Sei sulla pagina 1di 7

SPEEDAM 2008 International Symposium on Power Electronics, Electrical Drives, Automation and Motion

Analysis and Control of a Bidirectional Two-Stage Boost Converter


J.V. Gragger*, F.A. Himmelstoss**, and F. Pirker*
* Arsenal Research, Giefinggasse 2, 1210 Vienna, (Austria) ** University of Applied Sciences Technikum Wien, Hoechstaedtplatz 5, 1200 Vienna, (Austria) Abstract In this work a two-stage boost converter circuit is investigated. Relations for the dimensioning of the converter elements are given and a realistic converter model is derived and presented using state-space description. The transfer functions for the output voltage depending on the duty ratio and the input voltage are calculated and topologies of possible voltage control systems are suggested. The proposed converter can be used to boost up the battery supply voltage e.g. for electric auxiliary drives in automotive applications. Index TermsConverters, DC-DC power conversion, power supplies, voltage control

Fig. 1. Two-stage boost converter.

For these reasons a voltage control system for the twostage boost converter suggests itself in many cases. II. BASIC ANALYSIS Only idealized components are considered for the basic analysis. The ohmic component in the storage inductors L1 and L2, the equivalent series resistances (ESRs) of the buffer capacitors C1 and C2, and the conduction losses and the switching losses of the converter switches are neglected. Furthermore, the converter is assumed to operate in steady-state with continuous inductor current mode (CICM). With the additional assumption that the buffer capacitors are large enough to keep the output voltage of each stage nearly constant it is well known that the voltage transformation ratio of a single boost converter stage with the conventional topology MA can be found by
MA = VOUT 1 = VIN 1 d

I. INTRODUCTION The proposed two-stage boost converter is illustrated in Fig. 1. By using such a converter a high DC voltage transformation rate can be realized while the losses in the components can be kept at acceptable levels. In contrast to other step-up converters no transformer is used in the proposed topology [1]. A simpler version of the presented converter would need only two active switches (S1, S3) and two passive switches (S2, S4) realized by diodes. While the first converter stage is a low voltage (high current) unit and therefore can be built with active switches with low turn-on resistances the current in the second converter stage is considerably lower and the voltage is higher. Another advantage of the proposed circuit is that the switches S1 and S3, can be controlled without potential-free driver units. It is possible to use the presented converter for boosting a low battery voltage in order to supply e.g. electric auxiliaries in automotive applications such as an electrified air conditioning system. In such an air conditioning system the load of the converter can be an induction machine with integrated inverter. Also a DC motor or universal motor can be used in simpler drive versions. Adjusting the DC-link voltage of the drive according to the load point of the machine can decrease the losses dissipated in the boost converter significantly. Furthermore, it is possible to increase the efficiency in induction machines by lowering the stator voltage. This is because the iron losses are dependent on the stator voltage and also the magnetizing current is smaller if the stator voltage is low. Decreasing the stator voltage is state-of-the-art in inverter fed drives with low loads at small speeds. Such drives can be found in various applications (e.g. compressors, pumps, fans, etc.).
978-1-4244-1664-6/08/$25.00 2008 IEEE

(1)

where VOUT is the output voltage of the stage and VIN is the input voltage of the stage. d represents the duty ratio, which is defined by
d= ton TS

(2)

where ton is the time when the active switch is closed and TS is the switching period of the converter [2]. Two single-stage boost converters are connected in series as it is illustrated in Fig. 1. The resulting two-stage boost converter is operated in push-pull mode. Using (1) the voltage transformation ratio of the proposed two-stage boost converter MB can be written as
MB = VOUT 1 VOUT 2 1 1 = 1 d VIN 1 VIN 2 1 1 d2

(3)

with

667

VOUT 1 = VIN 2

(4)

where the variables indicated with index 1 refer to the first (low voltage, high current) stage and the variables indicated with index 2 refer to the second (high voltage, low current) stage. If the switch pairs S1 and S3, as well as S2 and S4, are switched on and off simultaneously the voltage transformation rates of both stages are the same. Hence, both stages are controlled with the same duty ratio signal, d, in any instance. Consequently,
d1 = d 2 = d

(5)
Fig. 2. Relationship between duty ratio and voltage transformation ratios.

and d can be calculated by

d=

t S1, S 3 TS

TS t S 2, S 4 TS

(6)

For such a switch control method the voltage transformation ratio of the two-stage boost converter MB can be written as
MB = VOUT 2 1 = . V IN1 1 d
2

(7)

The voltage conversion ratios according to (1) and (7) are illustrated in Fig. 2. III. DIMENSIONING OF THE COMPONENTS For the parameterization of the converter the equations resulting from the ideal circuit without consideration of losses are used [2]. From (7) the duty ratio can be found if the voltage conversion ratio is known. The average inductor current of a single boost converter stage is
IL = PN VIN

In the following, the component values of a two-stage boost converter for a battery driven air conditioning system in heavy duty vehicles are given. The open circuit voltage of the battery is 25 V and the nominal power of the compressor is 1500 W. The compressor is assumed to have a moderate breakaway torque. Using (8) (11) with a switching frequency of 50 kHz the inductances are L1 = 40 H and L2 = 270 H, and the capacitances are C1 = 760 F and C2 = 220 F. These values are used for all presented simulation results and frequency domain plots of the converter. IV. DYNAMIC MODEL REPRESENTATION Due to the proposed switching scheme and considering only CICM the converter can be described by two equivalent circuits. The circuit shown in Fig. 3.a describes the behavior of the converter in state 1 where S1 and S3 are closed and S2 and S4 are open. The circuit shown in Fig. 3.b describes the behavior of the converter in state 2 where S2 and S4 are closed and S1 and S3 are open. In the two circuits the ohmic contributions of the storage inductors are considered by RL1 and RL2, the ESRs of the capacitors are considered by RC1 and RC2 and the on-resistances of the four MOSFET switches are considered by RS1, RS2, RS3 and RS4, respectively. Switching losses are not considered in the model. If the switches S2 and S4 are replaced by diodes in order to build a unidirectional converter the circuit representing state 2 must be extended with two voltage sources in order to consider the threshold voltages of the passive switches. In the presented simulation results and frequency domain plots the ohmic contributions of the inductors are RL1 = RL2 = 10 m and the ESRs of the capacitors are RC1 = 58 m and RC2 = 69 m. The on-resistances of the switches are RS1 = RS2 = RS3 = RS4 = 20 m. Based on the two circuits representing state 1 and state 2 one can establish two describing differential equation systems, one for each state. The circuit representing state 1 can be described by

(8)

whereas PN is the nominal converted power and VIN is the input voltage of the stage. From L a current ripple can be estimated. Depending on the application a current ripple around
I L (0.1...0.3) I L

(9)

is a reasonable design rule. The storage inductance of a stage can be estimated by


L= VIN d I L f S

(10)

where fS is the switching frequency of the converter. By choosing a maximum output voltage ripple VOUT the value of the buffer capacitor can be estimated by
C= I L , MAX VOUT d fS

(11)

where IL,MAX is the maximum current through the storage inductor.

668

diL1 1 = [vIN 1 iL1 ( RL1 + RS1 )] dt L1 diL 2 1 [vC1 iL 2 (RC1 + RL 2 + RS 3 )] = dt L2 dvC1 1 ( iL 2 ) = dt C1 dvC 2 1 1 = vC 2 dt C2 RLoad + RC 2

(12.a) (12.b) (12.c) (12.d)


Fig. 3.a. Equivalent circuit of the converter in state 1.

where iL1 and iL2 are the currents through the storage inductors L1 and L2, and vC1 and vC2 are the voltages across the buffer capacitors C1 and C2. If the converter is in state 2 the corresponding equations are
diL1 1 = [vIN1 iL1 ( RL1 + RS 2 + RC1 ) + iL 2 RC1 vC1 ] (13.a) dt L1
RC 2 diL 2 1 = 1 vC1 + iL1RC1 + iL 2 Ra + vC 2 dt L2 RLoad + RC 2 (13.b) dvC1 1 (iL1 iL 2 ) = dt C1

Fig. 3.b. Equivalent circuit of the converter in state 2.

(13.c)

dvC 2 RC 2 1 1 1 vC 2 = iL 2 dt C2 RLoad + RC 2 RLoad + RC 2 (13.d) with 2 RC 2 . (13.e) Ra = (RC1 + RL 2 + RS 4 + RC 2 ) + RLoad + RC 2

Fig. 4. Simulation results of the switched converter model (black) and of the averaged system (grey). Voltage across C2 (top); current through L2 (bottom).

A22 =

1 d (RC1 + RL 2 + RS 3 ) + L2
2 RC 2 (RC1 + RL 2 + RS 4 + RC 2 ) R R + Load C2

The two state-space systems (12) and (13) can be combined in order to derive a nonlinear averaged statespace system for the converter. An averaged system is only accurate if the switching frequency is very large compared to the eigenfrequencies of the converter. Weighting (12) with d and (13) with (1-d) before combining them yields
iL1 A11 d iL 2 A21 = dt vC1 A31 v C2 0 A12 A22 A32 A42 A13 A23

+ (1 d ) A23 = A24 = A31 =

(14.f)

1 L2 1 d L2 1 d C1 1 C1
RC 2 1 R R + C2 Load

(14.g) (14.h) (14.i) (14.j) (14.k) (14.l) (14.m)

0 0

0 iL1 B1 A24 iL 2 0 v (14.a) + IN1 0 vC1 0 A44 vC 2 0

A32 = A42 =

with
A11 =

1 [(d 1)(RL1 + RS 2 + RC1 ) d (RL1 + RS1 )] L1 R A12 = C1 (1 d ) L1 d 1 A13 = L1 R A21 = C1 (1 d ) L2

(14.b) (14.c) (14.d) (14.e)

RC 2 1 (RLoad + RC 2 ) 1 A44 = C2 (RLoad + RC 2 ) 1 . B1 = L1

1 d C2

With the state-space model described in (14) one can quickly assess the behavior of the converter. A large signal analysis is possible since there are no limitations imposed on the state quantities. However, the ripple on currents and voltages caused by the switching between

669

the two states is not described in averaged converter models. In SMPS (switch mode power supplies) usually very large buffer capacitors are chosen because a smooth output DC voltage is of highest concern. Therefore, the signals generated by an SMPS model considering switching effects differ only insignificantly from the signals of an averaged SMPS model as described in (14). In Fig. 4 MODELICA [3] results of the uncontrolled converter, simulated with the averaged model (14) and simulated with a model considering switching effects are presented. Between time 0.02 s and time 0.04 s the duty ratio gets decreased from 0.76 to 0.6. It can be seen that, apart from the switching effects, there is no significant difference between the results. However, a realistic model has to include a current limitation. V. MODEL FOR CONTROL

A21 =

RC1 (1 D0 ) L2
2 RC 1 2 + RS 4 + RC2 RS 3 + D0 L2 RLoad + RC2

(16.e)

A22 =

A. Linearization The weighted state-space model (14) is nonlinear because in real operation the duty ratio is subject to changes. In order to utilize the possibilities of control design based on linear control theory (14) has to be linearized so that the duty ratio can be regarded as an input quantity in the state-space system. Linearization is only accurate if the disturbances (in a controlled system) of input quantities and state quantities are small compared to their steady-state values. State quantities and input quantities can be split into two components, one representing the steady-state value, and the other one representing disturbances. Steady-state values are written in capital letters and denoted with index 0 and disturbances are indicated with ~. For the proposed converter the ansatz is ~ iL1 = I L10 + iL1 ~ iL 2 = I L 20 + iL 2 ~ vC1 = VC10 + v C1 (15) ~ vC 2 = VC 20 + v C2 ~ v =V +v
~ d = D0 + d
IN 1 IN 10 IN 1

2 RC 2 + (RC1 + RL2 + RS 4 + RC2 ) RLoad + RC2 1 A23 = L2 D 1 RLoad A24 = 0 L2 RLoad + RC 2 1 D0 A31 = C1 1 A32 = C1 (1 D0 ) RLoad A42 = C2 (RLoad + RC 2 ) 1 A44 = C2 (RLoad + RC 2 ) 1 B11 = L1 1 B12 = [VC10 (I L 20 RC1 ) + I L10 (RS 2 + RC1 RS1 )] L1

(16.f)

(16.g) (16.h) (16.i) (16.j) (16.k) (16.l) (16.m) (16.n)

B22 =

RLoad 1 + VC20 L2 RLoad + RC 2

2 RC 2 + I L20 + RS 4 + RC 2 RS 3 RC1I L10 R Load + RC 2 I L10 B32 = C1 I RLoad . B42 = L 20 C2 (RLoad + RC 2 )

(16.o)

(16.p) (16.q)

Using (14) and (15) the linearized state-space model applicable for control design is ~ ~ iL1 A11 A12 A13 0 iL1 B11 B12 ~ ~ ~ d iL 2 A21 A22 A23 A24 iL 2 0 B22 vIN1 = + ~ ~ 0 B ~ 0 dt v 32 d C1 vC1 A31 A32 0 ~ v ~ C 2 0 A42 0 A44 vC 2 0 B42 (16.a) with
A11 = A12 = 1 [ (RL1 + RS 2 + RC1 ) + D0 (RS 2 + RC1 RS1 )] (16.b) L1

B. Transfer Functions Applying Laplace transform on (16) the transfer functions for VC2(s), given in (17.a) and (17.b), can be derived. In Fig. 5 8 the bode plots and the poles and zeros corresponding to (17.a) and (17.b) are presented. (17.a) can be used to design a linear voltage control system and (17.b) is the basis for a possible disturbance compensation. One has to bear in mind, however, that the transfer function is plotted only for one operation point. To construct an appropriate controller different operation points have to be studied and the whole set of transfer functions has to be considered.
VI. MODEL CONSIDERING SWITCHING EFFECTS WITH IMPLEMENTED CURRENT LIMITATION In a real application the currents through the storage inductors have to be limited in order to avoid saturation of the magnetic cores. A possible method is to control the switches such way that S1 and S3 turn-off and S2 and S4

RC1 (1 D0 ) L1 D 1 A13 = 0 L1

(16.c) (16.d)

670

Fig. 5. Bode plot of the transfer function GV

C 2D

( s ).

Fig. 7. Bode plot of the transfer function GV

C 2VIN 1

( s ).

Fig. 6. Poles and Zeros of the transfer function GV

C 2D

( s ).

Fig. 8. Poles and Zeros of the transfer function GV

C 2VIN 1

( s ).

GVC 2 D ( s ) =

VC 2 (s ) = D(s ) V IN 1 (s )= 0
A12 A13 B12 s A22 A23 B22 s B32 A32 0 B42 A42 0 A12 A13 s A22 A23 A24 0 s A32 0 s A44 A42

GVC 2V IN 1 ( s ) =
s A11 A21 A31 0 s A11 A21 A31 0

VC 2 (s ) = VIN 1 (s ) D (s )= 0
A12 A13 B11 s A22 A23 0 0 s A32 0 0 A42 0 A12 A13 s A22 A23 A24 0 s A32 0 s A44 A42

s A11 A21 A31 0 s A11 A21 A31 0

(17.a)

(17.b)

turn-on when either the limit of iL1 or the limit of iL2 is reached. The converter flips back to conventional pushpull mode as soon as a positive edge of the pulse-width modulated (PWM) signal for voltage control occurs. Since the converter works bidirectional also a current limitation for regenerative mode has to be implemented. Whenever one of the negative current limits is reached the switches S1 and S3 turn-on again and the switches S2 and S4 turn-off so that the negative amounts of the currents start to decrease again. Detailed simulation screenshots of the converter signals with implemented current limitation are given in [4].
671

VII. POSSIBLE CONTROL TOPOLOGIES In Fig. 9 four options for voltage control are presented. Please, note that the transfer functions of the plant (17.a) and (17.b) only apply for control (a) and control (c). In control (a) a single linear controller directly generates the reference duty ratio of both stages. Parameterization of this controller can be sensitive due to the current limitation of the converter. However, for loads with relatively large time constants compared to the converter time constants (e.g. compressor drives, pump drives, fan drives, etc.) this control topology can be calibrated with bearable efforts. Simulation results of a

Fig. 9. Possible control topologies for the two-stage boost converter.

Fig. 10. Simulation results of the converter with control (a) and control (b). Voltage across C2 (top); feedback controller output signal (bottom).

Fig. 11. Simulation results of the converter with control (c) and control (d). Voltage across C2 (top); feedback controller output signal (bottom).

full electro-mechanical application using this control method are presented in [4]. A different approach is control (b). In this case a feedforward control law is used to generate the reference duty ratio signal d1 for stage 1 whereas stage 2 is controlled with d2 generated by a conventional feedback controller. Rearranging (7) the feedforward control law can be written as
d CL = 1 VIN 1 * VOUT 2

(18)

where dCL is the output duty ratio signal of the feedforward control law, VIN1 is the input voltage of * stage 1 and VOUT 2 is the reference output voltage of stage 2. Since the two stages are controlled with different duty ratio signals a state-space model of the plant applicable for linear control design would have to include two different duty ratio signals as input variables. However, one can easily include different duty ratio signals into the state-space model. With

control (b) an improvement in terms of dynamic response to reference voltage changes can be achieved. Furthermore, the calibration of the controller parameters is easier and the robustness concerning input voltage changes is better. A drawback of control (b) is the increased system sensitivity to heat related resistance changes of the components. Control (c) can be regarded as an extended version of control (a). Like in control (a), both stages are controlled with the same duty ratio signal d. The difference to control (a) is that d is generated by summing the output signal of the feedforward control law (18) and the output signal of a feedback controller. In order to reach good results the limits of the feedback controller have to be small compared to the output signal of the feedforward control law. During the startup of the converter the dynamic of control (c) is limited due to the saturation of the feedback controller. Control (a) and (b) have better start-up dynamics. One has to keep in mind that during start-up and during large transients the current limitation function of the

672

converter is effective in order to protect the converter devices. On the other hand, the dynamic response on reference voltage changes is very good in control (c) and the system is very robust against heat related resistance changes of the components. A further advantage of control (c) is the fairly easy calibration of the controller parameters. Very important for the performance of the system is the right choice of the limits in the anti-windup controller. For some applications variable limits that change with the load point of the converter can further improve the system performance. Control (d) is a combination of control (b) and control (c). The performance of control (d) under ideal conditions is not significantly better than the performance of control (c). However, the feedforward control of stage 1 makes the system more robust with regards to input voltage changes. Furthermore, the specific control topology of stage 2 accounts for a good system behavior towards heat related resistance changes in the components of the converter. Since no losses are considered in (18) there is further potential to improve control (b) (d) by using a relation derived from the steady-state solution calculated with (14) and (15) as an extended feedforward control law. With such a feedforward control law, the ohmic contributions of the storage inductors, the on-resistances of the switches and the ESRs of the buffer capacitors can be taken into account and consequently the robustness of the systems can be further increased. The drawback of such an extended feedforward control law is the need of higher processing effort. Furthermore, it is possible to increase the efficiency of the converter if two optimum duty ratios dOPT1 for stage 1 and dOPT2 for stage 2 can be found. This is a multi variable optimization problem. dOPT1 can be used in the feedforward control law in control (b) or in an extended version of control (d), for instance. VIII. SIMULATION RESULTS OF THE CONTROLLED CONVERTER In Fig. 10 and Fig. 11 MODELICA simulation results of the control systems illustrated in Fig. 9 are shown. All four control topologies are simulated with the same load conditions, input voltage and reference output voltage signal. The converter start-up is performed under low-load with a dissipated power of 0.05 PN and a reference output voltage of 350 V. At time 0.06 s a load dissipating 0.6 PN gets connected to the converter. Between time 0.1 s and time 0.15 s the reference output voltage is changed from 350 V to 200 V. Figure 10 shows that control (a) and (b) have very similar start-up characteristics. Both control systems

regulate load changes very fast. Also full load steps can be controlled without stability problems. At time 0.15 s when the reference voltage changes back to 350 V the limit of the anti-windup algorithm in the controller affects the system response considerably. Figure 11 illustrates that with control (c) and (d) also the start-up characteristics are very much influenced by the limits of the anti-windup algorithm. With a finer calibration of the limits the small overshoot at time 0.02 s could be eliminated. On the other hand, the controller must be able to contribute the full difference between the output signal of the feedforward control law and the duty ratio, which is actually needed to reach the reference output voltage. If the controller limit is too small a static error in the output voltage occurs. IX. CONCLUSIONS The biggest advantage of the proposed two-stage boost converter is the high voltage transmission ratio achieved with a fairly low duty ratio, which is a good premise for efficient operation. No transformer is needed to build this converter and a basic control system can be realized with only one integrator in a single control loop. The implementation of an openloop controller for the low voltage (high current) unit decreases instabilities due to input voltage changes and also makes it easier to calibrate the closed-loop controller of the high voltage (low current) unit. Calculating the duty ratio by summing a closed-loop controller signal and an open-loop controller signal further increases the robustness of the system. The comparison of the four proposed control methods shows that control (d) has the best overall performance. It is fairly easy to calibrate and robust against input voltage changes as well as parameter detuning due to varying operation temperature. REFERENCES
[1] Edelmoser, K.H.; Himmelstoss, F.A.: DC-to-DC Solar Converter with Controlled Active Clamping System. Proc. of EPE-PEMC 06; Conference on Power Electronics and Motion Control - Portoroz (Slovenia) Aug. 30- Sept. 1, 2006, pp. 124-127. [2] Mohan N.; Undeland, T.M.; Robbins, W.P.: Power Electronics, 2nd ed., J. Wiley & Sons, Inc., New York, 1989, pp. 172-178. [3] Fritzson, P.: Principles of Object-Oriented Modeling and Simulation with Modelica 2.1, IEEE Press, J. Wiley & Sons, Inc., Piscataway, NJ, 2004. [4] Gragger, J.V.; Simic, D.; Himmelstoss, F.; Pirker, F.: Simulation of a battery powered air conditioning system with Modelica. Proc. of PCIM 2008; Conference for Power Electronics, Intelligent Motion and Power Quality - Nuremberg (Germany), May 2008.

673

Potrebbero piacerti anche