You are on page 1of 4

DESIGN AND ROBUSTNESS ANALYSIS OF A STATE FEEDBACK

CONTROLLER APPLIED TO A BIDIRECTIONAL BOOST CONVERTER

V.F. Montagner1, A. A. Ferreira2


1
Federal University of Santa Maria, GEPOC
Av. Roraima, 1000, 97105-900, Santa Maria - RS - Brazil
2
Federal University of Pampa, Campus Alegrete
Av. Tiarajú, 810, 97546-550, Alegrete - RS - Brazil
vfmontagner@gmail.com andre.cta.unipampa@gmail.com

Abstract – This paper investigates the application of a regulator that aims to drive to zero the deviations of the
state feedback control technique to a bidirectional boost states of the plant from the desired values, with a prescribed
converter. The proposed control system is dynamics given by pole location specifications. Simulation is
mathematically formulated such that the state feedback used to illustrate the good capacity of regulation of the
control gains become the gains of a proportional-integral closed-loop system and also its robustness to parameter
controller for the voltage loop and a proportional gain for uncertainty.
the current loop. The control design does not rely on
manual tuning procedures neither use complex adaptive
techniques. On the contrary, the control gains are rapidly
and automatically calculated off-line, ensuring good II. PROBLEM STATEMENT
transient and good steady state responses. A robust
analysis evaluated a posteriori demonstrate that the Consider the bidirectional boost converter subject to open
closed-loop system has good robustness against loop control, shown in Figure 1 (for details see, for instance,
uncertainty on the DC input voltage and on the load [1]).
parameter.
Keywords – Control design conditions, state feedback
control, boost converter, robustness.

I. INTRODUCTION

Modern control techniques, based on state space Fig. 1. Bidirectional boost converter with open loop control.
representation of the control system, are undoubtedly The duty cycle is 0.5 (output signal from voltage
important in many areas of applied control (see, for instance, comparator), the nominal load is Ro=188.3 Ω and the load
[2], [3], [5]). In this context, state feedback control strategies can be increased by turning on the switch sw1. The other
are particularly useful due to the fact that they allow to parameters of the converter are:
investigate controllability and observability of the system
and to use several tools already developed for control design, • 100 V of voltage source (Vi);
as pole placement, linear quadratic regulator, and others. • 114 mΩ of internal resistance of voltage source (Ri);
• 500 μH of inductor (L);
Power electronics applications have demanded high
• 300 μF of output capacitor (Co);
performance controllers in order to meet more and more
stringent specifications [4]. However, very frequently, the • 74 mΩ of internal resistance of output capacitor (Rseo);
control design for power electronics applications relies on • 0.5 V of voltage reference (Vref).
heuristic procedures carried out in frequency domain. The transient response of the open loop control system is
Certainly, this approach has lead to several good results in given in Figure 2. One can notice from Figure 2 that the
practice, providing to the control designer a deeper output voltage reaches 200 V in steady state but the transient
knowledge of the plant. On the other hand, it can be time response can be improved by reducing the oscillations, the
consuming and also can lead to a poor closed-loop behavior, overshoot and the settling time. In addition, the transient
depending on the expertise of the control designer. inductor current is too high and vary fast, that is not suitable
This paper presents a control design suitable to provide to electrochemical energy conversion devices, such as
state feedback gains that ensure good transient response and batteries and fuel cells. Improvements on the system
good steady state behavior for a bidirectional boost converter response can be achieved by means of closed-loop control
[1], [6]. This design tool can be useful for the control based on state space techniques.
engineer and for the power electronics engineer to compute
the control gains without iterative procedures nor trial and
error based methods. The controller is described as a

978-1-4244-3370-4/09/$25.00 © 2009 IEEE 385


D1 = D2 = [0 RseoCo]
250
Output voltage (V)

200

150
The operation point is given by
100
50
Inductor current (A)

X = − A−1 BE
0

Now, one gets matrices of system (1) by


-50

A(δ ) = A
Output current (A)

0
B(δ ) = ( A1 − A2 ) X + ( B1 − B2 ) E
0 0.05 0.1 0.15 0.2 0.25
Time (s)

Fig. 2. Response of the open loop system. The switch sw1 is C (δ ) = C + DA


turned on during the interval 0.1 to 0.2 seconds.
D(δ ) = (C1 − C 2 ) X + DB(δ )
III. STATE SPACE CONTROL DESIGN
Notice that system (1) is a representation of the deviations
The converter in Figure 1 admits the following small of the state variables from their steady state values. In steady
signal model (for details, see Chapter 4 in [1]) state, such errors must be driven to zero in order to ensure
~
good regulation. The control variable is d and the pair
~ ~
x = A(δ ) ~
x + B (δ )d (A(δ), B(δ)) is controllable. Then, by means of the state
~ (1) feedback control
~y = C (δ ) ~
x + D(δ )d
~
where the system matrices and the operation point depend on d = − K~
x (2)
the nominal parameters of the circuit and on the value of the one can place the closed-loop poles at any stable point on the
steady state duty cycle δ. The state space variables are the complex plane [5]. The control gains
inductor current (x1) and the the output voltage capacitor
K = [0.0067 0.0004 ] assign the closed-loop poles at the
(x2). Matrices A(δ), B(δ), C(δ) and D(δ)are obtained based on
the auxiliary matrices given below. points –1000 and –2000. The control system with these gains
was simulated, respecting the limitations of the signals using
A = δ A1 + (1 − δ ) A2 saturation blocks, as shown in Figure 3.

where

⎡− Ri / L 0 ⎤
A1 = ⎢
⎣ 0 − 1 /( RoCo)⎥⎦

⎡− ( Ri + koRseo) / L − ko / L ⎤
A2 = ⎢
⎣ 1 / Co − 1 /( RoCo)⎥⎦
and
ko = Ro /( Rseo + Ro)
B = δB1 + (1 − δ ) B2 Fig. 3: Closed-loop system with gains K = [0.0067 0.004].

where Notice from Figure 3 that the control signal is the sum of
0.5 (fixed part concerning the steady state duty cycle) with
⎡1 / L ⎤ ~ ~
B1 = B2 = ⎢ ⎥ d given by (2). d has the objective to eliminate the errors of
⎣0 ⎦ the state variables with respect to their desired values as fast
as imposed by the pole placement specification. The results
C = δC1 + (1 − δ )C2 for the system output are shown in Figure 4.

where The initial inductor current is still too high, but this effect
C1 = C 2 = [0 1] can be avoided by using a ramp output voltage reference
from 100 V to 200 V by 25 ms.
D = δD1 + (1 − δ ) D2
Comparing Figure 4 (closed-loop system based on pole
where location) with Figure 2 (open loop system) it is apparent the
improvement provided by the state feedback control strategy

978-1-4244-3370-4/09/$25.00 © 2009 IEEE 386


in terms of faster transient, without oscillation. However,
from 0.15 to 0.2 seconds, when the load is increased, one can
notice the loss of regulation of the closed-loop system. This
points out to the need of including an integral action over the
error of the output voltage.
Output voltage (V)
Inductor current (A)

Fig. 5: Closed-loop system with integral action and gains


Output current (A)

K = [0.0088 0.0039 2.4135].

Output voltage (V)


Fig. 4: Response of the closed-loop system shown in Figure 3.

Since the error on the output variable is given by


e=~
y (4)
Inductor current (A)

one has that


Output current (A)

Δ t t
σ~ = ∫ e dτ = ∫ ~y dτ (5)
0 0

which allows to write Fig. 6: Response of the closed-loop system shown in Fig. 5,
~ with output voltage ramp reference.
σ~ = ~y = C (δ ) ~x + D(δ )d (6)
It is worth to notice that the design using integral action
Notice that σ~ represents the integral of the error over the provides a straight and automatic procedure to obtain a PI
output variable. Including this variable in the description of controller for the voltage loop and a proportional controller
system (1), one has the following augmented system for the current loop. There is no need of iterative procedures
neither exhaustive simulation to obtain the PI controller
⎡~x ⎤ ⎡ A(δ ) 0⎤ ⎡ ~
x⎤ ⎡ B (δ ) ⎤ ~ suitable to meet the performance specifications, thanks to the
⎢ ~ ⎥ = ⎢ ⎥ ⎢ ~⎥ + ⎢ D (δ )⎥ d (7) state feedback approach used here to formulate the control
⎣σ ⎦ ⎣C (δ ) 0⎦ ⎣σ ⎦ ⎣ ⎦ problem.
System (7), with the parameters of the converter under
investigation here, is controllable, allowing to impose IV. ROBUSTNESS ANALYSIS
arbitrary pole location. The control gains
This section presents two analyses of robustness of the
K = [0.0088 0.0039 2.4135] (8) closed-loop system with gains (8). First, consider that the DC
input voltage E can assume any value in the interval
ensure the pole location at –800, –1000 and –2000. [90, 100]V. Since the parameter E affects the system
The control system with these gains was simulated, matrices, one has that the pole location imposed for the
respecting the limitations of the signals using saturation nominal situation is changed by variation on E. Figure 7
blocks, as shown in Figure 5, and the output of this system is shows the trajectories of the closed-loop poles when E varies
given by Figure 6. from 90V (lower DC input voltage) to 100V (nominal DC
input voltage). It can be observed that the operation under
The effect of the integral action, driving to zero the steady lower DC input voltages leads to slower settling times and
state error when the system operates with load Rop connected, oscillations on the transient response. However, for the entire
is apparent in Figure 6 (compare with Figure 4). Moreover, range of values of DC input voltage, the closed-loop poles
one can note the fast transient, with no oscillation, thanks to have real part lower than –500, and then the transient
the chosen pole location. responses remain fast in any case, as corroborated by Figure
8, illustrating that the control gains in (8) ensures robust
stability with good performance for this uncertainty on E.

978-1-4244-3370-4/09/$25.00 © 2009 IEEE 387


800
V. CONCLUSION AND PERSPECTIVES
600

400
This paper presented conditions to design state feedback
controllers for a bidirectional boost converter. Based on the
200
small signal state space representation of the system, one
Imainary axis

0 proposes a controller to drive to zero the deviations of the


-200
state variables with respect to the desired values. This
controller is designed based on pole location techniques,
-400
using available specialized routines, being an easy alternative
-600 for the control designer to obtain a controller with
-800
specification in terms of pole location. The design conditions
-2000 -1500 -1000 -500
Real Axis of section III compute the control gains in less than 1
Fig. 7: Closed-loop poles for E=90, 91, ..., 100 V. E=100V is the millisecond when solved in Core Due 1.66 GHz, 1GB RAM
nominal value of this parameter. computer, which points out for high saving of time during
the control design stage since there is no heuristic neither
manual tuning procedures to find the control gains. The
simulations carried out in the paper show good performance
and robust stability against parameter uncertainty. As a
perspective for future work, notice the augmented system
Voltage (V)

matrices (7) can be written as a polytope of matrices (see, for


instance, [7]) to encompass uncertain parameters affecting
the system and then advanced robust control techniques
Inductor current (A)

including robust pole location and rejection of disturbances


can be used.

ACKNOWLEDGEMENT
Fig. 8: Controlled output and DC input voltage (upper graphic) and
inductor current (lower graphic). Grants from UFSM - Programa de Auxílio à Pesquisa de
Now, consider the load as an uncertain parameter, allowed Recém Doutores (research project 024336).
to assume values in the interval [94.15,188.3] Ω. The closed-
loop poles are shown in Figure 9. The poles for the nominal REFERENCES
load (188.3 Ω) are represented by the filled small ellipses
and the poles for the increased load (94.15 Ω) are pointed by [1] A. A. Ferreira, Sistema supervisório de gestão de
arrows. Clearly, the performance become poorer for the múltiplas fontes de suprimento para aplicações em
situation where the load increases, but the closed-loop veículos elétricos, PhD thesis, in Portuguese,
system still remains fast enough for a good transient response University of Campinas, Unicamp, 2007.
(slower pole with real part lower than –600).
[2] C. T. Chen. Linear System Theory and Design. Oxford
From the robustness analysis of the proposed controller University Press, 3rd edition, 1999.
(8), one can conclude that the closed-loop stability is
[3] J. C. Doyle, B. A. Francis, A. R. Tannenbaum.
preserved for the uncertainties on the load and on the DC
Feedback Control Theory. Macmillan Publishing
input voltage tested here. Moreover, the transient responses
Company, New York, NY, 1992.
remain fast, thus pointing out for good performance of the
proposed controller. [4] G. Kassakian, M. F. Schlecht, G. C. Verghese,
150 Principles of Power Electronics, Addison-Wesley,
Boston, MA, 1991.
100
[5] K. Ogata. Modern Control Engineering. Prentice-Hall,
50
NJ, 1990.
Imaginary axis

[6] H. Tao, A. Kotsopoulos, J. L. Duarte, M. A. M.


Hendrix, “Family of Multiport Bidirectional DC-DC
0

-50
Converters”, IEE Proceedings Electric Power
Applications, v. 153, n. 3, p. 451 – 458, 2006.
-100
[7] V. F. Montagner, P. L. D. Peres, “Design of robust
repetitive controllers based on convex optimization
-150
-2000 -1800 -1600 -1400
Real axis
-1200 -1000 -800 -600
with an application to power electronics”, Proceeding
of 9th Brazilian Power Electronics Conference, p. 22 –
Fig. 9: Closed-loop poles for Ro=188.3 Ω (marked as ellipses) and
25, 2007.
for Ro=94.15 Ω (pointed by arrows).

978-1-4244-3370-4/09/$25.00 © 2009 IEEE 388

You might also like