Lago: A new complete non-linear simulation model of
a buck DC-DC converter, Proc. ISIE99, Vol. 1, pp. 257261,
1999.
[7] B. Arbetter and D. Maksimovic: Feedforward pulse witdh
modulator for switching power converter, IEEE Trans. Power
Electronics, Vol. 12, No. 2, pp. 361368, 1997.
[8] Y. B. Shtessel, O. A. Raznopolov, and L. A. Ozerov: Sliding
mode control of multiple modular DC-to-DC power converters
in conventional and dynamic sliding surfaces, IEEE Trans. Cir-
cuit & Systems 1: Vol. 45, No. 10, pp. 10911100, 1998.
[9] J.-H. Su, J.-J. Chen, and D.-S. Wu: Learning Feed-
bank Controller Design of Switching Converters Via MAT-
LAB/SIMULINK, IEEE Trans. Education, Vol. 45, No. 4,
pp. 307315, 2002.
[10] T. Gupta, R. R. Broudreaux, R. M. Nelms, and J. Y. Hung: Im-
plementation of a fuzzy controller for DC-DC converters using
an expensive 8-b microcontroller, IEEE Trans. Industrial Elec-
tronics, Vol. 44, No. 5, pp. 661669, 1997.
[11] H. S. Ramfrez: Nonlinear P-I controller design for switchmode
DC-to-DC power converters, IEEE Trans. Circuits & Systems
1, Vol. 38, No. 4, pp. 410417, 1991.
[12] T. Geyer, G. Papafotiou, and M. Morari: On the optimal control
of switch-mode DC-DC converters, Proc. HSCC 04, LNCS,
Vol. 2993, pp. 342356, 2004.
[13] A. G. Beccuti, G. Papafotiou, and M. Morari: Explicit hybrid
model predictive control of the boost DC-DC converter, Proc.
PESC, pp. 25032509, 2006.
[14] S. Almer, H. Fujioka, U. Jonssson, C.-Y. Kao, D. Patino, P.

Вам также может понравиться